ATTENTION MICROFICHE USER, The original document from which this microfiche was made was found to contain some imperfection or imperfections that reduce full comprehension or some of the text despite the good technical quality of the microfiche itself. The imperfections may be: - missing or illegible pages/figures; - wrong pagination; - poor overall printing quality, etc...

We normally refuse to microfiche such a document and request a replacement document (or page) from the national INIS Centre concerned. However, our experience shows that many months pass before such documents are replaced. Sometimes the Centre is not able to supply a better copy or, in some cases, the pages that were supposed to be missing correspond to a wrong pagination only. We feel that it is better to proceed with distributing the microfiche made of these documents than to withold them till the imperfections are removed. If the removals are subsequestly made then replacement microfiche can be issued. In Une with this approach then, our specific practice for microfiching documents with imperfections is as follows:

1. A microfiche of an imperfect document will be marked with a special symbol (black circle) on the left of the title. This symbol will appear on all masters and copies of the document (1st fiche and trailer fiches) even if the imperfection is on one fiche of the report only. 2. If imperfection is not too general the reason will be specified on a sheet such as this, in the space below.

3. The microfiche will be considered as temporary, but sold at the normal price. Replacements, if they can be issued, will be available for puchase at the regular price.

4. A new document will be requested from the supplying Centre. 5. If the Centre can supply the necessary pages /document a new master fiche will be made to permit production of any replacement microfiche that may be required.

The original document from which this microfiche has been prepared has these imperfections: missing pagesmgu»8i numbered: \ ————— wrong pagination poor overall printing quality combinations of the above other INIS Clearinghouse IAEA P.O. Box 100 A-1400, Vienna-AUSTRIA \ ^^"[OOJoT- fMC.TH.33D2-

THESE

PRESENTEE

A L'UNIVERSITE PARIS VII

POUR OBTENIR

LE GRADE DE DOCTEUR ES SCIENCES PHYSIQUES

par

Isabel le GRENIER

LES RAYONS GAMMA , TRACEURS OU MILIEU INTERSTELLAIRE

ET MESSAGERS DES PULSARS ET AUTRES OBJETS ENERGETIQUES

Soutenue le 24 mars 1988 devant la Commission d'Examen composée de:

MM. A. BRAHIC Président

L. GOUGUENHEIM Rapporteur

J. PAUL

R. PELLAT

G. VEDRENNE

J. AUDOUZE Examinateur

P.GOLDREICH

J. HEYVAERTS » THESE

PRESENTEE

A L'UNIVERSITE PARIS VII

POUR OBTENIR

LE GRADE DE DOCTEUR ES SCIENCES PHYSIQUES

par

Isabelle GRENIER

LES RAYONS GAMNA,TRACEURS DU HILIEU INTERSTELLAIRE

ET MESSAGERS DES PULSARS ET AUTRES OBJETS ENERGETIQUES

Soutenue le 24 mars 1968 devant la Commission d'Examen composée de:

MM. A. BRAHIC Président

L. GOUGUENHEIM Rapporteur

J. PAUL "

R. PELLAT

G. VEDRENNE

J. AUOOUZE Examinateur

P.GOLDREICH "

J. HEYVAERTS " " Joy -in looking and compteh&nding lt> natwiz'i moat beMUl£ut gli-t " MJbvU Einitzin

AVANT PROPOS

En premier lieu, je tiens à remercier vivement tous les membres du jury qui

m'ont fait l'honneur et la joie d'accepter de juger ce travail; André BRAHIC

pour avoir bien accepté d'assurer la présidence du jury, Lucienne GOUGUENHEIM,

Jacques PAUL, René PELLAT et Gilbert VEDRENNE qui ont bien voulu être les

rapporteurs de cette thèse ainsi que Jean AUDOUZE, Peter COLDREICH et Jean

HEYVAERTS pour avoir accepté de faire partie du jury.

Ce mémoire est le fruit de travaux menés depuis l'été 1983 en astronomie gamma. L'interprétation des données m'a permis d'aborder des domaines aussi variés que l'étude du milieu interstellaire en gamma, radio, millimétrique ou

infra-rouge et la physique des pulsars et des restes de supernovae. J'ai pu ainsi m'initier à plusieurs techniques d'observation et à quelques aspects théoriques. L'essentiel de ces travaux a eu pour cadre le Service d'Astrophysique du Centre d'Etudes Nucléaires de Saclay. Aussi voudrais-je exprimer ma gratitude à Daniel CRIBIER, Alain RAVIART, Charles RYTER et

Catherine CESARSKY pour n'avoir accueillie et offert la possibilité d'apprendre le métier de chercheur dans des domaines aussi variés et actifs de l'astronomie moderne. J'ai particulièrement apprécié l'ambiance chaleureuse du service. Les nombreuses discussions que j'ai eues avec tous ont été pour moi un encouragement constant. J'aimerais tous les remercier vivement.

L'un des charmes de l'astrophysique est de pouvoir nouer de nombreuses collaborations au niveau international. COS-B fut une excellente carte de visite

dans ce donaine et, grâce à elle, j'ai pu rencontrer en particulier Wim HEEMSEN,

Andy STRONG and Andy POLLOCK. Je voudrais les remercier de m'avoir transmis

"l'art" de déjouer les pièges de COS-B et du maximum de vraisemblance. Cette

thèse doit beaucoup à leur amitié.

Le défaut majeur de l'astronomie spatiale est de ne percevoir le ciel qu'au bout d'un ordinateur. Aussi aimerais-je remercier Patrick THADDEUS de la confiance qu'il m'a témoignée en mettant à ma disposition le télescope millimétrique de Columbia. Il m'a ainsi permis de connaître l'excitation (et les rigueurs) des observations "en direct". Un grand merci également aux membres de son équipe, Thomas DAME, Ronald HADDALENA et Sam PALMER, pour m'avoir enseigné les techniques de la radioastronomie millimétrique, même si je les soupçonne d'avoir été de très bons professeurs pour éviter d'être réveillés en pleine nuit par des questions du style "Pourquoi le télescope tourne-t-il à l'envers ou comment remonter à l'endroit la batterie de filtres et réaccorder le récepteur?", l'annonce des découvertes étant bien sûr toujours remise au lendemain matin.

Je voudrais exprimer ma profonde reconnaissance à Jacques PAUL pour n'avoir spontanément accueillie et initiée à l'astronomie gamma. Sans son soutien, la méthode d'analyse spectrale n'aurait jamais vu le jour. Jacques m'a également appris à oublier les réflexes d'écolière et à ne pas restreindre un travail d'analyse de données à une publication de chiffres, mais à tenter chaque fois de percer les conséquences les plus diverses de ces nouveaux indices. Il m'a

également démontré l'importance de la simplicité et de l'esprit critique dans un travail d'interprétation. Je me suis souvent appuyée sur ses vastes connaissances pour aller de l'avant. J'espère que son exemple m'incitera toujours à éviter l'écueil d'une trop grande spécialisation. François LEBRUN sait combien sa collaboration et son amitié me sont précieuses. Je voudrais le remercier plus particulièrement pour la patience avec laquelle il m'a enseigné l'art des statistiques. Malgré des rechutes fréquentes, je crois comprendre un peu mieux les effets malicieux du hasard. J'ai par dessus tout apprécié le jeu "d'avocat du diable" que nous avons souvent joué au cours d'interminables discussions. La plupart des travaux présentés ici sont nés de telles joutes. Elles m'ont appris la prudence et la remise en cause.

Je voudrais enfin remercier mes parents d'avoir cru une gamine qui, en

1969, avait brusquement décidé de devenir astronome et de m'avoir ensuite toujours soutenue dans cette voie exotique. Sans M. LEQUEUX, instituteur à l'Kay les Roses qui m'a révélé les merveilles du ciel, et M. DUCROCQ qui m'a fait découvrir les rayons gamma, cette thèse n'aurait jamais existé. Mes sincères remerciements vont également à André pour ses conseils, ses encouragements et pour m'avoir communiqué une partie de son immense enthousiasme. TABLE DES MATIERES

1. INTRODUCTION

1.1 Petit Guide de la thèse

1.2 "Les Rayons de la Colère"

2. L'ASTRONOMIE GAMMA AVEC COS-B

2.1 Un télescope gamma

2.2 Quelques méthodes d'analyse

a) analyse de l'émission gamma galactique

b) recherche de sources gamma ponctuelles

c) analyse temporelle et spectrale de sources isolées

3. LE MILIEU INTERSTELLAIRE

3.1 Les nuages moléculaires dans la Galaxie

3.2 Comparaison de traceurs du milieu interstellaire

4. LE RAYONNEMENT GAMMA GALACTIQUE DIFFUS

4.1 Les rayons gamma et le gaz interstellaire dans le disque galactique

4.2 Les rayons gamma et le gaz interstellaire dans Céphée - Cassiopée

5. L'EMISSION GAMMA DES PULSARS

5.1 Le pulsar du Crabe

5.2 Le pulsar des Voiles

6. PERSPECTIVES - I -

INTRODUCTION

" Vvwujjt. a. le m£me. cUiolt de i 'expumA que la \>iiUtt " CondoiceX - J -

INTRODUCTION

1.1 PETIT GUIDE DE LA THESE

Ce mémoire est le fruit de travaux menés depuis l'été 1963 sous la direction de Jacques PAUL au Service d'Astrophysique du Centre d'Etudes

Nucléaires de Saclay. avec pour fil d'Ariane l'analyse du rayonnement gamma observé dans notre Galaxie par le satellite européen COS-B. Un tel rayonnement ayant deux origines distinctes dans la Galaxie, ce mémoire s'articule autour de deux thèmes principaux. Il s'agit tout d'abord de l'étude du milieu

interstellaire en mettant à profit le fait que les rayons gamma diffus naissent de l'interaction des rayons cosmiques avec toute matière interstellaire.

D'autres sources d'information ont également été utilisées dans cette étude, notamment les cartes de nuages moléculaires obtenues pendant plusieurs mois d'observations avec le télescope millimétrique de l'Université Columbia de New

York. L'analyse détaillée de brillantes sources compactes de rayons gamma constitue l'autre thème du mémoire. Les photons gamma sont alorsexploités pour tenter de sonder les conditions physiques extrêmes qui gouvernent les objets très énergétiques tels que les pulsars.

Il peut paraître étonnant de présenter une thèse sur les observations de

COS-B treize ans après le lancement de celui-ci, alors que les 250 000 photons qu'il a recueillis ont déjà donné matière à cinq ou six thèses. Mais ce laps de temps est le simple reflet de la complexité des analyses eu astronomie gamma. - 2 .

Environ dix ans ont été nécessaires pour pleinement cerner et caractériser le

comportement du télescope durant sa longue vie et le même temps a été requis

pour apprendre à exploiter efficacement les quelques photons capturés grâce à

des analyses statistiques toujours plus poussées. L'exposé de ces difficultés

constitue le prochain chapitre. De plus, l'analyse du rayonnement gamma diffus

n'est jamais si fructueuse que lors de la confrontation avec les cartes des

nuages moléculaires. Or la couverture complète du disque galactique et de ses

abords dans le domaine millimétrique a été achevée seulement en 1986. Ainsi,

longtemps après l'excitation des premières observations et découvertes, COS-8 a

encore beaucoup à offrir et les travaux réunis ici ne représentent qu'une petite

partie de la moisson de résultats qui attendent encore.

En premier lieu sont présentées les observations millimétriques effectuées

à New York. Elles couvrent près du quart de la Voie Lactée (le premier quadrant

galactique) ainsi que la constellation de Céphée. Ces mesures ont permis de

confirmer l'existence en CO jusqu'ici très controversée d'un lointain bras

spiral de la Galaxie, situé à sa périphérie (15 kpc), et de cartographier pour

la première fois un long prolongement du bras spiral de Persée dans le premier

quadrant. Dans la direction de Céphée, les observations ont dévoilé l'existence d'un complexe moléculaire massif, proche du Soleil (300 pc). Une large et

étonnante "bulle" y a été décelée en CO ainsi qu'à d'autres longueurs d'onde.

Elle correspond vraisemblablement au reste de l'explosion d'une supernova qui

s'est produite il y a environ 40.000 ans près du Soleil (300 pc). Enfin,

l'ensemble de ces observations a été utilisé pour élaborer le premier panorama

C0 complet du contenu en gaz moléculaire de la Voie Lactée.

La seconde partie présente les résultats d'études comparatives entre divers

traceurs du milieu interstellaire. Sont abordées par exemple la calibration des mesures d'absorption interstellaire par les comptages de galaxies ou les

capacités de traceur moléculaire de l'émission infrarouge à 100 microns. D'autre _ 3 _

parti la corrélation des cartes du gaz interstellaire et du rayonnement gamma

diffus a été effectuée indépendamment dans tout le disque galactique et dans

Céphée. Toutes deux fournissent une mesure du rapport de conversion des

observations millimétriques en colonnes de densité de gaz moléculaire. De

faibles gradients de densité des rayons cosmiques ont également été ainsi

mesurés à l'échelle de la Galaxie. Dans Céphée plus particulièrement, l'analyse

a apporté la découverte d'une source que l'on peut probablement considérer comme

la dernière des rares sources gamma de haute énergie détectées dans l'Univers

jusqu'à présent. Mais sa nature véritable demeure, comme pour ses congénères* un

mystère.

Le dernier chapitre regroupe les résultats de l'étude des rayons gamma émis

par les deux seuls pulsars observés à ces énergies. L'analyse spectrale et

temporelle du pulsar du Crabe a été effectuée de manière traditionnelle tandis

que le pulsar Vela a été étudié en détail grâce à une méthode d'analyse que j'ai

mise au point dans les premiers temps de ma thèse, Cette approche a mis en

évidence des comportements inattendus du pulsar (forte variabilité à long terme,

émission puisée hétérogène comprenant plusieurs composantes de spectre et

d'évolution disparates) qui démontrent la complexité de sa magnétosphère

(coexistence de plusieurs sites d'activité et de divers processus générateurs de

rayons gamma, variabilité au coeur de la magnétosphère mais stabilité relative à

sa périphérie), Confrontés aux modèles théoriques et au comportement beaucoup plus calme du Crabe, ces résultats apportent de nouveaux indices et contraintes observâtionnelles sur les mécanismes fort mal connus d'émission des pulsars. - « -

1.2 "LES RAYONS DE LA COLERE"

Comment ne pas regretter l'étroitesse de la fenêtre que la Nature nous a accordée sur le nonde extérieur? En protégeant la vie. l'atmosphère de notre planète fait écran à la majorité des rayonnements que nous envoie l'Univers et nous prive ainsi d'une grande source d'informations. Car. à l'aube de l'aventure spatiale, la principale source d'informations dont nous disposons sur l'Univers provient toujours des rayons que nous captons. Fendant des millénaires, l'homme a dû se contenter de regarder le ciel dans la minuscule gamme de longueurs d'onde auxquelles son oeil est sensible, la lumière visible, en inventant lunettes et télescopes pour percevoir des détails de plus en plus fins et lointains. Hais, grâce à la technologie du XXème siècle, il a peu à peu appris à capter et à regarder d'autres lumières ,.. Les rayonnements radio ainsi qu'infra-rouge et ultra-violet proches furent les premiers et les plus faciles à enregistrer parce qu'ils traversent l'atmosphère terestre. Puis, avec l'invention des ballons, des fusées et des satellites, vint le tour des rayonnements infra-rouge et ultra-violet lointains et des rayons X. Néanmoins en

1967, aucun photon gamma extra-terrestre n'avait encore été aperçu,

Ces rayons, provenant de l'extrémité du spectre électrc.-magnétique située vers les courtes longueurs d'onde, transportent les plus grandes énergies créées dans la Nature. Ils prennent d'ailleurs naissance au cours des processus astrophyciques les plus violents rencontrés dans l'Univers, impliquant soit l'excitation de noyaux atomiques, ou la présence de particules ultra- relativistes ou encore l'existence de champs électro-magnétiques intenses. Par leur formidable énergie, ces rayons ont un pouvoir de pénétration de la matière important, Contrairement à beaucoup d'autres rayonnements, ils traversent sans encombre les milieux le* plus denses pour nous renseigner tur les événements qui se déroulent dans les régions reculées de notre Galaxie. Bien entendu, ils parcourent également les immenses régions intergalactiques sans être altérés ni - s- _

déviés pour nous révéler l'activité d'autres galaxies. Bien que fortement

apprécié des astronomes, ce pouvoir de pénétration pose en contrepartie un

problème épineux de détection. Pour arrêter un photon gamma, il faut un matériau

dense et une surface d'interception importante car il n'existe aucun système de

"miroirs" analogue à celui utilisé dans d'autres télescopes. Malheureusement,

poids et grande taille sont les ennemis jurés des lanceurs astronautiques ... De

plus, l'observation des rayons gamma se heurte à la rareté des photons, Il

emportent avec eux une énergie telle qu'ils sont suffisamment rares dans

l'Univers pour être captés un par un par les instruments. Ils épuiseraient

d'ailleurs vite les ressources de l'astre émetteur s'ils étaient plus nombreux,

Tandis que cette page reçoit environ 10" photons lumineux du Soleil par

seconde, le télescope gamma à bord de COS-B a intercepté en moyenne 3000 photons

par mois! La situation empire encore au delà de 100 GeV si bien que l'on est obligé d'utiliser l'atmosphère terrestre elle-mêne comme détecteur. Il est amusant de noter que l'astronomie gansa de ces très hautes énergies se pratique

à l'aide de télescopes optiques conventionnels, équipés de photomètres très

rapides pour déceler les brefs éclairs de lumière Cherenkov que provoque l'entrée d'un photon gamma dans l'atmosphère. 11 faut alors des semaines, voire des mois d'exposition pour obtenir une image! Mais, puisque très peu de ces photons extrêmes ont été utilisés dans le présent mémoire, je les écarterai de la suite de l'histoire gamma. Par souci de brièveté, j'écarterai également l'astronomie gamma de basse énergie (inférieure à quelques MeV) qui appartient davantage au domaine de l'astrophysique nucléaire (étude des raies nucléaires et nucléosynthèse) et à l'étude spécifique des sursauts gamma.

Ainsi faut-il des satellites, de grandes surfaces collectrices et beaucoup de patience pour étudier les rayons gamma célestes. C'est pourquoi l'astronomie gamma est la dernière née des branches de l'astronomie moderne. Il y a seulement

35 ans, on imaginait à peine la présence de rayons gamma parcourant l'Univers. - S -

Leur existence fut prédite d'un point de vue théorique en 1952 lorsqu'on

soupçonna les rayons cosmiques d'engendrer des photons gamma en traversant la

matière interstellaire, soit par collision des protons cosmiques et des noyaux

interstellaires puis désintégration spontanée des mésons n° ainsi formés

(Hayakawa, 1952, Prog. Theor. Phys., 8, 571), soit par freinage des électrons

cosmiques dans le champ électrique des noyaux interstellaires (Hutchinson, 1952,

Phil. Hag., 43, 847). Ces travaux furent repris et complétés à la fin des années

50 par l'américain P. Morrison, le français E. Schatzman et les soviétiques V.I.

Ginzburg et S.O. Syrovatskii. L'émission Compton inverse des électrons cosmiques sur le champ de lumière visible et infra-rouge galactique fut notamment ajoutée au rang des mécanismes de production d'émision gamma diffuse. Nul ne soupçonnait alors l'existence de "sources gamma" isolées dans la Galar.ie.

Jusqu'en 1968, tous les espoirs de détection furent déçus. Les photons gamma déjouèrent tous les détecteurs envoyés en haute altitude en ballon. Mais cette année là, la possibilité de placer un détecteur à bord du satellite 0S0-3 fut offerte à une équipe du M.I.T. Ils furent récompensés de leurs efforts par la découverte de 631 photons gamma d'énergie supérieure à 50 MeV (Kraushaar et al., 1972, Ap. J. • 177, 341) qui confirmaient modestement l'émission diffuse de la Voie Lactée. L'astronomie gamma naissait enfin. Peu après, elle portait à son actif la découverte d'une première source ponctuelle, le pulsar du Crabe, grâce

à un vol franco-italien en ballon stratosphérique. Puis, en novembre 1972, fut lancé un nouveau détecteur à bord du satellite américain SAS-2. Les observations furent malheureusement interrompues au bout de 8 mois à cause d'une défaillance technique. Mais les 8000 photons recueillis au dessus de 30 MeV permirent d'esquisser les grandes lignes de la carte gamma de la Voie Lactée et d'isoler deux nouvelles sources: le pulsar de Vela et un mystérieux astre de la constellation des Gémeaux, baptisé plus tard Geminga. Deux autres pulsars et la source Cygnus X-3 furent également repérés mais les observations suivantes ne les ont pas confirmés. Dès lors, le ciel gamma possédait toutes les - ? -

caractéristiques que nous lui connaissons aujourd'hui, c'est à dire d'une

émission diffuse ponctuée de sources puissantes. L'inattendu était la forte

luminosité à haute énergie des deux jeunes pulsars du Crabe et de Vela. En 1975,

l'analyse de l'émission diffuse observée par SAS-2 imposait l'idée des pionniers

qu'elle provient essentiellement de l'interaction des rayons cosmiques avec tout

gaz interstellaire, qu'il soit sous forme atomique, moléculaire ou ionisée. Dans

le même temps, la mise en évidence dans les données gamma d'une variation de la

densité des protons cosmiques à l'extérieur de la Galaxie apportait la preuve de

l'origine galactique des rayons cosmiques.

En août 1975, l'astronomie gamma a enfin connu un véritable essor grâce au

lancement du satellite européen COS-B. Porteur d'un télescope semblable à celui

de SAS-2 mais beaucoup plus performant, il a parfaitement fonctionné jusqu'à

épuisement de ses combustibles en avril 1982, soit pendant presque 7 ans. A son

tableau de chasse figurent plus de 240 000 photons qui ont permis de dresser une

carte beaucoup plus fine de l'émission galactique et qui ont surtout révélé

l'existence inattendue d'une population d'une vingtaine de sources

"ponctuelles". Avec COS-B est donc né le problème de la recherche et de

l'identification de nouveaux astres, les sources gamma, qui rayonnent le plus clair de leur énergie aux fréquences gamma. Les pulsars sont aisément

reconnaissables grâce à leur signature temporelle mais seuls deux, le Crabe et

Vela, ont été repéré avec certitude, La nature de la plupart des autres sources reste encore mystérieuse, l'identification étant rendue difficile par la mauvaise résolution du télescope gamma. D'après leur distribution concentrée le long du plan galactique, il s'agit probablement d'objets jeunes, situés entre 2 et 7 kpc du Soleil, mais dans la mesure où tout ce qui a une dimension angulaire inférieure au degré est "vu" par COS-B comme un point, les objets recherchés ne sont pas nécessairement de type stellaiie. Une accumulation de rayons cosmiques ou de matière interstellaire simulerait tout aussi bien une source ponctuelle. A très haute latitude galactique, la tâche est un peu facilitée. C'est pourquoi on - g - a pu reconnaître en 2CG289+64 le quasar 3C273. Il est cependant clair que de nouvelles observations, plus précises, seront nécessaires pour identifier les sources qui se regroupent près du plan galactique. Les instruments qu'utilise aujourd'hui l'astronomie gamma sont très imparfaits. Ils n'atteignent pas les performances des télescopes construits aux autres longueurs d'onde, loin s'en faut. Hais il ne s'agit là que de la première génération de télescopes gamma et de nouveaux vont suivre rapidement pour améliorer les observations présentes, compléter les régions du ciel non couvertes par COS-B et étendre l'astronomie gamma spatiale à de nouvelles énergies. Alors l'astronomie gamma quittera l'enfance. - II -

L'ASTRONOMIE GAMMA AVEC COS-B

. . . . On a iondé eu liglom vailéu. Lu bonnu du. pou-ibte. ont été temtéu! Un montit a pu. voiA. a/imé d'un o

2,1 UM TELESCOPE GAMMA

Après avoir evince COS-A de la liste des projets, le satellite COS-B eut

l'honneur d'être le premier fleuron de l'Agence Spatiale Européenne. Cinq

laboratoires européens se sont associés pour concevoir et réaliser l'expérience.

Leur étrange surnom "La Collaboration Caravane" a survécu depuis les débuts du projet en 1965, même si l'origine du nom s'est effacée des mémoires et si le gtoupe s'est joint plus tard un sixième laboratoire. Cette collaboration a persisté pendant vingt ans, d'une part à cause de l'exceptionnelle longévité de

COS-B mais aussi parce que la Caravane s'est chargée du long dépouillement des données, Le but de cet exposé est justement de rappeler pourquoi cette entreprise fut de longue haleine et en quoi l'analyse des données de COS-B est complexe.

"U C0U1B0RATI0N CJS1VJNE":

Laboratory for Space Research Leiden, Leiden, The Netherlands Istituto di Fiska Cosiica del CIO., Nilano, Italy Istituto di Fisica Cosiica e Inforiatica del C.N.t., Paleno, Italy Nax-Planck-Institut fur Physil und Astrophysil, Institut fur Eitraterrestrische Ph ysi h. Garching pei Hûnchen, 5enan» îerïice d'Astrophysique, Centre d'Etudes nucléaires de Sacl.ay, 5if/»yett», Franc* Space Science Department of the European Space Agency, ESTEC, Noordnijli, The Netherlands

Une première cause évidente vient du petit nombre de "bons" photons capturés comparé au grand nombre de photons de bruit qui polluent les données. Les problèmes statistiques inhérents à la nature du rayonnement observé seront abordés au prochain paragraphe. Une autre source de complexité provient directement du principe de fonctionnement du télescope. C'est pourquoi on peut espérer que ces problèmes s'atténueront, ou mieux disparaîtront, avec les nouvelles générations d'instruments. Le télescope gamma de COS-B est construit autour d'une chambre à étincelles (SC sur le schéma) dans laquelle les photons sont convertis en paires électrons/positrons. Ces particules ainsi créées filent au travers de la chambre et sont perçues à leur sortie par le "télescope à - ID -

déclenchement" (Bl + Cl + B2). Comme son nom l'indique, ce système de compteurs

commande alors le déclenchement de la chambre, c'est à dire la visualisation des

trajectoires des particules par des étincelles. La géométrie de ces traces

permet de reconstruire approximativement la direction incidente du photon

initial. Les particules vont ensuite se perdre dans un calorimètre (E ••• D) qui,

en absorbant leur énergie, renseigne sur l'énergie du photon qui leur a donné

naissance. Toutefois, pour ne pas déclencher inutilement la chambre sur le

passage d'un rayon cosmique (les protons cosmiques, par exemple, sont 10* fois

plus nombreux que les gamma), l'ensemble du détecteur est niché dans un dôme de

plastique scintillant (Al). Ce système d'anticoïncidence détecte le passage des

charges électriques et inhibe en conséquence le fonctionnement du télescope

pendant un bref instant. Ainsi COS-B a-t-il passé une partie non négligeable de

son temps à attendre l'autorisation d'observer. L'ensemble de ces équipements

constituent un télescope gamma sensible au dessus de 40 HeV. Sa gamme

d'énergie utile s'étend en fait de 50 MeV à 5 CeV et son champ de vue a environ

30° de rayon bien que les données au delà de 20° aient été rarement employées.

Contrairement aux instruments traditionnels de l'astronomie qui fournissent des flux, des intensités ou des images, COS-B nous délivre après traitement des photons individuels, chacun portant une étiquette avec l'instant de sa capture, son énergie et la direction d'où il provenait. Malheureusement, seul l'instant est connu avec précision (

Hais puisque certaines courbes illustrent bien les difficultés ou la fiabilité des analyses effectuées dans cet ouvrage, elles ont été réunies ci-dessous, Une partie de ces figures (4 à 6) montrent les caractéristiques du télescope dans les trois gammes d'énergie fréquemment utilisées: 70-150 MeV, 150-300 MeV et

300-5000 MeV.

COS - B iZ -

1) Surface sensible effective en fonction de l'énergie et de l'inclinaison de* photons. 2) Paramètres de la réponse angulaire en fonction de l'énergie et de a<= l'inclinaison: PSF (6) = exp {-(e/e0) }. 3) Dispersion en énergie relevée pour des faisceaux de photons d'énergie et d'inclinaison données.

CD C tï

—r -i i r i 11 ----i i T i r i i i | T" 1 ' . 0.8 ^\^^ • - •0. 0.66 —'- - 04 "

•< 1 3 8° p. inclination: . —- 0» - b" \\ 10° - XN 20° " '*. - \ W ?0 ^is " - ^*s^'- - - ^^^firT^TSTJ- " 0° 1 I 1000 100 EIM*vl

(3) 4) Surface sensible effective dans les gammes d'énergie 70-150, 150-300 et 300-5000 HeV en fonction de l'angle d'inclinaison et de l'indice spectral gamma. 5) Réponses angulaires différentielles pour les trois gammes d'énergie 70- 150, 150-300 et 300-5000 MeV. 6) Probabilité d'attribuer un photon d'énergie incidente E a l'une des trois gammes d'énergie 70-150, 150-300 et 300-5000 HeV. (O (5)

|- i — | -i -r r 1

• loi 70-150 MiV - Ibl 150-300 MlV - le) 300-5OOOH* . 1 . (cil - 1 - . 1 . .ibl \ *\\ - Jol >

. i . I . I . i

r-^—r i-i—, | •"' f'i f

:=== • \. l, — N «• ^y / 1 - «4 -

A la suite de claquages répétés, le mélange de gaz rares d'une chambre à

étincelles vieillit rapidement. C'est pourquoi il avait été prévu sur COS-B de

changer ce gaz régulièrement. Au début de la mission, celui-ci fut remplacé à

intervalle de quelques semaines, puis de quelques mois afin d'économiser les

réserves de gaz (COS-B n'était sensé vivre que deux ans). Le vieillissement du gaz e~ son renouvellement ainsi que des problèmes épisodiques dans le fonctionnement du détecteur ont eu pour conséquence de faire varier l'efficacité absolue du télescope au cours de sa vie. Ces changements ont pu être mesurés grâce à la stabilité de l'émission galactique diffuse en comparant les cartes de régions célestes observées à plusieurs mois d'intervalle. La plupart des observations de COS-B se recoupant ainsi. Strong et al. (1987. Astr. Ap. Suppl.,

67. 283) ont pu estimer à 10% près l'efficacité relative de chaque période d'observation (figure 7). Cette tâche se complique du fait de l'évolution simultanée de l'intensité du bruit de fond. Ce dernier est principalement composé de véritables photons gamma et non de fausses traces dans la chambre.

Mais ces photons n'ont pas une origine "astronomique". Ils proviennent de l'interaction des rayons cosmiques avec l'ensemble du télescope et de son satellite. Aussi l'évolution du taux de bruit de fond était-elle supposée suivre les effets de la modulation solaire au fil des ans. Les résultats obtenus par

Strong (figure 8) ont en effet conduit à une évolution du bruit dans COS-B similaire aux variations de flux des rayons cosmiques dans le sytème solaire

(Voyager 1 et 2, Pionneer 10 et 11. IMP 8). Il est clair que les évolutions de la sensibilité et du taux de bruit de fond étaient deux paramètres vitaux pour générer et exploiter la carte complète de l'émission diffuse vue par COS-B

(chapitre 4) et mettre en évidence la variabilité à long terme de l'émission gamma des pulsars (chapitre 5). _ ir -

; >- H 1—4

> 0 0 .8 ' © o « o f- o ©° CO # V .• * ° * 0° § [ « eo a " s o en L o .4 u r © > < .3 r UJ ' ' • • • • f f • • • • f ,.... i...... t ••. 0. r 0 9 10 15 20 23 30 39 43 90 99 W OBSERVATION NUMBER

- figur» 7 -

- Figura S -

I Z :_ J o . s S.1 I—I 1 few** < 3 1 ^^ > a z o OS ca 11.11 'Ml LU-UJ nihil < 10 19 » 2* 10 IS DQ OBSERVATION NUMBER - IS .

2.2 QUELQUES METHODES D'ANALYSE

La complexité du comportement du télescope et la rareté des photons

détectés ont nécessité le développement de méthodes d'analyse toujours plus

sophistiquées dans les trois domaines de:

- l'étude de l'émission galactique diffuse par corrélation avec divers

traceurs du milieu interstellaire.

- la recherche systématique de sources ponctuelles,

- l'analyse temporelle et spectrale de sources isolées.

Pour chaque application, les techniques ont évolué afin d'améliorer la

précision, la sensibilité et la fiabilité des résultats. Aujourd'hui, la plupart

repose sur le principe statistique de vraisemblance maximale. Cette notion a été

fort heureusement introduite dans la Collaboration Caravane par A. Pollock pour

rechercher des sources gamma extragalactiques au dessus d'un bruit uniforme

(Astron. Astrophys. 94. 116. 1981). Elle a été ensuite appliquée à de nombreuses autres analyses.

Elle consiste d'une manière générale à décrire l'émission gamma réelle d'une région du ciel à l'aide d'un modèle, à dégrader cette émission compte tenu des performances de COS-B et à comparer l'émission gamma ainsi "prédite" à celle effectivement observée. On ajuste alors les paramètres du modèle pour reproduire au mieux les données et aboutir ainsi à la description du ciel "la plus vraisemblable". Dans la pratique, la notion de vraisemblance est mesurée par la probabilité d'obtenir les photons observés à partir du modèle choisi. Bien sûr. ce dernier dépend essentiellement du contexte de l'analyse, des limites du problème et du niveau de sensibilité désiré. Cette approche présente malheureusement un défaut majeur. Il n'est en général pas possible de juger de la qualité absolue d'un modèle. La seule information que nous livre la vraisemblance est relative: un modèle rend mieux coopte des données qu'un autre, mais tous deux peuvent être fort éloignés de la réalité. Par contre, les valeurs - I* -

de vraisemblance obtenues pour les deux modèles permettent une comparaison

quantitative de leurs qualités respectives (voir les articles des chapitres >* et

5 et le livre "Statistical Methods in experimental physics" par Eadie et al.,

North Holland, 1971).

Je me suis attachée ci-dessous à exposer les principales méthodes employées

aujourd'hui pour analyser les données de COS-B. en mettant l'accent sur celles

qui reposent sur la puissante notion de vraisemblance et qui ont été utilisées

dans cet ouvrage. Mais l'exposé n'est en aucun cas exhaustif. D'autres approches

sont en cours de développement (comme la recherche de sources selon le principe

de "Bootstrap sampling"). De même, d'autres axes de recherche existent (étude de

la polarisation des photons, des spectres de l'émission diffuse, recherche de

périodicité ou de variabilité temporelle).

a) Analyse de l'émission gamma galactique:

Dans ce cadre, la méthode de vraisemblance maximale a rapidement supplanté

les analyses traditionnelles de corrélation entre les rayons gamma diffus et le milieu interstellaire à cause des faible^ comptages exprimés dans les cartes

gamma et du grand nombre de cartes à combiner (par exemple HT, CO et gamma). Une analyse traditionnelle provoquait aussi l'exclusion des larges régions

contaminées par des sources brillantes. Après un essai sur les données plus

simples de SAS-2, le principe de vraisemblance maximale fut donc appliqué à COS-

B par Lebrun et al. (1982, Astron. Astrophys. 107, 390) pour l'étude du milieu

interstellaire local. Dans ce contexte le modèle testé est simple. Il comprend un bruit de fond instrumental isotrope (Ib) plus l'émission diffuse née de l'interaction d'un flux uniforme de rayons cosmiques et du gaz interstellaire

(NM). Ce dernier est tracé par les cartes de comptage de galaxies. On peut alors prédire des cartes d'intensité gamma dans plusieurs gammes d'énergie DE du type:

1 1 -2 Ip,.a(l.b,DE) = (q/

q étant 1'émissivité gamma du gaz. Il suffit de prendre en compte les

performances de COS-B en effectuant la convolution de la carte interstellaire

par la réponse impulsionnelle de COS-B (ou PSF pour "point-spread-function")

dans la bande d'énergie considérée. Cette convolution est marquée par un

"tilda". Puisque l'aspect spectral n'est pas primordial dans ce genre de

corrélation un indice spectral moyen de l'émission gamma est utilisé pour

calculer les effets de la résolution spectrale de COS-B sur la PSF applicable à

DE ainsi que sur la sensibilité relative du télescope dans cette bande.

A partir des cartes d'intensité gamma prédites on peut alors produire des

cartes de comptage de photons attendus:

N„«a(l,b,DE) = T fi SA(DE,e) !„,.*(l,b.DE) (2)

où SA représente la surface sensible effective du télescope dans la gamme

DE choisie et dans la direction du pixel considéré (vu par COS-B sous l'angle

6). SA provient donc de la surface réelle de détection et de l'efficacité du détecteur. fi et T représentent l'angle solide et le temps d'exposition du pixel.

La taille des pixels des cartes est bien sûr choisie comme un compromis entre le désir d'avoir de nombreux pixels pour contraindre la corrélation et le souci de garder des comptages raisonnables dans chacun d'eux. Généralement, ces comptages sont faibles. Aussi peut-on faire appel à la statistique de Poisson

pour comparer les photons prédits (NP) à ceux effectivement observés (No). Dans chaque pixel:

N P(N„, No) = (N,,)"» e- P / (No!) (3)

Chaque photon étant indépendant, la vraisemblance L d'un modèle est mesurée par le produit des probabilités obtenues pour chaque pixel. Ce n'est autre que la probabilité de reconstruire la carte observée à partir du modèle considéré.

En jouant sur les paramètres qui décrivent le modèle (ici 1'émissivité q et le bruit Ib), on peut optimiser L(q.Ib) et obtenir le modèle le plus vraisemblable. _ IS _

Les barres d'erreur autour de ces paramètres sont déterminées en calculant les

rapports de vraisemblance:

maxxb L(q.Ib) max„ L(q,Ib) l(q) = l(Ib) =

max^.xt. L(q.Ib) max<,.Te, L(q.Ib)

Puisque la quantité -2 Ln(l) est distribuée comme Xax (Eadie et al.), les

niveaux la et 2a correspondent à une diminution de L telle que - 2 Ln(l) = 1 et

- 2 Ln(l) = 4 respectivement.

Cette méthode d'analyse offre l'avantage de s'adapter facilement à des

contextes astrophysiques plus complexes. On peut ainsi remplacer les cartes de

comptage de galaxies par des observations HI et CO. Ceci permet de mesurer

indépendamment les émissivités gamma du gaz atomique et des nuages moléculaires,

pour peu que leurs distribuitions spatiales, une fois convoiuees par la PSF de

COS-B, soit nettement différentes. Sinon la distinction n'est plus possible

statistiquement. Ce type d'analyse permet également dans certaines conditions de

calibrer le rapport N(Ha)/WCO de conversion des données CO en colonnes de

densité d'hydrogène moléculaire. En effet, si on écrit:

1 _l 2 Ipr«,a(l.b,DE) = A. NHlQ.b) + E. WCO(l.b) + Ib en s" sr cm" (4)

A représente directement l'émissivité q«i/47i du gaz atomique et B est relié

à l'émissivité de l'hydrogène moléculaire selon: B = 2 (qH2/4;i) (NMa/WCO). Un

tel modèle suppose évidemment une contribution négligeable des sources gaoma compactes et une répartition uniforme, des rayons cosmiques dans la région

étudiée. Si, de plus, ce flux est équivalent dans les régions HI diffuses et dans les nuages moléculaires plus denses, le rapport N(H*)/KCO devient égal à

B/2A.

Il est enfin possible de lever l'hypothèse sur les sources gamma compactes et l'isotropie locale des rayons cosmiques en incluant des termes source (des fonctions de Dirac) et des gradients d'éaissivité gamma dans l'équation (4). _ ZO _

Ceci permet alors d'explorer l'ensemble du disque galactique simultanément, sans

exclure les régions contaminées par de brillantes sources. L'importance de la

base de données de COS-B permet de découper la Galaxie en anneaux concentriques

afin de mesurer de faibles variations galactocentriques de l'émissivité gamma

(voir chapitre 4) que l'on relie à la présence de gradients de densité des

rayons cosmiques. Un brillant traitement mathématique a été mis au point par A.

Strong pour ajuster simultanément un grand nombre de paramètres (1985. Astron.

Astrophys. 150. 273).

b) Recherche de sources gamma ponctuelles

Avec COS-B est né le problème de la recherche de sources gamma ponctuelles.

Isoler et identifier une source est en soi passionant. De plus à l'échelle

galactique, la recherche systématique des sources apporte des indices sur la population toute entière (à défaut d'identification personnelle!) et permet une analyse "propre" de l'émission gamma véritablement diffuse. L'importance du problème s'est révélée avec l'analyse rapide de la structure fine des premières cartes de COS-B (Herœsen. Nature 269. 494, 1977) qui a mis à jour 13 sources gamma. La tâche est ardue pour de multiples raisons. La rareté des photons en est une. Hais le plus difficile consiste à isoler un excès d'allure semblable à la PSF au dessus d'un bruit de fond fort et structuré puisqu'il s'agit de la combinaison de l'émission galactique diffuse et du bruit instrumental. La première méthode d'analyse imaginée revient à corréler la carte d'émission gamma avec la PSF du télescope pour la gamme d'énergie sélectionnée, toute corrélation positive étant considérée comme une source. Cette analyse a été testée et présentée par W. Hermsen (thèse, Université de Leiden, 1980). Elle a abouti au catalogue 2CG (Swanenburg et al.. 1981, Ap.J., 243, L69). Malheureusement, toute structure de l'émission diffuse de dimension inférieure à 1 ou 2 degrés (selon l'énergie et les conditions d'observation) contrefait une source ponctuelle, La - 2.1 -

sensibilité avec laquelle on détecte un excès n'est pas homogène dans le disque

galactique selon l'intensité et l'allure de l'émission diffuse sous-jacente.

Enfin, la séparation de sources angulairement proches (moins de quelques degrés)

est malaisée avec cette méthode.

Entre-temps, l'analyse (par la notion de vraisemblance) de l'émission

diffuse a démontré que l'on pouvait reproduire empiriquement les cartes gamma à

partir d'une combinaison de cartes HI et CO et d'un bruit isotrope. On peut

alors rechercher tout excès ayant une distribution spatiale semblable à la PSF

au dessus de l'émission diffuse prédite par une telle combinaison. Cette

approche permet une recherche de sources beaucoup plus poussée. La sensibilité

est meilleure et plus uniforme. Elle rejette toute accumulation connue de gaz

baigné par un flux normal de rayons cosmiques. Par contre, elle isole en plus

des véritables sources compactes toute région interstellaire (étendue ou quasi-

ponctuelle) d'émissivité anormale. Cette technique a été mise au point par A.

Pollock (Astron. Astrophys. 146, 352, 1985). Elle consiste à calculer la

probabilité d'observer un photon dans une direction (l.b) à partir du modèle

suivant:

I„.a(l.b.DE) = r» 6(l-la,b-b») + A. NHI(l.b) + B. WCO(l.b) + C (S)

Le premier terme représente la source de flux rs à la position (ls. b,). Le

reste de l'expression prédit l'émission diffuse et le bruit. A, B et C

proviennent de l'analyse galactique préalable ou peuvent être ajustés lors de la

recherche.

Afin d'augmenter la sensibilité de l'analyse, les photons observés ne sont

pas rangés dans des cartes mais ils sont exploités individuellement. On ne

prédit donc pas des comptages par pixel mais on calcule directement la

probabilité d'observer un photon dans une direction selon:

p(l.b) = dfl Jj I„„„(l'.b') PSF(l,b,l',b') SAO') dn' (6)

La PSF, ou probabilité de reconstruire à l.b un événement venant de l',b', est déterminée pour une gamme d'énergie donnée (et un indice spectral moyen), _ 42 _

Mais pour simplifier l'expression (6), les variations de la PSF avec l'angle

d'incidence 6' sur le télescope ont été négligées. Par contre, les effets de

l'inclinaison 6' et du choix de la gamme d'énergie ont tous deux été pris en

compte pour déterminer la surface sensible effective SA.

La normalisation de cette probabilité et le formalisme du modèle conduisent

à l'expression:

r„ SAO») PSF(l,b,ls,bs) + SA(9) [A.NHI(l,b)+B.WCO(l,b)+C| p(l,b)= (7)

r» SA(6») J"J" PSF dfl + J*j" SA(0) [A.NHI(l.b)+B.WCO(l,b)+Cl d«

où 9= est l'angle entre la source et l'axe de COS-B, le tilda marque la

convolution des cartes interstellaires par la PSF.

La vraisemblance du modèle (r=. ls, b») est alors définie comme le produit

des probabilités estimées pour chaque photon observé par COS-B. Il faut bien sûr

établir un seuil statistique de détection d'une source. Soit:

1(1».b=) = Ln L(r„«,l,,bs) - Ln L(0,la,b=)

1 est nul en l'absence d'excès (émission diffuse seule). Par contre, 1

augmente en présence d'un excès à (1«, b3) et passe par un maximum pour le flux

r»=r,o«jt le plus "vraisemblable". Puisque trois paramètres sont laissés libres

2 dans la recherche de la source (rs, ls et bs), 2.1 suit une distribution en X 3

(Eadie et al.). Ainsi une valeur de 2.1 > 12 correspond à une probabilité < 0.01

que l'excès provienne d'une fluctuation aléatoire lors de l'observation de

l'émission galactique diffuse seule.

Déjà appliquée au premier et au deuxième quadrant galactique et à la région

de la Carène (Pollock et al., 1985, I.C.R.C, La Jolla, 1, 338) au delà de 300

MeV pour profiter de la meilleure résolution angulaire, cette analyse a permis de réviser une partie du catalogue 2 CG, d'isoler une source double dans le

Cygne (2CC078+01 et 2CG075+00) et de repérer des sources variables à long terme.

Employée à d'autres énergies dans tout le disque galactique, elle aboutira

bientôt au catalogue final des sources gamma de COS-B (3CG). _ 23 -

c) Analyse temporelle et spectrale de sources isolées

Lors de l'analyse de la luminosité de sources brillantes, l'énergie des

photons devient un paramètre crucial et les performances du télescope selon

cette énergie doivent être prises en compte avec précision. Le principal écueil

auquel se heurte ce type d'analyse provient du couplage des informations

"direction" et "énergie" de chaque photon (voir section 2.1). Il convient

ensuite de séparer correctement la source de toute émission sous-jacente

(galactique + bruit). L'analyse spectrale de sources gamma a été tout d'abord

tentée sur les pulsars en imaginant la méthode dite "de saturation" (Bennett et

al., 1977, Astron. Astrophys. 61, 279). On utilise pour cela la signature

temporelle de l'astre. Les instants d'arrivée des photons concernés sont

convertis en phase par rapport à la rotation du pulsar afin de construire la courbe de lumière de l'émission gamma puisée. Puisque le bruit galactique et

instrumental est stable dans le temps, ses photons se répartissent uniformément

à toutes les phases de la courbe de lumière. 11 en est de même pour une

éventuelle émission constante (par opposition à puisée) en provenance du pulsar,

On peut donc soustraire la composante plate de la courbe de lumière pour ne compter que les photons "puisés" qui viennent nécessairement de la source. Ce travail est effectué dans plusieurs gammes d'énergie. Pour augmenter la fiabilité du résultat, le nombre de photons puisés N„ est déterminé en ouvrant peu à peu l'angle d*acceptance des photons autour de la source. La distribution

des Np ainsi calculés suit la PSF intégrale de COS-B pour chaque gamme d'énergie sélectionnée. Elle "sature" lorsque tous les photons puisés ont été inclus. Les valeurs N„ obtenues à saturation pour chaque énergie sont ensuite converties en flux et en spectre connaissant la variation de la surface sensible effective de

COS-B avec l'énergie et sa résolution spectrale (voir thèse de W. Hermsen),

En l'absence de signature temporelle, on ne peut compter que sur les courbes de saturation pour isoler la source du reste de l'émission. De plus, on - 2H -

procède par comptages (dans des pixels aussi bien que dans des intervalles

d'énergie) et sans tenir simultanément compte des résolutions spatiale et

spectrale. C'est pourquoi la méthode de saturation est assez peu sensible.

Une fois encore, la noion de vraisemblance est venue à notre secours. Pour

séparer la source du bruit, un modèle similaire à celui de la section (b) a été

imaginé. Il comprend une source ponctuelle, une émission galactique diffuse dont

la structure est prédite par des cartes du milieu interstellaire ainsi qu'un

bruit instrumental isotrope. Cependant, l'aspect spectral est pris en compte en

explicitant le spectre en énergie de chaque composante du modèle. En cas de

périodicité de la luminosité de la source, il faut également introduire dans le

modèle la notion de phase des photons. Bien sûr, l'émission en dehors de la

source est toujours constante. Le modèle général s'écrit donc:

n(*.E.l,b) =

c Fp(*) E"°' " 6(1-1»,b-bs) pour l'émission puisée de la source aa + F* E" ô(l-l0,b-bs) pour l'émission constante de la source + BKI E"ox pour le bruit isotrope et constant + q E~oa NH(l,b) pour l'émission galactique diffuse.

n(*.E.l,b) est alors la densité de photons réellement émis dans le ciel par

unité de phase, d'énergie et d'angle solide.

Les performances de COS-B sont pleinement caractérisées par les trois

fonctions suivantes:

- SA(E.e) sa surface sensible effective en fonction de l'énergie et de

l'angle d'incidence sur le télescope.

- PSF(E,6.cO la probabilité de reconstruire un photon (d'énergie E

arrivant sous l'incidence Q) à un angle am de sa veritable direction.

- ER(E,e,E„) la probabilité de mesurer un photon (d'énergie E arrivant sous l'incidence 0) à l'énergie E».

L'indice m désigne toujours le résultat d'une mesure par COS-B. Il reste

donc à calculer la probabilité de détecter un photon à l'instant tm (ou la phase - zr.

*m), à la position In, b„ et à l'énergie E™. Puisque la résolution temporelle du

satellite est excellente, on peut écrire tm=t ou 4w=$. Alors:

B lb f JT n(*.E,l,b) dE dn CAL(E.01B,amib,E™) p(*.E„,l„,.b„) = - (9) J""8™ JT1"""" numérateur dE™ â£U

où CAL est le produit de SA, PSF et ER. Chaque point-source élémentaire

(l.b) du ciel envoie des photons sur COS-B sous l'incidence 9ib. La probablité

d'en retrouver un à (1„, b„) dépend de l'angle a.11, entre les directions (l,b)

et (lm,b„). Les intégrales portent sur tout le champ de vue de COS-B (60° de diamètre) et toute sa gamme d'énergie utile (50-5000 MeV) car au delà sa

sensibilité est négligeable ou nulle. La normalisation fournit une condition aux

limites. En effet, le dénominateur de l'expression (9) n'est autre que le nombre total de photons "détectables" dans tout le champ de vue et à toutes énergies à partir du modèle testé. Il doit être égal pour tout modèle au nombre total de photons détectés durant la période d'observation.

La vraisemblance du modèle est alors définie comme le produit des probabilités obtenues pour chaque photon (*.E™,lm.b™) observé par COS-B.

Plus d'un an m'a été nécessaire pour mettre au point cette analyse ou, plutôt, pour optimiser ces calculs. Pour donner un exemple de leur complexité,

50 10e octets sont requis pour décrire la fonction CAL (car il n'existe pas de forme analytique simple) et le modèle général pour Vela comporte 16 paramètres.

Mais la finesse des résultats qu'elle a apportés sur le comportement de Vela m'a largement récompensée (voir chapitre 5). Cette méthode est en effet extrêmement sensible. Contrairement à celle dite "de saturation", elle peut analyser les photons détectés à chaque observation indépendamment pour étudier la variabilité à long terme des flux et des spectres d'une source. Sa sensibilité m'a également permis de subdiviser chaque observation et d'étudier indépendamment chaque phase du pulsar,

Appliquée dans l'avenir au pulsar du Crabe, cette méthode offrirait en plus - 26 -

de l'analyse par phase une mesure très précise du spectre de l'émission gamma

non puisée du Crabe que l'on attribue probablement à sa nébuleuse (voir chapitre

5). Enfin malgré sa lourdeur informatique, cette analyse est toute indiquée pour

étudier les sources non puisées pour lesquelles les méthodes traditionnelles

sont peu fiables. N'ayant plus besoin de diviser les données par intervalles de

phase, tous les photons d'une observation peuvent être exploités simultanément

pour déterminer en détail la forme du spectre en énergie. Il n'est plus

nécessaire de se restreindre à une ou deux lois de puissance. Et si plusieurs

observations sont disponibles, on peut toujours étudier la variabilité à long

terme de la source. Geminga est un candidat parfait pour ce type d'étude afin de cerner sa nature physique. L'analyse est en cours. Enfin, la grande sensibilité de la méthode en fait un outil indispensable pour étudier les sources gamma moins brillantes. Paradoxalement, le terme source du modèle peut être abandonné.

On obtient alors une méthode puissante d'analyse des spectres de l'émission diffuse. Msiis il s'agit là d'une autre histoire ... et d'une autre thèse. -IU­

LE MILIEU INTERSTELLAIRE

UUUcm HzMdn.it, 17SS 3.1 Les mages moléculaires dam la Galaxie

Cette partie repose essentiellement sur les cartes CO que j'ai obtenues en sept mois d'observations quotidiennes avec le télescope millimétrique de l'Université de Columbia à New York. J'ai tout d'abord montré les cartes du premier quadrant car elles couvrent près du quart du disque galactique. Elles ont été présentées au Congrès de "l'A. A. S." en 1985 et feront l'objet d'un article de l'Astrophysical Journal cette année. Leur étendue en latitude est suffisante pour contenir la plupart des nuages de la ceinture de Gould dans cette direction. Mais elles révèlent surtout des indices nouveaux sur la structure de la Galaxie externe, comme le prolongement en CO du bras de Persée dans le premier quadrant et la découverte d'un impressionnant complexe (10 à 20 millions de masses solaires) formant une partie du bras extérieur de la Galaxie (15 kpc) déjà perçu en NI et en radio. La présence de molécules CO dans ce bras avait été détectée par Kutner et Nead, puis niée par Solomon et al. Aussi est-elle aujourd'hui confirmée mais également cartographie». Une telle découverte suggère que la périphérie de la Galaxie contient d'importantes quantités d'hydrogène moléculaire qui demeurent ignorées, sans doute parce que ces nuages lointains sont plus froids et diffus qut leurs cousins des régions internes. Leur métallicité >tt aussi probablement plus faible, auquel cas les molécules CO sont un traceur moins efficace. La recherche systématique de tels nuages et l'analyse de leurs caractéristiques physiques apparaissent donc essentielles. Enfin, la structure de la Galaxie externe n'étant pas aussi confuse qu'à l'intérieur du cercle solaire, la disposition et la forme des bras y sont plus faciles à apprécier. En offrant un contraste plus marqué qu'en HI, cette carte CO du bras externe apporte quelques premiers détails et des observations plus poussées, en sensibilité et en résolution, sont prévues dans un avenir proche. J'ai ensuite présenté la carte CO complète du disque galactique qui a été obtenue en combinant l'ensemble des observations du "Columbia Sky Survey" avec celles que j'ai effectuées dans le premier quadrant et Céphée (Dame et al., 1987, Ap. J. 322, 706). La comparaison avec d'autres traceurs montre que cette carte contient la plupart des complexes moléculaires attendus (sauf peut-être à la périphérie de la Galaxie ...). L'ampleur, la sensibilité et la finesse de ce panorama sont donc des atouts indéniables pour étudier les structures à grande échelle du milieu moléculaire et les caractéristiques ou la disposition des nuages proches.

Le dernier article (à paraître dans l'Astrophysical Journal) présente les résultats d'une étude du milieu interstellaire proche dans Céphée. Les observations CO révèlent deux nouveaux nuag»« du bras Local (800-900 pc) ainsi que Céphée et Cassiopée qui forment l'un des plus importants complexes de notre voisinage (300 pc). Ce dernier est situé à environ 50 pc au dessus du plan de la ceinture de Gould tel que nous le connaissons. Pourtant, par sa distance et la vitesse des nuages qui le constituent, le complexe semble bien appartenir à la ceinture ... Serait-elle gauchit ou inclinée un peu différemment! La question demeure en suspens. Enfin, les nuages de ce complexe apparaissent façonnés par une large "bulle" mise en évidence i diverses longueurs d'onde (X, radio, CO, HI) et qui représente vraisemblablement le reste d'une supernova ayant explosé près du Soleil voici 40 000 ans. Le choc semble interagir aujourd'hui avec le nuage massif de Céphée, fournissant ainsi un excellent candidat pour analyser en détail ce type d'interaction. - 2* _

A LARGE SCALE CO SURVEY OF MOLECULAR CLOUDS IN THE FIRST GALACTIC QUADRANT

I.A. Grenier1, T. Da*ea and P. Thaddeus2 1- Service d'Astrophysique, C.E.N. Saclay 2- Harvard-Smithsonian Center for Astrophysics

A large scale survey of molecular clouds in the first galactic quadrant has been carried out from 12° to 100° in longitude and - 10° to + 11° in

latitude in the 12C0 line at 2.6 mm using the 1.2 m Columbia Sky Survey telescope of Columbia University in New York City. The "superbean" technique, which reduces the nominal 8.7° resolution of the antenna to a 0.5°x0.5° square bean, and the use of a highly sensitive SIS receiver allowed to fully cover a very large area (= 1700 deg*) with a good sensitivity (0.28 K rms) and angular resolution (0.5°) during the 1984-1985 winter.

CO emission was detected froo about half the region mapped, It cones almost equally from local dark clouds belonging to the Great Rift in the Milky Way and fron distant clouds in the inner spiral arms of the Galaxy (4 to 7 Kpc from the galactic center). This partition is illustrated on figures 1 and 2. The first shows the entire CO content of the galactic disc in the first quadrant. The emission has been integrated over all velocities (-100, +140 km.s"1). The inner arms appear as the intense ridge along the plane up to 1=45°, On figure 2. data from -10 to +34 km.s"1 have been integrated to produce the map of the Great Rift clouds. On both figures, three maps are displayed. The first shows the old survey by Dame and Thaddeus (1985) (Ap. J. 297, 751). which had a spatial resolution of 1°, while the third nap presents the new survey with its higher sensitivity and resolution (0.5°). The intermediate map represents the new data - zt -

smoothed to 1° to compare with the old nap. The obvious general agreement

demonstrates the reliability of the observed features.

The longitude-velocity diagram integrated over latitudes smaller than 5° on

figure 3 reveals the complex dynamical structure of the inner molecular clouds

tracing the spiral arms, while the feature that runs through the plot around v =

10 km.s"1 arises from the Creat Rift clouds.

The coincidence of the nearby CO clouds with the dark nebulae which obscur

the northern Milky Way is striking (figure 2). The agreement has even improved

with the better quality of the present survey, so that nearly all the zones of

heavy obscuration do correspond to the presence of molecular clouds. The Great

Rift has been resolved into distinct clouds located from 200 to 2300 parsecs

away (Dame and Thaddeus. 1985), most of them being closely related to the HI

expanding ring observed by Lindblad (1973, Astr. Ap.. 24, 309), that is to the

Gould Belt. Of particular interest is the arc structure of the Rift in the sky

(which runs from Aquila to Cygnus) because at lower longitudes it is prolonged

by the famous Oph-Sag and p Oph clouds (both part of the Belt) and, especially,

because at higher longitudes it seems to connect to the Cepheus clouds (see the

whole galactic map in the next chapter of the thesis). One question raised by

the study of the latter was their position with respect to the Belt. Their velocity suggests that they are linked to the HI expanding ring of the Belt, whereas their elevation above the Belt plane (as traced by the stars) is too high. Their apparently smooth connection with the Rift arc structure gives an other argument in favour of their belonging to the Belt, however surprising it may be since it would then imply a slightly tilted or warped Belt (see Grenier et al., this chapter).

To fully discuss the numerous clouds mapped in this survey is beyond the

scope of this short note and will be the core of the article presenting these data. Interesting new features, however, deserve a few lines for they were not observed in the old survey. _ £3 -

The faint spots at 1= 80° - 100° and v= -70; -20 ka s"1 are in fact real

emission coming from a narrow ridge along the galactic plane. This is why a

longitude-velocity diagram integrated from -1° to +1° in latitude has been

displayed (figure 4). The structured emission between 80° and 100° around v= -40

km s~l is now apparent. It represents a new and long extension of the Perseus

spiral arm. lying 1 to 2 kpc away, well into the first quadrant.

Another feature lies at 1= 70° and v= -60 km s"1 (see figure 4 and 5) and

has been mapped on figure 6. These CO clouds are related to an HI concentration

(figure 7, Burton. 1970, ASA Suppl, 2, 261). Furthermore, in 1981 (ApJ, 249,

L15). Kutner and Head have suggested from CO sampling observations that the

emission running around -65 km s"x at 1= 55° to -75 ka s-1 at 1= 95° belongs to

a well-defined arm 15 kpc from the galactic center. Later, Solomon et al. (1983.

ApJ, 267, L29) failed to reproduce the detection in CO of this distant arm.

However, the present data seem to confirm the presence of molecular gas in these remote regions. Placing these clouds in the 15 kpc arm at 1= 70° implies a distance from the Sun of 15 kpc too. So the present clouds stretch along ~ 3 kpc of the arm. From the study of the diffuse gamma-ray emission (Strong et al.,

1988, Grenier and Lebrun, 1988) we may further assume a N(H2)/WC0 ratio of 2.3

10io aolec. cm-1 K"1 km"1 s implying an extremely large mass for this complex of the order of 1 to 2 10' solar masses (for a distance from the Sun of 10 to 15 kpc)!

Such impressive values are still questionable, but the existence of this emission is real. The discovery of these giant molecular clouds in the survey clearly urge deeper observations in CO in order to map their faint extensions, to probe how far in longitude they stretch and to confirm their remote position in the Galaxy. CO: ALL VELOCITIES • 111 . j..., |.,. ,,.,.. 1 .... 1 ... . | i i i i | i i i 1 ' " 'I ' ' "1 ""1 • ' "1 " • BEAM 10* Ka) l op f^T ; t- 5*

î—i—i—i_ Figure 2: Contour interval is 4 K km s"

71 i r" . ,-r. , , ,.. . "i""i"' - • i • • • • i • • • • i • " 'T'-"-| •—"!••- • 10* 0 0 .=>% =7 °Aj t-^S^PV^ 5» \ ° O ®

0*

&& . , I. I ^0 o O • <ËJ yTs& O - (/kiJlJ -5 o J — i — 1....1....1... .i . 1.... 1.... 1... . i...j. ri... s^. i.^. . 1, , , ,1, ,, .1 , , ^.'

îîîîïîuîîïïïuîîîîîîîïînîmzîiïtnuuîii• ' INTEGRATED TEMP 0.5 SUPERBEAM 03/22/85

Contour values: 3.5by 3.5 |( Integrated over: -5.0 to 5.0 i") Figure 3 Mapping off Interpolation on 5-0 -r~r~rT i i i i i i

95-0

85.0

75.0

65-0 i

55.0 .**-.*' à&ffk 15.0

35.0

25-0

15.0

5.0 I I I I I I 1 I I I I I 1 I I I I I I •?" --n.m n-:=r..n -Fn.n -

•1.00 TO 1.00 CONTOUBS: FROM 0.15 BY 0. 15

v-

Figure 4 - 3

3- 0.0 TO 1 .00 0.10 BY

1 1 i i 1 o ro a> — — H— CO o o ru en •c ro . o o km/: o a a o o o o o o o• a o o o a V

Figure 5 _ -iS -

Figure 6:

CONTOURS: FROM 0.75 BT 0.75

J7, i c io I [

v O

O ^ 0 Cb l P l I I en m g> oo at

— I Fï« ure? - 3> -

THE AS7HOPHY51CAL JOURNAL. 322:706-720,1987 November 15 . IW TheAirwnMnAHronomic«JS4ici«ty.Alln|htirewr.«d.PfiBiedinUSA.

A COMPOSITE CO SURVEY OF THE ENTIRE MILKY WAY

T. M. DAME.1 H. UNGERECHTS,1 R. S. COHEN,2 E. J. DE GEUS,3 I. A. GRENIER,* J. MAY,S D. G MURPHY,6 L.-Â. NYMAN,' AND P. THADDEUS1 Received I9H7 February4: accepted 1987 May 13

ABSTRACT Large-scale CO surveys of molecular clouds in the Milky Way, first undertaken with a 1.2 m telescope in New York City, have been extended along the entire Galactic plane with a similar instrument on Cerro Tololo, Chile, By combining five large unbiased surveys of segments of the Galactic plane with 11 surveys of particular local clouds (e.g., the Taurus dark clouds), a composite map 10'-30° wide in latitude has been obtained at an angular resolution of 0?5. It is clear from a comparison with dark nebulae, and from the IRAS far-infrared and COS B gamma-ray surveys that little molecular gas lies beyond the boundary of our compos­ ite map. The full composite survey contains more than 31,000 spectra and fully samples -~7700 deg", nearly a fifth of the entire sky. An I, v diagram has been produced for a subset of the survey, the data from a strip 6?5 wide in b centered on the Galactic equator. The main structural features of the molecular Galaxy are readily apparent at an angular resolution of 0?5. The molecular clouds in the inner spiral arms of the Galaxy, the so-called molecular ring, appear as an intense ridge of emission ~2J wide in b extending -60" on either side of the Galactic center. The /, v diagram reveals large departures from axisymmetry both in the molecular ring and elsewhere; most striking are the large arcs of the Carina and Perseus spiral arms. The local (low-velocity) emission mainly follows Gould's Belt of OB stars, gas, and dust: large clouds in Lupus. Ophiuchus, and Aquila mark the Belt's positive-latitude extension, while the Taurus and Orion clouds, directly opposite in the sky, lie at the negative-latitude extension. The complete longitude and the wide latitude coverage of the composite survey permits, for the first time, a fairly complete determination of the distribution of molecular clouds near the Sun. It has long been suspected from the distribution of dark clouds and rifts that local molecular clouds are more common in the northern Milky Way than in the southern, and the composite survey confirms this quantitatively: the molecular mass within 1 kpc is 4 times greater in the first and second quadrants than in the third and fourth. Furthermore, nearly all the clouds within I kpc in the first and fourth quadrants apparently lie on a fairly straight ridge more than 1 kpc long which may trace the inner edge of the Local spiral arm. The measured rms dispersion about the Galactic plane of the local molecular gas, 74 pc, corresponds to a Gaussian with half-thickness at -3 half-intensity of 87 pc. The mean surface density of molecular gas within 1 kpc of the Sun is 1.3 M0 pc ; on the assumption of a Gaussian : distribution with a half-thickness of 87 pc, the mean midplane density at the 3 3 solar circle is 0.0068 M0 pc" , or 0.10 H2 cm" . Subject headings: galaxies: The Galaxy — galaxies: structure — interstellar: molecules

(. INTRODUCTION sky (Kerr, Hindman. and Gum 19S9); done with inexpensive In spite of great advances in instrumentation, radio astrono­ antennas dedicated to hydrogen-line work for long periods, mical surveys of molecular clouds in the Milky Way since the these completely mapped at - T angular resolution the distant discovery of the ubiquitous tracer molecule CO in 1970 have spiral arms of the Milky Way around the Galactic equator and not proceeded as systematically or as rapidly as the early 21 cm followed local gas to fairly high latitudes ( ± 10°), establishing a surveys a generation ago. A year after Ewen and Purcell's secure foundation for subsequent higher resolution observa­ tions with larger antennas and interferometers that continue to (1951) detection of the 21 cm line, Christiansen and Hindman the present day. (1952) completed a preliminary large-scale 21 cm survey cover­ ing ±50° of Galactic latitude over 270v of Galactic longitude; The first Galactic CO surveys, in contrast, were confined to their angular resolution was poor but their sampling was good, a thin strip along the Galactic equator in the northern hemi­ and most of the important large-scale 21 cm features, including sphere and only a minute fraction of the molecular gas in the the Galactic center. Vela, and Cygnus X regions along the Galaxy was actually observed (Scoville and Solomon 1975; plane and the high-r extensions in Ophiuchus and Taurus are Gordon and Burton 1976; Cohen and Thaddeus 1977). Owing evident in their maps. Within a few years this preliminary work to the small number of millimeter-wave instruments, the small was followed by the Leiden survey of the northern sky (Muller beam widths of most (- O, and the close association of CO and Westerhout 1957) and the Sydney survey of the southern clouds with a variety of cataloged Population I objects, the progress of unbiased CO surveys has been slow, and, until now, no survey comparable in coverage and sampling to the 1 Harvard-Smithsonian Center for Astrophysics. Sydney-Leiden surveys had been undertaken. 1 Department of Physics. Columbia University. The Columbia millimeter-wave telescope in New York and 1 Sttrrewacht. Leiden. The Netherlands. * Service d'Astrophysique. CEN Saclty. France. its Twin instrument in Chile were designed to carry out CO 1 Departamenio de Asironomia.Universidadd; Chile, surveys of molecular clouds north and south in the spirit of the " European Southern Observatory. Sydney-Leiden surveys. The 1.2 m antennas of these telescopes 706 -3«

CO SURVEY OF THE MILKY WAY 707

yield a beamwidth of 8:7 at the frequency of the CO I — 0 line. adjusting integration times) and data analysis (by smoothing) equivalent to a 100 m antenna operating at 21 cm. With these and are not an important source of inhomogeneity in the final two telescopes it has been possible to produce well-sampled composite survey. For the work described here, two super­ surveys of the distant inner Galaxy and of specific local clouds ; heterodyne receivers were used in succession on the New York but to obtain the wide-latitude coverage required for compari­ telescope, the first with an uncooled Schottky diode mixer son with the IRAS and COS B surveys in a reasonable amount (Cong, Kerr, and Mattauch 1979) and a single-sideband recei­ of time, a lower resolution, closer to that of the Sydney-Leiden ver noise temperature of 900 K, the second, installed in the surveys, was required. summer of 1983, with an extremely sensitive liquid helium- Lower resolution was achieved not by stopping down the cooled SIS mixer (Pan 1984) and a single-sideband tem­ antenna (an inconvenient procedure with a CassegrainJ but by perature of 95 K. On the Chile telescope an inexpensive but the simple expedient of sequentially stepping through a square rugged and sensitive liquid nitrogen-cooled Schottky diode array of points about a beamwidth apart during a single obser­ receiver with a noise temperature of 385 K was used. For vation and summing the spectra obtained. Information is spectrometers, both instruments were equipped with 256 acquired at essentially the same rate as with a smaller antenna channel filter banks copied with minor improvements from an whose beam matches the array in size, and, implemented NRAO design (Mauzy L974), that in New York with a fre­ entirely by computer software, the procedure has the practical quency resolution of 250 kHz. or 0.65 km s~ ' at 115 GHz, that merit of allowing the telescope to revert at any time to observa­ in Chile with a resolution of 500 kHz, or 1.3 km s~' at 115 tions at full resolution without modification of feed or antenna. GHz; a second filter bank with 100 kHz resolution was used on the Chile telescope for two small regions. This technique was first used to map the first Galactic quad' rant over more than 10" of Galactic latitude (Dame and Thad- Spectral line intensities were calibrated and corrected for deus 1985); an 8 x 8 array was scanned to yield an angular atmospheric absorption by briefly observing before each inte­ resolution of V. Since the completion of this survey in 1980, gration a room-temperature chopper wheel in front of the feed the array scanning technique has been applied to a number of horn (Penzias and Burrus 1973). The effective temperature and different surveys that now cover most of the Milky Way in a opacity of atmospheric water vapor were determined every 6 somewhat irregular quilt. In this paper we present a synoptic hr, or more often during weather changes, by antenna tipping. view of the entire Milky Way produced from a synthesis of five Intensities were further corrected for main beam efficiency to major Galactic plane surveys and 11 surveys of particular yield radiation temperatures that agree to within 5% with all regions, such as the Taurus dark nebulae. Most are at an data from the New York and Chile telescopes reported since angular resolution of +', the rest smoothed to this resolution. 1980. In particular, the intensities reported here agree with The total composite survey contains more than 31,000 spectra those of the Wide-Latitude survey of Dame and Thaddeus and fully samples - 7700 deg2, nearly a fifth of the entire sky. (1985), used by Bloemen et ai (I986) to determine the ratio of

We will not attempt here either to derive the large-scale dis­ velocity-integrated CO intensity to H3 column density. tribution of molecular clouds in the Galaxy or to Study partic­ Although a recent réévaluation of the beam efficiencies of the ular regions, as that is being done elsewhere; we do derive, New York and Chile telescopes by Bronfman et al. (1987) sug­ however, some important parameters not previously very well gests that this intensity scale may be low by -20% in absolute measured, in particular, the mean density and scale height of units, we retain it here for consistency with the CO mass cali­ molecular gas within 1 kpc of the Sun, and we attempt to bration. locate the main molecular concentrations within this region on Most of the data were taken by position switching against the Galactic plane. one to three reference ("OFF") positions with a period of 30 s; The next section (§ II) describes the two telescopes, the such spectra toward emission-free regions typically were quite observing techniques used for the individual surveys, and how flat, so only linear baselines were removed. In some regions these were reconciled and joined together. In § III the compos­ well off the Galactic equator, however, with only one or ite survey is presented as a spatial map integrated over velocity occasionally two low-velocity CO lines (e.g, the Taurus dark and as a longitude-velocity map integrated over latitude; indi­ clouds), frequency switching by 10 to 20 MHz was used instead vidual clouds and large-scale features of the maps are briefly to increase the speed of data acquisition; higher order base- discussed. In § IV the distribution and properties of molecular tines were generally required to fit such spectra, but, because clouds within 1 kpc of the Sun are investigated and in § V we the lines in such regions are narrow, integrated line intensities demonstrate by a comparison with other Population I tracers were insensitive to the order of the fit. that the composite survey probably provides a nearly complete Within - 5' of the Galactic center intense emission extends inventory of nearby molecular clouds. over more than the 332 km s "l bandwidth of the spectrometer on the Chile telescope, so it was necessary there to observe IL OBSERVATIONS AND ANALYSIS each direction twice, shifting the spectrometer by -100 km s"1 and joining the resultant spectra (Nyman et ai 1987a). a) Observations The New York telescope has recently been described in The Chile telescope is a close copy of that in New York, and detail by Cohen, Dame, and Thaddeus (1986), and the Chile the two together provide the first means of observing molecu­ telescope by Cohen (1983) and Grabelsky (1985); the cali­ lar clouds over the entire sky with similar instrumentation. bration procedures used on both telescopes are discussed by Both telescopes have fast Cassegrain antennas with a primary Cohen, Dame, and Thaddeus (1986), and a detailed compari­ aperture of 1.2 m and beamwidths of 8:7 ±11 at 115 GHz, the son of their calibrations is given by Bronfman et af. (1987). frequency of the I -* 0 rotational transition of CO used as the standard tracer of molecular clouds. The main instrumental b) Synthesis of the Data differences were in receiver sensitivity and spectrometer The New York and Chile telescopes have now mapped a resolution, and these were easily reconciled in data taking (by strip J0°-20o wide in Galactic latitude around the entire - S3 -

DAME ET AL.

Galactic plane, and nearly all of the large, nearby regions of The noise levels of the surveys varied between 0.1 and 0.35 K optical obscuration and star formation at higher latitude as rms per channel. The noise in the composite spatial map there­ well These surveys, five large Galactic plane surveys, and 11 fore varies somewhat from region to region, depending on others are summarized in Table I; the sky coverage of each is channel width, beamwidth, and spectral baseline fit of individ­ shown in Figure i. These surveys were generally fully sampled ual surveys, and, further, on the velocity-integration range and at an angular resolution of 0?5, but some differences in coordi­ the amount of interpolation and smoothing applied. The nate systems, sampling patterns, beamwidths, and velocity lowest contour level for the composite map, S K km s" \ lies at resolution exist. We discuss below the general procedures used about 2.5 times the highest noise level. A contour map of the to combine these surveys to produce a single velocity- negative values in the composite spatial map verified that, integrated spatial map and a Galactic latitude-integrated except in a small region near the plane between / =* 128° and longitude-velocity map of the entire Milky Way. Specific 140° (perhaps contaminated by a bad OFF position), the details on the treatment of individual surveys can be found in number of positions with integrated temperature below -5 K the notes to Table i. km s"1 was consistent with an rms noise everywhere in the map of less than ^2Kkms"'. The individual surveys provide nearly complete velocity coverage of Galactic emission in all observed directions. To Compared with the spatial map, the longitude-velocity map produce the composite spatial map, the spectra at Low latitudes integrated over ±3!25 in latitude, involving only the Galactic in the first and fourth quadrants were integrated over the full plane surveys (numbers 2, 4, 9, 11, and 13 in Table 1), was bandwidth of the spectrometer to cover all emission within the relatively simple to produce. After integration over latitude solar circle and most beyond. (Between / = 12° and / = 60°, the the northern surveys (2 and 4) were smoothed in velocity to £ velocity coverage was not adequate to observe all material resolution of 1.3 km s~' to match the resolution of th< beyond the solar circle, but, owing to the rapid fall off in density southern surveys, and every second velocity channel from the of molecular clouds beyond the solar circle [Solomon, Stark, northern surveys was used for the composite map. The latitudt and Sanders 1983; Grabelsky et ai 1987; Kutner and Mead integration range includes essentially all material outside the 1985] and the large distance of this region from the Sun, its Local spiral arm. inclusion probably would have changed the map only very slightly.) In the second and third quadrants, the spectrometer III. THE COMPOSITE MAPS bandwidths were adequate to cover all velocities where 21 cm The composite spatial map is shown in contour form in the emission is significant, and the data were integrated over all Figure 2 foldout (Plate S). The burned out region of intense detected emission. More than 3° from the Galactic equator the emission near the Galactic plane is shown clearly with more data were generally integrated only over the low-velocity inter­ widely spaced contours in the maps in the lower portion of the val where emission was detected or expected. figure. The Figure 3 foldout (Plate 6) represents the composite Several surveys (as indicated in Table 1) required special longitude-velocity map, integrated over ±3°25 in latitude and treatment: local surveys done in equatorial coordinates or on covering 360s of Galactic longitude; Figure 4 is a blowup of an /, b grid shifted from the main map grid were added to the the Galactic center region, adding the small amount of high- main map by numerical interpolation, while others taken with velocity emission clipped in Figure 3. In Figure 5 the survey is an angular resolution of £° or i° were smoothed to the \° shown, integrated over both radial velocity and Galactic lati­ resolution of the composite map. In regions covered by more tude, as a function of Galactic longitude. ; than one survey, that with the lowest noise was used. The composite maps are discussed below starting from the

TABLE 1 INDIVIDUAL SURVEYS

No. Region Telescope Note Réf. I... '™ *~. *~. 1 Oph-Sfr 8" 40s r 24 NI 1 I •) 1st Quad 12 100 -10 10 N2 i 3 Aquila South 27 40 -21 -10 NI 1 3 4 2nd Quad 98 180 -4 10 NI 4 5 Cepheus 143 8 22 N2 5 6 Taurus 15»2 180 -25 -3 N2 2 6 7 Anticenter 175 210 -9 7 NI 7 Orion 180 225 -25 5 NI 2.3 8 9 3rd Quad 180 279 -5 5 C 9 10 Carina-deep 270 300 -2 7 C 3 10 CarinadeG«is.987. •3- I 180" 160° 140" 120° 100" 80* 60* 40* 20" 0" 340* 320* 300* 280" 260" 240* 220" 200- 180" GnLaCTIC LONGITUDE FIG. I.- -Areascovered by Iheindividual CO surveys used losyntliesiielhecompiMile CO maps ohhe Milky Way in Figs. 2 and 3, numbered according lo Table I Vol. 322

-200

15° 10° 5° 0° 355° 350° 345° GALACTIC LONGITUOE FIG. 4.—Expanded version of the Galactic center region from Fig. 3. covering the full velocity extent of the emission. Contours are the same as in Fig 3, except for an additional comcrar at Ô.4 K degr far left in Figures 2 and 3 (second quadrant), through the three major molecular clouds, one at » 140 pc associated with Galactic center, to the far right (third quadrant), and back to the dark nebulae in Taurus and Auriga, a second at - 350 pc the starting point at I « 180°. We note major clouds and associated with IC 348 and NGC 1333. and a third at about regions and cite sources or the data as well as previous surveys the same distance associated with the California nebula (NGC carried out with the New York and Chile telescopes. In Figure 1499) and NGC 1579. The second and third clouds mentioned 6 cloud complexes and other large-scale features are labeled by lie on either side of the Per OB2 association, also at 350 pc their common names, taken generally from constellations or (Fig 6). from well-known associated objects. The rest of the second Galactic quadrant is characterized by widespread filamentary émission lying almost entirely in two a) Individual Regions velocity intervals : between 0 and -10 km s " ' from the Local The dark clouds in Taurus and Perseus well below the spiral arm, and between -40 and -60 km »"' from the Galactic plane between 1-155' and 180° were observed and Perseus arm (Fig 3); the three largest molecular complexes discussed by Ungerechts and Thaddeus ( 1987), who identified associated with the Perseus arm in the second quadrant, which kl.

No. 2. 1987 CO SURVEY OF THE MILKY WAY

VaiilfUffl

HOO - " ",s"' MOtse

j" Ë

180° 150" 120° 90V 60° 30° 0° 330° 300° -* 270° 2*0° 210° 18i ^- Galactic longitude -v^

;^V^ 'AACV vV n ^w\^

10° 0° 050° 3-10° 320" D10J 300° Galactic longitude F(G. 5.—Longitude profile or the composite survey integrated aver ail velocities and over latitude. The instrumental noise varies from 2 to 6 K km s ' ' dcg. Most of the apparent noise is not instrumental—it is the "shot noise" in the distribution or molecular cloudi The lower curve is simply a blowup or the central 120" of longitude. we designate NGC 7538. W3, and S235, are indicated in Figure 1978) and diffuse y-ray emission (Lebrun and Paul 1983) there 6. This region was discussed both by Cohen et ai (1980) and are greater than expected from 21 cm observations. They dis­ Dame el ai. (1986). except for the Cepheus region, observed covered an appreciable amount of cold molecular gas, some so by Grenier. Dame, and Thaddeus 11987). faint in CO that it falls below the lowest contour of Figure 2. The first quadrant survey of Dame. Grenier, and Thaddeus South of the Aquila Rift is a region with a similar H r deficiency 11987) covers intense emission from the inner Galaxy spiral (region 3 from Fig. IX but Lebrun (1986) found only a few very arms and the local dark clouds that form the Great Rift in the small clouds there, all below the lowest contour of Figure 2. Milky Way. Although few individual clouds can be distin­ Well above the Galactic plane near the center of the map lie guished in Figure 2. two fairly local dark nebulae are visible, the similar, possibly related p Oph and Lupus molecular cloud Cyg OB7 and the Aquiia Rift, as well as three quite distant and complexes at 100 to 200 pc, each with masses of a few 10* M0. massive molecular complexes, W51, W44, and M17, all identi­ The p Oph clouds, which contain one of the nearest, most fied in Figure 6. Previously, Dame and Thaddeus (1985) closely studied regions of star formation, were recently mapped most of this region at V spatial resolution and showed observed by de Geus (1987) with the Chile Telescope at full that with the information provided by the CO radial velocities resolution but with i° spacing; an earlier, preliminary i° it was possible to resolve the Great Rift into 10 distinct molec­ resolution survey had been carried out from New York by ular clouds lying over a wide range of distances, 200 to 2300 pc. Bronfman (1980). The Lupus clouds, mapped from Chile The Ophiuchus-Sagittarius region above the Aquiia Rift (Murphy 1985; Murphy, Cohen, and May 1986), contain a (region 1 in Fig. 1) was studied by Lebrun and Huang (1984) large number of T Tauri stars sometimes referred to as the because the optical opacity (Heiles 1976; Burstein and Heiles Lupus T association (Schwartz 1977). ta) 3 ** .fl>aouK e io­

MR2 « *•*<** CAS! VPCRW 2 i«r no* 120* 240* 220"

—1 1 1 1 1—

« ** * n VILA

IM* 160" IW '20* 100" W* 60* W 20m IF MO" HO" 300" 2W 260» 240* 220" XXT IW GALACTIC LONGITUDE

Fw. 6.—Fwder durU for objects and regions discussed in ihc text :(u) corresponds to the velocity-integrated spatial map in Fig. 2 and (6) to the longitude-velocity map in Fig. i. The lowest contour is sketched schematically in both charts, and large regions with intensity greater than or c^ual to 4 K defearc shaded in (/>). - 1(4 -

CO SURVEY OF THE MILKY WAY 713

Far below the plane near the middle of the map lie the small, the Rosette nebula (Mon OB2; Blitz and Thaddeus 1980). isolated R CrA and Chamaeleon clouds, well apart from the another with the Cone Nebula (Mon OBI; Blitz 1978). The

dark clouds and other young objects in Gould's Belt which at quite massive ( - 10° Me) but strangely cold, quiescent cloud this longitude is above the plane (see § Mo). The R CrA cloud studied by Maddalena and Thaddeus (1985) is just below the was mapped at i3 angular resolution from Chile by de Geus plane between the CMa OBI and Mon OB2 clouds. Overlap- 11987); CO observations toward the young, luminous, pine lections of the region / = 180° to 230° have been observed emission-line star R CrA were previously reported by Loren, at i° resolution from New York by Huang (1985) and Madda­ Peters, and Vanden Bout 11974). Most of the Chamaeleon lena and Thaddeus ( 1985) and from Chile by Murphy ( 1985). region was mapped at full resolution from Chile by Koprucu, Three of the most extensively studied molecular clouds, Cohen, and Thaddeus (1987); Ketoand Myers(1986)observed those associated with the Orion Nebula. NGC 2024 (Orion B). three small clouds on the fringe of the main star-forming cloud. and Mon R2, are conspicuous well below the plane near The fourth-quadrant survey of Nyman et al. ( 1987a), cover­ / ~ 210°. This region, mapped and discusseu in detail by Mad- ing the intense emission from the Galactic center and the inner daleoa et al. 11986), also contains numerous smaller clouds and spiral arms, has a lower noise level (0.1 K rms) than most two interesting molecular filaments, each more than 10° long others ; essentially all features within its boundaries in Figure 2, and typically + ° wide, which reach toward the Galactic plane. even those so small as to be taken as noise, are real molecular These filaments, barely resolved and lying mainly below the clouds. For example, the lower two of the three small features lowest contour, are not visible in Figure 2. lying above the plane near / = 325s are intense regions of a 5 large (I0 .vf0) cloud that Nyman et al. (19876) found to lie b) Large-Scale Structure - 250 pc above the Galactic plane, nearly 3 times the molecu­ The intense ridge of CO emission along the central 120° of lar cloud half-thickness in z. Its situation above a conspicuous the Galactic equator is the most prominent feature in Figure 2. hole in the longitude-velocity distribution of CO in the plane As the longitude-velocity map (Fig. 3) shows, the high intensity suggests that both the large displacement of the cloud and the of the emission results from integration over many clouds hole itself may be due to a single cause, possibly multiple along the line of sight, most of them lying in the inner Galaxy supernova explosions and stellar winds. In contrast to the first spiral arms or in the region of the Galactic center. The rapid quadrant, blanketed by nearby dark clouds, the fourth quad­ falloff in density of molecular clouds beyond the solar circle is rant is comparatively free of low-velocity emission (Fig 3), with evident when the intensity of this ridge is compared with emis­ the conspicuous exception of the Coalsack, identified in Figure sion at low latitudes in the second and third quadrants, where 6 even though in the spatial map (Fig. 2) it is badly confused with more intense background emission from the inner nonlocal outer Galuxy clouds appear. Galaxy. Excluding the i'iner-Galaxy ridge, most of the emission in the spatial map (Fig. 2) arises from clouds within -1 kpc. The CO emission toward the Galactic center extends over a Except in small regions of severe velocity crowding near I » 0°. very wide velocity range, roughly ±250 km s"' (Fig. 4), and 90°, 180°, and 270°, these local clouds clearly are separated in the integrated intensity in Figure 2 reaches 948 IC km s'1. velocity from the more distant ones (see Fig. 3); in the first and ' about 4 times the highest intensities found elsewhere along the third quadrants they generally lie between —10 and 20 km s " ', Galactic equator (see also Fig 5). Among the most striking while in the second and fourth they fall between —20 and 10 features in this region are the objects near / = 3° and 5° which km s " '. The large clouds in Figure 2 more than 3° from the stand out in the lower panels of Figure 2 due to their high plane, excluded from the longitude-velocity map (Fig. 3), also integrated intensities and in Figure 4 due to their extraordi­ fall within these velocity ranges. nary velocity widths 1^100 km s"1). Their CO luminosities s A striking aspect of the local molecular cloud distribution is and velocity widths imply kinetic energies of order 10 * ergs, the much higher density of clouds in the northern half of the comparable to those of the 3 kpc and 135 km s" ' expanding Milky Way (( = 0° to 180°) than in the southern ( 180° to 360°). arms iBitran. Thaddeus. and Cohen 1987). This asymmetry stands out in the longitude-velocity map (Fig. The Carina nebula lies fairly near the tangent point at 280° 3), even though much of the local emission in the northern of the Carina spiral arm. one of the best denned structural Milky Way lies beyond ±3° in latitude and is therefore features of the Galaxy in molecular clouds. The tangent is in­ excluded. The preponderance of nearby dark clouds in the tense in CO, while just beyond it (near / « 275°) the CO inten­ north is very evident on large-scale optical surveys, the Milky sity drops abruptly (Fig 5). In the longitude-velocity map (Fig Way from Sagittarius to Cepheus being divided by the Great 3), the Carina emission clearly shows the loop characteristic of Rift of optical obscuration (Dame and Thaddeus 1985; Feit- a Galactic spiral arm. Grabelsky (1985) mapped this region at zinger and Stiiwe 1986a) and the region beyond, from Cepheus {' angular resolution and the inner 2° to 3° in latitude at full to the anticenter, covered by a complex system of filamentary resolution. The large-scale properties of the Carina arm are dark clouds. Seen clearly in Figure 2 between I - 100° and discussed by Cohen et al. ( 1985) and Grabelsky et al. (1987). 180°. these filamentary clouds have been called the Northern A large part of the third Galactic quadrant, from 210" to Dark Cloud System by Schlosser and Gornandt (1984), who 279°, including the rich region in Vela centered at about 265', suggested they constitute a coherent structure. Figure 2 lends was surveyed by May, Murphy, and Thaddeus (1987) and some support to this idea, since the extensive, filamentary emis­ Murphy (1985) at J'' resolution. In Vela, Murphy identified at sion of the region sharply contrasts with most other regions of least four massive ( -10' M0) hot clouds between 800 pc and the map. the fourth quadrant in particular. 2400 pc from the Sun, as well as closer gas, possibly related to One large-scale structure evident in Figure 2 is associated the "Vela Sheet" of Eggen (1980), at approximately the dis­ with Gould's Belt, the apparent disk of OB stars, gas, and dust tance of the Vela pulsar and the Gum nebula, - 400 pc. Spread in a great circle about the Sun inclined some 20° to the Galac­ along the plane over the rest of the third quadrant are several tic plane (e.g., Stothers and Frogel 1974). The maximum exten­ well defined, star-forming clouds, one of them associated with sion to positive latitude of the Belt lies in the general direction - I»F,

714 DAME ET AL. Vol. 322

of the Galactic center and is marked by the large clouds in depend only on the calibration relative to the survey of Dame Lupus, Ophiuchus. and Aquila; the Taurus and Orion clouds and Thaddeus (1985) used by Bloemen a ai. (1986). Observa­ opposite in the sky mark the maximum negative-latitude tions of reference CO sources along the Galactic plane confirm extension of the Belt. The CO maps provide an important that the relative calibration is accurate to 5%. To account for observational link between the optical dark clouds and associ­ helium and heavier elements in the clouds, we assumed a mean

ated surs of the Belt and the 21 cm "feature A " of Lindblad « molecular weight per H, molecule of 2.76mM, al. (1973) (see § IV). As emphasized by Dame and Thaddeus Individual local clouds or cloud complexes are listed in ( 1985), the CO clouds are closely correlated spatially with the Table 2 with their limits in / and b, mean velocities, distances, dark nebulae, while many are closely correlated in velocity masses, mass-weighted mean displacement from the Galactic with feature A. plane, ï, and rms dispersion about the plane, c.. Two sets of I Another possible large-scale symmetry, evident in both and b limits are given for the Aquila Rift, Pet OB2, and p Oph Figure 2 and Figure 3, is formed by regions of intense, complex because, owing to their irregular shapes, none rf these could be emission toward Cygnus X (I - 80°) and toward Vela enclosed by a single rectangular region without including other (i ~ 270°). which, as Murphy (1985) recently noted, lie in the clouds or large, emission-free regions. Masses were determined Galactic plane -180° apart and are fairly similar in overall directly from the CO emission within the rectangular regions intensity, angular size, velocity extent, and general structure. integrated over the total velocity extents of the clouds. Further, these regions coincide with strong concentrations of The cloud boundaries given in Table 2 are in places some­ Population I radio and optical tracers spread over a consider­ what arbitrary (e.g.. between the Vul and Cygnus Rifts) and able range of distances. The CO data are evidently consistent when so should be considered rough divisions between regions with the idea long discussed (e.g, Bok 1959; Humphreys 1970; with nearly constant velocity and distance rather than as defin­ Walborn 1973; Herbst 1975; Vega, Muzzio, and Feinstein ing individual clouds. Neither the Lindblad Ring nor the 1986) that the Sun is located in a Local spiral arm, sun tangen- " -12 km/s" emission is a single object; both appear to consist tially toward Cygnus and Vela (see § IV). rather of fairly complex systems of clouds at nearly constant distance. Similarly, the limits given for any other region, such as the Aquila Rift or Cyg OB7, might contain smaller but IV. LOCAL MOLECULAR CLOUDS presumably related clouds whose distances cannot be deter­ The wide latitude coverage along the entire Milky Way of mined individually. The derived masses are typically -50% the composite survey allows the thickness and density of larger than in the first quadrant analysis of Dame and Thad­ molecular gas in the solar neighborhood to be determined for deus (1985) as a consequence of the larger cloud boundaries; the first time from a direct inventory of nearby molecular adopted here. I clouds. Previous CO estimates have been derived by extrapo­ The complex, widespread emission in the second quadrant1 lating data from the inner and outer Galaxy to the solar circle requires special attention. As mentioned (§ HI), this emission is (e.g.. Sanders, Solomon, and Scoville 1984; Thaddeus and clearly divided between the Perseus arm near —40 km s " ' and Chanan 1985) or have largely been based on local observations local material between -20 and 10 km s"' (see Fig. 3). The in only one quadrant (Dame and Thaddeus 1985). local emission, however, appears further divided by a gap near Following Dame and Thaddeus (1985), we define local to —7 km s " '. The more negative local component, centered near mean within 1 kpc of the Sua Clouds this close lie in the —12 km s ™ \ contains several large clouds and smaller frag­ low-velocity lane in the longitude-velocity map (Fig. 3) at ments plausibly related to a string of OB associations—Cep absolute velocities less than 20 km s " *. In most directions such OB2, Cep OB3, Cep OB4, and Per OB3—between 800 pc and clouds can be readily distinguished from more distant ones by 1000 pc from the Sun (Humphreys 1978). The Cep OB3 cloud their low velocity and wide latitude extent; many are visible as has been studied in detail by Sargent (1977,1979), and parts of dark nebulae on optical photographs. Distances, all taken other clouds in the region have been observed by Sargent et al. from the literature, are based on associated Population I (1981), Elmegreen. Dickinson, and Lada (1978), and Casoli, objects or on star counts or on the relation of visual absorption Combes, and Germ ( 1984); the distances derived or adopted, as versus distance for stars within the cloud boundaries (eg., well as the kinematic distances to most of the clouds, are in Forbes 1985). Kinematic distances, although generally so inac­ rough agreement with those of the OB associations. Further, curate for such nearby objects as to be almost useless, were the CN absorption-line study of Munch (1964) toward the H u taken into account in a few cases (e.g., the " -12 km/s " clouds region S171, lying toward the Cep OB4 molecular cloud, and CepheusX but were never the sole or primary distance reveals strong CN absorption at a distance of 900 pc (Osterbrock 1957). We therefore adopt 800 pc as the distance indicator. The change in Rs from 10 to 8.5 kpc recommended by IAU Commission 33 (Kerr and Lynden-Bell 1986), there­ to all the clouds in the more negative velocity local component fore has virtually no effect on the results presented below. and call them the" - 12 km/s" clouds. As in the analysis of nearly all the large-scale CO surveys, Two local velocity components are also seen in 21 cm obser­ masses for the local clouds were computed by assuming a vations over most of the Galactic plane (Lindblad 1974). The linear proportionality between Wc0, the velocity-integrated 21 cm component that corresponds to the " -12 km/s " clouds

CO intensity, and NH,, the H2 column density. We adopted the is identified with the Local spiral arm by Lindblad (1974), and ratio NnJWtomM x 1020 cm"2 K"' km s"' derived by the other, closer to 0 km s~ ', is feature A, modeled by Lind­ Bloeman et ai. (1986) from intercomparison of CO, 21 cm, and blad et al. (1973) as an expanding shell of cold gas surrounding •/-ray surveys over large areas of the Galaxy (see Dame and the Sun. In the second quadrant, Lindblad's "expanding ring" Thaddeus 1985), since this determination agrees very closely lies on average at ~300 pc, in reasonable agreement with the with the results of a similar analysis of one of the most massive distances to the nearest dust clouds in this direction local molecular cloud complexes: the Orion region (Bloemen (Elmegreen 1982). The low velocity of the CO emission and the et at. 1984). Masses thus obtained are independent of uncer­ very large latitude extent of the 21 cm feature A suggest that tainties in the absolute calibration of the telescopes; they the material is close—significantly closer than the " -12 - Ht .

No. 2. 1987 CO SURVEY OF THE MILKY WAY

TABLE 2 MoLfcutA*. GAS «THIN I KltoPAUCC OF SUN

M Region *-. *~ (km J"1) (pel Ref. |I0'.)/.) 1S?5 34" 6' 10s 1 Aquila Rift .... 8 200 1 1.5 9 15 34 44 4 4 1 Cloud A 44 49.5 4 2 27 500 I 0.4' -7 14 Cloud B 44 54 4 5 7 300 1 0.4= 0 11 CloudC 50 55 1 3.5 24 500 1 0.3 5 12 VulRift 54 63 3 5 10 400 1 0.8 5 12 CygRifi 63 86.5 4 4 7 TOO 1 8.6' -4 24 CygOB7 87 99 3 8 -1 800 1 7.5 41 52 Lindblad Ring . 100 164 4 10 1 300 1.6 22 28 --l2kot.s" ... 102 161 4 10 -12 800 3-> 8.7 44 59 Cepheus 100 120 1 22 -5 450 4 1.9 131 133 Taurus 163 178 2 -9.5 5 140 5 0.3 -37 38 154 162.5 -3 5 -7 51 PerOB2 350 5 1.3 -84 92 163 171 9 -6 -3 1 Mon OB1 197.5 205 1 4 7 800 7 1.6 17 22 Orion A 208.5 218 -: 1 -14.5 5 500 6 1.6 -163 164 Orion B 202J 208 -! 1 -6 5 500 6 1.7 -129 132 Mon R2 210 218 4 -10 7 830 6 1.2 -182 183 Vela Sheet 272 279 3 8 0 425 8 0.8 9 19 Cham 295 305 -2 3 -12 4 215 9 0.1 -60 60 Coalsack 300 307 4 3 -4 175 10 0.04 _2 5 0317-4 315 320 6 -2 -6 170 11 0.03 -11 II Lupus 333 346 4 22 5 170 8 0J 32 35 350 2 1 3 24 1 3 165 12 0.3 35 39 <>Oph 356 5 3 12.51 RCrA 357 4 -22 -14 6 150 13 0.03 -55 55

'r - Em,*,!>!,; using bins lOpcwideins. 1 * a, - (Smt:t'/XmJ'' ; using bins 10 pc wide in :. ' The boundaries of this cloud are 2-3 times larger than the ones adopted by Dame and Thaddeus 11985) and include adjacent, presumably related gas. ' Following Dame and Tnaddcus ( 19851, the average CO intensity of the Cygnus Rift above ( - 74s, where use Rift is confused with, the more distant and intense Cyg X region, was assumed to be the same as that measured below 74e, where only the Rift is seen. DISTANCE REFEIENCES.—(I) Dame and Thaddeus 1985:12) Undblad a ul. 1973: 13) Humphrey] 1978: 14) Lebrun 1986: 15) Ungerechts and Thaddeus 1987:(6) Maddelena « d. 1986:(7) Blitz 1978: (8) Murphy 1985: (91 Hyland. Jones, and Mitchell 1982: (10) Rodger 1960:111) Neckel and Klare 1980:112) Chini 1981:113) Caposchkin and Creenslein 1936.

km/s " clouds. We therefore adopt a distance of 300 pc for the (Table 2) is shown in Figure 8 as a function of distance from the CO clouds coincident in velocity with the 21 cm feature A and Galactic plane z, broken down into the contribution from each designate them the Lindblad Ring clouds. Galactic quadrant. This figure is an extension to all four quad­ The local molecular clouds are shown in projection on the rants of Figure 11 of Dame and Thaddeus 11985), with the Galactic plane in Figure 7. The distribution is far from trivial difference that differential mass per parsec in z has been random: nearly all the clouds in the first and fourth quadrants converted to density. The disproportionate amount of molecu­ apparently lie on a fairly straight ridge more than 1 kpc long, lar gas in the northern Milky Way is evident: in the first and running from the Vela Sheet to the Cygnus Rift. The clouds second quadrants the molecular mass is 4 times greater than in from Chamaeleon to the Vul Rift may be related to Lindblad's the third and fourth. expanding ring, although the fourth quadrant clouds are much Since about half the local molecular mass is contained in five farther from the Sun than in the model of Lindblad el al. fairly compact objects (Cyg Rift, Cyg OB7, Cepheus, Orion B. (1973). The entire ridge may trace the inner edge of the Local and Mon OBI), smalt number fluctuations are significant, and spiral arm. and it is possible that the large Vela Sheet and Cyg within 1 kpc the average shape of the Galactic distribution in z Rift clouds at either end are superpositions of clouds which lie at the solar circle is not well sampled and therefore not deter­ even farther along the ridge. Trie overall distribution of clouds mined very accurately. Generally, a Gaussian distribution in : is consistent with the Sun lying near the inner edge of the Local has been adopted in the analysis of the large-scale distribution spiral arm; the large apparently empty region in the first and of Clouds in the Galaxy, and Grabelsky el al. (1987) showed fourth quadrants below the ridge is the interarm region that a Gaussian is indeed a good approximation to the z dis­ between the Local arm and the Sagittarius arm, which at I - 0° tribution of clouds in the Carina arm. Simply fitting a Gauss­ lies ~ 1700 pc from the Sun (Dame et al. 1986). The existence of ian to the clouds in Table 2, however, because of the poor this large interarm void in molecular clouds and presumably sampling makes little sense. Although this small sample of dark nebulae explains why in directions such as that toward clouds does not allow determination of all moments of the : M17 at I » 14°. where there happens to be little local obscur­ distribution, the first two moments can be determined: the ation in the Rift system, it is readily possible to observe optical mean z displacement of the distribution is + S pc and its rms objects in ''ie Sagittarius arm. dispersion about the plane is 74 pc. Given that from other The mem density of molecular gas within 1 kpc of the Sun evidence molecular clouds appear to follow a Gaussian ; dis- H» •

716 DAME ET AL. Vol. 322

1 kpc r- FIG. 7.—The distribution in the Galactic plane of molecular clouds within 1 kpc of the Sun (Table 2). The circle radii arc proportional to the cube roots of the doud masses and in most cases are dose to the clouds' actual radii The shadinf indicates distance from the Galactic plane. The gênerai regions of the * -12 kra/s " and Lindblad Ring clouds are indicated but individual douds are not shown. The widths of these regions in hdioccntric distance are unknown: the widths shown are arbitrary. tribution, our best estimate of the molecular layer half- Table 3 summarizes the local scale height and density of thickness at the solar circle is 87 pc, the half-thickness at molecular gas and compares our results to the first quadrant half-intensity of a Gaussian with a dispersion of 74 pc. results of Dame and Thaddeus (1985). As expected, the surface Similarly, our best estimate of the mean midplane density at density averaged over all four quadrants is lower than that the solar circle is obtained by-assuming that, on average, averaged over only the first, as that contains the large clouds of molecular gas has a Gaussian distribution in z. The total mass the Great Rift. The increased layer thickness also is not sur­ s of molecular clouds within 1 kpc of the Sun is 4.0 x 10 M0, prising: Dame and Thaddeus based their results on the disper­ l implying a mean surface density of 1.3 MQ pc " . If a Gaussian sion of only the first quadrant and Orion clouds; the present distribution in z with a half-Ouckness of 87 pc is then assumed, larger survey includes other high z clouds, in particular those 1 the mean midplane density is 0.0068 MQ pc' , or 0.10 H, in Taurus and Cepheus. With a smaller total mass spread over cm"3. The value at z — 0 pc in Figure 8 is about twice as high, a larger z range, the average midplane density derived here is presumably simply because of statistical fluctuations in the z almost a factor of 2 below the value derived from the first distribution of clouds. quadrant alone.

TABLE 3 MOUCULA* GAS WrnttN I KuoPAJtSEC

I - I2M0O" only: All quadrants: Dame and Thaddeus (198S) Present work

rms i dispersion 64 pc 74 pc Layer thickness (HWHM,. 75 pc 87 pc Mass 44 x 10» M-,' 4.0 x 10» M. r Surface density 2.0Mopc- 1.3 M0 pc"' J Midplane density 0.013 Ma pc"' 0.0068 M0 pc" a20H,cm-J 0.10 H, cm"1 ' Four timet the nuus in the quadrant observed (I- I2M00°). -Ut -

No. 2, 1987 CO SURVEY OF THE MILKY WAY 717

o Z

-IDD 100

Z (pc) Fio. 8.—ThcmeudensityoTmolecuUrpswithin I kpcoftheSimuaftmctionordistancefromtheGiJjcticplincz

OUT direct measurements of the local molecular scale height total mass in gas and stars n only about half of the Oort limit, and density agree well with results from the inner and outer the total mass in tbe solar neighborhood deduced from the Galaxy extrapolated to the solar circle. The inner Galaxy observed distribution of stars perpendicular to the Galactic axisymmetric analyses of Bronfman er al. (1987) and Sanders, plane. The present work is conclusive evidence that the Solomon, and Scoville ( 1984) yield a half-thickness at the solar "missing mass" in the solar neighborhood, tbe différence circle of - 75 pc, while tbe analysis of Carina arm clouds in the between tbe total mass observed and the Oort limit, is not in outer Galaxy by Grabelsky « at (1987) yields ~ 100 pc at the the form of molecular clouds. solar circle, nicely bracketing the present 87 pc. Extrapolations Finally, it is instructive to compare our results with those of the densities of Bronfman « al. (1987) and Grabelsky « at. from a completely independent study of the local Hj gas by (1987) to the solar circle yield midplane densities of -0.01 AJç Savage et al. (1977), who used Coptrniaa UV absorption 1 1 pc' , in rough agreement with our value of 0.0068 MQ pc' . spectra of the Lyman bands toward early-type stars to deter­ Although the value derived by Sanders, Solomon, and Scoville mine H, column densities. The 109 stars they analyzed are (1984) is more than 4 times higher, this difference can be attrib­ distributed fairly uniformly in Galactic longitude and about uted in part to differences in instrumental calibration and 85% fall within 25° of the plane; however, as they point out, adopted NHJWco ratios; Bronfman et al. (1987) show that the (heir survey is biased toward stars with less than normal remaining discrepancy is mainly due to a difference in fitting reddening per unit distance; all but 18 lie outside the lowest procedures. contour in Figure 2 and some may lie in front of our CO clouds. Because of this bias, the mean H density along lines of The agreement of our measurement of the local density with a extrapolations of azimuthal averages within and beyond the sights to these stars. 0.036 cm " ', a not a useful measure of the total H density locally; it is valuable, however, as an upper solar circle implies that, averaged over a circular area of 1 kpc 2 radius, the molecular density near the Sun is typical of that at limit to the amount of H, that lies outside the lowest contour the solar circle. Within this area, however, a large-scale asym­ in Figure 2, possibly in the low-density halos of clouds or in metry exists, the density in the northern half (/ -0° to 180°) clouds deficient in CO. Our midplane density might, at most, being 4 times larger than that in the southern half (180° to be corrected upward by one-third to account for such gas. 360°). In an attempt to remove their bias toward directions of low

The relative contribution of H2 to the total local interstellar Hj column density. Savage et al. used the large-scale extinction mass is evident from the insert to Figure 8. The surface density measurements of FitzGerald (1968) to extrapolate to higher of Hj is less than a third that of H I, and the total mass of gas density regions. They derived a mean H2 density in the Galac­ (H j + H I) is about a fifth of the total mass in surs. In turn, the tic plane within 500 pc of the Sun of 0.14 cm " '. In comparison, -HS -

718 DAME ET AL. Vol. 322

from Table 2 the average midplane density out to 500 pc (again complete out to at least ±20° of Galactic 'latitude. The dark- with 87 pc as the layer half-thickness) is round to be 0.13 cm "J. cloud map was adapted from a map with --4° angular This close agreement is undoubtedly fortuitous, but taken resolution produced by Feitzinger and Stiiwe ( 1986ft), while the loosely it implies that most of the matter in the local clouds is angular resolution of the COS B gamma-ray data is greater not highly opaque at visual wavelengths. It is evidence too for than 2°; the CO and infrared maps are smoothed to 2°5 the general paradigm that nearly all dark nebulae are molecu­ angular resolution for comparison. The three contours in each lar clouds and vice versa. map were chosen to emphasize the similarity of the maps. Some of the obvious differences among the maps are easily V. COMPLETENESS understood. The dark-cloud map lacks the intense inner- Since many of the large local clouds lie more than 10° from Galaxy ridge, because only relatively nearby molecular clouds the Galactic plane, beyond the latitude range of the composite are observed as dark clouds. The CO emission is more nar­ survey at some longitudes, it is appropriate to ask whether rowly confined in latitude because it traces the distribution of similar clouds might exist in unobserved regions (see the sam­ only molecular gas; the other maps also trace atomic gas, pling boundary in Fig. 2). Although many small clouds which has a scale height in z about twice as large as molecular undoubtedly exist beyond the composite survey, such as those gas. Further, the gamma-ray map contains several point associated with the IRAS high-latitude "cirrus' (Magnani, sources, notably the Vela pulsar near I - 263° and the Crab Blitz, and Mundy 1985; de Vries, Heithausen, and Thaddeus pulsar near / - 184°. 1987), comparison of our CO data with other Population I Allowing for these differences, the similarity in the distribu­ surveys indicates that no large clouds have escaped detection. tion of all the Population I tracers is striking. The CO and Figure 9 presents a comparison of four large-scale tracers of dark-cloud maps show the most detailed correlation, with an interstellar gas. With the exception of the CO map, each is almost one-to-one correspondence of clouds; even the small ' i | I r !-•—!• ' 1 • i—i ' ' r > 7- | i i | 1 -i" r-r T-r-r- -v |—r I I-' '_ DARK CLOUDS 20° Ss^j^ 0° -o ^ ©^Ka^ffi ^^T\I. ^3**!"3sD^cSs ^ 52>® <= R-- 20°

i . i i i i i i i l-l 1 , , 1 , i • 1 1 J_ 1 1 1 1 11

—i—i—|—i—r- I—i—t—r-r-i—i —r—1—r -1—r-| -I—I—r -1-1-1—r- i-r i -i—i—i-i pi—i- • co 20° V G> a o_^3 52s£3 ,5oo 0^*23 Oo 0

£ i i i i i 1 1 I 1 .L. 1 1 .1-1 1 1 J- 1 1 1 1 1 1 1 , i , 1 1 1 1 1 , '

1 I ' I I ' I I

. i.. i.. i

- •c^p't •A**3l

_1_ I I I I -I- I . • I 180° 150° 120° 90° 60° 30" 0* 330* 300° 270° 840* 210° 180° GALACTIC LONGITUDE F». 9—Dark clouii: The Gilutic dUlribuuon of 2622 dirk cloud, foldfd with • Giuiiun poinl-ipmd (unction of the form tip ( -616,1 win s„ - 2?3. Thf contour, «re it opacity duM 1,2, sod 5; adtptfd from Fig. 1 of Feilzinjtr ind SiUwt (19860). CO: v»locity.inuiriud CO imiuioD From tht Milky Wiy (Fis. 2) •aootkod to in infutir nMlulioa of V.i (FWHM). Cootoun in it 2,9.

No. 2. 1987 CO SURVEY OF THE MILKY WAY 719

dark cloud near I = 35' was detected in CO (Lebrun 1986). features as the distribution of dark clouds: the large Lupus. although its emission lies below the lowest contour in Figure 9. p Oph, and Aquila Rift clouds lie at the positive-latitude exten­ The absence of the Cepheus cloud in the dark-cloud map is sion of Gould's Belt, while opposite in the sky the Orion and somewhat puzzling, but a general flaring of the dark clouds Taurus clouds mark the negative-latitude extension. The much near I =• 110e is clearly seen in the unsmoothed data higher density of dark clouds in the northern Milky Way iFeitzingerandStuwe 1986ft). versus the southern, obvious on large-scale photographs, is A detailed comparison of the CO data and both the infrared evident also in CO. Within 1 kpc of the Sun. the molecular and gamma-ray data is beyond the scope of this paper, but. like mass in the first and second quadrants is 4 times greater than the dark-cloud map. the infrared and gamma-ray maps in in the third and fourth. Figure 9 offer little evidence for large objects outside areas The complete longitude and wide latitude coverage of the observed in CO. One exception may be the region above the survey provides a thorough inventory of large molecular plane near / = 0° : the other gas tracers suggest that low-level clouds near the Sun. The overall distribution of clouds within 1 emission unobserved in CO (Fig. 2) may connect the Aquila kpc is consistent with the Sun tying near the inner edge of a Rift, p Oph, and Lupus clouds. Further CO observations are local spiral arm or spur. The half-thickness at half-intensity of planned to fill these possibly significant holes in our composite the local molecular cloud layer is 87 pc, in good agreement survey. with results from the inner and outer Galaxy, and the mean midplane density, derived assuming a Gaussian : distribution, VI. SUMMARY 3 is ~0.0068Mopc- . Large-scale CO surveys of the entire Galactic plane and specific nearby clouds have been combined to produce a pan­ orama of the entire Milky Way in molecular clouds at an We thank D. Grabelsky, Y.-L. Huang, M. Koprucu. F. angular resolution of +!. Covering 10°-20° in latitude at all Lebrun, and R. J. Maddalena for providing data, some in longitudes and all or nearly all large, nearby clouds at higher advance of publication: F. Avilés, H. Alvarez, L. Bronfman, J. latitude, the composite survey is the only molecular line survey Montani, and E. S. Palmer for help in operating and main­ to date with sky coverage and resolution comparable to that of taining the New York and Chile telescopes; A. V. Smith for the early 21 cm surveys. help with data analysis; and E. Sarot for editorial assistance. The inner Galaxy spiral arms produce, as expected, a thin, L-A. N. acknowledges support from the Swedish Natural intense ridge of emission along the Galactic plane within ~ 60s Science Research Council and H. U. acknowledges support of the Galactic center. The local emission, covering a much from the Alexander von Humboldt Foundation through a wider latitude range at low velocity, shows the same large-scale Fedor-Lynen Fellowship.

Bahcall. J. N. 1981. Ap. J, 27t. 16». Grabelsky, D. A. 1985. Ph-D. thesis. Columbia University. Bitran. M.. Thaddeus. P. and Cohen, R. S. 1987, in preparation. Grabelsky. D. A.. Cohen. R. S. Bronfman. L, Thaddeus. P. and May. J. 1987, Blitz, L 1978. Ph.D. thesis. Columbia University. Ap. J. 315.121 Blilz.L.andTbaddeiu.P. 1980. ,4p. J.241.676. Grenier. I. A. Dame. T. M, and Thaddeus. P. 1987. private communication. Bloemen. J. B. C. M.. Caraveo. P. A, Hennsen. W„ Lebrun. F.. Maddalena. Heiles. C. 1976..4B.J..2S4.379. R. J. Strong. A. W„ and Thaddeus. P. 1984. .4str. .4p. 139.57. Herbst. W. 1975, A J, M. 503. Bloemen. I. B. Q. M.. « al. 1986. Am. Ap, 154.25. Huang. Y.-L. 1965. Ph.D. thesis. Columbia University. Boulanger, F. 1987. private communication. Humphreys, R. M. 197ft AJ, 7S. 602. Bok. B. J. 19S9. Observatory. 79.38. . 1978. Ap. J. Suppl, 31309. Bronfman. L. 1980. M.S. thesis. University of Chile. Hyland. A. R.. Jones. T. J., and Mitchell. R. M. 1982. M.H.R.A.S, 201.1095. Bronfman. L.. Cohen. R. S., Alvarez, H.. May. J., and Thaddeus, P. 1987, Ap. J, Keto. E. R.. and Myers. P. C. 1986. Ap. J, 304.466. in press. Kerr. F. 1, Hindman. J. V, and Gum, C. S. 1959. Australian J. Phyi., 12.270 Burstein. D, and Heiles. C. 1978. ApJ.. 225. 40. Kerr. F. J., and Lynden-Bell. D. 1986. M.N.R.A.S.. 221.1023. Burton. W. B. 1976. Am. Rev. Astr. Ap.. 14.275. Koprucu, M.. Cohen. R. S., and Thaddeus, P. 1987. private communication. Casoli. F. Combes. F, and Gerin. M. 1984. Aslr Ap., 1J3.99. Kutner. M. L.. and Mead. K. N. 1985. in IAU Symposium 106. The Milky Way Chini. R. 1981. Am. Ap.. 9», 346. Galaxy, ed. H. van Woerden. R. J. Allen, and W. B. Burton (Dordrecht: Christiansen, w. N„ and Hindman. J. V. 1952. Australian J. Sci. Res.. 5.437. Reidell. p. 209. Cohen. R. S. 1983, in Surveys of the Southern Galaxy, éd. W. B. Burton and Lebrun. F. 1986. Ap. J,30t. 16. F. P. Israel IDordrecht: Reidel), 265. Lebrun. F, and Huang, Y.-L. 1984. Ap. J, Ml. 634. Cohen. R. S., Cong, H.. Dame, T. M., and Thaddeus, P. 1980. Ap. J. (Letters). Lebrun. F, and Paul. I. A. 1983. Ap. /., 2M. 276. 23». L53. Lindblad. P. 0.1974. Highlights of Astronomy, 3, ed. G. Conlopoulos. 381. Cohen. R. S.. Dame. T. M., and Thaddeus. P. 1986. Ap.l. Suppl., 6*. 695. Lindblad. P. O.. Grape, K„ Sandqvist. Aa.. and Schober, J. 1973. Am. Ap, 24. Cohen, R. S.. Grabclsky, D. A., May. J, Bronfman. L.. Alvarez. H. and Thad- 309. deus,P l985.,4,..MLelrers).2»0.Ll5. Loren, R. B., Peters, W. L, and Vanden Bout, P. A. 1974, ,4p. J. {Letters). 194. Cohen. R. S. and Thaddeus, P. 1977, Ap. 1. (Letters), 217. L155. L103. Conn, H.-l.. Kerr, A. R.. and Matlauch. R. 1.1979. IEEE Trans., MTT-27.245. Maddalena. R. J, Morris. M, Moscowitz. J, and Thaddeus. P. 1986. Ap. J, Dame, Til. and Thaddeus, P. 1985. Ap. J, 297,751. 34)3.375. de Geus. E.J. 1987. private communication, Maddalena. R. J. and Thaddeus, P. 1985. Ap. J, 2*4.231. de Vries. H. W. Heilhausen. A., and Thaddeus. P. 1987. Ap. J, 31». 723. Magnani. L. Blitz. L.. and Mundy. L. 1985. .4p. J., 2»5.402. Eggen. 0.J.1980. Ap. J, lit. 627. Mauzy, B. 1974. NRAO Electronics Division Internal Kept, No. 146. Eftnegreea. B. G. 1982, in Submiliimeter Wave Astronomy, ed J. E. Beckmart May. J. Murphy. D. C and Taaddeus. P. 1987. Aslr. Ap. SuppL, in press. and/. P. Phillips (Cambridge: Cambridge University Press), p. 1. Mayer-Hasselxandcr. H. A. et al. 1982. Am. Ap.. IBS. 164. Smegma B. C Dickinson. D. F„ and Lids. C. J. 1978. ,4p. J, 220.853. Miller. G. E.and Scalo. J. M. 1979. Ap. J. Suppl, 41,513. Ewen. H. I. and Pureed, E. M. 1951. Nature, It* 356. Mullet, C. A, and WesterlwutG. 1957. Bull. .Aslr. Inst. Netherlands. 13.1S I. Fdtzinger. I. V. and Stavw. J. A 1986a, Ap. J., 305.534. Munch. G. 1964. .4*. J, 14». 107 . 19860. Vistas Aslr., 2». 291. Murphy, D. C. 198S, PhD. thesis, Massachusetts Institute of Technology. FilzGerald. M. P. 1968. A J, 73,983. Murphy. D. C, Cohen. R„ and May. J. 1986. Am Ap, 14>7.234 Forbes. D. 1985. AJ, •». 301. Nickel. T- and Klare. G. 1980. Am. Ap. Suppl, 4t 251. Gaposchkin. S- and Grecnitein. J. L. 1936. Harvard Obs. Bull., No. 904. Nyman. L.-A.. Alvarez. H, Cohen, R. S.. and Thaddeus, P. 1987a. in prep. FiizUerald, M. P. 196». AJ„ 73.983. aration. Forbes.D. l9»i.AJ,H,301. Nyman. L-A, Thaddeus. P.. Bronrman, L, and Cohen. R. S. 19876. Ap. J, 314. Gaposchkin, S„ and Graf nstein, J. L. 1936, Harvard Obs. Bull., No. 904. 374. Gordon. M. A, and Burton. W. B. 1976. ,4». J„ 201,346. Osterbrock. D. E. I9S7. .4». J, 125.622 - SI -

720 DAME ET AL i

Pao.S.-K. 1984. Ph.D. thesis, ColumbiaUniversity. Schwartz.ELD. 1977,/tp. J.SuppL.SS. 161. Penrias,A.A..andBurrus,C.A.1973..-iii».JÏfv..4«r. Ap^l\,Sl. Scovi]le.N.Z., and Solomon, P. M. WS.Ap.J.lLeuers\ I99.LI03. Rodgers. A. W. I960. M.N.M.S.. 120.163. Solomon. P. M. Start A. A, and Sanders. D. B. 1983. Ap. J. iUaer». 2*7 Sanders. D. B.. Solomon. P. M., and Scoville. N. Z. 1984. .4». J.. 27*. 182. L29. Sargent. A. 1.1977, .4p. J. 211.736. Stothen. R.. and Frogel. J. A. 1974, A J-7». 456. . 1979. .4p. J. 2M. 163. Thaddeus. P.. andChanan.G. A. 1985. Nature, J14.73. Sargent. A. U van Duinen. R. J. Nordh. H. L. and Aaldera. J. W. G. 1981. Asir. Ungerechts. H., and Thaddeus. P. 1987. Ap. J. Suppl, 63.645.

Ap.,94.377. Vega. E. I-, Muzzio, J. C and Feinstein. A. 1986, Rev. Mexicana Asir. Apn IZ Savage. B. D.. Bohlin. R. C. Drake. J. F., and Budich. W. 1977. Ap. J.. 216,291. 33. Schlosser. W.. and Gôrnandt. V. 1984. .4«r. Ap.. 137.287. Walbom. N. R. 1973. A J, 711067.

R. S. COHEN: 51 Seventh Ave. Apt 1. Brooklyn. NY 11217

T. M. DAME. L.-Â NYMAN, and P. THADDEUS: Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridg- MA 02138

E. J. DE GEUS: Sterrewaeht Leiden. P.O. Box 9513.2300 RA Leiden, Netherlands

I. A. GRENIER: CEN-Saclay, DphG/SAP91191. Gig-sur-Yvette, Cedex. France

J. MAY : Departamento de Astronomia. Universidad de Chile, Casilla 36-D, Santiago, Chile

D. C. MURPHY': European Southern Observatory, CasiUa 19001, Santiago, Chile

H. UNGERECHTS : I. Physikalisches Institut der Universitat zu Kôln, Zûlpicher Strasse 77,5000 Kôln 41, West Germany W^*^*ie€£&**

-i 1 i i i - ra -

CO OBSERVATIONS IN CEPHEUS: I- NEARBY MOLECULAR CLOUDS AND THE EXISTENCE OF A CLOSE BUBBLE

I.A. Grenier, F. Lebrun, M. Arnaud Service d'Astrophysique, c.E.N. Saclay

T. Daae and P. Thaddeus Harvard-Smithsonian Center for Astrophysics

ABSTRACT

Emission at 2.6 mn from 12C0 has been mapped in the Cepheus region over 490 square degrees at an angular resolution of 0.5° with the Coluabia Sky Survey telescope. We present in this article the spatial distribution, distances and masses of the surveyed molecular clouds. Two clouds, representing a total of 1,3 10s solar masses, are found in the Local Arm (800-900 pc) at a surprisingly high elevation. The observations also partly reveal a vast nearby molecular complex lying at a distance of about 300 pc. One object, the Cassiopeia cloud, stretches along an arc 30 degree long and contains about 10s solar masses while the second object, the Cepheus cloud, represents 0.6 10s solar masses at a similar distance. The situation of this complex wit* respect to the Gould's Belt is not clear. Furthermore, the ring-like disposition of these nearby clouds, the lack of atomic and molecular gas in the middle, and observations at soft X ray energies and radio wavelengths strongly suggest the presence of a close bubble betwwen Cepheus and Cassiopeia. A tentative interpretation as a 4 10* year old. type I supernova remnant is proposed. rij

1. INTRODUCTION

Although Hubble had noticed in 1934 a large zone abnormally void of

galaxies in the Cepheus constellation, which he named the Cepheus Flare, no

systematic study of the interstellar medium content of the region was undertaken

till the 21 cm observations by Heiles in 1967. His extensive survey (100° SIS

140°, 13° Sb< 17°) revealed the existence of a complicated, unexpected,

dynamical structure with two separate velocity components lying above a diffuse background emission. He interpreted the observations as two sheets of gas

extending over tens of degrees along the galactic plane and having a relative motion of expansion or rather collision. The distance of the sheets was poorly known, but 500 pc seemed a reasonable estimate, Later, the HI deficiency noted with respect to interstellar absorption measurements in this direction (Burstein and Heiles, 1978, Strong and Lebrun 1982) prompted Lebrun to map the molecular content of the region. The first 12C0 observations (Lebrun 1986) revealed a large cloud complex covering 140 square degrees between 100° and 115° in longitude and 11° to 20° in latitude. Its mass and proximity made it one of the major features of the solar neighbourhood, a cousin of Orion and Taurus.

Today, we present a more complete survey of the 12C0 emission in the

Cepheus region, that is from 100° to 140° in longitude and from 8° to 22° in latitude, obtained using the Columbia Sky Survey telescope. The large scale observations cover with a better sensitivity the clouds already mapped by

Lebrun. Furthermore, the map extends to much larger longitudes where the presence of the two HI sheets and also the relative HI deficiency with regard to the diffuse gamna-ray emission (Strong et Lebrun, 1982) strongly suggested the presence of other molecular concentrations. Indeed, the data has revealed a second cloud complex with a velocity distribution very similar to that of the atomic gas sheets.

Hence, after the presentation of the observations, we will study the - s r -

dynamical structure of the whole complex. Distance and mass estimates of the

different entities will be discussed and a special attention will be paid to the

low velocity component which accounts for a large majority of the emission.

Merged to the similar Columbia CO survey of the second galactic quadrant (Dame

et al. 1987), the observations show that the high longitude part of the local

component in fact represents the northern half of an important new nearby cloud.

Finally, the disposition of these local molecular clouds, which delineate an

interesting loop structure, will be compared to observations at other

wavelengths.

The study of the diffuse gamma-ray emission originating in this nearby

interstellar formation, which did prompt these observations, and its

implications on the calibration of the N(H2)/WCO ratio will be addressed in a

second article.

2. OBSERVATIONS

The observations of the "CO J=l-0 transition line at 115 GHz (2.6 mm) were carried out by one of the authors (I.A.G.) during the winter 1984-1985 with the

1.2 m Columbia University Sky Survey telescope in New York City. The full angular resolution of the antenna (FWHM of 8.7') was downgraded to a 0.5°x0.5° square beam by stepping the telescope through a square array of positions about a beamwidth apart and summing the data into a single spectrum. This "superbeam" technique (Dame and Thaddeus, 1985) associated with a new performing receiver allowed a well-sampled and fast coverage of a large area in the sky. The total of 1961 spectra taken along a regular (l.b) coordinate grid fully samples 490 square degrees of the Cepheus constellation. The outer boundaries are 99.5° to

141° in longitude and 8° to 22,5° ir. latitude and the detailed borders are drawn on figure 1. No data were recorded below 8° since these regions were already mapped in the frame of the second quadrant survey (Dame et al., 1987). _ ?6 -

The telescope was equipped with an extremely sensitive liquid helium-cooled

SIS receiver (Pan. 1984) with a typical single-sideband noise temperature lower

than 90 K. together with a 256 channel spectrometer providing a frequency resolution of 250 kHz. or a velocity resolution of 0.65 km/s at 115 GHz. The filter bank was centered to cover the velocity range -82.5 to +82.5 km/s. The emission intensities were calibrated and corrected for atmospheric absorption by briefly rotating a room-temperature blackbody wheel in front of the feed horn before each scan. To measure the temperature and opacity of atmospheric water vapor, antenna tipping was performed every 12 hours or more often when the weather was variable. As a final test, pointing and calibration were checked at least daily using a bright standard source (W51 at 1=110°, b=0.125°).

Because of the unusual large width of the lines observed in this region and because of the presence of two velocity components, frequency switching was not adopted but position switching against two reference "OFF" positions was preferred. Moreover, at these high latitudes, a large and well-spread set of emission-free positions could easily be found by frequency switching. With this observing procedure, only linear baselines had to be removed from the raw spectra. Integrating times were automatically set to achieve an rms noise temperature of 0.2 K. Under clear and cold winter conditions each scan so lasted

30 to 60 seconds.

To present the data, two quantities have been calculated,

- The integrated intensity WCO(l.b) = /T«.(v) dv, where T» is the radiation temperature measured in each channel of a spectrum. The velocity range of integration was chosen to cover all the emission lines, unless other boundaries are specified in the text. To check the quality of the data, a histogram has been constructed from the negative values of WC0 recorded in the survey where no real emission occurs, This distribution is completely consistent with an rms noise of 0,2 K,

- The longitude-velocity diagram LVCOd.v) = £i,T«-(b) showing the velocity -re­

structure of a region by summing the spectra over latitudes.

3. RESULTS

Figure 1 presents the contours of WCO over the whole region. The lowest

contour value has been chosen to delineate real emission with a 2.5cr confidence

level. The figure obviously reveals the existence of an impressive molecular

concentration which covers 120 square degrees at longitudes higher than 120°.

The tiny cloud situated at 1=117°. b= 11-12° had already been mapped in detail

by Huang (1985) for it lies near the direction of the CTA1 supernova remnant.

The good agreement between the latter observations and the present ones provides

a good verification of the quality of the data. Such a consistency is also found when comparing the low longitude part of the survey with the map of Lebrun

(1986). The present better sensitivity, however, reveals the faint extensions

and diffuse edges of these clouds. The majority of the molecular gas in Cepheus

evidently gathers within two broad regions. The large angular extension of the different concentrations indicates that they probably belong to the solar vicinity. But the absence of emission between the two groups raises a first question: do they form a unique complex?

In order to study the large scale dynamical structure of this medium, figure 2 gives the longitude-velocity diagram derived from the spectra. As the clouds rather stretch along the galactic plane, not much confusion is introduced by the summation over all observed latitudes. The striking partition of the material in two well-defined velocity components exactly recalls the two HI sheets observed by Keiles (1967). Again the separation appears as a dip around

-6 km/s easily seen in the high longitude part of the plot, At lower longitudes,

the separation is not as clear. Nevertheless, a spatial map of the emission situated between 1» 103° and 111°, and in the velocity interval -8 to -15 km/s, _ ri .

demonstrates that it comes from an isolated entity, distinct from the vast cloud

visible at lower velocities. To outline this high velocity component of the

emission. WCO integrated from -8 to -20 km/s has been displayed in figure 3.

The remarkable similarity of the velocity distributions of the atomic and molecular gas in the whole Cepheus region strongly suggests that the CO clouds are physically related to the HI sheets. Moreover, the spatial arrangement of the members of the two velocity families coincides in HI and CO. A fortuitous alignment, both in space and velocity, being quite unlikely at these latitudes, the presence of the molecular hydrogen within the atomic sheets seems a reasonable assertion. In fact, the present situation confirms what Heiles had suspected in 1967 when he noticed an HI deficiency in different dusty locations in the sheets. The problem now is to estimate the relative distances of the two components. A striking feature of the common disposition of the 21 cm and 2.6 mm emissions is the lack of any material, atomic or molecular, in the middle of the map (around 1= 120°, b= 15°). We will thoroughly discuss this peculiarity later in the article.

U. DISCUSSION

Cepheus has long been considered as a rather poor, unattractive part of the solar environment compared to the famous Orion, Taurus, Cygnus and others. It is true that the region possesses no spectacular nebula or splendid young star cluster. If the Ceph.us molecular clouds shelter star formation sites, the products must be in general low mass stars since no OB association appears within the boundary of the survey (Humphreys, 1978), nor any isolated OB or WE star (Cruz-Gonzales et al., 1974; Van der Hucht et al., 1981). Only a few T

Tauri stars were observed in the reflection nebulae NGC7023 and NCC7129. No others were ever noticed in the high longitude clouds (Herbig and Rao, 1972). In such a vast region, however, a few HII regions are found, SU, S133, S136. S137. - rs .

S150 and S174 (Marsalkova, 1974), some of which are of.particular interest since

Blitz et al. (1982) observed them in CO. As already mentioned, a notorious

supernova remnant, CTA1. lies at a distance of 1 to 2 kpc in the direction 1=

119.5°, b= 10°. On the whole, the distribution of the dust clouds, as observed

by Lynds (1962) or photographed by Schlosser and Gornandt (1984), agrees well

with the presence of cold molecular gas in Cepheus. A few faint reflection

nebulae, studied by Racine (1968) and Knapp et al. (1977), can be related to the

present CO data. They lead important clues to the distance of the gas. All these

objects are presented in figure 4 where the outlines of the clouds belonging to

the two velocity constituents have been drawn.

4.1 Distances to the clouds

First confronted to the problem of the distance evaluation of the two HI

sheets. Heiles reported color excesses measurements by Elvius (1956) which

indicated the presence of absorbing material at 300 and 700 pc in the directions

1= 108.5°, b= 20.8° and 1= 100.3°, b= 7.1°. Other optical measurements and

kinematical considerations (such as a -12 km/s velocity at 1= 135° implies a

distance of 800 pc in the frame of the galactic rotation) led him to choose 500

pc as a common distance for the HI emission and 300 pc and 500 pc as more

probable for the two sheets.

The value of 300 pc for the nearest component is fairly consistent with the

distances found for some nearby reflection nebulae. Racine (1968) gives 400 pc.

250 pc, 500 pc and 260 pc for VDE152, DC177, DG178 and Racl57 respectively.

Observed in CO by Knapp et al. (1977), the first three objects exhibit lines

centered on -4 km/s. A detailed map of the CO emission around this velocity

derived from the present data indeed shows small concentrations at the position of the three nebulae which could plausibly be related to them. VDB141, also

studied by Knapp et al., appears as well on the fine map but its distance is unfortunately no reported. The brightest nebula of the region, NGC7023, may _ 40 -

also belong to the local part. Its distance, estimated at 440 pc by Viotti

(1969) is now reported at 290 pc (Allen. 1983). It is obviously coincident with

a CO enhancement between 0 and +4 km/s. Such velocities are in very good

agreement with the detailed "CO and 13C0 observations of Elmegreen and

Elmegreen (1978) who recorded a line at 2.7 km/s and 2.4 km/s wide. In the high

longitude region this time, another reflection nebula seems a reliable distance

indicator of the nearest component since its illuminating star is a Cepheid, SU

Cas. It lies at 300 pc from us. Finally, a systematic search for interstellar

Calcium and Sodium lines against A stars with known distances is in progress.

Preliminary results show a strong absorption near 250 pc (Grenier, private

communication).

Beside the kineraatical argument already given by Heiles. very few optical

indicators at these latitudes may lead to the distance of the other velocity

component. A single reflection nebula, Racl46, located near NCC7129 and NCC7142,

seems related to cjr CO data. It lies at »1 kpc from the Sun. Another nebula,

Racl43, which does not appear on the CO map rests at =800 pc. Two HII regions,

Sll and S137, observed in CO at -10 km/s (Blitz et al.. 1982), seem related to each other. S137 is thought to be at 620±200 pc. Finally, the northern part of

the Cep 0B2 association extends up to b= 12.3° between 1= 97° and 108°. Its distance, 800-900 pc. is consistent with the other evaluations.EJvius measurements and the above values therefore seem to converge toward about 800

PC

Other arguments favouring the choice of 300 pc and 800 pc are provided by the situation at lower latitudes. For instance, the interstellar reddening measurements performed by Lucke (1978) detect two sharp absorbing fronts at 300 and 700 pc below 6° in latitude, with a faint intermediate one at about 450 pc.

Furthermore, the partition of the velocity distribution is well marked near the galactic plane. The nearby emission clearly contrasts with the gas present in the Perseus arm (-30 to -60 km/s in this direction) and it also breaks up into - 41 -

the sane two components as in Cepheus. The gap occurs near -7 km/s. Their

relative distances have teen exhaustively discussed by Dame et al. (1987). They

adopted 800 pc for the farthest component, called the Local arm, since it may

plausibly be related to the OB associations Cep 0B2, Cep OB3. Cep OB4 and Cam

OBI which all rest around 800-900 pc. Dame et al. also quote that the nearest j component was identified by Linblad (1974) as a close HI expanding ring lying at

' =300 pc.

For all the above reasons we shall adopt 300 pc and 800 pc as the best

estimates of the distances to the two velocity components found in Cepheus. The

800 pc value is somewhat higher than the 500 pc chosen by Heiles. But his

interpretation of the confused velocity distribution in the low longitude part,

where Che sheets seem to merge, in terms of colliding sheets implied a rather

narrow gap between them. The collision hypothesis may not be supported by the CO

data. An alternative interpretation, which will be discussed further, is that

the two formations are well separated in space throughout the region, but that

an unusually wide velocity dispersion in the nearest clouds below 120°

obliterates the velocity gap.

With this choice of distances, average heights above the galactic plane can

be derived for the different clouds (see table 1). The two main constituents of

the high velocity part are found at 130 and 170 pc above the plane. The

molecular layer half-thickness (HWHH evaluated including the high z clouds

Taurus and Cepheus) being 87 pc within 1 kpc of the Sun (Dame et al., 1987)

implies that these clouds are at a significantly unusual high elevation. On the

other hand, the average height of the Local arm CO members in the second

quadrant is 44 pc with a dispersion of 59 pc (Dame et al., 1987). The two

Cepheus members are therefore not really displaced with respect to the others.

They may represent the highest extensions of the Local arm. On the contrary, the

elevations estimated for the nearby clouds of the survey (50 to 90 pc) are

completely consistent with the average disposition of the molecular medium in - il -

the solar vicinity.

4.2 Masses of the clouds

Although laC0 molecules are so abundant in the cold clouds that the lines should not be optically thin, it is now believed, enpirically, that WCO is one of the most reliable tracers of molecular hydrogen column densities on a large

scale. Integrating WCO over a cloud and converting this value to N(Ha) therefore yields a mass estimate if one assumes a mean atomic weight per H atom (often

taken as 1.36). The conversion factor. N(H2)/WC0, has long been discussed.

Recently, the diffuse gamma-ray emission observed by the COS-B satellite has provided an interesting mean to calibrate this ratio on the same large scale.

The value used here. (2.3 ± 1.4) 102° raol. cm"2 K"1 km"1 s, has been derived from the study of the 150-5000 MeV gamma-ray flux in the Cepheus region itself.

This analysis is presented in paper II. The value fully agrees with the (2.3 ±

0.3) 10ao obtained from the comparison of the gamma rays and the gas content along the entire galactic plane (Strong et al.. 1988). Mass estimates have thus been calculated for the different clouds of the survey according to their distance. It should be noted that the largest uncertainty comes from the very uncertainty in the distance. Results are listed in table 1.

They show that the mass contained in the survey, =2.5 10s solar masses, contributes to »6X of the molecular gas known within 1 kpc. This gas is more or less equally distributed amongst the different entities. Each velocity component accounts for about half of the material observed in the region, that is around

10s solar masses each. The high latitude extensions of the Local arm seen in

Cepheus thus represent =10% of the CO in the arm. Within each velocity component, each cluster (high/low longitudes) again accounts for about 50% of the mass, So the large clouds in the survey all weight about 5 10* solar masses.

This average is quite typical of the nearby interstellar medium. Ignoring the

Cygnus Rift and Cygnus 0B7 giant formations, the masses found for the main _ t.% -

clouds within 1 kpc all range from 3 to 15 10' solar masses (Dame et al., 1987,

scaled to N(Ha)/WC0= 2.3 1020).

To test whether the clouds are gravitationally stable, 'virial' masses have

been calculated in the simplest case where a cloud is considered as a

homogeneous sphere without any magnetic field or any other internal source of

pressure. Such an estimate requires the knowledge of the 'radius' of the cloud.

It may be approximated to: (area of the cloud lx)x'*. Then the full width ôv of

the total CO line of the cloud is needed. It may be obtained by summing all the

spectra of the cloud and by assuming a gaussian shape: FWHM= 0.94 *WC0 /T^tPeak)

On the whole, the derived virial masses are systematically larger than the

CO estimates, reflecting the overall large velocity dispersions measured in

these clouds (see figure 2). One possibility is that, high above the galactic

plane, these clouds may not be gravitationally stable. He explicitely assumed

that no physical cause other than random inner motions is responsible for the

velocity dispersion. But systematic variations of the velocity, such as a strong

gradient within the cloud, will artificially increase the parameter 6v and

induce an overestimation of the mass. This is indeed the case in the clouds 3

and 5 of Cepheus. For them two virial masses have been calculated: one with 6v

derived from all the spectra simultaneously and a more realistic one using 6v measured in different locations along the velocity gradient. In the latter case,

the discrepancy between the CO and virial masses of these clouds is reduced to a more acceptable level. For two clouds however (2 and 4), one in the Local arm and one nearby, the discrepancy is quite large. No velocity gradient which night account for this disagreement being visible, these clouds must be rather unstable. No explanation was found for the farthest cloud but one will be proposed for the nearby cloud below in this chapter.

4,3 A vast cloud in the Cepheus-Cassiopeia region

So far we limited our presentation and discussion to the CO emission found - m -

within the strict boundaries of the survey. As already mentioned, the

observations were confined above 8° in latitude because the regions below had been mapped by the Columbia telescope with the same grid and angular resolution

(Dame et al., 1987), It is, however, obvious on figure 1 that the nearby cloud

(around 0 kra/s) discovered beyond 120° must be prolonged at lower latitudes. The full extension of this cloud can be seen from the synthesis of the two surveys.

As the same velocity partition between the Local arm and the nearby medium exists below 8°. the map of the nearby component has been produced by integrating WCO between -8 and +8 km/s. Figure 5 reveals that the cloud extends far beyond Cepheus, well into the Cassiopeia constellation. The "Cepheus" part represents in fact the northern half of a vast formation which stretches along an arc 30 degree long and covers about 200 square degrees. To be true, the cloud is barely in Cepheus and covers a major part of Cassiopeia. So we suggest to call it for convenience the Cassiopeia cloud as opposed to the Cepheus cloud which lies at the same velocity between 100° and 120° in longitude, and 13° and

22° in latitude. The vast angular extent of Cassiopeia confirms that it belongs to the solar neighbourhood. It is indeed one of its most prominent constituents.

Its full mass, derived from WCO, reaches 10s solar masses at 300 pc compared to the 0.6 103 solar masses concentrated in the Cepheus cloud. Both these values matches the masses found for the clouds in Orion and Taurus.

Although their CO emission appears in the same velocity range, a projected distance larger than 50 pc separate Cassiopeia and Cepheus (if they are 300 pc away), There is no way to conclude on geometrical grounds whether they form a unique, gravitationally bound, complex or not. But the existence of a possible loop between the two clouds (presented next) indicates that they are closely related.

Besides, one would like to place them with respect to the Gould's Belt since other close CO clouds are obviously related to this structure (Lupus, p

Oph. Orion, Taurus). The distance of 300 pc is compatible with the clouds - fef -

belonging to the Belt. As mentioned earlier, Linblad's expanding HI ring, which

is part of the Belt, is found at 300 pc in this direction. Unfortunately, the

ring has been studied only near the galactic plane toward these longitudes. No

clues are given to the latitude extent of the feature (Linblad et al., 1973 and

Linblad. 1974). However, in the longitude-velocity diagram displayed on figure

2, the general trend of the local component velocity strongly recalls what

happens in CO near the plane and what Linblad has observed of the HI ring

between 100° and 140°. Such a similarity in the velocity trend suggests that

Cepheus and Cassiopeia are related to the HI expanding ring. On the other hand,

the Belt as traced by the young stars is inclined some 20° to the galactic plane

along a 1= 0-180° axis. With its minor axis parallel to the galactic plane, the

Belt crosses the latter around 1= 90°. So Cassiopeia and Cepheus appear at a

fairly high elevation (£ 50 pc) above the Belt disc and the bright young stars

in the second quadrant do not suggest any tilt nor any warp of the Belt plane

towards higher latitudes in this direction (Stothers and Frogel, 1974, Lucke,

1978). It is true though that very few OB stars are available above 10° to trace

a possible extension of the Belt. As it seems difficult from the geometry to

think of Cepheus and Cassiopeia as members of the Gould's belt, we are facing

two conflicting arguments, More work has to be put .into the latitude extent of

Linblad's ring, the distribution of the young stars in this particular region

and in the actual distance (and hence elevation) of the two clouds before

placing Cepheus and Cassiopeia with respect to the Belt,

4.4 An old supernova reanant between Cepheus and Cassiopeia?

An Interesting aspect of Cassiopeia is its arch structure and the ring-like

feature it forms with the high longitude part of Cepheus and the tiny cloud near

118°. 10°, Moreover, the region confined between these clouds presents rather peculiar aspects, It is void of molecular gas, down to 0.35 K km/s in CO. It is also depleted of atomic gas with respect to its surroundings, with column _ *« -

densities below 6 1019 at/cma compared to an average of 1 to 2 10"° along the

borders. Towards Cepheus. the "hole" is bordered by a very sharp front in CO

where the emission drops from 18 K km/s to nearly 0 within 1.5°. that is within

10 pc or less. This edge also exhibits the widest lines found in the whole

region. The study of the area at other wavelengths reveals other striking

aspects.

First, the "hole" appears as a bright spot on the soft X ray maps of the

sky produced by McCamaon et al. (1983) (figures 7a, 7b & 7c). It is in fact the brightest part of the galactic plane in the B band (130-188 eV). Not much

structure is found in. the C band but the highest contours in the B and Ml bands nicely match the CO disposition (figure 6a). Because of their softness, the X rays must come from a nearby region, otherwise the diffuse atomic gas observed by Heiles would absorbe them rapidly. A simple explanation for the presence of many soft X rays and the simultaneous lack of molecular or even neutral gas is that the "hole" is filled with a very hot plasma. Its temperature must exceed a few 10s K since the medium does not emit in Ha (Sivan, 1974). At radio frequencies, the colour map of the sky at 408 MHz (Haslam et al., 1981) shows the hole as a clear notch in the smooth galactic emission. The detailed contours

(figure 6b) present the brighter rims which seem to form a small loop structure above 10° in latitude. Below 10°, too many well known bright sources, such as

CTA1, HI, S137, unfortunately disturb the large scale emission. The radio loop nicely follows the CO and X ray edges of the "hole". To help the comparison, the position of the X-ray rims have been reproduced on the radio nap (figure 6b).

This radio structure is not spurious. It is also clearly seen in the 820 MHz survey by Berkhuijsen (1972) and it is barely visible (barely because of the large angular resolution) on the 150 MHz map of Landecker and Wielebinski

(1970). The presence of the radio structure gives little weight to the hypothesis that the soft X rays do not originate between Cepheus and Cassiopeia but come instead from a diffuse emission behind the clouds which is enhanced - t» _

there because of the little absorption in the hole. The spatial coincidence of

the CO, HI, radio and X ray configurations looks striking, especially when

noting that all four present a sharp edge at low longitude and a much fuzzier

border at high longitude. Polarisation directions are also disordered in this

region compared to their regular alignment along the galactic plane below 10s.

For all these reasons we propose that a bubble exist between Cepheus and

Cassiopeia.

We have to say, however, that the X ray observations must be taken with

care. McCamnon et al. present this area as a highly contaminated one, with

possible sources of background being "auroral" X rays or X rays generated in the

payloaO 'tself during the rocket flight. It is true that the large part of the

sky covered during this flight appears slightly brighter on the B map than other areas. But it seems difficult to create from an internal source of noise a

structured enhancement in (l.b), structure which is moreover perpendicular to

the detector path in the sky. But, as a uniform flux is probably contaminating the data, we have considered only the highest contours and we have subtracted the surrounding flux in our determination of the bubble emission. McCaomon et al. were doubtful about the data in the B and Ml bands because the relative weakness of the C band emission. The weakness is real because it also appears on the SAS-3 map (Marshall and Clark, 1984). If the B and Ml enhancements are likely to be real because of the observations at other wavelengths and for the reasons just discussed, an alternative interpretation of the low C eraissivity can be found for instance in the chemical abundances inside the bubble. The X ray emission in the different bands is very sensitive to the density of few species (Arnaud and Rothenflug, 1986). A silicium depletion could explain the present B/C ratio and seems possible in an evidently dusty location. Another possible source of pollution is the soft emission from discrete X ray sources,

Four HEA0-A1 sources beside CTA1 are found in the area (they are displayed on figure 6a) but they are not likely to simulate the observed enhancement (Wood et _ it -

al.. 1984).

Ring-like features in the interstellar medium are numerous because they may have two different origins. Some are interpreted as bubbles blown by the strong winds of young massive stars, others as shells created by supernovae explosions.

Some authors also combine the two phenomena to explain the supershells. The present one has a reasonable size, approximately 30 pc in radius at 300 pc. and both phenomena can plausibly create such a bubble. The wind-driven shell, however, is not supported by the observations since no 0 or B star, nor any WK star, lies in this direction. Because of the closeness of the bubble, such a star would have been easily recognized during the systematic surveys of giant stars. The stars of an early type nearest to the bubble are found around (118°,

15°) and near (110°, 8°) at a distance of 800-900 pc for they belong to the

Cepheus associations. So the only explanation left is the supernova event.

The available observations have been used in the frame of a classical Sedov model of the evolution of supernova remnants in order to test the supernova origin. Such remnants are believed to be strong emitters of soft thermal X rays when they are old enough to contain a mass of swept-up interstellar matter equivalent or larger than the mass of the éjecta. This phase of the adiabatic expansion has been theoretically described by Hamilton, Sarazin and Chevalier

C1983), They showed that the X ray emission is completely characterized by the conserved energy E of the event, the ambient interstellar hydrogen density no and the shock temperature T. The temperature profile inside the remnant is not flat but, as a first approximation and because we observe the bubble from outside, we assume that the temperature derived from the X ray fluxes equals the shock temperature. The values of T and no measured from the X rays are then used in the expansion model to evaluate the energy of the explosion and its age.

As already explained, to limit the impact of any contamination from the experiment we derived the X ray count rates originating in the bubble by subtracting the surrounding emission. This was also the only realistic solution - fcS -

to eliminate the contributions of the Local Bubble along the line of sight and

of the poorly known extragalactic background. Count rates of 30. 40 and 10

count/s have been retained for the B. C and Ml bands respectively. As the C flux

is relatively small and as the emission in the B and Ml bands is strongly

correlated to the structure at other wavelengths, the latter have been preferred

to measure the temperature. The ratio B/Ml yields a temperature of 1.5 10s K and

an emission measure of slO"2 at cm"' pc. But, to take into account the weakness

of the emission in the C bandtwo extreme cases have been determined using the

B/C and C/Ml ratios. They yield temperatures of 0.5 and 2.2 10s K. and emission

measures of 1.5 and 0.35 10"2 cm"* pc. The supernova event has then been tested

in the three cases. Its characteristics have been estimated following Sedov's

equations (Hamilton et al.. 1983):

T The energy E= no 10s1 erg LlO'RJ U.62pc-. 62pc'' T The present expansion velocity v= 839 I 1 km s~l LIO'KJ -J -0.3 The age e= 4015 [ 1 [ 1 years LI07KJ U.62pc,62pcJ 1 T -," r E The swept-up ISM mass M= 96.8 M0 Ll0UO7K7KJ Ll0LioslergeraJ where T and r are directly given by the observations and no is related to the

emission measure by: E.M.= no2 r3 /da. The distance d and radius r of the bubble

are taken as 300 and 30 pc respectively.

Results for the three cases are listed in table 2. As the remnant seems old, the pure adiabatic expansion may be irrelevant. The remnant may have

reached the phase when radiative losses become important. This was tested by

calculating the critical size where starts the radiative phase. According to

Kahn (1985):

E a/7 n -a/7 R= ~ 53 T 1 \ - 1 pc LlO"ergJ Lo,427cm"3J The values found for R= indicate that the bubble has not yet reached this state

and though it is close to it, the Sodov model is still a reasonable - 10 - approxima t ion.

All the parameters of the remnant listed in table 2 must evidently be considered as orders of magnitude because of the large uncertainties in the measurements. The distance of the bubble and the absorption of the X rays are crucial points. But the two extreme cases presented largely include the uncertainties from the X ray measurements and the reader is referred to the formulas to see the impact of the distance knowledge. Unfortunately, the consequences of the model limitations cannot be inferred. On the whole, however, the observed bubble is consistent with a 4 10* year old supernova releasing 10s1 erg into the nearby interstellar medium. Such an age is realistic. The supposed event aust be older than a few thousand years otherwise it would have been recorded on historical documents for the star must have been splendid day and night near the north pole! On the other hand, the supposed supernova must be younger than a few 10s years before the remnant vanishes into the interstellar medium. Finally, the energy release found is compatible with a type I supernova.

These calculations leading to a plausible scenario, we are left with the present observations to believe or not in a supernova remnant between Cassiopeia and

Cepheus. We may further mention that the spectral index of the radio emission measured along the bubble rims at 408 and 820 MHz equals -2.2 and is similar to the indices found for nearby famous remnants.

Supernovae shells are not rare in this region. Hu (1981) has observed a large HI bubble at (105°, 17.5°). It certainly lies beyond 150 pc and more probably around 500 pc according to her model. Its radius is of the order of 26 pc, Another HI shell of the region has been interpreted as a =10' year old supernova remnant (Velden and Hirth, 1982), It is found at 1=130°, b=22.5°, at a kinematical distance of 450 pc and has a radius of 24 pc. At lower latitudes

(102.8s, 6.7°), a giant IB bubble, with an outer diameter of 7°, has been recently discovered (Kun et al., 1987) but it belongs to the Local arm (900 pc).

Finally, one has to remember that loop III dominates the second quadrant. To - Il -

account for the radio flux and polarisation measurements along the spur,

Spoelstra (1972) placed the loop center at (125°, 15.5°) and 150 pc away from

the Sun. Such a position and the observed angular radius of 45° imply that the

shock front is very close to the Cep-Cas clouds region. But the uncertainties in

the distances are again too large to try to guess whether loop III is actually

disturbing the two clouds.

5. CONCLUSION

The present CO observations, prompted by the gamma rays/interstellar gas

correlations study, has largely contributed to our knowledge of the local

interstellar medium. They have led to the discovery of two clouds, of 6 and 8

10' solar masses, belonging to the Local Arm (800-900 pc) but situated at a

surprisingly high elevation. The survey also partly revealed one of the major

clouds of Che solar vicinity. The vast formation of 103 solar masses probably

lies at =300 pc from the Sun. Its filamentary appearance creates an arch 30

degree long in the Cassiopeia constellation. Nearby, at approximately the same

distance, rests the 0.6 10s solar masses Cepheus cloud which presents unusually wide CO lines.

The two entities are close enough to possibly form a unique complex.

Besides, their common situation with respect to the Could's Belt is not clear.

On one hand, their velocity suggests that they are related to the HI expanding

ring of the Belt, but on Che other hand they are too high above Che known position of the Belt disc. A careful study of the latitude extent of the HI and

CO ring, so far observed near the galactic plane, and a systematic study of the young stars above the plane in the second quadrant would be welcomed to clarify the situation. Are the Cassiopeia and Cepheus clouds indeed not linked to the

Belt? Is the Belt slightly tilted towards Cepheus (a few degrees would be enough)? Or was it in the past (the Belt may have a "seesaw" motion perpendicular to the galactic plane, Frogel and Stothers, 1977)? -?i -

The region confined between Cassiopeia and Cepheus looks rather striking.

It is void of atomic and molecular gas and seems filled with soft X rays. The

borders of the hole form a ring-like feature in CO and HI that is also visible

on the X ray and radio maps. The bubble cannot be blown by the strong wind of a

young massive star for no such star has been found inside the hole. The

observations, however, are consistent with a 4 10' year old supernova remnant

releasing 10" erg in the surrounding medium. The existence of such a remnant

would establish the close relationship between Cassiopeia and Cepheus. It would

explain as well the large velocity dispersion observed in Cepheus since the

shock would now be travelling inside the cloud. But precise X ray observations of the bubble must be undertaken before we may confirm its supernova origin. The observations would also be valuable to study the physical conditions inside the bubble. The confirmation of a supernova event in this close region would provide an excellent and detailed example of the interaction of a supernova shock with a dense molecular cloud. . 73 -

a) Individual clouds:

n° longitude latitude v (km s"1) d (pc) z (pc) MO

1 103° - 111° 9° - 15° -20 ; -8 800 170 M,„= 7.5 10* Mvi^= 14. 10»

2 131° - 135° 8° - 11° -20 ; -8 800 130 M=== 5.5 10» tt,!r= 35. 10*

3 99.5°- 104.5° 12.5°- 22° -8 ; +10 300 90 M«,= 1.5 10» tt^„= 10. 10» M^^s 3,4 10»

1 1U3' - 120° 12.5°- 22.5° -8 ; +10 300 90 M„= 4.5 10» It,*,™ 15. 10*

5 122° - 141.5° 8° - 17° -8 | +10 300 70 M„= 5.2 10» &,*«= 41. 10* H,i»a 9.5 10*

6 114° - 121° 9.5°- 13.5° -8 ; +10 300 50 M„= 0.6 10*

M^»= 4.2 10*

b) Cepheus and Cassiopeia groups:

n° longitude latitude v (km s"1) d (pc) z (pc) M0

Cep 100° - 120° 13° - 22° -8 ; +10 300 90 R=„= 0.6 10s

s Cas 114° - 143° 2° - 16° -8 : +10 300 50 tUa= 1.0 10

Table 1: Characteristics of the individual CO concentrations (a) and of the two main groups (b). Position, velocity range, distance, elevation and masses are given. Mco and Hvi, refer to the masses derived from WCO and to the "virial" masses described in the text. - ?

X ray band B and C B and Ml C and Ml

Temperature (K) 0.5 10s 1.5 10* 2.2 106

Emission measure (cm"' pc) 1.5 10-2 1.0 10"2 0.35 10"2

Ambiant ISM density (cm"a) 0.22 0.18 0.11

Expansion velocity (km s"1) 188 325 394

Conserved energy (erg) 0.5 10" 1.2 10" 1.0 10"

Present age (years) 6.2 10' 3.6 10" 3.0 10*

Swept-up ISM mass (MO) 900 750 450 Radiative phase radius (pc) 30 40 50

Table 2: Charactistics of the supernova remnant derived from the X ray measurements (temperature and emission measure) and from Sedov's equations of an adiabatic expansion assuming a distance for the bubble of 300 pc and a radius of 30 pc. - «"-

REFERENCES

Allen. 1983, Astrophysical Quantities, The Athlone Press, 3rd edition.

Arnaud H.. Rothenflug R., 1986, Adv. Space Res.. 6, 119.

Berkhuijsen E.M., 1973, Astr. Ap., 24, 143.

Blitz L., Fich H., Stark A.A., 1982, Ap. J. Suppl.. 49. 183.

Burstein D.. Heiles C, 1978, Ap. J., 225, 40.

Cruz-Gonzales C., Recillas-Cruz E., Costero R., Peimbert M., Torres-Peimbert S., 1974, Rev. Mexicana Astr. Astrof., 1, 211.

Dane T.M. . Thaddeus P., 1985, Ap. J.. 297, 751.

Dane T.M., Ungerechts H., Cohen R.S.. de Ceus E.J., Grenier I.A., May J., Murphy D.C., Nyman L.-A., Thaddeus P., 1987, Ap. J., 322. 706.

Elmegreen D.M., Elaegreen B.G.. 1978, Ap. J., 220, 510.

Elvius T., 1956, Stockholm Obs. Ann., 19, 3.

Frogel J.A., Stothers R., 1977. A. J., 82, 890.

Hamilton A.J.S., Sarazin C.L.. Chevalier R.A., 1983, Ap, J. Suppl., 51, 115.

Haslam C.G.T., Salter C.J., Stoffel H. . Wilson W.E.. 1981, Astr. Ap. Suppl., 41, 1.

Heiles C., 1967, Ap. J.. 149, 97.

Herbig G.H., Rao K., 1972, Ap. J., 174. 401.

Hu E., 1981, Ap. J.. 248, 119.

Humphreys R.M., 1978, Ap. J. Suppl., 38, 309.

Kahn F.D.. 19

Knapp G.R., Kuiper T.B.H., Knapp S.L., Brown R.L.. 1977, Ap. J., 214, 78.

Kun M.. Balazs L.G.. Toth 1., 1987, Ap. Space Sci., 134, 211,

Landecker T.L., Wielebinski R., 1970, Australian J. Phys. Astrophys. Suppl., 16, 1.

Lebrun F., 1986, Ap. J.. 306. 16,

Linblad P.O., Grape K., Sandqvist Aa., Schober J.. 1973, Astr. Ap.. 24. 309,

Linblad P.O., 1974, Highlights of Astronomy, ed G. Contopoulos, 3, 381 - 74 -

Lucke P.B., 1978, Astr. Ap., 64, 367.

Lynds B.T., 1962. Ap. J. Suppl., 7, 1.

Marsalkova P., 1974, Ap. Space Sci. Rev., 27, 3.

Marshall F.J., Clark G.H., 1984, Ap. J., 287, 633.

McCammon D.. Burrows D.N., Sanders W.T., Kraushaar H.L.. Ap. J., 269, 107.

Pan S.K., 1984, Ph. D. Thesis, Columbia University.

Racine R., 1968, A. J., 73, 233.

Schlosser W.. Gornandt V., 1984, Astr. Ap., 137, 287.

Sivan J.-P., 1974, Astr, Ap. Suppl., 26,

Spoelstra T.A.T., 1972, Astr. Ap., 21. 61.

Stothers R., Frogel J.A., 1974, A. J.. 79, 456,

Strong A.W., Lebrun F., 1982, Astr. Ap., 105, 159.

Strong A.W., Bloemen J.B.C.M.. Dame T.M.. Grenier I.A., Hermsen W., Lebrun F.. Nyman L.-A., Pollock A., Thaddeus P. Î988, Astr. Ap., in press.

Van der Hucht K.A., Conti P.S., Luuds.-.rou I., Stenholm B. • 1981. Space Sci. Rev., 28, 227.

VeldenL.. Hirth W., 1982, Astr. Ap., 113, 340.

Viotti R., 1969, Mem. Soc. Astron. Italiana, Nuova ser., 40, 75.

Wood K.S., Meekins J.F., Yentis D.J., Smathers H.W., McNutt D.P., Bleach R.D., Byram E.T., Chubb T.A., Friedman H., Meidav M.. 1984, Ap. J. Suppl., 56. 507. T1 -

FIGURE CAFTIONS

Figure 1^ Map of the velocity-integrated 12C0 emission, WCO, with an angular resolution of 0.5°. The lowest contour is 2.2 K km s"1 and corresponds to a 2.5o confidence level. The contour interval is 2.2 K km s~x. The border of the survey is indicated by the dashed line.

Figure 2: Longitude-velocity map of the 12C0 emission integrated over all surveyed latitudes (8°-22.S°). The lowest contour and steps are 1.9 K deg. The velocity resolution is 0.65 km s"1.

Figure 3: Map of the laC0 emission integrated over velocities from -20 to -8 km s"1 with an angular resolution of 0.5°. The lowest contour and steps are 2 K km

Figure 4; Finder chart for objects and regions discussed in the text. The WCO contour sketched corresponds to the lowest contour of figure 1. The outlines of the Local Arm clouds (high velocity component) are represented by a dashed line.

Figure 5; Compositemap of the nearby molecular complex with an angular resolution of 0.5°. The present observations have been merged to the survey of the scond galactic quadrant and the CO emission has been integrated from -8 to +8 km s"1. The lowest contours and steps are 2 K km s"1.

Figure 6^ Observations at different wavelengths of the bubble structure. All naps are at the same scale. a) Map of the nearby WCO emission, as presented on figure 5, where the two highest contours of the soft X ray emission have been reproduced. The solid lines and dashed lines represent the emission in the B and Ml bands respectively. b) Map of the 408 MHz radio emission froa Haslam et al. Brightness temperatures are given on the contours. The two highest contours of the X ray emission in the B band have been reproduced.

Figure 7; Soft X ray intensity maps of the galactic anticenter from McCanmon et al. a) Count rates in the B band (130-188 eV). the contour interval is 7.5 cts s"1. b) Count rates in the C band (160-284 eV). the contour interval is 20. cts s~l. c) Count rates in the Ml band (440-930 eV). the contour interval is 5. cts s"1. - i »jn6îj -

i i i i i i i i i i i i r i i i i i i i i 4 i i i i i i i i—i—i—i—i—i i—i—L INTEGRATED 1EMP o 5 SUPERBEAV. 04/17/83

Contour values: ?,9by 2.9 Inteqr 8to 2?

Mapping, off Interpolation on 1 r "i "T T"-T ~i 1 1 r

136- 131 -

13? =

130 = 128 - 126 - 121 - 122 - 120 3 118 —

116 = 111 - 112 - 110 - 10B - 10G- 101 - 102t 100- 96 — _i_ _1_ -90

V (km/s)

- Figura 2 - 20. -

15. -

10. -

mo. 130. 120. no. 100.

< Figura 3 - I I I I I I I I I I I ' 1 I I i ; i i i i [ i i i- i - i" i i •]—T i i —T —i—i—i—r ~^> o P>o o o o

517*

CCD

<§3> 7'

..!_...l.l.-L I. I...I . I ..1 I I I 1 I I—1 I L \?1 1,V 1,11 | in 11FÏ 111 US llD 106 105 101 102 IO0

—«sa • •— I.

- Figure 4 - **• •••*&

^ ^ •oJ

- Figur» 5 - (•)

M RI M Ml WIW MP IM t IM > SrT»^ ^ ÇSjJ^ f\> > ^^àaN: (h)

^A^^

^SsCK >0^ ^v^ -?*-$& (&ï^

s V^£/2 ,/W—\\ï l'vL/ ^^B W A- eëfiaùc

figur» 6 I IJ»»W '

(>)

(O ?

<•)

- hi - 3.2 Comparais»* dt trieur* d« allien interstellaire

L'exploration du milieu Interstellaire utilité en particulier les observations HI (21 cm), CO (2,6 M), infra-rouge, las comptages d'étoiles ou de galaxies, les usures de rougissaient et l'émission gasaa diffuse. Cependant, leur qualité dt traceur reposa souvent sur une hypothèse: faible profondeur optique, estimation des rapport! NH./UCO et gaz tur poussières, constance de ces rapports dans la Galaxie, connaissance du flux de rayons cosmique» ... Confronter ces divers traceurs permet donc de s'affranchir de ces hypothèses pour mieux décrira le milieu, ou de tester leur bien-fondé, ou encore de calibrer les traceurs les moint bien maîtrisés. Ainsi, la comparaison des comptage» d'étoilet et de galaxiet dans le Taureau et Persée qui est tout d'abord présentée (Lebrun et Grenier, 1986, Attron. Astrophys. 154, 181) a permis de préciser la relation non linéaire qui relie l'absorption interstellaire aux comptages de galaxies. Associée aux travaux ultérieurs de F. Lebrun sur l'influence des étoiles du champ sur las comptages de galaxies, une telle calibration permet une mesure dorénavant fiable de l'absorption Interstellaire par les comptages de galaxies. La comparaison présentée entuite porte tur let cartas CO et infra-rouge des principaux complexes proches du Soleil (Boulangar et al.). Elle montre qu'après avoir soustrait la contribution attendue du gaz atomique et des régions HII, une forte corrélation existe entre l'émission infra-rouge à 100 microns et les cartes CO. Au delà de 10* en latitude, l'émission infra-rouge apparaît donc comme un traceur des nuages proches. En dehors des réglons optiquement épaisses ou activas du nuage (formation d'étoiles), l'émission infra-rouge peut également être convertie en masse moléculaire à un factaur 2 près. Enfin, la corrélation IR - CO permet de séparer l'émission infra-rouge provenant de l'absorption de la lumière "interstellaire'' qui pénètre dans le nuage de celle créée par l'absorption de la lumière d'étoiles nichées dans le nuage. Cette dernière composants permet alors de repérer les sites peuplés de jeunes étoiles (moins de 2 millions d'années). 6^L Astron. Astrophys. 154, 181-184(1986) »r- ASTRONOMY AND ASTROPHYSICS

Galaxy and star counts in Taurus and Perseus

F. Learaa and I. Grenier Service d'Astrophysique, CEN Saclay DPhG/SAp, F-9119I Gif sur Yvette Cedex, France

Received April 2, accepted June 14, 1985

Sammarv. The star counts of MacCuskey (1938) in Taurus and Perseus are compared with the Lick galaxy counts of Shane and N; Wirtanen (1967) for b < — 10°. It is found that the absorption y derived from star counts suffers from serious zero level errors. Shane and Wirtanen (1967) 0.47 89 This is due to the fact that none of the reference fields used by Heiles (1976) MacCuskey is really 1res of interstellar extinction. Furthermore, 1 100 Burstein and Heiles (1978) this comparison allows a more reliable calibration of the absorp­ 50 Strong and Lebrun (1982) 0.75 tion derived from galaxy counts Finally, a corrected map of the 50 de Vaucouleurs and Buta (1983) interstellar extinction in Taurus and Perseus is presented in galac­ 100 tic coordinates.

Key ««res: radio lines: 21 cm - Galaxy: solar neighbourhood - following equation: interstellar medium: clouds: Taurus-Perseus - interstellar me­ dium: extinction ^j (1/7) x iog(iv;/ny (i) where AT/ is the number of galaxies that would be observed in the absence of absorption. But no consensus has been reached on the value of y and \" (see Table 2). If the Shane and Wirtanen catalog 1, Iatiaaacnoa is magnitude limited, then y represents the slope at m ~ 18 of the m, logW(m) relation, i.e. y - 0.4-0.5 (Tyson and Jarvis, 1979; Star and galaxy counts have long been used as interstellar extinc­ Peterson et al., IP79; Shanks et aL. 1980). However, if this catalog tion tracers. Although these two methods are basically very simi­ is angular diameter limited, y should be between 0.75 and 1 (Phil­ lar, each of them has its own advantages. For sake of clarity, they lips «ci., I98i). Besides, Strong (1983) proposed a nonlinear relation between A and logfV On the other hand, the value of nave been listed in Table I and are discussed below. f% r N" is linked to the problem of the absorption towards the galactic poles (for a review, see Burstein and Heiles, 1982 and de Vaucou­ i.l. Galaxy counts leurs and Buta, 1983). If the dust is uniformly distributed in a With the completion of large scale, low resolution surveys of the plane parallel layer, N° should be equal to 100 and the absorption local interstellar medium (21cm line, y rays), the interest of the towards the poles would be between 0.25 and O.Smag. However, Lick galaxy counts catalog (Shane and Wirtanen, 1967) as an if, as indicated by some reddening studies, the absorption is negli­ interstellar extinction tracer has been renewed (Heiles, 1976; gible towards the poles, which means the solar vicinity is void of Burstein and Heiles. 1978, 1982; Kniflen and Fichtel, 1981; dust. N; should be equal to 50. Strong and Lebrun, 1982; Strong, 1983; Phillips et al., 1983). For The extent and uniformity of the Lick catalog and the large \b\ > 10°, it is generally assumed that galaxy counts obey the scale isotropy of rV* are main advantages. However, the galaxy counts method is limited by the rather poor statistics involved, which restricts the dynamical range and/or the angular accuracy of the absorption measurements. In particular, the validity of this Star counts' Galaxy counts* type of measurements is restricted to medium and high latitudes <|6|>I0°). Zero level N'(I,6)~I0* JV7~10' Scaling factor »(!.«)-a1 lofA/dm v? 1.2. Star counts Dynamical range ~ 6mag ~ 2mag Latitude limit '|o|>10° |*| > 10° The basic relation (often implicitly) used with star counts is simi­ lar to that for galaxy counts: -19 •(!/») xIoi(AÏ/«y (2)

Send offprint requests to: F. Lebrun where N' is the number of stars that would be observed in the -It ^ 182

absence of absorption. It follows from the definition of the inter­ The absorption so derived has been read off in 0.69 square stellar absorption that if - J(log JV,)/dm. In contrast with the gal­ degree cells from the contour map published by MacCuskey. axy counts case, both q and N* have to be measured for each Each cell has been divided into 64 subcells which were then rear­ direction in the sky. Nevertheless, the high star number density ranged in a 1" x 1° grid and also in the same grid as the galaxy allows a fairly good angular resolution and a rather large dynami­ counts. However, the procedure followed by MacCuskey assumes cal range in the absorption measurements. Again, this method is that the star distribution is symmetric with regard to the old not reliable at low latitude because star counts cannot reveal galactic plane. We preferred to extract the raw counts at the 15th absorbing material behind the bulk of the visible stars. The mean magnitude by reversing formula (3) and to derive again the ab­ distance of 15th magnitude stars is 700pc (Rossano, 1980). which sorption using only the reference curves obtained at negative lati­ limits the validity of MacCuskey counts to latitudes greater than tudes. Unlike Eq. (2), which assumes implicitly the linearity of the 10°. The main difficulty with star counts is the estimate of rj and togJv*(m) curves and a slope independent of latitude of r\ = 0.33, iV; which can only be obtained in a field completely free of ab­ we derived the absorption using all the curves as they are. sorption, the so called reference field. Because gas and dust ap­ pear to be fairly well mixed in the interstellar medium, this field should also be free of gas. Nowadays, one may wonder how it is 3. The flbaorptfaM • MacCutkey's reference fields possible to find such fields at medium latitude (10° g |6| g 30°) since gas. at least that revealed by the 21 cm line, is present every­ Below — 10°, MacCuskey used 15 "apparently" unobscured re­ where in appreciable quantities [N(Hi) ~ !02' at cm " 2\. gions of 3.1 square degrees each to determine the reference curves

logJV(m)rar. As stated in the introduction, these reference fields are most likely affected by some absorption. Two independent 2. The data ways of estimating this absorption have been investigated in this paper and are described in the following. The galaxy counts are those of Shane and Wirtanen (1967) as published by Seldner et al. (1977) and arranged in a galactic grid 3.1. Hi column density in the reference fields of 2" x 2° cells. All Seldner et al. correction factors have been Assuming the gas-to-dust ratio is sufficiently uniform over the applied except the plate correction factors for which we used sky, the absorption in a given solid angle can be estimated from those of Shane and Wirtanen, since, for this low Latitude region. its gas content. The entire sky has been surveyed for line emission they may be more accurate than those derived by Seldner et al. at 21cm, so that, assuming the Hi line is optically thin, it is (Groth. private communication). possible to obtain a reasonable estimate of the H1 column density From his star counts in Perseus and Taurus, MacCuskey in each square degree of the sky, at least at medium latitude. The (1938) derived the interstellar absorption using Eq.(2) in the fol­ H1 column densities for the 15 reference fields were derived from lowing form: the Hi survey of Heiles and Habing (1974). It seems that there is A„ = 3x [log N(15) „ - log iV(l5)I (3) an offset in the relation between gas and dust in the local interstel­ f lar medium, in the sense that significant H1 emission is observed where AT (15) represents the number of stars of the 15th magnitude when the reddening is estimated to be negligible, this emission

corresponding to an Hi column density, N(Hi0), of about and N(15)„r designs a reference value. The reference curves N(m) 2O 3 used by MacCuskey to estimate N(15) are averages of the 2l0 cm' (see e.g. Strong, 1983). The slope of the average rtf /V(Hi), A relation is quite well defined but can only apply to curves obtained in "clear" fields at the same old galactic latitude ft on both sides of the galactic plane. randomly selected regions, and this is clearly not the case of the

I11' I I

1.9

1.6

1.4

1.2

I .• I " l—1 *_ O O f—' ^-* j—T—<*, . Ftj.1. The absorption in MacCuskey reference field». The circles denote ihe absorption «li­ X mited, from Hi, and the histogram that adopted in the prêtent work. The crosses rep­ resent the absorption estimated from the gal­ axy counu -II -It -20 -22 -24 -a falactio totttuda F. Lebrun and I. Grenier: Galaxy and star counts in Taurus and Perseus isc. ;~o.

180. no. galactic longitude

Ftf.2. The distribution or absorption in Taurus and Peraeu». Contours are labelled in 0.1 mag

reference fields which have been selected by MacCuskey for their The rather good agreement between both estimates shows that apparently uniform star distribution. It is therefore unlikely that significant absorption is taking place in the reference fields and these areas contain a normal proportion of Hlt since their clumpy that the map published by MacCuskey is not only wrong in its aspect would have led the observer to reject them as reference zero level, but also in its shape. A corrected map was therefore fields. Then, one can assume that their absorption is entirely constructed by adding the absorption in the reference fields related to Hi and that the N(Hi)fA„ ratio in these areas equals derived from H1 to the la x 1 ° map obtained in Sect.2. This final the standard total gas-to-dust ratio: g - fAT(Hi) + 2N(Hi)\/Api. map is displayed in Fig. 2. Similarly, the 2° x 2° map was cor­ Assuming R - A„IE{B - V) * 4, the Copernicus results rected to allow the comparison with the galaxy counts. The gen­ 21 2 1 (Bohlin et al., 1978) imply Q * 1.45 10 cm" mag" . So, using eral relation between A„ and log Nt is illustrated in Fig. 3. A,t " [WH i) - iV(H to)] IQ, the absorption in the reference fields has been computed and is given in Fig. 1 as a function of the galactic latitude.

3.2. Galaxy counts in the reference fields If one tries to fit the data points displayed in Fig. 3 with a linear

Alternatively, the absorption in the reference fields could be relation similar to Eq.(l), Art being the independent variable, a derived from the galaxy counts in these areas. However, one of the y value of about 0.5 will result with a value of about 1.4 for log N(V goals of the present paper is precisely to determine the calibration This value for y is surprisingly small and, in fact, entirely compat­ of such a derivation; so, in principle, this method cannot be used ible with the value proposed by Shane and Wirtanen which would at this stage. Nevertheless, for comparison, the absorption was suggest that the Lick catalog is essentially magnitude limited. derived from the galaxy counts using the calibration of Strong However, it should be bom in mind that Phillips et al. (1981) have (1983) and is displayed in Fig. 1. But in the following, our estimate shown that the degree of clustering of the Lick catalog may indi­ of the absorption in the reference fields will be based on H i only. cate an angular diameter limited sample. Furthermore the value - «J -

1.0 Lebrun. 1982). Unlike Burstein and Heiles, we compare tracers intercepting the same solid angles in the sky. Moreover, our com­ parison escapes the necessary assumptions of uniform grain size when comparing reddening to absorption. And it escapes partly, except for the determination of the absorption in the reference l.J fields, the assumption of a constant gas-to-dust ratio as assumed • ^ by Strong and Lebrun (1982) and by Strong (1983). However, since the calibration proposed by Strong tums out to be correctly •"\\.' founded and is more accurate titan ours, we suggest to adopt it. The validity of Strong's calibration implies that our deter­

U.5 •.'A\- mination of the absorption in the reference fields from the galaxy counts is correct. As expected, the agreement between the average absorptions estimated from N, and HI implies that the amount of

H2 in the reference fields is negligible. Therefore, since the errors on JV(Hl) are negligible, the uncertainty on the absorption map o.o given in Fig. 2 results only from the errors on the star counts and that affecting the offset of the iV(Ht), A„ relation. This absolute • ' %s ' • uncertainty should be about 0.2mag. Acknowledgements. The authors wish to thank J.A. Paul, A.W. Strong and the referee of the present paper. D.A. Kniffen. for helpful comments. RPG IMHG)

Fig. 3. The relationbetwee n log Nt and Arr The solid line represents the calibra­ References tion proposed by Strong 0983). the interrupted line represents y • 1. rV*f * 100 (Heiles. 1976) and the dotted line represents y - 0.75, N* - SO (Strong and Lebrun. 1982) Burstein. D., Heiles, C: 1978, Astrophys. J. 225, 40 Burstein, D., Heiles. C: 1982, Astron. J. 87, 1165 de Vaucouleurs. G.. Buta, R.: 1983. Astron. J. 88, 939 of log Nf is in that case much lower than the expected value of 1.7. Heiles, C: 1976, Astrophys. J. 204, 379 One may wonder if, as in Oph-Sag, numerous stars in the field Heiles, C, Habing, H.J.: 1974. Asiron. Astrophys. Suppl. 14, 1 may hamper the visibility of galaxies (Lebrun, 1984). However, Kniffen, D.A., Fichtel, CE.: 1981, Astrophys. J. 250, 389 the star density in Tau-Per is much lower than in Oph-Sag. Lebrun, F.: 1984. 8th European I.A.U. meeting, Toulouse The first remark one can make about Fig. 3 is that the relation MacCuskey, S.W.: 1938, Astrophys. J. 88,209 appears non-linear with a break around 1.6mag. Below 1.6mag, Peterson. B.A.. Ellis. R.S., Kibblewhite. E.J., Bridgeland, M.T., the slope is about 0.7, very similar to the value given by Strong Hooley.T., Home, D.: 1979. Astrophys. J. 133, LU3 and Lebrun (1982). Above 1.6mag the relation flattens with a Phillips. S., Ellis. R.S., Strong, A.W.: 1981, Monthly Notices Roy. slope around 0.4. Such a behaviour is almost exactly that pro­ Astron. Soc. 197, 151 posed by Strong (1983) on the basis of a comparison of galaxy Rossano, G.S.: 1980. Astron. J. 85,1218 counts with H i. The calibration of Aet estimates from AT,, with Seldner, M.. Siebers, B., Groth, E.J., Peebles. P.J.E.: 1977, As­ R = 4, recommended by Strong is superposed on the data points iron. J. 82, 249 in Fig. 3. The agreement supports Strong's fundamental assump­ Shane. CD.. Wirtanen. C.A.: 1967. Publ. Lick Obs. 22,1 tion that the ratio of the averages over longitude of A„ and Af (Hi) Shanks, T.. Fong, R., Ellis, R.S., MacGillivray, HT.: 1980, is independent of latitude. If the gas-to-dust ratio is uniform, this Monthly Notices Roy. Astron. Soc. 192, 209 implies that the fraction of the gas in molecular form is indepen­ Strong, A. W.: 1983, Monihly Notices Roy. Astron. Soc. 2*2,1015 dent of latitude. We believe our determination to be more reliable Strong, A.W., Lebrun, F.: 1982. Asiron. Astrophys. 105,159 than the previous ones (Burstein and Heiles, 1978; Strong and Tyson, J.A., Jarvis, J.F.: 1982, Asirophys. J. 230, LI 53 - S3 -

Far-infrared Emission From Nearby Molecular Clouds

F. Boulanger1, R.S. Colien1, M. Gaida2, I. Grenier1'3, M. Koprucu1, R.J. Maddalena1,P.Thaddeus1, and H. Ungerechts1

1 Goddard Institute for Space Studies 2 I. Physikalisches Institut der Universitat zu Koln 3 Service d'Astrophysique, CEN Saclay OPhG/SAp

Abstract We present observations of the main nearby molecular clouds by the Infra-Red Astronomy Satellite (IRAS); these observations are compared with large scale CO observations and some l3CO and Ay data. In each field the foreground and background emission not associated with the molecular gas is subtracted from an analysis of the correlation between infrared and H I emission around the molecular cloud. A good correspondence between CO and IR pictures of the clouds is observed wherever CO observations are available. In most fields, the IR maps indicate the existence of new clouds not yet observed in CO, which shows that the IRAS observations can be of value in planning new molecular observations outside the Galactic plane. Outside regions of star formation, the infrared emission is correlated with the CO, 13CO, and Av data; this correlation indicates that far-infrared and CO brightnesses are equally good tracers of molecular column densities. The scatter from cloud to cloud of the ratio between far-infrared and CO fluxes indicates that both fluxes may be used to measure molecular masses to within a factor of 2. The ratio between H2 column density and far-infrared brightness can be derived from the slope of the infrared-H I correlation within a scaling factor that accounts for the attenuation of the interstellar radiation field by the molecular cloud, and the atomic envelope that protects the molecules from photodissociation. The straight comparaison of the results of the lOOftm-HI and 100/tm-integrated CO emission

20 -2 l (WCO) correlation gives NH, /WCO ratios in the range 0.3 to 1. 10 cm /(K kms~). The attenuation of the ISRF probably accounts for the factor of 4 difference between these values and the previous calibration of the AH-,/WCO ratio (Bloemen et al. 1986). The contribution of embedded stars is negligible for all clouds that are not associated with HII regions. For clouds like Orion which have recently formed high mass stars, the infrared emission represents only a small fraction of the overall luminosity of the associated O and B stais; only the radiation of the youngest stars is significantly absorbed by the molecular cloud. The infrared emission of these clouds is only a measure of the very recent star formation activity, integrated over the last few 10° yrs. . 90 _

I. Introduction Since dust in molecular clouds radiates in the IR energy absorbed from the Inter-Stellar Radiation Field (ISRF) and embedded stars, IR observations of molecular clouds can provide information on star formation, dust distribution and properties, and the ISRF. Myers et al. (1986) studied the stellar content of the most prominent clouds in the first quadrant of the Galaxy from an intercomparison of far-IR, CO and radio-continuum surveys. Far-IR observations have also been used to measure the luminosity of embedded stars in the cores of several nearby molecular clouds (e.g., Sargent et al. 1981; Nordth et al. 1982; Cudlip et al. 1984), and Keene (1981) observed globules in order to study dust properties and the ISRF. The IRAS survey allows an extension of these initial investigations with higher sensitivity and over a larger set of clouds. This paper presents a comparative study of CO and IR observations of the main nearby molecular clouds. Since the spectral distribution of the IR emission of the different components of the interstellar medium are discussed elsewhere (Boulanger and Pérau It et al. 1986), we limit our investigation to lOOftm IRAS observations. We derive general properties of the IR emission of molecular clouds and discuss the extent to which IR observations can be used to study dust distribution and the overall luminosity of embedded stars. The observations and data reduction are described in section II and III. Section IV presents an intercomparison of IR, CO, and Av data, which results are used in section V to discuss the possible use of IR observations to measure molecular column densities and in section VI to estimate the contribution of embedded stars to the IR emission. The conclusions are summarized in section VII.

II. Observations Extensive CO observations of nearby clouds and the Galaxy have been obtained with the Columbia University and Goddard Institute for Space Studies Northern and Southern 1.2m telescopes. These observations recently were synthesized in a single Galactic map by Dame et al. (1986). In this paper, in order to minimize confusion along the line of sight, we limit the IR-CO comparison to the nearby clouds located away from the Galactic plane listed in Table 1. References for the observations are given for all clouds except Chamaeleon; for which an integrated-CO map is presented for the first time in Figure la. Optical extinction which measures dust column densities, is related to IR emission independently of the dust-to-gas ratio and the correlation between CO intensities and gas column density. Our IR-extinction comparison is limited to the Chamaeleon I cloud for which we derived the optical extinction from star counts made by eye on the ESO-SERC plates. A contour map of the extinction relative to reference fields located around the cloud in regions where no CO emission is observed, is presented in Figure lb. The IRAS mission and data products are described in the IRAS Explanatory Supplement(1985V For each cloud, we build IR maps with 1/2° resolution from the - Si -

Zodiacal History File and higher resolution maps by piecing several Sky Flux Images together. To allow direct comparison, all IR maps are smoothed to the resolution of the CO observations.

III. Foreground and Background Emission in the IRAS Data

a) Zodiacal Light The IR emission of the solar system is subtracted from the IRAS data by using the empirical model of Boulanger et al. (1986). Since the zodiacal light depends both on the direction of observation and the position of the earth along its orbit, the subtraction must be made before different observations are combined to produce maps. Such a subtraction can be performed only for the 1/2° resolution maps built from the Zodiacal History File. Since we do not know which observations were used to build the Sky Flux Images, we cannot directly directly subtract the zodiacal light from the maps with 8' resolution. To subtract the zodiacal emission from these maps, we combine 8' and 1/2° resolution observations as follows: an initial 1/2° resolution map is built with zodiacal light subtracted from the Zodiacal History File. A second 1/2° resolution map is then obtained by smoothing the map with a resolution of 8' to a resolution of 1/2°. The Galactic emission is the same in both maps and the difference between the second and first maps is a smoothed map of the zodiacal emission present in the 8' map. Since the structure of the zodiacal tight is essentially smooth, this difference represents a good approximation of the zodiacal emission present in the 8' resolution map. b) Emission Associated with Atomk Gas Emission from atomic gas is subtracted by using H I data and the correlation between the IR brightness and the emission of atomic hydrogen (Boulanger et al. 1985, 1986). The diffuse galactic emission outside molecular clouds and H II regions comes from dust associated with neutral atomic gas and is given by the following integral along the line of sight:

1=1 nH [(s) xd Ed(s) ds where nm, Xd, and Ed are, respectively, the density of atomic hydrogen, the dust-to-gas ratio and the IR emissivity per grain. E,i depends on the absorption properties of dust and the radiation field. At high Galactic latitudes the IR emission is well correlated with the velocity integrated H I emission (Boulanger et al. 1985; 1986) indicating that the dust-to-gas ratio, dust properties, and the interstellar radiation are approximatively uniform along the line of sight. Close to the plane, the line of sight has a longer path through the galaxy over which the ISRF varies significantly, and variations of IR emissivity along the line of sight need to be considered in order to estimate the atomic component of the IR emission from H I data. - 9 2. -

The H I distribution around each cloud is obtained from the Hat Creek, Parkes, and Argentina surveys (Heiles and Habing 1974; Weaver and Williams 1973; Henderson et al. 19S2; and Colomb et al. 1980). We analyse independently the IR-H I correlation in each field by comparing the IR brightness and the H I emission for all lines of sight that do not cross molecular clouds and H II regions. Figure 2a illustrates the result of the comparison in Orion. More than 10 from the galactic plane, we find a good correlation in all cases. However, some clouds in Taurus and Orion extend to lower latitudes. In these two fields, the remote gas observed at low latitudes is located in the outer parts of our Galaxy, in regions where the ISRF is much less intense than in the solar neighbourhood. The IR emissivity of this distant gas is much lower than of the gas in the solar neighbourhood. The difference is such that at more than 5 from the plane, we can neglect emission associated with distant gas. In Orion we exclude distant gas by integrating the H I emission only over velocities lower than 15 km/s. Thr data at high galactic latitudes show that our velocity limit rather uniformly excludes 20% of the local emission. We correct for this exclusion by scaling the integrated H I emission by 1.20. Since Taurus is located in the direction of the anticenter, velocities cannot be used to separate local and remote gas in this region. We base our subtraction in this field on the IR-H I' comparison in figure 2b, which shows that the H I emission above 1200 K kms'1 has a very low IR emissivity. Above that threshold, the H I emission must come predominantly from gas located in the outer parts of the galaxy. In the subtraction, the IR emission associated with this gas is ignored. The slope of the IR-H I relation is derived by a linear least- squares fit (Table 2). The variations of the slope of the IR-H 1 relation from field to field is probably related to changes in the interstellar radiation but might also partly result from changes in the dust-to-gas ratio and dust properties (Boulanger et al. 1986). The accuracy of our subtraction of emission associated with atomic gas is limited by small scale inhomogeneities of the IR emissivity per H atom, the noise of the H I data (typically 15 K kms~l, Heiles 1975), and structures in the distribution of the atomic gas on scales between the IR and H I resolutions. The dominant uncertainly in every map derives from subtraction of the atomic component of the emission. The noise in each field is estimated by computing the root mean square of data points located outside molecular clouds and H II regions (Table 2). _ 93 -

c) Emission from Ionized Gas Part of the molecular clouds in Perseus, Lupus, Ophiucus and Orion overlap with H II regions. For the first three clouds, the emission associated with the ionized gas is so much larger than the emission of the underlying molecular gas that it is difficult to derive a reliable estimate of the emission of the molecular gas from the data. In these cases, we thus limit our anlysis to the parts of the clouds which do not overlap with ionized gas. The Orion A and B clouds lie behind an extended low density H II region associated with the Orion OBI association (Reich 1978). The emission of this H II region does not represent a large fraction of the emission of the molecular clouds and it is possible to subtract it from the data with sufficient accuracy. Since Lyman photons are a major source of heating of dust in H II regions, the emission of ionized gas is subtracted by using radio-continuum observations. We build a radio-continuum map of the Orion region from the 408MHz all-sky survey of Haslam et al. (1982). A smooth galactic background representing the galactic synchrotrpn emission is subtracted from the radio-continuum map by fitting a linear function of the Galactic longitude and latitude over all data points outside the H [I region and the molecular clouds observed in the IR data. The IR and radio-continuum data are then compared over regions where the ionized gas is observed outside of the molecular clouds. The good correlation we find between the IR and radio-continnum brightness permits us to use the radio data to subtract the emission of the low density H II region all over the clouds. The emission of the Orion A and B nebulae is not subtracted from the IR data.

IV. IR, CO, Av Comparison a) Qualitative Comparison After subtraction of the different background and foreground emissions, the IR maps can be compared with the CO and Av data. IR contour maps of the Orion-Monoceros, Taurus-Auriga-Perseus and Chamaeleon clouds presented in figure 4a,b,c, are overlaid on grey scale images of the CO observations to show the good aggreement between the CO and IR pictures of the clouds where CO observations are available. Discrepancies in Perseus and around the California nebula are due to the presence of H II regions. The good coincidence observed in Figure 4 indicates that outside of the Galactic plane, IRAS observations can be used to show or exclude the presence of molecular gas in some region of the sky, what makes them particularly helpful for planning new molecular observations. For example, the IR map of Chamaeleon indicates the existence of several new clouds beside the one already mapped in CO. Further, in the IR data the various clouds in Chamaeleon appear clearly connected by a low brightness envelope. The origin of this low-intensity envelope is not clear: it may be evidence for the existence of a transition layer between the atomic and molecular gas consiting of a mixture of cold atomic and molecular hydrogen without any CO molecules. Deep observations of CO could confirm this suggestion or rule it out. . 9lf-

b) Quantitative Comparison

Since dust in molecular clouds is heated by both embedded stars and the external radiation field, the IR emission studied has two distinct origins. General features of the respective contributions to the IR from embedded stars and external radiation to the IR emission have been discussed on the basis of simple models (Boulanger et al. 1986). The contribution of embedded sources depends on the presence or absence of a star- forming region along the line of sight and does not depend in a simple way on the column density of dust. On the other hand, the emission that results from absorption of the external radiation is directly dependent on the column density of dust along the line of sight so long as the cloud is not totally optically thick to the interstellar radiation. Up to such an optical depth, this last part of the IR emission is expected to be correlated with the column density of dust and, provided tftat the dust-to-gas ratio is uniform throughout the cloud, to the column density of gas. Therefore, a quantitative comparison between the IR emission and the column density of dust or molecular gas appears as an obvious way to discriminate between the respective contributions of embedded stars and external radiation.

Observations of CO and l3CO are most commonly used to map the column density of molecular gas on large scales outside very dense regions, while the distribution of dust is generally inferred from extinction measurements. For each cloud, we compare the IR data with these gas and dust tracers along the same line of sight. Figure 4a,b shows the pixel by pixel correspondence between the CO emission integrated over velocities, WCO, and the lOOpmbrightness for the Chamaeleon I and II clouds; the results of the l3CO - 100 /tmcomparison for the same clouds is presented in figure 4c; Figure 4d shows the correspondence between the lOOjtroand Ay data. The correlation between the IR emission and the various tracers of column density is obvious in the four diagrams in Figure 4a,b,c,d. In 4a, a few points lying well above the correlation line correspond to lines of sight in the direction of the two reflection nebulae visible on the optical plates of the Chamaeleon I cloud. For these particular lines of sight we observe, above the emission associated with the external heating, the radiation absorbed from the stars illuminating the reflection nebulae. Other than these few positions, thee are no embedded stars sufficiently luminous to contribute significantly to the IR brightness of the cloud on the scale of the spatial resolution of this study. Over most of the cloud the IR emission essentially comes from the heating of dust by the external radiation.

The results of the IR - CO comparison for Orion A and Mon R2 are illustrated by Figure Sa,b. Since star formation occurs more actively in Orion than in Chamaeleon, in Figure Sa many more points are scattered above the line of correlation than in Figure 4.a,b., although, as in Chamaeleon, most points are at the bottom J( th? diagram, around the line of correlation. The IR - CO diagrams of Figures 4a.b, and 5a,b present the same basic features, and the analysis of the other clouds of our sample gives similar results. All of these diagrams show that the IR emission can be written as the sum of two terms: the first, correlated _ ir-

with the column density of gas, represents the contribution of the interstellar radiation and the second associated with embedded stars, is significantly positive only over a limited set of regions of star formation. Clouds, like Chamaeleon II, that are not associated with any significant star formation enable us to estimate uncertain ities in the IR-CO correlation. In Figure 5b, the errors in the subtraction of the atomic component of the IR emission (Table 2) account for most of the scatter of the points about the global correlation for the lowest value of WCO. The larger scatter on the right part of the diagram probably results as much from point to point changes in the NH2/WCO ratio than in the IR/NH2 ratio. The IR - CO diagrams provide a way to separate the respective contribution of embedded stars and the interstellar radiation to the IR emission of the molecular clouds studied. For each cloud, we fit a straight line in the IR-CO diagrams, ignoring all points obviously outside the correlation. Since the points excluded represent only a small fraction of the total set of data points, the results of the fits do not depend much on the criteria for excluding points outside the correlation. After subtraction of the IR component correlated with WCO, the IR data can be used to measure the total emission coming from embedded stars and to localize the most important star-forming regions in the clouds. For each cloud, the total IR flux of embedded stars is given in Table 4, with the fraction of the total surface of the cloud where there is a nonnegligible contribution of embedded stars to the IR emission.

V. IR Emission and Molecular Column Densities The integrated intensity of the CO J=l-o line, WCO, is used in most large scale studies to measure molecular column densities. Estimates of the NH2/WCO ratio, derived from independent empirical studies (e.g. Dickinan - 1978, Frerking et al. 1982, Bloemen et al. 1986), are mostly within the range 1 to :i I02°cm~2/(K kms~l). The IR-CO diagrams (Figs. 4.a,b. and S.a,b.) show that within a cloud, the 100 pmbrightness and the CO emission integrated over velocities are equally good tracers of the molecular column density outside star forming regions and optically thick condensations. Changes from cloud to cloud in the IR emissivity per hydrogen molecule are analysed in order to investigate the extent to which IR observations can provide reliable estimates of masses of local molecular clouds. The slopes of the IR-CO relation vary from cloud to cloud by as much as a factor of 3. The scatter in the IR/WCO ratio probably comes partly from changes in the NH2/WCO ratio not related to the IR data, and partly from variations in the IR/NH2 ratio which depends on several parameters that may vary from cloud to cloud: the intensity and spectrum of the interstellar radiation, the dust-to-gas ratio, dust properties, and the spectral distribution of the IR emission. However, as long as the IR/NHI ratio measured for the atomic gas surrounding the clouds and the IR/NH2 ratio depends similarly on these parameters, we can empirically convert the IR fluxes in molecular masses by using the results of the IR-H I comparison (Sect. Ill), independently of the physical conditions within the clouds. Changes in dust properties -st­

and the dust-to-gas ratio which probably are the same in the molecular cloud and in the atomic medium surrounding it, will modify in the same way the molecular and atomic IR emissivity per nucléon for the atomic and molecular gas. On the other hand, since the interstellar radiation gets attenuated by absorption and scattering as it penetrates into the cloud, the IR emissivity per nucléon decreases from the external to the inner parts of the cloud. Further, the attenuation of the radiation field can change the spectral distribution of the IR emission if the grains become colder. Since the 60/*m/100>mcolor temperature of the IR emission has been shown not to change significantly from the molecular cloud to the surrounding atomic medium (Boulanger et al. 1986), we will ignore this last effect. Consequently, the molecular to atomic ratio of the IR emission per nucléon only depends on an average attenuation factor that accounts for the absorption and scattering of the interstellar radiation by the molecular cloud. The attenuation of the interstellar radiation depends on the geometry and the optical depth of the cloud. The CO observations presented in this paper show that the bulk of

the molecular mass is observed in regions where 1 < Av < 3. The molecular to atomic ratio of the IR emission per nucléon has been computed for spherical clouds of various optical depths (Boulanger et ai. 1986). For a total visible extinction through the center of the cloud of 1 and 3, the ratio is 0.67 and 0.4S, respectively. On the basis of this result, we assume that the attenuation factor is approximately 0.5 for all clouds in our sample. With this assumption, we convert the IR fluxes in molecular column densities. The ratios between the CO emission and the molecular column densities derived from the IR data are listed in Table 3. The limited scatter observed from cloud to cloud indicates that both CO and IR data can be used to measure masses of local molecular clouds within a factor of 2.

20 2 _1 We find values of the NH, / WCO ratio in the range 0.6 to 1.910 em~ /{K km s ), which are slightly lower than previous empirical determinations of the NH,/WCO ratio (see review by Bloemen et al. 1986). The lower values we find do not contradict previous determinations because there is probably an additional attenuation of the ISRF due to the atomic envelope which protects the molecule from photodissociation. Assuming that this protective envelope has an optical depth of 1 in the ultraviolet we estimated that at the surface of the molecular cloud the intensity of the interstellar radiation field is half less intense than in the surrounding atomic gas. Taking into account this additional source of

20 -J -1 attenuation we derive iVH2/WCO ratios in the range 1.2 to 3.810 cm /(/f ferns ), in agreement with previous empirical calibrations. There is a tendency for small clouds to have lower JVH2/WCO ratios than the largest clouds of our sample. - »* - VI. Luminosity of Embedded Stars

For each cloud, the total flux coming from embedded stars is given in Table 4; for several clouds we do not give any number in the Table because the contribution of embedded stars is too small to be separated from the emission of dust heated by the external radiation field. The numbers in Table 4 show that the overall luminosity of embedded stars is comparable or greater than the energy absorbed from the external radiation only for clouds like Orion or Monoceros which have recently formed high mass stars. The other clouds do contain many embedded sources but these objects account for a negligible fraction of the infrared luminosity of the clouds (see Boulanger and Pérault 1987). Infrared luminosities have been commonly used to measure the efficiency of star formation in molecular clouds in the inner parts of the Galaxy (Caux et al. 1985, Myers et al. 1986, Solomon et al. 1986). However, detailed studies of molecular clouds associated with O and B stars (Leisawitz 1986, Boulanger and Perault 1987) show that the infrared luminosity of the clouds represent only a small fraction of the luminosity of the OB associations. Most of the O and B stars lie far away from the cloud which, consequently, absorbs only a small fraction of the stellar radiation. This result implies that IR observations measure only the luminosity of the youngest stars associated with the clouds; therefore they do not provide information about the star formation efficiency of molecular clouds over their whole life-time. In the following we illustrate this statement by analyzing in detail the origin of the infrared emission of the Orion A and B molecular clouds. The total infrared luminosity of the Orion A and B cloud is 1106 igMost of this radiation comes from the brightest parts of the cloud; by integrating the emission within a circular aperture of 1.5 deg2 centered around the Orion A and B nebulae we get about

s s 6095 of the total luminosity of the clouds, 4.210 LQ, and 1.8 10 £sfor Orion A and B, respectively. The Orion A nebula is heated by the Trapezium stars and the embedded cluster in the Kleinman-Low Nebula. The overall luminosity of stars in the Trapezium cluster is about 4 105 (Reeves 1978); the embedded cluster has a luminosity of the order

5 s of 10 L0(Werner 1982). The infrared luminosity, 4 10 £0, comes clearly from the embedded cluster plus about half of the radiation of the stars of the Trapezium which lie just at the surface of the cloud. For the Orion B Nebula the infrared luminosity represents only 1/3 of the luminosity of zeta Orionis (an 09.71 star, Humprheys 1978) which is the main heat source for the dust. Outside the Orion A and B nebula, the infrared luminosity

6 of the molecular cloud is 4 10 I0. This luminosity which represents a small fraction of the total luminosity of the Orion OB I association, 5 10s (Boulanger and Pérault 1987). Therefore, the radiation of most of the stars belonging to the OB association does not get absorbed by the molecular cloud. There is an efficient conversion to infrared radiation only for the embedded cluster and the stars of the Trapezium which lie very close to the cloud; as all these stars are younger than about 2 10° yrs (Blaauw 1964) we see that the IR emission of the molecular clouds measures only the luminosity of the stars formed - 31 -

during the last 2 10s yrs while the oldest subgroup of the OB association are about 107 yrs old (Blaauw 1964).

VTI. Summary Our large-scale intercomparison of IR, CO, and Av data over a large sample of nearby molecular clouds leads to the following conclusions: 1. After subtraction of background and foreground emission associated with H I gas, molecular clouds located outside of the plane can be clearly identified in the IRAS data. 2. A good correspondence between the corrected IR and CO pictures of the clouds is observed away from H II regions, indicating that IRAS observations can be valuable in planning new CO surveys outside the galactic plane. 3. For all lines of sight, other than a limited number that cross star forming regions, the IR emission is linearly correlated with CO, l3CO and Av measurements. The correlation indicates that within a cloud which is not too massive, IR and CO are equally good tracers of total molecular density. 4. Outside molecular clouds and H II regions, the IR emission is well correlated with the H

I emission. This correlation enables us to determine the Nn2/IR ratio in molecular clouds within an average attenuation factor that accounts for the absorption and scattering of the interstellar radiation by the cloud. Simple models of radiation transfer suggest that the molecular to atomic ratio of the IR emission per nucléon is about 0.5. The atomic enveloppe which protects the molecules from photodissociation produces an additional attenuation of about O.S. 5. The ratio between the CO emission and the molecular column densities derived from the IR data is constant from cloud to cloud within a factor 3. Such limited scatter demonstrates that both IR and CO data can provide reasonable estimates of the masses of local molecular clouds. 6. The straight comparison of the results of the IR-HI and the IR-CO correlation leads

20 2 1 to //H,/WCO ratios in the range 0.3 to 110 cm' /(K kms' ). The difference between these values which the previous empirical estimates of the NH„ /WCO ratio can be accounted by the fact that the ISRF is attenuated inside the molecular clouds. 7. The IR-CO comparison enables us to separate the respective contributions of embedded stars and the external radiation to the IR emission of molecular clouds. The overall contribution of embedded stars is negligible for all clouds that do not contain any high mass stars. For clouds like Orion which have recently formed high mass stars, the infrared emission represents only a small fraction of the overall luminosity of the associated O and B stars; only the radiation of the youngest stars is significantly absorbed by the molecular cloud. The infrared emission of these clouds is only a measure of the very recent star formation activity, integrated over the last 10° yrs. 99 -

References

Beichman, C., A., Neugebauer, G., Habing, H. J., Clegg, P. E., and Chester, T. J., 1985, IRAS Explanatory Supplement. Bloemen, J. B. G. M., Strong, A. W., Blitz, L., Cohen, R. S., Dame, T. M. , Grabelsky, D. A., Hermsen, W., Lebrun, F., Mayer- Hasselwander, H. A., and Thaddeus, P., 1986, A. A., 154. 25. Boulanger, F., Baud, B., and van Albada, G. D., 1985, A. A., 144. L9. Boulanger, F-, Falgarone, E., Perault, M., and Puget J. L., 1986, in preparation. Colomb, F. R. , Poppel, W. G. L. , and Heiles, C., 1980, A. A. Suppl. 4.0, 47. Cudlip, w. , Emerson, J. P., Furniss, I., Glencross, W. M., Jennings, R. E., King, K. J., Lightfoot, J. F-, and Towlson, W. A., 1984, M.N.R.A.S., Hi, 563. Dame, T. M., et al., 1986, In preparation. Dickman, R. L. , 1978, Ap. J. Suppl., 3_7, 407. De Grus, E., et al., 1986, in preparation. Frerking, M. A., Langer, W. D., and Wilson, R. W., 1982, Ap. J., 262. 590. Grenier, I., et al., 1986, in preparation. Haslam, C. G. T., Salter, C. J., Stoffel, H., and Wilson, W. E., 1982, A. A. Suppl., 42, l. Heiles, C, and Habing, H. J., 1974, A. A. Suppl., 14., 1. Heiles, C. , 1975, A. A. Suppl. 2J), 37. Henderson, A. P., Jackson, P. D., and Kerr, F. J., 1982, Ap. J., 263. 182. Keene, J., 1981, Ap. J., 245. 115. Maddalena, R. J., Morris, M., Moscowitz, J., and Thaddeus, P., 1986, Ap. J., 303, 375. Murphy, D. C, May, J., and Cohen, R. S., 1986, A. A., Submitted. Myers, P. C., Dame, T. M., Thaddeus, P., Cohen, R., Silverberg, R. F., Dwek, E., and Hauser, M. G., 1986, Ap. J., 301. 398. Nordth, H. L., van Duinen, R. J., Sargent, A. I., Fridlund, C. V. M., Aalders, J. W. G., and Beintema, D., 1982, A. A., 115. 308. Reich, W., 1978, A. A., M, 407. Sargent, A. T., van Duinen, R. J., Nordh, H. L., and Aalders, J. W. G., 1981, 94, 377. Ungerechts, H., and Thaddeus, P., 1986, Ap. J., submitted. . 100 -

TABLE 1 CO OBSERVATIONS

Clouds Angular Resolution Reference

Cepheus 30 Grenier et al. (1986) Chamaeleon 8 Lupus 8 Murphy et al. (1986) Ophiuchus 8 de Grus et al. (1986) orion-Monoceros 8 Maddalena et al. (1986) Taurus-Auriga-Perseus 30 Ungerechts and Thaddeus(1986 - 101 -

TABLE 2 IR-H I COMPARISON

Clouds I100uu/NHI a100um (1) (MJy/sr)/(1020cm"2) MJy/sr

Cepheus 0.75 2.0 Chamaeleon 1.3 1.2 Lupus 1.9 4.0 Monoceros 1.0 1.6 Ophiuchus 1.9 2.6 Orion A and B 1.3 1.6 Orion Filaments 1.0 1.6 Taurus-Auriga-• Perseus 1.0 2.6

(1) : Noise (W) in the IR maps after subtraction of zodiacal light and emission associated with atomic gas. - 10» _ TABLE 3

IR-CO COMPARISON

I wco Clouds 100um/ NH2/wco (MJy/sr)/(Kxkm/s) l(Tcm": /(Kxtan/s)

Auriga 0.8 0.8 Cepheus 0.7 1.0 Chamaeleon I 1.3 1.0 Chamaeleon II 0.8 0.6 Lupus 1.2 0.7 Mon R2 1.2 1.2 Northern Orion Filament 1.9 1.9 Ophiuchus -

Orion A 2.2 1.7 Orion B 2.5 1.9 Southern Orion Filament 1.6 1.6 Taurus 0.8 0.8 - I03- TABLE 4 RESPECTIVE CONTRIBUTIONS OF EMBEDDED STARS AND INTER-STELLAR RADIATION fISR) TO THE IR EMISSION

Clouds F100um(ISR) F100um(s_tars> S MJy/sr MJy/sr (1)

Auriga 0.12 2.1 10"2 0.01 Cepheus 0.19 - -

Chamaeleon I 1.3 10"2 3.7 10"3 0.11

Chamaeleon II 2.0 10"2 - - Lupus - -

Mon R2 2.3 10"2 4.3 10"2 - Northern Orion Filament 2.1 10"2 - - Ophiuchus - - - Orion A 0.14 0.74 0.30

Orion B 0.15 0.36 0.21

Southern Orion Filament 1.5 10"2 - - Taurus 0.29 — _

(1): Fraction of the total surface of the cloud rfhere the contribution of embedded stars to the oberved brigthness is not negligible. - I 0 U - Figures Captions Fig. 1 -a) Contour map of integrated CO emission, WCO, of the Chamaeleon I and II clouds. The lowest contour is 2 K*lan/s, and the separation between contour is 2 K*km/s. The border of the surveyed region is indicated by the thick, continuous line, b) Contour map of the visible extinction of Chamaeleon I. The extinction is measured relative to reference fields located around the cloud in regions where no CO emission is observed.

Fig. 2 -Comparison of lOOum brightness, liooum' and H * integrated emission, WHI. (a) IiQOum brightness versus WHI in Orion at Galactic latitudes smaller than -10°. (b) The same comparison in Taurus at Galactic latitudes smaller than -3 . In both plots, each cross represents a 1/2"pixel.

Fig. 3 -Contour maps of the lOOum emission of (a) the Orion- Monoceros, (b) Taurus-Auriga-Perseus, and (c) Chamaeleon molecular clouds. The zodiacal light and the emission associated with the atomic gas are subtracted from the maps. In Orion, we have further subtracted the IR emission of the H II region associated with Barnard's loop. The IR maps are overlaid on grey scale images of the CO emission to show the good agreement between the CO and IR pictures of the clouds. In the three maps, contour values are: 2, 4, 6, 8, 10, 13, 16, 20, 24, 29, 36, 45, 62, 78, 128, and 228 MJy/sr.

Fig. 4 -Pixel-by-pixel inter-comparison of lOOum, CO, 13C0, and

Av measurements of the Chamaeleon clouds. Fig. 4 a.b presents •^lOOum versus WCO for Chamaeleon I and II. The 100um-13CO and 100um-Av comparisons are respectively shown in figures 4c. and 4d. In each of the four plots, a cross represents a 1/2° pixel.

Fig. 5 -Pixel-by-pixel comparison between lOOum brightness and CO integrated emission for (a) the Orion A and (b) Mon R2 clouds. Each cross represents a 1/2° pixel. Fi. la.

O" «•-iH.ia » IK.H C«NIOIMS< mon ;.oa Il J:. 00 CO EMISSION* V -l*B 10 |£C Ml S ' »• - 1 1 1 | 1 1 1 1 1 ' i i i -

•• ^L r?°

AS »•

1 1 1 I I I 1 i ii i

AK n

CHAMAELEON AVIS A VIS. DATA FROM 1.81 TO 5-35 . CONTOURS FROM £.00 BY 0.50 TO 5.CO GAUSSIAN AVERAGE . FUHM - 0.13 3-D

1.0 3.0 •« (1950) - 10? - F;' V 2-i

3.0

:2.0 XR- HX û(w>*s I

Jl.O

50.0 i9.0

!8.0

17.0

16.0

1S.0

1U.0 o effîGVa pLp.-w^7r-—*>rd; *^& eus

13.0

12.0 s 11.0 4ïff 10.0

9.0

6.0

7.0

6.0

5.0

t (Jl o o o o o o o o 1} o o o o s l 3 o - I lilt

Fiv 2.h_

Xlo«/vm (Wiy/A^

50.0 X*- H X Tftu^^,

10.0

+ + +

30.0

++ *$ ++ •*** +

20.0

10.0

i\) J= m a ro OJ u) o o o o o •c en o ru •e CO oo o ru o o o o Q o o o o o o o o o o o o o o o o o o o o o

\WH3 (K hm«"'i (09 -

fi ; M n t i î i i i i i i i M i i M i i i i i i N n i

CVI 61 to CI O T- u. o

tD

D

G3

3)

'' I I I I I I I I I I I 'III nnninininininnininnriinininininininin 8 Bi)Tii).ipi]io;Niyi i i i i. Ms n a -

1—i—i—i—i—i—i i i i r ~i—I—I—I—I—I—I—I—1—I—I—I S

o

1 ;\ .... \. ' r-"., I o 1 u

J I I L_l I I I I I i i i i L inuiintnininmtnminintn inr

v> _ Ml -

CHBHRELEON I

IJ00UM,HJT/SR) i—i—m—r i i r

* +

I I I I I I I l I I l L 1 o IU •c a> CD ro IU IV» l\J M ro o IV) •C 07 co o lu •c co CD o o o o O o• o o o o o O o o o O NCO

F^ */£• I I 2.

CHBMflELEON II

IIOOUHWJT/SH1

«0.0

30.0 _

20.0 _

10.0 _

0.0

Fîe\. ^k -113-

CHHMRELEQN I û.t\c\ 3L

IlD0UHtHJT/5m

F»-»->->— »~ IM n> i\> lu i\> c» ai ai u>° o ctu oryiCcnooorOiCcncDOiyc a>o O O O O O £ 2 oo ooooooooooooo oi 3>

0"S - IV -

LE RAYONNEMENT GAMMA GALACTIQUE DIFFUS

" Ce qui empêche, de. conpnend/ie. c'ut que. c'e&t t/iap iûiple " Jean Cocteau Ce chapitre regroupe deux étude* qui ont pour but de confronter l'émission gamma diffuse et la distribution du gaz Interstellaire atoaique et moléculaire. La première analyse (Strong et al., 1988, soumis i Astron. A»trophy».) traite l'ensemble du disque galactique découpé en anneaux concentriques afin de déceler un éventuel gradient d'émissivité gamma. Le résultat prouve en effet que la densité des rayons cosiiques diainue faiblement vers l'extérieur de la Galaxie. Les statistiques montrent par ailleurs une légère différence entre les spectres des éiisslvités gamma liées aux milieux atomique et moléculaire. Le dernier semble plus mou. Les problèmes de propagation des rayons cosmiques dans les nuages denses (piégeage, effet d'écran, production d'électrons secondaires, etc..) pourraient jouer un rôle dans ce phénomène. Hais une population de sources gamma de basse énergie distribuée comme le CO simulerait le mime effet. Malheureusement, à cause de la mauvaise résolution de COS-B, la corrélation entre 70 et 150 MeV est très délicate. Les émissions gamma liées aux régions HI et H, se ressemblent fortement et toute différence de comportement a nécessairement une signification statistique faible. C'est pourquoi seules les mesures au delà de 150 MeV ont été utilisées pour calibrer le rapport 20 7 11

Mf2/WC0. La valeur de (2,3 +/- 0,3) 10 molécules cm K km r ainsi déterminée est certainement l'estimation à grande échelle la plus fiabla à ce jour, au moins en tant que valeur moyenne. La seconde analyse (Grenier et Lebrun, soumis à Astrophys. J.) concerne plutôt le milieu proche, dans le complexe de Céphé* et Cassiopée. Au delà de 150 MeV, elle conduit à une valeur de 20 NH./WCO identique à la moyenne galactique bien que moins précise (2,3 •/- 1,2 10 ). Cil* aboutit également à un spectre de l'émissivité gamma liée au gaz atomique en parfait accord avec celui mesuré au cercle solaire dans l'étude précédente. Aussi la densité des rayons cosmiques apparaît-elle "normale" au voisinage du Soleil. Elle ne diffère pas sensiblement de la moyenne galactique à 10 kpc. Un tel accord montre de plus que l'émissivité gamma du gaz atomique proche est à peu près connue entre 70 et 5000 MeV. Il n'en est pas de même pour celle du gaz moléculaire au dessous de 150 MeV. En essayant de comprendre sa valeur anormalement élevée, un fort excès ponctuel a été isolé. Sa signification statistique (4 o-) est suffisante pour le considérer comme la dernière-née des sources COS-B, mais elle devra être confirmée par de nouvelles observations. Comme toute source COS-B qui te respecte, elle n'est pas identifiée. Une radlogalaxie serait pourtant un candidat intéressant, d'autant que la source apparaît comme la cinquième du ciel en dessous de 150 MeV. Une fois la source prise en compte dans la corrélation gamma/gaz à basse énergie, l'émissivité des nuages moléculaires s'effondre. L'effet est statistiquement faible mais il mérite d'être mentionné car 11 semble cette fois difficile è expliquer par un* limitation instrumental*. Il paraît donc important aujourd'hui d'étudier systématiquement l'émission gamma diffuse de* nuage* proche* afin d'établir clairement l'origine de ce* résultats inattendu* * bas** énergie. En effet, Céphée-Catliopée et 1* Taur**u-P*r*é* sont dis complexe* calmas oil I** observations CO doivent tracer correctement la matière. Or il* appiraident mal en gammi au dessous de 150 MeVI - nr -

Proc. XXth Int. Cosmic Ray Conf., Moscow, 1, 1987.

GAMMA-BAY/GAS CORRELATIONS OVER THE WHOLE GALAXY A. W. Strong1, J.B.G.M. Bloemen2 , T.M. Dame6 LA. Grenier3, W. Hermsen4, F. Lebrun3 L.-A. Nyman*. A.M.T. Pollock?, P. Thaddeus* 1. Mmx-Plmnck-Iutltnt fttr Phjriik sad Aâtrophyiik, lnatitut (Or Extr*t«rr»trUc&* Physlk, Garch- in( - b*i - Mttschra, F.R.G. 2. A»tronomy Dtpartmut, Uaivmltjr of California, Barkalay, U.S.A. 3. Sarrica d'Aatropbyalqua, Caatra d'Btado NucMaina da Saclajr, Franca 4. Laboratory for Spaea Raaaareh, Laldan, Tha Natharlands 5. EXOSAT Obaarvmtory, Aatrophyaica DlTialon, Spaca Sclanca Dapartmant, ESTBC, Noordwijk, 0. Harvard^anutaaonian Cantar for Aatrophyaiea, , Cambridge, U.S.A. ABSTRACT A new compilation of CO and HI survey* covering the entire Galactic plane hat been used together with the final COS-B data to study the distribution off-ray emisaivity and the gas content of the Galaxy. The agreement between the model and the observations is excellent. A summary of the results for i-ray energies >150 Me V is given. 1. INTRODUCTION. The correlation between Galactic T-rays and gas tracers (ET and CO) was studied by Bloemen et aL (1986: hereafter B86), using COS-B data and taking advantage of large-scale CO surveys using the Columbia 1.2m telescope. The distribution of n-ray emissivity per H atom in the Galaxy as a function of energy was determined, together with a 'n-ray value' for the conversion factor from the integrated temperature of the CO line to molecular hydrogen (Hi ) column density and an estimate of the total Hi mass of the Galaxy. The analysis of B86 used the COS-B data from the first 55 (out of 65) observation periods together with the 1st and 2nd quadrant and Carina Columbia CO surveys and various HI surveys. The latitude range used, -4.5' < b < 6.5s, had essentially complete CO coverage in these quadrants. The gas surveys were first divided into Galactocentric distance bins using a Galactic rotation curve, and the analysis was done in three energy ranges (70-150,150-300 and 300-5000 MeV). The emissivity gradients (and by implication therefore the cosmic-ray gradient) were found to be surprisingly small, especially at high energies (>3O0 MeV), when compared to the distribution of candidate cosmic-ray sources in the Galaxy. Since B86 various developments have enabled the analysis to be extended. First the final COS-B database with its final calibrations and corrections (Strong et al. 1987a) is now available. Second (and more important), new CO surveys by the Columbia group (see Dame et al. 1987) both in the Southern and Northern hemisphere now provide essentially complete CO coverage of the Galactic plane for \b\ < 7", together with extensive coverage at higher latitudes. In addition the fitting and error analysis has been refined to allow the rapid appraisal of models even when the number of parameters is large. The correlation analysis using these data is the subject of a forthcoming paper (Strong et al. 1987b); here we summarize some of the conclusions. 2. METHOD. The method consists of fitting the observed

A - £ £&*'-* + ***oo*) + fiche +1% + £ /*A (1) > « where « is the -y-ray emissivity in the rth ring, Nsi,i is the column density of atomic

hydrogen, WCo,i •* the velocity-integrated CO brightness temperaure, Y is the effective 1-ray value for the conversion factor from CO integrated temperature to Hi column density, ficUc •» the inverse-Compton emission (we will skip further details here), 1% - 116 -

is an isotropic background (cosmie+instrumental) corrected for temporal and angular variations (see Strong et al. 1987a for details), /* is the flux of the k'th point source included in the model and h is the angular distribution for a source of unit strength situated at the k'th source. The tilde indicates the convolution with the COS-B point- spread function, which is quite accurately known from studies of the Vela pulsar (see Mayer-Hasselwander 1985). The Galactocentric rings chosen for this analysis were : 2-4,4-8,8-10,10-12,12-15 and >15 kpc (we take Ao=10 kpc). This choice of rings is a compromise between the known limitations of the T-ray data and the requirement to resolve variations in the inner Galaxy. The choice of the 4-8 kpc ring is based on our wish to separate the main concentration of molecular gas in the inner Galaxy. The basic assumption characteristic of this analysis is that the 7-ray emissivity (and thus the cosmic-ray density) is a function of radius only; no attempt is made explicitly to distinguish ann/interarm contrasts, for example. However if such spiral structure effects are present, they will still be covered to some extent by the model via the YWco.i term, since the CO is at least in part a tracer of spiral structure. In this case the value of Y derived from 7-rays will of course include any cosmic-ray enhancement in spiral arms and therefore be an overestimate of the true Wco-to-iVs, factor, which we denote by 'X'. Equation (1) can be applied to each energy range separately, but we have to con­ sider also cases where all energy ranges are considered together, in order to investigate models with energy-independent parameters. For example we treat the case of energy- independent shape for q(R) and energy independent Y by making fitst o the three ranges simultaneously. The four moat intense "y-ray sources were explicitly included in the model via the final term. The sources are: the Vela pulsar, the Crab pulsar, Geminga (2CG195+04) and 2CG78+01 (Swanenberg et al. 1981). Note that the fluxes of these sources are free parameters, and in fact the resulting spectra are a useful by-product of this analysis. The error analysis is performed using the information matrix technique described by Strong (1985a). The theoretical basis of the likelihood ratio method used is described inB86. The fits were done principally for |4| < 9.5" because the CO coverage is generally sufficient in this range. Since however the convolution means that coverage beyond this range is strictly necessary, most of the fits were also done for \b\ < 5.5" to check for differences. In this range the coverage is complete for the convolution but the statistics and dynamical range are smaller.

8. RESULTS. The general analysis was done in three energy ranges in order to examine any energy variation of the parameters. For simplicity however we here restrict attention to the two high ranges (150-300 and 300-5000 MeV) within which there is no indication of energy-dependent effects. Fig 1 compares the result of the model prediction with the observations for the energy range 150-5000 MeV. The fit gives a good representation of the data in all four quadrants; of particular interest is the fourth quadrant, where die CO coverage is new and a detailed comparison with 7-rays is done for the first time. An exception to the good agreement is in the range 350* < I < 10* where the relation between 7-rays and the CO tracer of Hj is known to ba anomalous (see Blitz et al. 1985); this region was excluded from the fits. Other less conspicuous discrepancies are usually attributable to known 7-ray sources not included in the model; for example in the Carina region (around l m 284") there is a clear excess. A detailed study of such excesses for the first quadrant has been given in Pollock et al. (1985), and the extension of this work to the whole Galaxy is in progress. Fig 2 shows the radial distribution of emissivity in this model. - M? -

The value of y in equation (1) is a measure of the 'physical' CO-to-JTj conversion factor X. In B86, the entire 70-5000 MeV range was used to derive this factor, giving JC=(2.75±0.35) xlO20 mol. cm-3 (K km s~l)~l , but a more reliable estimate may follow from the 150-5000 MeV range. One of the main reasons for believing this is the poor angular resolution of COS-B at low energies. For the 70-5000 MeV range, our analysis of the whole Galaxy gives the same X-value as found by B86; for the 150-5000 MeV range we find X = (2.3±0.3) xlO20 mol. cm"1 (K km s"1)"1 , so slightly lower than in B86.

' ISO 140 120 100 SO 60 40 20 0 340 320 300 2S0 260 240 220 200 ISO GALACTIC LGNGITUOE

FIG 1. Longitude distribution of observed and predicted -y-ray intensity in the range 150-5000 Me V, for \b\ < 5.5". Lower curve; inverse Compton emission, second curve: HI contribution, third curve: total including Ha contribution, top curve total including Ha and the Jour sources mentioned in the text. The isotropie background is included in all predictions. Vertical bars represent COS-B data with ±lo statistical errors.

The effect of possible variations in Y with radius R was considered; no significant variation was found. We can show that even if y is forced to be a function of R m the fitting procedure (by fixing V(28 kpc] = 0.5 for example) the resulting Y value for the inner Galaxy (2

In B80 the value for the total mass of the molecular hydrogen in the inner Galaxy

(2-10 kpc) was found to be ~ 10* M@ ; this assumed however that the first and fourth quadrant* are fully symmetric. Bronfman et al. (1987) did not make this assumption s and derived 1.2 10 Ms from a combination of the Northern and Southern hemisphere Columbia data and applying the X-value obtained in BB6. Scaling to the X-value obtained here gives 1.0 10* Afs . Ill

Galactic Ejnissivity Gradient

CM

(J, ISO- 300 M«V JK 300-9000 M»V

Galactocentrlc radius lepc

FIG a. Radial dependence of 7-roy emittivity per B atom for 150-S00 and >S0OO MeV. Fig 2 confirms the remarkably flat emissivity distribution found in B86 which is quite unlike that of potential cosmic-ray sources such as supernovae or SNRs. A theo­ retical approach to explain this (and the variations with energy not discussed here) in terms of cosmic-ray propagation effects has recently been proposed by Bloemen (OG2.2- 5: this conference ). The emissivity values at R = 10 kpc in Fig 2 agree well with those found in the completely independent determination for the local region using COS-B data, HI and Galaxy counts at intermediate latitudes (Strong et al. 1985b).

REFERENCES Blitz, L. et al. (1985) Attwn. Attrophy». 143, 267 Bloemen, J.B.6.M. et al. (198S) Attron. Attrophy». 154,25 Bronfman, L. et ai, (1987) Attrophy». J. in pre»» Dam*, T.M. et al. (1987) Attrophy». J. tubmitted Mayer-Hast elwasder, H.A. (1985) Explanatory Supplement to the COS-B Databate Pollock, A.M.T. ft al. (1985) Attron. Attrophy: 140, 352 RUey, P.A., Wolfendale, A.W. (1984) J. Phyt. G 10,1269 Strong, A.W. (1985a) Attron. Attrophy». ISO, 273 Strong, A.W. «t of. (1985b) Proe. XIXICRCX, 317 Strong, A.W. tt al. (1987a) Attron. Attrophy». Supp 07, 283 Strong, A.W. ft al. (1987b) in preparation Swanenberg B. ft al. (1981) Attrophy». J. Lett. 243, L69 - 113 -

THE RADIAL DISTRIBUTION OF GALACTIC GAMMA RAYS IV: The Whole Galaxy

A.W. Strong1, J.B.G.M. Bloemen3-*-8, T.M. Dame7 LA. Grenier3 W.Hernuen* P. Lebrun3 L.-A. Nyman7 A. Pollock5 P.A. Thaddeu»7

1. Max-Planck Institut fur Phyiik und As trophy tik, Institut fur Extraterrestrische Phyiik, Garching-bei- Mûnchen, Germany 2. Dept. of Astronomy, University of California, Berkeley, CA 9-4720, U.S.A. 3. Service d'Astrophysique, Centre d'Etudes Nucléaires de Saclay, France 4. Laboratory for Space Research, Leiden, The Netherlands 5. Space Science Department of the European Space Agency, ESTEC, Noordwijk, The Netherlands 6. Leiden Observatory, Leiden, The Netherlands 7. Harvard-Smithsonian Centre for Astrophysics, Cambridge, MA 02138, U.S.A.

For submission to: Astronomy and Astrophysics Main Journal

Section heading: Galactic Structure and Dynamics

Running Title: Radial distribution of galactic gamma rays.

Send proofs to: A.W. Strong, Max-Planck Institut fur Physik uud Astrophysik, Institut fur Extrater- restrische Physik, Garching-bai-Mûnchen, Germany

Send offprint requests to: A.W. Strong, Max-Planck Institut fur Physik und Astrophysik, Institut fur Extraterrestrische Physik, Garching-bei-Mûnchen, Germany

Thesaurus Codes: 04.01.1,07.36.1,19.94.1

Keywords: COS-B, gamma rays, cosmic rays, interstellar clouds

2 _ /20 -

THE RADIAL DISTRIBUTION OF GALACTIC GAMMA RAYS IV. The Whole Galaxy

ABSTRACT The correlation between diffuse galactic gamma rays and gat tracers is studied using the final COS-B database and HI and CO surveys covering the entire galactic plane. A good quantitative fit to the T-rays is obtained, with a small gradient in the "y-ray emissivity per hydrogen atom. The average ratio of Hi column density to integrated CO temperature is determined, the best estimate being 2.3 ± 0.3 1030 molecules cm~a(K km t-1)"1 . The corresponding matt of molecular hydrogen in the inner galaxy, derived using both 1st and 4th quadrants, is l.Ox 10' MQ . It is shown on a statistical basis that the softer 7-ray spectrum towards the inner Galaxy found in previous work can be attributed to a steeper emistivity gradient at low energies and/ or to a softer 7-ray spectrum of the émission distributed like molecular gat. Our statistical tettt suggett the latter is the better model. A steeper emittivity gradient at low energies could be related to cosmic -ray spectral variations in the Galaxy, to different distributions of cosmic-ray electrons and nuclei, or to a contribution from discrete sources. A softer spectrum for the emission associated with molecular clouds may be physically related to the clouds themselves (i.e. cosmic-ray spectral variations) or to an associated discrete source distribution.

3 . 12.1 -

submitted to Astronomy and Astrophysics 19 Feb 1988

1. INTRODUCTION The good quantitative correlation between galactic it-rays and gas tracers (HI and CO emission) in the Galaxy is the main evidence that interactions of cosmic rays with interstellar gas are responsible for most of the diffuse Galactic "y-ray emission for ener­ gies between 100 MeV and a few GeV. Prior to 1982 it was impossible fully to exploit this correlation for lack of adequate coverage ( in particular in latitude) in the CO surveys, nec­ essary because of the relatively low angular resolution of T-ray telescopes. The situation improved dramatically with the availability of the first large-scale 'superbeam' surveys in CO from the Columbia 1.2m telescope, since they provide complete sampling over a large enough latitude range to allow a reliable comparison with the T-ray data. Lebrun et al. (1983) used Columbia CO data (Dame 1983, Dame and Thaddeus 1985) and HI data to analyse the -/-ray emission observed by COS-B above 300 MeV in the first galactic quadrant and to place limits on the CO-to-ifj conversion factor. In this analysis the density of cosmic rays (in this energy range mainly nucléons are involved) was assumed to be uniform in the Galaxy. In papers I and H of the present series (Bloemen et al. 1984a,b) the correlation between the EI and -y-ray distributions beyond the solar circle (2nd and 3rd quadrants) was studied, ignoring molecular gas. The analysis of Paper III (Bloemen et at. 1986; see also Bloemen, 1985) used low-latitude COS-B data taken from the first 55 (out of 65) observation periods together with the 1st and 2nd quadrant and Carina Columbia CO surveys, and various HI surveys. The latitude range used, —4.5" < I < 6.5", has essentially complete CO coverage in these quadrants. The gas surveys were first divided into galacto- centric distance bins using the rotation curves of Gordon and Burton (1976) and Blitz et al. (1980), as modified by Kulkarni et ai. (1982); the bins chosen were 2-8,8-10,10-15 and >15 kpc. The analysis was done in three energy ranges and allowed for spatial variations in ir-ray emissivity. In Paper HI the main result was the determination of the T-ray emissivity gradients in the three energy ranges, the establishment of the requirement for an energy-dependent model, and the derivation of a value for the CO-to-Hi conversion factor together with a new estimate for the mass of molecular hydrogen in the inner Galaxy. The emissivity gradients were found to be surprisingly small, especially at high energies (>300 MeV), when compared to the distribution of candidate cosmic-ray sources in the Galaxy. Several developments have now made it possible to improve the analysis of Paper III. These include the availability of the total ÇOS-B database with its final calibration and the updating of the CO database to include the 3rd and 4th quadrants, a new 1st quadrant survey and various other new surveys from the Columbia telescopes. The new CO surveys, as well as covering the entire longitude range, now allow the analysis of a greater latitude range than before. In addition the fitting and error analysis has been refined to allow the rapid appraisal of models even when the number of parameters is large.

An important and independent indication that the underlying assumption is correct, viz, that the bulk of the Tr-ray emission originates in interactions of cosmic rays with

4 - 121 -

submitted to Astronomy and Astrophysics 19 Feb 1988

gas, is provided by the observation of 7-rays from the Orion cloud complex (Bloemen et al. 1984c). The derived ratio of molecular hydrogen column density to integrated CO brightness temperature was 2,6± 1.2 1020 molecules cm~J(K km s-1)-1 , which is in agreement with the large-scale analysis (see Section 2) and indicates that -/-ray sources are not playing a major role. Similarly studies at intermediate latitudes (Strong et al. 1982a, Strong 1985a, Strong et al. 1985) show that most of the emission can be explained as the sum of emission from atomic and molecular gas (plus a small inverse Compton component). 2. METSQP The method is the same as in Paper HI; it consists of fitting the 7-ray data to the function: A = £ &*"•'+ 2yvïrco,,) + fiche +1% + Y. hh W * k where g,- is the 7-ray emissivity in the i'th ring, Nui.i is the column density of atomic hydrogen, WCo,i » the velocity-integrated CO brightness temperaure, Y is the apparent 7-ray value for the conversion factor from CO integrated temperature to Hi column density, (assuming that g, applies equally to atomic and molecular gas), ftchc a the inverse-Compton emission, 1% is an isotropic background (cosmic-(-instrumental) corrected for temporal and angular variations (see Strong et al. 1987 for details), /* is the flux of the k'th point source included in the model, h is the distribution for a source of unit strength situated at the k'th source. The tilde indicates the convolution with the COS-B point-spread function, which is quite accurately known from pre-launch calibrations and studies of the Vela pulsar (see Mayer-Hasselwander 1965). The free parameters of the model are, for each energy range: 9t,»_i-e, Y, fie, 1% &&d fk,k=i-*.

The galactocentric rings chosen for the present analysis were : 2-4, 4-8 8-10, 10-12, 12-15 and >15 kpc (we take R©= 10 kpc). This choice is a compromise between the known limitations (statistics and angular resolution) of the 7-ray data and the requirement to resolve variations in the inner Galaxy. The choice of the 4-8 ring is based on our wish to separate the main concentration of molecular gas in the inner Galaxy. The basic assumption characteristic of this method is that the 7-ray emissivity is a function of radius only; no attempt is made explicitly to resolve non-axisymmetric structure in the emissivity distribution (e.g. related to spiral arms) or to distinguish arm/interarm emissivity contrasts. However if contrasts are present, they will still be covered to some extent by the model via the YWco.i term, since the CO is at least in part a tracer of spiral structure. In this case the value of Y derived from 7-rays will of course include the cosmic-ray contrast, and therefore be an overestimate of the true Ng,IWco ratio factor, which we denote by 'X'. The relation of y to X is further discussed in Section 4. Equation (1) is valid for each energy range separately, but we have to study also cases where all energy ranges are considered together, in order to investigate models with energy- independent parameters. For example we treat the case of energy-independent shape for q(R) and energy independent Y by making fits to the three ranges simultaneously. The

5 - 113 -

submitted to Astronomy and Astrophysics 19 Feb 1988

number of parameters to be determined then becomes quite large but the fitting routines are sufficiently robust to locate the maximum likelihood solution in about 20 iterations. The inverse Compton emission was calculated using the model described in Strong (1985a). Although this differs slightly from that used in Paper HI, this has a small effect, and in any case the absolute level is determined via the free parameters he derived from fits to the 7-ray data. The four most intense 7-ray sources were explicitly included in the model via the final term. The sources are: the Vela pulsar, the Crab pulsar, Geminga (2CG195+4) and CG78+01. Note that the fluxes of these sources are free parameters, and in fact the resulting spectra are a useful by-product of this analysis. The error analysis is performed using the information matrix technique described by Strong (1985b). The theoretical basis of the likelihood ratio method used is described in Paper III, and a detailed treatment can be found in Kendall and Stuart (1973), (Ch. 24). Here we use the notation: AN = number of parameters of model relative to the general model, A In L = log-likelihood of model relative to general model. Then -2 A In £ is distributed as X\N- We define p as the probability of obtaining a value of -2 AlnL greater than that observed, for a given model. The 'significance' is then 1- p . This method is a formally exact treatment of the statistical problem once the model has been defined. The 'general model' as given by equation (1) is found to be good enough to reproduce the observations as well as could be expected given the quality of the data, and so is probably sufficient for the purposes of the likelihood method, which requires that the 'general model' actually is the 'true' model for some particular set of values of the parameters (Kendall and Stuart (1973), Ch. 23). Actually we may have rather more parameters than are strictly necesssary to get a satisfactory fit; however this in no way affects the validity of the tests made with respect to this model. The fits were done principally for |6| < 9.5° since the CO coverage is almost complete in this range. Since however the convolution means that coverage beyond this range is strictly necessary, most of the fits were also done with |i| < 5.5° to check for differences. In this range the coverage is complete for the convolution but the statistics and dynamical range are smaller. Additional runs with 10° < ( < 270° and 90° < I < 350° were done to check for differences between the 1st and 4th quadrants. 3. EATA 3.1 Gamma rays The final COS-B database (Mayer-Hasselwander 1985) contains a total of 65 obser­ vations, of which 55 include regions near enough to the galactic plane to be of interest to be used in the present analysis. The baseline instrumental response as determined from pre-launch accelerator calibration is given by Mayer-Hasselwander (1985). The relative sensitivity of the instrument for these observations as well as the variations in background with time during the COS-B mission was determined by Strong et al. (1987). The method by which skymaps (consisting of counts and effective exposure) are constructed from the database is described by Strong et al. (1987). The maps used are those included with

6 . 12.4 -

submitted to Astronomy and Astrophysics 19 Feb 1988

the final COS-B database. These assume an input spectrum of E~l-BS for the exposure calculation; expected variations from this value have negligible effect on the exposure. 3.2 HI surveys The 21-cm line surveys of Weaver and Williams (1973; 10° £ l .$ 250°) and Kerr et al. (1986; 240° £ I £ 350°), and part of the survey of Strong et al. (1982b; 350° J$ I £ 10°), were used to obtain HI column densities for |6| S, 10°. At \b\ £ 10°, the survey of Heiles and Habing (1974) was used for the northern sky {6 Z, - 30°) and that of Heiles and Cleary (1979) for the southern sky. The surveys were first corrected to obtain the brightness temperature and then column densities were derived assuming a uniform spin temperature of 125 K, as described in Papers I Se II. The construction of column density maps for separate galacto-centric rings is described in Paper III. Maps of the total HI column density from these data are shown in Fig 1(a), convolved with the COS-B point-spread function in the three energy ranges.

3.3 CO surveys The CO (J=l-0) data base used for our analysis is a composite of several surveys carried out over the last 5 years with the Columbia 1.2 m telescopes in New York City and on Cerro Tololo in Chile, and includes basically all the data in the large-scale panorama presented by Dame et al. (1987). The combined survey covers the entire galactic plane and is essentially complete up to |6| s 7s(particularly for the 1st and 4th galactic quadrants), with several large extensions to higher latitudes. The spatial resolution is 0.5°. The construction of velocity-integrated CO maps for separate galacto-centric rings is described in Paper HI. Maps of the integrated CO temperature are shown in Fig 1(b), convolved with the COS-B point-spread function in the three energy ranges. 4. DISCUSSION OF FITTING RESULTS Table 1 summarizes the fits performed, and Table 2 summarizes the resulting param­ eters, the log-likelihood values and the error estimates as determined from the information matrix. Table 3 gives the fitted intensities of the four point sources included in the model. In the following discussion the reader should refer to these tables for details. Note that the Galactic Centre region 350" < I < 10" was excluded in all fits, for the reasons discussed in Section 5. 4.1 General A general statement can be made that, within the quoted uncertainties, the parameter values are consistent for different choices of the fitting region both in latitude (\b\ < 9.5° or |6| < 5.5° , cases 1 and lb) ) and longitude (10° < / < 350°, 10° < / < 270°, 90° < I < 350°, cases IN and IS ). A systematic effect can however be seen comparing the analyses using the separate regions (case IN and IS ) with that for the total (case 1 ): the separate regions give a larger q(R) gradient and a correspondingly lower Y value. This effect is most marked in the 70-150 MeV range, and may be related to the fact that when the separate regions are

7 - I7.S-

submitted to Astronomy and Astrophysics 19 Feb 1988

used, less weight is given to the inner Galaxy in the fits (which always include the whole 90° < / < 270° region). This effect is discussed later. It seems most natural to give equal weight to the inner and outer galaxy and to use all the data together. Therefore in what follows we shall concentrate on the results from the total region (\b\ < 9.5°, 10°

8 - IZl -

submitted to Astronomy and Astrophysics 19 Feb 1988

1 , we have A lui =8, AN =10 and p =0.1. Hence there is barely an indication of an energy dependence in the shape of q(R). This model (case 3 ) is therefore both an improvement over the totally energy inde­ pendent one, and not significantly worse than the general model. 4.3.3 Tests of case 2 The next model in order of increasing complexity comes from relaxing instead the condition of energy independent shape for q(R), case 2 . We can test the energy dependence of q(R) by comparing case 2 with 4 . Then A In L = 3.3, AN =10 and p = 0.8. There is less evidence for an energy dependence than in the test of section 4.3.2. Since Y remains energy independent in this case, comparing to the general model amounts to testing for energy dependence of Y. Comparing case 2 with 1 we have A In X = 8.2, AJV =2 and p =3 10-4. Thus there is an indication for energy dependence of Y, but it should be noted that the effect is actually due to a difference between the ranges 70-150 MeV and >150 MeV. The model studied in case 2 is therefore unsatisfactory by both criteria: it is not an improvement over the totally energy independent model, and is significantly worse than the general model. Note however that the cases 2 and 3 have practically the same value otlnL, indicating that the fit is 'equally good' in the two cases. The difference in the significance of the energy dependent effects only arises from the different number of parameters in the two models. The nature of the low energy Y value is further studied in Section 4.6, where radial variations of this parameter are considered. 4.3.4 The 'best' model From the results of the tests described above, case 3 is the only one which is both an improvement on the simple energy independent model and not significantly worse than the most general model. It thus qualifies for the classification 'best' model. As mentioned above, the energy dependence in this model is actually due to a difference in Y between the ranges 70-150 MeV and >150 MeV (see Fig 2): Y(70-150 MeV) = (3.3± 0.5) 1020 molecules cm"J(K km s"1)-1 and Y(>150 MeV) = (2.5± 0.3) 1030 molecules cm~2(K km s"1)"1 . It is gratifying that this 'best' model is actually rather simple, with the same shape of q(R) in each energy range, and all the energy dependence in the parameter Y. However the systematic effect discussed in Section 4.1 remains a worry; we return to this problem in sections 4.6 and 4.7 where the nature of the low-energy Y value is discussed further. 4.4 The value of the CO-to-/^ conversion factor. The choice of a 'best' value for the 'physical' conversion factor from CO integrated temperature to Hi column density must depend on how we choose to treat the fact that the fitted value of Y in equation (1) is larger at low energies (cf. case 3 ). Since X is by definition "energy-independent, there are (at least) four possible situations:

9 _ 11? -

submitted to Astronomy and Astrophysics 19 Feb 1988

(i) systematic effects due to the poor angular resolution for 70-150 MeV (ii) other components (e.g. sources) with steep spectra which also correlate with the CO distribution (iii) effects related to cosmic-ray propagation or production in molecular clouds (iv) energy-dependent spiral arm/interam cosmic-ray contrasts In each case it is clear that the value of Y in the low energy range is the least re­ liable indicator of X. The value also depends slightly on whether we assume an energy independent shape for q(R) or not. Our best estimate is made using cases 1 and 3 for the 150-300 and 300-5000 MeV ranges; the value of Y ranges from 2.0 to 2.5 1020 molecules cnrs(K km s"1)-1 , so we adopt X = 2.3± 0.3 1020 molecules cm"2(K km s-1)"1 . However, it should be kept in mind that this value is an upper limit if a population of unresolved f-ray point sources exists with an angular distribution similar to that of the molecular gas, or if the cosmic-ray density is enhanced in molecular clouds. A detailed discussion on the application of X and a comparison with other determinations in given by Bloemen et al. (1986). 4.5 Radial dependence of emissivity. All the results in Table 2 show that q(R) is relatively insensitive to the choice of data or model, the differences being within the quoted error bars. Fig 3a shows q(R) in each energy range in the case of gradient and Y energy dependent (case 1 ). The radial variation is small as found in Paper in, amounting to at most a factor 2 between the inner and outer Galaxy, and 1.5 between the inner Galaxy and the Solar circle. Fig 3b shows q(R) in the case of an energy independent shape (case 3 ), which is consistent with the data as shown previously. Comparing with Fig 3a the differences are within the quoted errors for each case. The main difference is in the inner Galaxy at low energies (see the discussion in Section 4.6). Fig 3c shows q(R) for the case of a free shape with energy-independent Y (case 2 ). Again the main difference is in the inner Galaxy at low energies. The effect on q(R) of possible variations of Y with radius are considered in the next section. 4.6 Radial variation of Y. In the models discussed so far, Y has been assumed independent of galactocentric radius. Since this is not necessarily true, and indeed claims of substantial variation have been made together with the suggestion that this has a large effect on q(R) (Bhat et al. 1985 a,b), we explicitly test the variation of Y with R against the 7-ray data. We define

r = y(2 - 8Jfepe)/r(> 8kpe)

and repeat the fits for a range of values of r (case lr ). Fig 4 shows L(r) for the 3 energy ranges and for the total range. This figure shows that the data above 150 MeV are consistent with r=1.0 (no variation of Y with R). Since in the low energy range Y is not a good indicator of X, we use the two higher ranges to determine the formal limits on r : r = 0.8±0.2. Further using these ranges we find that A In I (r=0.5)=1.5, corresponding

10 _ l« -

submitted to Astronomy and Astrophysics IS Feb 1988

to p =0.08 (1 degree of freedom) so that a factor 2 variation in Y is improbable but not excluded. Although this argues against an X variation as large as that adopted by Bhat et al. al. (1985a,b) we show in Fig 3d the corresponding q(R) for r = 0.5. The gradient is only slightly larger than for r = 1.0, the difference being limited by the fact that the HI component is not affected by the X value. We conclude that variations of X in the Galaxy, even if present, do not have a major influence on our derived emissivity gradients. This may be important since the existence of a factor 2 increase in the O/H ratio in the inner Galaxy has recently been clearly established by IR emission-line observations (Lester et al. 1987), although CO intensity may not scale with CO abundance (see Kutner and Leung 1985, Solomon et al. 1987).

The 70-150 MeV case is somewhat different, Fig 4 showing a clear indication for r > 1. (The likelihood value actually continues slowly to increase aymptotically as Y increases .) * The formal significance is given by Ami =4.8 with AN =1, giving p = 2 10-3. This means that in the inner rings essentially all the emission is attributed to the CO-Iike component. To illustrate this, Fig 3(e) shows q(R) for the range 1.0

Also note that for r>1.2, q(R) increases with R in the inner galaxy for 70-150 MeV ; although not impossible a priori it does seem very unlikely that such a sharp change in behaviour with respect to the higher energy ranges occurs. If we insist that q(R) does not increase with R, then we find that r < 1.2, i.e., consistent with higher energies. We therefore adopt the assumption that r = 1.0 in drawing conclusions in this paper. Under this assumption we are left with the conclusion that Y is larger in the 70-150 MeV range. The preferred model is then case 3 (shape of q(R) energy independent), with Y(70-150 MeV)= 3.3±0.5 10M molecules cm-*(K km s"1)"1 .

* The effect is evidently related to the energy dependence of Y found in Sec 4.3.3. This effect was already evident in the fits using the 1st and 4th quadrants separately (Sec 4.1), where more weight is given to the anticentre and the average Y value appears smaller than when these quadrants are both included in the fit.

11 . 1*9 -

submitted to Astronomy and Astrophysics 19 Feb 1988

4.7 Again, the 'best' model. Although we are still left with the conclusion that case 3 (with r=1.0) is the preferred model after the test in the previous section, we have lost some confidence in this model. The reason is the finding in Sec. 4.6 of a formally high significance for an effect (an increase in q(R) with R for 70-150 MeV) which can almost certainly be excluded on physical grounds. This may indicate an incompleteness of the model for the low energy range, but it is hard to improve on it because of the wide COS-B point-spread function in this range. Another indication of the same problem is the systematic difference (see Section 4.1) between the fits using the 1st and 4th quadrants separately and the fit including both quadrants. In summary: although our tests indicate that an energy-dependent Y is preferred of an energy-dependent shape of q(R), we have indications that the energy dependence may in fact be due to another effect not encompassed by our model. 5. COMPARISON OF MODEL WITH THE OBSERVATIONS. It is not possible nor desirable to show the comparison of the data with each of the model fits iisted in Table 2, so the 'best' model (in the sense of section 4.3.4) is chosen, corresponding to an energy independent shape of q(R) (case Z ). Fig S shows longitude distributions for |6| < 5.5". The four fitted point sources are included in the plot. In this presentation the other cases appear almost indistinguishable from this one. Considering that systematic uncertainties in exposure can lead to fluctua­ tions of up to 10% on top of the statistical noise, the fits are generally quite satisfactory, and show that the model is adequate to account for the bulk of the diffuse emission along the entire galactic plane. Towards the inner Galaxy, \l\ < 60°, HI and H% produce approximately equal con­ tributions to the -7-ray intensity, indicating approximate equality of the mass in these components. Arm-interam contrasts in emissivity are not explicitly included in our fit, and there is no indication that they are required by the data. However note that to some extent the model can adjust to such contrasts via the Y term, since the CO is at least in part a tracer of spiral structure. Despite the general good agreement there are some significant deviations deserving attention. Most prominent is the galactic centre region, |(| < 10°, which is well known to deviate strongly from the behaviour along the rest of the galactic plane (Blitz et al. 1985, Bania 1986, Stacy et al. 1987), and which was not included in the present fits for this reason. Stacy et al. (1987) have recently shown that this discrepancy may be accounted for by wide-line molecular clouds in the vicinity of the Galactic centre. Also prominent are the excesses in 330" < I < 345" for >300 MeV. These are approx­ imately coincident with the positions of the sources 2CG333 and 2CG342 (Swanenburg et al. 1981). There are corresponding peaks at lower energies, but they are much less evident in the longitude plots, consistent with the hard spectrum found by Swanenburg et al. Also noticeable is the excess around 2CG235+1 and in the Carina region (280° < 1 <290°). A

12 - 130 -

submitted to Astronomy and Astrophysics 19 Feb 1988

significant dip visible at all energies around 60 ° < 1 < 70s corresponds to a well-known interarm region and may be the only hint in favour of an arm-interarm emissivity contrast. Fig 6 shows latitude profiles of predicted and observed "y-rays for the inner (300° < I < 60°) and outer (60° < / < 300°) Galaxy. The agreement is satisfactory, except that there seems to be an excess relative to the model for |A| >10° (this latitude range was not included in the fits). 6. COMPARISON WITH LOCAL EMISSIVITIES PROM INTERMEDIATE LATITUDES The average emissivity in the distance range 8 < R < 12 kpc from the present work can be compared with the independent estimate of the local value within a few hundred parsecs using 7-ray , galaxy count and gas data at intermediate latitudes (10° < |&| < 20°). Strong et al. (1985) derived the values (1.10 ± 0.14,0.76 ± 0.09,0.68 ± 0.09) lO-28 sr-1 s-1 for the same three energy ranges as used here. By comparison the present values for 8 < R < 12 kpc are, taking ease 3 as representative, (1.02 ± 0.10,0.65 ± 0.06,0.62 ± 0.06) in the same units. The agreement (within the quoted errors) is remarkable given the very different latitude range and survey data used, and known problems with the galaxy counts calibration (see Leburn 1986). 7. THE MASS OF MOLECULAR HYDROGEN IN THE INNER GALAXY As noted in Section 5, Fig 5 shows that the contributions to the "7-ray emission from HI and Hj are about equal in the inner Galaxy, and this immediately leads to Mg, ~ Mm. Henderson et al.(1982) find Mm = 0.9 109 Af© in the inner galaxy. In Paper III the value for the total mass of the molecular hydrogen in the inner Galaxy s (2-10 kpc) was found to be — 10 AfQ ; this assumed however that the first and fourth quadrants are fully symmetric. Bronfman et al. (1987) did not make this assumption and derived 1.2 10s MQ from a combination of the Northern and Southern hemisphere Columbia data and applying the X-value obtained in Paper HI. Scaling to the X-value obtained here gives 1.0 10s M® . Again, for the reasons mentioned in Section 4.4, this value should in fact be regarded as an upper limit. Further discussion on this point and a comparison with previous determinations is given by Bloemen et al. (1986). 8. COMPARISON WITH PAPERS I-m Since this is the fourth paper in a series addressing the same general theme, it is important to point out and resolve any differences in conclusions between this and earlier papers. Papers I and II were concerned only with the outer galaxy and longitudes 90° < / < 270°, and used only total HI column densities. Paper I concluded that the emissivity gradient is undetectable in the outer galaxy above 300 MeV, but significant for 70-150 MeV. Paper II came to a similar conclusion using three galactocentric rings (10-12,12-15 and >15 kpc). The present work confirms the small gradient at high energy, but finds no detectable energy dependence of the gradient, even in the outer galaxy. Reference to Paper I shows that the conclusion was based on a local value for q(70-150 MeV) from early f-ny -galaxy count correlation studies (Strong et al. 1982a); subsequent analyses have reduced the local q-value, so that the anti-centre gradient is correspondingly reduced,

13 _ 131 —

submitted to Astronomy and Astrophysics 19 Feb 1988

and is consistent with the present paper. In Paper II the effect found was not extremely significant, and the limits on the gradient are fully consistent with the present analysis. In any case the neglect of molecular hydrogen in Papers I and II means that these analyses are superseded by the present one. In Paper III the first three Galactic quadrants and the Carina region were used, and the analysis principle was the same as in this paper. It was concluded that Y is energy- independent to within the accuracy of the analysis, while the shape of q(R) was found to depend on energy. This was then interpreted as a steeper gradient for cosmic-ray electrons compared to nuclei. The present paper confirms the small gradient at high energies, but ascribes the energy-dependence instead to the Y-value, although there may be some incompleteness in our model (Sec. 4.7). There are therefore some differences in the conclusions of Papers in and IV which require comment. In Paper III the evidence for the energy-dependent gradient was not overwhelming, and the energy-dependent Y was stated as also acceptable. There is there­ fore no real inconsistency between the two analyses; the better data and treatment has simply produced a more reliable result. The improvements include the extension of the CO surveys to include the entire southern Galaxy, as well as extensive upgrading and filling -in for many other regions. Also , the final COS-B database with its additional observations and final corrections has been used. Further, the analysis has improved technically with fully automated minimization and error calculation for the very many parameters involved. For all these reasons, we regard the present analysis being the most reliable up to the present time. 9. COMPARISON WITH OTHER RESULTS 9.1 The 'Durham' Analyses The Durham group have published many results on the topic studied here (e.g. Bhat et al 1985a,b 1986). Although their findings are not drastically different from ours, there is some discrepancy, at least in the conclusions. This is largely a result of the difference in philosphy between our apporach and that of the Durham group. We determine Y and q(R) simultaneously from the large-scale -y-ray -gas correlation study (via the different spatial structures of HI and CO in selected Galactic rings) whereas the Durham group generally uses non- 7-ray estimates of X and then sets Y=X in order to determine q(R), and vice-versa. A detailed comparison of the methods is given by Bloemen (1988). They find a lower Y value for some local clouds, and this was considered as support for their independent (non- "7-ray ) estimates. The most recent value derived by this group (Bhat et al. (1986)) is X = 1.5 1020 molecules cm"2(K km s"1)"1 locally, falling with decreasing R. Their local value is thus barely consistent with ours; further, their implied inner Galaxy value of 0.8 10" molecules cm-2(K km s_l)~l is inconsistent with our value discussed in Section 4.4 (see Fig 2), even considering our case where Y is a function of R. Note however that in a later work (Szabelski et al. 1987) they use as input to a 7-ray analysis a value of 2.0 1020 molecules cm~2(K km s~x)~l , which is consistent with our value, and naturally leads to emissivity gradients similar to ours.

14 - 131 -

submitted to Astronomy and Astrophysics 19 Feb 1988

9.2 The 'Great Ring' The emissivity gradients derived here can be compared with the results of Harding and Stecker (1985). They claim a large peak in JI(R) for 3.5 < R < 5.5 kpc for the 4th quadrant, based on COS-B data >100 MeV and >300 MeV. Fig 5 shows however that we get an excellent St to the data >150 MeV with no such peak in q(R). The large peak found by Harding and Stecker appears to originate from the effect on their unfolding procedure of the sharp edge in the longitudinal T-ray distribution at I = 330", which is due at least in part from discrete sources in this region as described in Section 5. The fact that the excess in this region is not present in the two lower energy ranges argues against attributing it to enhanced emissivity in the '5 kpc ring*. The result of Harding and Stecker can be attributed to their use of integrated |6| < 10" Tf-ray profiles without considering the latitude information, and the absence of a treatment of the instrumental resolution in their analysis. 10. CONCLUSIONS The diffuse galactic T-ray emission in the range 70-5000 MeV is well represented by the sum of contributions from atomic and molecular hydrogen with a small inverse- Compton component. The required emissivity gradient is small, with a maximum variation for the solar circle to the inner Galaxy of a factor 2. The gradient is much smaller than that of the distribution of supernova remnants or pulsars.Therefore the arguments sometimes made invoking the K-ray gradient as direct support for cosmic-ray origin in such objects cannot be substantiated; cosmic-ray propagation effects have to be taken into account. Our 'best model' is one in which Y is energy dependent, while the shape of the emissivity variation is energy independent. The steeper f-ray spectrum towards the inner Galaxy is then accounted by a Y-vaiue increasing at low energies by about 40%. Although an energy dependent emissivity variation can also fit the data, our statistical tests indicate that it is less satisfactory when the number of degrees of freedom are taken into account. The energy dependence of Y leads us to doubt the reliability of Y as a measure of X in the 70-150 MeV range, and hence we use the fits for > 150 MeV for our 'best value' of X: 2.3±0.3 1010 molecules cm~3(K km s'1)-1 . This leads to an estimate s for the mass of molecular hydrogen in the inner Galaxy of 1.0 10 M0 in good agreement with our previous estimates and now based on the CO surveys of both the northern and southern Galaxy. The energy dependence of Y is an interesting result perhaps related to physical pro­ cesses associated with cosmic-ray propagation in molecular clouds, such as production of secondary electrons. Alternatively, steep-spectrum 7-ray sources spatially correlated with clouds would give a similar effect.

ACKNOWLEDGEMENT

15 _ 12% -

submitted to Astronomy and Astrophysics 19 Feb 1988

L.-À. N. acknowledges financial support from the Swedish National Science Reseach Council. We thank J.G. Stacy for useful comments on this paper.

ie _ 134 -

submitted to Astronomy and Astrophysics 19 Feb 1988

REFERENCES Bania T M (1986) Astrophys. J. 308, 868 Bhat C I, Mayer C J, Wolfendale A W (1985a) Nature 314, 511 Bhat C L, Mayer C J, Wolfendale A W (1985b) Proe XIX Int. Cosmic Ray Conf. 1, 336 Bhat C L et al (1986) Phil. Trans. R. Soc. Lond A 319, 249 Blitz L, Bloemen J B G M, Hermsen W, Bania T M (1985) Astron. Astrophys. 143, 267 Blitz L, Fich M, Stark A A (1980) in Interstellar Molecules ed. B. Andrew, Reidel, Dordrecht, p. 213 Bloemen J B G M (1985) Thesis, University of Leiden Bloemen J B G M (1987) Astrophys. J. Lett 31T, LIS Bloemen J B G M, Blitz L, Hermsen W (1984a) Astron. Astrophys. 279,136 (Paper I) Bloemen JBGM, Bennett K, Bigami G F, Blitz L, Caraveo P A, Gottwald M, Hermsen W, Lebrun F, Mayer-Hasselwander H A, Strong A W (1984b) Astron. Aatrophys. 135,12 (Paper H) Bloemen JBGM, Caraveo P A, Hermsen W, Lebrun F, Maddalena R J, Strong A W, Thaddeus P (1984c) Astron. Astrophys. 139, 37 Bloemen JBGM, Reich P, Reich W, Schlickeiser R, (1988 ) Astron. Astrophys., submitted Bloemen JBGM, Strong A W, Blitz L, Cohen R S, Dame T M, Grabelsky D A, Hermsen W, Lebrun F, Mayer-Hasselwander H A, Thaddeus P (1986) Astron. Astrophys. 154, 25 (Paper HI) Bloemen J B G M, (1988 ) Ann. Rev. Astron. Astrophys, in press Bronfman L, Cohen R S, Alvarez H, May J, Thaddeus P (1987) Astrophys. J. in press, Dec 15? Dame T M (1983) Ph.D. Thesis, Columbia University Dame T M, Thaddeus P (1985) Astrophys. J. 297, 751 Dame, T.M.,Ungerechts, H., Cohen, R.S., de Geus, E., Grenier, I., May, J., Murphy, D.C., Nyman, L.-A, Thaddeus, P.: 1987, Astrophys. J., 322, 706 Gordon M A, Burton W B (1976) Astrophys. J. 208, 346 Harding A K, Stecker F W (1985) Astrophys. J. 201, 471 Heiles C , Cleary, M N (1979) Australian J. Phys. Suppl.47,1 Heilet, C , Habing, H J (1974) Astron. Astrophys. Suppl.U, 1 Henderson, A P , Jackson P D, Kerr F J (1982) Astrophys. 7.263,182 Kerr, F J , Bowers, P F , Jackson, P D , Kerr, M (1986) Astron. Astrophys. Suppl. 66, 373 Kendall M G, Stuart A (1973) The Advanced Theory of Statistics Vol II, Charles Griffin and Co., London Kulk&rni S R, Blitz L, Heiles C (1982) Astrophys. J. Lett. 250, L63 Kutner M L, Leung C M (1985) Astrophys. J. 201, 188 Lebrun F (1986) Astrophys. J. 300,18 Lebrun F , Bennett K, Bign&mi G F, Buccheri R, Caraveo P A, Gottwald M, Hermsen W, Kanbach G, Mayer-Hasselwander H A, Montmerle T, Paul J A, Sacco B, Strong A W, Willi R D (1983) Astrophys. J. 274, 231

17 _ isr -

submitted to Astronomy and Astrophysics 19 Feb 1988

Lester D F.Dinerstein H L, Werner M W, Watson D M, Genzel R, Storey J W V, (1987) Astrophys. J 320, 573 Mayer-Hasselwander H A(1985) Explanatory Supplement to the COS-B Database (avail­ able from K.Bennett, Space Science Dept, ESTEC, Noordwijk, The Netherlands) Reich P, Reich W (1988 ) Astron. Astrophys., in press Solomon P M, Rivolo A R, Barrett J ,Yahil A (1987) Astrophys. J 319, 730 Stacy J G, Dame T M, Thaddeus P (1987 ) Proe. XX Int. Cosmic Ray Conf. 1, 117 Strong A W (1985a) Astron. Astrophys. 145, 81 Strong A W (1985b) Astron. Astrophys. 150, 273 Strong A W, Bloemen J B G M, Hermsen W, Mayer-Hasselwander H A (1985 ) Proc XIX Int. Cosmic Ray Conf. 1, 317 Strong A W, Bignami G F, Bloemen J B G M, Buccheri R, Caraveo P A, Hermsen W, Kanbach G, Lebrun F, Mayer-Hasselwander H A, Paul J A, Wills R (1982a) Astron. Astrophys. 115,404 Strong A W, Bloemen J B G M, Hermsen W, Lebrun F, Mayer-Hasselwander H A, Buccheri R (1987) Astron. Astrophys. Supp 87, 283 Strong A W, Riley, P A , Osborne, J L , Murray, J D ( 1982b) Monthly Notices Roy. Astron. Soc. 201,495 Szabelski J, Mayer C J, Richardson K M, Rogers M J, Wolfendale A W (1987 ) Proe. XX Int. Cosmic Ray Conf. 1, 133 Swanenburg, B N , Bennett K, Bignami G F, Buccheri R, Caraveo P A, Hermsen W, Kanbach G, Lichti G G , Masnou J L, Mayer- Hasselwander H A, Paul J A, Sacco B, Scarsi L, Wills R D ( 1982b) Astrophys. J. Lett. 243, L69 Weaver, H F , Williams, O R W (1973) Astron. Astrophys. Suppl.S, 1

18 . 136 -

submitted to Astronomy and Astrophysics 19 Feb 1988

TABLE 1 Summary of fits. Case longitude latitude gradient Y r No.parmNc s range range energy energy (d«g) (dee) depend­ depend­ ence ence

1 10/350 -9.5/9.5 dep dep 1.0 36 lb -5.5/5.5 IN 10/270 -9.5/9.5 IS 90/350 2 10/350 -9.5/9.5 dep. ind 1.0 34 2b -5.5/5.5 3 10/350 -9.5/9.5 ind. dep 1.0 26 3b -5.S/5.5 4 10/350 -9.5/9.5 ind. ind 1.0 24 4b 10/350 -5.5/5.5 ind. ind 1.0 24 lr 10/350 -9.5/9.5 dep. dep 0.5-3.0 37 Blank entries indicate previous value applies.

19 . i a-» -

submitted to Astronomy and Astrophysics 19 Feb 1988

TABLE 2a

70-150 MeV

Case ?i 92 93 94 ?B 96 *BT" Y -InL -£ln£ 1 0.98 0.87 0.96 0.71 0.84 0.83 6.39 5.01 13182.1838008.8 lb 1.08 1.03 1.05 0.77 1.07 1.11 5.30 4.38 8097.26 23772.24 IN 2.12 1.35 1.12 0.89 0.96 0.80 6.04 3.22 10131.64 IS 1.82 1.12 1.23 0.93 1.08 0.90 5.67 2.93 9642.39 2 1.62 1.45 1.23 0.97 1.06 0.75 5.86 2.60 38016.97 2b 1.53 1.46 1.25 0.96 1.29 1.08 4.79 2.80 23775.53 3 1.39 1.26 1.11 0.94 0.83 0.70 6.15 3.30 38016.71 3b 1.40 1.34 1.16 1.02 0.90 0.80 5.56 3.12 23779.9 4 1.60 1.40 1.23 1.04 0.93 0.79 5.74 2.70 38020.28 4b 1.55 1.47 1.27 1.11 0.99 0.89 5.06 2.70 23780.99 lr 1.72 1.29 0.91 0.67 0.80 0.85 6.44 5.50 13186.78 * 0.54 0.54 1.00 0.73 0.86 0.82 6.35 4.67 13179.44 ** 150-300 MeV 1 1.28 0.92 0.74 0.60 0.59 0.42 2.60 1.97 12433.84 lb 1.07 0.79 0.69 0.56 0.61 0.44 2.64 2.45 7838.63 IN 2.28 1.17 0.84 0.68 0.67 0.41 2.37 1.41 9538.85 IS 3.60 1.36 0.81 0.75 0.70 0.43 2.29 0.46 9040.39 2 1.07 0.76 0.68 0.54 0.56 0.43 2.73 2.60 2b 0.98 0.72 0.66 0.52 0.58 0.44 2.74 2.80 3 0.88 0.79 0.70 0.59 0.52 0.44 2.64 2.51 3b 0.81 0.78 0.67 0.59 0.52 0.46 2.61 2.66 4 0.86 0.76 0.66 0.56 0.50 0.43 2.71 2.70 4b 0.81 0.76 0.66 0.58 0.51 0.46 2.62 2.70 lr 1.97 1.20 0.71 0.59 0.59 0.43 2.61 2.15 12433.22 * 300-5000 MeV 1 0.77 0.81 0.66 0.63 0.43 0.40 2.22 2.25 12392.75 lb 0.71 0.84 0.66 0.66 0.42 0.41 2.11 2.22 7836.35 IN 0.00 0.87 0.67 0.64 0.42 0.40 2.19 2.22 9575.19 IS 1.37 0.91 0.70 0.68 0.46 0.40 2.08 1.68 9048.53 2 0.69 0.74 0.63 0.59 0.41 0.40 2.29 2.60 2b 0.59 0.71 0.61 0.60 0.38 0.41 2.29 2.80 3 0.83 0.76 0.67 0.56 0.50 0.42 2.28 2.44 3b 0.78 0.75 0.65 0.57 0.50 0.45 2.14 2.48 4 0.82 0.71 0.63 0.53 0.47 0.40 2.37 2.70 4b 0.76 0.72 0.62 0.54 0.49 0.44 2.26 2.70 lr 1.29 1.08 0.64 0.62 0.42 0.40 2.22 2.44 12394.15 *

20 _ 13» -

submitted to Astronomy and Astrophysics 19 Feb 1988

Notes to TABLE 2 *in this fit the value of r was fixed at 0.5, and the value of Y is that for R > 8 kpe **i'n this fit the value of r was fixed at 2.0, and the value of Y is that for R > 8 kpe fies = 1.0 in all fits shown here. Units: ?, : lO-^atom^sr-1»-1 1% : lO-'em-V-1»-1 Y: 1020 molecules cm"2(K km s"1)"1 £ In I refers to the summed likelihood over the three energy ranges.

21 - 139 -

submitted to Astronomy and Astrophysics 19 Feb 1988

TABLE 2b

Case ?i 72 ?3 94 «5 ?e 1B Y 70-150 MeV 1 0.28 0.14 0.09 0.10 0.20 0.16 0.25 0.92 lb 0.35 0.20 0.11 0.13 0.26 0.21 0.53 0.92 IN 0.85 0.26 0.11 0.13 0.23 0.17 0.29 0.71 IS 0.67 0.23 0.13 0.15 0.23 0.18 0.33 0.88 2 0.42 0.07 0.05 0.09 0.21 0.17 0.21 0.25 2b 0.42 0.07 0.06 0.11 0.25 0.22 0.46 0.32 3 no error anal. 0.28 0.47 3b 0.54 0.58 4 0.26 0.33 4b 0.50 0.45 lr 0.47 0.18 0.09 0.10 0.19 0.16 0.26 0.99 0.16 0.10 0.09 0.10 0.20 0.17 0.25 0.85 150-300 MeV 1 0.33 0.11 0.04 0.06 0.11 0.08 0.13 0.38 lb 0.31 0.13 0.06 0.07 0.12 0.10 0.26 0.55 IN 0.99 0.16 0.06 0.06 0.11 0.08 0.15 0.34 IS 1.14 0.24 0.06 0.08 0.12 0.09 0.17 0.35 2 0.24 0.04 0.02 0.04 0.10 0.08 0.10 0.25 2b 0.23 0.04 0.03 0.05 0.11 0.09 0.21 0.32 3 no error anal. 0.14 0.33 3b 0.27 0.43 4 0.15 0.37 4b 0.27 0.45 lr 0.45 0.11 0.04 0.06 0.11 0.08 0.13 0.39 300-5000 MeV 1 0.25 0.08 0.04 0.05 0.09 0.06 0.11 0.31 lb 0.26 0.10 0.04 0.06 0.09 0.08 0.21 0.37 IN 0.64 0.10 0.04 0.05 0.09 0.06 0.12 0.35 IS 0.41 0.12 0.05 0.06 0.09 0.07 0.14 0.38 2 0.21 0.03 0.02 0.04 0.08 0.06 0.09 0.25 2b 0.20 0.03 0.02 0.04 0.09 0.08 0.16 0.32 3 no error anal. 0.11 0.28 3b 0.21 0.35 4 0.12 0.33 4b 0.22 0.42 lr 0.35 0.12 0.04 0.05 0.08 0.06 0.11 0.33

22 _ mo -

submitted to Astronomy and Astrophysies 19 Feb 1988

TABLE 3 Results of fits: spectra of point sources

Energy range Source CG78+1 Crab Geminga Vela 70-150 MeV 1.15±0.38 4.8±0.45 2.63±0.4 7.72±0.61 150-300 MeV 0.76±0.16 1.27±0.17 1.29±0.16 4.06±0.29 300-5000 MeV 0.55±0.11 0.82±0.12 1.59±0.15 4.59±0.26 Units: 10-8 photons cm ~2 s -1

24 _ It, I _

submitted to Astronomy and Astrophysics 19 Feb 1988

Figure Captions 1. Illustrations of gas data used: (a) HI total column density contour interval: 5 (lO01" -1) x 1020 atoms cm-2 (b) CO integrated temperature contour interval: S (10olB - X) K km s"1. Both maps are shown as convolved with the COS-B point-spread function in the three standard energy ranges used. 2. Log-likelihood ratio as function of Y for the three energy ranges, and for the two higher ranges summed. The model and data correspond to case 1 . 3. Radial distribution of 7-ray emissivity (a) for q(R) shape and Y energy dependent (case 1 ) (b) for q(R) shape energy independent, Y energy dependent (case 3 ) (c) for q(R) shape free, Y energy independent (case 2 ) (d) for r = 0.5 (case lr ) (e) for 1.0 8 kpc, for each energy range (case lr ). 5. Longitude distributions of predicted and observed "y-ray intensity, averaged over |A| < 5.5s. The model corresponds to energy independent shape of emissivity gradients (case 3 ), but appears almost identical in this presentation for the other cases. Vertical bars: T-ray intensities measured by COS-B: ±lc error bars Continuous lines: predicted f-ray intensities as follows - Lower line : inverse Compton emission Middle line; emission from HI Upper line: total model, including HI, Hi, inverse Compton and four point sources All predictions include the fitted isotropic background (celestial + instrumental) 6. Latitude distributions of predicted and observed -y-ray intensity. (a) 300* < I < 60°, omitting 350° < I < 10" (b) 60" < / < 300"

25 Total HI convolved to 70-150 MeV resolution FILE: HBLO-HÏT FITS UNITS 0.10E020 ATOM CM-2 NAXIS1-720 CONTOURS! C + r«(10*«(l 1-1 )>G) - II NAXIS2-96 C-0. F-50.00 G-0.10 NAXIS3-1 NAX3-1 NAXIS4-3 NAX4-1

ihmhtiiliiMliiii|illilllil|i,,,lii,,liiti|.,l,li,lil,ii,l,l|il.,iiliAjlli|iiliiii

i

£-20 iiii|ihi|ini|ii1ii|ii'ii|riii|i'iii|riii|iiii|liiii|iiiiriiMP'Ti'|iiiiiiiiiiiin|iiir|iii 80 60 40 20 0 340 320 300 280

260 240 220 200 ISO 160 140 120 100 GALACTIC LONGITUDE la 6) Total HI convolved to 150-300 MeV FILE: HBLO-HIT FITS UNITS 0.10E020 ATOM CM-2 NAXIS1-720 CONTOURS: C • F«llO»«(l 1-1 )»G) - 1) NAXIS2-9B C-0. F-50.00 G-0.10 NAXIS3-1 NAX3-2 NAXIS4-3 NAX4-1 UlllUlll

j ^^,l^"Tl,l^","",nT,|,,nTr'|,"^fi^Ti^T^n""^""ffT,"^,"fTi,l^"T^i,'|^^^^^^r,^ffin^,i^HHi^^^ u 80 60 40 20 0 340 320 300 280 1111111111111111111111111111 n 111111111111111111 y 111 n 11 M 1111 JJ_jj^ < IJ^VJ I " I |J

1111 f i ifi 111 p i PI fi il 111 ifi 1111 i\ ri 111 m i'| i n 111 260 240 220 200 180 160 140 120 100 GALACTIC LONGITUDE Total HI PSF >300 MeV FILE: HBLO-HIT FITS UNITS 0.10E020 ATOM CM-2 NAXIS1-720 CONTOURS: C • F«l 10**< ( 1-1 )«G) - 1) NAXIS2-98 C-0. F-50.00 G-0.10 NAXIS3-1 NAX3-3 NAXIS4-3 NAX4-1 "'.'Il I_l""_' " '.' L1 •••.'• • "J1 l"|"Ml""l il l'ILLl "•?''""

i

0 —' ~ XX — Q Ç -20 i n 111 ri 111 I'TTJI 111111 I*I 1111 ill 11111111111111 ri 111111111111111 iTpi 11111*? 1111 iTï| ^? 11 fi i rt 80 60 40 20 0 340 320 300 280

rp 220 200 lflû 160 GALACTIC LONGITUDE lo.(ft) Total CO convolved to 70-150 MeV resolution FILE: HBLO-COT FITS UNITS 0.10EOOO K KM S-l NAXIS1-720 CONTOURS: C • F*{ 10««(( 1-1 )»G) - 11 NAXIS2-96 C-0. F-50.00 G-0.10 NAX1S3-1 NAX3-1 NAX1S4-3 NAX4-1 IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

u

J Iu111il 1111111111111111111111111111111111 n | u 111111111111111 H 1111111111111111111111111 60 60 40 20 0 340 320 300 280 I- u< llllllllllllilllli il.ii ill n ill in li i i i li Milm i| i ml in •!• m II m In ii !• i iilinil ii i i -1 20

-20 - 1111111111111111111111111111111111111 u 111111111111111 |lf 11| 11111111111 il 1111111111111111 260 240 220 200 160 160 140 120 100 GALACTIC LONGITUDE Total CO convolved to 150-300 MeV resolution FILE: HBLO-COT FITS UNITS 0.1OE0O0 K KM S-l NAXIS1-720 CONTOURS: C + F«[ 1QMM( u-1 )>G) - 1) NAXIS2-98 C-0. F-50.00 G-0.10 NAXIS3-1 NAX3-2 NAXIS4-3 NAX4-1 II III III II III III ll llllll II I I I I I I I II I I II I I I I llllll I

GALACTIC LONGITUDE Integral CO PSF >300 MeV FILE: HBLO-COT FITS UNITS 0.lOEOOO K KM S-1 NAXIS1-720 CONTOURS-. C • FM 10««<

I III | I I I I | I I I l| I I I I | I I I l| I I I I | I I I I | I I I I | I I I I | I I I l| I I I I | I I I I | I I I I | I I I I | I I I I | I I I I | I I I I | I I I I 80 60 40 20 0 340 320 300 260 i m In 11111 nil II 11 n i I|I 11111 n 111111 h 111111 • i| i II . 1111111111111,111111 |n 111111111111 20

10

0

10

20 111111111111n1111111111111111111111111111111111111111 |Yi 111111111111111111111111111111 M 260 240 220 200 ISO 160 140 120 100 GALACTIC LONGITUDE -log likelihood for Y

70-150 MeV 150- 300 MeV

TJ 6 300-5000 MeV "ioo-Sooo O 70-5000 MeV O

iSo-3* DO O

1 • ' • ' i '

Y/1.0E20 mol. (K kms-ll-1 Fig 2 I m Galactic Emissivity Gradient

Fit lo 0 < I < 350, tbî < 10 Y free, gradient en. dep

[TJ 70-150 MeV «>

<

>2 |" •••••"•••••••••'••"•|" ..t.... tn..i....*....|.tuli.*»!•...I»...li»i.|i»»li |.lr i 0 2 4 E 6 10 12 M 16 IB 20 22 Ga1actocentric radius kpc

FIG 3a 1 (Il Galactic Emlsslvity Gradient

Tit to 0 < I < 350. tbl < 10 Y free, gradient en. Ind

* (TJ 70-150 MeV en 0 150- 300 MeV vjç 300-5000 MeV + ci * m di

* %

2 ItHil^nlnnlimlnnlinilniilvil^itMniilnnlimli^ilimliniliitiIltitlm.liinltmlitÉilmllm 0 2 t 6 B 10 12 H 16 la 20 22 Ga1actocentric radius kpc

FI G 3b I II) Galactic Emissivity Gradient

Fit to 0 < I < 350. Ibf < 10 Y en. Ind.. gradient, free

[I |T) 70-150 MeV + (T) 150- 300 MeV i * Q [] ^ 3Q0-5000 MeV n

\i u n o

is

1 1 .2 I"" »"' miiii«»|mtltMiliiitlÉtiitt«tijitttltMtlÉiiil.niliiit|Ét.ilintl..nltn.lft^|*iHfi...l..t

FIG 3c Galactic Emissivity Gradient

Flt to 0 < ! < 350. Ibt < 10 Y free, gradient free.r - 0.5

70-150 MeV H 150- 300 MeV £ 300-5000 MeV [] []

2 |nnii»«tiiiiiini»imi|i-ufirrtlii,ll""'""|""'""'""'""'""|""'""'""l-ii'''iii|i"iii"»hll'1'"|l 0 2 * 6 a 10 12 H ]6 I] 20 22 21 GaIacLocentric radius kpc

FIG 3d ProflrAn cosbl gradpl57

I m Galactic Emissivity Gradient

Fit to 0 < I < 350. "b" < 10 Y free, gradient free.r - 1-3

r - 1.0

• r - 1.4 «I * Id X r - 2.0 o r - 3.0

[ u I

' I ' I....l....l....l....l....[ llllll.l.ll.ll.l....l»iIll....l.l,. 0 2 4 6 8 10 12 14 16 10 20 22 24

Galactocentric radius kpc ? => Log-likelihood for Yl0-8kpc)/Yl >8 kpc)

70-150 MeV 150- 300 MeV 300-5000 MeV 150-5000 MeV O O

oo o

l.S 2.0

yto-a kpc )/y(>a kpc ) Fig 4 FREE FADMAP FADM0002 GAMMA- 1.65

160 140 120 100 80 40 20 0 340 320 300 280 260 240 220 200 160 GALACTIC LONGITUDE AT y FREE FADMAP FADM0002 GAMMA--1.65

160 140 120 100 60 40 20 0 340 320 300 260 260 240 220 200 160 GALACTIC LONGITUDE S^ y FREE FADMAP FADM0002 GAMMA--1.85

LONGITUDE RANGE 100.0 TO 100.0 LATITUOC «ANGE -9.9 TO 5.S DBKV RANGE 100 TO 9000 HEV EHISSIVITYi 0.03 0.T6 0.67 0.56 0.90 0.42 E-26 SR-1 H2-T0-C0 RATIO - 2.44 E20 HOLS KH-1 S CM 35 OM-AXIS BACItGROUND-2.26 E-S CM-2 SRI S-l les rAcnw-i.oo u

»H »

U

>- < a:

<

160 140 120 100 80 60 40 20 0 340 320 300 280 260 240 220 200 180 GALACTIC LONGITUDE INTENSITY DEF: FLUXDEFO FADMAP FADM0002 GAMMA--1.65

i

-5 0 5 GALACTIC LATITUDE 6*0 INTENSITY DEF. FLUXDEFO FADMAP FADM00O2 GAMMA--1.85

-5 0" 5 GALACTIC LATITUDE ( ...(-A INTENSITY DEF: FLUXDEFO FADMAP FADM0002 GAMMA--1.85 T- -t" UMCITUOE «MCE 300.• TO EO.O ÏIWJ-3S&, lo^io' LATITUDE MUGE -20.0 TO 20.0 ENEMiY UMCE 300 TO 5000 KEV EHISSIVITYi 0.03 0.76 0.67 0.S6 O.SO 0.42 E-26 SS-1 S-l K2-TO-C0 IATIO • 2.44 C20 HOLS KM-1 S OM-AXIS IACKGaoUND-2.26 C-S CN-2 3R-I S-l I X ICS FACm-1.00 u

U o

in

i ' i i -T—r fc i ., i—i r • ' * ' J—'—L- —I I L_ -20 -15 10 -5 0 5 10 15 20 GALACTIC LATITUDE b«(ùï) INTENSITY DEF= FLUXDEFO y FREE FADMAP FADM0002 GAMMA--1.85 -i—I—i- • i I -I 1 1 1 1—J 1 r- LONGITUDE tAHGE 60.0 TO 300.0 LATITUie «AWE -20.0 ID 20.0 OS EHEM» «MICE 70 TO ISO HE» EHISSIVITT, 1.39 1.26 I.II O.M 0.03 0.70 E-26 SR-I S-1 in H2-T0-C0 1AT10 • 3.30 E20 HOLS KM-1 S OH-AXIS OACKOHOUNO-O.lS C-S CH-2 31-1 S-1 2: ICS FACTOB-l.tl u

«} IS b]

>- < OS <. <

-15 -10 -5 0 5 10 15 20 GALACTIC LATITUDE f. .rV INTENSITY DEF: FLUXDEFO y FREE FADMAP FADM0002 GAMMA--1.85

GALACTIC LATITUDE Uc«) INTENSITY DEF: FLUXDEFO y FREE FADMAP FADM0002 GAMMA--1.85 -i— T^ -t- -I- LONGITUDE «ANSE 60.0 TD 300.0 LATITUOC RANGE -20.0 TO 20.0 DCKV KANGE 300 TO 5000 HEV in EHISMVITYi 0.S3 O.Tfi t.ET 0.56 0.50 0.42 E-26 5»-l S- H2-T0-C0 IAT10 - 2.44 E20 HOLS KH-1 S ON-AXIS 1ACXGIOUND-2.25 E-5 CM-2 SR-1 S-l 1CS rACTDR-1.00 CJ 16

W >«

>- z. zi-

< OS

-20 -15 -10 -5 0 5 10 GALACTIC LATITUDE . im -

CO OBSERVATIONS IN CEPHEUS: II- GAMMA RAÏS / GAS CORRELATIONS AND A CANDIDATE FOR A NEW COS-B SOURCE

I.A. Grenier and F. Lebrun Service d'Astrophysique, C.E.N, saclay

ABSTRACT The recent CO survey of the Cepheus area (100°S1S141°, 8°Sb£22°) has allowed a detailed study of the correlation between the diffuse gamma radiation as observed by COS-B and the atomic plus molecular gas content of this nearby region. The analysis yields an estimate of the gamma-ray emissivity spectrum of the atonic gas between 70 MeV and 5 GeV, and a value of (2.3 ± 1.2) 10*° sol. 1 1 1 cm" K" km" s of the local N(H2)/WC0 ratio in these quiet and cold nolecular clouds. The correlation study also reveals a significant (3.9c) point-like excess which stands above the diffuse gamma radiation at 1= 111°. b= 19.8° with an error box 1.8° in radius. This excess represents a good candidate for a new bright and extremely soft gamma-ray source, but its nature remains mysterious. - Ifcf -

l. mntoDocnoN

As revealed by the telescopes aboard the SAS-2 and COS-B satellites, the

diffuse gamma-ray emission has proven to originate in interactions of cosmic

rays and interstellar matter and thus to be a valuable total gas tracer. It is

indeed valuable because it provides large scale maps of atomic plus molecular

gas and because it acts as a deep probe since interstellar matter, however dense

it may be, is highly transparent to the gamma rays. These properties, first

tested in the local interstellar medium (Lebrun et al., 1982, Strong et al.,

1982, Lebrun and Paul, 1983). have been extended to the entire galactic disc

(Lebrun et al,, 1983, Bloemen et al., 1986). Recently, Strong et al. (1988) have been able to correctly reproduce the bulk of the diffuse gamma-ray emission observed by COS-3 from the atomic and molecular content of the disc and with limited variations of the cosmic ray density through the Galaxy. An important

by-product of such analyses is the necessary calibration of the ratio of H2 column density to integrated CO line intensity (often noted as N(H»)/WC0). An accurate estimation is essential to evaluate the masses and densities of molecular clouds.

Of particular interest is the study of gamma rays / gas correlations on a smaller scale at medium latitudes. The complete CO and HI surveys of extensive nearby complexes provide a good mean to explore the local gamma-ray emissivity and cosmic ray density, and the local N(Ha)/WCO ratio. Moreover, because of their apparent narrow distribution along the galactic plane, the contamination of the diffuse emission by galactic gamma-ray sources is greatly reduced.

Bloemen et al. (1984) first studied in this way the Orion-Monoceros region. The good agreement they obtained between the observed gamma-ray flux and the fluxes predicted from the gas implied that no faint sources were significantly contributing to the diffuse emission. Then the emissivities derived above 100 and 300 HeV suggested no cosmic ray variations in this direction. On the contrary, though within 1 kpc the cosmic ray density is expected to be about - Hi -

uniform, the study by Lebrun and Paul (1985) and Lebrun (1986) of the Oph-Sag

region has established the existence of a gamma-ray enhancement with respect to

the gas prediction. The extensive excess has been attributed to an emissivity

enhancement rather than to a source contribution. Loop I. which lies in this

direction, might be responsible for this effect.

la this context, the study of the Cepheus region appears as a necessary

step in the understanding of the high energy behaviour of the solar environment.

The region has been observed by COS-B and because Strong et al. (1982) bad

noticed there a large gamma-ray flux excess over that expected from HI alone, a

broad survey of the region in the "CO line has been undertaken with the

Columbia Sky Survey telescope. The observations at 2.6 mm cover 490 square

degrees between 100° and 140° in longitude, 8° and 22° in latitude, with an

angular resolution similar to that of the HI and gamma-ray maps. The details of

the observations and of the molecular clouds found in the survey have been

presented in the first article of this series (paper I). We now concentrate on

the spatial comparison of the gamma ray and gas distributions by giving the

results of correlation studies. They have been performed using a likelihood analysis in three energy ranges (70-150, 150-300, 300-5000 MeV) and yield an estimate of the local N(Ha)/WC0 ratio and of the gamma-ray emissivity spectrum.

Finally, as the comparison has revealed the existence of a strong point-like gamma-ray excess which stands out above the diffuse emission at low energies,

its significance and identity are discussed.

2. DATA AND AHALÏSIS

2.1 CO data

The recent CO observations at 115 GHz of the Cepheus region have been carried out during the winter 1984-1985 with the 1.2a millimeter-wave telescope of the Columbia University in New York City. The "superbean" technique, reducing the nominal 8.7' angular resolution of the antenna to a 0,5°x 0,5° square beam •- I6ï _ along an (l,b) grid, and the use of an extremely sensitive SIS receiver allowed a complete and fast coverage of a broad area in the Cepheus constellation.

99.5°-141.5° in longitude and 8°-22.5° in latitude are the outer boundaries of the survey. Its detailed borders are drawn on figure ?. But. because of the wide spatial resolution of the COS-B telescope, CO data below b=8° and outside the covered longitude interval were needed. So the present survey has been merged to the similar observations of the 1st and 2nd galactic quadrants which are part of the Columbia composite CO survey of the Milky Way (Dame et al., 1987).

Unfortunately, towards high latitudes, no CO data are available and the medium outside the limit of the different surveys has been assumed to be void of CO.

This hypothesis is supported by the reddening measurements which do not suggest any dense CO concentration above Cepheus but only a diffuse absorption (Heiles,

1976, Lucke, 1978).

The Columbia telescope was equipped with a 256 channel spectrometer providing a velocity resolution of 0.65 km s"1 from -82.5 to +82.5 Ian s~r, As the interstellar matter in the Cepheus direction obviously breaks up in two velocity components, a nea-J.j complex at =300 pc and clouds belonging to the

Local Arm («800 pc) (see paye: I), the spectr? have been integrated over the whole velocity range to produce a total intensity (WCO) map of the CO emission.

More details about the observational procedure and the data calibration and sensitivity are given in paper I.

2.2 HI data

In the Cepheus region and outside, total N(HI) column density maps with an angular resolution of 0.5° have been constructed from the observations at 21 cm of Weaver and Williams (1973) below b=10° and of Heiles and Habing (1974) above

10°. These data are part of the Berkeley HI survey. Near the galactic plane, a uniform spin temperature of 125 K was adopted while above 10° the 21 cm emission was considered as optically thin. . lit -

Both the CO and HI maps were available with an angular resolution identical

to the gamma-ray map but the gas (l.b) grid was shifted by 0.25° with respect to

the gamma-ray grid. So. both maps have been re-arranged into the gamma-ray grid

after their convolution by the COS-B spatial resolution.

2.3 Gamma-ray data

Maps of the gamma-ray intensities in different energy ranges have been

derived from the Final COS-B Database (Hayer-Hasselwander, 1985). The Cepheus

region as delimited by the CO survey stayed in COS-B field-of-view during six

observing periods (11, 16. 28, 35. 56 and 57). Each lasted typically one month

but only one, period 35 centered on 111.6°, 27.2°, largely contributed to the

recorded fluxes. For each period, all the photons detected within 20° of the

telescope axis have been selected. The database also contains information on the

instrument response and behaviour. Its sensitive area, energy resolution and

point-spread-function are provided for different inclinations and energy

ranges. The point-spread-function given by the database is an in-flight

parameter calibrated on the Vela pulsar which has an average spectral index of -

1.84 ± 0.03 (Grenier et al.. 1988). Thus for sources with a quite different

index, or simply for energy ranges not included in the database, effective point-spread-functions have been produced from the pre-flight calibration data of the telescope (Hermsen. 1980).

The evolution of the instrumental sensitivity during its lifetime has been estimated from the study of overlapping observations along the galactic plane

(Strong et al.. 1987). As corrections are accurate to about 10%. effective exposure maps can be constructed from the database, Input spectra of E"1'*3 for

the diffuse emission and of E"ao for the isotropic background have been assumed, The details of the exposure calculations and photon selections are given by Mayer-Hasselwander (1985), For the three classical energy ranges (70-

150, 150-300, 300-5000 MeV) the gamma-ray intensity maps correspond to those . 169 -

presented with the database (figure la, lb and lc).

2.4 Analyses

To test the correlation between the diffuse gamma rays and the total gas distribution in different energy ranges we have attempted to reproduce observed gamma-ray maps by adjusting a linear combination of the atomic (N(HI)) and molecular (WCO) gas maps plus an isotropic emission. The latter takes into account the instrumental background, any extragalactic background radiation and the inverse-Compton emission from the interstellar photon field. No attenpt was made to set the inverse-Compton component as a free parameter since the studies by Strong et al. (1985) and Bloemen (198S) suggest that no marked spatial structure of this faint emission is expected in the Cepheus region, in longitude as well as in latitude. It should blend into the isotropic background.

Therefore, in one direction in the sky, predicted counts for the diffuse gamma radiation are calculated as:

ND = [JT [ A 8(HI) + B WCO" + C ] (model 1) where tilde indicates the convolution of the gas maps with the energy dependent point-spread-function of COS-B and SIT refers to the effective exposure in a particular energy range. C directly represents the background intensity. A gives

the gamma-ray emissivity per H atom in its atomic form by A = qBi/4n whereas B leads to the emissivity of H atoms in their molecular form by B =

2 (qH2/4n) (N(Ha)/WCO). The meaning of A and B of course supposes a negligible contribution from discrete sources.

As a compromise between the need for detailed maps and significant dynamical ranges on one hand, and the requirement to maintain good statistics on the other hand, the analysis has been performed in 2° x 2° bins from 100.5s to

140.5° in longitude and from 9° to 21° in latitude (the values refer to the

center of the extreme bins). In such bins, the predicted counts NP and observed counts No follow a Poisson distribution. So the values of A, B and C have been . 170 .

adjusted to maximize the likelihood:

T-r N„<» ) e-CM„) L= J | 0 bins N= ! The formal la errors on the best values of A. B and C are determined from the

shape of the likelihood ratios 1*(A)= max,.o(L)/L*, 1„(B)= raax*,c(L)/L* and

1*(C)= max,,C(L)/L*, where L* is the maximum likelihood value reached for the

best set of parameters. Since the quantities -21n(l«) all have a X2 distribution

with one degree of freedom, the la confidence levels are obtained for a decrease

of 1 unit in -21n(li).

As a first step, the analysis has been performed for the sole diffuse

emission. But the discovery of a strong gamma-ray flux excess over that expected

from expression (1) has led to Che addition of a point-source term in the

prediction:

N„ = SIT [ A S(HI) + B WCÛ" + C + F ISOURCE 1 (model 2)

where F represents the luminosity of the tested source and ISOURCE is the

spatial distribution of a point-source of unit strength as given by the COS-B

point-spread-function for the particular energy range. The position and obvious

softness of the excess suggested the choice of 8° as an average source

inclination on COS-B and E"3 as an input spectrum for the calculations of point-

spread-functions and exposures. This choice has been verified a posteriori.

The likelihood analysis has also been used to localize the point-source.

Maps of predicted counts in 2° x 2° bins have been constructed for regularly

spaced positions of the source (with 0.5° by 0.5° steps in 1 and b). For each position, the likelihood value L"(l,b) has been maximized over the remaining parameters A, B. C and F. In these conditions, the likeliest position of the

source in the sky is found where L*(l.b) reaches its very maximum L* and the la confidence region is derived from the two-dimensional likelihood ratio 1,(1,b) = maxi.»L*(l.b)/L*, where L* this time has been maximized over all possible parameters A. B. C. F, 1 and b. The la boundary corresponds to a decrease in _ 141 -

-21n(L„(l.b)) of 2.4 since L, has now a X' distribution with two degrees of

freedom.

3. RESULTS

3•1 Study of the diffuse emission

The correlation between the gamma-ray and the gas maps has first been studied for the sole diffuse emission according to expression (1). that is any contribution from discrete sources was explicitely neglected. However, if some are present in the area, they may be covered to some extent via the B.WCO term which is the more contrasted part of the emission. The A.N(HI) component is indeed very smoothly distributed in the Cepheus region. The analysis has been performed for the three classical COS-B energy ranges (70-150. 150-300, 300-5000

HeV) so that the derived emissivities can be compared to the values found in other parts of the Calaxy. The likeliest values of A, B and C in each range and their respective errors are listed in table la. Considering that in the Cepheus direction the bulk of the interstellar matter lies within 1 kpc (1.5 kpc at most), we may suppose a uniform distribution of the cosmic rays. He may further assume that the same cosmic ray flux pervades the diffuse atomic clouds and the

denser molecular cores, that is qKi equals qMa. Bearing in mind these important restrictions and following our first assumption on the sources, the best parameters A and B yield estimates of the N(H»)/WC0 ratio for each energy range,

N(H2)/WC0 = B/2A, which are also presented in table la.

At medium and high energies, the derived A and B emissivities nicely agree with the values found in the Orion-Honoceros complex, at medium latitudes in general or near the galactic plane at the solar circle. A thorough comparison

will be discussed in the next ciiapter. Moreover, the two estimates of N(Ha)/WC0 above 150 HeV are fairly consistent. The derived background intensities do agree with the values proposed in the COS-B database which come from the study of the entire set of 65 observations in the whole Galaxy. As an internal check of the statistical reliability of the analysis, the

same likelihood has been computed for l°xl° maps instead of Che 2°x2° binning

that was first considered. Because of the wide angular resolution of COS-B at low energy, the l°xl° analysis has been restricted to energies higher than 150

MeV. The best parameters obtained at medium energy (A= 0.7±0.2, B=3.±2.. C=

24±5) and at high energy (A= 0.57±0.19, B= 3.5±1.5, C= 22.5±4.5) confirm the results given in table la. But, despite the increase by a factor of 4 in the number of data points, the quality of the likelihood fit has not improved because of the reduced statistics in smaller bins. The analysis in 2°x2° bins has therefore been systematically preferred for the next studies.

At low energy, however, very surprising results have been obtained. When one would expect enissivities slightly higher than above 150 MeV to follow a spectral index of -1.85, A drops dramatically while B rather increases. The

combination of both tendencies leads to an unreasonably high value of N(H2)/WC0 compared to the estimates at higher energies. The background intensity as well do not correspond to the COS-B database value. Although the errors on all parameters are large, the results at low energy look doubtful. A similar behaviour is observed at even lower energies. The maximum likelihood is reached between 50 and 150 MeV for A= 0.6, B= 18 ..nd C= 118, and for the full 50-5000

MeV interval, where the high energy data is supposed to highly constrain the fit, values of 2.1, 26 and 153 are derived for A, B and C respectively. All these tests suggest that a structured excess is put into the B.WCO tern and the consequent rise in B induces a drop in A.

To seek the origin of this striking behaviour, Che low energy gamma-ray data have been confronced Co a "normal" gas prediction. What values can be considered as "normal" gas emissivities in the solar vicinity? The scudy of the entire galactic plane (Strong et al., 1988) yields a 70-150 MeV gamma-ray emissivity of 1.02 ± 0.10 at Che solar circle in good agreement with the value

1.1 ± 0.14 measured at medium latitudes (Strong et al., 1985). So, an emissivity . 113 -

A of 1.05 10"2' at"1 s"1 sr"1 has been adopted as "normal". Then, the. "normal" B

parameter has been determined from the knowledge of A and using the N(H2)/WC0

ratio measured at high energy in the galactic plane (2.3 ± 0.3 1020, Strong et

al., 1988). A "normal" B values thus equals 5 10"' cm"2 sr"1 K"1 km"1. As a

final choice, the background intensity provided by the COS-B database was

adopted; C= 72 10"' cm"2 s"1 sr"1. The comparison of the observed data and the

flux expected from those average eaissivities is illustrated on figure 2a and

2b, The scatter plot has been produced for 4°x4° bins while the map of the

differences between the two quantities has been constructed for 2°x2° bins. The

latter differences are expressed in a values and all discrepancies within ±0.5o

have been ignored. Such a comparison obviously reveals a deviant region around

1=110°, b=20°, which stands on the scatter plot at 3o above the prediction

without any optimization in the binning size. The excess appears as a bright

spot on figure la and its shape, because of the wide point-spread-function in

this range, strongly suggests the presence of a point-like excess. The absence

of any coincident enhancement above 150 MeV on figures lb and lc reveals an

unusual softness of the excess.

As the presence of such a localized and significant deviation certainly

disturbs the likelihood estimation of the gas emissivities. an extra point-

source component has been added to the model to test (expression 2) and the whole analysis has been performed again. As already explained, the position and

luminosity and the source and the diffuse emission parameters have been fitted

altogether. Table lb gives the results on the gas and background intensities.

The fits for the three energy ranges are illustrated on figure Î where the

convolved HI and CO data, the observed gamma-ray flux and the differences between the observations and predictions are displayed in 2sx2° maps,

At low energy, the model including a source is a significant improvement on

the previous one. At the best position and source flux, the amelioration

corresponds to a 3,75a effect (the determination of this number is presented _ im -

with the results on the source). On the contrary, at medium energy, no

improvement could be detected from the addition of a faint source in this region

and at high energies, a very small increase in the likelihood was observed (0.6o

effect). This is why the inclusion of a source had quite a small impact on the

A, B and C estimates above 300 HeV and no impact at all at 130-300 HeV. At low energy, however, the parameters describing the diffuse emission have changed.

The background intensity C and the emissivity A 1 inked to the atomic gas are now in complete agreement with the values we have just selected as "normal". Only B remains surprising. From a previous unusually large value, it has now dropped to a very low one. A scatter plot of the observed versus predicted flux in 2°x2° bins indeed shows that the majority of the data points lie below the "normal" prediction. The effect is marginal but still visible on the 40x4° scatter plot

(figure 2a). This means that the low B value is not induced by the gamma-ray deficiency in a particular molecular region, but that the two extensive CO concentrations present in Cepheus are barely seen on the low energy gamma-ray

map. Because of B, the N(H2)/WC0 estimate is also low. As the errors on both are large, the discrepancy between the 70-150 HeV parameters and the estimates above

150 HeV amounts to only lo and is not really significant. But the fact that we do not see the molecular clouds on the 70-150 MeV COS-B maps remains puzzling.

3,2 The point-source component

The addition of a source term in the model leads to a significant improvement in the quality of the fit at 70-150 MeV. To quantitatively compare the two models (with and without a source), we consider the increase in likelihood between the best values obtained in each case (L* and L* respectively, see section 2.4), that is when all parameters are optimized. The increase is measured by the log-likelihood ratio. -21n(L*/L"), which equals

13.0. This quantity being distributed as Xi2, the probability of obtaining by pure chance -21n(L*/L*) higher than 13.0 is p= 2 10"\ Therefore, the - i?r -

improvement in the fit corresponds to a 3.75o effect. (Details on the likelihood

ratio and how it permits a quantitative comparison between two hypotheses, namely here "does the model with a source better accounts for the data or not?", are explained by Eadie et al., 1971). On the contrary, no excess was detected by the likelihood analysis at 150-300 HeV around (110°. 20°). Above 300 MeV, a very small increase in the likelihood was observed when placing a faint source at the position indicated by the low energy investigations. The value of 0.35 obtained for -21n(L"/L*) leads to a probability p= 0.55, that is a 0.6ct significance of the amelioration of the fit. The effect, however, was too weak to allow a precise localization of the faint excess. The gamma-ray data in the three energy ranges being independent, the total significance of the point-like excess is

0.99989 or 3.9o.

We used the likelihood analysis tu question "is there a significant point­ like excess in this area?". Hence, the 3.9o value is the actual answer, But to be conservative, we may ask how many such excesses can be found within Che whole survey. No true estimation exists. Nevertheless, we may make use of the spatial resolution definition to give an order of magnitude. In order to distinguish the», two point-sources must be separated by an angular distance larger than the full width of the point-spread-function, So, 10 to 15 sources could be isolated in the whole 70-150 MeV map depending on the choice of the point-spread-function width (FWHM= 7° or width at the SOX area limit= 9°). The significance of the present excess when asking "is there an excess somewhere in Che 70-150 MeV map?" is then reduced to 3c.

Other tests have been performed to measure the impact of the rather low B values obtained for the best low energy fit. The diffusion emission parameters in the model have been forced to their "normal" values, as chosen in section

3,1, In fact, while A and C were set to 1.1 and 72, two values have been

selected for B (4.7 and 5.3) because they correspond to N(H2)/WC0 ratios of 2,1 and 2.5 which enclose the different values measured above 150 MeV in Cepheus and . fît -

near the plane. In the two cases the likelihood analysis gives a 3,4a and 3.7a

significance for the excess at low energy. Finally, the analysis has been

applied to the 50-150 MeV interval for which the discrimination between the point-like excess and the structured emission from the nearby CO cloud is poorer because of the wider angular resolution. This effect is illustrated by a higher

B value for the best fit (A= 1.3-1.4. B= 7-9, C= 115-120. N(H2)/NC0= 3). The significance of the excess is still found at 2,9a. In conclusion, the different analyses indicate that a statistically significant point-like excess lies around

(110°, 20°) above the diffuse emission expected from the interstellar matter.

As explained in section 2.4, the likelihood analysis has been used at low energy to localize the source. The result is presented on figure 4. The likeliest position is 1= 111° and b= 19.8° with a radius of the lo confidence region of 1.8°. Such a small error box with respect to the point-spread-function half width of 3.5° and its regular circular shape reflect the strength and point-like quality of the excess. Its sharp profile could not be attributed to the CO component.

The map of the likelihood values obtained when testing different positions in this area is similar to the output of the general method applied to search for gamma-ray point-sources in the C0S-B data (Pollock et al., 1985). Therefore, the increase in likelihood measured here with respect to the null hypothesis of only gaseous emission can be compared to the statistical threshold which was considered by Pollock et al. as a reasonably weighty indication of the presence of a source. Because of the three degrees of freedom embodied in the strength of an excess and its position in 1 and b. they adopted a value of -21n(L"/L") of 12 as a detection threshold. As a value of 13.0 has been measured in the present case, the excess at 1= 111° and b= 19.8° can be considered as a new C0S-B source.

The source fluxes are presented in table 2. The strong excess at 70-150 Mev indeed has a large flux since (2.3 ± 0,7) 10"* cm"2 s"1 corresponds to the 5th _ n? _

brightest source in the 70-150 MeV sky, after Vela, Crab, Geminga and 3C273, The

analysis at 150-300 MeV only leads to a lo upper limit and the extreme softness

of the excess, pointed out on figure 1, is illustrated by the weak 300-5000 MeV

flux. The differential photon spectrum thus obtained is fitted by the power law;

S(E) = 0.277 E-3-3*0* cm"2 s~x MeV-1

The absence of detection at 150-300 MeV is not significant when compared to the

spectrum expectation (1.8cr). The softness of the emission, however, is striking.

The source shows the steepest spectrum ever observed by C0S-B. This steepness,

besides, reinforces the idea that the excess does not originate in the random

clustering of photons born in the interstellar medium or in the experiment because the diffuse emission and the instrumental noise have significantly different spectral indices (-1.85 and -2.0).

4. DISCUSSION

4.1 The diffuse galactic emission

The eaissivity per H atom in the atomic form (A) measured in Cepheus in the three energy ranges (table lb) are quite consistent with the values derived from the complete set of gamma-ray observations available at medium latitudes

(1.10±0.14, 0.76±0.09, 0.68±0.09, Strong et al., 1985). The latter, however, should be considered with care since the molecular gas contribution has been evaluated from galaxy counts not yet corrected for the bias introduced by field stars (L86) and because the deviant Oph-Sag region mentioned in the introduction was included in this fit. Closer to the galactic plane, similar emissivities have been derived in the outer Galaxy, in the 10-12.5 kpc galactocentric ring

(1.39±0.25, 0.56±0.14, 0.53±0.14, Bloemen et al., 1984b) although the molecular gas content has been neglected in this fit, More important is the full agreement between the emissivities obtained in Cepheus and those derived recently at the solar circle from the correlation of the gamma rays and the atomic plus molecular gas along the entire galactic plane (|b|<10°, 1.02±0.10, 0.65±0.06, - m -

0.62±0.06, Strong et al., 1988). And particularly interesting is the consistency

with the values measured from the SAS-2 observations at medium latitudes,

excluding the Oph-Sag region (Lebrun and Paul, 1985). Finally, the detailed

study of the Orion-Monoceros complex above 300 HeV has also given a similar value for A (0.52±0.13, Bloemen et al., 1984). Such a consistency between

emissivities measured in different areas, both in latitude and longitude, with different 21 cm and CO surveys, different COS-B and SAS-2 observations and various fitting programs shows that we can most probably trust the gamma-ray eraissivity spectrum of the atomic hydrogen lying within 1 kpc. The knowledge of this spectrum is essential to the determination of the cosmic ray spectrum outside the solar system, Moreover, the agreement between values measured at medium latitudes and averaged along the solar circle suggests that the cosmic ray distribution in the solar vicinity'is not peculiar.

At medium and high energies, the B parameter, which is related to the molecular gas eaissivity, is also found in very good agreement with the values evaluated locally in the plane (3.26±0.30, 3.03±0.29, Strong et al., 1988), the latter being highly constrained by the numerous nearby clouds in the 1st and 4th quadrants. Above 300 MeV, the present B estimate is consistent with that derived in the Orion-Monoceros clouds (3.0±1.2, Bloemen et al., 1984). Hence, the measurements of B above 150 MeV in the solar neighbourhood seem reliable too.

This is an important point since the knowledge of B ensures a calibration of the

N(Ha)/WCO ratio on a large scale. The present estimates of the ratio in the two higher energy ranges are fairly consistent. So, we will adopt (2.3 ± 1.2) 102°

s 1 -1 molec. cm" K" km s as the best estimate of N(H2)/WC0 measured in the Cepheus region. This value fully agrees with that determined within the whole Galaxy in

ao the same energy range (>150 MeV: N(Ha)/WC0 = 2.3 ± 0.3 10 , Strong et al.,

1988) and is comparable to the ratio measured in Orion-Monoceros above 300 MeV

(2.6±1.2. Bloemen et al., 1984). This is why a. ratio of 2.3 10" has been used in paper I to evaluate the masses of the molecular clouds found in the present - lia .

CO survey.

The calibration of the N(H2)/WCO ratio in the Cepheus region is

particularly interesting because these nearby clouds are generally quiet and

cold. All the observed brightness temperatures fall below' A K and the clouds

possess no apparent OB associations, no other young massive stars, nor any H1I

regions which might locally heat the gas, The observed bubble (see paper I)

could have an impact on the gamma-ray emission above ISO MeV, but no coincident

enhancement is seen on the COS-B maps. A recent theoretical work by Kutner and

Leung (1985) suggests that the N(H*)/WCO ratio should be reliable, i.e. WCO is a valuable tracer, below 3 mag of absorption. In this case, the Cepheus clouds are

indeed good candidates to calibrate the ratio since the absorption in their densest parts does not reach 2 mag (A„ < 1.4 - 1.6 mag, Lebrun 1986). On the other hand. Leung et al. (1982) have proposed on theoretical grounds that the

1 3 N(H2)/WCO ratio should be temperature dependent (T*" ' ), but the agreement between the ratios aeasured in Cepheus and Orion do not support this idea.

Cepheus is significantly colder than Orion (brightness temperatures from 0.5 to

4 K against 5 to 30.8 K in Orion A and B. Maddalena, 1986), but the ratio derived in Orion is comparable, even slightly higher, than in Cepheus.

At low energy, however, the systematic deficiency of the structured gamma radiation related to the observed molecular clouds is puzzling though barely

significant. A natural explanation would be that WCO does not trace the true H2 mass in these clouds, but the arguments just presented on their quietness and coldness does not favour this possibility. Moreover, unusually excited CO lines would not lead to a normal N(H»)/WCO ration above 150 MeV either. Another idea is that a large part of the CO contribution from the cloud at low longitudes has been unduely attributed to the source. But the statistical tests and its spectral slope do not support this. Anyhow, the weak deficiency is also observed in the high longitudes cloud where no source can be responsible for it. Another cause could be instrumental, An unexpectedly wide point-spread-function at low . IÎO -

energy would smooth out the contrasted radiation born in the molecular clouds.

But such a wide point-spread-function is not supported by other studies (Vela

and Crab pulsars for instance), nor by the profile of the present source excess.

And the contrast in the convolved CO map presented in figure 3a, demonstrates

that the 0.5° uncertainty on the point-spread-function width cannot explain the

smoothness of the observed gamma radiation. Because the region has merely been

observed once by COS-B, low statistics could be another explanation.

Nevertheless, the low energy range provides larger counting rates than the two

higher energy intervals where the analysis has led to coherent results. Strong

et al. (1988) in their study of the whole Galaxy faced similar problems at low

energy, but in their case B was found too high. Because of the complexity of

their analysis, the origin of the effect could not be isolated but the parallel with our first study of the diffuse emission (without a source) is interesting.

Is Cepheus an example of what disturbs the Galaxy analysis? Is the weak low energy deficiency in the Cepheus clouds a real effect implying cosmic ray variations inside the molecular clouds? Clearly, the detailed study of all the nearby complexes which have been fully surveyed, such as Taurus, Orion,

Cassiopeia, is required before we can understand the gamma rays/ gas correlations below 150 MeV and interpret them in terms of cosmic ray propagation inside the dense clouds.

A.2 A candidate for a new COS-B source

The statistical tests presented in the last section lead to a significance of the excess at 1= 1110 and b= 19.8° of 3.9o above the diffuse radiation predicted from the best fit obtained in the Cepheus region. But the 70-150 MeV emissivity of the molecular gas being rather low in this area, the excess has been confronted to the diffuse emission expected from known average emissivities of the local interstellar matter (< 1 kpc). The significance is still around

3.5o. The snallness of the position error box compared to the spatial resolution _ HI .

and the unusual softness of the excess, which is not compatible with the spectra

of the diffuse and background radiations, are other arguments in favor of the

existence of a discrete source in this direction. It may seem surprising that

this candidate never belonged to any COS-B source catalogue. But isolating

gamma-ray sources below 150 MeV had never been attempted because of the poor

instrumental resolution and because only very bright sources could in principle

be detected • (the low energy point-spread-function dramatically reduces their

significance) and strong emitters are supposed to be visible at higher energies

too. The recent. CO data were also essential to distinguish this excess for any

gas clump can mimic a point-source at low energy. The present investigations

therefore reveal a late candidate for a new gamma-ray source. It is clear, however, that the source is at the limit of detectability and its existence must be confirmed by new observations. In a near future, the Camma-I and Sigma telescopes could establish the existence of the source while measuring its precise position and studying its spectrum at lower gamma-ray energies.

While waiting for these crucial observations, we may discuss the possible identity of the candidate. An error box 1.8° in radius should contain many plausible counterparts at other wavelengths, even at 20° in latitude.

Surprisingly, very few have been found. Exploring the stars in this area, no peculiar objects have been noticed except the variable star VW Cepheus, at 1=

109.5°. b= 20°, which appears as a weak X ray source (3U2041+75 or 1H2041+756,

HEA0-A1, Wood et al., 1984). The star lies 1.5° away from the likeliest source position, on the edge of the error box. No other X ray source from the Uhuru.

HEA0-A1 and A2 surveys have been found in the error box (Forman et al., 1978.

Wood et al., 1984, Marshall et al., 1979), nor any supernova remnant (Ilovaisky and Lequeux, 1972), pulsar (Manchester and Taylor, 1987) or HII region

(Marsalkova, 1974) which are often considered as likely gamma-ray sources. At

408 MHz, however, a bright radio source is almost coincident with the gamma-ray position. It corresponds to the radiogalaxy 3C427.1 (also called 4C76.13. . Ut -

QSO2104+763, EQ2104+763) and its position, 1= 111.04°, b= 19.28° (see figure 4),

fully agrees with the gamma-ray one. The radiogalaxy possesses two bright lobes

(Strom and Kronberg, 1976) and was identified in the visible light as a 5"

object by Smith et al. (1976). Without optical emission lines, its redshift was

first poorly known. Recently, deep spectra allowed to situate the object at z=

0.572 (Spinrad et al., 1985). The IE magnitude of the radiogalaxy at 2.2 u is normal for such a redshift whereas the galaxy looks unusually red in the visible. Unfortunately, 3C427.1 does not have an X ray counterpart. A short observation by the Einstein satellite only led to an upper limit of 2.4 10"13 erg cm-2 s"1 (F. Seward, private communication). Another quasar has been detected in gamma rays by COS-B. It was one of the nearest known, 3C273, at z=

0.158. Of particular interest is the comparable softness of the gamma-ray emission from 3C273 and the present source. 3C273 has indeed the steepest- spectrum recorded so far in the COS-B catalogue (2.5±0.6, Bignami et al.,

1981). Nearly 2a separate the two slopes but, bearing in mind the difficult conditions of the present detection, the two spectra are comparable. In contrast, under the assumption that 3C427.1 is at the origin of the gamma rays, there would be a factor of 5.7 between the intrinsic gamma-ray luminosities of the two quasars (the flux ratio at 70-150 MeV is 2.3+1.6 while the distance ratio is 13). Besides, the analysis has been applied at 300-5000 MeV at the position of 3C427.1 and the excess has been found slightly more significant than at the position determined at low energy (0.92a against 0.6a). Hence, the presence of highly relativistic electrons responsible for the extended radio lobes make the radiogalaxy a likely candidate for a gamma-ray source. This idea is marginally substantiated by the soft flux observed from 3C273, although no X ray have been detected from 3C427.1 while 3C273 is a 3 UFU source (4U1226+02.

Forman et al. 1978). In conclusion, only two objects have been noticed as possible counterparts of the observed gamma-ray source and for its nature and remarkable spatial coincidence, the radiogalaxy 3C427.1 is the most interesting. _ l?3 -

5. CONCLUSIONS

The study of the diffuse gamma radiation in the nearby Cepheus region has

yielded an emissivity spectrum of the atomic gas, (1.1 ± 0.3, 0.7 ± 0.18, 0.67 ±

0.17) 10"" "-1 s"1 sr~x, in very good agreement with Che values measured in

other close regions at medium latitudes (as observed by COS-B and SAS-2) and

near the galactic plane. The knowledge of this spectrum seems therefore reliable

in the solar neighbourhood. Above 150 MeV, the study of the quiet and cold

Cepheus molecular clouds has allowed a calibration of the N(H2>/WC0 ratio, with

an estimate of (2.3 ± 1.2) 1C2P molec. cm"* K"1 km"1 s identical to the value

determined from the study of the entire galactic plane. Below 150 MeV. however,

a statistically weak but systematic deficiency of the gamma radiation

originating in the CO clouds has been observed.

The analysis has revealed the existence of a significant (3.9o) excess

which stands out above the diffuse emission. The excess represents a good

candidate for a new bright gamma-ray source. But its softness limited the

significance of the detection and further observations are needed to confirm its

existence and position. Only two possible counterparts have been noticed in the

rather large error box (1.8° in radius): a variable X ray star and a more likely

radiogalaxy. The spatial coincidence of 3C427.1 with the source is remarkable but no conclusion can be reached on a possible relationship. Nevertheless, if confirmed by future experiments, the latest COS-B source by its position at high

latitude and by its extreme softness will add but a piece to the puzzle of the unidentified gamma-ray sources. _ cm -

REFBBESCES

Bignami G.F., Bennett K,. Buccheri R., Caraveo P.A., Hermsen W., Kanbach G., Lichti G.G., Masnou J.L.. Mayer-Hasselwander H.A., Paul J.A., Scarsi L.. Swanenburg B.N., Wills R.D., 1981, Astr. Ap., 93. 71.

Bloemen J.B.G.M., Caraveo P.A,, Hermsen W., Lebrun F., Maddalena R.J.. Strong A.W., Thaddeus P., 1984, Astr. Ap., 139, 37.

Bloemen J.B.G.M., Bennett K. , Bignami G.F., Blitz L. , Caraveo P.A., Gottwald M., Hermsen W., Lebrun F,, Mayer-Hasselwander H.A.. Strong A.W., 1984b, Astr. Ap,, 13S, 12.

Bloemen J.B.G.M., 1985, Proc. XlXth Int. Cosmic Ray Conf., La Jolla, 1, 313,

Bloemen J.B.G.M., Strong A.W., Blitz L., Cohen R.S.. Dame T.M., Grabelsky D.A., Hermsen W,, Lebrun F., Mayer-Hasselwander H.A., Thaddeus P., 1986, Astr. Ap., 154, 25.

Dame T.M., Ungerechts H., Cohen R.S., de Geus E.J., Grenier I.A,, May J., Murphy D.C., Nyman L.-A., Thaddeus P., 1987, Ap. J., 322, 706.

Eadie W.T.. Drijard D., James F.E.. Roos M., Sadoulet B., 1977, Statistical Methods in experimental physics, North Holland, Amsterdam,

Forman W., Jones C, Kominsky L., Julien P.. Murray S.. Peters G., Tananbaum H., Giacconi R.. 1978, Ap. J, Suppl., 38, 357.

Grenier I.A., Hermsen W., Clear J., 1988, Astr. Ap., in press.

Heiles C, 1976. Ap. J., 204, 379.

Heiles C., Habing H.J., 1974, Astr. Ap. Suppl., 14, 1.

Hermsen W., 1980, Ph. D. thesis, University of Leiden.

Ilovaisky S.A., Lequeux J., 1972, Astr. Ap., 18, 169.

Kutner M.L., Leung C.H., 1985, Ap. J., 291, 188.

Lebrun F.. Bignami G.F., Buccheri R., Caraveo P,A., Hermsen W., Kanbach c. Mayer-Hasselwander H.A., Paul J.A.. Strong A.W., Wills R.D., 1982, Astr. Ap. , 107, 390.

Lebrun F., Paul J.A,, 1983, Ap. J., 266, 276.

Lebrun F.. Bennett K., Bignauii CF., Bloemen J.B.G.M., Buccheri R., Caraveo P.A., Gottwald M., Hermsen W.. Kanbach G., Mayer-Hasselwander H.A., Montmeil» T., Paul J.A., Sacco B.. Strong A.W., Wills R.D., 1983, Ap. J,, 274. 231.

Lebrun F., Paul J.A., 1985, Proc. XlXth Int, Cosmic Ray Conf.. La Jolla. 1, 309. - JÏJ- -

Lebrun F., 1986, Ap. J., 306, 16.

Leung CM.. Kutner M.L., Mead K.N., 1982, Ap. J., 262, 583.

Lucke P.B., 1978, Astr. Ap., 64, 367.

Maddalena R.J., 1986, Ph. D. thesis. Columbia University.

Manchester R.N., Taylor, 1987, Ann. Rev. Astr. Ap., 25.

Marsalkova P., 1974, Ap. Space Sci. Rev., 27, 3.

Marshall F.E., Boldt E.A., Holt S.S., Mushotzky R.F., Pravdo S.H., Rothschild R.E., Serlemitsos P.J. . 1979, Ap. J. Suppl., 40, 657.

Mayer-Hasselwander H.A., 1985, Explanatory Supplement to the COS-B database.

Pollock A.M.D., Bennett K., Bignami G.F., Bloemen J.B.G.M., Buccheri R.. Caraveo P.A., Hernsen W.. Kanbach G., Lebrun F.. Mayer-Hasselwander H.A. . Strong A.W., 1985. Astr. Ap., 146, 352.

Smith et al., 1976. Ap. J., 210, 627.

SpinradH., Stauffer J., Butcher H., 1985. Ap. J., 296, 784.

Strom , Kronberg. 1976,

Strong A.W., Bignami G.F., Bloemen J.B.G.M., Buccheri R., Caraveo P.A., Hermsen W., Kanbach G., Lebrun F., Mayer-Hasselwander H.A., Paul J.A., Hills R.D., 1932. Astr. Ap.. 115, 404.

Strong A.W., Bloemen J.B.G.M., Hernsen H,, Mayer-Hasselwander H.A., 1985, Proc. XlXth Int. Cosmic Ray Conf., 1, 317,

Strong A.W., Bloemen J.B.G.M., Hermsen W., Lebrun F.. Mayer-Hasselwander H.A., Buccheri R., 1987, Astr. Ap. Suppl., 67, 283.

Strong A.W., Bloemen J.B.G.M,, Dame T.M., Grenier I.A.. Herrasen W., Lebrun F.. Nyman L.-A., Pollock i... Thaddeus P., 1988, Astr. Ap.. in press.

Weaver H., Williams D.R.W., 1973, Astr. Ap. Suppl.. 8, 1.

Wood K.S., Meekins J.F., Yentis D.J., Smathers H.W., McNutt D.P.. Bleach R.D., Byram E.T.. Chubb T.A., Friedman H.. Meidav M., 1984. Ap. J. Suppl.. 56. 507. . lit -

la) model without a point-source component

Energy ABC N(Ha)/WCO

MeV 10-**at-1s-lsr-1 lO-'cm-^sr"1^1!™-1 10-6cm-2sr-1s-1 mol.cm-2K-1km-1s

70-150 0.35 ± 0.35 7. ± 4 82. ±9. 10 ± 11

1S0-300 0.70 ± 0.18 3.25 ± 1,75 22. ± 5. 2.32 ± 1.38

300-5000 0.63 ± 0.17 3.1 ± 1.5 22.0 ± 4.5 2.46 ± 1.36

lb) model with the gamma-ray source

Energy ABC S(H2)/HC0

MeV 10-"at-1s-1sr-1 10-scm-2sr-1K-lkm-1 lO-'cm-^sr^s"1 mol.cm-2K-1km-:ls

70-150 1.1 ± 0.3 0.5 ± 5.0 72. ± 10. 0.23 ± 2.3

150-300 0.70 ± 0.18 3.25 ± 1.75 22. ± 5. 2.32 ± 1.38

300-5000 0.67 ± 0.17 3.0 ± 1.4 21.0 ± 4.5 2.24 ± 1.19

Table 1: Maximum-likelihood values obtained for the parameters describing the diffuse emission for the three energy ranges and the consequent ratio N(H=)/WCO= B/2A. a) model without a point-source component b) model with a point-source at 1= 111°, b= 19.8° - lî» -

Energy Source flux

MeV HT6 cm-2 sr"1 s"1

70-150 2.3 ± 0.7

150-300 < 0.18 (la)

300-5000 0.07 ± 0.09

Table 2: Maximum-likelihood fluxes of the proposed gamma-ray source at 1= 111°. b= 19.8° in the three energy ranges. . m -

FIGURE CAPTIONS

Figure 1: gamma-ray intensity maps of the 2nd galactic quadrant from the COS-B database. The instrumental background has been subtracted. a) intensities in the 70 - ISO MeV range. The contour interval is 5 10"3 cm"2 s"1 sr"1. b) intensities in the 150 - 300 MeV range. The contour interval is 3 10"s cm"2 s"1 sr"1. c) intensities in the 300 - 5000 MeV range. The contour interval is 3.5 10"' c«"a s_x sr"1.

Figure 2; Comparison of the observed intensities and predicted from the average gas eaissivities measured in the solar vicinity, plus the instrumental noise. a) scatter plot produced for 4°x4° bins b) map of the differences expressed in a. All discrepancies within ±0.5o are set to 0.

Figure 3: 2° by 2° maps of the convolved HI and CO data, the observed gamma-ray intensity and the differences between the observed flux and expected from the best fit presented in table lb. The latter are expressed in a. a) 70 - 150 MeV b) 150 - 300 MeV c) 300 - 5000 MeV

Figure 4_^ Position of the proposed gamma-ray source. The likeliest position (1110. 19.8°) lies at the center of the la confidence region (circle). The eccentric cross marks the position of the radiogalaxy 3C427.1. . l!9 -

=^E=T ? 180* 170° 160" 150° 140° 130° 120* 110° 100° 90° Longitude

-i—i—i—i—i—i—i—i—i—:—i—i—i—|—i—i—i—r

.20° .sat-

.10°

180° 170* 160* 150* HO» 130° 120° 110° 100° 90° Longitud*

- riguri 1 - - 150 MeV: erved Intensity (10" en' t~ sr" )

10 Predicted Intensif

- Figure 2a- - I9T _

latitude

25* •

*

1 ' ' • 186' 111* 116* 121' longitude

- Figur» 4 - - V -

L'EMISSION GAMMA DES PULSARS

" Twinkle., Xuiiniilz, CUtle. itaA. How 1 wondvi what you ate " Jane TayloA. - X .

LES PULSARS ET LES RAYONS GAW1A: DE NOUVEAUX INDICES

Isabelle Grenier

Service d'Astrophysique Centre d'Etudes Nucléaires de Saclay

Ecole, de GocUtùu de. 19S7; "Vhotoni X et gcana et pKocuius, non th.va1tf4u.ei" - 33 -

La nature des deux plus éclatantes sources f du ciel a été rapidement ^masquée grâce â l'extrême régularité de leurs signaux. Dès 1974, elles furent reconnues comme le célèbre pulsar du Crabe, né en 1054, et celui des Voiles, vieux de 30000 ans et qu'on appelle aujourd'hui familièrement Vela. Par la suite, les sept années d'observations assidues de COS-B n'ont pas permis d'agrandir la famille des pulsars ? (voir l'article de J. Paul). Il se peut que les effets géométriques aient une importance telle que nous manquions les faisceaux émis par les autres pulsars. Mais avoir intercepté ceux du Crabe et de Vela tiendrait alors vraiment de 1'extraordinaire! Il semble plus facile d'imaginer que seuls les jeunes pulsars ont les capacités de briller à haute énergie. L'étude de ces deux sources œ est essentielle pour la •-1 ,1 I l| GammnT(50 -GOOD H«V) firenler et al, 87 compréhension des étoiles à ISO neutrons et des pulsars car . les rayons 7 emportent avec VELA eux une fraction appréciable r n Mi l de l'énergie de l'objet (quelques pour-cent de / •"VfillILMv- l'énergie rotationnelle). Or ,IM nous n'avons que deux Gamma ( 1, - 30 H«V) jumer et al. 8» spécimens à ausculter. ^ : J •^JK^J Comment discerner 0.» l'essentiel du cas A 1 \ op'*" particulier? D'autant que,o .10 si tous deux émettent un - puissant rayonnement 7 ; Ji Radio (2295 MHz) caractérisé par deux ; impulsions, là s'arrête cf ~o7 1 toute ressemblance... FJmJ ph*" Le premier a été continûment détecté du domaine radio aux rayons 7, avec une parfaite synchronisation des impulsions à toutes les longueurs d'onde. Par contre, aucun signal X puisé de Vela n'a encore été capté et ses impulsions radio, optiques et 7 apparaissent désynchronisées (Harnden et al.1985, fig.1, Bignami et Hermsen 1983). Enfin, COS-B a révélé deux pulsars aux comportements nettement - JH .

différents, tant en énergie qu'au fil des ans. Au calme et à l'homogénéité de l'émission du Crabe s'opposent l'étonnante complexité et la forte variabilité de celle de Vela.

SOUFFLEES, LES CHANDELLES DE CALIBRATION GAWA ! Le flux 7 du Crabe est resté globalement assez stable sur sept ans, ce qui correspond bien au calme observé en X ou en optique. On a pu cependant noter une légère et régulière évolution de l'intensité relative de la seconde impulsion 7 au cours des observations (fig.2a, Clear et al.1987).

T -t 197» 1979 19J0 1981 1982 fig. 2a: CMB 50-3000 M«V puisa ratio P2/P1 Dans le même temps, Vela a Iflg. 2b: VELA 50-5000 M«V total flux par contre connu de violents soubresauts. Sa luminosité 7? » s'est révélée nettement s 3» variable sur une échelle de | a temps de la dizaine de jours î «

â l'année, avec une° ,5 amplification du phénomène m vers les basses énergies , (fig.2b, Grenier et al.1987 ICRC). Cet aspect du pulsar n'était pas complètement inattendu puisque l'émission radio était elle-même sujette à de fortes variations (McAdam 1981). Même s'il n'est pas encore possible de se prononcer sur une variabilité optique à cause de la très faible magnitude de Vela (proche de 24), il s'agit maintenant de comprendre l'origine d'une telle agitation. Le "flash" 7 observé en 1975 survenant deux à trois semaines après une brusque accélération de la rotation du pulsar, un "glitch", on pourrait s'interroger sur un éventuel lien de cause à effet. Malheureusement, la rareté des photons 7 empêche de suivre l'évolution de la luminosité sur moins d'une semaine et la date exacte du glitch est inconnue dans un intervalle de 15 jours. - isr_

D'un point de vue temporel et énergétique, les deux phénomènes ne sont pas incompatibles, mais on ne peut rien conclure de plus précis, d'autant que la forte luminosité de 1981 ne correspond à aucun glitch répertorié. Néanmoins, les principaux modèles d'émission des pulsars (cf exposé d'E. Asseo) devront désormais tenir compte et tenter d'expliquer une telle variabilité et pourquoi ces deux pulsars réagissent si différemment.

L'ASPECT SPECTRAL DE L'EMISSION GAMMA DES PULSARS Là encore, il paraît difficile de réconcilier nos deux spécimens. La distribution en énergie de l'émission puisée du Crabe peut être décrite entre 50 MeV et 3 GeV par une unique loi de puissance d'indice 2.00 + 0.10 (fig3a, Clear) qui rejoint correctement les observations effectuées a plus basse énergie. L'allure du spectre complet et la coïncidence des impulsions â toutes les énergies semblent donc favoriser les modèles qui utilisent un pur rayonnement synchrotron de l'infrarouge aux ? (Knight 1982), ou bien le 10" ' I I ] ' ' l_L.- ^1M m-2 NT1 io° (GeV) synchro-Compton cher à Cheng et al. où les particules donnent un petit coup de pouce aux photons X pour peupler la gamme du GeV. Toutefois, il reste â expliquer l'origine de la composante "interpic" que l'on voit culminer dans le domaine X: son spectre, identique à celui des pics en 7 mais différent en X, indique-t-il une diffusion Compton? 0e nouveau, Vela dévoile une plus grande complexité. Les premiers travaux de Kanbach et al. en 1980 la suggéraient déjà et les dernières analyses l'ont pleinement confirmée (Grenier et al.1987). Il n'est plus possible de décrire le spectre 7 de Vela par une simple loi de puissance entre 50 et 5000 MeV. Deux sont nécessaires, d'indices 1.72 + 0.07 et 2.12 + 0.07 respectivement en-dessous et . 3T -

au-delà de 300 MeV environ. Puis il est devenu indispensable de dissocier cinq composantes de l'émission y, reliées aux 5 intervalles de phase fléchés sur la figure 1. En plus de leurs différentes luminosités, ces composantes sont reconnaissables à leurs allures de spectres diverses, avec des cassures plus ou moins prononcées, et à leur évolution temporelle propre. Globalement, les rayons y émis dans la région interpic et pendant la seconde impulsion sont plus durs que pour le premier pic. Hais il faut surtout porter une attention particulière au rayonnement interpic qui se sépare en deux composantes apparemment en phase avec les deux impulsions optiques. La distribution en énergie très "bombée" de l'émission "interpicl" s'oppose à la forme convexe du spectre de sa voisine en phase (interpic2). flg. 3b: Les 5 périodes (n* croissant avec Cheng et al. ont également lëtôips) correspondant aux flux figure 2b.

raisonnablement reconstruit VELA: Spectral variability le spectre complet de Vela î.E-a mais ils ne mentionnent pas l'existence de plusieurs I.E-7 composantes à haute énergie. En 1983, Morini avait » i.e-9 cependant montré par un modèle géométrique simple que 3 l.E-9 1 la seconde impulsion y IO pouvait naître de la : i.E-io superposition de photons émis près du cylindre de lumière I.E-ll (synchro-Compton) et de y wnole «Used Mission issus de régions proches de 100 loco (nev) la calotte polaire (Compton inverse sur les rayons X thermiques de l'étoile). La première impulsion y correspondait alors à un pur effet synchro-Compton dans la magnétosphère externe et les impulsions optiques provenaient de régions polaires à moyenne altitude. Ce modèle, étayé en 1986 par F. Smith, vient aujourd'hui appuyer l'idée que les diverses composantes y récemment décelées émanent de plusieurs processus qui oeuvrent en divers lieux, en altitude et en latitude, de la magnétosphère du pulsar. Cette idée est d'ailleurs renforcée par le fait que la variabilité temporelle affecte •différemment les cinq phases. Cette dernière résulte en fait d'une . m.

distorsion des spectres à basse énergie au fil du temps (fig.3b), dont l'ampleur varie selon la phase. Tandis que les émissions interpic changent considérablement (le flux du 1er interpic peut subir un facteur 7 en 15 jours!), le second pic se montre un peu plus calme et le premier pic assez stable. On peut donc suivre les modèles de Smith et Morini pour placer le site responsable des deux composantes interpic dans les régions médianes de la magnétosphère (0.5 x Rcyl-lum) au-dessus d'un pôle. Le phénomène de variabilité concernerait alors plutôt les profondeurs de la magnétosphère d'un pulsar. Le Crabe, dont les impulsions naîtraient plutôt dans les régions externes, y serait de ce fait moins sensible. Ainsi 1'astronomie y a-t-elle certainement perdu Vela comme source de calibration. Et le Crabe ne vaut guère mieux. Mais il lui reste l'émission galactique diffuse comme étalon et en compensation, elle a gagné deux pulsars animés qui devraient nous aider de manière capitale à sonder les magnétosphères des étoiles à neutrons et à clairement établir le fonctionnement des pulsars.

•EFEIEKCES: Blgnaai G.F., Horasen W., 1983, Ann. Rev. Astr. Astrophys., 21, 67. Cheng K.S., Ho C, Buderaan M., 1986, Astrophys. J., 300, 522. Clear J., Bennett K., Buccheri R., Grenier I.A., Hansen W., Mayer-Hasselwander H.A., Sacco B., 1987, Astron. Astrophys., 174, 85. Grenier I.A., Herasen W., Clear J., 1987, Proc. 20th Int. Cosaic Bay Conf., 0G 2-9. Grenier I.A., Hernsen Y., Clear J., 1987, Astron. Astrophys., souals. Harnden F.B., Grant P.O., Seward P.O., 1985, Astrophys. J.,299, 828. Kanbacn G. et ai. (COS-B Caravane Collaboration), 1980, Astron. Astrophys., 90, 163. Knight F.K., 1982, Astrophys. J., 260, 538. Manchester R.K., Wallace P.T., Peterson B.A., Elliot K.H., 1980, M.N.B.A.S., 190, 9P. McAdaa V.6., 1981, Proc. Astr. Sac. Austr., 4, 219. Morini M., 1983, Mont. Hot. Boy. Astr. Soc, 202, 495. Siith F.G., 1986, Mont. Not. Boy. Astr. Soc, 219, 729. TDaer O.I., Oayton 8., Long J., O'Neill T., Zycb A., White S., 1984, Nature, 310, 214. I9t Astron. Astrophys. 174. 85 94 11987) ASTRONOMY AND ASTROPHYSICS A detailed analysis of the high energy gamma-ray emission from the Crab pulsar and nebula J. Clear6, K. Bennett6, R. Buccheri3, I.A. Grenier3, W. Hernuen1. H.A. Mayer-Hasselwander*, aad R Sacco3 The Caravane Collaboration for the COS-B satellite 1 Laboratory for Space Research Leiden. Leiden. The Netherlands 2 Istituto di Fisica Cosmica del CNR. Milano, Italy 3 Istituto di Fisica Cosmica e Informatics del CNR. Palermo, Italy * Max-Planck-lnstitut fur Physik und Asirophysik. Institut fur Extraterrestrische Physik. D-8046 Garching-bei-Munchen. Federal Republic of Germany 5 Service d'Astrophysique. Centre d'Etudes Nucléaires de Saclay. Gif-sur-Yvette. France " Space Science Department of the European Space Agency. ESTEC, 2200 AG Noordwtjk. The Netherlands Received July 21. accepted September 11, 1986

Summary. The results of a detailed analysis of the gamma ray by approximately 0.4 in phase. The radiation from the pulsar is emission from the Crab pulsar and nebula are reported. The data thought to result from the interaction of high energy particles are the combination of 6 observations of the galactic anti-centre with the intense magnetic fields that are known to exist at the region made by COS-B in the energy range 50 MeV to 3000 MeV. surface of neutron stars, (see Knudt and Krotscheck. 1980 for a The accumulation of the data over a period of 6.7 years from detailed review). 1975 to 1982 has allowed a study of the temporal behaviour of The Crab Nebula was the first identified celestial gamma ray the pulsed gamma ray flux from PSR 0531 + 21. A study of the source and its emission has been observed up to the highest ratio of the two gamma ray peaks in the pulsar's lightcurve shows gamma ray energies (KnifTen et al., 1974; Lichti et al.. 1980; evidence at a 3tr level for time variability over a scale of years. Cawley et al., 198S). The emission spectrum from the nebula can Fluctuations in the gamma ray Dux from the pulsar have also be represented by a single power law of index 2.08 from X-ray been observed. For the first pulse the statistical significance of the energies (Toor and Steward, 1974} up to energies of several MeV fluctuations is not large enough (2.0a) to claim variability. In the (Walraven et al.. L975), thus supporting the idea that synchroton case of the second pulse a decrease in the flux during the period radiation is the dominant emission process over this energy range. 1975-76 is observed. The significance or this variability is 3 30 MeV; Bennett et al. Key wordK gamma rays-PSR 0531 + 21 - Crab nebula - COS-B U977). 50-5000 MeV; Porter et ai. (1974). > I TeV. At gamma ray energies the total emission spectrum from the pulsar may be re­ presented by a single power law of index 2.1 iBennett etal.. 1977). I. Introductioa However, at the highest energies (above 1 TeV) some softening of the spectrum is required (Dowthwaite et al.. 1985). The Crab Nebula is one of the most unusual and well studied A systematic variation in the pulsar spectral index with phase celestial objects. Its emission has been observed from radio waves has been noted by Pravdo and Serlemitsos (1981) and it has been up to high energy gamma rays. The emission mechanism from the suggested that the pulsar emission may be represented by two nebula is well understood, with synchrotron radiation from very independent components; (i) the emission from the two main high energy electrons moving in a magnetic field providing a good peaks which dominates the pulsed emission over the entire energy explanation for the observed spectrum. The discovery of the Crab spectrum and (ii) the emission from the inlerpulse which is just pulsar has provided a source for the energy .supply which is visible at optical energies but represents 15-30°,, of the pulsed required to sustain the nebular emission, while the pulsar itself emission from X-rays to high energy gamma rays. The emission has also been observed throughout the electromagnetic spectrum spectrum from the first component is consistent with a syn­ with the emission being characterised by two pulses separated chrotron model, w*hile the second component may be fitted by a bremmstrahlungorCompton model I Knight, 1982), or by curva­ Send offprint requests to: J. Clear ture radiation (Hasinger. 1984). - IM

Table 1. Instrumental characteristics of each observation

Observation Pointing Aspect Date Relative Exposure period direction it.b) angle efficiency (10'cm's)

0 184=, -6' 1° 75/08/17 1.00 3.74 75/09/17 14 195'. +4= 15° 76/09/30 0.97 1.94 76/11/02 39 190=. 0= 8= 79/02/22 0.69 2.60 79/04/03 44 172°.-12= 14= 79.(08/29 0.66 1.13 79/10/10 54 188",-3= 5= 80/09/04 0.47 2.09 80/10/17 64 190°, 0= 8° 82/02/18 0.55 3.19 82/04/25

This paper reports the results of a detailed analysis of the total normalised to unity at the beginning of the experiment. The sys­ COS-B observations of the Crab region. With a total source tematic uncertainty in the determination of this value is of the exposure of L.S 10a cm2 s the data provide the largest gamma ray order of 10% (Strong et al., 1985). observations from the Crab region in this energy range. The long duration over which the data were accumulated (6.7 years) facili­ 3. Temporal analysis of PS R 0531 + 21 tates a study of the temporal behaviour of the gamma ray emis­ The arrival times of the selected gamma rays were transformed sion, while the good statistics available allow a measurement of to Solar System Baryoentric Time T within an accuracy of the spectral index of the emission from both the pulsar and the 0.25 ms using the optical position of the pulsar (RA 82.88095; nebula. dec 21.9817778). The phase

2. ObservitioM epoch 7*0 according to COS-B operated virtually continuously from its launch in 1975 until it was finally switched off in 1982. During these 6.7yr 65

individual observations were made each lasting for a duration of where AT =T -T0 and 0O is the phase offset. For observations approximately 30-40 d. Instrumental characteristics and opera­ 0, 14, 39 and 64 radio parameters were available for an epoch tional details of the experiment have previously been given by close to the observation period (Gullahorn et al., 1977, and pri­ Scarsi et al. ( 1977) and Mayer-Hasselwander ( 1985). By a selection vate communications from V. Boriakoffand D. Ferguson, J.M. of high quality gamma rays the observed data-set has been con­ Rankin, and A.G. Lyne). For observation 54 the parameters were densed to produce 'the final COS-B database1 (Mayer-Hassel- determined by scanning the data from a 2-12 keV X-ray detector wander et al., 1985) which has been used for this analysis. on-board COS-B, while for observation 44 the actual gamma ray The Crab region was in the COS-B field of view during six data was scanned about an extrapolation of the parameters from of the observation periods. The details of each observation are another epoch. The long duration of observation 64 (68 days) given in Table 1. The observation period refers to the number results in a degradation of the pulsar lightcurve when a single set of the particular observation within the COS-B data-set The of radio parameters are used for the analysis. Subsequently the relative instrumental efficiency depicts the change in the detector data set for this observation were subdivided into 3 time intervals sensitivity during the mission life-lime. It is determined by a for which 3 corresponding sets of radio parameters were used. The comparison of multiple observations of the galactic plane and is pulsar parameters are given in Table 2.

Table 2. Timing parameters of PSR0531 + 21 used for analysis of data Obs. Epoch / / f Offset period JD-2440000 (s-'l ll0-'°s-!) (10-2Qs-J) *o

0 2647.0 30.13744554816328 -3.8348446531 1.1413101714 0.352344 14 2982.0625 30.12634886 -3.8310414 0.3956 0.376843 39 3946.442 30.094467179 -3.82125043 1.225 0.958065 44 4128.5 30.088458052 -3.81956 1.224 0.86 54 4508.5 30.075925066 -3.81527838 1.22 0.8S 64 5015.500000167454 30.05922409965165 -3.81011540574 0.965891824 0.579803 64 5043.500000145255 30.05830242482511 -3.80960798375 0.965664183 0.58075! 64 5074.500000244583 30.05728207409174 -3.80936211028 0.965572313 0.573312 - IS» .

87

35 In order to generate phase histograms, gamma rays were 1975 AUG.-SEP. selected over an energy lange of 50 MeV to 3000 Me V from within 30 an energy dependent cone of half angle 25 »,„„ = 12.5£-°"'(dcgrees) 20 about the pulsar position. This expression optimises the gamma ray 'signal to noise ratio' for the energy dependent instrumental 15 point-spread-function (Buccheri et al., 1983). Figure 1 shows the resultant phase histograms for each observational period. The 10 variation in the background level represents not only the differ­ 5 ence in useful observation time but also the angular dependence of the effective sensitive area and the gradual degradation of 0 relative instrumental efficiency from 1.0 at observation 0 to 0.55 at observation 64. The summation of the individual phase histo­ 15 grams yields Fig. 1 which represents all COS-B observations of the Crab region. The separation and widths of the two peaks have 10 been measured by Wills et af. (1982) and are found to be consis­ tent with corresponding measurements at other wavelengths. In S

0 1979 FEB.-MAR. 30

25

20

15

10

S

0 1979 SEP.-OCT 20

15

10

5

0

20

15

10

5 0.0 0.1 0.2 0.3 0.A 0.5 0.6 0.7 0.8 0.9 1.0 0 PHASE 20 Fig. 2. The résultant phase histogram from the summation or the six observations given in figure I 15

10

5 Fig. 1. The gamma ray phase histograms iron? PSR0531 + 21 during each of the COS-B observa­ 0„ tion!. Background levels and phase boundaries are indicated 0.2 0.t 0.6 0.8 1.0 PHASE - OS

TaMe 3. Definition of phase intervals of Crab pulsar and is found to be 1.33 ± 0.39. Combining the SAS-2 result with those from COS-B gives a probability of 0.0] that the distribution Region Phase interval of the data is due to statistical fluctuations. This result cannot be attributed to a change in the instrument response during the First pulse 0.10-0.20 lifetime of the experiment, unless this change was energy depen­ fnter-region 0.20-0.48 dent and a difference in the spectral index of the two pulses Second pulse 0.48-0.62 existed. In Sect. 4 it is shown that the differential spectrum of Background 0.62-0.10 both pulses may be well represented by a single spectral index value. (It should be noted that for observation 64 the difference in the pulse ratio from that shown by Ozel and Mayer-Hassel- the region between the peaks evidence exists for 'interpulse' emis­ wander arises from the use of a single set of timing parameters sion from tbe pulsar. This is consistent with observations at X-ray to derive the phase histogram). energies where a significant fraction of the total pulsed emission Using the instrument's calibration parameters, the effective comes from this region (Pravdo and Serlemitsos, 1981). For de­ sensitive area was determined for each observation over an energy tailed analysis of the pulsar emission the phase histogram has range of SO MeV to 3000 MeV for an assumed source spectral been divided into four regions as given in Table 3. These values index of 2.00 (this is the value determined for the total pulsed are the same as those used by Wills et al. (1982) in analysis of emission). Figure 4 shows the resultant temporal behaviour of the COS-B data. The definition of these phase intervals is somewhat flux from each pulse and the total pulsed emission during the arbitrary and is chosen solely for the COS-B data analysis. period 1975-1982. The errors indicated in the ligure are due only AJI examination of Fig. 1 indicates a possible relative temporal to statistics and do not include uncertainties in the value of the variation in the count rate of the two pulses. This effect has been relative instrument sensitivity used for each observation. These reported previously by Wills et al. (1982) and Ozet and Mayer- uncertainties are less than 10% along the whole galactic plane Hasselwander ( 1984). The number of gamma rays in each pulse (Strong et al., 1985). However, in the region of the Crab there arc (above tbe background) was determined using the 'saturation many overlapping observations and a more realistic estimate of method* described by Bennett et al. (1977). The integral COS-B the uncertainty for these observations is 5%. The first pulse point-spread-function

1975 1976 1977 1978 1979 1980 1981 1982 OBSERVATION EPOCH Fig. 3. Temporal niruimn of the counts ratio of second pulse to the lust pulse over an interval from 1975 io 1982 in the energy range 50-3000 MeV 89 3). The data can be well represented by a single power law of FIRST PULSE IP, F(£) = (2.86 ± 0.50) lO^E'200-0 l0photons,cnrsGeV t The combined observations provide sufficient data to make a detailed measurement of the energy spectrum of the three emis­ sion regions defined in Table 3. The results are shown in Fig. 7. The individual spectra are seen to agree within the statistical

SECONO PULSE IP,I uncertainties with the spectral index calculated for the total emission from the pulsar. The derived spectral indices from the individual data have values of 2.08 ±0.13 for the first pulse 1.99 ± 0.17 for the second pulse and 1.98 + 0.32 for the inter region. 1 ' Jr- ; f : T 1 TOTAL(P, -Pj-II ; c) 5. The unpubed gimmi ray emission from the Crab regioa T l ,af" i Lichti et al. (1980) reported the detection or un pulsed emission 1 1 1 from the Crab region up to an energy of 400 MeV using the data t t from the first two COS-B observations. This emission was found OK i 1 1" to contribute 45 + 10°„ to the total emission. In order to de­ termine the differential energy spectrum of the steady emission, "H skymaps of the pulsed and impulsed emission from the Crab were generated for four energy intervals. The background region OBSERVATION EPOCH defined in Table 3 is used to represent the unpulsed emission, while the remainder of the pulsar cycle is taken as being pulsed. Fî(. 4a-c Temporal variation of the gamma ray flux above SOMeV from i the first pulse h the second pulse and c the total pulsed emission. Figure 8 shows the resulting longitudinal profiles determined by Only statistical errors are indicated summing the Mux in galactic latitude over a region centered on the pulsar position. The extent of the latitude summation is de­ termined by the instrument's point-spread-function over the par­ ticular energy range and is indicated in the ligure. The angular lions from either of the two pulses and the remainder of the width of the point-spread-function is seen to decrease with in­ pulsed emission. Table 4 shows the pulse ratios and flux values creasing energy. Using a method similar to that described by during each observation. Kan bach et al. (1980) the total flux from the pulsed profile Fp and

the total flux from the unpulsed profile Fup may be expressed as

F = P+fiS+B) 4. Energy spectrum of the pulsed emission from PSR0831 + 21 p

Fup = |l-/)(S + B| In order to calculate the differential energy spectrum or PSR 0831 + 21 the data set was sub-divided into 6 energy inter­ where P is the flux from the pulsar. 5 is the unpulsed source flux, vals. In Fig. 5 the resultant phase histograms Tor each of the B is the diffuse galactic background and / is the duty cycle of the energy intervals is shown and pulsed gamma ray emission is seen pulsar. The combination of the pulsed and unpulsed longitudinal up to an energy of 3 GeV. Using the same method as described profiles gives the total gamma ray emission from the Crab region in Sect 3 the actual number of pulsed counts was determined for in each energy range. Assuming that the background flux may each energy interval. Using the energy dispersion probability and be well represented by the emission from the regions outside the sensitive area of the instrument the count rate spectrum was point-spread-function then a test may h* made using the pulsed deconvolved to give the actual differential energy spectrum of the and unpulsed profiles for the value i such that pulsar. Figure 6 shows the derived pulsed spectrum for the com­ bined emission from both pulses and the inter-region (see Table x = S;P

Tabled. Relative pulse strengths and flux values (50-3000MeV) during each observation

Observation period 0 14 39 44 54 64

Pulse ratio P./P, 0.96 + 0.19 0.43 + 0.17 0.32 ± 0.09 0.39 + 0.21 0.39 + 0.15 0.91 ± 0.29 Pulse flux ( 10 "6 photons/cm2 s) H, 2.54 ± 0.27 2.63 ± 0.41 3.65+0.31 2.51 +0.44 2.96 ± 0.38 1.73 + 0.28 P, 2.35 ± 0.24 1.13+0.26 1.15 + 0.23 0.99 + 0.36 1.14 + 0.29 1.57+0.25

P,+P2 + l 5.54 + 0.41 4.47 + 0.46 5.30 + 0.46 4.52 + 0.79 4.31 ±0.48 3.59 ± 0.34 toi -

10

CRA8 PULSAR

10 5 "«•'or'E^, ID S 10"6

o a.=c 10- 7 x =3

„-9 10"' 10"' 10u 10 ENERGY (QeV) Fig. 6. The differential pulsed spectrum from PSR0531 + 21. The line is a maximum likelihood fit to the data. Only statistical errors are indicated

h should be noted that this emission may not be entirely attrib­

si. n 800-1500MeV uted to the Crab nebula as the angular resolution of COS-B is not sufficient to discriminate the emended nebula emission from i^l m-, nEMn n Pnn a possible steady component from the pulsar. However, results from X-ray observations show the pulsar completely 'switched 1500-3000MsV off" during the background phase interval lHarnden. 1983).

An n nnn 6. Discuuio* and conclusion» C 05 Q6 0,7 n n , ri The extensive observations by COS-B of the galactic anti-centre PHASE region have facilitated a detailed study of the gamma ray emis­ Fif. 5. Phase histograms of the total gamma ray emission over all obser­ sion from the Crab pulsar and nebula. With a total exposure of vations in six energy intervals

Table 5, Calculated values of parameter x and d.c. flux values

The value P has been determined for each energy inierval by using Energy interval Steady flux S the energy spectrum of the pulsar given in Sect. 4. Table 5 shows (MeVI (photons'cm ~ s GeVl the derived values of s and the corresponding value of the steady gamma ray flux S from the Crab region for each energy interval. 50-100 1.67 ± 0.50 9.55 ±5.67 10' Figure 9 shows the resultant differential energy spectrum. This 100-200 1.13 ±0,35 1.62 ±0.9210" has been fitted with a single power law spectrum of 200-500 0.33 ±0.16 9.44 r 6.S9 10- 500-3000 < 0.60 \2o)

nTmj—r i i imi| 1 i i i nu i\r 111JEL| 1 i iniiij—r—rrnTç °' \ CRAB PULSAR : CRA8 PULSAR FIRST PULSE INTER-REGION -

1?7*10 E|GeV)

s: 10

• "••' ' • ' '

ENERGY IGeV) Rj, 7i-c The differential spectrum from a the first pulse b the second pulse and c the imerpulse. The lines are fiut o the data for an assumed spectral index of 2.00

1.5 10" cm2 s this gives the largest gamma ray data-set on this significance of the features to 2.0tf for the first pulse, 3.0

165 175 185 195 205 145 175 185 155 205 LONGITUDE LONGITUDE

500-3000MeV \'? -8°sbs-i.°

60 h

<»Q - WJL

3 up to L

175 185 195 165 175 IBS 195 205 LONGITUDE LONGITUOE Fig. 8. Galactic longitudinal profiles of the pulsar and nebular gamma ray emission for four energy intervals: 50-100 MeV: 100-200 MeV: 200- 500 MeV; 500-3000 MeV. The extent of the galactic latitude summation is indicated and is chosen to maximize the signal-to-noise of the profiles

results from HEAOA-4 observations {Jung, 1986) indicate a sets are in good agreement, thus supporting the spectral break at break in the spectrum from the nebular emission at an energy hard X-ray energies and the continuation of the softer spectrum of 150 keV with the spectral index increasing from 2.13 ± 0.05 up to 500 MeV. However, due to the poor spatial resolution of to 2.54 ± 0.08 and is in agreement within the statistical uncer­ both HEAOA-4 and COS-B experiments this spectrum may tainties with the value calculated from the present data. The contain a contribution from a possible steady emission from the extrapolated spectrum of Jung is indicated in Fig. 9. Both data pulsar during the non-pulsed phase interval. At very high energies. A04 -

93

wander, H.A., Paul, J.A., Scarsi, L., Swanenberg, B.N., Taylor, m-m 1 i i HUH—i—i 11 nui B.O., Wills, R.D.: 1977, Astron. Astrophys. 61, 279 CRAB NEBULA Browning, R., Ramsden, D., Wright, P.J.: 1971 Nature 232, 99 Buccheri, R, Bennett, K., Bignami, G.F, Bloemen, J.B.G.M, Boriakov, V„ Caraveo, P.A., Hermsen, W, Kanbach, G-, Manchester, R.N., Masnou, JX„ Mayer-Hasselwander, HA., Ozel, M.E., Paul, J.A., Sacco, B., Scarsi, L., Strong, A.W.: 1983, Astron. Astrophys. 128, 245 Cawley, MJ, Fegan, DJ, Gibbs, K., Gotham, P.W, Lamb, 0.63. NT7 E-|^ R.C., Liebing, D.F., MaciCeotvn, P.K., Porter, N.A., Stenger, VJ., Weekes, T.C.: 1985, Proc. 19th Int. Cosmic Ray Conf, La Jolla, USA, 1, 131 Dowthwaite, J.C., Harrison, A.B., Kirkman, I.W., Macrea, H.J., McComb, TJ.L., Orford, KJ., Turver, K.E., Walmsley, W.: 1985, Astrophys. J, 286, L35 Eadie, W.T., Drijard, D„ James, F.E., Roos, M., Sadoulet, B.: 1971, Statistical Methods in Experimental Physics, North Holland, p. 230-232 Fazio, G.G., et aL: 1972, Astrophys. J. Letters, 175, L117 Gullahom, G.T, Isaacman, R., Rankin, J.M., Payne, JJ.: 1977, Aslron. J. 82, 309 Graser, IL, Schônfelder, V.: 1982, Astrophys. K 263, 677 Greisen, K., Ball, S.E., Cambell, M.. Gilman, D, Strickman, M , McBreen, B., Koch, D.: 1975, Astrophys. J., 197, 471 Harnen, F.R.: 1983, Supernova Remnants and their X-Ray Emis­ sion, p. 131. (eds. Danziger, J. and Gorenstein, P., IAU) Hasinger, G.: 1984, Proc Int. Symp. on X-ray Astron, Bologna, p. 305 Haymes, R.C., Glenn, S.W., Fishman, GJ, Harnden, F.R.: 1969, J. Geophys. Res., 74, 5792 10' Hermsen, W.. 1980, Ph.D. thesis, Univ. Leiden ENERGY (GeV) Jung, G.V.: 1986, Ph.D. thesis, Univ. of California, San Diego Kanbach, G., Bennett, K., Bignami, G.F., Buccheri, R., Caraveo, Fie* 9. The difTeretitUtl spectrum of the Crab nebula. The solid line is the P, D'Amico, N., Hermsen. W„ Lichti. G.G., Masnou, J.L., maximum likelihood fit to the data. The shaded area is an extrapolation Mayer-Hasselwander, H.A, Paul, J.A, Sacco, B., Swanenberg, of the energy spectrum from hard X-ray energies (Jung, 1986) B.N, Wills, R.DJ 1980, Astron. Astrophys. 90, 163 Kniffen, D.A., Hartman, R.C., Thompson, DJ., Bignami, G.F., Fichtel, C.E., Ogelman, H., Turner, T.: 1974, Nature 2S1, 397 above 500 GeV, there have been reports of the detection of non- Knight, F.K.: 1982, Astrophys. J. 260, 538 pulsed gamma ray emission from the Crab region at flux levels Knudt, W„ Krotscheck, E.: 1980, Aslron. Astrophys. 83, 1 of ~510"" photons/em's (Fazio et al., 1972; Cawley et al., Laros, J.G., Matteson, J.L., Peiling, R.M.: 1973, Nature Phys. Sci. 1985). An extrapolation of the spectrum derived above gives a 246. 109 value which is more than an order of magnitude below the Lichti, C.G., Buccheri, R., Caraveo, P., Gerardi, G., Hermsen, reported flux value. W., Kanbach, G., Masnou, J.L., Mayer-Hasselwander, H.A., Paul. J.A., Swanenburg, B.N, Wills, R.D.: 1980, Non-Solar Gamma Rays, Advances in Space Exploration Vol. 7 p. 49 Acknowledgements. We wish to thank Drs J.M. Rankin, V. (eds. Cowsik, R. and Wills, R.D. Pergamon, Oxford) Boriakoff, D. Ferguson and A.G. Lyne for the communication Mayer-Hasselwander, H.A.: 1985, Explanatory Supplement to of unpublished radio data. J. Clear acknowledges receipt of a the COS-B Final Database research fellowship from the European Space Agency. The Lab­ Mayer-Hasselwander, HA., Bennett, K., Bignami. G.F., Bloemen, oratory for Space Research Leiden is supported financially by ZWO, the Netherlands Organisation for the Advancement of J.G.B.M., Buccheri, R., Caraveo, P.A.. Hermsen, H., Kanbach, Pure Research. G, Lebrun, F., Paul, J.A., Sacco, B, Strong, A.W.: 1985, Proc. 19th Int. Cosmic Ray Conf., La Jolla, USA, 3, 383 McBreen, B., Ball, S.E., Cambell, M., Greisen, K., Koch, D.: 1973, Aarophytt. J. 184. 571 References Ôzel, M.E., Mayer-Hasselwander, H.A.: 1984,. Inter. Workshop on Data Analysis in Astronomy, Erice, Italy Albatt, P., Frye, G.M., Jr, Mace, O.B., Hopper, V.D., Thomas, Porter, N.A., Delaney, T., Weekes, T.C- 1974, Proc 9th ESLAB J.A.: 1972, Nature 240, 221 Symp., ESRO SP-106. 295 Bennett, K„ Bignami, G.F., Boella, R„ Buccheri, R.. Hermsen, W, Porter, N.A, Weekes. T.C.: 1978, Smithsonian Ap. Obs. Spec. Kanboch, G., Lichti, G.G., Masnou, J.L., Mayer-Hassel- Rept., No. 381 94

Pravdo, S.H., Serlemitsos, PJ.: Astrophys. J. 246, 484 Walraven, G.D.. Hall, R.D, Meegan, C.A., Coleman, PL, Strong, A.W.. Bloemen, J.B.G.M., Buccheri, R-, Hermsen. W., Shelton. D.H., Haymes, R.C.: 1975. Asirophys. J. 202, 502 Lebrun. F., Mayer-Hasselwander. H.A.. 1985. Proc. 19th Int. White, R.S., Sweeney, W.. Turner, T„ Zych. A: 1985, Aslrophvs. Cosmic Ray Conf., La Jolla, USA, 3, 387 J. 299, L23 Thompson. D.J., Fichtel, CE., Hartman, R.C., Knilîen. D.A., Wills, R.D., Bennett, K., Bignami. CF., Buccheri, R„ Caraveo. Lamb, R.C.: 1977. Astrophys. J. 213, 252 P.A., Hermsen, W„ Kanbach, G., Masnou. J.L.. Mayer- Toor, A, Seward. F.D.: 1974. Astwn. J. 79, W5 Hasselwander, H.A.. Paul. J.A., Sacco, B.: 1982, Sature 296. Trimble, V.: 1968, ,4«ron. J. 73, 535 723 . 10 6 -

High Energy Gamma Rays from the Vela Pulsar: Long-term Variability and Energy Distribution

I. A. Grenier1, W. Hermsen2 and J.Clear3

1 Service d'Astrophysique, Centre d'Etudes Nucléaires de Saclay, France.

2 Laboratory for Space Research Leiden, Leiden, The Netherlands.

3 Space Science Department of ESA, ESTEC, Noordwijk, The Netherlands.

Send offprint requests to: I. A. Grenier

Send proofs to : I. A. Grenier (address above).

Thesaurus codes: 07.36.1; 16.13.1; 16.14.2; 19.50.1

Main Journal, Section 6 (Formation, structure and evolution of stars)

Aifuinomy and KifiophytÀ-d : in pnjui

1 - zoi .

Summary

New results on the temporal and spectral characteristics of the high energy (50 MeV to 5 GeV) gamma ray emission from the Vela pulsar are presented. A new sensitive analysis method using maximum likelihood techniques has been applied to the final COS-B dataset. The whole pulsed flux is found to exhibit long term variability. The energy spectrum of this total emission cannot be described by a single power law and the data show strong evidence for a spectral break at approximately 300 MeV. Five discrete emission regions within the pulsar lightcurve have been identified, with the spectral characteristics and long term behaviour being different.These results support the idea that various source regions simultaneously exist in the pulsar magnetosphere and that the physical processes generating the gamma rays in these sites differ with the location (in attitude' and latitude).

Key Words: PSR 0833-45 - Vela - Pulsars - Gamma Rays - COS-B - Pulsar Glitch

1. Introduction

The study of the high energy gamma ray emission from celestial objects provides an excel­ lent method in helping our understanding of the physical processes which occur within such sources. The various possible emission mechanisms are well understood, with the gamma ray emission being closely related to the central power house of the source. A detailed knowledge of the emission characteristics therefore leads to such important properties as the nature of the process, particle densities and spectra, and the strength of the magnetic field. The Vela pulsar (PSR 0833-45) is a most unusual galactic object and is one of the fastest pulsars known. First discovered by its pulsed radio emission (Large et al., 1968), it was at the time of its detection the weakest object ever observed in the visible light, while being the brightest source in the gamma ray sky. The pulsar luminosity reaches a maximum in the gamma ray domain but it has not been detected at X-ray energies.

2 - lot -

Furthermore, its lightcurve structure varies considerably with the energy range. Clearly, the study of the high energy gamma ray emission from the pulsar provides a useful method in understanding the complex energetic processes which occur in its magnetosphere.

Figure 1 shows the pulsar lightcurve at radio, optical and gamma ray energies.The radio signal (Buccheri et al., 1978) shows a single pulse arriving about a third of a period before the common mid-point of the other lightcurves. Its profile depends on the frequency and on the intensity of the emission, and could be the result of four components generated at different altitudes just above the polar cap (Krishnamohan and Downs, 1983). The pulsar was observed at optical wavelenths by Wallace et al. (1977) and Peterson et al. (1978) as a 24"1 magnitude star with two broad optical pulses seperated by 0.25 in phase on *top of a background light level which accounts for half the luminosity of the pulsar. Manchester et al. (1980) And evidence for a slight modulation of the baseline seen in the optical lightcurve. The pulsar has long been recognized as the X-ray source, 3U 0833- 45, over a wide range of energies (Kellogg et al., 1973, Culhane et al., 1974, Pravdo et al., 1976 and references therein). Accurate observations by the Einstein satellite recently resolved the source into a point-source coincident with the pulsar position and embedded in a tiny (4') nebula of diffuse emission (Harnden et al., 1985), but no pulsed emission was observed. Results from other experiments (Knight et al., I982 and references therein) report only upper limits to the pulsed emission. However, two positive detections have been reported in the past (Harnden at al., 1972, and Harnden and Gorenstein, 1973) but the results have not been confirmed. Contrarily, the reported fluxes are not consistent with these upper limits and the derived phase distribution and pulsar period respectively, are inconsistant with the other reports. The first detection of pulsed gamma ray emission from the Vela pulsar was reported by Albats et al. (1974) in the 10 - 30 MeV region. The lightcurve showed a single peak almost coincident in phase with the radio peak, however this result has never been confirmed. Turner et al. (1984) at 0.3 - 30 MeV, Thompson et al. (1977) above 35 MeV and Bennett et al. (1977) at 50 MeV - 5 GeV all report a similar lightcurve structure which is characterised by two sharp peaks separated by 0.42 in phase and bracketing the optical pulses. At TeV energies pulsed emission has been reported

3 . 209 .

from Vela at a low statistical significance, with the position and number of the peaks in the lightcurve varying (Bhat et al., 1980; Gupta et al, 1982; and Bhat et al., 1987). The energy spectrum of the pulsar still remains uncertain. An extrapolation to X-ray energies of the COS-B results of Kanbach et al. (1980) lies several orders of magnitude above the reported upper limits, while at TeV energies the flux values reported by Bhat et al. (1980) clearly require a break in the spectrum. Earlier results from COS-B data by Bennett et al. (1977) and Kanbach et al. (1980) indicate a possible break in the spectrum at a few hundred MeV, hut the level of significance of these results is low. Finally, Caraveo et al. (1987) have recently reported a possible detection of linear polarisation in the high energy gamma ray emission using COS-B data. If confirmed, this result will strongly support an electromagnetic production mechanism of the gamma rays from the pulsar.

This article presents the latest results on the Vela pulsar in the 50 MeV to 5 GeV energy range. The data have been collected using the COS-B satellite which observed the Vela region during 10 observation periods, that is for more than 300 days and a total exposure of 1.48 10s cm2 s. Whereas most of this dataset has been used previously to study the pulsar timing parameters and lightcurve structure (Wills et al., 1981), only a small fraction has been used to examine its spectral characteristics (Bennett et al., 1977, Lichti et al., 1980, Kanbach et al., 1980), and no previous study of the long term behaviour of the pulsar luminosity and spectral properties has been reported. Initially, the details of the obser­ vations and the results of a preliminary study using the traditional analysis method are presented. Subsequently, a detailed description is given of a sensitive maximum likelihood analysis method developed to study the spectral characteristics of the pulsed emission and its long term behaviour. Finally, the results of this analysis are presented and a discussion is given of how this detailed description of the Vela pulsar's behaviour at high energies may constrain the theoretical modelling of the source.

4 . 210 -

2. Observations

The gamma ray satellite COS-B operated in the 50 - 5000 MeV range. Details of the spark chamber experiment are given by Scarsi et al. (1977). For this analysis, data from the 'final COS-B database' have been used (Mayer-Hasselwander et al.,1985). The database contains information on each selected photon, together with the results of pre-flight and in-flight calibration data on the instrumental point-spread-function (PSF), on the energy dispersion probability (both as a function of energy and source aspect angle), and on the relative instrumental sensitivity for each observation. The latter shows a generally decreasing trend during the COS-B lifetime due to the degradation of the spark chamber efficiency. It's evolution was estimated from the study of overlapping observations along the galactic plane (Strong et al., 1987a), so that the sensitivity during an observation is known with an accuracy of 10% (or better when the observation refers to a repeatedly surveyed region). During its 6.7 years of almost continuous operation a total of 65 individual observations were made, each lasting typically 30 to 40 days. The Vela pulsar was in the COS-B field of view during the 10 observations in table 1.

3. Preliminary Results

For temporal analysis, the data have been folded with the pulsar timing parameters appli­ cable to the particular epoch of the observation. These were derived from contemporary radio measurements for periods 2, 3, 12, 21 and 45, and for the remaining observations parameter scanning using the COS-B data itself was performed. In order to generate lightcurves, events have been selected from within an energy dependent cone of half an­ gle ^rnu = 12.5 EQ^VB degrees about the pulsar position. This expression maximizes the signal to noise ratio within the lightcurve for the energy dependent PSF (Buccheri et al., 1983). Figure 1 shows the resultant lightcurve obtained by summing the data over all observations and compares it to the pulsed emission recorded at other wavelengths. Also indicated in figure 1 are the boundaries which are used in this analysis to study the background and phase dependence of the emission. The boundaries are specified in table

5 - 2.11 -

2.

The entire data set was initially analysed using the 'traditional' saturation method de­ scribed in detail by Bennett et al (1977). It exploits the information given by the back­ ground region of the lightcurve to determine the number of pulsed photons for a particular energy interval and phase region. Flux values may then be calculated directly from the useful observing time and effective sensitive area for an assumed source spectral index. The saturation method provides a well established quick-look analysis method which can handle data up to large aspect angles, and also facilitates verification of results derived using the maximum likelihood method which is described in detail in section 4.

To study the long-term behaviour of the pulsar, the total pulsed flux in the energy interval 50 - 5000 MeV was determined for each observation period for an assumed source spectral index of-1.89 (the index determined by Kanbach et al., 1980, using observation periods 2 and 3 only). Figure 2 shows the evolution of this flux from 1975 to 1981. The error bars indicated in the figure are statistical only and do not account for uncertainties in the instrument response. The results show significant flux variability from the pulsar. Including the uncertainty in the instrument response (10%) the chance probability of this effect lies well below 10-7.

To determine the time averaged spectral characteristics of the source, the pulsed flux was calculated using the data from all 10 observations and for 9 energy intervals (the effective sensitive area was derived assuming again a source spectral index of -1.89). Figure 3 shows the resultant spectrum. A single power law fit to the points, yields a spectral index of -1.92 for the entire energy range. However, the probability that the data is well represented by such a spectrum is less than 10~3. By restricting the fit to energies above 300 MeV a single power law fit of

K, (E) = 2.62±0.13 < KT3 £ M.V"0'07 photons/cm2 s MeV has been found to well represent the data with a probability of 0.8. Clearly, a second power

6 . i> 2. -

law or an even more complicated model is required for energies between 50 and 300 MeV.

Finally, as part of the preliminary analysis, the background region of the lightcurve has been analysed for possible steady emission from the Vela region lying above the diffuse galactic emission. The analysis method has been described in detail by Clear et al. (1987). Table 3 shows the results for 4 energy intervals. The value a represents the 95% confidence limit to the ratio of the pulsed to unpulsed emission from Vela. For each energy range, no evidence exists within the statistical uncertainties for the detection of steady emission from Vela. Over the total energy range a 95% confidence level upper limit of 1.15 10-6 photons/cm2 s has been determined.

The results of the preliminary analysis show the Vela pulsar to exhibit states of high and low activity during the epoch of the observations. The energy spectrum cannot be described by a single power law and there is evidence for a spectral break at approximately 300 MeV. In order to study the phase and spectral characteristics in greater detail a more sensitive analysis method is required. For this purpose a maximum likelihood analysis method has been developed, which simultaneously considers all COS-B information for each event, and which forms a complete model of the gamma ray emission from the Vela region. The details of this method are now described.

4. The Maximum Likelihood Method

A maximum likelihood analysis has been applied to the data to estimate the flux and spec­ tral properties of the source. The newly developed approach utilizes all COS-B information, that is the arrival time, energy and direction of each photon. It takes simultaneously into account the limited spatial and energy resolutions of the instrument, and their dependency on the inclination and energy of an event. Finally, it separates the pulsar emission from any background radiation by exploiting the pulsar lightcurve shape against the flatness of the background, and the source spatial structure (the COS-B point-spread-function) against the structure of the underlying diffuse emission originating in the interstellar medium and the isotropy of the instrumental noise. The statistical processing uses the maximum likeli-

7 . 213 -

hood principle with a probability density function of detecting a photon at an energy Em,

position lm,bm and phase m (where mis for 'measured' against 'true') expressed as:

SEI f,bn{,E,l,b) dE dXl Ckh[Em,E,amu,Ab)

/Blii//;,„b„ Numerator dEm dUm

The true celestial gamma ray distribution is described by n[,E,l,b). CAL represents the COS-B convolution, given by the product of its effective sensitive area, point-spread- function and energy resolution. These parameters have been determined from pre-launch calibration data (Hermsen, 1980, Mayer-Hasselwander, 1985). % represents the true incli­ nation angle with respect to the detector axis of the photons coming from a point source

at (l,b) and resconstructed at an angle aml,b away from this source, f) denotes the solid angle. The likelihood value can then be computed from the product of the probabilities of finding the photons actually recorded by COS-B.

For the Vela gamma ray source, the general model to test may include up to four major components:

t 1. the pulsed emission from the pulsar: F„v(cj>).E~''•'*'.£(/ - lv,b- b„)

1 2. a possible steady component from Vela: Fva.E ~~ '.8(1 - f„,6- bv)

3. a constant and isotropic instrumental background: B[.E "

4. the steady diffuse galactic emission around the pulsar: q.NH^rty.E"1»

The latter term results from the interaction of cosmic rays with interstellar matter. Its spatial structure can be traced by the HI and CO surveys (Dame et al., 1987; Strong at al., 1987b). To predict the galactic gamma ray distribution, 3" •/. 3° interstellar maps showing NH(l, b) within 30° around Vela have been constructed assuming a value of 2.8 1030 mol

-2 _1 -1 cm K km s for NH2/WCO (Bloemen at al., 1986). For each observation the entire COS-B field of view has been covered.

Testing different models is performed by evaluating the likelihood expression for the param-

8 eters of the different components: Bi, q, Fvi, Fvp(), and the different spectral slopes: -y,-, "fy> 1st •>(<£)• The highest likelihood value gives the most probable model for n(4>,E,l,b) that explains the data. As a boundary condition, the total number of photons from a selected model is assumed to equal the total number of photons detected within the entire COS-B field of view (60°) and energy range (50 - 5000 MeV). This number depends only on the phase and is not altered by COS-B. The good timing resolution of the instrument (1 ms) allows the phase of each event to be derived to an accuracy of 0.01. Thus for the analysis, the measured phase may be taken as the actual phase of each event. This is an important condition as the spectral characteristics from adjacent phase intervals may therefore be analysed independently. Table 2 and figure 1 give the details of the six phase domains which have been selected for the analysis. The specific choice of these domains is based on the apparent structure in the gamma ray lightcurve following Kanbach et al. (1980), the possible physical relationship between the gamma ray structure in interpulse-1 and the first optical peak and on the requirement to maintain good statistics. The back­ ground region (phase 0.77 - 0.07) has been analysed as part of the preliminary analysis to search for a possible steady emission from Vela above the diffuse galactic emission. The resultant upper limit for the total energy range (see section 3) shows that such emission accounts for less than a few per-cent of the pulsed luminosity, and may thus be neglected from the model. Ignoring possible steady emission in the model implies that outside the pulsed phase range, the photons only come from the two background components of the model. As a firbt step the likelihood analysis is performed for the events from the back­ ground phase domain to give the best emissivities and spectral indicies of the instrumental

background and galactic emission (Bi,~ti,q,*ig). These values are then fixed in the model and the analysis is applied to the photons from the other five phase intervals indepen­

dently to derive the emissivities Fvp(4>) and spectral indicies ~iy(

This type of analysis is more precise and powerful than the traditional saturation method

9 . zisr.

such that the total pulsed spectra as well as the spectra for the selected phase intervals can be determined for individual observations. However, for those observations with large source aspect angles (> 15°), where the efficiency and the energy resolution of the instru­ ment are poor, this likelihood analysis is limited by the available statistics and has not been performed, For the remaining observations (2, 3, 12, 45 and 59) a detailed analysis has been made for the 5 phase intervals. Within each phase interval the analysis has been performed for 3 energy intervals (50 - 5000 MeV, 50 - 300 MeV and 300 - 5000 MeV). The division of the data into these energy intervals is based on the evidence for a break in the pulsed spectrum at approximately 300 MeV found in the preliminary analysis. Due to computing time considerations the position of the spectral break has not been used as an additional free parameter within the model, however the results of the independent modelling of the spectrum above and below 300 MeV strongly supports the choice of this value.

5. Results

The likelihood analysis has been performed for the 5 selected phase intervals, for the 5 best observation periods and for 3 energy ranges, giving a total of 75 spectra and flux values. The results have been summarized in table 4. The spectral slopes have been scanned over a wide range (typically 0 to 4) and the derived regular parabolic shapes of all the likelihood functions denote that the parameters have been well optimized. The quoted errors have been determined from the shapes of the likelihood ratio functions A (defined as the ratio between the likelihood maximized over the background parameters and the likelihood maximized over all parameters). Since in the present case the quantity

2 -2 ln(A) has a x distribution with one degree of freedom, fp(), the 68% confidence level corresponds to a decrease in —2 ln(A) of 1.0. Probability estimates are given in table 4 to evaluate the chance probability of a change in the spectral shape with phase against the assumption of a homogeneous pulsed spectrum, and to examine the time variability of the spectral slopes. Details of how the likelihood analysis permits a qualitative comparison between two models are given by Eadie et al. (1982). To study the long-term evolution of

10 . Zlt, .

the fluxes, x2 values are given. In the following discussion these results shall be considered in detail giving a coherent image of the behaviour of the Vela pulsar in the gamma ray domain.

Initially, consider the time averaged, phase averaged behaviour of the pulsar. The single power law which best describes the pulsed emission over the energy range 50 - 5000 MeV has an index of 1.84 ± 0.03. This can be compared with the value of 1.89 ± 0.06 calculated by Kanbach et al. (1980) using periods 2 and 3. The spectrum of the whole pulsed emission derived independently at energies above and below 300 MeV and using all observations is:

F{50< E<300) =2.74±0.21x lO"4^^2*007 photons/cm2 s MeV

F(300 < E < 5000) = 2.71 ± 0.04 x lO-3.^^2*0'07 photons/cm2 s MeV

As an internal check, it is noted that there is complete agreement between the high energy spectrum calculated by the saturation method and the above spectrum. The significant difference between the high and low energy spectral indices clearly shows that the time averaged Vela pulsed spectrum from 50 to 5000 MeV cannot be described by a single power law. At least two are required and in this representation the good connection between the two spectra at 300 MeV strongly supports the choice of 300 MeV as the energy of the break.

The luminosity of the pulsed emission from Vela has been determined by integrating the two-power-law phase averaged spectrum for each observation. The result, displayed in figure 4a, corroborates the striking long-term variability of Vela's brightness presented in section 3. In addition, the likelihood analysis provides simultaneously the background intensity for each observation, and its stable behaviour from one observation to the next, displayed in figure 4b, confirms the true source origin of the pulsed variability phenomenon. Further confirmation of this effect comes from the phase dependency of the variability (this

11 - 11? -

shall be discussed later in this section).

To study the energy dependence of this variability, figure 5a shows the evolution of the pulsed flux for the energy ranges 50 - 300 MeV and 300 - 5000 MeV. The main contribution to the flux variability is clearly due to the low energy emission. The different evolutions of the flux from the two energy ranges denote a distortion of the Vela spectrum from one epoch to the next. The observed fluctuation in the spectral ratios (the ratio of low to high energy flux) has a probability of 10"10 of being due to a random effect. Figures 5b and 5c show the spectral indices as a function of epoch for the 3 energy ranges. As expected, the spectral index for the entire energy range is seen to vary with time with a probability of 2 10-3 that this effect is due to chance f, whereas a simultaneous analysis of the background spectra shows no indication for variability. The spectral variability corresponds to an overall steepening of the spectrum when the Vela flux is high, and to a flattening when the flux is low. The correlation still holds when the low energy part of the spectrum is considered, see figure 5c, although the effect is less significant due to the reduced number of photons in this smaller energy range (chance probability of 1.3 10-2) . For the high energy part the time averaged spectral index of 2.12 ± 0.07 is statistically consistent with each observation but a similar trend is seen. Finally, the systematic difference between the spectral slopes for the two adjacent energy intervals is clearly observed in figure 5c.

To study the phase dependence of the pulsed emission, the likelihood analysis has been performed independently for the 5 phase intervals defined in table 2. From the study of flux ratios from observations 2, 3 and 12, Kanbach et al. (19S0) found evidence that the second peak and the total interpulse exhibit a harder spectrum than the first peak, but they were unable to quantitatively measure these differences. Figure 6a shows the time averaged spectral index as a function of phase. The probability that the emission may be best represented by a homogeneous spectrum over all phases (E-1"8"1) is 10-4. Although, t Although it has been established that the Vela spectrum cannot be described by a single power (aw over

the COS-B energy range, the comparison of the evolution of the spectral index for the single power law

fit is used to support the evidence for spectral variability during the observations. Obviously, this test of

variability is less sensitive than the variation in spectral ratio.

12 . lit _

ideally a single power law is not the best lit to the individual phase regions, the result confirms the spectral differences noted by Kanbach et al. (1980). The second peak is much harder than the first (3.75c), as are to a lower degree of significance the interpeak emissions, while the trailer is seen to exhibit an extremely soft spectrum. A further understanding of the phase dependency of the pulsed emission may be obtained from figure 6b, which shows the spectral characteristics of the high and low energy emissions as a function of phase. Although, the existence of a spectral break in the whole pulsed emission has been clearly established in the above results, it is not possible, due to the reduced statistics, to draw such a firm conclusion for each phase interval separately. In all cases, the difference between the indices from the low and high energy ranges lies within the 2.5 to 3.0

Analysis of the spectral characteristics of the high energy emission shows that there is no significant variation in its spectral index with phase, giving a phase averaged value of 2.12 ± 0.07. Thus, the spectral index of the GeV radiation from Vela displays both a stable and homogeneous behaviour. In contrast, the spectral index of the low energy emission shows evidence for phase dependence (chance probability 5 10-3). It is evident from the change in the spectral index at low energies (from the first peak to interpulse-2, 4.5tr and interpulse-2 to the second peak, 3.6o-) and from the softness of the first peak and trailer over the entire energy range, that 5 separate phase components exist in the Vela gamma ray lightcurve. Unfortunately, due to limited statistics, the boundaries of the phase components could not be treated as additional free parameters. However, the selected boundaries are found to be in good agreement with the results of a cluster analysis study of the Vela lightcurve topology using observation period 2 by Buccheri et al. (1987). The two approaches (spectral and topological) concur to conclude that the pulsar gamma ray emission is composed of discrete components.

Figure 7 shows the evolution of the flux from each of the phase components for each observation and for both high and low energy ranges. The long-term trend is obviously similar for each phase component, however the relative amplitude of the variability is

13 - 219 -

significantly larger for the two interpulse emissions. In the low energy range, the flux from interpulse-1 is seen to increase by a factor of 7 within 2 weeks while the first peak remains stable during the same interval. This phase dependent effect gives weight to the hypothesis that the variability truely originates from Vela and is not an instrumental effect. It is of interest to investigate whether the flux variations in the separate energy intervals are the result of spectral distortion or are purely changes in absolute flux. Unfortunately, the low statistics reduce the sensitivity and no significant spectral index variability is observed for any of the phase components, see table 4. Figure 8 shows the spectral ratios, calculated from the Huxes derived for the high and low energy ranges, as a function of epoch for each phase component (for the trailer the statistics are too poor to produce a meaningful result). For the first peak, no significant alteration with time in the relative spectral properties from the two energy ranges is observed (chance probability 0.56). For the remaining phase components, a significant distortion of the spectrum is detected (chance probability interpulse-1: < HT25, interpulse-2: 10-20 and peak 2: 7 10"°).

To illustrate their long term evolution, the energy spectra giving the maximum likelihood fits to the data for each phase component and for each observation are displayed in figure 9. Good agreement is observed between the low and high energy fits at approximately 300 MeV for most phase components. The two power law representation is therefore a good approximation to the real spectra (for the trailer, the poor statistics make interpretation of the results rather difficult, therefore the reader is referred to the single power law fit in table 4 to judge the trailer evolution). The actual position of the break or bend in the real spectrum will vary somewhat in energy following the apparent long term variability. The phase averaged spectra, figure 9f, show the high energy emission to be rather stable throughout the observations, while the low energy emission is seen to vary significantly and is responsible almost entirely for the observed flux variability.

In conclusion, the results of the analysis show the Vela pulsar as a most complicated gamma ray source. The emission is found to be highly variable with time. Five distinct phase components are identified which are individually variable and show different spectral characteristics. The energy spectrum over the 50 - 5000 MeV region is best represented

14 . 220 _

by a two power law fit with a spectral break at approximately 300 MeV. The observed variability is found to be almost entirely due to spectral changes in the low energy part of the emission with the high energy component remaining comparatively stable throughout the period of the observations.

6. Discussion

6.1 Total PSR0833-45 Spectrum

The two-power-law-fit time averaged spectrum of the whole pulsed emission has been plotted in figure 10 together with the data recorded at other wavelengths in order to show the full energy distribution of the Vela pulsed radiation over the entire electromagnetic spectrum. Over this spectrum, the pulsar reaches its maximum luminosity in the MeV- GeV domain. At very high energies, the recent results of Bhat et al (1987) indicate a spectral slope of -3.5 at TeV energies which does not require an excessive steepening when extrapolating over three decades of energy from the COS-B gamma ray range. The observed time variability presented in this article may support the proposed time variability at TeV energies, e.g. Bhat et al. (1980), however, it should be noted that the variations recorded above a few hundred MeV are relatively small (< 25%).

In order to take into account the large time variability at lower energies, the gamma ray spectra derived in the extreme cases of periods 3 and 45 have also been displayed in figure 10. The result between 35 and 100 MeV obtained by SAS-2 in February and April 1973, is in good agreement with the COS-B time averaged spectrum, as is the spectrum derived by Turner et al. (1984) in November 1981 at lower energies (1 - 30 MeV). Their lowest energy point (0.3 - 1.5 MeV) together with the monotoneous increase in the spectral index from 1 MeV to 5 GeV strongly suggest that the shape of the whole pulsed energy distribution in the gamma ray domain is a smooth curve. The very low upper limits set by the Einstein and HEAO 1 satellites (energy ranges 0.1 - 4.5 and 15 - 175 keV) suggest that the energy distribution in the X-ray domain is still bending down before rising again towards the optical and radio wavelengths, Since the pulsed X-ray emission may also be

15 - ZZI .

time variable, the bulk of the other negative detections has been added to the figure as well. The relatively low X-ray emisBivity of the Vela pulsar is in clear contrast to that of the Crab pulsar. The latter exhibits an X-ray luminosity which is about equal to that ' in the gamma ray domain from 1 MeV to 1 GeV. This difference might indicate different physical or geometrical conditions in the magnetosphere.

6.2 Long-term Time Variability of PSR0833-45

The flux evolution presented in figures 2 and 4 demonstrates the gamma ray variability of the pulsar over several years. The abrupt increase in luminosity observed during the first two periods indicates that the variations may occur also on a much shorter time- scale. Figure 11 presents a close-up view of this epoch where the first two observations have been sub-divided into time intervals of about 10 days. The corresponding fluxes have been estimated using the "saturation method" to derive the count rates and the telescope sensitivity averaged over the full observation period. Strictly taken, the latter might not be correct, as the sensitivity might fluctuate within one observation period. However, the COS-B behaviour has proved to be very stable during the first four periods (only a 3% change in sensitivity over the first four months after launch, Strong et al., 1987a). To quantify the time scales of the evolution, two exponential functions have been fitted to these data and are displayed in figure 11. The resultant rise and decay times are 11.4 ± 1.0 and 137 ± 39 days respectively, which means that the pulsar high energy behaviour evolved on a time-scale of about 10 days and that the effect lasted up to a few months. According to the fit, the maximum flux was reached on November 16 1975, that is about one month after a large discontinuity in the pulsar rotational parameters, a so-called glitch. The glitch occurred sometime between September 25 and October 15 (Manchester et al., 1976) while the gamma ray observations started on October 20. Although, the exponential function starts rising at the epoch of the glitch, one has to keep in mind that the exponential model is rather arbitrary and has no a priori physical support. The decay time of the gamma ray perturbation differs significantly from the rotational one, which was found to be 640 days (Kanbach et al,, 1980). This difference is not surprising

16 . 224 .

since the gamma ray decay corresponds to a magnetosphere-reaction time scale while the rotational decay time is linked to the relaxation of the inner star phenomena. From an energy point of view, the relationship between the two events is not excluded. Assuming that the pulsar energy stockpile is the-kinetic energy of rotation, the spin down luminosity is of the order of 1037 erg/s (using a iypical moment of inertia of 1.4 - 2 1045 g cm2, Shapiro and Teukolsky (1983). The increase in luminosity caused by the glitch just after its occurrence amounts to 0.8 - 1% of this value (Manchester et al., 1976; Downs, 1981). On the other hand the power radiated through the 50 - 5000 MeV gamma rays during the " quiet" behaviour of the pulsar is also of the order of 1035 erg/s (assuming that Vela is 500 pc away and that it radiates over one pole), and the gamma ray outburst observed after the glitch corresponds to an additional output of approximately 0.4 1035 erg/s radiated by the pulsar at its maximum luminosity. These approximate values indicate that if the gamma ray flare is related to the glitch event, then the burst observed by COS-B in 1975 required initially an appreciable part of the increase in rotational energy release caused by the pulsar spin-up. However, the recurrence of such a phenomenon may be questioned. A further giant glitch occurred in July 1978 (Downs et al., 1978) when no COS-B data were taken from the source (see figure 2). According to the time scales quoted above, a coincident gamma ray flux increase could have happened without being noticed in later COS-B observations. What is surprising, is the high gamma ray flux recorded in April and May 1981 when no glitch was reported. Such an event occurred five months later, in October 1981 (Gorenstein et al., 1981), when Vela was not in the field-of-view. Thus, the gamma ray enhancement is recurrent but it does not necessarily follow a glitch event. For this reason, it is not possible to draw any firm conclusion from the COS-B data on the relationship between the gamma ray variability and the glitch phenomenon.

The Vela pulsed radio flux is not stable either. As shown in figure 12, three large irregu­ larities were recorded at 408 MHz between 1971 and August 1978 when the observations at the Molonglo Observatory were stopped (McAdam, 1981). Unfortunately, additional 408 MHz data are not available beyond this date, but further studies at 843 MHz seem to corroborate the long term radio variability (McAdam, private communication). These

17 . Z2 3 -

secular changes could not be explained by the interstellar scintillation, nor do they appear to be related to the glitches. Their rather regular aspect over the years suggested that such fluctuations are caused by a precession in the beam angle with respect to our line of sight (McAdam, 1981). Despite the minimal overlap of the radio and the gamma ray data, one could conclude that the two variation patterns are not strongly correlated. The small irregularity in the general radio decrease, in late 1975, seems more related to the third Vela glitch (the two eccentric radio points were recorded at the beginning and the end of October) than to the gamma ray (lux, which peaked more towards mid-November. Therefore, the precession effect proposed for the radio evolution is not supported by the gamma ray data, since in this case it would not explain the stability of the first peak while the other phase components all vary. It would also be difficult to account for the energy dependence of the variability (merely the 'steadiness' of the high energy part of all the dif­ ferent components). No other clues to the origin of the long-term variability can be found from the pulsed emission at other wavelengths. In the optical band, the 24'* magnitude Vela object is too faint to provide so far precise information on possible apparent mag­ nitude changes. Such information is very important and should become available soon. In the X-ray domain, the only long-term data are obtained with the Einstein satellite. Ten observations were made in the period from November 1978 to June 1980, during a quiet phase in the gamma ray activity but including the 1978 radio glitch. No pulsations were observed and the X-ray point source coincident with the pulsar position showed no significant variation in its flux.

Other radio pulsars have proved to be variable sources, but usually on a much shorter time-scale. In the Crab case, after a continuous weakening of the 74 MHz pulsed emission was observed between 1971 and 1975, its radio flux remained steady over the next six years (Rickett and Seiradakis, 1982), as did its optical and X-ray flux (Knight, 1982). Throughout the COS-B observations, spread over seven years, only a marginal change in the ratio of the two pulses has been observed (Clear et al., 1987). Hence, in the long term, the Vela and Crab pulsars do behave differently and the large variability observed from Vela, both in radio and gamma rays, looks unusual compared to the measured behaviour

18 . 214 of all other pulsars.

0.3 Discrete Components of the Lightcurve

The present results lead to the identification of at least five discrete components of the Vela pulsed gamma ray emission. These exhibit different spectral distributions at low energies and different long term evolutions. Of particular interest is the interpulse radiation which is found to be composed of two discrete phase components, with the first being approximately coincident in phase with the first optical peak. Such a phase coincidence is not clear in the case of interpulse-2 and gamma ray emission in phase with the second optical peak would largely disappear beneath the strong second gamma ray peak (see figure 1). Furthermore, while the two optical pulses exhibit a similar structure in phase and have similar luminosities and colours (Peterson et al., 1978), interpulse-1 shows a broad peak and is brighter than the flat interpulse-2 and they clearly have different spectral characteristics. The spectrum of interpulse-1 exhibits a pronounced break around 300 MeV and highly variable low energy emission. In contrast, the interpulse-2 emission seems to present a surprisingly "convex" energy distribution. Unfortunately, it is impossible to conclude on the reality of such a reversed shape because the discrepancy between the average low and high energy spectral indices does not reach a 2c significance level. It is striking, however, that the remarkably steep distribution at low energies is present during four independent observations (the interpulse-2 radiation during period 45 was too faint to be analysed). The high values of the interpulse-2 spectral ratios corroborates the extreme softness of its radiation. The SAS-2 data, at energies above 35 MeV, may also support this characteristic. In the lightcurve of Thompson et al. (1977) the flux from the two interpulse phase regions is about equal. This relative increase of the interpulse-2 component may be explained by the extreme softness of its spectrum as measured in the COS-B data. Unfortunately, it is not possible from the two-power-law representation that has been adopted here to know whether the steep low energy part starts flattening around 50 MeV. The relative fluxes of the interpulse-2 and the second peak as observed by SAS-2 and extrapolated from COS-B do not suggest such a bending. However, a sharp turn-over must exist around a few tens

19 - 22* _

of MeV, otherwise interpulse-2 would be clearly visible in the soft gamma ray lightcurve of Turner et al. (1984), see figure 1. Aa expected, the extrapolation of the harder interpulse-1 spectrum to medium gamma ray energies (1-30 MeV) disappears within the statistical fluctuations of the lightcurve. Further observations in this energy range leading to a precise lightcurve are of importance for the study of the interpulse emissions and their physical origins. Such observations will be provided by the COMPTEL experiment on-board the Gamma Ray Observatory which is due for launch in 1990.

At TeV energies the detection of pulsed gamma rays at phase 0.27 (Bhat et al., 1987) indicates that the interpulse-1 phase component, which is by far the most variable part, could also be related to the very high energy gamma ray activity of the pulsar. Further observations at these energies are also required to determine the complete gamma ray spectrum of the interpulse emission.

6.4 Implications for Vela Models

After the demonstration by Goldreich and Julian (1969) that a pulsar magnetosphere may be filled with plasma, various models have been proposed to explain how pulsed radiation can be produced at different wavelengths within such magnetospheres. Many agree on the existence of accelerating regions called "gaps", which arise from a local charge depletion and provide the primary population of ultra relativistic particles streaming along the magnetic field lines. However, the models disagree on the location of the active regions (e.g. polar gaps, outer gaps etc.), on the type of physical processes and on the sequence of processes which convert the primary particles into pulsed radiation. An important consequence of the new description of the Vela gamma ray behaviour is that the activity in its magnetosphere is much more complex than imagined by these models. The identification in the lightcurve of discrete high energy components with their own evolution implies that several active sites exist simultaneously in the magnetosphere, and the different spectral characteristics of the components indicate that the physical processes differ with location. It is beyond the scope of this article to test the numerous pulsar models against our results, However, as several discuss the high energy emission in more detail and provide useful tests to evaluate

20 . . lit -

interesting ideas, some of these shall be briefly considered.

Salvati (1986) has recently pointed out that secondary synchrotron emission would pro­ duce a spectral shape in the soft gamma ray domain which bears some information on the source location in the magnetosphere if the secondary particles originate in magnetic photoabsorption. The predicted synchrotron spectrum has a slope of -1.5 between two characteristic photon energies. This slope is consistent with the data reported by Turner et al. (1984) but the upper break is not visible in figure 10. The new COS-B spectrum suggests that the total gamma ray spectrum is a smooth curve, continuously steepening with increasing energy. Adopting the upper break energy at either 15 MeV (following Salvati) or 300 MeV (as suggested by the COS-B data), the test leads in both cases to a source location extremely close to the star surface and in complete contradiction with the models that are discussed below. Moreover, one should be very cautious comparing the soft and hard gamma ray data, since the spectral variability observed below 300 MeV may extend down to the soft gamma ray range. Although the main peaks are more stable than the interpuises, contemporary data at different energies would be preferable for future discussions on the details of the spectral shape.

A spectral break would not be predicted in the soft gamma ray range if the assumption of magnetic photoabsorption is not relevant in the Vela case, as mentioned by Cheng et al. (1986b). In the frame-work of their outer gap model, the primary photons and the secondary pairs are produced respectively by inverse Compton scattering and collsions with a tertiary population of soft synchrotron photons (mainly IR) which illuminate the outer magnetosphere. The predicted pulsed spectrum is in good agreement with figure 10. In the gamma ray domain, the spectral shape is in good agreement with the data of Turner et al, (1984) and with the time averaged COS-B spectrum up to 3 GeV where their model induces a break which is not seen in the COS-B data (see figure 3). Although Cheng et al. do not present any lightcurve, they calculate that their model can reproduce the different gamma ray and optical phase separations between the peaks, since the tertiary optical photons are emitted somewhat closer to the star (about half way) than the secondary gamma rays. However, the outer gap model cannot explain the existence of the two gamma ray interpulse

21 . 22? -

components. Because of their position, they should also originate from a part of the gap closer to the star than the section responsible for the main peaks. It is then difficult to conceive how the inner part of the same gap may be subject to a significant long term variability while the outer part remains comparatively stable. Furthermore, since in this framework the two main peaks are produced by two symmetrical outer gaps, it is difficult to explain why one is slightly more variably than its twin and why they produce different spectra (the possibility that the orientation of the pulsed emission highly depends on the energy would also contradict on statistical arguments the mere fact that we see the Vela and Crab pulsars in gamma rays). Finally, the spectral shape predicted by the outer gap model is not consistent with that of the two interpulse spectra. An alternative explanation is that active sites, other than the outer gaps, created along the last open field lines exist within the magnetosphere.

According to Cheng et al. (1986a), the main peaks come from two symmetrical gaps, one situated over each pole. In contrast, Morini (1983) ind Smith (1986) propose a geometrical scheme where the two main gamma ray pulses are produced in the two sets of terminal field lines above the same pole and close to the light cylinder. This scheme assumes that the photons are emitted along the field lines and that the dipole is orthogonal to the spin axis. As in the other models , no correction could be introduced to account for the influence of the particle flow on the geometry of the lines. Morini uses this geometry to show that the gamma rays from the second peak may be contaminated by some polar cap radiation (both are seen at the same phase), whereas the first pulse is due entirely to radiation from the outer-magnetosphere. Assuming that relativistic particles produce synchro-Compton emission near the light cylinder (for the first peak and part of the second) and gamma rays near the polar cap by inverse Compton scattering of the thermal X-rays from the hot surface (for the rest of the second peak), Morini calculates spectra for these peaks which are consistent with the COS-B results. However, as the majority of the second peak gamma rays come from the polar cap component they should practically disappear below 50 MeV. This is' not observed in the soft gamma ray lightcurve and therefore the mechanisms responsible for the high energy emission or their relative contributions require

22 - 229 .

revision.

This scheme has been refined significantly by Smith (1986). By following this purely geometrical approach some of the new aspects of the pulsar behaviour may be discussed. As already indicated, the difference in the spectra between the two gamma ray peaks may come from the phase reconstruction. Only those photons created near the leading terminal field lines close to the light cylinder contribute to the first peak. On the other hand, "outer" photons from the trailing side may combine with photons from the same field lines deeper in the magnetosphere and also with photons from the polar cap region to produce the second peak. Furthermore, the geometrical pattern indicates that the optical pulses are produced in the open magnetosphere above the pole, at a distance which is about half the light cylinder radius, and that the gamma ray interpuises originate along the same lines at a slightly higher altitude. The mixing in interpulse-2 of photons created well inside the magnetosphere and near the polar cap could again explain the spectral differences with interpulse-1.

Assuming that the gamma ray interpuises do originate in the open magnetosphere, the spectral shape of interpulse-1 has been compared with the predictions of various polar gap models (Daughterty and Harding, 1982; Heyvaerts and Signore, 1981; Ayasli, 1981). The lack of agreement between the predicted and measured shapes renders no clues to identify the mechanism responsible for the interpulse-1 radiation. Using the grid of spectra pre­ sented by Massaro and Salvati (1979) for various pulsar parameters, a good fit is obtained in the case of an initial Lorentz factor of 3 107 for the primary particles and a low surface magnetic field value of 3 1011 G. However, the energy output from their model is not consistent with the COS-B observations of the interpulse.

The recent study by Caraveo et al. (1987) of possible polarisation of the Vela gamma ray emission using COS-B data, also suggests that the peak and interpulse emissions originate in different parts of the magnetosphere. Their result, if confirmed, indicates a high degree of linear polarisation for the gamma rays comirig from the entire interpulse, while the effect is much weaker for the second peak and not observed in the first peak.

23 . 229 -

In the scheme proposed by Smith, a highly polarised gamma ray component deep in the open magnetosphere, created from synchrotron radiation or surviving selective absorption processes (Caraveo et al., 1987), would be consistent with the present observations.

The long-term variability of the gamma ray lightcurve reported in section 5 also implies different source locations for the highly variable interpulse emissions and for the relatively stable main peaks. It is unlikely that the physical conditions may be variable on a time- scale from weeks to months, or stable, in regions along the same field lines. In the model presented by Smith, the various evolutions observed for the different gamma ray compo­ nents would signify that the region sensitive to the variability is the open magnetosphere in the polar region, while the conditions along the last open field lines in the outer mag­ netosphere would be more steady. This situation may be tested with precise data on the evolution of the visual magnitude of the Vela pulsar. Alternatively, the inferred stability of the phenomena related to the outer gaps agrees with the steadiness of the Crab pulsar emission at all wavelengths. The synchronisation of its pulses from the optical range to gamma rays indicates that they are all emitted close to the light cylinder along the outer gap (Smith, 1986). For Vela, the details of the radio pulse profile and polarisation involve a source located close to the star polar cap (Krishnamohan and Downs, 1983). Therefore, the variability observed in radio supports the idea that the conditions in the inner magne­ tosphere may change in the long-term. The same conclusion is reached when considering that the largest contribution to the variability in the gamma ray domain comes from the low energy part, provided that in the cascade of photons the softer are produced deeper in the magnetosphere.

The qualitative discussion of the new aspects of the pulsed gamma ray emission in the geometrical model proposed by Smith suggests that the origin of the interpuises exists above one of the poles, deep in the magnetosphere or at medium latitudes, with the source of the peaks lies close to the co-rotating magnetosphere and to the light cylinder. However, further study is required into the possible co-existence of "polar" and "outer" gaps, and into the mechanisms which can explain the observed spectra. The model must also explain the non-existence of second pole radiation except in the radio range, the greater stability

24 . 230 .

of the "outer" gaps over the "polar" gaps, and the different evolutions of the gamma ray and radio emissions. Finally, any model which tries to describe the Vela pulsar must at the same time explain the basic similarity of the Crab and Vela lightcurves at high energies and the marked differences in their overall behaviour.

Conclusions

The identification of the different components in the Vela gamma ray lightcurve supports the idea that several source regions exist in different parts of the pulsar magnetosphere. The significant contrast observed in the characteristics of the peak and interpulse emissions suggests that there may be at least two types of sources in the magnetosphere according to their location. The different spectral characteristics of the components indicate that different generating processes exist in each source. The origin of the variability observed in the low energy part of the gamma ray emission on time-scales from weeks to months must still be explained. In particular, with the present data, it is not possible to absolutely associate this phenomenon with the occurrence of giant glitches. Future satellite exper­ iments, such as Gamma-1, Sigma and GRO, have lost their last gamma ray calibration candle, but will undoubtably provide important information on the spectral characteristics of the various phase components and on their evolution. For the present, we have gained important informalion on the fundamental processes which govern the magnetosphere of a young pulsar.

Acknowledgements: We are grateful to Drs R. Buccheri and B. Sacco for the communi­ cation of the Vela timing parameters before the COS-B database release and to Dr. P. Thaddeus for the use of the Columbia CO survey. We thank Drs R. Buccheri and F. Lebrun for useful discussions. J. Clear acknowledges receipt of a research fellowship from the European Space Agency. The Laboratory for Space Research Leiden is supported by NWO, the Netherlands Organisation for the Advancement of Scientific Research.

25 . 231 -

Observation Pointing Vela Start, End Relative Vela Exposure Period Direction (1,1 Aspect Dates Efficiency 50 MeV - 5 GeV Angle (d/m/y) (107 cm2 s)

2 264°,-3" 1° 20/10/75 08/11/75 1.02 2.44

3 263°, +4° 6° 08/11/75 28/11/75 1.03 2.06

5 292", 0° 29" 24/12/75 23/01/76 0.94 0.25

12 263°, +3° 6" 24/07/76 24/08/76 0.72 2.61

20 260°, +18" 21° 14/04/77 02/05/77 0.66 0.56

21 243", -2° 20" 02/05/77 08/06/77 0.68 1.26

40 243", -3" 20" 03/04/79 0.84 0.82 09/05/79

42 278", 0" 17" 20/06/79 27/07/79 0.72 0.73

45 263°, +4" 7" 10/10/79 0.73 2.25 15/11/79

59 269°, -10° 9° 09/04/81 0.53 1.80 03/06/81

Table 1: Characteristics of Vela observation.

26 231 .

Region Phase interval

Peak 1 0.07 - 0.15 Interpulse-1 0.15 - 0.33 Interpulse-2 0.33 - 0.47 Peak 2 0.47 - 0.58 Trailer 0.58 - 0.77 Background 0.77 - 0.07

Table 2: Definition of the selected phase intervals of the Vela pulsar lightcurve.

Energy Range Unpulsed/Pulsed (MeV) (a)

50 - 100 0.08 100 - 200 0.09 200 - 500 0.07 500 - 5000 0.06

Table 3: Upper limits to the ratio (a) of the unpulsed to pulsed emission for four energy ranges.

27 . 233 .

Gisimnoi FUI1 inumm IlTllllUll M» : TUllll IEQII 101SID si. mil: Ming mum rreb ol 13 PÏ1H WilUll

si.iitn ir l.t! t l.ll l.lltl.ll l.lltl.ll 1.71 S0.il l.tltt.Sl l.H 11.11 0.111 ! 11 2.11 ± 1.21 I.I tl.l 2.1 tl.) 1.1)11.11 1.) tl.i I.I!! 0.11 1)1 •I 1.11 t 1.1) l.li 1 1.2! 2.11 t 1.1! 1.1) t 1.11 1.11 1 i.l! 2.11 1 1.11 17 1

l.lltl.ll 2.1111.11 l.Sltl.li 1.11 !0.II 2.11 i l.ll l.lltl.ll III 1 l.tl t l.ll 2.11 t 1.2! 2.11 t 1.11 I.Il t 1.2! 1.1! t t.ll 2.11 t 1.11 lit 1.1) 11.1) 2.11 i 1.11 1.11 t I.1T 2.1! t 1.21 1.1 t 1.) 1.11 t l.ll 11 »

1.!! t 0.1) l.ll t l.ll l.il t l.ll l.ll t l.ll 1.1 t 1.1 l.ll t l.li II 1 11 l.il 11.21 1.!! 11.1! 2.! t 1.1 I.l! t l.li l.ll t l.li lit 2.1! i l.!l 1.1! t I.l! 2.1! t 1.)! 2.11 t 1.21 2.11 t l.ll I7.il

l.ll t l.ll l.i! t l.ll I.l! 1 l.ll 1.» t I.l! 1.1! 1 l.ll l.H t l.ll 111 II 1.1! t l.ll 1.15 t 1.51 l.ll t Ml 1.21 t t.il lit l.ll 11.1! l.ll 11.21 l.il t l.li Ml t 1.11 1.1 t I.l 2.11 t 0.11 111

1.1! 11.12 1.1111.11 1.1! 11.1! 1.1211.11 1.1! t I.l! l.ll t l.li 111 51 1.1! t l.ll 1.1111.» 1.)! 11.!! l.il 10.» l.ll 11.» 1.» t I.l! HI t.U t l.ll l.ll t 1.11 1.! t I.l 2.1! 11.21 2.11 t l.ll 111

HSU: l.lltl.ll 1.1! t l.li l.lltl.ll l.lltl.ll 2,2! 11,1! l.lltl.ll l.ll 1 lilt l.ll 11.11 l.ll t l.li 2.1! 1 l.ll l.il t l.ll 2.12 11.12 l.ll t l.ll III 1 limit 2.11 1 l.li l.ll t 1.12 1.11 11.22 2.11 t l.ll 2 1! t l.ll 2.11 1 1.11H I IIDR: pros 11 » 111 21 t 1! 1 11 1 l.ll 1 1.11 1 11 UN 11 1 II »! 1 1! 1 II 1 1.1 t l.i 1 wUlilltT 11 1 I) t 11 1 il t 12 1 1) 1 ill :isniAnn I1UI imiiïïisi I xi;?iisi: nu: mr.zt «KM ran» mii:»

E:II: ET 1.1 t I.l ll.l i I.l 15.1 1 l.t li.i t 1.! 11.1 1 2.1 21.! 1 1.1 ! II 11.5 11.! i.l t 2.! ll.l t I.l 11.2 t I.l 2.1 1 2.1 ll.l t 1.1 II 11.11 I.l 1.7 11.1 l.t 1 1.2 12.1 t t.ll 1.1 1 I.l! l.il t I.U

12.11 t.l 15.71 I.l 21.1 t 2.! ll.l 11.2 25.111.! 12.1 i 2.1 1 ll.l t I.l ll.l 11.7 11.1 t i.l U.I i 1.! ll.l 1 l.i 11.) t I.) ll.l t I.l 1.1 1 t.l I.l t I.l U.I i II 1.! 1 1.2 1.2 11.2

11.11 I.l ll.l t I.l l.i t 1.7 ll.l S l.i 1.1 1 I.l li.l 1 1.1 a ll.l 1 i.l IS.! 1 2.1 1.! t M 17.1 t I.l U.I 11.1 I.l 11.1 i.l 11.2 III t LIS ll.l t I.2S l.ll t Ml

11! 11.1 ll.l 1 1.1 2.1 t 1.! U.I t 2.S 1.1 t 1.7 ll.l 1 I.l il 11.117,1 1.2 ll.l 21.7 ll.l 1.1 tl.l ll.l t 1.2 S.l 11.1 I.» t I.l! 17.1 i 1.2! 1.1 I t.lS I.l t I.l

ll.l 1 1.1 12.1 t 1.2 21.1 1 1.1 11.1 t 1.1 I.l 1 1.1 ll.l t 1.1 » 51.1 t 1.2 ll.l t l.t ll.l 1 l.i ll.l t 1.1 1.1 1 1.1 12.1 1 1.1 11.2 t I.l I.l t I.l 1.1 1 1.1 U.I I t.l 1.! 1 t.l

HIS: 11.1 tl.l ll.it t.l ll.lt I.! ill 111 t.l t M ll.l tl.l till ll.l 1 1.1 21.2 t I.I ll.l t 2.2 ll.l t 2.1 1.2 t 1.1 li.l t 1.1 mn;i M.l 1 I.l 1.12 t I.K 2.2S 1 t.ll ll.l t I.l I.l t 1.1 l.M t l.H

nn: 21 III III II III 111 Q11 IB tlU 1 11 11 li 11 21 wlilllltT HI 111 111 111 III II

Table 4: Spectral index and flux values of the Vela pulsar emission for energy selec­ tions 50 - 5000 MeV (ET), 50 - 300 MeV (LE) and 300 - 5000 (HE), and for differ­ ent pulsar phase intervals (Table 2) and observation periods. Fluxes are given in units of 10~° ph ttn"1!"1 (phase unit)-1 and 10~° ph cm"V for the whole pulsed emission. Also given are probabilities (%) or \2 values testing for no variation with phase or time.

28 . 23q .

References

Albats,P., Frye, G.M.,Jr., Thomson, G.B., Hopper, V.D., Mace, O.B., Thomas, J. A., Staib, J.A.: 1974, Nature 251, 400.

Ayasli, S.: 1981, Astrophys. J. 249, 698.

Bennett, K., Bignami, G.F., Boella, G., Buccheri, R., Hermsen, W., Kanbach, G., Lichti, G.G., Masnou, J.L., Mayer-Hasselwander, H.A., Paul, J.A., Scarsi, L., Swanenberg, B.N., Taylor, B.G., Wills, R.D.: 1977, Astron. Astrophys. 61, 279.

Bhat, P.N., Gupta, S.K., Ramana Murthy, P.V., Sreekantan, B.V., Tonwar, S.C., Viswanath, P.R.: 1980, Astron. Astrophys. 81, L3.

Bhat, P.N., Gupta, S.K., Ramana Murthy, P.V., Sreekantan, B.V., Tonwar, S.C., Viswanath, P.R.: 1987, Astron. Astrophys. 178, 242.

Bloemen, J.B.G.M., Strong, A.W., Blitz, L., Cohen, R.S., Dame, T.M., Grabelsky, D.A., Hermsen, W., Lebrun, F., Mayer-Hasselwander, H.A., Thaddeus, P.: 1986, Astron. Astrophys. 154, 25.

Buccheri, R., V., Caraveo, P.A., D'Amico, N., Hermsen, W., Kanbach, G., Lichti, G.G., Masnou, J.L., Wills, R.D., Manchester, R.N., Newton, L.N.: 1978, Astron. Astrophys. 69, 141.

Buccheri, R., Bennett, K., Bignami, G.F., Bloemen, J.B.G.M., BoriakofF, V., Caraveo, P.A., Hermsen, W., Kanbach, G., Manchester, R.N., Masnou, J.L., Mayer-Hasselwander, H.A., Ozel, M.E., Paul, J.A., Sacco, B., Scarsi, L., Strong, A.W.: 1983, Astron. Astrophys. 128, 245.

Buccheri, R., Di Gesu, V., Maccarone, M.C., Sacco, B.: 1987, Proc. 20th Int. Cosmic Ray Conf., Moscow, U.S.S.R., 1, 70.

Caraveo, P.A., Bignami, G.F., Mitrofanov, I., Vacanti, G.: 1987, Astrophys. J. (In Press).

Cheng, K.S., Ho, C, Ruderman, M.: 1986(a), Astrophys. J. 300, 500.

Cheng, K.S., Ho, C, Ruderman, M.: 1986(b), Astrophys. J. 300, 522.

29 Clear, J., Bennett, K., Buccheri, R., Grenier, I.A., Hermsen, W., Mayer-Hasselwander, H.A., Sacco, B.: 1987, Astron. Astrophys. 174, 85.

Cruise, A.M., Newton, A.C.: 1973, Nature Phys. Sci. 244, 121.

Culhane, J.L., Cruise, A.M., Rapley, C.G., Hawkins, F.J.: 1976, Astrophys. J. Letters 190, L9.

Dame, T.M., Ungerechts, H., Cohen, R.S., DeGeus, E.J., Grenier, I.A., May, J., Murphy, D.C., Nyman, L.-A., Thaddeus, P.: 1987, Astrophys. J. (in press).

Daugherty, J.K., Harding, A.K.: 1982, Astrophys. J. 252, 337.

Downs, G.S.: 1981, Astrophys. J. 249, 687.

Downs, G.S., Manchester, R.N., Newton, L.M.,: 1978, IAU Circ. 3274.

Eadie, W.T., Drijard, D., James, F.E., Roos, M., Sadoulet, B.: 1977, Statistical Methods in Exper­ imental Physics, North Holland, Amsterdam.

Goldriech, P., Julian, W.H.: 1969, Astrophys. J. 157, 869.

Gorenstein, M.V., Shapiro, I.I., Rogers, A.E.E., Cohen, N.L.: 1981 IAU Circ. 3644.

Grindlay, J.E., Helmken, H.F., Hanbury-Brown, R., Davis, J., Allen, L.R.: 1975, Astrophys. J. 201, 82.

Gupta, S.K., Ramana Murthy, P.V., Sreekantan, B.V., Tonwar, S.C.. Viswanath, P.R.: 1982, Work­ shop on VHE Gamma-ray Astron, Ootacamund, India, Ed. Ramana Murthy and Weekes, 157.

Harnden, F.R. jr, Johnson, W.N. Ill, Haymes, R.C.: 1972, Astrophys. J. Letters 172, L91.

Harnden, F.R. jr, Grant, P.D, Seward, F.D., Kahn, S.M.: 1985, Astrophys. J. 299, 828.

Harnden, F.R. jr, Gorenstein, P.: 1973, Nature 241, 107.

Hermsen, W.: 1980, Ph.D. thesis, Univ. Leiden.

30 . £3« .

Heyvaerts, J., Signore, M.: 1981, Astron. Astrophys. 96, 36.

Kanbach, G., Bennett, K., Bignami, G.F., Buccheri, R., Caraveo, P., D'Amico, N., Hermsen, W., Lichti, G.G., Masnou, J.L., Mayer- Hasselwander, H.A., Paul, J.A., Sacco, B., Swanenburg, B.N., Wills, R.D.: 1980, Astron. Astrophys. 90, 163.

Kellogg et al.: 1973, Astrophys. J. 183, 935.

Knight, F.K.: 1982, Astrophys. J. 260, 538.

Knight, F.K., Matteson, J.L., Peterson, L.E., Rothschild, R.E.: 1982, Astrophys. J. 260, 553.

Krishnamohan, S., Downs, G.S.: 1983, Astrophys. J. 265, 372.

Large, M.I., Vaughan, A.E., Mills, B.Y.: 1968, Nature 220, 340.

Lichti, G.G., Buccheri, R., Caraveo, P., Gerardi, G., Hermsen, W., Kanbach, G., Masnou, J.L., Mayer-Hasselwander, H.A., Paul, J.A., Swanenburg, B.N., Wills, R.D.: 1980, Non-Solar Gaftma Rays, Advances in Space Exploration Vol. 7 p49 (eds Cowsik, R. and Wills, R.D. Pergamon, Oxford).

Manchester, R.N., Goss, W.M., Hamilton, P.A.: 1976, Nature 259, 291.

Manchester, R.N., Wallace, P.T., Peterson, B.A., Elliott, K.H.: 1980, Mon. Not. R. Astr. Soc. 190, 9P.

Massaro, E., Salvati, M.: 1979, Astron. Astrophys. 71, 51.

McAdam, W.B.: 1981, Proc. Astr. Soc. Austr. 4, 219.

Moore.W.E, Agrawal, P.C., Garmire, G.: 1974, Astrophys. J. Letters 189, L117.

Morini, M.: 1983, Mon. Not. R. Astr. Soc. 202, 495.

Mayer-Hasselwander, H.A.,: 1985, Explanatory Supplement to the COS-B Final Database.

31 . 2a? _

Peterson, B.A., Murdin, P.G., Wallace, P.T., Manchester, R.N., Penny, A.J., Jorden, A., King, D., Hartley, K.F.: 1978, Nature 276, 475.

Pravdo, S.H., Becker, R.H., Boldt, E.A., Holt, S.S., Rothschild, R.E., Serlemitsos, P.J., Swank, J.H.: 1976, Astrophys. J. Letters 208, L67.

Rappaport, S., Bradt, H., Doxsey, R., Levine, A., Spada, G.: 1974, Nature 251, 471.

Ricker, G.R., Gerassimenko, M., McClintock, J.E., Ryckman, S.G., Lewin, W.H.G. : 1973, Astro phys. J. Letters 186, Llll.

Rickett, B.J., Seiradakis, J.H.: 1982, Astrophys. J. 256, 612.

Salvati, M: 1986, Astron. Astrophys. 154, 379.

Scarsi, L., Bennett, K., Bignami, G.F., Boella, G., Buccheri, R., Hermsen, W., Koch, L., Mayer Hasselwander, H.A., Paul, J.A., Pfeffermann, E., Stiglitz, R., Swanenburg, B.N., Taylor, B.G. Wills, R.D.: 1977, Proc. 12th ESLAB Symposium, Frascati, ESA SP-124, p. 3.

Shapiro,S.L., Teukolsky,S.A.: 1983, Black Holes, White Dwarfs and Neutron Stars; The Physics o Compact Objects (John Wiley and Sons, New York).

Smith, F.G.: 1986, Mon. Not. R. Astr. Soc. 219, 729.

Strong, A.W., Bloemen, J.B.G.M., Lebrun, F., Hermsen, W., Mayer-Hasselwander, H.A., Buccher' R.: 1987, Astron. Astrophys. Suppl. Ser. 67, 283.

Strong, A.W., Bloemen, J.B.G.M., Dame, T.M., Grenier, I.A., Hermsen, W., Lebrun, F., Nymar L.-A.,Pollock, A.M.T., Thaddeus, P.: 1987, Proc. 20th Int. Cosmic Ray Conf., Moscow U.S.S.R., 1, 125.

Thompson, D.J., Fichtel, CE., Kniffen, D.A., Ogelman, H.B.: 1977, Astrophys. J. Letters 21- L17.

Turner, O.T., Dayton, B., Long, J., O'Neill, T., Zych, A., White, R.S.: 1984, Nature 310, 214.

32 . 23g .

Wallace, P.T., Peterson, B.A., Murdin, P.G., Danziger, I.J., Manchester, R.N., Lyne, A.G., Goss, W.M., Smith, F.G., Disney, M.J., Hartley, K.F., Jones, D.H.P., Wellgate, G.W.: 1977, Nature 266, 692

Wills, R.D., Bennett, K., Bignami, G.F., Buccheri, R., Caraveo, P.A., Hermsen, W., Kanbach, G., Masnou, J.L., Mayer-Hasselwander, H.A., Paul, J.A., Sacco, B.: 1981, Proc. 17th Int. Cosmic Ray Conf., Paris, France, XG 1.2-8.

Zimmermann, H.V.: 1980, Astron. Astrophys. 88, 309.

33 - 2Î3 -

Figure Captions

Figure 1: The lightcurve from PSR 0833-45 at radio (Buccheri et al, 1978), optical (Manch­ ester et al, 1980), soft energy gamma rays (Turner et al., 1984) and high energy gamma rays (this work). The arrows indicate the position of the phase boundaries which are used in this analysis.

Figure 2: Integrated gamma ray flux (50 - 5000 MeV) from the Vela pulsar at various epochs from 1975 to 1982. The flux values are calculated using the saturation method and an assumed source spectral index of -1.89.

Figure 3: Phase averaged differential gamma ray spectrum from the Vela pulsar using all COS-B observations. The dashed line represents the best fit to the spectrum at energies above 300 MeV.

Figure 4: Integrated gamma ray flux/intensity (I) values (50 - 5000 MeV) calculated using the likelihood analysis for 5 observations from 1975 to 1982. a) the pulsed gamma ray flux for PSR 0833-45; b) the diffuse galactic background and instrumental emission.

Figure 5: a) Vela pulsed gamma ray flux values integrated over 50 - 300 MeV (x) and 300 - 5000 MeV (o) for each observation period; b) phase averaged pulsed spectral index over the total energy range (50 - 5000 MeV): c) phase averaged pulsed spectral indices for high and low energy components.

Figure 6: Time averaged pulsed spectral index as a function of phase; a) one power law fit over total energy range; b) for high and low energy ranges.

Figure 7: Integrated gamma ray flux from each phase component as a function of epoch; a) for low energy range b) for high energy range (o peak 1, * interpulse-1, + interpulse-2, x peak 2).

34 . mo -

Figure 8: Ratio of low to high energy flux as a function of epoch; a) peak 1; b) peak 2; c) interpulse-1; d) interpuise-2.

Figure 9: Differential pulsed gamma ray spectra (50 - 5000 MeV) from Vela for 5 phase intervals and the total phase averaged emission. The power law spectra giving the maxi­ mum likelihood fits are shown for the 50 - 300 MeV and 300 - 5000 MeV energy ranges, respectively, for each observation; period 2 , period 3 — , period 12 , period 45 _ _, period 59 .

Figure 10: Total energy spectrum from the radio up to the very high energy gamma ray domain. The upper limits (3

Figure 11: Expanded view of the high energy gamma ray flux from the Vela pulsar. Observations periods 2 and 3 have been sub-divided into 10 day intervals to study the short term structure of the variability. The solid line represents two 'back to back' exponential functions fitted to the data. The dashed vertical lines represent the uncertainty in the epoch of a large discontinuity in the pulsar period (Manchester et al., 1976).

Figure 12: Comparison between the evolution of the radio flux (a) at 408 MHz (McAdam, 1981) and the high energy gamma ray flux (b) for the Vela pulsar. Gamma ray data points: * COS-B likelihood analysis; • COS-B saturation method; + SAS-2 (Thompson et al., 1977). The epochs of the pulsar glitches are indicated.

35 - Z.4 I -

GAHHA [S0-6000H*Vt

U2>-

: Î.S • ^-THryi.

0.2 •

JV^UT/"

RADIO I2295HHZ]

Figure 1.

36 . 24 Z -

1 ' 1— '- 1

» } " * * •ft 1 . 1 1976 1978 1980 1982 EPOCH

Figure 2.

37 - ZM3 -

IV \ '. \ • \ • -\ 1.*-!0 \ \ i r -•-\ '. -\ • \ ««-S -\ 10 - A- i \ \ -\ io-9 t- \ ; -v \ io-10 r \ : V\ i• \ • m-11 , , \ 10' 10' 10' 10" ENERGY (MeV) Figure 3.

38 - 2.4M -

©

©• 1076 1877 1071 1979 ISM 1SS1 Epoch Figure 4.

39 . 2ltf -

1S70 1977 1970 1979 1980 1991 Epoch

Figure 5.

40 . m 6

T 1 1 1 1 1 1 1 r • 50-500D HaV

0 I I 1 1 1 1 1 1 1- i' ® i i i i i i i .1 .2 .3 .4 .5 .8 .7 .9 .9 Pulsar Phas*

Figure 6.

41 - tm _

_ ! , , r V:;:::::.;

©

300-5000 M.V

1978 1977 1978 1978 1990 1991 Epoch

Figure 7.

42 . Z4t -

-i 1 1 r VU- T--+-1 ® H 1 1 I 1 h-

.t..t. * © û a •H 1 I 1 1 1 a.

VT © 1976 1977 1971 1979 19H 1991 Epoch

Figure 8.

43 n

6 amfiTj

0001 001 0001 001 U*1M|M pM|l* * £^ IKM «1 ®

H 1 1 H 1 1 H 1 V- H 1—I H 1 1 H 1 H ® ©

s,

H—I—I—H—I—I—H—I—V- 1—I—I—H—I—I—H—I- © ®

V-

' • ' ' • • ' U__

- Sh7 s*

I09 E NIE) l»ra cm"1 ••' k«V"'|

-1—1—1—1—1—1—1—1—1—1—1—1—1—1—r

\% 444

ifT TH E VEL / i

_I L ' I I I 1 1 I i_

4 4 4 4 jjniu î 1 » S S ! 'r

- - Il 4- 4- ^ !i!H! 1 " Î ' « =

- 017 z.n

1 • «ip •-y —^->-r-'-.-. . , • • i • i . • • •

11 11

I.JMV - il/ ^^—-^ if '"' £ 197S 1975 5 1Ï76 1976 5 1977 1977.5 1971 EPOCH

Figure 11.

• r* ~"7-~ I"™ ' -r 1 1 .*"•. * Radia 40t MHz >> D • r • - X 3 £ 6 I - *• •

3 1*

f , i 11 , t I t 1 1 i 1 ' ' 1 50-5000 HaV

E 30 u a. \ I' 2 2D 1 \ T> Flu x — I 'I, ® 1072 1873 1974 1975 1976 1977 1971 1979 1990 1991 Epoch

Figure 12.

46 - 2T2 -

Proc. XXth Int. Cosmic Ray Conf, Moscow, 1.

-1- OG 2.1-9 HIGH ENERGI GAMMA RAÏS FROH THE VELA PULSAR: EMEHGÏ DISTRIBUTION AMD FLUX VARIABILIS!!

IA. GRENIER', W. HERMSEN' and J. CLEAR1

1- SERVICE D'ASTROPHYSIQUE. CENTRE D'ETUDES NUCLEAIRES DE SACLAY. FRANCE 2- LABORATORY FOR SPACE RESEARCH LEIDEN, LEIDEN, THE NETHERLANDS 3- SPACE SCIENCE DEPARTMENT OF E.SJ*.. ESTEC, NOORDWIJK. THE NETHERLANDS Abstract New results on the temporal and spectral characteristics of the high-energy (50-5000 Mev) T-ray emission from the Vela pulsar, using all available COS-B data on this source, are presented: the total pulsed flux exhibits long-term time variability, the degree of variability being different for the emission from different phase intervals ; the energy spectrum of the whole pulsed emission cannot be described with a single power law; the spectral shape depends on the pulsar phase and shows long-term variability as well.

1. Introduction. The Vela and the Crab pulsars (PSR0833-45 and PSR0531+21) are thn only radio pulsars detected so far in the high-energy f-ray (50-5000 Mev) domain. While the emission from the Crab has been studied in detail over the entire electromagnetic spectrum, only limited information exists on the Vela pulsar. Pulsed emission from this source has been detected in the radio, optical and the total T-ray domain (> 1 HeV). In the 100 HeV region, where the pulsar reaches its maximum luminosity, all available data has not been fully exploited. The numerous Vela observations in the final COS-B database have been analysed to derive the PSR0S33-45 timing parameters (Kanbach et al.1980 alias K80, Mayer-Hasselwander 1985), but only ~ 20% of this data has 66 jn used to study its spectral behaviour (Bennett et al. 1977, K80) and long-tern time variability. However, detailed information on the spectral shape and time variability is required to constrain the theoretical modeling of the source.

2. The Method. A maximum likelihood analysis has been applied to the data to estimate the flux and spectral properties of the source. The developed new approach utilizes all COS-B information, that is the arrival time, energy and direction of each photon. It takes simultaneously into account tne limited spatial and energy resolutions of the instrument, and their dependency on the inclination and energy of an event. Finally, it separates the pulsar emission from any background radiation by exploiting the pulsar lightcurve shape and its spatial structure (the COS-B point-spread-function) against the structure of the underlying diffuse emission originating in the interstellar medium and the instrumental noise isotropy. The statistical processing uses a probability density function of detecting a photon at an Em, position lm, bm and phase ?m (where m is for "measured" against "true") expressed as:

N(?,E,l,b) describes the true celestial ^-ray distribution. n{ ,E,l,6) dE dO CM,(Em,E,«Blb,aiS) CAL represents the COS-B f convolution, given by the lb product of its effective I 11 Nuiwrator OEm oOm sensitive area, point-spread- JaaJJuusm function and energy resolution. The»» parameters were determined from the pre-launch calibration data (Hermsen 1980, Mayer-Hasselwander 1985). 91b represents the true inclination angle with respect to the detector axis of the pnotons coming from the peint source at (l,b) . 4f3 -

-2- OS 2.1-9 and reconstructed at an angle amlb away from this source, il denotes solid angle. The likelihood value can then be computed from the product of the probabilities of finding the photons actually observed by cos-B.

Th« modal to test may include up to four major.components pulsed component from Vela: PVP(o). E9f1""!'t' S(l-lv, b-bv) 3S » a steady component from Vela: TVS. E . $(l-lvr b-bv) « the steady diffuse galactic emission in the field of view around Vela, which has the spatial structure of the interstellar medium as traced by HI and CO (see Bloeraen et al.1966, Strong et al.1987 + this conference): g. NH(l,b). E 99 » a constant and isotropic instrumental background: BKI. E9

The analysis The gorjd timing resolution of the instrument allows an accurate determination of tne pulsar ligntcurve. In figure l is presented the lightcurve from all observations (31S days) using the pulsar timing parameters given with the final COS-B database. For the study of the pulsar behaviour as a function of phase, six phase intervals have been selected, similar to those used by K80: PEAK l OD? - a.is INTERPULSE J 0.15 - 033 fig. 1: « INTERPULSE 2 033 - 0.47 - PEAK 2 0.47 - 0S8 ~Z"' TRAILER OSS - 0.77 | BACKGROUND 077 - 0J>7 ""!

A study of the galactic longitudinal profiles of the Vela region, selecting events of the background phase part, confirms that any possible non-pulsed emission contributes less than a few percents of the total source emission, and may thus be neglected from the study. The likelihood analysis of these background events therefore gives the best spectral and flux parameters of the total background radiation (BKI, gi, g, gg) which, in return, are used to study the pulsed Vela emission in the other 'five phase intervals independently. Changing this set of parameters within the 1» level around its most likely value , ., ,s NUSE U has been checked not to affect tne results of the pulsar analysis within the quoted errors. Finally, as boundary condition, tne total number of predicted photons from a selected model is assumed to equal the total number of photons detected within the entire field of view (30°) and energy range (50-5000 HeV) of COS-B.

3. Result». Integrating the 5 observations listed in table i, which viewed the source within 10° of the pointing direction, we first found that the iingle-power-law spectrum which best describes the whole pulsed emission in the range 50-5000 MeV has an index of 1.84 t 0.03. This time averaged value is in very good agreement with that derived by K80 (1.89 ± 0.06, using only the first two short observations of vela). In the latter paper, one could see an indication of a spectrum flattening below 300 HeV (index 1.77 ± 0.15) and steepening above 300 HeV (index 2.00 ± 0.20). The possible existence of such a break was first noticed by Bennett et al.(1977). Exploiting our extended database, we repeated our analysis in two separate energy ranges. We conclude now that tne total tine averaged Vela spectrum cannot be described by a single power law because the spectral indices are significantly different: 1.72 ± o.07 in the 50-300 HeV range and 2.12 ± 0.07 for 300-5000 H*V. These values nicely match the 1.6 ± 0.2 index . ZfH .

-3- OG 2.1-9 derived by Turner et al. 1984 between 1 and 30 HeV and the 2.2 slope for 2000-5000 GeV reported by Gupta et al.1982. To study in detail the Vela pulsed e-ission, we derived the spectral slopes gp(ip) and intensities FVP(q>) within.the S selected phase regions, for the s listed observations and in 3 energy ranges (50-300, 300-5000 and 50-5000 MeV) separately. A full report on the analysis and the results will be the the subject of a forthcoming article (Grenier et al.1987); we will highlight here some of the most interesting findings. Figure shows the first 1 Vili: lotil 50-5000 HeV rim striking result: the total Vela " 3S •j-ray flux has been highly D a© - variable between 1975 and 1981. • 1 9 25 (The values have been derived C - fitting the data below and i ^a above SOOMeV separately). Such i - f 15 - a phenomenon cannot be ascribed • f 1 co an instrumental effect since I •-.he simultaneously derived 1C - intensity of the galactic 5 - emission appear stable within the statistical uncertainties. I97B >B" ,9'3 l9,s 1S0° ,SB1 The Vela f source was discovered by the SAS-2 team in 1973 (Thompson et al.1977). Extrapolating their measurement, assuming a 1.84 spectral slope, gives a 50-5000 HeV flux of 22 10-' cm-1 s-1 close to our average.

OSS EPOCH PEAK 1 INTERP.l INTERP2 PEAK 2 TRAILER

02 20.10 08.11.7S 82.3*83 13JO*2S 17.1*4.0 523*4.0 11.6*22 03 08.11 28.11.75 84.4*8.1 472*4.7 355*6X1 71.2*65 25.1*35 12 24J17 24JI8.78 66.4*8.7 22.1*2.6 72*2.6 485*4.6 07*0.4 45 09.10 15.11.79 60.8*7JS 11.4*3.1 23*05 417*5.4 1.4*02 59 09J04 03'JOE.81 69.8*92 36.4*5.0 35.2*5.6 64.7*6.7 8.7*2.6

Tao.l: Vela 50-5000KeV fluxes (10-' f cm-1 s-1 q>-') for each period and phase. A second striking result is shown in table 1. The long -erm flux variability qualitatively affects each phase similarly. However, its relative amplitude appears to be significantly larger for the inferpulse emission. Note in particular the jump in flux between the first two observations. The latter high luminosity is apparently reproduced five years later. K80 used observations 2, 3 and 12 to show qualitatively that the second peak and the combined interpulse emissions exhibit harder spectra than the first peak radiation. We merged our five periods to show quantitatively that the Vela energy distribution depends on pulsar phase. Assuming single power law spectra over the full energy range, the probability that the spectral indices in the five phase intervals are identical is 10 -•. Fig.3a and b display the variation of the index with phase, fitting respectively the entire energy range and the intervals below and above 300Mev with single power laws:

All Obftrvitioni 50-5000 Ml» All Oburvatloni 0. 3C0-500O HJ'( «. 50-300 Htï

PHASE 57!o,2 o, t o. « a. s 575 57?oTi a.o PHASE °. I 0-2 0.3 o. 4 o. 9 o. s o, 7 - zrr .

-4- OG 2.1-9 Finally/ we studied the long-term evolution of the spectral index: of the whole pulsed emission (fig.4) and of the radiation from each selected phase. Fitting the full energy range with one power law (fig.4a), the chance probability of the detected spectral variability is 2 10-'. Again, the simultaneous estimation of the background spectra indicates that this effect is not due to COS-B. Fig.4b shows the variability for the differential energy ranges, emphasizing the systematic difference in spectral slope below and above 300 HeV. Besides, this variability affects the emission from each phase interval similarly (i.e. an overall steepening of the spectra during the high emission state) But the interpulse emissions appear most variable (see Grenier et al.1987).

3. Conclution. While these results give a new, lively and more complex picture of the Vela pulsar, T-ray astronomy has lost its last calibration candle. But they give a handle to study in detail the physical processes responsible for the f radiation which undoubtedly differ with the location (in latitude and altitude) in the pulsar magnetosphere. of particular interest seems to be the strong variability of the interpulse emission, of which the first component approximately coincides in phase with the first optical peak.

4. Aknowl«dg«n«nt«. We are grateful to Drs R. Buccheri and B. Sacco for the communication of the Vela timing parameters before the COS-B database release and to Dr. P. Thaddeus for the use of the Columbia CO survey. J. Clear aknowledges receipt of a research fellowship from the European space Agency.

References Bannatt at si. . 1977, Astron. & Astrop., 61, 279 Grenier at al. 1987, Astron. & Astrop., in preparation Banaian, i960, Ph. D. Thesis, University of Leiden JCanUach at al. 1980, Astron. & Astrop., 90, 163 Hayar-BMMlMandar, 1985, Explanatory Supplement to the COS-B database Strong at al., 1987, Astron. & Astrop., in preparation Bloaaan at al., 1986, Astron. & Astrop., 154, 25 Thompion at al., 1977, Astrop. j., 214, u.7 Tuaar at al., 19B4, Nature, 310, 214 Gupta at al., 1982, Proc. of the Int. Workshop on VHE T-ray Astronomy, ootacamund, India, 157. - art -

WonJuhop on Huclexui Sptafioicoptf oi A&tAophyiicat SOUAC&A, Wa&tUngton, 19t1

HIGH ENERGY GAMMA RAYS FROM THE VELA PULSAR: LONG-TERM VARIABILITY AND ENERGY DISTRIBUTION

I.A. Grenier!, W. Hermsen2, and J. Clear3 1 Service d'Astrophysique, Centre d'Etudes Nucléaires de Saclay, France. 2 Laboratory for Space Research Leiden, Leiden, The Netherlands. 3 Space Science Department of ESA, Noordwijk, The Netherlands..

ABSTRACT

New results on the temporal and spectral characteristics of the high energy (SO MeV to 5 GeV) gamma ray emission from the Vela pulsar are presented. A new sensitive analysis method using maximum likelihood techniques has been applied to the COS-B data. The whole pulsed flux is found to exhibit long term variability. The data show strong evidence for a spectral break at approximately 300 MeV in the energy spectrum of this total emission. Five discrete emission regions within the pulsar lightcurve have been identified, with the spectral characteristics and long term behaviour being different. These results support the idea that various production regions simultaneously exist in the pulsar magnetosphere and that the physical processes generating the gamma rays differ with location.

INTRODUCTION

The study of the high energy gamma ray emission from celestial objects provides an excellent method in helping our understanding of the physical processes which occur within such sources. Tiiaer et al. (1984) at 0.3 - 30 MeV, Thompson et al. (1977) above 35 MeV and Bennett et al. (1977) at 50 MeV - 5 GeV all report a similar Vela lighcurve structure which is characterised by two sharp peaks separated by 0.42 in phase and bracketing the optical pulses. The energy spectrum of the pulsar still remains uncertain. An extrapolation to X-ray energies of the COS-B results of Kanbach et al. (1980) lies several orders of magnitude above the reported upper limits, while at TeV energies the flux values reportée by Bhat et al (1980) clearly require a break in the spectrum. Earlier results from COS-B data by Bennett et al. (1977) and Kanbach et al. (1980) indicate a possible break in the spectrum at a few hundred MeV, but the level of significance of these results is low. This article presents the latest results on the Vela pulsar in the 50 MeV to 5 GeV energy range. The data have been collected using the COS-B satellite which observed the Vela region during 10 observation periods between October 1975 and June 1981. The total COS-B gamma ray lightcurve together with the radio and optical ones are given in figure 1.

THE MAXIMUM LIKELIHOOD METHOD

A maximum likelihood analysis has been applied to the data to estimate the flux and spectral properties of the source using a - Ztl -

TABLE I Definition of the selected T • 'ft' A lil-sSMK»VI phase intervals of the Vela pulsar lighteurve.

Region Phase interval

Pea* 1 0.07 - 0.15 Interpulse-1 0.15 - 0.33 / ^ V*^ ; Interpulse-2 0.33 - 0.47 Peak 2 0.47 - 0.58 J Trailer 0.58 - 0.77 Background 0.77 - 0.07 \ i T Qrr,CAt-

UDtO I2MSI.HH Fig. 1. Vela lightcurve». testmodel, which may include up to four major components: 1. the pulsed emission from the pulsar; 2. a possible steady component from Vela; 3. a constant and isotropic instrumental background; 4. the steady diffuse galactic emission around the pulsar. The latter term results from the interaction of cosmic rays with interstellar matter. Its spatial structure can be traced by the HI and CO surveys (Dame et al., 1987; Strong et al., 1987), For further details on the method, see Crenier et al., 1988. Table I gives the details of the phase domains which have been selected for the analysis. The background region (phase 0.77 - 0.07) haa been analysed for a possible steady emission from Vela. The resultant upper limit shows that such emission accounts for less than a few percent of the pulsed luminosity, and may thus be negelected from the model. For those observations with large source aspect angles (>15°), where the efficiency and the energy resolution of the instrument are poor, this likelihood analysis is limited by the available statistics and has not been performed. For the remaining observations a detailed analysis has been made for the 5 phase intervals. Within each phase interval the analysis has been performed for 3 energy intervals (50 - 5000 MeV, 50 - 300 MeV and 300 - 5000 MeV). The division of the data into these energy intervals is based on the evidence for a break in the pulsed spectrum at approximately 300 MeV found in the analysis of Kanbach et al. (1980) using only 2 observations. Due to computing time considerations the position of the spectral break has not been used as an additional free parameter within the model, however the results of the independent modelling of the spectrum above and below 300 MeV strongly support this choice of value.

RESULTS

The likelihood analysis has been performed for the 5 selected phase intervals, for the 5 best observation periods and for 3 energy - Ztl -

-1 1 1 I

Fig. 2. Integrated flux/intensity (X) values (50-5000 HeV) for 5 obaervationa: a) the pulsed Vela flux; b) the diffuse galactic background and instrumental emission.

ranges, giving a total of 75 spectra and flux values. Only the main results will be highlighted (for detailed results see Grenier et al). The spectrum of the whole emission derived independently st energies above and below 300 HeV and using all obaervations is: F(S0 < E < 300) = 2.74 ±0.21 x 10"*.2"^"""photons/cm» s MeV

s ±,,0T F(300 < E < 5000) = 2.71 ±0.04 x 10- .£-Jv" photons/cm* s MeV The significant difference between the high and low energy spectral indices clearly shows that the time averaged Vela pulsed spectrum from 50 to 5000 HeV cannot be described by a single power law. At least two are required and in this representation the good connection between the two apectra at 300 HeV atrongly supports the choice of 300 HeV at the energy of the break. The flux of the pulsed emission from Vela has been determined by integrating .he two-power-law phase averaged spectrum for each observation. The result, displayed in figure 2a, indicates the striking long-term variability of Vela. In addition, the likelihood analysis provides simultaneously the background intensity for each observation, and its stable behaviour from one observation to the next, displayed in figure 2b, confirma the true source origin of the pulsed variability phenomenon. To study the energy dependence of this variability, figure 3a shows the evolution of the pulsed flux Eor the energy ranges 50 - 300 HeV and 300 - 5000 HeV. The main contribution to the flux variability is clearly due to the low energy emission. The different evolutions of tha flux from the two energy rangea denote a diatortion of the Vela spectrum from one epoch to the next. The observed fluctuation in tha spectral ratios (the ratio of low to high energy flux) has a probability of 10~10 of being due to a random affect. Figures 3b and 3c show the spectral indices aa a function of epoch for the 3 energy ranges. As expected, the spectral index for the entire energy range is seen to vary with time with a probability of 2 10~3 that thia effect is due to chance, whereas simultaneous analysis of the . «S -

background spectra shows no indication for variability. Figure 4a shows the time averaged spectral index as a function of phase. The probability that the emission may be best represented by a homogeneous spectrum over all phases (E~l*o4) is 10~*. A further understanding of the phase dependency of the pulsed emission may be obtained from figure 4b, which shows the spectral characteristics of the high and low energy emissions as a function of phase. Analysis of the spectral characteristics of the high energy emission shows that there is no significant variation in its spectral index with phase, giving a phase averaged value of 2.12 * 0.07. Thus, the spectral index of the GeV radiation from Vela displays both a stable and homogeneous behaviour. In contrast, the spectral index of the low energy emission shows evidence for phase dependence (chance probability 5 10~3). It is evident from the change in the spectral index at low energies (from che first peak to interpulse-2, 4.5o, and interpulse-2 to the second peak, 3.6o and from the softness of the first peak and trailer over the entire energy range, that 5 separate phase components exist in the Vela gamma ray lightcurve. The evolution of the flux from each of the phase components for each observation and for both high and low energy ranges has been studied.The long-tern variability is obvious and the relative amplitude of the variability is significantly larger for the two

H • to t Z0 » •

* » t •

Î'i i ' ©• 2. 1 I I i ' i © ? ' i } t ' a 50-900 *mi* *'•' \ ® '1 ! \ 1I r* t • • , 1,• © © in* itrc itrt in .1 .Z .J .* .3 .• .7 .% .• Ptri*«r Ph«i» Fig. 3.a) Vela pulsed flux: Fig. 4. Time averaged spectral 50-300 MeV(X),300-5000 HeV(O); index as a function of phase; b) phase averaged pulsed spectral one power-law fit ovsr a) total index (50-5000 M«V);c) as b), for energy range, b) high and low 50-300 MeV(X) and 300-5000 MeV(O). energy ranges. . ZtO -

Fig. 5. Differential pulsed gamma-ray spectra from Vela for S phase intervals and the total phase averaged emission for each observation period. The power law spectra giving the max. likelihood fits for the 50-300 MeV and 300-5000 MeV energy ranges are shown. Period 2 ... Period 3 Period 12 a- I* Period AS Period 59

s \ interpulse emissions. i » ' Furthermore, the flux variations in the separate I- energy intervals are different -I—I—H—h—I—H—I—•— I I II I—I II I I resulting in spectral distortions (probability levels are given in Grenier et al. 198B). To illustrate their long term evolution, the energy spectra giving the maximum likelihood fits to the data for each phaae component •r.J for each observation are displayed in figure 5. Good agreement is obaerved between the low and high energy fits at approximately 300 HeV for most phase components. The two power law representation is therefore a good approximation to the real spectra (for the trailer, the poor statistics make interpretation of the results rather difficult). The actual position of the break or bend in the real spectrum will very somewhat in energy following the apparent long term variability. Total PSR0833-45 Spectrum: The two-power-law fit time averaged spectrum of the whole pulsed emission haa been plotted in figure 6 together with the data recorded at other wavelengths in order to show the full energy distribution of the Vela pulsed radiation over the entire electromagnetic spectrum. Over this spectrum, the pulsar reachas its maximum luminosity in the MeV-GeV domain. At very high energies, the recent results of Bhat et al. (1987) indicate a spectral slope of -3.5 at TeV energies which does not require an excessive steepening when extrapolating over three decades of energy from the C0S-B gamma ray range. The observed time variability presented in this article may support the proposed time variability at TeV energies, e.g. Bhat et al. (1980), however, it should be noted that the variations recorded above a few hundred HeV are relatively small ( Î25Z). . Zfcl .

T 1—i r 1—T 1 1 1 1 1 1 1—i—i 1 1 r

log E |.V| Fig. 6. Total energy epectrua from the radio up to the very high energy gamma ray demain. The upper limita (3a) diaplayed in the X-ray range are calculated for an assumed duty cycle of 0.2 similar to the optical one.

CONCLUSIONS

The identification of the different components in the Vela gamma ray lightcurve supporta the idea that several source regions exist in different parts of the pulsar magnetosphere. The significant contrast observed in the characteristics of the peak and interpulse emissions evidently suggests that there may be at least two types of sources in the magnetosphere according to their location. The different spectral characteristics of the components indicate that different generating proceases exiat in each source. The origin of the variability observed in the low energy part of the gamma ray emission on time-scales from weeks to months muse still be explained. In particular it ia not possible to absolutely associate this phenomenon with the occurrence of giant glitches (Grenier et al. 1988).

REFERENCES

Bennett, K. et al.: 1977, Astron. Astrophys. 61,279. Bhat, P.N. et al.: 1980, Astron. .Astrophys. 81,1,3. Bh*-, P.N. et al.: 1987, Astron. Astrophys. 178,242. Dama, T.M. at al.: 1987, Astrophys. J. 322,706. Grenier, I.A., nermaen, W., Clear, J.: 1988, Astron. Astrophys. (in press). Kanbach, G. et al.: 1980, Astron. Astrophys. 90,163. Strong, A.M. at al.: 1987, Proc. 20th ICRC, Moscow, 1,125. Thompson, D.J. at al.: 1977, Astrophys. J. Letters 214,L17. Turner, O.T. et al.: 1984, Nature 310,214. - VI -

PERSPECTIVES

" TouAx. étoile, poitz e« i€\iz an 6oltU dam le. coeusi " AndKi VtAdeX . 262 .

J'ai commencé cet ouvrage en essayant de répondre à l'étonnement que j'ai

souvent provoqué en parlant d'exploiter encore aujourd'hui les données de COS-B.

Pourtant les résultats présentés ici sur la mesure des émissivités gamma du gaz dans le disque galactique et dans le milieu proche, la calibration du rapport

NH(2)/WCO, la nouvelle source gamma brillante, la variabilité et la complexité de l'émission à haute énergie des pulsars du Crabe et de Vela justifient pleinement une telle entreprise. Ces résultats nous ouvrent aussi de nouvelles perspectives. L'analyse spectrale des données est loin d'être achevée. Geoùnga est un objectif important, L'analyse fine de son spectre et de son évolution à long terme pourrait apporter de nouveaux indices sur sa nature et des contraintes aux divers modèles en cours de développement. L'émission diffuse est un second objectif d'importance. La nouvelle méthode d'analyse spectrale est toute indiquée pour parvenir à séparer les spectres des composantes

"bremstrahlung" et "n°" de l'émission diffuse et pour remonter ainsi aux spectres des rayons cosmiques de basse énergie. Toujours dans le domaine de l'émission gamma diffuse, les deux études présentées dans cet ouvrage ont montré qu'il fallait procéder à une analyse systématique de l'émissivité à basse

énergie des nuages moléculaires. Les complexes proches apparaissent bien sûr en tête de liste puisque leur vaste étendue angulaire permet de réduire les effets de la mauvaise résolution du télescope gamma. On pourra alors décider de la réalité de l'anomalie observée dans Céphée et, si tel est le cas. rechercher l'origine de ces variations d'émissivité du gaz moléculaire à toins de 150 MeV,

Enfin la détection de la nouvelle source gamma dans Céphée (surnommée familièrement Dragam) soulève le problème de la recherche de sources à basse

énergie. Elle n'a pas encore été tentée au dessous de 300 MeV à cause des performances de COS-B. Mais descendre en énergie serait souhaitable pour capter d'éventuelles sources au spectre très mou comme Dragam, COS-B a donc encore - ta .

beaucoup à offrir dans un bref délai.

D'autres questions soulevées dans cette thèse devront par contre attendre

le lancement des prochains télescopes gamma. J'espère, par exemple, que Gamma-I

et Sigma (prévus pour la fin 1968) pourront confirmer l'existence de Dragam,

mesurer et prolonger son étonnant spectre et cerner sa position avec une

précision suffisante pour pouvoir l'identifier. S'il s'agit bien de la

radiogalaxie. pourquoi celle-ci émet-elle autant d'énergie? Je compte également sur ces deux télescopes, ainsi que sur GRO (prévu pour 1990) pour poursuivre l'analyse des pulsars du Crabe et de Vela, ou de tout autre qu'ils pourraient découvrir. Etablir la cause de leur variabilité me paraît primordiale pour comprendre le fonctionnement des jeunes pulsars. Il serait donc essentiel d'observer Vela en gamma lors d'un "glitch" et de le surveiller régulièrement à plusieurs longueurs d'onde simultanément (de la radio aux gamma) car chacune révèle probablement une région différente de sa magnétosphère.

Les observations millimétriques présentées dans ce mémoire nécessitent

également quelques compléments et éclaircissements. J'aimerais, par exemple, connaître un jour la position relative de Céphée-Cassiopée et de la ceinture de

Gould. La continuité observée en position et en vitesse entre ce complexe et les nuages de la ceinture dans le premier quadrant n'est-elle qu'une coïncidence ou bien la ceinture est-elle différente des modèles actuels? Un léger gauchissement, une faible inclinaison de son petit axe ou un lent mouvement de va-et-vient pourraient réconcilier les observations, d'autant que la position de la ceinture est fort mal connue dans cette direction du ciel pauvre en étoiles 0 et B.

D'autre part, la carte CO du bras extérieur de la Galaxie est tout-à-fait sommaire, Des observations beaucoup plus sensibles et précises sont indispensables dans cette région, ainsi que dans le reste du premier et second quadrant afin de rechercher l'étendue complète du bras et d'étudier le - 1*4 .

confinement et l'état physique des nuages qui peuplent ces contrées lointaines.

Ce projet devrait aboutir bientôt.

Enfin, c'est avec plaisir que je continuerai l'étude de la bulle de Céphée.

Des observations sont à nouveau nécessaires, en X afin de contraindre davantage le modèle du reste de supernova et de préciser les conditions physiques qui y régnent, et en CO pour étudier le comportement du nuage de Céphée sous l'action du choc. Une étude infra-rouge de cette région "choquée" est également en projet. Aura-t-on jamais la chance d'y apercevoir la formation d'étoiles tant souhaitée des théoriciens?

En attendant, j'aurais plaisir à participer à la suite de l'aventure de l'astronomie gamma. Enfant, elle nous a tendu bien des pièges et réservé bien des surprises. Adolescente, elle promet déjà beaucoup. Souhaitons donc beaucoup d'imagination aux astronomes! - zir -

IMAGINONS ...

Imaginons des fantômes, des dieux et des démons. Imaginons des enfers et des paradis, des villes flottant dans les cieux et des villes englouties sous la mer. Licornes et centaures, sorcières et magiciens, djinns et farfadets. Anges et harpies, charmes et incantations. esprits élémentaires, familiers, démoniaques. C'est facile à imaginer tout cela: depuis des millénaires, les hommes 1'imaginent. Imaginez des astronefs et l'avenir. C'est facile à imaginer: l'avenir approche vraiment et il sera peuplé d'astronefs. Sf a-t-il quelquechose qui soit difficile à imaginer? Oui, bien sûr.

Imaginez un peu de matière, avec vous enfermé dedans, vous qui avez conscience d'exister, qui pensez et savez donc que vous existez, vous qui êtes capable de faire remuer la matière dans laquelle vous êtes, de la faire dormir et s'éveiller, de lui faire l'amour et monter les côtes.

Imaginez un univers - infini ou non, comme il vous plaira de vous le figurer - avec un milliard de milliards de milliards de soleils pour le constituer.

Imaginez une boule de boue qui tourne comme une folle autour d'un de ces soleils.

Imaginez-vous debout sur cette boule de boue, tournant avec elle tournant dans le temps et 1'espace vers une destination inconnue

Imaginez-le,

F. Brown A cause de la complexité des analyses en astronomie Y. treize ans après le lancement de COS-B, les quelques photons Y capturés ont encore beaucoup à nous offrir. Aussi le fil d'Ariane de cette thèse est-il l'étude du rayonnement r observé dans notre Galaxie par ce satellite. Un tel rayonnement ayant deux origines distinctes, les résultats présentés se regroupent selon deux thèmes principaux. Il s'agit d'une part de l'étude du milieu interstellaire .à grande échelle en mettant à profit le fait que les rayons r diffus naissent de l'interaction des rayons cosmiques avec toute matière interstellaire et les comparaisons effectuées avec divers traceurs comme les observations Co (2.6mm), HI (21cm), IR et les comptages d'étoilas et de galaxies. Cette étude repose également en grande partie sur les cartes de nuages moléculaires obtenues avec le télescope millimétrique de l'Université Columbia. La recherche et l'analyse de brillantes sources r compactes constitue l'autre thème abordé. Les photons sont alors exploités pour sonder les conditions physiques extrêmes qui gouvernent l'émission des objets très énergétiques comme les pulsars. Parmi tous les résultats peuvent être cifés : la mise en évidence en Co d'un lointain bras spiral de la Galaxie (ISkpc) et d'un important complexe moléculaire proche (300pc), le prêter panorama Co de la Galaxie, une mesure du rapport HB./SCo et des faibles gradients galactiques des rayons cosmiques, l'étonnant comportement à haute énergie "du pulsar Vela, la détection d'un? nouvelle source Y et la découverte du large reste d'une supernova qui a dû probablement retenir l'admiration de nos ancêtres lorsqu'elle a explosé à 300 pc du Soleil, voici 40 000 ans.