T i jii, i—iin ••*,
CONF-8909210—
Proceedings of
The 12th Werner Brandt International Conference on the Penetration of Charged Particles in Matter
September 4-7,1989
Universidad del Pais Vasco/Euskal Herriko Unibertsitatea San Sebastian, Basque Country, Spain
eman ta zabal zazu Printed in the United States of America. Available from National Technical Information Service U.S. Department of Commerce 5285 Port Royal Road, Springfield, Virginia 22161 NTIS price codes—Printed Copy: A-99 Microfiche A01
This report was prepared as an account of work sponsored by an agency of the United StatesGovernment. Neither theUnitedStatesGovernment nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the UnitedStates Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. The 12th Werner Brandt International Universidad del Pais Vasco/Euskal Hcrriko Unibertsitatea Conference on the Penetration of Charged September 4-7, 1989 Particles in Matter San Sebastian, Basque Country, Spain i I o OD to e o w w 10 CONF-8909210
THE 12TH WERNER BRANDT INTERNATIONAL CONFERENCE ON THE PENETRATION OF CHARGED PARTICLES IN MATTER
s..§ g-1, S- * ** if- 3 K'
September 4-7, 1989 San Sebastian, Spain i||ffj||f|
Date Published - April 1990 f 111 J g f f { g
3 &§ =• 5. r=. 3 P** -?=.} 5) *? i l*tJiff
Prepared by the OAK RIDGE NATIONAL L/rORATORY Oak Ridge, Tennessee 37831 operated by MARTIN MARIETTA ENERGY SYSTEMS, INC. for the U.S. Department of Energy Under Contract No. DE-AC05-84OR21400 TABLE OF CONTENTS
Page
Introduction and Overview
R. H. Ritchie vii
Toasts and Responses ix
Conference Participants xvii
Program xxi
Localised Elastic and Inelastic Scattering Processes for Electrons in Multilayer Structures A Howie. C. A. Walsh, and Z. L. Wang 1
Electromagnetic Resonances in Self Similar Arrays of Particles or Voids Francisco Claro and Ronald Fuchs 9
Influence of the Wake Potential on Inelastic Scattering Using a Small Probe in Confined Volumes P. E. Batson and J. Bruley 15
Electron Microscopy Study of the Electronic Properties of Small Silicon Particles Daniel Ugarte. Christian Colliex, and Pierre Trebbia 25
Electron Compton Scattering in the Electron Microscope P. Schattschneider and P. Pongratz 35
Dispersion Effects and Spatial Resolution in STEM N. Zabala and P. M. Echenique 61
Stopping Power for Protons in Al for the Whole Range of Velocities F. Flores, A. Arnau, P. M. Echenique, and R. H. Ritchie 69
Stopping Power for Protons Revisited Gregory Lapicki 85
Stopping Power Theory in Inhomogeneous Many Electron System M. Kitagawa „ 101 On the Z,' Correction of the Stopping Power I. Nagv and P. M. Echcnique Ill
Variational Approach to the Scattering Theory and its Application in Stopping Phenomena B. Apagyi and I. Nagy 125
Stopping Power and Ranges in Ion Implantation J. Bausells 139
Dynamical Effects in Electron Tunneling M. §unjic and L. Marusic 153
Photon Scanning Tunneling Microscopy R. J. Warmack. R. C. Reddick, and T. L. Ferrell 175
Extended Transfer Hamiltonian Method for Resonant Tunneling S. P. Apell. E. Albertson, and D. R. Penn 183
Laser Field Effects on the Interaction of Charged Particles with Degenerate and Non-Degenerate Plasmas N. R. Arista. R.M.O. Galvao, and L.C.M. Miranda 201
Screening Effects in Nuclear Fusion of Hydrogen Isotopes in Dense Media Nestor Arista, Raul Ba.-agiola, and Alberto Gras-Marti 211
Contribution of Ripplon Excitation to Interaction Energies of Particulates Rafael Garcia-Molina. Nuria Barberan, and Alberto Gras-Marti 221
q-Dependent Excitations of Metal Spheres Nuria Barberan, Manuel Barranco. F'ancesca Garcias, Jesus Navarro, and Lloreng Serra 231
Wake-Riding Electrons Emitted by Antiprotons Traversing Solid Targets J. Burgdorfer. J. Wang, and J. Muller 239
Non Perturbative Behaviours in Excitation and Ionisation of Atoms by Fast Charged Particles R. Gavet and A. Salin 255
Target Thickness Dependent Convoy Electron Production of Aligned Molecular HeH+-Ions J. Kemmler. K. O. Groeneveld, O, Heil, K. Kroneberger, and H. Rothard 267
ii Radiative Electron Capture by Channeled Ions J. M. Pitarke and R. H. Ritchie 273
Direct Methods in Structure Determination by LEED J. B. Pendry 289
Non-Local Exchange-Correlation Potential at a Metal Surface from Many-Body Perturbation Theory A. G. Eguiluz 303
Multiphonon Energy Exchange in the Collision of an Atom with a Surface J. R. Manson and R. H. Ritchie 313
Positronium Formation and Photoemission Spectroscopy for Surface Akira Ishii 329
Inelastic Positronium Formation at Metal Surfaces Toshiaki litaka 343
Energy Loss Measurements of He and H Scattered at Grazing Incidence off Ni A. Narmann, K. Schmidt, U. Imke, and W. Heiland 351
Electronic Response of Metal Surfaces to Electron Exchange with Doubly Charged Ions A. Niehaus 361
Transient Adsorption, Energy Loss, and Momentum Transfer in Low Velocity Ion Surface Scattering K. J. Snowdon. D. J. O'Connor, M. Kato, and R. J. MacDonald 375
The Neutralization Process of Highly Charged Ions Near a Metal Surface L. Folkerts and R. Morgenstern , 389
Some Basic Phenomena and their Importance in Relation to Techniques for Sputter- and Vapor-Deposition of High Temperature Superconducting Thin Films Orlando Auciello 397
Effective Mass of Bulk and Surface Positrons Due to Plasmon Excitation Shigeru Shindo 421
Variation of DOSD Moments with Bond Length in H2 and N2 Jens Oddcrshede, John R. Sabin. and Geerd H.F. Dicrcksen 435
iii De Structura Spectrorum Encrgiae Electroniorum Sccundariorum Ioniis Inductufum Prorsus Rcctorum c Corporibus Solidis Kar1-Ont(es Groeneveld. Robert Maier, and Hermann Rothard 445
Orbital Local Plasma Calculation of Mean Excitation Energies and Stopping Numbers S. B. Trickey. David E. Mcltzcr, and John R. Sabin 449
Secondary Electrons as Probe of Prccquilibrium Stopping Power of Ions Penetrating Solids K. Kroneberger. H. Rothard, P. Koschar, P. Lorenzcn, A. Clouvas, E. Veje, J. Kemmlcr, and K. O. Groeneveld 477
Secondary Electron Emission with Molecular Projectiles K. Kroneberger. H. Rothard, P. Koschar, P. Lorenzen, A. Clouvas, E. Veje, J. Kemmler, N. Keller, R. Maier, and K. O. Groeneveld 483
Direction-Dependent Stopping Power and Beam Deflection in Anisotropic Solids Oakley H. Crawford 489
Independent-Particle Model for Fusion in Cluster Impact Oakley H. Crawford 507
Bremsstrahlung Induced by 50 MeV H° Bombardment C. Stein, K. W. Habiger, B. R. Smith, Oakley H. Crawford, and R. H. Ritchie 517
Excited Substates Mixing of Fast Heavy Ions in Solids J. P. Rozet. A. Chetioui, K. Wohrer, C. Stephan, F. Ben Salah, A. Touati, M. F. Politis, and D. Vernhet 533
Radiation Mechanisms for Microscopic Relativistic Electron Beams Michael J. Moran 537
Interaction of Low-Energy Electrons and Positrons with Condensed Matter: Stopping Powers and Inelastic Mean Free Paths from Optical Data J. C. Ashley 551
Cross Sections for K-Shell Ionization by Electron Impact Ricardo Mavol and Francesc Salvat 573
The Theory of Track Formation in Insulators Due to Densely Ionizing Particles R. H. Ritchie, A. Gras-Marti. and J. C. Ashley 595
xv Track Structure: Perspectives, Progress, Problems Robert Katz 615
The Irradiation of Glycylglycine in Aqueous Solution - A Case Study of Calculations from Track-Structure to Biochemical Change J. E. Turner, Wesley E. Bolch, H. Yoshida, K. Bruce Jacobson, O. H. Crawford, R. N. Hamm, and H. A. Wright 633
Collective Aspects of Charged Particle Track Structure G. Basbas. A. Howie, and R. H. Ritchie 651
Calculations of Auger-Cascade-Induced Reactions with DNA in Aqueous Solution R. N. Hamm. H. A. Wright, J. E. Turner, R. W. Howell, D. V. Rao, and K.S.R. Sastry 673 THE 12TH WERNER BRANDT INTERNATIONAL CONFERENCE ON THE PENETRATION OF CHARGED PARTICLES IN MATTER
Introduction and Overview
R. H. Ritchie
It is in order to make a short sketch of the origin and philosophy of these workshops. The first was held at New York University in 1977, organized and inspired by our late colleague, Werner Brandt, who was Professor of Physics and Head of the Radiation Physics Laboratory there.
Werner was the driving force in the first and subsequent meetings until his untimely death, although it was my privilege to help in these and succeeding endeavors. Werner recognized the need for small and informal workshops in the community of workers concerned with the interaction of charged particles in matter. Since physical problems in this broad area often involve interdisciplinary expertise, it was expected that participants would be drawn from many different institutions and branches of knowledge and that smallness and informality would be essential for success in such gatherings. He emphasized that the main goal of such meetings should be to arrive at some sense of current problems, not merely to repeat results that have already been published.
In particular, all participants were encouraged to contemplate beforehand how their knowledge and scientific needs would relate to modern problems in penetration physics so that the workshops could address some of the outstanding questions of recent research. In a real sense, Werner here began a new mode of information exchange that involved timely and exciting discussions, required minimal effort by participants through the issuance of very informal summary reports, and included active workers with a broad spectrum of viewpoints. It has been a rare privilege for my colleagues and me to continue these workshops in Werner's honor and, in the tradition begun so ably by him, to attempt to convey through them the fascination to be found in research on charged particle penetration phenomena.
A list of past Workshops and their venues follows:
1977 - Wake Phenomena - New York University 1978 - Current Stopping Power Problems - New York University 1979 Low-Energy Particles - New York University 1980 - Matter Under Extreme Conditions - New York University 1981 - Exotic Projectiles - New York University 1982 - Dynamic Screening and Effective Charge - Honolulu 1983 - Properties of Ion-Induced Tracks in Matter • Oak Ridge National Laboratory
vii 1984 - Inelastic Near-Surface Interactions - Oak Ridge National Laboratory 1985 - Photon Emission from Irradiated Solids - Oak Ridge National Laboratory 1986/87 - Dynamic Interactions of Probes with Condensed Matter - University of Alicante, Alicante, Spain 1988 - Charged Particle Penetration Phenomena - Oak Ridge National Laboratory
The present Proceedings represents a wide spectrum of topics, but the freshness and interest of contributions to this Workshop more than compensate for the broadness of the coverage. The stimulation and excitement of exposure to the research of workers in other areas has led to interactions that may be important for the future.
Professor Pedro Elxenikc and his colleagues did an outstanding job of making arrangements for this Workshop and showed magnificent hospitality to the attendees. Pedro was ably assisted by Professor J. Colmenero, Drs. A. Rivacoba, A. Arnau, J. M. Pitarke, N. Zabala, J. Garcia, and Miriam Penalba, and by the conference secretaries M. Alvarez, A. Peon, C. Aycart, and E. Marti. Welcoming addresses by Dr. Angel Galindez, BBV; Dr. J. A Garrido, Consejero Director General de Iberduero SA; Xabier Albistur, Mayor of San Sebastian; and J. M. Aguirre, Deputy of Economy of the Province of Gipuzcoa, are given below.
Research sponsored by the Office of Health and Environmental Research, U.S. Department of Energy, under contract DE-AC05-84OR21400 with Martin Marietta Energy Systems, Inc.
viii Toast bv Dr. Angel Galindcz. BBV
Mis buenos amigos Pedro Echcniquc y José Antonio Garrido me han pedido que pronuncie unas breves palabras al término de esta cena, prólogo de la 12 Conferencia Werner Brandt que se va a celebrar aquí, en San Sebastián.
Quiero en primer lugar darles la bienvenida más cordial y afectuosa. Realmente, estoy seguro que todos Uds., que han acudido a la invitación de la Universidad del Pais Vasco, van a pasar unos días gratos, agradables y provechosos, envueltos en la hospitalaria amistad de la gente de esta tierra y dentro de este marco incomparable que es Donosti, una verdadera perla de nuestro Pais.
Deseo, también, aprovechar la ocasión que ofrecen estas jomadas científicas, en las que están implicados empresarios, para transmitir a todos Uds., inquietudes actuales de personas que han trabajado muchos años en las fronteras científicas de la tecnología, la industria y los servicios.
Estas preocupaciones se centran en la formación, la maduración y el reciclaje de las personas, principalmente los más jóvenes.
La Universidad y la Empresa son Instituciones muy vinculadas a esa tarea, exigida principalmente por la velocidad del crecimiento actual de los medios de comunicación y de los cambios tecnológicos y que, en resumen, trata de conseguir formar debidamente más gente en menos tiempo.
Tenemos la experiencia, vivida intensamente en los últimos años, de que es fácil, en el cambio de una empresa, reducir a la mitad el tiempo que históricamente se ha considerado necesario para preparar a los más jóvenes a hacer frente a ¡as mayores responsabilidades.
Igualdad y eficacia son téminos que es muy difícil conciliar, pero lo cierto es que la igualdad de oportunidades en el arranque y el estímulo en la carrera son los mobiles más adecuados para lograr la pronta maduración de las personas jóvenes.
Me parece que tanto la Universidad como la empresa deben de tener presente que hoy es esta la máxima prioridad. Formar adecuadamente más gente en menos tiempo.
Hasta ahora parecía que la acción de la Universidad y de la Empresa sobre la formación de una persona estaba saparada por una barrera infranqueable que se situaba en tomo a la edad del final de la graduación académica.
ix Creo que esta ban-era debe ser demolida. Creo que ambas, la Universidad y la Empresa, deben de colaborar desde los límites de la enseñanza secundaria, impartiendo al mismo tiempo y en el mismo tono, la ilusión, el conocimiento y el comportamiento colectivo.
Yo estoy seguro de que así será y de que todos Uds. van a trabajaren esa gran orea que es el formar y madurar a una juventud que quiere ansiosamente alcanzar, en menos tiempo, todo aquello de lo que se siente capaz.
Por último quiero levantar mi copa y brindar por esa colaboración de la Universidad y de la Empresa, por estas Jornadas, que son un ejemplo de esa colaboración, por Uds. y por sus familias.
ÁNGEL GALINDEZ
BBV Toast bv Dr. Jose Antonio Garrido. Consejero Director General de Iberduero SA
Nuestra Compañía tiene sus raices más sólidas en los grandes avances de la ciencia y en la cultura empresarial de sus recursos humanos. Queremos contribuir desde nuestro trabajo a la creación de un entorno científico tecnológico empresarial de calidad que nos permita competir en esta nueva Europa que con dificultades, pero con ilusión estamos construyendo.
En este proyecto su labor, la de los Europeos y la de nuestos amigos Americanos del norte y del sur, y Japoneses que nos acompañan es importante. No sólo para los países desarrollados. Como hemos discutido en otros foros es una responsabilidad ética de los que pertenecemos a los países desarrollados, empeñamos en desarrollar nuevos conocimientos y técnicas que permitan una solución mejor de los problemas humanos de todos los Países.
Al ofrecer esta cena de bienvenida al Duodécimo Congreso W. Brandt estamos queriendo hacer público nuestro profundo agradecimiento a todas las personas que, como Vds., hacen avanzar el conocimiento humano en aquellas cuestiones que nos permiten a nosotros mejorar la calidad de nuestro servicio a la sociedad.
Yo brindo, en nombre de mi Compañía, para que todos Vds. continúen en su esfuerzo de abrir nuevos horizontes a nuestras posibilidades.
JOSÉ ANTONIO GARRIDO
IBERDUERO
XI TOAST IN RESPONSE TO DRs. GALINDEZ AND GARRIDO
On behalf of the attendees at this 12th Werner Brandt International Conference on the Penetration of Charged Particles in Matter, I thank you for your kind remarks of welcome. In his autobiography, the famous physicist, George Gamow, tells about an interesting incident. At that time he was giving a popular lecture, open to the public, on the prospects of extracting energy from the fusion of nuclei. This was quite early in the history of nuclear physics, and the possibility of nuclear fusion had just been suggested. After his lecture, the director of the local energy district approached Gamow. He indicated great interest in the subject and made a very ingenious suggestion. He said that if Gamow would furnish him with some deuterium, presumably several decagrams in the D2O form, embedded in a copper conductor, he, the director, would make available all of the electrical power of his district to be sent through this material at an early hour on any given morning. Fortunately, Gamow was wise enough to decline to implement this fascinating suggestion.
I relate this incident to draw a parallel with the imaginative and farsighted interest in science that Dr. Garrido and Dr. Galindez have shown in their support of this Workshop and in many other instances in the past. Their remarkable vision and interest in the basic sciences are of great significance to the future of science and technology in the Basque Country and in Spain. We all thank you for this and for your magnificent hospitality.
R. H. Ritchie Oak Ridge National Laboratory and University of Tennessee
xii Opening Address of the Conference hv J. M. Apuirre. Deputy of Economy of the Province of Gipuzcoa
Buenos días Señoras y Señores.
Quiero agradecer a los organizadores de este Congreso, en particular a mi querido amigo el Profesor Echenique, la oportunidad que me brindan de hacerles esta salutación, en mi nombre y en el de todos los guipuzcoanos.
Es un orgullo poder dirigir unas palabras a un auditorio tan cualificado de profesores, investigadores y científicos de varios países del mundo. De la simple lectura de los títulos de las ponencias se deduce un eievadísimo nivel científico que espero les sea provechoso, a la vez que les deseo una agradable estancia entre nosotros, en los pocos ratos libres que les quedan para otras actividades.
sería muy presuntuoso por mi parte hablarles de la importancia de la ciencia para el progreso material de la humanidad. Cada vez más, estamos esperando de ustedes soluciones a los grandes problemas que tenemos planteados : sanitarios, de medio ambiente, energéticos, etc.... Sabemos que los resultados no se producen a corto plazo, sino con el esfuerzo continuado, con ilusión y con medios a todos los niveles: en la Administración, en la Empresas, en las Universidades y en los Centros de Investigación.
Es destacable el empuje que se ha dado en estos años a la ciencia y a la tecnología, en una sociedad como la nuestra, tradicionalmente algo reacia al conocimiento de las nuevas ideas, pero que ha comprendido perfectamente que nuestro progreso y nuestro futuro como pueblo, van unidos a otros, hasta el momento más avanzados. Multitud de problemas que tenemos, desde el déficit de la balanza tecnológica, hasta las altas tasas de paro, tienen una explicación última en el insuficiente esfuerzo en años pasados a los temas de la cultura, la formación, la investigación y el desarrollo tecnológico.
Esta tipo de Congresos tienen para mí una significación de una especie de símbolo de que nos enganchamos el tren de la modernidad y del futuro, con nuestra propia aportación, ciertamente modesta, pero que esperamos y deseamos creciente.
Esta Facultad de Químicas, reciente y joven, está alcanzando un grado de madurez en su presido científico que le permite tanto ser anfitrión de un evento tan cualificado como es éste, como de hacer sus propias aportaciones, tal como lo viene haciendo en los más prestigiosos foros y revistas científicas internacionales. Sepan Ustedes que, desde la Diputación y si bien con unas cifras modestas pero comparativarnenie con otras Administraciones con mayor competencia y responsabilidad administrativa, elevadas, apoyamos y seguimos apoyando este vivero de ideas y
xiii conocimientos que constituyen sus diversoso Departamentos.
Muchas gracias por su atención y espero que esta tarde tengamos ocasión de compartir un brindis con Ustedes, en la recepción que les ofrecemos en la Diputación, y a ia que gustosamente quedan invitados. Muchas gracias.
J. M. AGUIRRE
DEPUTY OF ECONOMY OF GIPUZCCA
xiv Toast by Xabier Albistur. Mayor of San Sebastian
I am very happy to welcome this distinguished group of scientists to the beautiful city of Donostia. Our fair town has existed since remote times, developed a prosperous maritime trade in the 14th century and is one of the loveliest and most picturesque cities in the world. You have seen some of the scenic attractions already and will understand why it is a highly popular resort area renowned for its restaurants, touristic and sporting activities.
As with most cities in the industrialized world, commercial developments in this area have grown out of the industrial revolution and the rapid increase in the resulting technology. Now we all ride on the crest of a new, high technology revolution in physics, electronics, computers, informatics, and molecular sciences. As Mayor of Donostia, I am proud to welcome a group of working scientists who arc closely coupled with these fascinating and timely developments.
The great philosopher, Seneca, wrote that "An age will come after many years when the ocean will loose the chains of things and a huge land lie revealed; when Tiphys will disclose new worlds and Thulc no more be the ultimate." Though this has been taken by many, including Columbus, as prophesying the discovery of ihe New World, it may be read as a revelation of the new world opening now in which man achieves increasingly complete control of the material world and his own genetic, physical, and intellectual nature.
Seneca also said that "a good mind possesses a kingdom." This workshop convenes many excellent minds engaged in important research. I hope that your conference will be highly profitable scientifically, that you enjoy your stay in Donostia, and that you will return in the near future.
XABIER ALBISTUR
MAYOR OF SAN SEBASTIAN THE 12TH WERNER BRANDT INTERNATIONAL CONFERENCE ON THE PENETRATION OF CHARGED PARTICLES IN MATTER San Sebastian, September 4-7, 1989
CONFERENCE PARTICIPANTS
Isabel Abril George Basbas Universidad Complutcnse Physical Review Letters Madrid, SPAIN Ridge, NY USA
J. F. Annctt Philip E. Batson University of Illinois IBM/T.J. Watson Research Center Urbana. IL USA Yorktown Heights, NY USA
Barnabas Apagyi Joan Bausells Technical University of Budapest Universidad Autonoma de Barcelona Budapest, HUNGARY Bellaterra, SPAIN
Peter Apcll G. Borste! Umea University University of Osnabriick Umca, SWEDEN Osnabriick, FRG
Nestor R. Arista Hidde H. Brongersma Universidad de Alicante Eindhoven University Alicante, SPAIN Eindhoven. THE NETHERLANDS
Andres Arnau Joachim Burgdordcr Universidad del Pais Vasco University of Tennessee San Sebastian, SPAIN Knoxville", TN USA
James C. Ashley A. Chelioui Oak Ridge National Laboratory Section de Physique el Chcmie Oak Ridge, TN USA University P. M. Curie 75231 Paris Cedex 05 Orlando Aucicllo CNRS URA 1379 Microeleclronic Center of North Carolina FRANCE Research Triangle Park, NC USA Francisco Claro Raul Baragiola Pontificia Universidad Calolica de Chile Rutgers University Santiago, CHILE Piscataway, NJ USA Vincent Cobut Manuel Barranco Universite de Sherbrooke Universidad dc Balearcs Quebec, CANADA Palma de Mallorca, SPAIN
:-:vn Juan Colmencro R. N. Hamm Universidad del Pais Vasco Oak Ridge National Laboratory San Sebastian, SPAIN Oak Ridge, TN USA
Oakley Crawford W. Hciland Oak Ridge National Laboratory Univcrsitat Osnabriick Oak Ridge, TN USA Osnabriick, FRG
Adolfo G. Eguiluz Archibald Howie Montana State University Cavendish Laboratory Bozeman, MT USA Cambridge, UK
P. M. Etxenike Toshiaki litaka Universidad del Pais Vasco Waseda University San Sebastian, SPAIN Tokyo, JAPAN
Fernando Eores Akira Ishii Universidad Autonoma de Madrid Tottori University Madrid, SPAIN Tottori, JAPAN
L. Folkerts Robert Katz Kernfysisch Versmallcr Instituut University of Nebraska Groningen, THE NETHERLANDS Lincoln, NE USA
Javier Garcia Jiirgcn Kcmmlcr Universidad del Pais Vasco Institut fur Kcrnphysik dcr Univcrsitat San Sebastian, SPAIN Frankfurt/Main, FRG
Rafael Garcia Molina M. Kimura Universidad de Murcia Argonne National Laboratory Murcia, SPAIN Argonne, IL USA
Robert Gayet Mitsuo Kitagawa Universite Bordeaux I North Shore College Talence-Cedex, FRANCE Atsugi, JAPAN
Alberto Gras-Marti Kurt Kroncbergcr Universidad de Alicante Institul fur Kernphysik dcr Univcrsitat Alicante, SPAIN Frankfurt/Main, FRG
Karl Ontjes Groencveld Norma J. Kwaak Institut fur Kernphysik der Universitat Oak Ridge National Laboratory Frankfurt/Main, FRG Oak Ridge, TN USA
Luis A. Guzman Gregory Lapicki IRST Materials Science Division East Carolina University Povo (Trcnto), ITALY Greenville, NC USA
xviii Janos Laszlo Miriam Pcnalba Max-Planck-Institut fur Plasmaphysik Universidad del Pais Vasco Garching, FRG San Sebastian, SPAIN
Ronald D. MacFarlane Jesus Perez Conde Texas A&M University Orsay U.P.S.-XI College Station, TX USA Ccdcx-Orsay, FRANCE
Joseph Manson Jose Maria Pitarke Clemson University Univcrsidad del Pais Vasco Clemson, SC USA Lcioa, SPAIN
Ricardo Mayol Sanchez Joseph Remillicux Universidad de Barcelona Institut Physique Nucleaire Lyon Barcelona, SPAIN Villeurbanne Cedex, FRANCE
Catherine J. McNeal Alberto Rivacoba Texas A&M University Universidad del Pais Vasco College Station, TX USA San Sebastian, SPAIN
Jose Luis Minchole R. H. Ritchie Dtor. Radiaciones Oncologico Oak Ridge National Laboratory San Sebastian, SPAIN Oak Ridge, TN USA
Rosa Monrcal J. P. Rozct Universidad Autonoma de Madrid Section de Physique ct Chemie Madrid, SPAIN P. M. Curie University 75231 Paris Cedcx 05 Michael Moran CNRS URA 1379 Lawrence Livcrmore Laboratory FRANCE Livermore, CA USA John R. Sabin Istvan Nagy Odcnse University Institut of Physics Odense, DENMARK Budapest, HUNGARY P. Schattschncider Arcnd Nichaus Technische Universitat Ryksunivcsileit Utrecht Vienna, AUSTRIA Utrecht, THE NETHERLANDS Lloreng Serra Crespi Hideo Nitta Univcrsidad dc Baleares Tokyo Gakugci University Palma de Mallorca, SPAIN Tokyo, JAPAN Shigeru Shindo John B. Pendry Tokyo Gakugci University Imperial College Tokyo, JAPAN London, UK
xix Ken Snowdon Ncrca Zabala Universitat Osnabrvick Universidad del Pais Vasco Osnabriick, FRG Leioa, SPAIN
Charles Stein Eugene Zarcmba Air Force Weapons Laboratory Queen's University Albuquerque, NM USA Kingston, CANADA
Marijan Sunjic University of Zagreb Zagreb, YUGOSLAVIA
Samuel B. Trickey University of Florida Gainesville, FL USA
J. E. Turner Oak Ridge National Laboratory Oak Ridge, TN USA
Jesus Ugalde Universidad del Pais Vasco San Sebastian, SPAIN
Daniel Ugarte Univcrsite Paris XI Orsay, FRANCE
• Michel A. Van Hove University of California Berkeley, CA USA
Matesh N. Varma U.S. Department of Energy Washington, DC USA
R. J. Warmack Oak Ridge National Laboratory Oak Ridge, TN USA
Helmut Winter Instilut fur Kcrnphysik Miinster, FRG
Yasunori Yamazaki University of Tokyo Tokyo, JAPAN
xx THE 12TH WERNER BRANDT INTERNATIONAL CONFERENCE ON THE PENETRATION OF CHARGED PARTICLES IN MATTER
San Sebastten, September 4-7, 1989
PROGRAM
Sunday. 3 September 1989 21:00 Dinner offered by Iberduero, S.A., at the Hotel Maria Cristina
Monday. 4 September 1989
08:00-09:00 Registration
09:00-09:30 Opening
09:30-11:30 Session I: Electron Microscopy. Chairman. M. Moran A. HOWIE: "Localised Elastic and Inelastic Scattering Processes for Electrons in Multi-Layer Structures"
F. CLARO: "Electromagnetic Resonances of a Self-Similar Array of Metallic Particles"
P.E. BATSON: "Near Edge Fine Structure: Differences Between Electron Energy Loss Scattering and Photon Absorption?"
D. UGARTE: "Electron Microscopy Study of the Electronic Properties of Small Silicon Particles"
P. SCHATTSCHNEIDER: "Electron Compton Scattering in the Electron Microscope"
N. ZABALA: "Dispersion Effects and Spatial Resolution in STEM"
11:30-12:00 Coffee Break
12:00-14:00 Session II: Energy Loss m. Chairman. J. R. Sabin
F. FLORES: "Stopping Power for Protons in Al"
xxi G. LAPICKI: "Formulas for Stopping and Range Revisited"
M. KTTAGAWA: "Stopping Power Theory in Inhomogcncous Many Electron System"
I. NAGY: "On the Z3 Correction of the Stopping Power"
B. APAGYI: "Variational Approach to the Scattering Theory and its Application in Slopping Phenomena"
J. BAUSELLS: "Stopping Power and Ranges in Ion Implantation"
16:00-17:00 Session HI: STM. Chairman. F. Flores
M. SUNJ1C: "Dynamical Effects in Electron Tunneling"
R. J. WARMACK: "Photon Scanning Tunneling Microscopy"
P. APELL: "Extended Transfer Hamiltonian Method for Resonant Tunneling"
17:00-17:30 Coffee Break
17:30-18:30 Session IV: Other Topics m. Chairman. M. Sunjic
N. ARISTA: "Laser Field Effects on the Interaction of Charged Particles with a Degenerate Electron Gas"
R. GARCIA MOLINA: "Contribution of Ripplon Excitation to the Interaction Energies of Particulates"
M. BARRANCO: "Multipote Response of Metal Spheres to q-Dcpendent Excitation Operators"
19:00 Welcome Reception
Excelentisima Diputacion Foral dc Guipuzcoa
Tuesday. 5 September 1989
09:00-11:00 Session V: Fast Ions. Chairman. J. Remillicux Y. YAMAZAKI: "Interaction of Anliprotons with Condensed Matter"
xx ii J. BURGD6RFER: "Forward Electron Production in Ion-Solid Collisions"
R. GAYET: "Non Pcrturbative Behaviours in Excitation and Ionization by Fast Charged Particles"
II. NTTTA: "On the Line Broadening of Okorokov Peak Profiles"
J. KEMMLER: "Electron Production and Transport in Thin Solids"
J. M. PITARKE: "Radiative Electron Capture by Channeled Ions"
11:00-11:30 Coffee Break
11:30-13:30 Session VI: Surface Physics. Chairman. A. Niehaus
J. B. PENDRY: "Direct Methods in Surface Structure Determination by Low Energy Electron Diffraction"
A. EGUILUZ: "Evaluation of the Exchange-Correlation Potential at a Metal Surface from the Knowledge of the Electron Self-Energy"
J. R. MANSON: "Multiphonon Energy Exchange in the Collision of an Atom with a Surface"
J. F. ANNETT: "New Universal Features of Surface Charge Density. Bonding and Stress"
A. ISHII: "Positronium Formation and Photoemission Spcctroscopy for Surfaces"
T. IITAKA: "Inelastic Positronium Formation at Metal Surfaces"
21:00 Basque Dinner at Sociedad Gastronomica Donostiarra
Wednesday. 6 September 1989
09:00-11:00 Session VII: Ion Surface. Chairman. J. B. Pendrv
W. HEILAND: "Energy Loss Measurements of He and H Scattered at Grazing Incidence off Ni"
R. MONREAL: "Energy Loss of Light Ions and Neutrals from Surface Scattering"
xxiii H. WINTER: "Coulomb Explosion of Fast Molecular Ions after Grazing Surface Scattering"
A. NIEHAUS: "Electronic Response of Metal Surfaces to Electron Exchange with Doubly Charged Ions"
K. SNOWDON: "Transient Adsorption, Energy Loss, and Momentum Transfer in Low Velocity Ion Surface Scattering"
H. H. BRONGERSMA: "New Possibilities of Low-Energy Ion Scattering"
11:00-11:30 Coffee Break
11:30-13:30 Session VIII: Other Topics(2). Chairman. A. Gras-Marti G. BORSTEL: "Calculated Photoemission and Many Body Effects in 3D- Ferromagnets"
K. J. McNEAL: "^Cf-Plasma Desorption Mass Spectromctry of Gold Clusters"
L. FOLKERTS: "The Neutralization Timescale of Highly Charged Ions on a Tungsten Surface"
R. BARAGIOLA: "Mechanisms for Multiply Charged Ion Emission in Electron Bombardment of Insulators"
O. AUCEELLO: "Some Basic Phenomena in Plasma and Ion and Laser Beam Interaction with Surfaces Related to the Synthesis of High-Tc Superconducting Films"
S. SHINDO: "Effective Mass of Bulk and Surface Positrons Due to Plasmon Excitation"
21:30 Conference Dinner at El Palacio de Miramar
Thursday. 7 September 1989
09:00-11:00 Session IX: Energy Lossf21. Chairman. J. C. Ashley
J. R. SABIN: 'Calculation of Slopping Powers in Some Aggregated Systems"
K. O. GROENEVELD: "Secondary Electron Emission as a Probe for the Non- Equilibrium Energy Loss"
xxiv S. B. TRICKEY: "Calculation of Mean Excitation Energy and Stopping Cross- Section in the Orbital Local Plasma Approximation"
K. KRONEBERGER: "Secondary Electron Emission with Molecular Projectiles"
O. H. CRAWFORD: "Direction-Dependent Stopping Power and Beam Deflection in Anisotropic Solids"
C. STEIN: "Brcmsstrahlung Induced by 50 MeV H° Bombardment"
11:00-11:30 Coffee Break
11:30-13:30 Session X: Other Topicsf3V Chairman. H. H. Brongersma J. P. ROZET: "Excited Substates Mixing of Fast Heavy Ions in Solids"
M. MORAN: "Radiation Mechanisms for Microscopic Rclativistic Electron Beams"
J. C ASHLEY: "Energy Loss of Electrons and Positrons in Solids"
R. MAYOL: "Cross Sections for K-Shell Ionization by Electron Impact"
A. GRAS-MARTI: The Theory of Track Formation in Insulators Due to Densely Ionizing Particles"
M. N. VARMA: "Radial Dose Distributions and their Application to Micro- and Nanodosimetry"
15:00-16:20 Session XI: Track Structure. Chairman. Y. Yamazaki
R. D. MacFARLANR "Fast Chemical Reactions in Heavy Ion Tracks"
R. KATZ: Track Structure: Perspectives, Progress, Problems"
J. E. TURNER: The Irradiation of Glycylglycine in Aqueous Solution, a Case Study of Calculations from Track Structure to Biochemical Change"
R. N. HAMM: "Calculations of Physical and Chemical Reactions with DNA in Aqueous Solution from Auger Cascades"
16:20-16:40 Coffee Break
XXV 16:40-17:25 Session XII: Cluster Fusion. Chairman. P. Echenique
G.BASBAS: "Ouster Fusion" R. BARAGIOLA: "Ouster Fusion" O. CRAWFORD: "Cluster Fusion" R. RITCHIE: "Ouster Fusion"
17:20 CLOSING REMARKS; P. M. Echenique and R. H. Ritchie
19:30 Social Party at Ritchie's Villa in Fuenterrabia
xxvi LOCALISED ELASTIC AND INELASTIC SCATTERING PROCESSES FOR ELECTRONS IN MULTILAYER STRUCTURES
A. Howie. C.A. Walsh and Z.L. Wang. Cavendish Laboratory, MadingJey Road. Cambridge CB3 OHE ABSTRACT: Characterisation of the microstructure of many modern composite materials depends on development and efficient use of new elastic and inelastic electron scattering processes which yield a high degree of spatial localisation. This is illustrated with reference 10 recent studies of semiconductor multilayer systems as well as natural layered crystals. In reflection siudies surface oxidation or damage seriously complicates the results. 1. INTRODUCTION Now that high resolution HREM and STM images showing individual atoms are commonplace in the scientific and popular literature, one couid be forgiven for thinking that the problems of materials characterisation are at an end. Rarely however do these images provide very direct information about local chemistry or electronic structure. This is well illustrated by the problem of characterising the artificial multiiayer structures which are of increasing importance both for semiconductor technology and metallurgy. Perhaps the most formidable challenge for electron microscopists, is posed by GaAs / AlAs superiattices where the small scattering difference between the two materials exacerbates the usual problems of quantitative HREM smicture imaging by phase contrast methods. The main effort so far has been in high resolution dark field imaging (Kakibayashi and Nagata, 1986: Bithell and Stobbs, 1989) based on the 200 Bragg reflection which, for structure factor reasons, maximises the scattering difference. Convergent beam diffraction methods have also been applied (Eaglesham and Humphrevs. 1986).
Scanning transmission electron microscopy CSTEM) is a potentially powerful combination of imaging and analysis for investigating many multilayer samples. Core loss speciroscopy CBuilock et ai 1986: Petford-Long and Long 1988) and annular dark field (ADF) imag3ng (Walls 198*: McGibbon et ai 1988) have both been usefully employed. The size of the focussed electron probe is a basic limitation in these methods but Pennycook f 1989) has recently demonstrated the exciting potential of the ADF mode in a high resolution STEM for Z- contrast imaging of multilayer and other structures at atomic resolution. Preparation of the thin multilayer samples needed for these edge-on siudies is non-trivia] and worth avoiding. Vincent et ai (1988) have extracted useful information from high-angle TEM diffraction patterns taken in plan view. Multilayers have also been imaged in bulk" at 2nm resolution by SEM fOgura. 1988) using both the back scaitered primaries and the secondary electron (SE) signals. An alternative possibility of reflection electron microscopy CREMi imaging at grazing angies (Eoothroyd et ai 1987: Buffat et al 1987) has been pursued here. combined with reflection energy loss spectroscopy using STEM. These methods have bee:: applied to samples of GaAs / A.'As muitiiayers cisaved in such a way (see fig. 1 > as to perm:: some transmission imaging and valence loss spectroscopy as well. We also report REM. vaience ioss spectroscopy ar.d SE imaging of cieaved surfaces of alum:na which are :-. l:mc of rarural uirrafine mu hi I aver sample. TEM REM
Fig. 1. 2. ELASTIC SCATTERING AND IMAGING Since AlAs surfaces oxidise much faster than those of GaAs and REM images are rather sensitive to surface disorder, it was found necessary to transfer the samples to the STEM as quickly as possible after cleaving them. Figs. 2 (a) and (b) show the bright field and ADF image of an AlAs / GaAs multilayer sample where, in a sequence of equally thick 2.8nm layers, an anomalous 5.6nm thick layer of Al Ga / As was accidendy introduced between iwo of the GaAs layers because the Ga shutter remained open. As expected, the ADF image which is dominated by incoherent scattering and Z contrast effects, shows the Ga rich regions as bright. The fact that the STEM bright field image fig. 2(a> is complementary to this, despite the considerable variations in thickness across the wedge sample, suggests that the intensity here is controlled mainly by absorption effects rather than by diffraction contrast
Figure 2. Bright field (a), ADF (b) and SREM (c) images of GaAl / As multilayer samples. Figure 2(c) is a SREM image of a different GaAl / As superlattice sample with lOnm thick layers of each material taken with the (10.10,0) reflection under surface resonance conditions. Here the contrast is similar to the ADF image, suggesting that large-angle incoherent scattering also makes an appreciable contribution to REM contrast. In some cases a large background was visible in the diffraction patterns, suggesting that the specimen had not been transferred quickly enough and that an oxide overlayer had built up.
Another case where surface layers exercise a profound effect in REM images is the a-alumina (0,1,1) cleaved surface which shows almost uniform bright contrast when first observed. After about three seconds of illumination however, some damage products start to appear on the surface. Figs. 3(a) .and 3(b) compare the same surface area before and after being illuminated by the electron beam for about 5 seconds. Inevitably the image shown in fig. 3(a) has already been slightly damaged. The damaged surface sometimes appears with three different contrast levels or domains labelled A, B and C and separated by some surface steps. The A domain appears with a very bright reflected intensity and keeps quite stable throughout. The B domain is also initially as bright as the A domain but is sensitive to beam irradiation and becomes dark showing some signs of damage products at the surface, see fig. 3(b). The products formed can cover the whole B domain but leave the A domain almost unchanged. The C domain, which is not present on all surfaces, may also undergo a damage process in a similar short period of time but cannot be visualized clearly due to its apparent roughness.
a-AJumina (0,-1,1) Surface
Fig. 3. REM images of cleaved Al (0,1,1) surface showing different surface termination domains (a) before and (b) after about 5 seconds of illumination.
Examination of the crystallographic structure of cc-alumina shows that the (O.T.I >• cleavage surface can have two or three different atomic terminations depending on the surface normal direction. For the surface normal f 0,1.1 ], the termination can be a single oxycen layer on the top of an aluminium ion layer or a single layer of A3+5 on the top of a O2 layer. Besides ihese two terminations, the surface can terminate with a double layer of O"2 on ihe top of an A3+3 layer if the surface normal is [0,1,1]. This is most probably the C domain. STEM studies of a-alumina by Bergcret al (1987) showed electron beam v;t3. VALENCE LOSS TRANSMISSION SPECTROSCOPY
Valence loss spectra obtained in transmission from individual layers sn ar. A;As I GaAs superlattice are shown in fig. 4. In contrast to previous results of Walls f 1^8,1 v.iio r-pon'id no detectable difference in the spectra, we find a consistent downward sh;f: of ;h'; r;-j3k piasmon loss combined with increased intensity near 30eV in one of :be layers. Although reliable independent identification of the layers was unfortunately unavailable at the time, both of these effects are consistent with the damped Lorentz oscillator mode] preoictior; for ::/- dielectric response of the lower band gap material with extra d-JeveJs associated w::r. GJ atoms.
Fig. 4. Transmission valence loss spectra from individual iayers of &r. AIGa /' As :jperla:::r_e
It is interesting that the transmission spectra obtained hue from the individuaj izyzri differ i>o:r: from pure GaAs as weii 2S from each other, even horn layers as narrow as 2.%r.rr. As fojr.c by Walls (1988) for the Si - S1O2 ir.:erf2c».:here car, be features of va'e-c» io',=. -pe-'j-a v,r.:cr. car. arise from sample structure finer th2rs y;s probe s:ze. As pointed ou: by fi:tch:e a::o ll'/xit • 19%%;. sensitivity to sucn fine 'tructure co-ic bt e-nar-c^: •:>• rr:ov:r." •:.-• 'I'fzv.: "• jr. -,:': ax:s posinori to emphasise n:gn-c excitations. /-.". adsqjate tneor. o: i:e.e::r;: ^•.•:::u'::.-:/: .:: 3'jperia'tices has vet 10 rs. worked out to :r.tsrcre; tr,'*st rss-.tv Hortr.;* <^ c '--'•. ",j.-- d sheets of eisctror. cas separated •:-• c:elet:tr:: .w-i~-.. ".'..' '.'z-irr.' :r ~.i r::~'-". •• ".••:•••"-• lz with orJv er.o'jsh erersv to excite cor.ajcf.Dn band txc::at;o''. 4. VALENCE LOSS REFLECTION SPECTRA
Fig. 5 shows some reflection valence loss spectra taken in the (10,10,0) specuiar reflection under resonance conditions in the geometry of fig. 1 from GaAs crystals (full lines in fig. 5) as well as from a superlattice of Alo.3Gao.7As/GaAs in equal lOnm layers (broken lines). One of the GaAs crystals was transferred rapidly and gave the light full line spectrum with a rather intense surface plasmon peak at 12eV. A few minutes were used in transferring the other GaAs crystal which yielded the heavier full curve with a surface or interface loss at 10eV and much higher intensity in the bulk plasmon peak as well as at higher losses. These differences all seem consistent with what can be expected from the presence of an oxide overlayer. A transfer time of a few minutes was also needed for the superlattice sample and it is noteworthy that both of the relevant (broken) curves in fig. 5 exhibit die high intensity in the bulk plasmon and higher loss regions apparently characteristic of the oxidised sample. The dotted curve with the most intense bulk loss and a surface loss at 1 leV comes from the interface between layers; the dashed curve with a surface loss at lOeV comes from a Alo.3Gao.7As layer.
• i.w-
EKERGr lOSS It.-)
Fig. 5. Reflection spectra of valence losses from individual layers of an AlGaAs supcrlatiice. The lower, thin solid line is from a clean GaAs sample and the upper, thick line curve from an oxidised GaAs sample. The dotted curve is from an Alo.3Gao.7As layer and the fourth, dashed curve is from the interface between the layers. Reflection valence loss spectra were also collected from the different A,B and C domains of the alumina surface. Despite the large differences in total reflected intensity noted ab^ve, it was found that the differences in the loss spectra in this case were very small though reproduceable. 5. SECONDARY ELECTRON fSE) IMAGES. Secondary electron (SE) images can be collected in the STEM and used to show surface morphology because of the influence of generation efficiency and escape depth. These images can have a resolution of lnm which is quite surprising if one considers the possible role of deiocalised processes like plasmon generation, propagation and decay in the chain of events leading to the generation of a secondary electron by the fast incident electron. Fig. 6 is a comparison of the scanning REM (SREM) image and the SE image taken under the same illumination conditions and almost simultaneously from the same area of an a-alumina cleaved surface after being slightly damaged. The bright A and the dark B domains are visible in the SREM image. A rather similar bright and dark contrast is observed in the SE image, with the bright A domain having a larger secondary electron emission rate but the dark B domain less. The origin of this effect is not entirely clear but three faaors which may affect SE emission are the step between domains, the change of surface work function and the change of surface conductivity depending on the nature of the terminating layer and the loss of oxygen ions.
400A :••'•-*
Fig. 6. SREM image (a) and SE image (b) of a slightly damaged a- alumina (0,1,1) surface* 6. CONCLUSIONS Although considerable progress is being made in developing specimen, characterisation techniques with high spatial resolution, the precise degree of localisation available with different excitations is still far from clear. The present work illustrates the particular difficulties of REM or SE observations where the results can be profoundly affected cy contamination, oxidation or radiation damage effects. In particular, the compiex radiation sensitivity of cleaved alumina surfaces has been demonstrated. We thank the SERC for financial support. :he Cavendish Laboratory for a Clerk Maxv.»i: Studentship Berger S , Salisbury I, Milne R H, Imeson D and Humphreys C J 1987 PhiL Mag. B55 341 Bithell E G and Stobbs W M 1989 Phil Mag A60 39 Bleloch A L, Howie A, Milne R H and Walls M G 1989 Ultnunicroscopy 29 175 Boothroyd C B, Britton E G, Ross F M, Baxter C S, Alexander K B and Stobbs W M 1987 Microscopy of Semiconducting Materials 1987 (Inst of Physics London) pl95 Buffat P A, Stadelmann P, Ganiere J D, Martin D and Reinhan F K 1987 Microscopy of Semiconducting Materials 1987 (Inst of Physics London) p207. Bullock J F, Titchmarsh J M and Humphreys C J 1986 Scmicond Sci Technol 1 343 Eaglesham D J and Humphreys C J 1986 Proc Xlth ICEM (Ed T Imura, S Marcuse and J Suzucki) (Jap. Soc. EM Tokyo) p209 Horing N J M, Tso H C and Gumbs G 1987 Phys Rev B36 1588 McGibbon A J, Chapman J N and Cullis A G 1988 Proc EUREM 88 Conf (Inst of Physics London) p403 Pennycook S J 1989 Ultramicroscopy 30 58 Petford-Long A K and Long N J 1988 Analytical Electron Microscopy (ed G W Lorimer) {Inst of Metals London) p201. Ritchie R H and Howie A 1988 Phil Mag A 58 753 Ogura K 1988 Proc F/th Asia-Pacific EM Conf (EM Soc of Thailand Bankok) p 165 Vincent R, Wang J, Chems D, Bailey S J, Preston A R and Steeds J W 1988 Proc EMAG87 (Inst. of Physics London) p233 Walls M G 1988 PhD Thesis (University of Cambridge) Yagi K 1987 J Appl Cryst 20 147 Yao N, Wang Z L and Cowley J M 1989 Surface Science 208 533 ELECTROMAGNETIC RESONANCES IN SELF SIMILAR ARRAYS OF PARTICLES OR VOIDS
Francisco Claro Facultad de F/sica Pontificia Universidad Catolica de Chile Casilla 6177, Santiago, Chile
Ronald Fuchs Ames Laboratory, U.S. Department of Energy and Department of Physics, Iowa State University, Ames. Iowa 50011
Abstract. The dielectric response of a self similar cluster of metallic particles or random array of voids is considered. The response function obeys a recursive relation that yields a self similar distribution of modes. Their strength is also found to exhibit self similar properties. 10
I. INTRODUCTION. It is well known thai a variety of surface and bulk inhomogeneities can be well de- scribed by self similar models [1]. These models are constructed so that through several stages of amplification the details in the structure have the same geometrical features (2}. In this work we consider two such cases, a self similar cluster of metallic spheres and a random array of voids in a metallic background. Examples of such structures occur is DLA samples {3] and porous rocks (4). We are interested in the response of the system to the passage of a charged particle [5] or as electromagnetic wave, as described by an effective polarizability or average dielectric constant. n. SELF SIMILAR CLUSTER. Consider a nearly spherical cluster (the cell) of N metallic spheres of radii a, distributed following some geometrical rule. The radius of the cluster is ai =
We have assumed the particles are not too close (center to center separation greater than 11
three particle radii {8]) so that the dipole approximation may be applied, although a
multipolar treatment is equally possible. Ct and n, are the strength and depolarization factor of mode 8. These numbers depend only on geometry and are defined by the original geometrical ruie for building the original cluster, the cell For instance, the cell might be a regular octahedron with a sphere at each vertex. In such case there are only two active modes so that s = 1,2, only. The strengths obey the sum rule JT, C, = 1. The relation between the poiarizability of the effective sphere at stage t-f 1 and its dielectric susceptibility is
Equating (1) and (2) we obtain
Vi+i =
3 l where F ~ N/cr is the stage filling fraction, b, = ns - 1/3 and y - -\{4*x)~ + Equation (3) represents an iterative map that is to be iterated up to i = I -1. The normal modes are the poles in the effective poiarizability, or the zeroes of yj. ffl. SELF SIMILAR ARRAY OF VOIDS. Consider a continuous metallic medium where tiny holes distributed at random and covering a volume fraction F are punched Larger holes covering the same filling fraction F are punched next also at Because of disorder there is no exact theory to obtain the dielectric response of the system. In a dilute sample the effective medium theory may be used. Defining aa effective dielectric constant for the medium at each stage of iteration this theory- gives [9].
where as before, F is the filling fraction at the stage (voids punched at stage i in a 12
background assumed uniform). Some algebra yields from (4)
where z = -{e — I)"1 = -(4arx)~1> Ci — 2/(3 - ^) is tne strength of the percolating
mode, C2 = (1 - f )/(3 ~ F), and ft = 1 - F/3. This expression is identical in form with Eq. (3), with « = 1,2, and must be iterated up to t = / - 1. The zeroes of zj are the resonances of the system.
IV. SPECTRAL DENSITY. Equations (3} and (5) have identical structure. Their physically significant solutions are the zeroes contained between the fixed points. They constitute a self similar set |10] and at stage I, number 2(2' -1). The strengths of these modes may be displayed through the spectral density function G[u) defined by jll],
where u is either y or z, and i is the overall filling fraction at state I. Note that G(u) is the
analog of the strengths Cs in Eqs. (3) and (5), and for discrete modes may be understood as a distribution. It depends only on geometry and its integral equals unity. One obtains form (6) by inversion 1 Gt(u) = Ijmlmfriutiu - iS)]- . (7)
For the cases discussed in Sees. II and HI this function has self similar properties {10]. Figure (1) shows the spectral density for a porous metal (Eq. (5)). It was obtained after
14 iterations with F = 0.05 {V. CONCLUSIONS.
We have shown that a self similar cluster of spheres, or array of voids gives, rise to an also self similar dielectric response. This property is exhibited by the spectral density function in terms of which an effective polarizability or dielectric constant may be defined. Details of this work and further discussion of the properties of the spectral density will be published elsewhere [10].
The Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under Contract No. W-7405-Eng-82. This investigation was supportedby the Director for Energy Research, Office of Basic Energy Sciences, U.S. Department of Energy, and by the Fondo Nacional de Investigaciones Cient/ficas y Tecnologicas (Grant No. 0679).
REFERENCES 1 For a review see S. H. Liu, Solid State Physics 30, 207, F Seitz, D. Turnbull editors (Academic Press, 1986); R. Orbach, Annu. Rev. Mater. Sci. 19, 497 (1989). 2 B. B. Mandebrot, "The fractal geometry of nature" (Freeman, San Francisco, 1983).
3 T. A. Witten, Jr. and L. M. Sander, Phys. Rev. Lett. 17, 1400 (1981). 4 A. J. Katz and A. E. Thompson, Phys. Rev. Lett. 54, 1325 (1985).
5 T. L. Ferrel and P. M. Echenique, Phys. Rev. Lett. 55,1526 (1985).
6 R. Rojas and F. Claro, Phys. Rev. B34, 3730 (1986).
7 R. Fuchs and F. Claro, Phys. Rev. B39, 3875 (1989).
8 F. Claro, Phys. Rev. B_25. 7875 (1982). 9 For a review see R. Landauer, AIP Conference Proceedings No. 40, C. W. Garland and D. Tanner editors (AIP, New York, 1978) p. 2.
10 R. Fuchs and F. Claro, to be published.
11 D. J. Bergman, Phys. Rev. 14, 4304 (1976). 14
o © OT CD Kti Ml US »i' C o "o 3
0.00 0.25 0.50 0.75 1.00 factor de depolarizacion 15
Influence of the Wake Potential on Inelastic Scattering Using A Small Probe in Confined Volumes
P.F. Batson and J. Brulcy
IBM Thomas J. Watson Research Center Vorktown Heights, New York 10598
When a fast electron penetrates a solid at a velocity, v, greater than the Fermi velocity, v,., a wake potential is confined within l-2nm in the charge density "wake" is created. The wake results lateral direction. Thus, if inelastic scattering from the hydrodynamic response of the system to occurs with an impact parameter smaller than the coulomb potential of the fast electron. In l-2nm, we might expect the wake potential to Figure 1, we reproduce results from Ritchie and influence the amount and type of the scattering. Fchcniquc1 for the potential and differential charge density which corresponds to the wake. These results are for a heavy, positive ion, but should be applicable to the light, fast electron with somewhat different length and density scales. The potential is highly anisotropic, having a wavelength of 2nvjwp behind the fast electron and parallel to its trajectory, and a wavelength of h/mv ahead and to the side of the trajectory. In effect, the wake is a real space representation of the plasmon and quasi-particlc continuum. For -as instance, at a certain distance behind the fast electron, the potential is mainly plasmon-like. Moving to the side, this plasmon potential becomes shorter in wavelength. When we reach the wake position at the wake angle 0 = sin-'vf/v, the plasmon potential has a wavelength similar to single quasi-particies, and the plasmon merges with the single particle continuum to the side of and in front of the fast electron. oca Scaling from results for fast ions, we expect that the depth of the wake potential will be of order e«>p/v ~ 0.25 eV for Silicon at lOOKcV incident electron energy, increasing to 0.5 eV for diamond. The integrated response charge which •• j. is responsible for this potential equals the fast electron charge within 10-20nm behind the incident electron. Therefore, 5-10nm behind the Figure 1. Wake potential and differentia! charge fast electron, the total potential will be highly density reproduced from Ritchie and Rchcniquc, distorted by the wake. On the other hand, the (1982). 16
Figure 2 shows typical energy loss results for silicon over the energy range 0-140 eV. The no loss electrons arc confined to about 0.3eV about zero energy-. The bulk plasmon occurs at 16.7 eV, and in this case we have a second peak due to multiple plasmon scattering. The Si 1^ 3 edge occurs at 99.84+0.02cV, and is sitting on a decaying backg-ound comprised of multiple single particle excitations. The present work is loosely aimed at using the Si l^-? edge as a probe of changes in the conduction band density of states as we shift the electron beam about the specimen. Figure 3 shows the l^j edge alone after stripping the background intensity. There I have stripped also the 1^ part away to reveal the shape which results from variations in the final states. In principle we want to interpret these in terms of final density of states in the crystal. In practice, there is a difficulty due to the presence of the charged hole left behind by the excitation. Therefore, the final states are relaxed downwards, 50 0 50 100 150 and a discrete hydrogenic series of lines may be formed below the ground state conduction band Energy Loss (eV) edge. A recent analysis of this core excitonic effect has been given by ("arson and Schnathirly2 Figure 2. Typical energy loss results for Si.
1 1
Si L2<3 Figure 3. Background stripped vi Si I.J7 absorption edge. The spin 3/2 component has been NX" Part extracted by a Fourier deconvolution of two impulse i functions having a reparation 0.61 cV and a relative weighting VB of 2:1.
> . • ••• CB L, L3
i • 1 , , i 1 , 98 99 100 101 102 103 104 Energy Loss (eV) 17
for several semiconductors. In Si the excitonic Winding energy is about 5()nicV, while in diamond it is about 0 2eY. I'hus, we expect that in silicon the excitonic distortion should be small on a scale of 0.2-Silicon
100 105 110 11 Energy Loss (eV)
Figure 5. Variation of the L3 part of the absorption with the local thickness of a thin wedge. The fine structure at the edge becomes less prominent in the thinner areas, but the edge docs not shift. The presence of SiOj cannot account for the differences.
In Figure 5 we present results for the Si edge 98 100 102 104 106 108 as a function of specimen thickness. A curious Energy Loss (eV) result emerges. Ilie structure at the near edge, identified above as being due to the conduction Figure 4. Comparison of measured Si absorption band DOS, becomes less distinct in specimen with the s- and d-projected density of final states areas which are thinner than about 15-20 nm. from Reference 3. The differentiated spectrum The shape differences do not appear to be due to shows the criterion used to determine the the Si()2 nor, from other comparisons, from absolute position of the onset with an accuracy amorphous Si. High resolution images of the of + 20meV. specimen in this region showed good quality lattice images, as well. Certainly, a 15-20 nm Figure 4 shows calculated1 s- and d-projectcd thickness comprises many unit cells, so this is not DOS for Si compared to the measured edge. a real electronic structure confinement effect. It Quite a thorough comparison is possible. The is known that a conduction electron having an bottom part of this figure shows the criterion used energy 2-3eV above the conduction band edge for determining the edge onset position. This will have an inelastic mean free path of several differential method is not sensitive to shape 10's of nanometers, so it may be reasonable lhat variations in the no loss beam, and so may be damping by collision with the surfaces would used to obtain the position of the edge with an occur at these thicknesses. But evidence below accuracy of about ;f20meV. I he edge intensity suggests that something more complicated may can be obtained witli better than 2% be occuring. reproducibility when the probe i* not moved. 18
In a thick region of the sample, near a broken edge, a similar behavior is observed. In Figure 6, as the probe is moved closer lo the edge, the fine structure near the absorption onset is lost. fcj l< ••> 6 Si L2>3 y/AV i b=10nm ^*** - s*"VVv*' 5nm ,-y* 0 50 100 15 2nm Energy Loss (eV) /*^ Figure 7. Ixw loss plus core absorption for several thickness's. Multiple plasmon scattering i . . . i . . . is evident. 95 100 105 110 115 Energy Loss (eV) Figure 7 shows multiple scattering results for several thickness's in the energy range 0-110 cV. Figure 6. Variation of the I ^ scattering with In Figure 8, we have performed the single 3 5 impact parameter relative to a cleaved edge in a scattering analysis from Miscll and Jones to thick region. Fine structure disappears near the obtain the single scattering. This is commonly edge. referred to as a log deconvolution" today and is given in Fourier space (denoted by ~) But in this geometry, the probe must be moved very close to the edge. (I have pushed this limit down to <0.5nm recently. 1/90) How can ann\. = In rz— we account for the difference in behavior between the "thickness" in the directions parallel and h. perpendicular to the incident beam direction? Perhaps, the anisotropic wake potential where the a is the specimen thickness, n is the influences the final state of the core absorption, number density of scattercrs, where P^ is the scattering per unit length for an 0.06 infinite foil, Pb, is the bulk plasmon correction, P(,2 is the surface intensity, and -Pbi-jfc-ttfn.f-i.S7J. 0.04 - UJ PM = 2-jfc-t[Ai-j--l.57] = -2Pw.
Therefore we may write 0.02
; P|*lasmon ~ 2 0.00 40 60 80 100 120 so that Energy Loss (eV) 1 r» Figure 8. Single scattering analysis of the data in Figure 7. Thus, the true thickness, a, can be obtained from thc arithmetic mean of thc thickness's derived When the specimen area is fairly thick, these from the measurements of thc plasmon and total two methods yield the same answer. 'ITiercforc, scattering intensities. Figure 9 shows plots of r tnc we may integrate anos between 12-21cV and then aP^, PTot and Ppiasmnn f° range of multiply by 12()nm to obtain a plasmon thickness's in this experiment. Measured determined thickness of 35nm for the thickest intensities for thc total scattering (closed circles) area. Conversely, we obtain 36nm using the total and for thc plasmon scattering (open circles) arc integrated scattering. When we go to thinner included. As expected, thc plasmon intensity areas, thickness's derived from the two methods deviates below thc linearity, and thc total diverge. Thus, the bottom curve of Figure 8 intensity deviates above in a manner consistent yields 2.2nm using the plasmon intensity and with Ritchie's analysis. 9.1nm using the total intensity. 1.0 1 In Ritchie's 1957 paper" on inelastic scattering in thin films, he pointed out that the surface !B O.B- y. o - Plot plasmons on the top and bottom surfaces reduce 2 \ the intensity of the bulk plasmon below the value \ en 0.6 -, expected. This reduction has a logarithmic "Core Loss—y divergence in the film thickness given in units of the plasmon wavelength. On the other hand, the o 0.4 - V) surface scattering itself increases by twice as much 'Ptasmon as the reduction of the bulk scattering. This was I t described as a "depolarization" effect of the 5 0.2 or surfaces. A result is that the total scattering per unit thickness goes up logarithmically as the film 0.0 gets thinner, while the plasmon scattering per unit 100 200 300 400 thickness is reduced. Ritchie wrote for the total Specimen Thickness (Angstrom) scattering probability
: Figure 9. Results for thc scattering probability I lot bl as a function of thickness following thc analysis in the text. 20
from the fast electron trajectory fan now be understood as well, bemuse tin1 wavelength of ihc wake potential is much '•boiler in this direction. Thus, normal dipole selection rules are totally inadequate to describe the scattering.
What are some of Ilie implications of ihis picture? First, since the plasmon wake has a short range in the direction perpendicular to the fas! electron trajectory, its influence should be small lor large impart parameter scattering. Thus, small angle scattering should not be complicated by this effect. I nfnrtunatclv, the small scattering angle experiment requires large 98 100 102 104 106 108 110 areas of uniform thickness. These are Energy Loss (eV) experimentally difficult to obtain, and so a detailed check as a function of thickness would Figure 10. Si 1/j results normalized for scattering be difficult. It is also difficult to obtain a high probability at the various thicknesses derived enough angular resolution in the STFM using Figure 8. The data for 21.6 and 21.9 nm geometry. are superposed to show rcproducibilily of the results. Another possibility involves ihe core exciton. As discussed above, this is a localized state near the core hole. If the final state electron binding Jn Figure 10, we show the Si I,3 edges for the various spectra shown in Figures 7 and 8 above. energy is comparable to, or smaller than, the These have been normalized in intensity to the plasmon potential, we might expect to see some core intensity visible in Figure 8. For clarity in modification of the c.xcitonic contribution in presentation, the each result has been displaced favor of propagating collective final states. upwards. Notice the decay of the electronic structure information at thickness's less than Figure i 1 shows results for the K core about I5nm. The reproducibility shown in thai excitation for diamond. Two results arc shown, case is typical of what can now be obtained. In one from the partial photoyield experiment,'' and Figure 9, we have plotted the integrated core one from the spatially resolved energy loss scattering between 104 to 108 cV, normalized to experiment. 'The photoueld experiment, in the the total valence scattering at the 35nm thickness simplest interpretation gives the optical (closed diamonds). We find that the core loss absorption. It shows a classic excilonic line scattering closely follows the total valence which can be characterized with a binding energy scattering. Thus, it seems possible that the final of about 0.2 cV below a conduction band edi>e state for the core loss participates in the collective at 2R9.2 cV and with a lifetime broadening of behavior of the solid in the same way thai the 0.17cV. The line fit to this data in Figure 11 was valence electrons do. Therefore, the excitation generated by an Flliott"1 model for the exciton probability for the core loss is influenced by the following Altarelli and Dexter.11 The spatially top and bottom surfaces in the same way that the resolved energy loss case docs not show the sharp valence electron scattering is influenced. This is line. An Flliott model analysis yields a good fi1 precisely what we might expect if the applied to the data, but the conduction band edge and coulomb potential is modified by the presence of excitonic binding energy do not show a simple the plasmon wake within the material. correspondence to the phoioyield results. In particular, the edge is at a similar energy, but the modeled exciton binding energy is smaller, and The behavior of the core scattering when the the band edge is lower. Perhaps more significant, surface is located at some lateral distance avvav the lifetime broadening is greater in the electron 21
There is a possible problem with this analysis that should be mentioned. The Flliott theory was Photoyield written for electron hole interaction i/i the valence bands of semi-conductors. It was therefore set E= 0.189 up for a p-Hkc initial stale and an s-likc final state. r= 0.17 The diamond initial slate is the Is level. )t is not CBM= 289.2 clear if this invalidates the photoyicld analysis. At first glance, the relevant excitonic series should be the 2p, 3p, ... levels, with the n=l level missing. This may not be a problem fora weakly bound exciton in the Wannicr picture.
284 286 288 290 292 294 Energy Loss (eV)
Figure 11. Comparison of partial photoyield (Reference 9) with the spatially resolved energy loss for the diamond Is absorption. An Flliott model fit has been applied to each results. The fits indicate binding energy, damping, conduction band edge position and spectrometer resolution respectively. 100 101 10 case. This behavior might be understandable if Energy (eV) the plasmon wake can "compete" with the core hole potential in the selection of final states for Figure 12. Comparison of photoyicld (Reference the excited core electron. In effect, the core 12) and energy loss results for Si. Below, each electrons may be excited into an excitonic level, result is differentiated to obtain the edge position. but then immediately scattered away into The valence band maximum (VFJM) is obtained propagating states - almost as if the plasmon from XPS data. The SRFHLS result obtains the potential has "ripped" apart the hydrogenic final conduction band edge (CI)M). state of the exciton. Alternatively, the periodic potential of the wake might force excitations to As a result of this problem it is best to return final states at higher angular or linear momentum to the Si [-2 3 edge where the initial and final state that the excitonic ground state. These long symmetries arc similar to the case that FHioll wavelength states might still be perturbed by the modeled. Figure 12 shows a comparison of the core hole, bul would not be bound. It is photoyicld results with the energy loss results. significant here that the exciton binding energy Fach of the results arc differentiated to show the here is 0.2 eV, while the plasmon potential depth precise edge locations. The photo yield results is 0.25-0.5 cV. Thus we don't expect a clear-cut were obtained by Mimpscl12 with an energy divergence between the photon and electron case. resolution of about O.I5cV. The absolute energy The intermediate result that we show in Figure was derived from Fbcrlmdt1* with an accuracy } I does seem reasonable. of ±50mcV. The energy loss results were obtained for the spatially revived case with an 22
energy resolution of ().28eV and an accuracy of be able to excite the core electron without ±20meV. interference from the piasmon wake.
lx)oking at the photoyicld results first, we find We have done this and find that the piasmon an onset energy of 99.76cV. The differentiated intensity is dramatically smaller. (A factor of two results show two peaks, one corresponding to the in our particular case.) 1'igure 13 shows a typical edge onset and one 0.62 eV higher. The valence result. band maximum may be obtained from XPS [ i i i results and is placed at 98.72 cV. Adding the Plasmon Intensity Intensity Mormoiized To Incident Dose room temperature band gap of 1.12 cV to the \ valence band edge position yields a conduction band edge at 99.84 eV. Using the fvlliott model with this conduction band minimum, we can analyze the edge shape to derive an excitonic binding energy of 50meV. This is in good agreement with the results by soft x-ray emission.2
The energy loss results display similar differentiated structure, but the edge position is clearly shifted to higher energies. The reproducibility here is perhaps the best we can do, 0 20 40 60 and care is required. The average of many results Energy Loss (eV) places the edge onset at 99.85+0.fl2cV. Hgurc 13. Piasmon scattering intensity for the Now we should be convinced that the energy two-beam geometry described above. The loss results arc different, but we arc not sure of conditions arc indicated graphically with the the precise reason. We need to find a link vertical line indicating the (220) Kikuchi band, between the piasmon scattering and the shape of the x indicating the entrance beam angle and the the core edge. We have arranged an unusual o indicating the exit collection angle. Notice that scattering geometry, which, to our knowledge, interchanging the exciting and collection angles has not been tried before in probing plasmons. docs not alter the result. The peak at 17cV in the We tilt the crystal in selected area diffraction so bottom curve is due to quasi-clastic plus piasmon that the exciting electron beam is aligned just multiple scattering and should not be affected by outside of the (220) two-beam diffracting the channeling conditions. position. We place the collection aperture so that it defines a direction just inside of the two-beam Figure 14 shows the core loss intensity for the condition, the crystal wavefunctions thus have, parity flipped excitation compared to the respectively, odd and even symmetry with respect symmetric results — again indicating excitation to the atom columns in the crystal. (This is a and collection conditions graphically. The results well characterized result of dynamical diffraction arc normalized between 104-106cV on the theory.) Since the bulk piasmon is a scalar response and since the coulomb interaction is assumption that cxcilonic differences will be symmetric about the origin, we expect the matrix negligible in that range. Notice the dramatic element for piasmon excitation to be zero for enhancement of the near edge structure. There initial and final fast electron slates which have is also a minisculc shift downwards of the edge opposite parity. However, a core excitation position. A simple subtraction of the symmetric wherein the angular momentum changes by results from the parity flipped results yields a A/; = + l, will be allowed. Therefore, we should remarkable agreement with the photo-yield data. Both the position and the shape of the results agree nicely. 1 i 1
Si L2,3 Photoyield ,. .
cri Y . co i J. EELS Enhancement /I During Channeling •
98 99 100 101 102 102 104 106 Energy Loss (eV) Energy Loss (eV) Figure 15. The difference between the results in Figure 14. The I^3 core loss scattering in the Figure 14 is displayed here compared with the off-two-bcam geometry described above partial photoyield. The similarity is striking and compared with the symmetric scattering results. the edge onset position is reproduced. The "flat" Notice that the edge is enhanced in the parity top of the photon case is reproduced also. flipped channeling case -- when the plasmon is depressed. There is a small shift in the edge also. In conclusion, the inelastic electron scattering appears to be qualitatively different from the In Figure 15, we show the difference plotted photon absorption, because the collective on and expanded scale compared with the response of the solid can take part in the inelastic Photo-yield results. The rcsemblcnce is striking. scattering in a way which favors transitions to Thus, it appears that the difference shown in final states which ordinarily arc not accessible in Figure 14 above, may be an isolation of only the the optical experiment. The results, at least for core excitonic part of the Si 1^ 3 edge. It seems well screened materials where the core exciton is astonishing that a 50meV exciton is able to distort weakly bound, closely resemble the ground state the spectrum by this amount. Future work will projected density of slates. 'i"hus, the energy loss aim to try to understand the precise origin of the may well be uscablc for determining conduction difference. bandstructure in very small volumes. References: 1. R.H. Rilchic, and P.M. Kchcriiquc, Phil. Mag. A 45 347 (19R2). 2. R.D. Carson and S.H. Schnatlcrly, Phys. Rev. led. 59 319 (1972). 3. X. Wcng, and P.Rez. Unpublished. 4. P.r. Balson. Ullramicroscnpy. 9. 277 (1982), and Phys. Rev. 1 .ell., 49, 936 (1982). 5. D. Misell and A.F. Jones, J.Phys. A 2 540 (1969). 6. P. Xu and J. Silcox, private communication. 7. M. Sarikaya and P.Rcz, Proc. 40lh HMSA (cd. Bailey, San Fransisco Press, 1983) p.486. 8. R.H. Ritchie, Phys. Rev. |06 874 (1957). 9. J.F. Morar. F.J. Ilimpscl. CJ. llollingcr. G. Hughes, and J.I.. Jordan, Iliys. Rev. Lett., 54,19M5 (1985). 10. RJ. F.lliolt, Phys. Rev. 108 1384 (1957). 11. M. Altnrelli and 1X1 . OcxUr. Phys. Rev. l.clt. 29 1100 (1972). 12. F.J. Ilimpscl, unpublished. 13. W. Fbcrhardl. (i. Kalkoircn, C. Kun/. I). Aspncs. atid M. Cardona, PJiys. Slat. Sol. (h) fiR 135 (I97R). 25
Presented to 12° Werner Brandt Workshop "PENETRATION OF CHARGED PARTICLES IN MATTER". 3-7 September 1989, San Sebastian, Spain.
Electron microscopy study of the electronic properties of small silicon particles
Daniel UGARTE, Christian COLLIEX, and Pierre TREBBIA. Lab. de Phys. Sol., Univ. Paris XI, Bat. 510,91405 ORSAY-FRANCE
INTRODUCTION
The problem of the energy lost by a fast electron interacting with a small particle has been studied for many years. Small objects have shown a behaviour substantially different from bulk materials. Moreover the electronic properties of small particles arc very important from a technological point of view because of their possible application in catalysis. Recently the advent of the Scanning Transmission Electron Microscope (STEM), which provides a focussed 100 kV electron beam within a diameter of less than 0.8 nm, has renewed the interest in the energy losses suffered by a fast electron travelling along a well defined trajectory in the vicinity of a small particle. Most of the theoretical work related to STEM measurements has been devoted to the excitation of plasmons in small spherical particles[l-9] or related gcometriesf 10,11]. Although many of the theoretical formalisms developed for small spheres are generally valid for any arbitrary dielectric function, the STEM experimental measurements have dealt mostly with metals, and their dielectric constants were usually modelized by the Drude expression for a free electron gas [11-13]. The aim of this work is to test the validity of the classical dielectric theory [14] for a more complex case. Here, we will be concerned with the electronic properties of spherical silicon particles. This material and its oxides arc very well characterized, because of their
applications in microelectronics. The dielectric properties of Si and its oxides (SiO2, SiO) present complex structures [15], and provide a stringent test of the theory developed up to now. We study in a STEM the low energy loss spectrum of silicon particles, and also the spatial localization of the most important spectral features observed, using digitally acquired energy filtered images. The results are compared to the predictions of the classical dielectric theory. An anomalous peak at 3-4 eV is observed at the outermost surface of ihc SiOjj layer surrounding the particles. We conclude that this phenomenon is due to a beam induced process, which renders the SiO2 surface conducting. We modelize this effect by including a very thin coating of amorphous silicon, supposed to be produced by electron induced oxygen desorption. The theoretical spectra and images then fit the experimental results corrccdy. 26
MATERIALS AND METHODS
The experimental work was performed on the VG HB501 dedicated STEM operating in Orsay. The microscope is fitted with a serial second order corrected spectrometer (Gatan 607). Both the EELS spectra and the energy filtered images were acquired with a 0.7 run probe of 7.5 mrad convergent semiangle, and a collection aperture providing a 6.5 mrad collection semiangle. The small semiconducting panicles were prepared by £ gas evaporation method [16]. Figure 1 shows an electron micrograph of a group of Si particles. They are nearly spherical in shape, their size range is 10-300 nm in diameter, and they are covered by a thin (2-6 nm) natural amorphous oxide layer, produced by exposure to the atmosphere during preparation for microscope observation. The particles are deposited on a holey carbon film, and the experimental work was performed only on particles over a hole, in order to eliminate the carbon substrate contribution to the spectra.
EXPERIMENTAL RESULTS
A typical low loss spectrum is shown in Figure 2. In order to study the spatial distribution of the main spectral features observed, we acquired the energy filtered images shown in Figure 3. For an easy comparison of the localization of the excitations, we have extracted line profile intensities from the images (Figure 4).
The analysis of these data, together with our previous knowledge of Si-SiO2 plane interfaces [17]), yields the following immediate conclusions:
- The 17 eV and the 23 eV peaks are due to the bulk plasmon of Si and SiO2 respectively. - The 9 eV contribution is due to the normal surface plasmon of the Si particle. As predicted in emulations applied to Al oxidized parbcles[5], the surface plasmon energy of a coated sphere should be found at an energy lower than that of an isolated free sphere in vacuum (for a Si
sphere the dipole mode (£=l) frequency is - The 10.5 and the 12.0 eV minor features are due to SiO2 interband transitions. - Finally the low energy peak at 3-4 eV is an unexpected result, since a careful look at Figure 4, shows that it is localized at the external surface of the Si-Oxide layer surrounding the particle. 27
DISCUSSION
A similar experiment has been reported by Batson [12]. He measured a low energy loss peak at 3-4 eV, on oxidized aluminium particles, and showed that it was the result of the coupling of the surface plasmons of neighbouring particles [10]. In our case the spatial localization of this low energy excitation does not show the dipole distribution that he observed, and it can be seen on all the particles. The localization of this energy loss at the outermost surface of the oxide layer is an anomalous phenomenon, because it is well known that insulators cannot have any
absorption at energy values in the gap (8.9 eV for SiO2), and certainly cannot sustain surface plasmons in this energy range. This suggests the existence of some land of induced surface phenomenon (charge, contamination, radiation damage, etc.). Using the classical dielectric theory [14] in the non retarded case, we have extended the coated particle model [6] to include the effect of a second coating layer L2 (Figure 5). The expressions derived are rather long and complicated and they will be presented in a forthcoming publication. If we modelize the second coating as a contamination layer of amorphous carbon, no important effect is observed in the low loss range. This fact can be understood easily by considering the surface plasmon at a plane interface between two materials 0 and b. The
condition for resonance is £ia(ca)+e1^((o)=O [18], i.e. if we consider the medium a as
SiO2 (£i(0-7 eV) = 2.2) in order to have a plasmon at the interface in the 3-4 eV region, the medium b must be a conductor or at least a semiconductor whose dielectric permittivity presents negative values (£jfc(3-4 eV)< 0). Although the conditions of resonance in the spherical geometry are more complicated, this consideration still holds. The amorphous-C is a semiconducting material [19], and in the low energy region £j > 0, so an amorphous carbon layer on the coated silicon particle will not create a new oscillation in this energy range. The effect of oxygen desorption induced by electron irradiation of oxides is a well known phenomenon in electron microscopy work. The STEM analysis usually requires quite a high dose in order to get a good signal to noise ratio, in our work for example one can estimate doses as high as 108 e'/nm2 for the series of images in Figure 3. If we make the
assumption that the SiO2 surface undergoes a beam damage process, and lose some amount of oxygen, a thin conducting film could be produced on the surface. In order to modelize this phenomenon, we assume that we get an amorphous surface silicon layer as a result of oxygen
desorption from the SiO2 surface. The calculation using the experimental parameters of Figure 3 for the panicle radius, and the Si core, assuming a 0.4 nm amorphous Si surface layer is shown in Figure 6 for the spectrum, and in Figure 7 for the probability of excitation as a function of the impact parameter, at the energy losses of the experimental energy filtered images (Figure 4). The thin semiconducting film has a dramatic effect, and the theoretical results fit correctly the experimental measurements. The intensity of the 3 eV experimental profile (Figure 4) is higher than predicted in Figure 7, this difference may be accounted for, by considering the 28
important contribution of the zero-loss peak tail in this energy *ange. In all the calculations of this work many multipoles have been included (&15 for the curves of Figure 6 and &10 for that of Figure 7), and experimental values have been used [IS] to describe the dielectric permittivity. A conducting surface layer covering an insulating sphere sustains two surface plasmons [20], one at the external surface, and a second at the internal interface, like that of a void If the thickness of this film becomes small enough, the two surface excitations are coupled, as for the case of a thin film [21]. Thus two new frequencies are created co " and a>+; the co" branch goes to a>=0 when the film thickness goes to 0. The low energy peak observed here can therefore be interpreted as the co " branch of the coupled surface excitation on the amorphous-Si surface layer covering the particle. This would explain its localization on die outermost surface of the sphere.
CONCLUSIONS
The classical dielectric theory has been tested in the case of small semiconducting particles. The predictions of the loss probability show a close agreement with the experimental results, even for minor details, provided that we use the experimental values to describe the dielectric permittivity of the analysed materials: the Drude model must be replaced by a more complex representation taking into account real damping. The STEM is specially well suited to the study of small volumes. The energy filtered imaging method applied to surface plasmons is a very sensitive tool for interface visualization. The ability to make the correlation between spectrum and energy filtered image features is of great help for a correct understanding of the physical processes involved. The high dose needed in STEM studies may induce important damage effects, and much care must be taken in order to avoid incorrect conclusions.
ACKNOWLEDGEMENTS
Thanks are due to S. Iijima for providing the specimen for this work. The invaluable help of M. Tencc" for the manipulation of the digital control of the STEM is gratefully recognized. The authors wish to thank M. Walls for his critical reading of the manuscript. REFERENCES
1.- SCHMEITS M., J. Phys. C: Sol St. Phys.,14, 1203, (1981). 2.- KOHL H., UInran.,11, 53, (1983). 3.- FERREL T.L. Figure 1. Conventional electron microscopy image of a group of silicon particles.
Figure 2. Typical low energy' loss spectrum of silicon particles. 31
3 eW 9 eV
Figure 3. Family of energy filtered images of a small silicon particle lying on a bigger one. The image marked IS eV actually corresponds to 17 eV (miscallibratJon of the spectrometer).
xO.33 (arb)
>
— 3eV — 9eV — 17 eV "" 23eV
•
Figure 4. Line profile intensities extracted from the energy filtered images of Figure 3. Figure 5. Electron moving with a velocity v at a impact parameter d of a double coated sphere of exterior radius rj, with a first coating layer LI of thickness (r2-r: -t) and a second coating layer L2 of thickness t.
Figure 6. Calculated spectrum (r2=17.1 nm, r3 =31.63 nm, d=37.3 nm, electron energy 100 kV). a) Si core coated with a SiO2 layer. b) Idem, including also a 0.4 nm thick amorphous silicon layer (L2) on the surface. I .. *0.33 (arb) — 3eV — 9eV — 17 eV —' 23eV
30 [nm]
Figure 7. Calculated energy loss probability as a function of impact parameter, for the case of a Si sphere covered with a layer (LI) of SiCfe, and an amorphous-Si layer (L2) on the surface. (r2=17.1 nm, rj=l 1.36 nm, t= 0.4 nm, electron energy 100 kV). 35
Electron Compton scattering in the electron microscope
P. Schattschneider and P. Pongratz Inst. f. Angewandtt und Teehnitehe Physik, Tcchn. Univ. A-1040 Vienna, Auttria
Abstract The distribution of electron momenta (electron density in momentum representa- tion) of gases can be probed by Compton scattering of either photons fa-rays or X-rays) or electrons. Recently it has been shown that Compton scattering of electrons is suited to the study of the electron momentum densities of solids on a microscopic scale. This technique, known as ECOSS, Electron Compton Scattering from Solids can be done in the electron microscope by Electron Energy Loss Spectroscopy (EELS). The advantages and disadvantages of ECOSS as compared to photon Compton Scattering are discussed. It is shown that ECOSS is a method suitable to obtain infor- mation on the ground state wave function in condensed matter. After a discussion of inherent approximations, especially the impulse approxima- tion, the reciprocal form factor is introduced. A method is proposed in order to cope with the main difficulty, namely multiple scattering. important applications of ECOSS are the study of anisotropy of momentum den- sities; correlation effects of conduction electrons in metals; and charge transfer in alloys.
Keywords: EELS, Electron Compton scattering, impulse approximation, electron momentum density, reciprocal form factor, multiple scattering, anisotropy, electron cor- relation. 36
Introduction
Electron Compton scattering is inelastic scattering of fast electrons at large an- gles (~ lOOmrad) off core or valence electrons. Like in the photon case, the energy of the scattered electron is increasingly lowered with scattering angle; the energy distri- bution has a halfwidth proportional to the mean squared momentum of the scatterer (Doppler broadening). The shape of the distribution can be shown to be an image of the momentum distribution of the electron in the ground state. The electron microscope is well suited to the study of the electron momentum densities of solids on a microscopic scale by this technique, known as ECOSS, Electron Compton Scattering from Solids. (10-12], In terms of electron microscopy, ECOSS is EELS in diffraction mode in the energy range of the Bethe ridge. Since the cross section for scattering of electrons is some five orders of magnitude larger than that of photons at the respective scattering angles, and since the intensity of electron sources is ~ 102 larger than for typical photon sources, it is evident that the sensitivity of ECOSS enormously exceeds that of photon Compton scattering, or to put it another way, the duration of an experiment can be reduced from days (for photon scattering) to hours or minutes. Another, and maybe the advantage is that, contrary to the photon case, ex- tremely high spatial resolution can be achieved in conventional Transmission electron microscopes (TEMs), which is important for the investigation of anisotropy in fine cry- stalline material. For instance, the anisotropy of the electron distribution in graphite was measured on a single graphite grain [9]. It is obvious that the electron microscope offers a number of instrumental features such as simultaneous monitoring of diffraction patterns, dark field images and variable magnification which are not available in the photon case. Modern electron optics have momentum resolutions of up to ~ 0.04a.u. which is a factor of five better than that of photon sources for Compton experiments. ECOSS is a relatively new technique. Radiation damage of the specimen and mul- 37
tiple scattering restricted application mainly to gaseous specimens. The first ECOSS experiment was reported in 1981 [10]. The experimental and theoretical work perfor- med so far concentrates on basic questions, simulations and comparison with photon Compton scattering {10-11,13, 8].
Theory
In first order perturbation theory the dynamical form factor S is
w relates to the energy E lost by the scattered electron as E = hut, the Ha-
miltonian B = Ho + V consists of an unperturbed part Ho and the potential of the binding forces V, p, is the density operator, and \
the ground state wave function of the scatterer. It can be shown [5] that for high energy transfer, in the power series
±iHt of the exponentials e , all terms cancel except for that containing Ho. Physically, this meaqns thatthe potential in which the electron moves can be considered constant for the short time of interaction in a high energy transfer, and eq. (1) becomes
S = ~ / t~lut <
t'Kp-,\tp > di. (2) 2irn J
By use of the time dependent Heisenberg operator
this is
The momentum density operator p, (t) is the Fourier transform of the ordinary density operator n(r) n(r) = v>(r)v>+{r) (5) where xj){r) are field operators. From the Wiener-Khintchin theorem, we know that the Fourier transform of the product in eq. (5) is the autocorrelation of the fieltl operators in momentum representation, so
< *(«)P; (o) >= £ < < (*K-M(*K-+«V > (•) p.r1
where the sum is over p-states within the Fermi sphere, a*, o are creation and an- nihilation operators. For large momentum transfer q, which is assumed henceforth, the final state electron can be considered to be free. For free electrons, a*{t) = a* (6)exp(—ip2/2m)t, and we may write
(7}
The second aspect of large momentum transfer is that the operator «£«+, creates a par- ticle in the previously empty state p' + q far beyond the Fermi momentum. Subsequent annihilation of a particle in state p + q is only possible when there is a particle, i. e. only for p = p' otherwise the probability amplitude for this process vanishes. See fig, 1. Hence J°
Eventually, (K (o) >=
Replacing the sum in eq. (9) by an integral (j£p -»/ ^fr) «»d the time integral in eq. (2) by an appropriate ^-function yields
where we have introduced the momentum distribution 39
The momentum distribution in the many-particle system is defined as
3 ...d pnx{Pi, -Pn )x' (Pl» -A) ( .
x{p) is the one electron wave function of the ground state in momentum repre- sentation. It should be noticed here that the last equality does not strictly hold in an interacting n-particle system where a one-electron wave function does not exist in a strict sense. Rather, one should use a density matrix formulation. Though, we identify p(p) with the square of a one-electron wave function henceforth. In the last step of the derivation, we integrate out the ^-function and use the relationship [11] between 5 and the differential cross section d7a/ddEil, d'o _\2mt>]2 k fd?p m kb f xy Em
Here, ka kb are the electron wave numbers of the incident and scattered electron respecti-
vely, E is the energy loss, and q is the momentum transfer. The variable pt is the momentum component in the direction of q. The scattering angle 0 relates to the quantities on the right by the scattering geometry (fig. l.b).
The quantity J{px) in eq. (13), called Compton profile
np,) (14) can be derived directly from experiment. It is a projection of the 3-dimensional mo- mentum density of the scatterer onto the direction of the scattering vector g. It is, in principle, possible to obtain the complete 3-dimensional distribution p{p) from a series of Compton experiments [6]. For fast probe electrons, then, and when the energy and momentum transfer in the interaction is large, the target electrons in the ground state can be treated as if they were free, but having a momentum distribution as if they were bound. This is the essence of the impulse approximation (IA), valid for large energy and momentum transfer in the Compton event. The reciprocal form factor
Compton scattering provides information on the momentum density of electrons in the specimen. However, the same is true for elastic scattering which yields the modulus of the static form factor (the Fourier transform of the particle density):
(15)
Taken for granted that the phase problem can be solved (determination of the phase of F) the question is legitimate whether F[p) and p{p) contain different physical informa- tion. Eq. (15) says that F is the Fourier transform of p. Identifying p with the particle density in terms of one-electron wave functions,
P(r)=¥>(r>-(r), (16)
(see the comment given after eq. (12)) , we can apply the Wiener-Khintchin theorem stating that F is the autocorrelation function of the wave function x in momentum representation I (17)
On the other hand, the quantity derived from Compton scattering is
i.e. the diagonal element of the density matrix in momentum representation. F and p contain cemplementary information. The latter is the probability of finding an electron with momentum p, and doesn't give any clue to the phase of x- The static structure factor F on the other hand contains information on the phases of wave functions, via the autocorrelation integral, but x cannot be derived uniquely from F. The Fourier transforms of eqs. (17) and (18) are
' (f) B(f) = JNote the formal symmetry of eqs. (18, 19) and eqs. (17, 20). B is called reciprocal form factor because of its similarity to F. In complete ana- logy to the statement on the momentum space quantities F, p it can be said that elastic scattering yields the diagonal element of the density matrix in real space (probability of finding an electron at position r) without any clue to the phases of wave functions. Compton scattering yields, via the reciprocal form factor B, information on the phase relations of the wave function
Contributions to the Compton Profile
Core electrons
The simple relationship between the momentum density and the differential cross section does no longer hold when the impulse approximation breaks down . One has to use a more accurate wave function for the final state in the matrix element. The exact hydrogenic (EH) approximation takes both <£>, and Conduction electrons
For the conduction electrons the simple free electron model can often be used. In the following example, this model was applied to polycristalline aluminum. From eq. (14) we get for the CP, by projecting the states within the Fermi sphere onto the direction p,, the characteristic free electron parabola falling to zero at the Fermi mo-
mentum ±pr. See fig. 3. In contrast to the L-electrons, the IA is valid for conduction electrons.
The Problem of Multiple Scattering
Aside from the Compton effect, other electronic excitations within the solid must be considered. Low-angle scattering off valence electrons gives rise to plasmon excita- tions peaking in the low energy loss region. Such a plasmon excitation can be followed by a Compton event. An equally dominant contribution to multiple scattering in the energy and momentum range of typical Compton experiments arises from combined Bragg-Compton scattering. The Bragg spots or rings and plasmon peaks can be consi- dered as new sources for Compton events. In the typical range of Compton scattering, multiple scattering is caused by elastic scattering into high angles followed by a Compton event. Since these Compton events correspond to scattering angles different from the single Compton scattering angle new Compton profiles are generated with different maxima and width. They overlap with the single Compton profile and alter the position of the maximum and its width. A third coupling mechanism which may be important involves diffuse quasi-elastic scattering which produces a slowiy varying intensity distribution in the diffraction plane. All these scattering events give rise to a background of 50% to 100% of the Compton peak value, and to serious masking of the line profile. The measured spectrum may 43
look completely erratic as was shown by a numerical simulation of CPs by Williams et al [13]. See fig.ll.
At present, a satisfactory solution of the multiple scattering problem is
lacking. However, there are some promising approaches which will be discussed in
the following.
Bragg scattering
To determine the Bragg scattering intensity, a density distribution in the dif- fraction pattern can be used. When there is a continuum of scattering angles for the coupling of Bragg and Compton events—as is the case in polycrystalline specimens with well defined Bragg rings—the double scattering contribution forms a smoothly decre- asing background dominated by the ionisation event with the smallest scattering angle. This is why the background approximately follows a power law dependence in energy A • E~', well known from inner-shell losses. See fig. 4 The coefficient A takes on a wide range of values, but s is generally in the range 2-6 [12]. Increasing specimen thickness lowers the value of s due to plural scattering contributions, and increasing energy loss increases this value. Due to this dependence of the coefficients A and s, they should be determined at each ionization edge. Usually the energy dependence of the background is measured over a fitting region immediately preceding the edge. This procedure is only approximately valid in practice since the io- nization threshold cannot be identified exactly due to multiple scattering contributions. One method, although not rigorous, to investigate the background behaviour is to fit a model calculation of the Compton profile, including multiple scattering contri- butions and a power law background, to the measured profile. (See the demonstration experiment). In most cases, single crystals are of interest for ECOSS. So, one has to face the difficulty of extracting a single Compton profile from a measured superposition like fig. 11. Note also the difference to fig. 4. This problem could be tackled as described in the following: The intensity M measured at a particular energy loss E is a linear combination of coupled Bragg-Compton events
E), (21)
where P3 is the intensity of the j-th Bragg (elastically scattered) beam, and / is the inelastic intensity which would be measured if only the (000) beam were excited. P, can be calculated with standard methods of dynamical diffraction theory, and they can be measured experimentally, ft, is the angle for inelastic scattering from the j-th Bragg beam. (See fig. 22).
It is always possible to choose n angles H'k such that the system of linear equations
(22)
can be uniquely solved with respect to I. For instance, in the two-beam case (only two Bragg beams excited) we have
[, E) = Po J(na, E) + PG I(n2, E) (23) = P0I(-n3,E) + Peli-n^E)
Because (at least in centrosymmetric crystals) J(fi) = /(—ft) we can solve for /(fti.2, E)
in the unsymmetric case (Po ^ Pa)- By choosing the geometry such that the scattering
angles i?a = jn\ |, t?2 = |H31, are equal, one obtains immediately the anlsotropy of the Compton profile /.
& HE) = MpZ™*- (24)
For many applications, this is the relevant .outcome which can be directly compared with theoretical predictions (See fig. 9b). 45
Plaamons
The intensities of successive plasmon events can be found by integrating plasmon spectra in the bright field mode. Alternatively, a calculation based upon independent scattering events results in some Poissonian distribution for the plasmon spectrum. The channel coupling between plasmon and Compton events can be simulated by a weighted superposition of Compton profiles shifted by multiples of the plasmon energy. This was done in the following demonstration experiment on aluminum. A direct method for removal af the plural plasmon contribution is in principle possible since the convolution integral for the measured intensity can be solved for the single Compton scattering intensity by any deconvolution method (e.g. [4,14]). To our knowledge, this has not yet been tried.
Quasielastic background
Quasielastic scattering followed by a Compton event produces a smooth and slowly varying background in the angular and energy range of the Compton profile. When a fitting procedure is used to extract the Compton profile from measurements (as was done in the following demonstration experiment) the quasielastic contribution need not be considered separately since the smooth background shows up in the power law dependence (A • E~')., Removal of the quasielastic background should be possible using a method originally proposed by Batson and Silcox [15]. The coupling of the quasi-elastic background to the Compton event is given as
, E) = I tfd'Qiti - &)![&, E) (25) where M is the result of the coupling, Q is the quasielastic intensity and / is the purely inelastic intensity. For slowly varying Q, we have approximately
, E) *Q{n)f tPn'un'^E). (26) The integral is available experimentally as an energy loss spectrum in image mode, and Q can also be measured. So, the background due to quasielastic coupling could be removed. It should be noted that this procedure is only valid for radially symmetric intensities.
Applications
Polycrystalline aluminum
In ~ demonstration experiment [8] CPs of polycrystalline aluminum were taken at a scattering angle of 5.2 degrees. Measurements were done on a cylindricai mirror analyzer (CMA) attached to a Siemens Elmiskop IA at 40keV.The angular resolution was ±2Amrad, corresponding to a wavenumber resolution of ±0.25 A l. The finite wavenumber resolution broadens the profiles by ±15eV which amounts to & momentum uncertainty of 0.23a.u. This is considerably higher than accuracies obtained from photon CS experiments (~ OAa.u. typically). The film thickness was 240 A. Measured CPs were fitted to calculations including L-shell- and conduction elec- trons as well as a background ~ AE~'. From figs. 4, 5 it can be seen that the quality of the fitting depends on the fitting range. If the region of valence electron contribution is chosen, a deviation of Z% — 4% is found. In cases of a wider energy loss range for the fitting procedure the deviations increase up to almost 100%. These results clearly show that background subtraction (i.e. choice of an appropriate fitting region) is a serious problem. From our calcuiational results it is clear that, although the IA is valid for the conduction electrons, it is not for L-shell electrons. The maximum of the L-shell CP is ~ 50eV higher than predicted by the IA. This effect causes the maximum of the measured profile - which is in essence a superposition of both contributions - to He between the IA and the exact maximum. Consequently, it does not make much sense to interpret the total CP as the electron momentum density. The asymmetry of the profile brings a commonly used practice in CP data pro- cessing into question, viz. to split the measured profile at its maximum and substitute the low energy part by the symmetrically extended high energy part. The reasoning is that multiple scattering contributes mainly in the low energy region of the profile. Given the experimental conditions encountered in ECOSS this procedure is not gene- rally valid, both because the asymmetry caused by the core contributions is not taken into account, and because multiple scattering is not negligible even in the high energy part of the spectra. In practical applications the situation is rather better, since multi- ple scattering and core contributions—both flat and broad distributions— don't much influence the widely used reciprocal form factor B{r). It is evident from fig. 6 that the L-shell contribution alters B{r) only at small wave number.
Anisotropy in graphite
In an electron Compton scattering study of graphite, Vasudevan et al. |9] found strong anisotropy of the momentum distribution within the basal plane, (fig. 7). Photon CS cannot give these results because the beam cannot be focused onto a single platelet of graphite the orientation of which is random in the basal plane. For comparison, fig. 8 shows predictions of various theories. The disagreement is striking. The authors speculate that antibonding sp2 ir' orbitals—which should be empty theoretically—may cause the large negative value of B at the lattice vector L100- One fact is obvious: Even todays most refined model calculations are too poor for the prediction of CS data. This may well be since these calculational methods have in general not been used to predict wave functions but energy levels and band structure rather. In any case, one should be cautious about the results of these first ECOSS exoeriments. Electron correlation in copper
In order to demonstrate what can be expected for future ECOSS experiments we quote some results from photon Compton scattering: Fig. 9.a shows a Compton profile of the valence electrons in Cu, along the < 110 > direction [7]. The theory which is an SCF local density approximation (LDA) with exchange/correlation correction using linearly combined Gaussian orbitals predicts the experiment quite well. However, when the anisotropy of the profile is plotted, the situation is different: fig. 9.b shows that theory overestimates the amplitude of oscillations. In later papers fl, 2] the local density approximation was re-investigated thoroughly. The periodic deviations of LDA predictions from measurement, given in fig. 10 were traced back to electron correlation effects in the inhomogeneous electron gas. They act so as to reduce the occupation of the Fermi sphere relative to any model calculation based on a single particle concept. In an extended zone scheme, this effect gives rise to the oscillations in fig. 10. Again it can be said that present theories of the electronic structure in the solid are too poor to predict Compton profiles accurately.
Conclusion
After a presentation of the theory of electron Compton scattering, the advantages and disadvantages with respect to established photon Compton scattering experiments are discussed. Examples show that a) the main obstacle of strong multiple scattering can be overcome by a careful analysis of the various contributions to the Compton pro- file; b) anisotropies in the momentum distribution of valence electrons can be measured in microscopic samples, thus opening the way to electron Compton experiments in poly- crystalline specimens or microscopic segregates; c) Compton scattering is an extremely precise method for the investigation of otherwise undetectable solid state effects of the ground state, such as electron correlation. 49
References
[1] Bauer G£, Schneider JR. (1984). Nonlocal Exchange-Correlation Effects in the Total Compton Profile of Copper Metal. Phys. Rev. Lett. Vol.52,23, 2061-2064. [2] Bauer GW, Schneider JR. (1965). Electron correlation effect in the momentum density of copper metal. Phys. Rev. B Vol.31,2, 681-692. [3] Bloch BJ, Mendelsohn LB. (1974). Atomic L-shell Compton profiles and incohernt scattering factors: Theory. Phys.Rev.A,Vol.9,No.l, 129-154. [4] Egerton R F. (1986). EELS in the Electron Microscope. Plenum Press, New York, 229-241 and 357-361. [5] Eisenberger P, Platzman PM. (1970). Compton Scattering of X Ray from Bound Electrons. Phys. Rev. A Vol2.2, 415-423. [6] Eansen NK, Pattison P, Schneider JR. (1987). Analysis of the 3-Dimensional Electron Distribution in Silicon Using Directional Compton Profile Measurements. Z.Phys.-Conden.Mat.66, 305-315. [7] Pattison P, Hansen NK, Schneider JR. (1982). Anisotropy in the Compton Profile of Copper. Z.Phys. B Cond.Matter 46, 285-294. [8] Schattschneider P, Hohenegger H. (1987). Electron Compton Scattering from Po- lycrystalline Aluminum. Analytical Electron Microscopy. San Francisco Press, 270-274. [9] Vasudevan S, Rayment T, Williams BG. (1984). The electronic structure of gra- phite from Compton profile measurements. Proc. R. Soc. Lond. A3S1,109-124. [10] Williams BG, Parkinson **1P, Eckhardt CJ, Thomas JM. (1981). A new approach to the measurement of the momentum densities in solids using an electron microscope. Chem. Phys. Lett. Vol.78,3, 434-438. [11] Williams BG, Sparrow TG, Thomas JM. (1983). Probing the structure of an amor- phous solid: Proof from Compton scattering that amorphous carbon is predomi- nantly graphitic. J. Chem. Soc. Chem. Commun., 1434-1435. [12] Williams BG, Sparrow TG, Egerton RF. (1984). Electron Compton scattering from 50
solids. Proc.R. Soc. Lond. A393, 409-422. [13] Williams BG, Uppal MK, Brydson RD. (1987). Dynamical scattering effects in elec- tron scattering measurements of the Compton profiles of solids. Proc.R.Soc.Lond.A 409, 161-176. [14] Schattschneider P, Zapfi M, Slcalicky P. (1985). Hybrid deconvolution for small angle inelastic multiple scattering. Inverse Problems 1, 381-391. [15] Batson P, Silcox J. (1983). Experimental energy-ioss function for aluminum. Phys. Rev. B 27, 5224-5239. 51
Tigure Captions
Fig. l.a: Upper part: Geometry of Compton {scattering, p is the initial -wave -vector of the target -electron, A5*is the momentm: transferred in the interaction, pj is the final -wave -vector of the target electron. The quantity measured in Compton scattering is the distribution of-wave -vectors projected (Onto the q direction, p,. The dotted area schematicaliy indicates the spatiaJ distribution of valence electrons. .Lower parr: Sketch of the Term: sphere with allowed creation And .annihilation processes (ful3 Jines). The process indicated by z. dashed line is forbidden. Fig- 2 : Scattering geometry .and relation between momentum density and Compton •profile.
Tig. S: Single scattering-profile for conduction electrons (Al, 5.2 deg)
Fig. 4: .Measured profile as compared with ieasz squares nt over entire energy range. (24nm Al, 5.2 deg)
Fig. 5: ieast squares it excluding background as compared with measurements for 24nm Al-at SJideg. Fitting region USeV — 6S0eV (a), 290cF — 470eV (b).
Fig. 6: JReciprocal form factor JB(r) of measured Compton profile incJuding and exciuding (dotted) ir-sheU contribution.
Fig. 7: HeciprocaJ form factor in < 100 > direction in the basal piane of graphite, derived 52
from ECOSS. Lattice points are denoted Lijk. From {9j.
Fig. 8: Calculated reciprocal form factors in < 100 > direction in the basal plane of graphite. Full line: Tight binding, dashed line: pseudopotential. From J9j.
Fig. 9a: Compton profile for the valence electrons of Cu. Dots: experiment; full line: theory. From [7]. Fig. 9b: Experimental anisotropy in the Compton profile of Cu in directions < 110 > - < 100 >. Dotted and dashed lines: experiment; full line: theory (SCF-LDA) in the Hohenberg-Kohn-Sham formalism. From [7].
Fig. 10: Difference in the Compton profile betweeen SCF-LDA theory and experiment. In the lower part, a projection of the reciprocal lattice onto the < 100 >, < 010 > plane with the Fermi surface of Cu is depicted. The oscillatory difference can be explained by the removal of states within the Fermi body to outside. From [2]. Fig. 11: Simulation of Bragg-Compton channel coupling in Si. From [13]. Fig. 12: Scattering geometry for Bragg-Compton coupling in the two-beam case. 53
core
OL. cs
CD
cs
= r\i~ a ( \ u^ t
— CM o> \
/ .6 8 1 6
/ \ 3 8 0.8 8 . 1e.ee ise.ee 2se.ee 35e.ee Mse.ee sse.ee 6S8.ee 75 .ee es .ee $s .ee te B.ee Encrov Ios6 toV] Scoltcring profile from conduction electrons I PI 1 5.2de9l
\ 1 i CD i i
1 «O
\ rb.un i t o • V c N • to . 5 as h ' — •/->-
<•
..« 3 // \ ^ o . > /; 2« c / \
s A CD
i r-4
H
•—*
\ \ V\ \\
IS ! t v 0 2 8 e . s c \
CB
ca «
Enerev loss (eV] Profile of conduction electrons conoored with calcuioicd orof«le( (fl( 2tl0fl..5.2ae9l f i 11 ins iniervol 1 = 1190ev-478evi 56
-e.2 J2 16 28 2H 28 32 36 V« 18
Torn Todor froa ncosured orofiie including ond excluding (dotted) L-sheil coniribuiion (fll 218 A. S.2 dc9)
010 0.1
0.05 - \
B{r) 0 H^W- Blr) 0 .ll2 '
-0.05 k I 1 I ' i *. : -cnop : : 57
Cu valence profile < 110 >
n
Pz(a.u.)
-0.04- 58 AE/keV ff g (*olid line), multiple (dotted line) and total (broken line) scattering from a silicon sample 136 A thick oriented in the (100) direction with 100 leV incident electrons. In (a), (b). (d) and (e) the scattering vector is in the {020) direction with a scattering angk of 5.5* {(a) and id)) and 3.9* ((6) and <«)). In (c) and (/) the scattering vector is in the (022} direction and the scattering angle is 5,5*. The data are plotted against both energy loss (ai-ic) and the momentum of the electrons in the sample (
-Q
000 00G 61
DISPERSION EFFECTS AND SPATIAL RESOLUTION IN STEM
:s, Zabala and P.M. Echenique*
Dpto. Electricidad y Electronica , Facultad de Ciencias, Euskal Herriko UnibertsiuueaApdoM*\48080 Bilbao Spain. * Dpto. FlsicadeMateriales, Facultad de Quimica, Euskal Herriko Unibensiuuea, Apdo. 1072,20080 San Sebastian. Spain.
Abstract The excitation probability of a STEM electron moving paralell to a vacuum-metal interface is studied. Expresions for the energy loss probability are obtained character/zing the solid by a spatial dispersive dielectric function e(ct)Jc), within the well known specular reflection model and computations are performed for 100 keV electrons, using the Mermin dielectric function which includes both single particle and damped collective excitations in the medium. The study is focussed on the variation of surface and bulk piasmon excitation probabilities with impact parameter, specially around the interface, at distances shorter than 10 A.
INTRODUCTION
In the last few years, developments in the field of Scanning Transmission Electron Microscopy (STEM) have renewed interest in the interaction of high energy electron beams with surfaces at glancing incidence <-h9\ The surface piasmon excitation probability is determined by analyzing the inelastic scattering of fast electrons aligned paralell to the surface, at a given impact parameter, which can be systematically varied. Probe sizes of 0.5 - lnm and 100 keV electrons are usually employed in this kind of experiments. At such energies, the classical dielectric theory in which the target response is described by a local dielectric function e(co), has shown qualitative agreement with many experimental data. Anyway, when STEM experiments are performed to study the nature of the surface, a high spatial resolution is required.I n these cases, an accurate knowledge of the dependence of die energy loss probability on the position of the probe is needed in order to interprcte the experimental data. Howie & Milne <5> and Krivanck et ai. C7) pointed out the insensittviry of high energy inelastic scattering to surface reconstructions. In contrast to this, Ichinokawa et al. W observed, with 100 eV electrons, changes in the spectrum when oxidation or reconstruction of Si surfaces occurred. Sheinfein et al. W pointed out the detection of changes in the loss 62
spectrum with 100 keV electrons at a spatial resolution better than 8 A as the probe was moved across an AI-AIF3 interface. This last result was justified by Cheng <10), who calculated the spatial resolution fcr 15 eV energy loss in Al as the plasmon localization distance. The spatial resolution can be analyzed by studing the variations of the EELS spectrum with impact parameter, using dielectric theory, which also includes plasmon lifetime. In the local approach, when the medium is characterized by a non-dispersive dielectric function £{co), rapid variations of the spectrum near the the surface occur due to the logarythmic divergence of the loss probability as the impact parameter goes to zero, which has no physical sense. A more realistic srudy of the problem can be done by considering spatial dispersive response function £(oojc) to characterize the target ^1>12).
DISPERSIVE ENERGY LOSS PROBABILITY
V
Figure 1 When the wave vector dependence of the dielectric response function of the medium is considered, the induced potential can be calculated making use of the specular reflection model 03,14) The p]ane boundary problem to bs solved, presented in fig. 1, can be replaced by that of a surface charge density a and two punctual charges simetrically placed at ZQ and - ZQ , in vacuum, to calculate the potential at z>0 (fig. 2a). For z<0 the potential is due, as it is shown in fig. 2b, to a negative charge density --a
<») (b) Figure 2 The energy loss by unit length is obtained from the induced potential as:
dW (1) dy •x-0 ; y«vt ;
and is expressed as a double integral in energy and momentum kx, paralell to the surface and perpendicular to the particle's trajectory:
dW = f d(o to f dk P(CD,k ) - J d» o P((o) (2) dy s ]| 0 0 0 , where in atomic units, the differential probability to loss energy co and transfer momentum
kx is given by: [e(Q,a,)- P(C0,k) (3)
, where Q is the momentum paralell to the interface
(4) v and £(Q,co), the surface response function defined as an integral in perpendicular momentum
(5)
, with (6) When dispersion effects arc neglected in e(k,co), after integration in (3) the surface response function gives the local function E(0)>, so equation (2) becomes 2 e22"0 J e(a»-l P(co,k ) = —- —pr- InJ — -I (7) ~"J^ Q l e(o)) + l TIV2 Q I e(0)) + !J
In this case the integration in kx can be calculated analytically, giving the well known expression <15)
2 2cozo I £(co) -11 P((0) = __ Ko(—21) InJ t t I <8)
, where KQ(X) is the zero order modified Bessei function, which presents the asymptotic behaviour:
Kn(x) ~ -ln(x) , x«1 (9)
To evaluate the dispersive loss probability we have considered the Mermin dielectric function (l6), which is an extension of the Lindhard function to include plasmon damping co [ e,(z,0)) -1] e(z,o))=l+ 1 (11)
, with z = k / 2kp, to = x + iy, x = co/cop and e^ the Lindhard dielectric function or random-phase approximation for the longitudinal dielectric function of a zero temperature electron gas. To compute the locai j robability we have considered the Drude dielectric function which also includes damping:
e(co) = l 2 (12) d) (a) + i y)
In fig. 3 we represent the differential probability P(co,kx) given by equation (3) for the surface plasmon energy ws and different values of the momentum kx, to show how the high momentum (associated to high deflection angles) give better spatial resolution of the surface but, on the other hand, their intensity is decreased and the contribution to the total probability losses importance.
In fig. 4, an energy loss spectrum is presented for Al (rs=2, y=l eV), produced by a 100 keV exterior electron at a distance of 1 A from the surface. The continuous line represents the dispersive spectrum and the dashed one .he local probability. The effect of 65
dispersion is to decrease the intensity of the maximum and to displace it slightly towards higher energies. In fig. 5 the probability to excite surface plasmons as a function of the reduced impact parameter B«2to ZQ/ v is represented until 100 A. The continuous plot corresponds to the dispersive probability and the dashed line to the local one. In the upper side we show both probabilities in the first 5 A. It is shown that the rapid variations near the interface can also be explained using dielectric theory. Near the interface those rapid variations of the spectrum occur at small distance variations but ,may be, not as small as the 4A quoted by Cheng, as it can be seen in the upper plot. Furthermore, the dispersion corrects the divergence at the boundary and the dispersive probability becomes soon the local one , before - 20 A. The rapid variations for B<1 are due to the logarythmic behaviour, but for
B>1 the behaviour is exponential and the spectrum changes in units of B. For uy=ACKNOWLEDGEMENTS
The authors wish to thak Prof. A. Howie and Prof. R.H. Ritchie for many stimulating discussions and Eusko Jaurlaritza for financial support.
REFERENCES
1.- BatsonRE., Ultramicrosc., 2 (1982) 277; Ibid, 11 (1983) 299. 2.- Cowley J. M., Surf. Sci., 144 (1982) 587. 3.- Fan Cheng-gao, Howie A., Walsh C.A. and Yuan Yun, (to be published). 4.- Howie A., Ultramicrosc., H (1983) 299. 5.- Howie A. and Milne R. H., J. Microsc., 12fi (1984) 279. 6.- Ichinokawa T., Ishikawa T., Awaya N. and Onoguchi A., Scanning Electron Microsc., Part 1, (Chicago: SEM Inc.) (1981) 271. 66
7.- Krivanek O. L., Tanishiro Y., Takayanagi K. and Yagi K., Ultramicrosc., 11 (1983)215. 8.- Marks L. D., Solid St. Commun., 42(1981) 727. 9.- Schcinfcin M., Muray A. and Isaacson M., Ultramicrosc., 1£ (1985) 233. 10.- Cheng S. C. Ultramicrosc., 21 (1987) 291. 11.- Echcnique P. M, Phil. Mag. B, 52 (1985) 9. 12.- Núñcz R., Echcniquc P. M. and Ritchie R. H., J. Phys. C, 12 (1980) 4229. 13.- Ritchie R. H. and Marusak A. L., Surf. Sci., 4 (1966) 234 . 14.- Wagner D., Z. Naturf. (c), H (1966) 634. 15.- Echcnique P. M. and Pendry J. B., J. Phys. C, £ (1975) 2936. 16.- Mermin N. D., Phys. Rev. B, 1 (1970) 2362. 17.- Lindhard, K. Dan. Vidcnsk. Selsk. Mat. Fys. Medd., 2S, No. 8 (1954). 67
Eo=l00KeV} rs = 2 - y= 1 eV
:. 3r
a. Sir
0 5
impact pcrcmeter 2a(A)
Figure 3
O. 045 r
0. 005 r
r,, ; 0.5 2. S C. 7 0.8
Figure 4 68
0.02 0.04 3.06 0.0S 0.1-
Figures
0. lp
0.08
0.2
Rfure6 69
STOPPING POWER FOR PROTONS IN Al FOR THE WHOLE RANGE OF VELOCITIES
by
F. Flores Departamento de Fisica de la Materia Condensada Universidad Autonoma de Madrid Cantoblanco, Madrid 28049, SPAIN
A. Aruau, P. M. Echenique Departamento de Fisica de Materiales, Universidad del Pais Vasco Facultad de Quimica Apartado 1072, San Sebastian 20U80, SPAIN
R. H. Ritchie Health and Safety Research Division Oak Ridge National Laboratory Oak Ridge, Tennessee 37831 USA 70
ABSTRACT
The electronic stopping power of aluminium metal for protons has been calculated with explicit inclusion of the different proton charge states inside the medium. The total stopping power is obtained by adding to the calculated stopping power of each charge state, the energy loss associated with the electronic exchange processes. Good agreement is obtained between our first principles calculation and the available experimental data. 71
1. INTRODUCTION
The problem of the energy loss of ions moving through solids has attracted the attention of physicists since the beginning of the century1. Fig. 1 shows schematically the different regimes for the ion stopping power as a function of its velocity. In the high velocity regime, the ions is stripped of its electronic charge and the theories of Bethe^ and Bloch3 of the electric stopping for a charge Z give a good description of the energy loss process. In the low veloci- ty regime, nuclear processes between the projectile and the target give an important contribution to the total stopping; the electronic contribution shows, however, a linear behaviour in this regime which is the result of the ion being surrounded by a cloud of polarization charge. The density functional theory allows one to calculate this cloud of polarization and the energy loss at those low velocities in a selfconsistent way. THe first explicit calculations of the stopping power in this formalism were perfomed by Echenique et al4 for hydro- gen and helium, and later were extended to higher ionic char- ges5.
At intermediate velocities (v ^ z2/3 i=. ) there is no rigorous theory of the stopping power for ions mov?.ng through solids, even in the case of light ions. Effective charge theories have been proposed** to explain the stopping data. The processes by which a partly stripped ion interacts with a target are numerous and complex7; in particular, in the intermediate velocity regime electronic exchange processes between the projectile and the target must play an important role in stopping. For this reason, the development of first principles methods to calculate electronic capture and loss cross sections^ should also be introduced in the calculation of the total energy loss. 72
In this work we have calculated the stopping power for protons in aluminium as a function of the intruder velocity, with explicit inclusion of such solid state effects as the electronic capture and loss cross sections of the moving ion.
2. THE MODEL
The different mechanisms included in the calculation of electronic exchange processes are shown in fig. 2. In the Auger process illustrated in fig. 2a, an electron is captured (or lost) by the ion to (or from)a bound state assis- ted by a third body that may be a plasmon or an electron- hole pair. Condensed matter effects are important here since electrons in valence band states are involved.
A second mechanism leading to capture and loss is the coherent resonant interaction. Electronic exchange processes are induced by the time dependent crystal potential a* seen by the fixed ion (fig. 2b). The coherent resonant interaction resembles a close atom-atom collision at high velocities. Al low velocities condensed matter effects come into play.
The third mechanism included in the electronic exchange processes is due to the shell interaction (fig. 2c). In this case, the inner electron of a target atom can be capture by the moving ion.
A main assumption in the calculation of all these exchange processes is that the proton has a well defined bound lovel over the whole range of proton velocities. Vinter and Guinea and Flores9 have shown, using a selfenergy approach, the reality of the H~ state in an electron gas. proving in this way that a proton can bind two electrons inside aluminium. 73
Moreover, Norskov^O, using the local density formalism, has concluded that the H~ configuration in aluminium has an energy that is 9eVs. lower that the one for H°, with a single electron bound to the proton. This result suggests that the H" and H° configurations would be mixed for an ion velocity, v, such that ,2 a atomic units,
namely, around V= 0.8 a.u.
The equilibrium charge fractions of bare protons ( neutral atoms ( 4>° ) and negative ions ( $~) as determined by the different mechanisms discussed above are given in terms of the capture and loss cross sections** as:
+ 1 • =oioss "loss(H) D" (la)
where:
D= "loss oioss + aioss ocapt + °capt
In these equs. oioss (H~) and aioss (H° ) are the loss cross sections of an electron by the negative ion and the + neutral atom respectively, while oCapt (H ) and <*capt "ft are the corresponding capture cross sections of an electron by the bare proton and the neutral atom.
Fig. 3 shows the results of our calculations for •+.#° and $~ as a function of the ion velocity. For very low veloci- ties, v « 1 a.u., only negative ions survive in the Metal; for high velocities, v »1 a.u., only positive ions survive. 74
while for intermediate velocities v = 1 a.u., *° reaches its maximum. Notice that this intermediate velocity is in good agreement with our rough estimation as discussed above using Norkovs calculations.
Finally, the stopping power is calculated by sunning the stopping power for each charge state weighted by the respective charge state fractions, and adding the energy lost per unit length in capture and loss processes. Fig. 4 shows how two succesive loss and capture processes leave the ion in the same state, creating an electron-hole pair excitation in the electron gas.
Thus, we write for the stopping power: £
In equ.O) ^£ an while it understimates at the sane time -he stopping power of H at intermediate velocities, 0.9 sv £1.3. He can expect* however, --hat both effects -e:::: to ccrr.r er.sate each other as the results of the total s-t-cpr-r-7 power '=« below) s-jegest.
3. RESULTS FOR THE STOPPIKG POVrSP.
The results obtained form ecu.(3) for the stopping power of protons moving with velocity v in hi ire shown in fig. 5 as a thick solid line (TOTAL). The curve labeled LT is the stopping power for a bare proton calculated in linear theory using an RFA dielectric function11, tc represent valence
electron excitations in aluminium with r£*2. Inner shell corrections from the 2 s and 2p electrons of the Al ions have also been included. The different contributions to til* curve labeled TOTAL have been separated to show the relevance of the various terms as a function of the ion speed. Mben V* 1.2 a.u. the contributions from the negative ions In conclusion, our first principles calculation shows the relative contribution to the stopping power of the diffe- rent charge states of a proton moving with velocity v in aluminium. Combining dielectric and density functional results in the appropriate velocity range with the energy lost in the capture and loss processes, we have obtained good agreement with experiments. 76
ACKNOWLEDGMENTS One of us (AA) is grateful for careful and stimulating conversation with J Nagy. The authors gratefully acknowledge help and support by Eusko Jauralaritza, Gipuzkoako Foru Alcheudk, the Spanish Comision Asesora Cientifica y Tecnica (CAICYT) and the NATO Research Grant 0142187. The authors would also like to thank Iberduero S.A. for this help and support. Support is also acknowledged from the Office of Health and Environmental Research, U.S. Department of Energy, under contract DE-AC05-84OR21400 with Martin Marietta Energy Systems, Inc. 77
IBrBKBHU
1.- N.Bohr, Philos. Mag. 26, 1, <1913) 2.- H.A.Bethe, Ann. d. Phys. 5, 325 (1930) 3.- F.Bloch, Ann. d. Phys. 16, 285 (1933). 4.- P.M.Echenique, R.M. Nieninen and R.H.Ritchie, Solid State Comm. 37, 779 (1981). 5.- M.J.Puske and R.H.Nieminen and R.H.Ritchie, Phys. Rev. B 27, 6121 (1983); A.Arnau, P.M.Echenique and R.H.Ritchie, Nucl. Instrum. Methods B 33, 138 (1988). 6.- S..Kreussler, C.Varelas and W.Brandt, Phys. Rev. B. 23, 82 (1981); W.Brandt and M.Kitagawa, Phys. Rev. B 25, 5631 (1982). 7.- M.A.Kumakhov and F.F.Kumarov "Energy Loss and Ranges in Solids" (Gordon and Breach Science Publishers, New York, 1981); Y.H. Ohtsuki "charged Beam Interactions with Solids" (Taylor, London, 1983) 8.- S.K.Allison, Rev. Mod. Phys. 30, 1137 (1958) 9.- B.Vinter, Phys. Rev. B. 17, 2729 (1978); F.Guinea and F.Flores, J.Phys. C 13, 4137 (1980). lO.-J.X.Norskov, Phys. Rev. B 20, 446 (1978). ll.-J.Lindhard, K.Dan.Vidensk, Selsk. Mat. Fys. Medd. 2«, no 8 (1954) ; R.H.Ritchie, Phys. Rev. 114, 644 (1959). 12.-H.H.Andersen and J.F.Ziegler, in "Stopping Powers and P.anges in all Elements: Hydrogen" (Pargamon, J.'ev.- York, 1977). 73
FIGURE CAPTIONS
Fig. 1 Shows schematically the stopping pcv.*er of an icn of charge Z as a function of its velocity. The intermediate regime is defined by the statistical velocity Z2/3 —
Fig. 2 i.) Shows a:; Auger process for an ion moving ir. a uniforr. electron gas. (b) £hc-..£ the coherent resonant process for an ion moving in a crys-f.l. (c) Shcv.s a shell process whereby an ion moving in a crystal captures cne electron from an inner level of the target.
Pig. 3 Different, charge fractions for H as a function its velocity.
Fig. 4 Shows schematically how loss and capture processes create an electron-hole pair excitation in the electron gas.
Fig. 5 Stopping power in atomic units of Al for protons as a function of the ion speed. The circles are the experimen- tal data. The different contributions to the curva labeled TOTAL from the fractions of bare protons (H+), neutral atoas (H°), negative ions (H~) and capture and loss processes Fig. 1 80
ORNL-DWG 87-11006 CAPTURE LOSS
V
EF
Fig. 2 C '
Equilibrium Charge Fractions o o o * * • •f^ O> 00
< 15*
T8 ORNL-DWG 9M584
+ ( v/ 1 1 11 H° H" loss capture electron-hole pair excitation OflNLOMBMMBH 0.3 . 1 . 1 •
TOTAL 7 /^ /o 0.2 S/ A"r /
LUl X TJ|"O Jrv\ 0.1 -/A
in 0.0 0.4 0.8 1.2 1.6 2.0 v(au)
Fig. 5 85
stopping pomp for protons revisited
Gregory Lapicki Department of Physics, East Carolina University Greenville, North Carolina 27858, USA
1. Introduction For a projectile of atomic number Zj (I.e., in effect of charge Z^«) and velocity v^, the stopping power in a target characterized by the electron density n ( n = number of electrons in an atom of atomic
number Z2 per unit volume), the stopping power can be written in atomic units (aee Fig.l) as 2 S = (4» Z\ n e aQ/ vj )L , (1) where the dimensionless, reduced stopping power L is known to be proportional to v^ at low velocities and it converges to Bethc's (1930) L» to(2»{/«) - B2 - In (1 -62) , (2)
with 8= Vj/c and n> = o>(Z2), in the high velocity limit. The L func- tion for protons (Z,=l) will be revisited in the low and high velocity regimes; the proposed formulas will be tested against empirical fits and fused into one function that reproduces the data at moderate velocities. The velocity ranges of 0.1-1.0, 1.0-10, and 10-100 are designated as low, moderate, and high velocity regimes. The low velocities mean proton beams of energy 25 keV or less; they are comparable or lets-than the Fermi velocity of the target electrons. The high velocities of v1>10 represent the protons with energy above 2.5 HeV. The stopping power peaks around v.= 2 i.e., for about 100-keV protons. 86
Figure 2 shows the L for stopping of protons in hydrogen, aluminum, silver, and bismuth (from the highest to the lowest curve) according to the Andersen and Ziegler (1977) fits to experiments. Since these fits were constructed to mimic the SLindhard and Scharff (1961) found in a Thomas-Fermi treatment (see Fig. 3; note the a of Lindhard-Scharff was multiplied by the square root of 2 to ensure that the TF unit of length has the saae magnitude in a homonudear quasimoleclue as in the TF atom) L . 2 vj / (Zf3 • 4'3 )3'2 . (3) if a strange factor i^ is deleted. Subsequent publications of Lindhard et al. -- known as Notes on Atonic Collisions: III (1963), 11(1963), and 1(1968) -- refer to Eq.(3) as a "result quoted and uti- lized in a summarizing article" i.e., in their 1961 Fhys. Hev. paper. The derivation of Eq.(3) has never been explicitly presented. It would appear that to obtain SCv, , Lindhard and Scharff had to choose an 1/r interaction in their power law description of the scattering poten- tial. Such a selection, however, results in S being proportional to a instead of a that is required to derive Eq.(3). Bonderup's (1978) remarks are worth quoting here: " we are faced with the problem of possessing only fragmentary information on thm function -dE/dx. A more detailed function has been undertaken by Lindhard and Scharff many years ago ... The details of the derivation. 87
however, remained a secret and we can provide no more than the following reference: J. Lindhard and M. Scharff: Notes on Atomic Collisions, IV, Hat. Fys. Hedd. Dan. Vid. Selsk., to be published." The details of the Lindhard and Scharff derivation are still se- cret. As their Notes-Part IV have remained to be published for the last three decades, the enigmatic L of .Lindhard and Scharff -- baring an unaccounted'(albeit irrelevant for protons of Z^=l) factor of Z^11'6 -- is "rediscovered" here (see Fig.3) with a scaling of Eq.(l) as follows. Two assumptions are made: 1) with L being proportional to v., the constant of proportionality simply set to 1 i.e., L = v., 2) the n of Eq.(l) is scaled to a , the unit volume in a Thomas-Fermi treatment, rather than a , the cube of the Bohr radius as the unit volume assigned conventionally to all atoms. This gives L =(v^a/a ) which, with the specific definition of a (see Fig.3), is identical with Eq.(3). The points added to Fig.7 from Echenique et al. (1986) represent (see Fig.4) the ratios of the experimental L from Andersen and Ziegler to the L of Eq.(3). The Lindhard-Scharff low-velocity L clearly uttfe-res- timates the data, increasingly so with the increasing Z-. By contrast, the sophisticated density-functional approach of Echenique et al. (1981) results in satisfactory agreement for most of the target elements. In this work, the Lindhard-Scharff formula is modified (see the top of Fig.5). Using the method by which Eq.(3) was "re-discovered" but with a different expression for the TF unit of length in a quasimolecule of
/3+ /3 the stopping medium: a =2*0.8852-*O/(Z^ z\ ) instead of a
/3 /3 1/2 0.8853-ao/(zJ + z| ) , it is found that
Liow = 555 vi 88
As seen in Fig.*, Eq.(4) is as good as the predictions of Echenique at
al. (If«1). Z£ oscillations ar« th« defect of a statistical description of an a ton; the nagnitude of this oscillations in the experiaent-to-the- ory ratios, however, could be partly in error because the stopping power data happen to be scarce for the targets where oscillatory peaks are seen. As illustrated in Fig.7, Eq.(4) agrees with the Andersen-Ziegler fits up to v, * 1, and it overshoots the neasured stopping powers as they decline fro* the linear-in-velocity dependence and reach ns«iess around v, * 2. 5. Hlsjh velocities
The Bethe-Bloch formula with "few = Kg* Z2 converges to the data remarkably well for large Z« (see Fig.7) provided the Bloch constant is tailored to then by choosing K^ - 10 eV; it fails, however, as Z2 —> 1 since it employs "few * 10 eV for Z.*l whereas the exact calculations of Dalgarno (19(2) for hydrogen give 15 eV. A significant improvement is obtained at high velocities when « * 0.275** Z2*(l * l/Zt ) or, in •», 3 -L * 15 eV-Z2(l • l/z}' )/2 (5) so that Hm * 15 eV for Z2*l and the effective Bloch constant is about 9 eV in heavy target elenents. With Eq.(5), th« Bethe's L reads (see the middle of Fig.5) Hiigh * ln I726 vl' Z2(1 * 1/Z2/3 M " *2 " ln (l It should be noted that the last two terns are insignificant in their contribution to L (see Fig.2). Equation (6) gives the stopping power in hydrogen within 2.51 of the data (see solid curves in Fig.t), while eh* Bethe-Bloch formula (see Fig.7) overestimates stopping power measure- nents in hydrogen by an unacceptable it. With the increasing Zg, the L 89
of Eq.(6) and the Bethe-Bloch become mutualy indistinguishable and fall within 11 the experiment i.e., within uncertainties of high-velocity data. At this level of experimental errors, a few percent overestimatiom of the stopping power measured in the lightest targets is significant. This could not be a symptom of an ill-description of the hydrogen atom in statistical treatment since Cq..(5) adjusts *&> to the exact value of 15 eV in hydrogen. However, the Andersen-Ziegler fits for stopping power in hydrogen were aade up to the highest available energy of 3S-MeV protons. Only truly high-velocity experiments, in which the stopping power of hundreds HeV if not GeV-protons would be measured, could yield clues as to the optimal selection of w = w (2g) for the "best" applica- tion of the Bethe formula in the high-velocity limit. 4. Moderate velocities At moderate velocities, the Lindhard and Scharff prescription of 1953, 1/2 3 2 L = i.36 x - 0.01« x ' , whore x s. vJ/Z2 , (7) seems to fill the gap between the low- and high-velocity formulas pre- sented in this work (see dashed curves in Fig.6). It links with Boa.(4) and (6), however, somewhat abruptly. Hence, a formula is proposed that is forced to transform smoothly into Eqs.(4) and (6) and that approxi- mates empirical stopping power, with deviations which are comparable with experimental uncertainties in the 1 < v,< 10 measurements. The low- and high-velocity formulas are bridged using the natural logarithms with a relatively simple argument (see the bottom of Fig.S and the dashed curves in Fig.9) : 3 2 3fl VS L = In |1 + Llo||/[1 -0.3 v j[ /(l • 2£ ) * 0.447 v|/( 1*1/^*1 | - 8* - In (1 - 82) , (•) 90
where Llo|# is given by Eq.(4) with Z^* 1.
5. conclusions and plans Tha low-velocity stopping power foraula of Lindhsrd and Scharff (1961) waa radarivad; whan sealed with a different definition of the TF unit of length for the projectlie"target quasiaolecule, it agreed with the senienpirical fits of Andersen and Ziagler (197?) up to Vj« 2/3 — the agreement was no worse than seen in prerious comparisons of these fits with the calculations of Echenique et al. (19S1.1986). Bathe's high-velocity stopping power was employed with a seaitheoretical func- tion for the average ionization potential -- this function was con- structed to equal 15 eV for hydrogen. The Andersen-Ziegler fits were found to converge to this high-velocity foraula; measurements with GeV-protons stopped in the lightest targets are needed to verify the universal utility of Bathe' stopping power in the high-velocity limit. Further work in progress will search for an expedient but store justified (if not siapler) L for all proton velocities. Extensions of Eqs.(4) and (6) to stopping power for other projectiles (Z^^l) will be investigated.
Andersen, H. H. and Ziegler, J. F., Hydrogen Stopping Powers and Samges in All Elements (Pergaaon Press, New York, 1977). Bethe, H. A., Ann. Phys. 5. , 325 (1930). 91
Bonderup, E., Lecture Motes on PMMtratien of Chargad Particle* through Matter (Institute of Physics, Aarhus University, 197S), p.142.
Dalgarno, A., in Atonic and Molecular Proceses, edlteo oy D. t. Bates (Aeadenic Press, New Tork, Iff2), Ch.15, p. Me*.
Echenique, P. M., Vieninen, 1. M., and Ritchie, ft. «., Solid State CoasKin. 12, 779 (1M1).
Echenique, P. M., Kieninen, R. H., Ashley, J. C, end Ritchie, ft. ft., Phys. Xev. A 31, 897 (1986).
Lindhard, J. and Scharff, M., Hat. Fys. Medd. Dan. Vid. Selsk. 21, #13 (1953).
Lindhard, J. and Scharff, M., Phy«. Rev. J2±, 121 (ltel).
Lindhard, J., Nielsen, V., Scharff, H., Thonsen, P. V., Nat. Pja. Dan. Vid. Selsk. U, #10 (19f3).
Lindhard, J., Scharff, M., and Schiott, H. E., Hat. Pys. Medd. Ban. fid. Selsk. 11, #14 (19C3).
Lindhard, J., Kielsen, V., and Scharff, M., Mat. Pys. Medd. Ban. Tid. Selsk. 14, #10 (19«t). 92
V,* Z, ««J *,
d»>»c^/•»'•>»Itxj lor
0.1 — f o -: io : 2.5 rvv O - j LAPICKI FIG. 1 93
cqUr fun)
190
FIG. 2 94
Zf r> e»g. jE
' "
a«
FIG-3 95
LAPICM FIG. 4
FIG. 7. Rmtioi of stopping powers for from Anderson and Ziegkr (AZ) OUf. 26), idW/4*)** to the density-functioiul predictions of Rcf. S, (rf*Yrf*W The solid circles represent experimenul diu for 16 eknealal solids, while the crosses are interpolations to materials for which no ex- perimental data are available. 96
-A
(MS.) L* » IO«V- (\ 0. L h,-.,k s In 27?7 b n 51 5*5" x?
LAPICKI - FIG. 5 97
LAPICKI . 6 L
14 1 1 1 1 1 «•• 4 > V .• ". • • I \
• 1i a • • • .^^B^— —^ s tf~- •V a •
Out - I • II 04 1 1 1 1 | 1 1 10
FIG. 7. lUtki of stopping pd«cn lor from Andenoa and Zkfkr IA2) OUT. M. density-functiootl predictioot of Rcf. 5V while the cronci are perimenta) diu are available. 98
LAPICKI FIG. 7 99
LAPICKI
•*• « v " - * "^ -J 100
LAPICKI FIG. 9 101
Stopping Power Theory in Inhomogeneous Many Eieetron System M. KITAGAWA Department of Electronics, North Shore College Atugi 243, Japan
A dielectric function £ for an inhomogeneous aany-electroii gas was obtained analytically as the first order formula under the condition of the high frequency response (1]. Using the lowest order of-approxiaation in an iteration for the integral equation of £ , we also showed that the lowest order formula of stopping power exactly corresponds to one called as the "local density approximation" (LDA). In this paper, we derive the integral for- mula of stopping power for the system, in which the electron density changes one—deaensionally. Such a kind of formula is applied to the problem of the solid surface.
1. Introduction Theoretical research concerning dielectric function £ has taken the position of the fundamental theae in the field of elementary excitations in solid for a long time. Under the system of free electron gas, various kinds of theoretical works on £ were per- formed since the work by Lindhard [2], and applied to the calcu- lation of stopping power for ion passing through solid [3-6]. Comparisons between theoretical and experimental works were also performed, and LEA theory, which was introduced phenomenologica- lly in the theory of stopping power, gives a relatively good agreement with the experimental data [5]. As far as the general theory of £ , which should be extended to an inhomog/eneous many- electron gas, concerns, a basic progress has been done recently. Lundqvist reported the general framework of theory on £ for the 102
lnhomogeneous many-electron gas in 1963 f7J. By us* of the Green function's met bod, we obtained the first order of analytical fo- rmula for e under the condition of the nigh fraqiwncy r«spons«. and also showed the theoretical position of LD* theory in ISM 11]. Zn this paper, we confirm the basic progress of the theory on e In inbomogeneous aany-electroa gas by use of the density matrix formalism, and derive the theoretical formula of stopping power for the system in which the electron density is treated to change one-denensionally. Zn the following discussions, we use the notations m, -e and Z^m for the mass of the electron and the charges of the electron and the projectile, respectively.
2. Dielectric function in inhomogeneous many-electron gas The first order of analytical formula for the dielectric func- tion 6 lnr-ffl space, which is proportional to 1/tf* (« :angular frequency of electric field), was represented as follows [1J
1 e2
i 2 - 12 ^ rx *|*V^| (1) «xezn(F?) Vcft-rt —i'-|
In the above, r* and r* refer to the position vectors, and n(r^) Indicates the electron density at r*. The important property of the above formula is that not oaly the term Indicating the role of the contribution due to the localised mode e)_ (r^)t*(4Xe2m{r*) /ml1'2) but also the term denoting the effect of boundary condi- tion determined from the many-electron system are included 1m eq.(l). in deriving eq.(l), the Irreducible polarisation propa- gator K makes an important role through the following equation 103
Zn eq.(2), an abbreviated notation (1) indicates the space-time coordinates which includes spin, and v means the Coulomb intera- ction between two charged particles. The detaied explanations of eq.(2) are refered to ref.(l). By use of the representation of density matrix formalism, It is given by
and
J
where n(rf,r,) and f.(r? ,r2) mean the density matrix and the
density matrix element. 0 , % and Er are the Heaviside step function, the positive infinitesimal and the last filled energy of the electron in the ground state of many-electron gas. f>. and E. are the wave function and the eigenenergy of the j-th elect- ron, which satisfiles the following Schrodinger equation
From eqs.(4) and (5), and using the notation that the operator of Bamiltonian H operates r^, we have
Expanding TT by I/a) in the complex energy plane under the high frequency approximation, we have 4 Z r*.r7) f j/r*.^) t
l 2 [*~ (EirE.)l
Using eqs.(4) and {«}, the first term in eq.(T) reduces to 104
and th* second ter» becoaes
I7Hr»n(r.?t ;
a«> 2 2 -(n(r».f*)+n(f|.r*))CV S"C^-^>1I
Above are exactly same with the results obtained in ref.(l). in which the sum-rule of the wave function is used in calculating 7J- . If we use the such a kind of sua-rule. there appears a eff- ort ing calculation, which aeans the usefulness of the density aatrix foraalisa in solving the problem of the inhoaogeneous aany-electron gas. The result of eq.(10) in ref.(l) is recoa- firaed by eq.(9) by use of the density aatrix foraalisa, and we also obtain eq.(l) for the first order £• in inhoaogeneous aany- electron gas froa eqs,(2) and (?). It is wirthy to be noted that the tera corresponding to the second term in eq.(l) doesn't app- ear in the theory of t under the usual free electron gas. It is »ell known that the dispersion relation of plaeaon and the contribution of the single-electron excitation are derived froa the fourth tera in eq.(7), if we apply eg,. (7) to the free electron gas under the proper approximation shown la refs.(3} and (4). The aethod used in the calculation of 6 in free elec- tron gas gives a theoretical guide in deriving the higher 105
formula of £. in inhoaogeneous many-electron system. However, there appears the mathematical difficulty in obtaining such a kind of higher order term exactly. The derivation of the higher order formula is now under construction. The orthogonality between £, and £ (inverse dielectric function) is given by eq.(5) in ref.(l), and it becomes as follows if we use the "r*-tU representation
Generally speaking, £"' is derived from the integral equation, which is obtained by substituting eq. (1) to eq. {10}. Applying an iteration to the above integral equation, we obtain the first order of analytical formula for i- as follows
(11) |r>r?l
3. Stopping power for one dimentional case By use of the electro-magnetic theory, the stopping power S is given as follows
eXt f (2) «5"(r^~^) t^-(t.Vt2)] (14)
In eqs. (12)-(14). 4ff{%) means the potential from the polarisation cloud, in which the second term denotes the subtraction of the pure Coulomb part because such a part doesn't contribute to the energy loss process. j>*xt(2) indicates the number density of ternal charge if we consider a classically straight trajectory 106
of the projectile, "v* and T? denote tlw velocity and tlw iapaet parameter of the projectile. Bow we consider the systea, in which the electron density is treated to change one-dlaentionally in order to derive the sto- pping power formula for the cases of the surface and planar cha- nneling. Me take the electron density n in Inhomogeneous systea aa a function of X, and s^ la eq.(l4) as
. (15) la the above, v is the two-dlmentionally Fourier component for the Coulomb interaction, which is gi
In eq.(15), it is noted that the traaslatlonal lnvarlance bet- ween X and b in £"' is not satisfied, although such a kind of lavarlance is satisfied in y and s coordinates, q and M)/V corre- spond to the Fourier components over y and z. In £ given by eq. (11), (0 depends only on x, and (i£ -7f ).^ua(r^) has only the x-coaponent. Substituting ve and £ discussed here, and using the following two dlaentlonnal Fourier transformation
we obtain the integral foraula of stopping
"• 107
(u)
U)_(X) -*-*• |X-b|cosh» e v J)
l where 0J|1|X»sinh"' (q|MBeV/Wp(x)J and q^^ga*. The first tarn is corresponding to tbs usual Bathe formula, and other terse repre- sent the density-changing effect. (X-b)/lX-b| denotes the step function. It is wirthy to be noted that eq.(ls) Includes aa ar- bitrary density function. Therefore, It Is possible to apply the above foraula to one-dlsentlonally lnboaogeneous systea, for cxaaple, the surface and planar channeling problems. If we con- sider the case of 2aV2/nu_In the above, X0(2»p{x> lX-*l/V) denotes the nodi fled Beeeel fun- ction of second kind, which Is the typical function appearing in the theory of sealclasalcal aodel. If we use eq.(lf) in eq.(l«) within the fraaework avoiding the singular point at X-b, eq. (is) is reduced to one-diaentional problea.
4. Concluding remarks The theory of the Inhnaagenecms aany-electron system la the new and charenging field. In which there are many interesting prob- leas relating with the inelaetlc scattering. In 1M». Luadeylst reported the general fraaework of the theory oa fc for each a kind of system [7]. In IMS. we derived the first order of ana- lytical foraula for 6 in the inboaogeaeous syatas under the COB-
^^B^T ^taa^sT -la^Las^aa A>a^ a^^^aTavmv^^PBjaave^F £ ^Bj^B9as^ evs^s^^Ps/w^areBi ^Cmeles^ [1] M. Kitagawa, Mucl. Instr. and Math. B13 (19SS) 133. [2] J. Lindhard and M. Scharff, K. Dan. Vld. Salak. Hat. Pya. Madd. 27,.NO.IS (1*83). [3] J. Lindhard and A. fflntar, K. Dan. Vld. Salsk. Nat. Pya. Madd. 34, No.4 <1««4). (4] L. Radln and S. Lundqvlat, Solid Stata Phya. 23 (lMt| 2. (5] ff.k. Chu and D. Powara, Pnya. Latt. 40A (1972) 23. [6] T.L. Parrall and X.H. Ritchla, Phys. Rav. BIS (1977) US. [7] S. Lundqvlst and N.H. March. Thaory of tha Electron Gas (Plemu, Naw York, 1983), Ill
ON THE Z]3 CORRECTION OF THE STOPPING POWER
L Nigy* ml P. M. Edtemque
DepansmattD de Hsica. Facukad de Quunica UniversidaddelPafe Vasco.Apdo. 1072.20010 S«i Sebastian. Spsm
ABSTRACT
Higher-order correction to the stopping power proportional to Zj* is evaluated. The screening phenomenon of the bare projectile in an electron gas is characterized by simple Yukawa potentials. Second order Bom approximation for the scattering amplitude is employed to describe the elastic collision cross section. For low velocity intruder a comparison of the analytical remits with those obtained from a phase-shift calculation is made. For high velocity of the bombarding particle oar final expression for the Zj3 correction corresponds to the classical resnlt the polarization effect of a harmonic oscillator.
'Permanent address: Quantum Theory Croup. Institute of Physics. Technical University. Budapest. Budafbki ut 8. 1111 HUNGARY. 112
1. INTRODUCTION
Since the original paper of Bcthe1 the energy Iocs of a charged panicle in matter has been studied thoroughly for many yean because of its practical implications. The intrinsic difficulty of the problem resides in the dynamic many-body character of the interaction of the projectile with the electrons and nuclei of the material through which the projectile passes. In the theoretical description of the electronic stopping there are two. in principle essentially, different models for the target. The atomistic description treats the stopping medium as a dense atomic fas and uses the standard method for excitations with an implicit cumulative assumption. In the homogeneous electron-gas model the stopping can be studied by response function formalism.2 In these classic first-order perturbation theories the stopping is proportional to the projectile charge squared. Zj2. because of the applied first Born approximation for the transition matrix elements. The possible (initial and final) scattering states are determined by kinematicai constraints. The problem of the divergency of the Coulomb-logarithm is avoided in different manner in the two above mentioned model. In the atomistic description, where the electrons are not free, the minimum value of the transferred impulse is related to the binding energy of the electron and the perturbing potential is pare Coulombic one. The free electron-gas model "eludes" die difficulty of the zero impulse transfer for pure Coulomb potential by the screened potential picture. It should be emphasized, that these pictures have physical grounds. The final forms of the electronic stopping obtained in the two models show significant similarity and thus there is a broad overlapping region between the dielectric and the atomic treatments of the problem. On the other hand it is know from measurements3 that the stopping power exhibits a dependence on the sign of the charge in contradiction of the "rigorous" Zj2 proportionality. The explanations for the so-called Barkas effect are different, depending on the model of the target. In the atomistic description the first Born approxiaution for the particular scattering amplitude is justified because of the pure Coulomb potential of the bare projectile (Rutherford scattering). Consequently one needs to modify the kinematicai constraint. In this model the Barkas effect has been interpreted as a polarization effect.4 In the standard (simplified atomic model} harmonic oscillator description the effect originates in the nonnegligible displacement of the atomic electron during the collision. This calculation fives a term proportional to Z|3 for the Barkas effect in the expression of the stopping power. In the electron-gas model the validity of a perturbation approach (first Bom approximation) is not obvious a priori. Here the kinematicai constraint represents the well-described pan of the 113
theory. The higher order Born terms in the transition matrix elements are related to the screening nonlinearities in a self-consistent description. The role of this nonlinearity was demonstrated is the scattering theory approach of the stopping power, for positive io». This theory uses density-functional method for the determination of the screened scattering potential and correct phase shift analysis at low velocity of the projectile.5 This approach is inherently correct to all orders in Z[. The nonlinear stopping power differs markedly from the standard dielectric (linear) result for the interesting (metallic) density of the electron gas. Note, however, that the dominant pan of this is simply due to the different scattering description. For "linear" potentials ike phase shift calculation gives results which are in acceptable agreement with the nonlinear ones.Funhermore the deviation is partly related to the different schemes for treating the screening problem itself (exchange-correlation and static local field problems). The dielectric formalism retains the momentum-space representation of the (static) scattering potential. Consequently, the transition matrix elements are calculated in first Born approximation for the corresponding (immediately given) screened potential. For arbitrary velocity the quadratic response function theory gives a possibility for the determination of the Z|3 term. This theory is aa alternative approach to nonlinearities expanding the stopping power (the induced density fluctuation) in powers of the projectile charge. Unfortunately this expansion technique rapidly becomes unmanageable, even in piasmoa pole6 or random phase approximation (RPA) of the system7. In this work, as an intermediate approach, we investigate the efficiency of the second Bom approximation for simple Yukawa type potential within the framework of the scattering theory for the electron-gas model. For low velocity projectile an additive "reaormalizMion" of the screening parameter is employed to get an approximate self-consistency. In Sec.II we present the derivation of the stopping power within, a second order Born approximation. Our numerical results for proton and aatiprotou ate summarized in Sec.ni . Finally Sec.IV, is devoted to the
II. SCATTERING DESCRIPTION IN SECOND BORN APPROXIMATION
The energy loss per unit length dE/dR for an ion moving with constant velocity (v) in a homogeneous electron gas is given by* 114
an .... (2.1)
and
for snail [v « vF, Eq. (2.1)] and high [v » vF. Eq. (2.2)] velocity of the projectile, 2 1 3 respectively. Here n i« the density of the system. vF * (3 K a) ' the Fermi velocity
and 0tr the momentum transfer (transport) cross section. Atomic units ate used throufhout this paper. The "remaining" problem is the determination of the screened scattering potential for which we calculate the phase shift factors. Toe density functional method solves this problem in a self-consistent way. In order to establish the improvement we retain the momentum-space representation for the scattering amplitude f(q). Thus our method is directly comparable to the dielectric (linear and quadratic) formalism. This latter makes use the RPA. Our simple Yukawa potential
V(r) — h-t'at (2.3)
represents the real-space equivalent of the RPA within its Thomas-Fermi limit. The transport cross section in momentum representation is given by
2 V, °w where q is the transferred impulse and vr • vF for v « vF and vr « v for v » vF, (see Eqs. (2.1) and (2.2)]. Fortunately, the second Bora expansion of the scattering amplitude for the Yukawa potential is a standard result9 42?
2 4 2 2 2 2 where A « a + 4vr a + vr q . The first term is the Fourier transform of the potential and it is the scattering amplitude in the first Bom approximation. Close to 115
this first Born limit the arctg (x) may be replaced by it* argument x and after simple integration in Eq. (2.4) we obtain (up to the Z\ 3 order)
In this equation the corresponding expressions are
a|
2 where g = ( a / 2 vr ) .
It is easy to show thai in the g-*0 limit
(vj-li Z?(l+^.) in 2V*
is the asymptotic formula for the momentum transfer cross section in which the so-called Zj3 (Barkas) factor is
B=li2? 2(t In 2VT tr v4 J v2 a
In the high velocity (v » vF , vr • v) limit and using a « att./v we obtain for ike stopping power with Barkas correction (an approximation for 'a' is X/2, see Ref. 10)
This corresponds to the harmonic oscillator (atomic model for the target) description, where the Zj3 term originates in polarization effects (Me the Introduction). Urns for a positive charge (Zj positive) the stopping is bigger than for a negative charge (Z| negative). 116
It is very inteicninf to investigate the relevant rcnonnalizatiea of a in the static limit. We know that in this limit the Friedel. SUB rale is a strong self-consistency *1 requirement for the regular scattering potential. In order to obtain (at least approximate) self-consistency we calculate the a valve in Bom approximation by using the *2
(2.9)
Friedel sum rule. We suppose that the phase shifts are small. Then the expression for the forward scattering amplitude is given by
(2.10) VF
for elastic scattering of an electron with impulse vF on the screened potential. Note that 8*0 scattering angle corresponds to q»0 impulse transfer. Comparing these last two expressions we obtain
for the sum rule in Born approximation.
In the appropriate (close to the first Bon) limit, after sobthuting [ from Eq. (25)] the relevant formula for the forward scattering amplitude, we fieri
I/2 for the determination of a. In this equation OTF * ( *vF/it ) is the (approximate) RPA result. The solution of Eq. (2.11) is not difficult. A simple relation
(2.12) 117
fives a very good approximation even for vF«l. Note that this expression exact in ike
vF -» •• ; high density limit lor the Yukawa potential. In the following section our numerical results for proton aad aatipiwoa are summarized.
HI. RESULTS FOR PROTON AND ANTIPROTON
First, we sbow in Fig. 2 the screening parameter (a) as a fuactkw of the
density parameter r, • 1.92 / vF. The first Bora value (Thomas-Fcrni expiesskM) is
denoted by aTF(3). For proton and aatiprotoa the second Bon results (From Eq. 2nd 2nd (2.11)] are ap (2) and alp (5). respectively. For completeness this figure includes the "exact" values of a for these panicles; obtained by partial wave analysis and numerical integration requiring that the phase shifts satisfy Eq.(2.9). (These Fr Fr results are denoted by ap (l) and alp (4)] For the practically interesting density
range (1.5 £ rs £ 2.5) our momentum-space calculation gives fairly good results for the screening parameters. Our numerical results for the stopping power in the static limit (dE/dR * vQ)
are summarized in Fig.2. The curve 'a' is the first Bom result with aTF. (See Bos. (2.1) and (2.6) with XQ (g)]. For proton and antiprotoa there are two-two curves in secoad Bom approximation. Curve b\ is obtained with unmodified a * Ctjf valve; while bj 2n2 otJ -il£ 0 +1) sin [ 6\ (vp) - SM (Vp)] (3.1)
VF
Fr Fr and numerical phase shifts (with ap and a«p )• The deviations from curve V are notable for the density range of Funhennoie, for the case of proton the analytical (02) aad pure — erica! (03) results are in surprisingly good agreement. The carves b\ and b% differ nurfcedly. For the case of antiproton there is a big deviation from curve C3 at awtallic ihmitifi 2a- Fr of the electron gas. (Remember the parameters aap and aip were in better agreement!). The curves cj aad C2 remain very close. These latter curves "break down" at about r, « 2.3. This is uaphyskat. The poteatial is mote aad more "Coulombic", therefore a simple secoad Bon approximation for the scattering amplitude is not adequate in the case of the aaUanwoa. Now. however, that for r, « 23 118
the screening parameter of the proton potential becomes smaller tkan the
well-known critical vaiue Ocr « 1.19. In principle (because of the appearance of a possible bound state) a penurhative
approach is not valid for higher r( values. Finally, the curves 03 and c3 are nearly symmetric with respect to the 'a' curve. The last section is devoted to the comments.
IV. COMMENTS
In this paper we have investigated the efficiency of the second Bom approximation for simple Yukawa potentials to distinguish the proton and antiproton stopping power. We have found that the validity and applicability of this approximation is bounded. Only for very high velocity of the projectile or very high density of the system (in the static limit) gives the perturbative method meaningful results (g->0 in Eq.2.6). In the low velocity range the nearly symmetric deviations for the proton and aatiproton (curves 03 and C3) with respect to the first Bon result (curve "a") ace somehow "extremely regular". In the light of the above a recent result for the nearly symmetric and big («40%) deviations close to the stopping maximum (typical velocity is 2-3 a.u.) needs further justification. Without physical constraint between the first order (Zf2) aid higher order (Z\*) terms, the results and statements are only acadsnuc for intermediate and smaller velocities. At this state of the theoretical description we can say that the deviations from the "onodox" result* are and might be notable. The range of validity of the lymmnrir deviations must be restricted. The choice of any model potential requires a special care in the case of the autiparticle. because of the "growing'' Coulomhk name of the scattering potential. With decreasing density the source of this is the eflickat charge density depletion, at the static limit. The detailed self-consistent investigation for antiproton. using density functional method, is in progress.14 119
ACKNOWLEDGEMENTS
The authors are thankful for many useful discussions to Prof. Eugene Dr. Andres Amau. The author* gratefully acknowledge help and support by Eusko Jauriaritza, Guipuzkoako Fora Aldundia and UPV/EHU. One of us (P.M.E.) would like to Iberduero S.A. for its help and support. 120
REFERENCES
1 HA. Bethe. Ann. Phy«. 5, 325 (1930). 2 J. Lindhanl, K. Da. Vidensk. Mat. Fys. Medd. 28. so. S (1954). 3 W.H. Bttfcas. N.J. Dyer, and H.H. Hecks***. Pfcys. Rev. Lett. 11, 26 (19*3); LA Andenen et aL. ibid «. 1731 (19t9). 4 J.C. Ashley, R.H. Ritchie, and W. Bnadt. pays. Rev. MS. 2393 (1972): JJ5. Jaefcwa mi ILL. McCinhy, ibid M, 4131 (1972).
Ecbcniquc, R.M. Nieminea, J.C Ashley, and RJL Ritchie. Fhys. Rev. A33, S97 (1996). J.C Ashley, A. Giw-Marti, and P.M. Ecbeaiqoe; ibid A34,2495 (1986); A. Anito, PAt Lchrmpc, and RJH. Ritchie, Nucl. Lutnm. Methods B33, 138 (19M); L Nagy, A. Anua, aad PJf. Echenique, Fhys. Rev. B3t. 9191 (198t). 6 C.C. Sung Mad RJL Ritchie, Fhys. Rev. A2S. 674 (1983). 7 CD. Hu and E. Zaremba, Fhys. Rev. 137,926S (198S). 8 P. Sifmond, Phys. Rev. AU, 2497 (19S2); I. Nagy, A. Anwu. and P-M. FrbfiqT. ibid A4f, 987 (1989); L. Booij and K. .Srhwthnmmrr. ibid B39.7413 (1989). 9 R.G. Newton, Scanerint Theocy of Waves and Panicles (Sprinfer, New Yoric 1912) pp. 291-292. 10 j.LindhantNucLInstnun. Methods 132,1(1976). " C Mael, Quantum Thewy of Solids (Wiley, New Y«k, 1987) pJ43. 12 j. Fricdd, Philos. Mat> 43.153 (1952). ^^ H. Fibffnicn and P. Sisjiimiid, Ann. Phys. (to be pwbtished). 14LNaiy,A.Antaii,P.M.Ecfaenk|tte,andE.2^tipba.Phys.Rev.B(iobejirtiliifc«i0 121
For the notations: see the text 122
For the notations: see the text 123
For the notations: see the text 124
For the notations: see the text 125
Variational approach to the scattering theory and its application in stopping phenomena *
B. Apagyi and I. Nagy
Quantum Theory Group, Institute of Physics, Technical University of Budapest, 1581 Budapest, Hungary
Abstract.
The spurious singularities arising in the Kohn and Schwinger variational methods (applicable, e.g., to the calculation of quantum reactive scattering of atomic systems) can be avoided by using a particular least-squares procedure referred to as Ladanyi variational method. With the aid of the latter method, phase shifts of very high accuracy can be computed for low energy electron scattering by impurities. Some simple one- and two-parameter models are presented for proton and anti-proton stopping by an electron gas. The models fulfill K&to'E cusp condition and/or Friedel's sum rule.
" Invited paper presented at the 12th Werner Brandt Workshop on the Penetration of Charged Particles in Matter held in San Sebastian, 4-7 September 1989. 126
1. Introduction
In recent decade, considerable interest has been devoted to the solution of
quantum mechanical scattering problems by variational methods. This in-
terest is motivated by a number of important investigations such as, e.g., the
establishment of reference-standard cross sections for calibration of experi-
mental electron scattering apparatus |l-2j, calculation of vibrational popu-
lation inversion in chemical reaction J3j forming the basis of chemical lasers,
or computing electron-molecule collisions {4,0} and photoionixation processes
Compared to this activity, in atomic physics,.application of variational
methods in condensed matter (solid state) physics is relatively rare. It can be explained by the fact that the interaction used in solution of problems of solid state physics is seldom non-local and channel dependent. An advantage of variational methods over direct numerical integration techniques is the easier handling complicated interactions. For example, the dimension of scattering equations to be solved by variational method remains exactly the same if one replaces local potentials by non-local ones. By use of direct integration technique, however, the dimension becomes squared and this can tax even the largest computer in case of calculating complicated multichannel reaction processes. Since variational methods treat scattering wave functions on the same footing as quantum chemical calculations do bound states, another advantage of variational methods may be to facilitate the use of standard target functions determined earlier by quantum chemistry. 127
In scattering theory there are two baiic types of variational functional* providing stationary expressions for scattering K-matrix. One is the Kohn variational functional ]7] associated with scattering state Schrodinger equa- tion plus boundary condition. The other is the Schwinger variational prin- ciple |8] which is based on the Lippmann-Schwinger equation. Since, in practice, one always deals with finite set of basis functions, both methods suffer from anomalies, the so called spurious singularities [9-lOj. Because of these spurious singularities, neither of these functional becomes a maximum or minimum principle. Therefore, the stationary principle may be replaced by a stability requirement [11]. It means that simple nonvariational methods may also yield stable results in a fairly large region of the nonlinear scale parameter characterising the basis functions. Concerning detailed analysis of variational schemes in scattering theory, the reader is referred to the lit- erature [12].
In the next section, a particularly simple expansion method, the Lada- nyi least-squares variational method will be outlined along the lines discussed in 113). In section 3 the Ladinyi method will be applied to some simple scattering problems including scattering of electrons by local and non-local potentials. Calculation of stopping power of an electron gas for slow pro- ton and anti-proton is also presented here for models fulfilling Kato's cup conditions and/or friedel sum rule. Section 4 contains a short summary. 126
2. Ladanyi varational method
The simplest (and most practical) form of Ladanyi functional reads for one channel scattering as
where / denotes the trial scattering radial wave function which is expanded in terms of basis functions as
f{r) = £ Oi «For an 5 — wave scattering the basis consists of the following two continuum functions (3)
(4) and of N discrete (square-integrabie) functions which can be chosen, e.g, to be
^•si^e-"- for * = l,2t..,tf (5) with a and fi being nonlinear scale-parameters and Jt denoting the wave mimb«r which is r«lat«d to tht scattering energy in atomic units by E = 1^/2.
The functional (1) minimises the deviation
(6) within a test-function space {XH} involving only square-integrabie functions which are chosen for the present application to be
h r Xh{r) = Bhr e~f for h = 1,2,..,A# > N +2 (7) 129
where Bh is a normalisation factor and 7 is a scaling parameter
In Ladanyi functional (1) Wk* s are positive weighting factor* and tl» normalization is necessary to exclude the trivial (/ = 0) solution.
The variation of (1) with respect to the coefficients a_i leads to a simple eigenvalue problem which can be converted, by renormal- ising the coefficients to have a-i = 1, into a system of linear mhomogeneous equations (the Ladanyi equations) N Li3aj = -I,,_i, 1: = 0,..., N, (8)
and a simple expression for the eigenvalue A (referred to as the measure of ike error of the approximate solution)
(9)
where the least-squares matrix is defined as follows
(10) JC^i with the matrix elements
Bhi = p dr Xk(r)(H - E)^{r). (11)
The tangent of the approximate phase shift of Ladanyi method is obtained after the solution of (8) as K[N) = 00. (12)
In the calculation we will employ the following simplifications in the ap- plications of Ladanyi method: «% = 1 (h = 1,...,M), fi = 7 = o, and M = AT + 10. 130
Filially, it should be pointed out that Ladinyi method avoids the Kohn- type anomalies. That is because of the appearance of the square of the Hamiltonian matrix in equation (8).
3. Application of Lmdinji varsatiousd method
3.1 Scattering of an electron by nonlocal potential.
Let us consider the s- wave (I = 0} scattering of an electros by the one-channel nonlocal potential
(13)
representing the s-wave interaction between an electron (projectile) and a hydrogen atom (target) in the static exchange approximation. The local part of the potential reads
and the nonlocal part representing exchange of the two electrons can be written in the form
/ M («)
where the nonlocal kernel is given by
wiry) = 4(-lfe-'{[l + *a)/2 - l/r^t^' (16)
with 5 = 0 for singlet scattering process as, and 5 = 1 for triplet scattering states. 131
In Table 1 the tangents of the triplet scattering phase shifts due to potential (13) are listed for various approximations containing N basis func- tions together with the eigenvalue A indicating the measure of the error of the approximate solution. The remarkable convergence and stability property of Ladlnyi method as well as the non-appearance of spurious singularities within this method (13] make it reasonable that this procedure may be used as a practical computational tool in solution of scattering problems.
S.2 Stopping power of an electron gas for slow proton and anti-proton.
Let us consider a jeilium as stopping medium with no = Q.75/(jrrJ) electrons
per unit volume, and take a proton (Z = I) or anti-proton [Z = -1) inter-
acting with the electrons via pure elastic collisions. If the velocity v of the intruder entering the jeilium is much smaller than the Fermi velocity (Jbp in atomic units) of the electrons, the stopping power can be written as vQ(kp) where the transport collision frequency Q is computed from the formula
Q=^ where r\i is the /th partial-wave phase shift of the electrons scattered by the screened potential V of the intruder.
In the remaining part of this subsection we consider two models for slow proton/aati-protou stopping in jellium. Both models fulfill Friedel sum rule 132
and the second one satisfies also Kato's cusp condition
n'(r) = -tZ (19) n{r) r=0 with n(r) = no + 6n(r) being the total density around the intruders and 6n(r) denoting the induced density. Both models automatically fulfill the total screening condition = Z. (20)
In both cases one assumes the Poisson equation to be valid between induced density and scattering potential
AK(r) = -4*6n(r). (21)
3.&.1 One-parameter model for stopping of $low proton mnd anti-proton kg jellium.
Let us consider the Yukawa potential (ft > 0)
V{r) = z£^- (22)
as a crude model of the potential induced by a proton/anti-proton entering the jellium with slow velocity.
Although such a simple Yukawa potential model leading to infinite den* sities at r = 0 is obviously unrealistic, it is interesting to compare Q values obtained for proton and anti-proton stopping at various densities r,. The requirement that Friedel sum rule (18) be satisfied can be made by adjusting the free parameter ft at every densities. Such an adjustment has already 133
been done by Ferrell and Ritchie (14] for a proton (Z = 1), therefore the calculation will be extended only to the case of an anti-proton {Z = — 1).
The results are contained in Table 2. The marked difference between the Q values for proton and anti-proton can be interpreted as a manifestation of the well known Barkas-effect in this simple model.
3.2.2 Two-parameter model for stopping of slow proton by jellium.
Following ref. [15], consider the ansatz
2 6n[r) = V0(p, v; r,)^-*" + (1 - Vo^r.))^" " (23)
with the obvious conditions fi^v and ft,u>0. This ansatz automatically fulfills the total screening condition (20).
By invoking cusp condition (19), we get an explicit expression for the
"strength" Vo as
,r,) (24)
3 with a = fi {fi -1) and b = J/3(I/ -1).
The effective electron potential is derived by solving Poisson equation (21) with the above charge density. The result
r V{r) = -Vue-*» {p. + \) - (1 - l'0)e-*"{* + \) (25) is the superposition of exponential and Yukawa potentials with different "range" parameters p aud v. 134
Fixing v at some reasonable value (e.g., v - 0.3), one can adjust the parameter p so that equation (18) is fulfilled.
In this way one obtains a two parameter description of proton stopping in jellium which obeys the exact rules (18)-(21). Therefore it is reasonable to compare the values of 6n(0) and Q with those calculated by more exact density functional methods. This comparison is presented in Table 3. One can conclude that the results are satisfactory in the light of the simplicity of the calculation.
4. Summary
Application of variational methods is scattering theory is of great importance from practical point of view. They keep the calculation at a manageable level and facilitate the use of standard target wave functions from quantum chemistry.
A particularly simple expansion method, the Ladinyi method has been applied to solve some simple scattering problems including stopping of slow protons and anti-protons by homogeneous electron gas of various densities. The impurity problem has been solved within several models which fulfill Kato's cusp condition and/or friedel sum rule. 135
Referencei
11] R. K. Nesbet, Phys. Rev. 20, 58 (1979)
[ 2] R. K. Nesbet, Variational Methods in Electron-Atom Scattering Theory (Plenum, New York, 1980}
[ 3] Y. C. Zhang, J. Z. H. Zhang, D. J. Kouri, K. Haug, D. W. Schwenke, and D. G. Truhlar, Phys. Rev. Lettere 80, 2367 (1988)
[ 4] M. A. P. Lima, K. Watari, and V. McKoy Phys. Rev. A 30,4312 (1989)
[ 5] C. W. McCurdy and T. N. Resdgno, Phys. Rev. A 40, 1297 (1989)
[ 6} R. R. Lucchese, K. T^katsuka, and V. McKoy, Phys. Rep. 131, 147 (1986)
| 7} W. Kohn, Phys. Rev. 74, 1763 (1948)
( 8] J. Schwinger, Phys. Rev. 72, 742 (1947)
{ 9] C. Schwartz, Ann. Phys. NY 16, 36 (1961)
[10] B. Apagyi, P. Levay, and K. Ladinyi, Phys. Rev. A 37, 4577 (1988)
[11] K. Ladanyi, P. Levay, and B. Apagyi, Phys. Rev. A 38, 3365 (1988)
[12] M. A. Abdel-Raouf, Phys. Rep. 108,1 (1984)
[13] P. Levay and B. Apagyi, J. Phys. B 21, 3741 (1988)
[14] T. L. Fen-ell and R. fl. Ritchie, Phys. Rev. B18,115 (1977)
[15] B. Apagyi and I. Nagy, J. Phys. C 21,3845 (1988)
[16] C. 0. Almbladh, U. von Barth, Z. D. Popovk and M. J, Stott, Phys. Rev. B 14 2250, (1976)
[17] P. Jena and K. S. Siugwi, Phys. Rev. B 17, 3518 (1978)
118] J. Gondzik aud H. Stacuowiak, J. Phys. C 18, 5399 (1985)
119] M. J. Puska and R. M. Nieminen, Phys. Rev. B 27,6121 (1983) 136
TABLE 1. Triplet reactance matrix elements K — tanijo and the eigenvalues A com- puted by Ladanyi methods at several basis-set sixes JV for the nonlocal potential (13). The wave number k is fixed at 0.5 a.u. and two values of the nonlinear scale parameter a have been chosen to exhibit the stability of the method.
a = i.O a = 1.5
A* K{N) X(N) K{N) A(JV)
2 2.07761368 7.4 x 10~5 2.07028445 1.1 x 10~* 4 2.06977848 2.6 x 10~7 2.07005340 1.6 x 10~10 6 2.07011720 2.2 x 10"10 2.07006643 8.5 x 10-w 8 2.07006473 2.3 x 10"l4 2.07006656 1.6 x 10"ir 10 2.07006721 5.7 x 10-" 2.07006665 1.4 x 10-" 12 2.07006646 1.3 x 10"11 2.07006663 2.1 x 10-ai 14 2.07006663 8.4 x 10~31 2.07006663 6.9 x 10-" 137
TABLE 2. Calculated values of Q for Yukawa potential (22) with parameter fi adjusted so that Priedel sum rule (18) be satisfied within an accuracy of one percent. 2 = 1 means proton, Z = -1 stands for aiiti-proton.
Z=-l Z=l
r, p Q p Q
0.5 0.998 0.328 1.03 0.435
1 0.718 0.173 0.800 0.309
1.5 0.585 0.107 0.688 0.248
2 0.496 0.074 0.620 0.205
3 0.390 0.042 0.566 0.136
4 0.325 0.027 0.550 0.089
5 0.280 0.019 0.550 0.061
6 0.248 0.014 0.554 0.043
10 0.175 0.006 0.580 0.016
15 0.133 0.003 0.585 0.007 138
TABLB3. potential (2$). p kinds* OJiPhafbrn adjwtod to that «qa»- tion (IS) hokk Tbe vaiset pndictod by tdf-couwtent cakuia- tioo» am lilted lor computes.
1 1.331 0.722 0.327 OJ03 — — — 1.S 1.155 0.4S9 0.264 — — — 0.301 2 1.081 0^70 0J07 0.492 — 0.461 0.269 3 1.020 0^21 0.127 OJ04 OJK 0^72 0.163 6 1.006 0.298 0.041 0J34 — 0JI9 —
*AhnbUdh * JtM ud Sincwi (17) e Gcadnk and Stadwmk 139
•To*»iMa tons jua> BSJMU x* io» muu
J. Bausells
Trabajo presentado «n «1 "XII W«rn«r Brandt workshop on th* Penetration of Charged Particles in Matter", San Sebastian* 4-7 Septieabre 19<9. 140
STOPPING POWER AND RANGES IN ION IMPLANTATION
J. Bausells Centro Nacional de Hicroelectrbnica (CSIC) Universitat Autdnoma de Barcelona, 08193 Bellaterra, Spain
1. Introduction Ion implantation is the introduction of atoms into a solid substrate by bombardment of the solid with ions. It has important applications in the fields of materials science and microelectro- nics, where it is a fundamental technique for semiconductor device fabrication. In most cases succesful use of ion implantation requires an accurate knowledge of the implanted ion range distri- bution inside the target. The ion range distribution is obtained from the stopping of many ions that follow different trajectories inside the solid, and is therefore a process of statistical nature. Two different sta- tistical approaches are widely used to calculate ion range distri- butions: the Boltzmann transport-equation1'' and the Monte Carlo method*-'. In both cases the ion stopping power inside the solid is needed to calculate the range distribution. The ions lose energy by elastic ion-ion interactions and by inelastic electronic stop- ping. The elastic ion-ion interaction is a well-known classical two-body collision where the main uncertainty comes from the calculation of the interatomic potentials. Statistical models of the atom have been widely used in calculating elastic stopping powers. These include the Thomas-Fermi', Moliere*, Lenz-Jensen* and Bohr1' potentials. Recently, improved interatomic potentials have been obtained"'" by using Hartree-Fock atomic distributions. Inelastic stopping involves complicated processes of electron capture into (and loss from) the bound states of the moving ion. These processes determine the charge state and effective stopping- power charge of the ion as it moves through the solid. At high ion velocities, v » Ztv, (where Zl is the ion atomic number and v, the 141
Bohr velocity), the ion is stripped of its electronic charge. In this case the well-known Bethe-Bloch" linear theory for the stop- ping power can be used. At very low ion velocities, v ^ v,, the ion is dressed with quasi particles. In this velocity region large nonlinear effects are important, and the inelastic stopping power can be accurately calculated within the density functional forma- lism""". The stopping power is in this case proportional to the ion velocity. At intermediate velocities (of the order of the average electronic orbital velocity in the Thomas-Fermi atom, v * Z"*v.) an equilibrium ion charge state is reached and the stopping power calculation becomes more difficult. Ion implantation usually involves relatively light ions (8,11, o, p, si, Ar, Ge, As,...). For many years, standard ion implanta- tion applications have covered the energy region from 5 to 200 KcV. At 200 KeV, the lightest (and therefore, the fastest) ion considered, B, has a velocity of v-0.857 v., and therefore all possible ion implantation calculations are restricted to the low energy region, where the inelastic stopping power can be accura- tely calculated. Good agreement can be obtained in this case between experiment and theory. In particular, the TRIM Monte-Carlo code'-", which uses the universal Hartree-Fock interatomic poten- tial of ziegler et al." and a semi-empirical low-velocity electro- nic stopping power that agrees with density functional calcula- tions, provides quite realistic predictions, with an overall accuracy of a few percent for the range of ions in solids. In recent years, however, "high energy" ion implantation (usually up to *4 MeV) has been heavily investigated and is expec- ted to play an important role in semiconductor applicationsu>>*. At 4 MeV, boron, phosphorus and arsenic ions, for example, move with velocities of 3.83v(, 2.28v, and 1.47v,, respectively. Thus most ions used in ion implantation (and specially the lightest ones) are well inside the intermediate velocity region, where ion charge Jtate and stopping power calculations are not very well defined. As will be discussed in the next section, inelastic stopping power theories in the intermediate velocity region have been using average ion charge states to calculate ion effective charges. Recently, Sols and Floras" have obtained the distribution of charge states for light ions in the upper part of the intermediate 142
velocity region. As the implanted ion range distribution is a statistical result, the use of this distributions of charge states instead of average values in the calculation of stopping powers can lead to a different shape in the ion range distributions. In this work we study this effect in order to ascertain whether the use of charge state distributions can add new features to ion range distributions in the MeV-energy range or whether its effect is snail and need not to be considered.
2. Charge states and stopping powers at intermediate velocities
The stopping power Sn of an ion of atomic number Z, can be related to the proton stopping power S, through the concept of the ion stopping power effective charge1* Z'
jr (l) where S,, and S, are determined in the same material and at the same velocity. The fractional effective charge % of the ion is defined as z: -lzt (2) As proton stopping powers are well known"'", knowledge of the effective charge of an ion allows the calculation of its stopping power. The effective charge of an ion moving in a solid depends on the ion velocity, and is related to the charge state of the ion, i.e., to the occupation of the ion's bound states. If an ion of atomic number Zt moves in the medium carrying N electrons, then its ionic charge is Q«Zt-K, and the degree of ionization is Q M « - — - 1 " IT <3) Brandt and Kitagawa"-" calculated the relation between the fractional effective charge X and the charge state q of the ion. They assumed that the electronic charge bound to the ion extends over a radius A. Then the electrons in the medium at impact parameters greater than A encounter a fully screened ion of charge Q, but at smaller impact parameters they penetrate into the screening cloud and find an ionic charge greater than Q. Therefore one expects that 1 >q. within linear response theory, Brandt and 143
Kitagawa found" the result
K - q + C(r.) (1-q) In 1 +
where kr is the Fermi momentum of the solid, a, is the Bohr radius, and C(r.)» 0.5 is a weakly dependent function of r., the electron gas density parameter. From extensive data analysis it has been found" that a good approximation is C(r.) * \ J^-j The next step needed is the evaluation of the ion's degree of ionization (or charge state) q, which depends on the ion velocity and the target material. Bohr" suggested that the ion is stripped of all electrons with classical orbital velocities lower than the ion velocity. Kreussler, Varelas and Brandt** showed that at velo- cities of the order of v, this stripping criterion should be ap- plied to the relative velocity vr between the ion and the conduc- tion electrons of the medium. This gives, in the Thomas-Fermi model of the ion, v, which can be written in an explicit font using the result on the Bohr stripping criterion obtained by Northcliffe1* and substituting v by vr ''
1 - exp - „ , „, 1 (6)
This expression agrees with Brandt's calculation" of q(vr), and fits remarkably well the experimental data for several thou- sand stopping powers11. At low ionization fractions (tj<0.3), howe- ver, there is a deviation between the theoretical expression and the experimental results. Ziegler et al." fitted an expression based on the font (6) to the experimental data, which gives M q - 1 - exp[.8O3 y/-» -1.3167 yr -.38157 yr -.0089*3 y/1 (7) whera yt« vr/(v. z,'"). He will call this expression the Brandt- Ziegler, Biersack and LittaarJc (B-ZBL) formalism to the charge stat«. The above theories provide average values for the ion charge 144
states . Recently, Sols and Floras" (SF) have calculated the cross sections and charge states associated with the electronic capture and loss of B, C, N and 0 ions Moving in aluainua with velocities
Z.v, ^ v .5 3Z,v0, where the is orbital of the projectile ion con- trols the capture and loss processes. Their results for the dis- tribution of charge states for a B ion as a function of its velo- city are shown in fig. 1.
1.0
o.a
o.s
0.2
0.0 it V/5VO Fig. 1. Distribution of charge states for B ions noving in Al, as a function of the reduced ion velocity. The solid lines are the results of Sols and Flores". The dashed lines are an extension of these results to the low-velocity region.
The charge state of an ion greatly influences its stopping power, as can be seen in fig. 2, where we show the inelastic stepping cross sections calculated froa eqs. (1) and (4) for the six possible charge states of boron and froa the B-ZBL theory. At a given velocity, two different B ions can have different charge states according to the probability distribution shown in fig. 1, although on average they would have the charge state given by the B-ZBL theory. This aeans that individual ions will have, during their slowing down, stopping powers different from each other and froa average stopping given by eqs. (1), (4) and (7), and this can lead to a different result for the ion range distri- 145
bution. This is a kind of energy loss straggling, because we get unequal energy losses for ions that travel under identical condi- tions.
6OOOr
Fig. 2. Inelastic stopping cross sections for boron ions in Al. Dashed lines: cross sections for B ions with charge Q. Solid line: ions with average charge stats given by eq. (7).
As the B ion is the lightest (and therefore the fastest, at a given energy) of the ions usually employed in ion inplantation, this difference in range distributions will be, at a given energy, maximum for boron. Therefore we will concentrate our study on B ions. The difference between the two formulations can be seen by calculating the energy distribution of 25 HeV boron ions (v» 9.5S v.) after traversing 24 pm of Al. This gives an average final velocity of 4.86 v,, and thus covers the velocity rang* of the SF calculation, we have used the TRIM Monte-Carlo code", which in its standard fora follows the B-ZBL formulation, and a modified ver- sion to take into account the SF distribution of charge states in which, for each ion and each velocity, the charge state is calcu- lated according to this charge state probability distribution. Mote that this kind of calculation is implemented very easily in a Monte-carlo formalism, that follows individual ion trajectories 146
in the target. The results are shown in fig. 3. It can be seen that the energy distribution from the SF charge state theory is much broader than the B-ZBL theory distribution, due to the va- rious charge state possibilities that each ion has available in the former case. Also, the Bean final ion energy is lower in the SF formalism, due to the fact that the Q«5 charge state, which is the most probable over most of the energy range covered by the ions as they traverse the target, gives a stopping power greater than the B-ZBL formalism, as can be seen in fig. 2. It is therefo- re to be expected that there will be differences in the final ion range distributions, when using the actual distribution of charge states of B instead of the B-ZBL formalism.
10000 r
1000
in O
E/EO (*) Fig. 3. Energy distribution of B ions of initial energy Eo- 25 MeV after traversing 24 ft of Al. Solid line: ion charge state given by the B-ZBL average formulation. Dashed line: ion charge states given by the Sols and Flores" distribution.
In order to make an estimation of the importance of this effect in ion range distribution calculations, we need to know the distribution of charge states from v-0 to v» 5v,, where the SF calculation starts. If ft(v) is the probability of the Q*i charge state am a function of the ion velocity, we know that 5 Z f,(V) - 1 (•) i-0 147
on the other hand, the average charge state at a given velo- city v must agree with the empirical result q^Jv) given by (7) 2 £ fAv) - g».(v) (9) i-l D Indeed, the SF probabilities verify eq. (9) in their velocity range of validity, between Sv, and 15v,. Taking into account these two conditions, we have extended the results of SF down to v»o with the results shown in fig. l. He do not intend that this is the true charge state distribution for boron. Our purpose is to obtain a reasonable charge state distri- bution in order to make an estimation of the differences in ion range distributions that can be obtained when using it instead of an average value for the charge state. It is clear that (8) and
(9) alone do not completely determine the form of the six ft(v). But we have to note that f, and f« are almost completely determined by the SF results and, if we assume that f, (i>l) do not extend to v-o, then f, is fully determined by (8) when fx is made to agree with (9) near v=0. This leaves ft, f, and f, to be determined. But (8) and (9) are not easily fulfilled simultaneously, so that we believe that, although our solution is not the only possible, it should not be very different from the actual distribution of charge states for B ions.
3. Results We have calculated, using the TRIM Monte-Carlo code, the ion range distributions of boron ions implanted at 4 MeV (v» 3.83vJ in aluminum at normal incidence, using the B-ZBL averaged charge states and the extended SF charge state distributions discussed above. 4 MeV is the maximum energy of standard high-energy implan- tations, and therefore this calculation should give an upper limit to the differences that can be expected between the two formalisms in actual implantations. The results are shown in fig. 4. The maximum of the ion range distribution moves from about 4.8 jn (B- 2BL) to about S.O jm (SF), which is a noticeable 4.2 % shift. Ibis fact can be understood in terms of the ion stopping powers shown in fig. 2 and the ion charge state distributions shown in fig. 1. The mean total range of the ions can be estimated from 148
(10)
and it is easy to verify that an ion has a greater probability of going deeper (than when using average charge states) than of going shallower. Also, the SF ion range distribution is less skew and peaked (skewness—2.3, kurtosis- 18.4) than the B-ZBL one (skew- ncss*-l2.7, kurtosis- 230).
100000
30000 40000 50000 Depth (A) Fig. 4. Range distributions of boron ions implanted at 4 MeV in Al. Solid line: ion charge states given by the B- ZBL average formulation. Dots: ion charge states given by the extended Sols and Flores charge state distribu- tion.
There is no available experimental data on MeV boron implan- tations in aluminum, but there are several results for boron in silicon3*"*. In order to verify whether the effect we are discussing gives a better agreement between theoretical calculations and experiments, we have calculated the ion range distribution of B ions implanted at 2 MeV in Si. Me have used the same extended SF charge state distributions as in Al, which should be « good appro- ximation because the electron gas parmeters of both materials at* very similar, and thus their electron stripping properties from the moving ion should also be similar. 11M results are shown in 149
fig. 5 and table 1. Zt can be seen in fig. 5 that the extended SF range distribution gives a better agreement with the experiaental SIMS data of Ingram et al.M than the B-ZBL range distribution. The use of more accurate values for the charge state distribution of B ions in Si would probably give an even better agreement with the experimental data. There is a disagreement in the deeper part of the distribution due to channeling effects which are not included in our calculation. The results of table 1 for the momenta of the range distribution show that its overall shape is in much better agreement with the data of Hong et al." in the SF calculation than using the averaged B-ZBL formalism.
1000-d
zbcioo 22600 ' I _ Depth (A) Fig. s. Range distributions of B ions implanted at 2 KeV in Si. Solid line: ion charge states from the B-ZBL average formulation. Dashed line: ion charge states given by the extended SF charge state distribution. Dots: experimental data".
4. Summary and Conclusions Ion implantation in the MeV-energy range is expected to play an increasing role in semiconductor applications. This energy region involves, for light ions such as B, the intermediate ion velocity range, where capture and loss charge exchange processes between the ion and the medium determine the ion charge state and stopping power. Presently there are no reliable theories to calcu- 150
Table 1. The first four momenta of the rang* distribu- tion of 2 MeV B ions implanted in Si: Mean projected range (Rp), range straggling ( ARp), skewness and Jcurto- sis. The two theories discussed in the text are compared with experinental results.
exper." SF B-ZBL
Rr (A) 29720 27872 27092 AH, (A) 2060 1706 2383 Skew. I -1.94 -2.68 -11.24 Kurt. 6 15.8 18.8 160
late ion range distributions in this energy range1*. We have calculated range distributions for MeV boron ions implanted in Al and Si. We have used a standard average approxima- tion to the ion charge state as a function of its velocity, and compared this result with a calculation involving an approximation to the actual B ion charge state distribution. This distribution has been obtained from Sols and Floras" for v>5 v, and has been extended to lower velocities in such a way that the average ion charge state at each velocity agrees with the semi-empirical average charge state of Ziegler et al." We expected that, by allowing different ions to have different charge states (and therefore different stopping powers) at the same velocity, the use of charge state distributions would modify the shape of the final ion range distribution, in comparison with that obtained with the standard approach. The results of our calculations show that the differences between the ion range distributions obtained with the two formula- tions are relevant for B at 4 M«*V. Furthermore, by using charge state distributions we get a better agreement with available experimental range distributions for B in Si at 2 MeV than when we use th« standard average ion charge states. We conclude that first-principles calculations of charge states for ions moving in solids between v, and *Zv« can be very helpful to the understanding of implanted ion range distributions in the MeV-energy range, a field that is becoming increasingly important in semiconductor technological applications.
References 1. J.Lindbird, H.SdJirff and l.E.Schiott, K.Du.VidMsk.Stlks.Mat.rys.Mi. 33, lo.11 (1963). 2. K.B.HisUrbot, Cu.J.fbyi. 46, 2479 (1961). 3. O.littnark ud J.F.Ziogltr, Phys.lw. 123, 64 (MM). 4. J.F.CibbOM, Hucl.Instr. ud fcta. B21, 13 (1917). 5. X.T.lobiMM and IXftntM, Phys.lM. M, 5001 (1974). «. J.P.Bitruck ud L.G.IMpitk, MucLIostr. ud Httk. 174, 257 (19M). 7. l.So«erftld, Z.Physik 71, 213 (1932). I. C.ltolitrt, I.ltatutforscb. 12, 133 (1947). 9. tf.Unz, Z.Physik 77, 713 (1932); LJeasea, Z.Pbysik 77, 722 (1932). 10.1.Bobr, K.Du.VidtMk.Stlks.lat.Fys.ltodd. II, KO.I (1941). II. S.D.Bilson, L.S.lMptck ud J.P.Bitrsack, ftys.m. BIS, 2451 (1977). 12. J.F.ttigltr, J.P.Bitnack ud O.Littiirt, "Ttit Stoppia; ud tuqt of Km u Solids'. Vol. 1 of lit Stopping ud tups of ions in Kitttr", Ed.J.r.Zioql«r, ftrqm* Ms, tm York (1915). 13. I.l.Bttbt, Z.Pbyiik 7C, 293 (1932); F.Bloch, ino.Paysik 16, 217 (1933). 14. P.R.Edmiqut, E.H.»i«tiaM ud IXtitchit, Sol.St.CMM. 37, 779 (1911). 15. P.M.Ecbttiqut, l.X.Vieuwi J.C.lsUty ud I.l.Iitdiit, tbft.Uf. 133,197 (19M). IS. V.Biibtru ud P.H.Ecbttiqut, J.Phyi. B19, Lll (19M). 17. P.I.EcbMiqut, Ihiel.Iattr. ud Btth. B27, 2SS (1917). II. J.P.Bitrsack, fucl.Instr. ud Nttb. B3S, 205 (1911). 19. P.I.Siapr, Seucond.lBttmt., I, *>.9,92 (Sept. 1917). 20. H.Brndt, in 'ltotic Collitieot ii Solids', Vol. 1. p.»l, Eds. S.Dtti, B.l.lpplttoB aid CO.IOik, Kmm, tm York (1975). 21. H.I.lndtrstt ud J.F.Zitqltr, "lydroon Stopping Powrs ud lugts in All Hants', 7ol. 3 of Tbt Stopping ud lugts of ioas in lUtttr', £d.JJ.Zi*gltr, PetoHW frtss, «w lock (19IS). 22. H.Brudt ud U.liUqm, Pnys.teT. B25, 5631 (1912). 23. H.Brudt, tucl.Instr. ud fcth. 194, 13 (1912). 24. l-.Bobr, Pbys.ltT. 51, 654 (1940); P&ys.Btr. 59, 270 (1941). 25. S.KrwHltr, C.Tanlas and ff.Brudt, Pnys.Btr. B23, 12 (1911). 26. L.C.IortncliK*, Pnys.Ier. 120,1744 {I960). 27. F.Sols and F.Flocw, Pbys.Brr. 137, 14(9 (1911). 21. D-CIngm, J.l.Baktr, D.l.italsb ud E.StratlmaB, lud.Instr. and fcth. S21, 460 (1917). 29. I.Bang, E.0ug, I.H.Cbtag, P.K.Cbu, t.X.Stratmm ud I.O.Stratbnu, lael.Iistr. ud HO. Btt, 447 (1917). 30.1.Tuun, I.totsnaki, T.aada and E.Iitani, lucl.Isstr. ud Ittk. B21, 131 (1917). 153
DYNAMICAL EFFECTS IN ELECTRON TUMNEUNO
and L.MaruJic Department of Physics, University of Zagreb, POB 162, ;i001 Zagreb, Croatia, Yugoslavia
Recent advances in materials science, particularly in high precision fabrication (molecular beam epitaxy) of heterostruetu- res and quanttn wells with up to atosdc •onolayer precision, have led to rapid development of aesoscopic physics - studies of phe- noaena on a subnanoaeter scale. This paper will treat one parti- cular aspect in the theory of electron tunnelinc in these structures: the role of djmaaical screening and potential* in very thin carriers (e.g. doped Al Ga As layers). The aodel is based on the exact treatment of surface/interface plaaajon excitations in layered structures and their nonlocal coupling to the electron current, and is solved in the W£B approxismtion. Analytic expression is obtained which contains the effects of screening dynamics, electron energy and decay, parallel velocity and recoil. Several cases are discussed, including self-consis- tent results and tunneling rates. 154
Recent experimental advances in the ••ainra—nt of timeline through well characterized barriers, and the observation /I/ of possible dynamical effects for tunneling tiaes coaparable to the plaaaon/screening tiaes have again emphasized the need for a theoretical description of effective barriers seen by the tunneling electrons, that would take into account the dynamics of the charge fluctuations and their coupling to the tunneling electrons.
Very early, indications of deviations froa classical potential were observed experimentally, and several atteapts were made to find sesdeapirical corrections /2/. More recently it Mas realized that the origin of iaage potential is in the electron interaction with polarization aodea in the solid, in particular surface {or interface) plasaons /3,4/ which also led to the seaiclassical foraulation of the dynamical iaage potential for uniformly moving particles /5-8/. Application of this surface plasaon model to electron timeline Mas attempted by several authors /9-15/, for various experisantal situations and trying to explain different physical phenoasna, like currant oscillations in photoassisted field emission /I1,12/. Two papers /9.15/ are especially relevant for the present work; Jonson /*/ formulated the iaage potential in teraa of a non-local selfenergy which he approximately calculated for dispersionlcsa surface plasaons (SP) in a barrier outside the seaiinfinite aatal. In the first (nonselfconsistent) version his results contain soaa difficulties, probably arising froa the problaa of treating nonlocality in different approximations at various stages of the calculation, as we shall discuss later. Persson end Baratoff /15/ avoided these traps by choosing carefully their raraatfrs and calculating the tunneling rates directly, without discussing the iaage potentials and barrier shapes. This is probably sore correct physically, because the role of fluctuating potentiala in this situation cannot be fully represented by their (static) averages, but nevertheless we feel that it should be also useful 155
to calculate anl discuss the shape of the tunneling barriers. Therefore in this paper /W we set up • dynamical image potential aodel haeed on the electron coupling to two eurfaoe (or interface) plaamons, in the metal-insulator metal geoastrjr, appropriate to the GaAs-AlGaAe-aaA* systaa /I/, taking into account their dispersion and exact coupling functions. Ihe potential is calmlatort in the seoond order perturbation theory* and then eade aelfooosistant {e.g. bjr iteration). He systematically apply MB approximation, first bsonuae of its simplicity which enables us to evaluate analytically all results, and also in order to connect with the recant work on tunneling, especially related to the concept of tunneling (or dwell, or phase, or delay) tiaes /16.17/.
The analytic results for the dynsnical potential barriers are presented in Sacs. II. and III., and discussed in Sees. IV. and V. in Sec. VI. we briefly coapare them to the nonlocal results of refs. /9/. In Sec. VII. we study the Modification of tunneling rates due to the dynamical effects, again in the MB approximation.
The hamiltonian of the system
H x H# + H>f> •
The electron kinetic and potential energies are
H9 s K • V Is) «> where v includes the tunneling barrier. The energy of surface/ /interface plammons of the mstml-liiiJmtnr ms*sl symtsm (Fie.II is 156
where i » (Q,p) Q is the parallel wavevactor and p::l is the SP mode parity (in a syaoetric configuration). Fie2A shows the
frequencies of the syanetric ( where, inside the barrier ,'18,4,8/
The shape of the coupling functions T as shown in Fii.lb.
As the problem is essentially nonlocal, we define an effective local dynamical potential as
«yz,E> = J2k(z,z';E) «k(z')dz7«k (6)
where k is the parallel electron noMntua and «k is the solution of the nonlinear self consistent equation
»a *' (7)
The effective potential
V«ff(z> = V(z) + VZtE) (8)
contains the external potential barrier V(z) and the selfconsistent induced potential W. •* is the effective electron MUM, and E is its "perpendicular" energy. 157
The self energy in <6) is (Fig.2a)
1 Zk i
where the total energy E» = E ••hV/a/ (11)
includes the parallel kinetic energy, and
In the WKB approxissition the Green's fraction of an electron v z with energy E iifn a slowly varying potential ,ff ( ' can calculated to be:
* G(z,z';E) = e \ (121 / M{Z)MiZ')
where * 1 «For a constant potential VQ this reduces to the version used by Jonson /9/.
It is worth stressing again that this WKB Green's function becoaes less accurate near the turning points, i.e. near the edges of the potential barrier, and for electrons tunneling with energies close to the top of the barrier. However, as we shall be ultimately interested in the tvameling, i.e. electron attenuation, where the boundary regions give —11 ir contribution, we feel that WKB, as usual, provides a flood approximation. 158
III.
Mow we shall aaltc two crucial stapa, again justified by the assuaptions underlying the HKB acthod. First, for slowly varying x(x) the exponent in (12) can be approxiaated by
Now one could generalize the jcethod of Jonaon /9/ and study non- local potentials for the case of slowly varyinc potentials, becauae his constant potential aodel cannot be aid* self consistent.
Instead, we shall here notice that the decay constant
^•j- fv(z) - E + *». + 4^(0)]
>*ere iE^Q) = fi2 (Qa + 2k-Q)/2B*, corresponding to the Graen's function in the self energy (10), is large, and therefor* aakc a local approxiaation
G(z,z';qV/2«*) = - -^-S(z-z') (15)
q -» « this Mounts to averaging (6) over z' on a scale "q. In tens of ^i•glass this aeans that we have appradsated the diagrm (a) in Fig.3 by the diagram (b), by contracting z'=z, in the usual saaiclassical approxia»tion.
Doing so, we have given up the possibility to treat the fluctuations in the tunneling current due to quantua aacfesaical interferences and siadlar |fiaiiiB»Ma which are outside the MB spproxiaation anyway. 159
Now the effective local potential (6)
d* , e"2O-(ch2Q* + p) E) =-r-fiu* Y (cW f 2=- 2 P £ J J Z» where £ is given in (13) and (8). and
:v(-«t ;• (17)
Before proceeding with the discussion of self-consistent results, let us briefly discuss (16), which is the central result of our theory. (It can be generalized to any tunneling geometry and potential V(z) by changing the coupling functions I\ and SP frequencies « ). Hie character of the dynamical corrections is shown explicitly in the denominator in (16). the first tens defines the dynamics of the screening machaniaa. The sacoad tarn can be written in terms of the local energy dependent decay constant x(z), given by (13), and is thus related to the local electron decay. The third term AE^(Q) contains the influence of parallel electron motion (v, ) and the recoil term /5-8/. It is obvious that « here does not enter as the "timeline velocity": in this situation * plays a role of the decay
To emphasize this point we should rgsmmfair that in the came of an electron moving freely with the normal velocity vx the corresponding term was
va-Q * iQvx (IS) i.e. the normal velocity was also coupled to the SP Instead, * in (16) should be coneiderad as a spatial parameter, leading to the time decay oonatant T^ S 4k/c(z). In this way the dynaucs in (16) enters via the ratio n = T./T. of the two characteristic times, T and r , what* r^ depend! an the local decay of the tunneling electron.
He specifically notice that thia n differs from the ratio
T#/rt, usually tafcan /I.9,15/ to describe the 160
corrections to the image potential in tunneling.
IV. Xonaelfconaistenfr result* / first iteration
In the following we shall study the k = 0 case, which is simpler, but contains all essential physic*. In this cue the
potential (16) depends on two ratios: n(z) = c(z\/fum and < = E,,/hu , and scales with e*/4a. Here a *
ft3 E = —; (191 " 2m*(2m)2
is a characteristic energy of electrons in a barrier of width 2a.
It is instructive to discuss the dynamical tmafc potential (16) by treating n and K as parameter*, and thus ignoring the fact that r?(z) has to be determined self consistently. This in fact corresponds to the first iteration of the potential (16), with c (z) = Vo-E. Fig.4 shows how the image potential reduces a constant tunneling barrier of height VQ. For n - 0 one recovers the classical result, modified at the interfaces by the introduction of a short wavelength cutoff which prevents divergencies /4/.With increasing n, i.e. the ratio of screening vs. decay times, the screening becomes less effective, the potential is reduced and the barrier become! less penetrable.
The recoil tern £ is usually very saall, except for very thin barriers and low SP frequencies, and in particular saall effective Masses. In any case, Fig.4, shows that the dominant reduction coaes froa the dynamics, described by n. Fig.5 shows how the top of the barrier increases with increasing Q and. If. However, for low i? this is the case uhere it becosee essential to include selfconaistency. 161
V' Self-consistent results
Obviously 77(z) is not a free parameter but chances inside the barrier, and should be treated selfconsistently. Already in the first iteration the electron builds up the image potential, so the energy E locally appears above the potential barrier, as the turning points tz move inwards from z - *a.
In Fig,6 we show both the self consistent and first iteration
solutions for the top of the barrier (2=0) as functions of e0 = = VQ- E, for two SP frequencies.
In Figs. 7a and 7b we show a series of potentials for two SP frequencies and different electronenergies, evaluated selfcon- sistently. Selfconsistency always lowers the barrier in comparison with the nonselfconsistent results. Fig.4.
Instead of making a local approximation (15). Jonson /9/ proceeded to calculate the self-energy (10), integrating over z" and using constant «. For a charge tunneling out of a metal into an insulator with the dielectric constant cQ he finds the modification of the classical image potential at distance z froa the surface in the fora
V(z,E) = f- I dQ — <20> 0 « *• or, after integration:
a f 2«'« V(E,E where the function g is 162
g(x) = x e~* Ei(x) (21)
For larce x g(x) -* X, as expected, reducing to the classical result, but for finite x it shows two strange features. For 0l, and the potential is in fact enhanced with respect to its classical value.
Neglect of surface sensitive corrections in Jonson's approach cannot be responsible for this. We have generalized the nonlocal theory to the tunneling in a symmetric M-I-M structure, and the same unphysical behaviour can be observed even in the middle of the barrier where the surface cannot influence the T T results, now as a function of v = «*/2a» *>t, which is the #/ t ratio. For n?-i the barrier is not lowered due to screening as
one would expect, but increases above its VQ value.
All these features do not agree with our intuitive picture of dynamical screening of tunneling electrons, and one suspect that some uncontrolled approximation produced these unphysical results.
VII. Tunneling rates
The shape of the potential barrier - its width and height, determines the conductance of the barrier, and this i*s one of the reasons for this study (as well as the justification for the use of VKB approximation). The tunneling rate in MJB is
T(E) a e"2B <22) where
B(E) = | J« where « is given by (13), and ±zQ arc the turning points. In the 163
absence of the screening the exponent in (22) would have the fora
BQ(E) =
2 i/z where *>= i2m*lVg-E)/h ) . Both B(E), where the potential W in
*(z) was evaluated self consistently! and B0(E) are shown in Fig.8
for the barrier width 2a = 200 aQ, and for two SP frequencies. Me also show the tunneling rate B (E) calculated for the classical (i.e. static) image potential.
At this point we have to emphasize that WKB theory applies to the tunneling through a static barrier, and it would be easy to find arguments against this treatment for the case of a fluctuating potential. In Fig-2b we illustrate it by showing relative contributions of various Q components of the symmetric/antisymmetric surface plasaons to the iaage potential (albeit in the static case) for electrons at three different positions in the barrier. It can be seen that the dominant contributions to the barrier cose froa SP nodes which oscillate with frequencies often comparable or even larger than other characteristic frequencies in the problem. The use of the static barrier provides only an estimate of the dynamical effects in electron tunneling, and for further work one probably has to adopt a store sophisticated path integral/instanton approach, as in ref ./15/. On the other hand, in the latter approach one cannot study explicit fora of the iaage potentials in the barrier, as presented in this paper.
VIII. Conclusions
In conclusion, in this work we have calculated analytic expressions for the dynamical iaaae potentials for tunneling electrons in the MSB approxiaation, taking into account SP dispersion and coupling in a particular aetal-insulator-aetal geometry. We have shown how finite SP frequencies, wbsa comparable to other characteristic energies, prevent formation of 164
a full (static) img* potential, and thus scdify the barrier conductance.
This work Has partially supported by the US-Yugoslav Joint Board on Scientific and Technological Cooperation. Grant PS 85KNIST).
References
1. P.Gueret, E.Marclay, and H.Meier. Appl.Phys.Lett. 52, 1617 (1988); 2. J.G.SinBons, J.Appl.Fhys. 31, 2581 (1963); Z.A.MeinberS and A.Hartstein, Solid State Coaaun. 22, 179 (1976); A.Hartstein, Z.A.Keinber*. and D.J.OiHaria, Ftoys.Rev.B 25» 7174 (1982) 3. A.A.Lucas. Phys.Rev.g£, 2939 (1971) ;A.A.Lucas and M.Sunjw*. Phys.Letters 3S&, 413 (1972), J.Vac.Sci.Technol. 2* ?25 (1972) and Surface Sci. 3J. 439 (1972); R.H.Ritchie, Physics Letters 38A.. 189 (1972). 4. Z.Lenac and M.Sunjid, N.Ciawnto 23, 681 (1976) 5. M.Sunjid, G.Toulouse, and A.A.Lucas. Solid State Coaaun. U. 1629 (1972) 6. R.Ray and G.O.Hahan, Phys.Letters 42o, 301 (1972); see also G.D.Mahan, in Collective Properti,— of Hursical Systeas. Nobel Syaposiui 24. ed.by B.I.Lundqvist and S.Lundqyist (Acadeauc. New York. 1974), p.164. 7. P.M.Echenique, R.H.Ritchie, N.Barbetan. and John Inkaon. Phys.Rev. S22, 6486 {1981); J.R.Manson and R.H.Ritchie, Phys .Rev. BJi, 4867 (1981) 8. F.Sols and H.H.Ritchie, Solid State Cc—w. §2, 245 (1987) and Phys.Rev. Bj£, 9314 (1987) 9. M. Jonson. Solid St*te Cbwun. 22, 743 (1980) 10. Ashok Puri and M.L.Scteich, Phys.Rev. BZt, 1781 (!St3) 165
11. R.A.Your*. Solid State Consul. 4£, 263 (1983) 12. J.W.Wu and G.D.Mahan, Phys.fiev. Jgg, 4839 (1983) 13. M.C.Payne and J.C.Inkaon, Surface Sci. 152. 4C5 (1985) 14. P.M.Echenique, A.Gras-Marti, J.R.Manson and R.H.Ritchie, Phys.Bev. §22, 7357 (1987) !5. B.V.J.Persson and A.Baratoff, Phya.Rev. B2g, 9616 (1988) 16. M.BUttiker and R.Landauer, Phya.Rev.Lett. &, 1739 (1982); Phys.Scr. 12, 429 (198S); IBM J.Res.Develop. 22. 451 (1986). 17. E.H.Hauge, J.P.Falck and T.A.Fjeldly, Phys.Rev. 82fi, 4203 (1987); C.R.Leavena and G.C.Aers, Phya.Rev. §29, 1202 (1989) 18. M.Sunjitf and A.A.Lucas, Fhys.Rev. B2, 719 <1971) 19. See also M.Sunjid and L.Muruiiti, subaitted for publication 166
Fisury ctcticra
Fig.l (a) Ceoaetry of the aetal-insulator-aetal systaa.
(b) Coupling functions Tt (Q,z) for the syaaetric Fig.2 (a) Frequencies of syaattric <»t) and antisyaaetric («_ ) surface/interface plasaom of the H-I-M systaa. (b) Relative contributions of syaastric («. ) and antisia- eetric (w_) surface plasaons to the (static) iaage potential, for several positions of the electron in the barrier: z/a = 0, 0.5, 0.8 and 1. Fig. 3 (a) Second order nonlocal electron self-energy diagrea, and (b) its local {or seaiclasaical) approxiaation. Fig.4 Dynamical iaage potent .als in the barrier of width 2a =
= 200 a0, in the first iteration. The SP energies are
(a) fc>t= 0.1 By and (b> fiu^ = 0.01 By, a = 0.07 a, and •lactron energies are given on the figure. Fig. 5 Top of the tunneling barrier, in the first iteration,
as a function of vo= eo/Kus, for several values of? (denoted on the figure). W is given in units of ea/4e. Fig.6 Top of the tunneling barrier - selfconsistent cal-
culations. The barrier is 2a = 200 atf wide, a = 0.07a, and the surface plasson energies are (a) *«,= 0.01 By and (b) fii^s 0.1 By. Dotted lines denote the first iteration. Fig.7 Selfconaistently *^Fig.8 Tunneling rates for the barrier with 2as 200 aQ, a*= 0.07 a, ~ir».i«»»i with no iaage potential (lo). classical isage potential (B^) and the dynaaical
potential (B) for tuo SP frequencies: f*>m= 0.1 Rr
0.01 By. The ratea mn given in unita (2a/a0 )/a /a . x67
M
-*• z
(*) 168
^. 2 170
Q. Z 2'
-*.»
0.0 169
-o.io -10 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0-» 1.0
(a)
0.00
-0.09 -
-0.10 -10 -O.I -O.« -0.4 -0.2 0.0 02 0.4 0.* 04 10 t/a 171
0.000
-0.001
-0.002'
-0.003
-0.004
-0.005
-0.MK
-040? 0.00 0.02 OM 0M 0.M 172
-0.00
-0.02
-0.04 I
-0.06 A
-0.08 A
-0.10 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.+ 0.6 0.8 1.0
(«•)
-0.00
-0.02 .i
-0.04 4
-0.06 1
-0.08 H
-0.10 I i. i i i i ' ' _' -' , - -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 10 I/O 173
0.35
0.30
0.2*
0.20
0.19
0.10
COS -
0.00 0.1*
0.16
0.1*
0.12
0.10
0.M
0.M
0.04 •
0.02 -
0.00 0.000 0.004 0.00* 0.012 0.01*
<="•)•« 175
Photon Scanning Tunneling Microscopy
R.J. Warmack. R.C. Reddick. and T.L FerreH Oak Ridge National Laboratory P.O. Box 2006, Oak Ridge, Tennessee 37831
The PSTM is shown to be the photon analogue of the electron STM. In the PSTM, externally-supplied plane waves undergo total internal reflection at a dielectric interface. This creates an exponentially-decaying evanescent field adjacent to the interface. Photons can be made to tunnel to a nearby optically conducting fiber optic tip. The transmitted light signal is strongly dependent upon the tip to interface distance and can be regulated by feedback control to stabilize this distance. Perturbations caused by irregularities in the interface can therefore be imaged by scanning the tip laterally in the same way as STM. Subwavelength lateral and vertical resolution is demonstrated. Lateral resolution is governed by the decay of the evanescent field and the probe shape. Vertical resolution is governed by electrical and mechanical noise. Controllable parameters include the angle of incidence, wavelength, polarization, index of refraction, and probe shape. Direct imaging of standing waves caused by the interference of two coherent surface waves at the interface wiN also be shown. JCamttnn rfobe I
FORCE.
(0) w
MAGNETIC CAPACITANCE TON )L M. Si Photon Scanning Tunneling Microscope
Photo- multiplier Tube
Piezoelectric Computer Control Translator and Image Processing XYZ Motion and Fiber Optic Feedback Tip
Prism Fiber Optic Tip
Evanescent* Totally Internally ... Field Reflected light . Sample 178
t, n
- Hi - hi.
«A\J
Uw NP048703.C01
20 Photomultiplier Current vs. Increasing Distance from Surface. ^ 15
10 v
200 400 600 800 1000 AZ(nm) 180
, k ' 1?
W
n* 1
% iE* 181
Direct Observation of Interference Fringes by PSTM An interference pattern caused by the intersection of two laser beams is not ordinarily observable since the fringe spacing is so small. The photon scanning tunneling microscope (PSTM) is used here to image the interference pattern at sub-wavelength resolution. The fringe spacing displayed is 307 nm or about half the wavelength of the light used. These images ciso demonstrate that the PSTM rcsoJuiion is an even smaller fraction of a wavelength.
ORNL Submicron Physics Croup 182
Holographic Grating Profile (PSTM)
c
-120
PSTM ADVANTAGES
Sub-Wavelength Resolution Three-Dimensional Mapping Images Insulators Non-contacting/Nondestructive Samples under Ambient Conditions Wavelength Selection: Spectroscopy 183
Extended transfer Hamiltonian method for resonant tunneling
S.P. Apcll Department of Theoretical Physics Umea University, S-90187 Umea, Sweden
E. Albertson Institute of Theoretical Physics Chalmers University of Technology S-41296, Goteborg, Sweden
and
D.R. Penn Radiation Physics Group National Institute of Standards and Technology Gaithersburg, MD-20899, USA August 24, 1989 184
Abstract A scheme originally due to Heitler is utilized in extending the transfer Hamil- tonian description to resonant tunneling, for calculating transition probabil- ities and general frequency characteristics of tunneling coupled systems. We give examples for a single barrier in the context of the Scanning Tunneling Microscope and for a double barrier in connection with an irradiated quan- tum well. In this way we can very conveniently account for the saturation of the contact resistance in the STM and give a simple explanation for the high-frequency response in an irradiated double junction. 185
Introduction One of the most used schemes for calculating tunneling probabilities is based on the so called transfer Hamiltonian formalism [1]. Recently this formalism has been discussed from the point of view of extending it to situations where there are real intermediate states, like in double barriers [2,3]. This prob- lem is of current interest in the resonant tunneling through double barrier structures, with or without applied electromagnetic fields, and in the treat- ment of the Scanning Tunneling Microscope (STM) [4]. In the present paper we wiil give a description of theses processes within an extended transfer Hamiltonian formalism which is based on a scheme originally developed by Heitler [5], many years ago. We illustrate our formalism with two systems of interest in device applications: the single barrier and the double barrier shown schematically in Figure 1. Basically the single barrier has two states, left (L) and right (R), and the double barrier has also an intermediate state (I). We will treat situations having both discrete energy levels as well as a continuous range of energies since they will exhibit very different response characteristics. Let us first discuss a Fourier-space representation of transition probabili- ties in a quantum mechanical system to get the spirit of Heitlers treatment. The virtue of going over to a Fourier space representation of the physics is that we can utilize schemes developed for calculating Green's functions to infinite order in the coupling between subsystems. Those can be different electronic parts of a system but they can also be the interaction with ex- ternal radiation or an applied potential. Furthermore working in Fourier space characteristic response frequencies, such as Rabi oscillations and their strength, come out directly. We start from the time-dependent Schrodinger equation for the state vector
= HT* (1)
wnere Hy is the total Hamiltonian. Let /fr=Ho+H,-Bt Ho describes the un- perturbed system and H,nt the interaction. We denote an eigenstate of Ho by #n for short. The exact solution of Eq.(l) can then be expanded
(2) 186
2 6n | is the probability for the system to be in the state n at time t, and
3 Dn(t) = jt\K\ (3) defines the transition probability per unit time. Notice that Um«^«, 0n(t) gives the standard Golden Rule formula for the system. We have a general constraint on our system in that: £A.(O = o (4) n since the normalization of the b's is conserved; the particle we are describing has to be somewhere. Making a Fourier expansion of bn(t):
we can finally express Dn(u>) as the following convolution
+ u/) (6)
In what follows we will be mainly concerned with D/j(o>) which we will treat as a response function of our system. We will study a system which is in equilibrium at time t=0~ in which we inject an electron into a particular state o at time t=0+. The injection procedure can be included in Eq.(4) by letting £„ Dn(t) = 6{t). We therefore have a sum-rule £„ Dn(u>) — 1 in Fourier space. From equation (3) we find the interesting static limit:
a ZJn(« = O) = /*/>»(<)=! **(*>) | (7) using the fact that bn(0~) = 0. The smalt energy limit of our response function Dn(u) is thus a measure of the stationary occupancy of the state n. In the next section we first introduce Heitlen scheme for calculating the amplitudes (bn) and we apply the formalism to a single barrier, illustrating the Rabi frequency concept and the saturation of the tunneling for strong coupling. Then we go on to study the double barrier both with and without an external electromagnetic field. We end with a small commentary to the general response properties of real irradiated double barrier devices based on the physical principles of single and double barriers we have illustrated earlier. 187
Single barrier In the previous section we set up the formalism for calculating the occupation amplitudes bn(w) and here we will use the single barrier a* an illustrative example. From Heitler [5] we can write down the followingrepresentatio n for b,(t):
6w(t) = y- f~ dEGn^Eyl**-1*/* (8) £X J-co yielding (c.f. Eq.(5)) 6.(u;)«iGW + £.) W where the Green's function G*» can be written as:
Gno(u>) = CU~)G~(«K(w - En) (10) with and the " injection" state Green's function (?„(«)« (a,-£.-£M)-1 (12) with its self-energy
expressed in terms of a renormalized coupling strength
) - ff^ + £ ffw»{(u» - £,)£/«.(«) (14)
HmB is the matrix element of the interaction part of the Hamiltonian be- tween the states m and n and can be calculated e.g. within the trans- fer Hamiltonian formalism. Now specifying to a two-level system in the form of a single barrier having only one discrete level on either side there is only one coupling matrix element Hmll«Hj|» s H. Therefore UJW«H. S(w) =H3/("+ic - £A) with e * 0* and we get: 188
defining
wl>2 = -ERO - -c ± -WH (16) where 2 WH = v1^ + 4i/ ] (17) is the so called Rabi frequency [6]; the characteristic frequency for oscillating back and forth between the states o and R. Inserting bn(w) into Eq.{6), picking up the two poles in the upper half plane, we get:
(18)
Dfl vanishes with the coupling Hz as it should. D/}(0)=2^j- and it reduces of course to ^ when En — EQ. Comparing to dielectric functions the structure of DR(U), Eq.(lS), is char- acteristic of a bound electron system having a transition energy w^. Going into real time space we see that during a time proportional to w^1 the elec- tron has left the o-state and come back. Clearly u>^1 can be viewed as a typical time of " tunneling". Furthermore letting t—• oo
DR{t^ oo) - j- | H \*6(En.) (19)
to lowest order in H2, i.e. we retrieve the Golden Rule result for the transition rate between the states o and R. In a real situation we do not have the very sharp energy levels of the example above, instead we have a continuous distribution of states to the left and right of the barrier which we can characterize with their density of states PL and pn respectively which we take to be the same in what follows; PL — PR — P- Going through the same steps leading to Eq.(18) and letting all H/j=H we now find that the Rabi-oscillations are masked because the self-energy T.(u) acquires a finite imaginary part F. We also find that Dfl(O), measuring the occupancy of the right hand side state R after very long time, is now:
£^T5 (20) where T = rpHTI (21) 189
and
v <22> is a new renormalized coupling strength (tunneling probability) instead of H. To arrive at Eq.(22) we have approximated the energy-integrated (-function defined in Eq.(ll) by its imaginary part, to give the main physics of the problem. For thin barriers, where irpH Double barrier As our second example we consider the double barrier as shown in Figure lb. A crucial feature of the double barrier is the broadening of the intermediate state (I) due to its coupling with the continuum at either side. This broad- ening makes it qualitatively different from an atomic three-level system with well-denned Rabi frequencies which we will see in what follows. As before we focus on i>n(u>):
bR(u) = iGRo(u> + ER) (23) where
GUu + ER) = Utoiu + ER)GM((J + ER)t («) (24) and URO{U> + ER) = HRIt(u> + EmWtciu + ER) (25) with 2 1(" - ER)UIo + £ | Hu \ e(u; - EL)UIo (26) H L using
ULO{") = Hu&u - Ei)Uj0 (27) in Eq. (26), c.f. Eq.(14). The summations over L and R represent the coupling of I to the contin- uum we have to the left and to the right (EL —* fdELpl) and E/j = E/-Ej in Eq.(25) and what follows. Neglecting the shift of the energy levels we will replace the summations with a constant life-time for the intermediate state: 2 T/ = T1R + Tu with r/t = xfdEiPi \ Hu \ . This gives finally for U/(((w):
(28)
As usual letting all H/j be the same (=H) and inserting Eq.(28) into Eqs.(23- 27) we get when E0=E/i:
*«(«) = / ^ • w ^ M T (29) (w + tc)(w w)(w w) where \ 2 2 (30) 191
Notice that comparing to Eq.(16) we can identify in Eq. (30) an effective energy Ej — iT/ for the intermediate state. For T/ <\ E{o |, b«(o>) shows that we have a high frequency mode u>\ ~ Ej0 when E/o > 0 or a broad 2 low-frequency structure around Re[u/2] a H IEoj when E/o < 0. From Eq.(29) we now get for
; - ) + __- :_ ] (3i)
When 0/ = 0 we have the limit:
DM0) =1 — I2 (32) WW2 This is the same as saying that the long time response of the junction will yield a non-vanishing electron density to the right. This is clearly what one associates with a tunneling process through a junction where .depending on the coupling a certain probability exists for finding the particle to the right 2 of the barrier. For E/o > H and using T/ « 2*pH we get:
In Figures 2a-c we show DA(W) calculated from Eq. (31) in three different situations: E/ < £„, E; ~ Eo and E/ > Eo for Tf/ff si. We see thai Dfl(w) vanishes for large frequencies, as it should since this corresponds to short times and there is not yet any probability amplitude to the right then. In general DR(U) exhibits a broadened oscillator behaviour around energies ~j 7 I, 7 = EiojTi. Furthermore DR(UJ) quickly vanishes for increasing J 7 | as a comparison between Figures 2a) and c) together with b) shows. Notice also that there is of course very little left of the Rabi frequency concept when we have continuous energy states both to the left and right of the barrier as well as a broadened state inside the barrier. In the following section we will extend our treatment of the double barrier to include an external potential. 192
Irradiated double barrier We have illustrated the use of a Fourier transform scheme to appreciate the characteristic energy-scales involved in a single and a double barrier respec- tively. To make contact with real devices we have furthermore to include a static bias and/or an applied microwave field. In this section we will give a brief outline of the main results for the case of an external electromagnetic field of the form E0cosut [10]. First we make an important observation. In semiconductors with a typ- ical doping, the energy scale is in the 10-100 meV range. This corresponds to very long wavelengths or penetration depths, being much larger than the tiny extension of the well region. It is therefore possible to neglect the part of the absorption which has to do with the well. This picture is in contrast to traditional approaches in this area [2,11-19] which tend to stress the im- portance of the coupling of the well with the incoming light. In our view we have absorption in the bulk electrodes outside of the well, creating electron- hole pairs which in a bias will be swept through the system and collected in the form of an I-V characteristic. The only influence is in giving an energy- selective constriction to electron transport. We thus conclude immediately that any intrinsic cut-off of such a system is determined by the optical prop- erties of the electrodes, which in reality often means the optical properties of bulk n-doped GaAs. Within the transfer Hamiltonian scheme we have to do the same steps as in the previous section, but now with an added perturbation in terms of an incoming electromagnetic field £Acosu>t. We call the coupling term in the Hamiltonian for W which we take at a constant for simplicity. Based on the arguments given above it only couples states on the left hand side of the junction with each other or on the right hand side with each other. After some straightforward algebraic manipulations we can write down a Golden Rule formula going from the state Eo to E* [10]:
3 2 oo) = y[| /(/?„) | 6(E,u)+ \ g{ER) \ 6{E,u ± M] (34) having two contributions being one the static part:
/(£„)= *** <35) ERI + *!/ 193
corresponding to two tunneling events enhanced close to resonance (E/i/=0) and the other one being the dynamical part:
corresponding to light absorption and two tunneling events or vice versa. In Eq. (36) we have set pi =s pn = p lot simplicity and again used the imaginary part of ( to exhibit the general behaviour. In what follows we neglect the interference between the two terms in g(E/i) and use the fact that for the situation we are investigating it is a good approximation to replace the energy-denominators containing the intermediate state with delta functions, since F/ is smaller than the other energy-scales involved [11]. This gives:
2 | g(ER) | * -Jr/WW&t,) + S(EM ± M] (37) 4 and ^ (38)
yielding a general ratio between the dynamical and the static parts (i 1 for ordinary electromagnetic fields. In device physics it is appropriate to calculate the current I which flows through the junction. Since I is an energy-integration over DR{1 —• oo) the first part of Eq.(34) gives the static I-V characteristic of the junction [20], which we call IO(V). It is then easy to see that the dynamic part of the current can formally be written as:
2 I4{V,w) at (xpW) [ro(V + u;) + I0(V - «)] (39)
in terms of the static I-V characteristic Ie(V). Not surprisingly this is * result which is very similar to the general treatment for irradiated Josephson junctions by Tien and Gordon [21]. The pref&ctor in Eq.(39) contains W3 which is a measure of the optical aborption in the electrodes. If we had treated this part more carefully if would of course contain the absorption frequency characteristics of the bulk electrodes. This has a characteristic cut-off frequency being the plasma fre- quency of the doping electron gas or for that matter roughly its Fermi level [22]. We therefore anticipate an upper cut-off in the response of an irradiated 194
double barrier around 40 meV (or 10 THz) based on a doping of This result therefore confirms the experimental findings in [11] of the possi- bility of observing a THz response from an irradiated double barrier. Notice also that since F/ is typically 3-5 meV it plays almost no role whatsoever for the validity of our conclusions because it is much less than 50 meV. At the end we should finally acknowledge the situation that whereas there is a response time-scale giving a cut-off around 10 THz for the isolated de- vice most experiments on double junctions are in oscillator configurations which give a cut-off in the GHz range [23-26] determined primarily by the capacitance and the series resistance of the device and not directly by its intrinsic response properties. 195
Acknowledgement This project was supported by a grant from the Swedish Natural Science Research Council. 196
References 1. See e.g. C.B. Duke, Solid State Physics, Suppl. 10, Academic Ptess N.Y., 1969. 2. M.C. Payne, J. Phyt. C19,1145 (1986). 3. L. Brey, G. Platero and C. Tejedor, Phy». Rev. B38,10507 (1988). 4. T.E. Feuchtwang and P.H. Cutler, Physic* Script* 35,132 (1987). 5. W. Heitier, The Quantum Theory of Radiation, Ch.IV:16, Dover, N.Y. 3" edition (1984). 6. See e.g. P. Stehle, Physics Reports 156, 67 (1987). 7. J.K. Gimzewski and R. Moller, Phys. Rev. B3«, 1284 (1987). 8. J. Ferrer, A. Martin-Rodero and F. Floras, Phys. Rev. B38, 10113 (1988). 9. L.V. Keldysh, Sov. Phys. JETP 20,1018 (1965). 10. D.R. Penn and S.P. Apell, to be published. 11. T.C.L.G. Sollner, W.D. Goodhue, P.E. Tannenwaid, CD. Parker and D.D. Peck, Appl. Phys. Lett. 43, 588 (1983). 12. K.W.H. Stevens, J. Phys. C17, 5735 (1984). 13. S. Luryi, Appl. Phys. Lett. 47, 490 (1985). 14. J. Kundrotas and A. Dargys, Phys. Stat. Sol. bl34, 267 (1986). 15. W. Frensley, Phys. Rev. B36,1570 (1987). 16. J.F. Young, B.M. Wood, G.C. Aers, R.L.S. Devine, H.C. Liu, D. Land- heer, M. Buchanan, A.J. SpringThorpe and P. Mandeville, Phys. Rev. Lett. 00, 2085 (1988). 17. D. Sokolovski, Phys. Rev. B37, 4201 (1988) and Phys. Lett. A132, 381 (1988). 197
18. Y. Nomura, S. Nut, S. Maruno, M. Gotoda, Y. Morithita and H. Ogata, Superlattices and Microstructures 8, 73 (1989). 19. M.Jonson, Phys. Rev. B30, 5924 (1989). 20. See e.g. the presentation in H.C. Liu and G.C. Aers, Solid State Com- mun. 67, 1131 (1988). 21. P.K. Tien and J.P. Gordon, Phys. Rev. 129, 647 (1963). 22. W. Szuszkiewicz, K. Karpierz and Vu Hai Son, Physica Scripta 37, 836 (1988). 23. T.C.L.G. Sollner, P.E. Tannenwald, D.D. Peck and W.D. Goodhue, Appl. Phys. Lett. 45, 1319 (1984) 24. T.C.L.G. Sollner, E.R. Brown, W.D. Goodhue and H.Q. Le, Appl. Phys. Lett. 50, 332 (1987). 25. J.F. Whitaker, G.A. Mourou, T.C.L.G. Sollner and W..D. Goodhue, Appl. Phys. Lett. 53, 385 (1988). 26. E.R. Brown, W.D. Goodhue and T.C.L.G. Sollner, J. Appl. Phys. 64, 1519 (1988). 198
Figure Captions 1. a) Single barrier with states L,R and o. o is the injection state. b) Double barrier with states L,R,I and o. I is the intermediate state with broadening F/. 2. Real (full line) and imaginary (dashed line) parts of the general trans- mission function DR(UJ) is plotted for the double barrier as a function of detuning 7 = (Ej — E0)fTi. w is measured in units of H. The cal- culation is done for Fr/H = 1.0 and a) 7 = -10, b) 7 = 1.1 and c) 7 = 10. Notice that a) and c) are magnified a factor of 1000 as compared tob). 199
FIGURE 1
b) H 200
Figure 2
D x 103 0
DxlO3
-10 201
LASER FIELD EFFECTS ON THE INTERACTION OF CHARGED PARTICLES WITH DEGENERATE AND NON-DEGENERATE PLASMAS
N.R. ARISTA*. R.M.O. GALVAO**. AND L.GM. MIRANDA**
Universitat d' Vacant, Depanamento de Fisica Aplkada. E-03080 Alicante. Espafia
A general description of the effects of a strong laser field on the inelastic inieractkm between charged panicles and a plasma is presented, both for classical (dilute) or quantum mechanical (dense) plasmas. The dynamical response of the medium is recalculated, including the effects of the laser field, using a Dawson-Obennan transformation for classical systems. and the RPA formulation for quantum plasmas. We find that the energy exchange and the particle scattering rate are modified by mold- photon processes, and become anisotropic with respect to the direction of panicle motion relative to the direction of laser polarization. The RPA formalism describes the excitation of ptosmons and electron-hole pain, with simultaneous emission or absorption of photons. We calculate the contribution of these processes to the energy exchange and »the meao- freepath of the panicle, in the range of a typical low density plasma, and in me range of solid-state electron densities. New effects due to the laser field become particularly important when the laser frequency becomes close to the plasma frequency. In this case. pUsmon excita- tions can be produced below the normal velocity threshold through photon-assisted processes. This gives place to an anomalous low-velocity behaviour of the stopping power.
* Centro Atomico Bariloche, Division Colisiones Atomicas. 8400 Barilocbe, Argentina **Insntuto de Pesquisas Espaciais (INPE), 12201 Sao Jose dot Campos. 5.P.. Brazil 202
VI K I •M 1 * * -w 4 - 8 i
u i x
H W 4: a
H Ui
o 41 I i 11 •N I
5
I o ^ i s J
2 w i 3 « 1 * i rx i 1 » 1 i i; ii j F X- CLASSICAL 3. Fsurier
pUt»* l',3,-t) (cU»iu( J.siaV. function)
^e•. j>0 = Z« 6(f-* t)
<«<» Vy :
Source :
s
R«..tt:
ire 204
4 s ?• £ V) It IM II H at | -» 9 •+ IE -T III )l It (3 2 5 X "3" 2 -I OJ n
V*)
ut
u 1b
•HI
it 205
eV
VOAT 206
5 Jta,
3. wf ST u« 1i j- /— '*• 1 «•» j*
jt t *- *+* n •A
"i** ) I -»• y _j
•» So k
' >- 6
1W
1 u a. O -2 (J o J
II E t'- KJ1 I
o s Ul x I 4 < I ? O. ct It 207 ^v^
Ml
j
a. 3 f I I» pi <|1
3 c H i
OIIJC f -oJ-» m
1——1 3* X I «^ i>
7 Ul H
5 H
Ul
z
*c
UJ 208
VJ
s 0
«t
a $ He •/ — k
T
i • < \ 1 -|K' in 209
I oi c 210
=frT)tscri|>t»oH (classical or a.m.) df p*tidt-bias**- laser inierjctioM processes i» tferwo of
£ Processes : —i«eU«ti'c suttee M.t M pUsMA, / " K
or
las&r
I$I$ lt IK iU ^e. ^ IO - IIOO
3. flugubr ie^evi J«wtcc
ike
^ (on^ if (^ ^ 211
Screening Effects in Nuclear Fusion of Hydrogen Isotopes in Dense Media
- Nestor Arista, Centro Atomico Bariioche, R.A. - Raul Baragiola, Rutgers State University, USA - Alberto Gras-Martft Universitat d'AIacant Spain.
ABSTRACT
Nuclear fusion rate* of isotopic hydrogen nuclei embedded in
dense screening media are calculated. Ue consider the case* of a
uniform degenerate electron gas and the inhomogeneou* electron density
in solids. Ue derive an exact wave function for the screened nuclear
interaction, and an analytical expression for the barriei—penetration factor, in the case of homogeneous screening. Tor qualitative estimates of the screening in solids, we use a Thomas-Fermi description of the electron density.
A cross-over of the fusion rates of the various isotspic pairs analyzed (p-d, d-d, p-t and d-t) is predicted for increasing screening length, velocity or effective temperature of the medium.
The effects of variable screening length and local effective temperatures are considered for electron gases in the range of metallic densities and compared with previous experimental and theoretical studies. 212
Present Scattering Model:
* exact wave function
> repulsive potential
* screening in electron gas
> homogeneous
> inhomogeneous
* TF model
* statistical isotope distribution Screening Effects in Nuclear Fusion of Hydrogen Isotopes Present Scattering Model: in Dense Media
- Nestor Arista, Centra Ato'mico Bariloche, R.A. - Rat' Baragiola, Rutgers State University, USA - Alberto Gras-Marti, Universitat d'Alacant, Spain.
AMTIMCT * exact wave function N3 Nuclear fusion rate* of tutopic hydrogen nuclei embedded in > repulsive potential screening aedta are calculated. We consider the cases of 4 uniform degenerate electron gas and the 1nhomogeneous electron density In solids. Ue derive an exact nave function for the screened nuclear * screening in electron gas Interaction, and an analytical t-priiimn far the barrier-penetration factor, in the cave of homogeneous screening. For qualitativ* > homogeneous efttimate* of the ecreeninQ in ftolld*, w» u»e a Thdmat-Fertti description of the electron dennty. > inhomogeneous A cro*»-Dver of the fusion rates of the various isotopit pai?s »TF model anaJyted (p-d( d-d, p-t and d-t> is predicted for increasing screening length, velocity or effective temperature of the medium. * statisticai isotope distribution The effects of variable screening length and local effective temperature* are considered for electron gases in the range «f Metallic densities and compared with previous experimental aria theoretical studies. Degenerate Electron Gas: Repulsive potential a.u.: e = m - ti - 1 a » 0.529 A 18 v0 - ?.«8.|,> cm /s kT = 27.2 eV ~ 300.000 K
i* relativistic: Corrections to q#: • partial degeneracy: * K,
n« ~* rf: o"e-electron radius. SC
-" 2 0.6 0.3 0.01
metals Jupiter Sun W. dwarf FUSION RATES IN A HOMOGENEOUS ELECTRON GAS * Attenuation Factor: T (
1 = A s*
Hulthen Potential: Strong function of q. ,v Fusion rate / scattering center:
Fusion rate / volume:
* Attenuation Factor: Fusion rate / pair: A" r = Fusion rate = n. A E = CM energy of relative motion
fusion rates depend very strongly on the screening conditions and on the collision energy
Fusion rates for p-d are larger at low screening and low-energy - fusion rales for d-d take over in the opposite limit
- Obviously; screening of nuclear charge by the surrounding electron gas leads to orders of magnitude change In rates. O\ • highest q« values: strongly screened interactions as they would apply, for instance, for bound p-mu or d mu muonic systems studied in muon-catalyzed fusion reactions.
- the large variations of the attenuation factor as a function of the screening parameter are similar to those found by Koonin t Nauenberg, or Van Stolen and Jones' predictions for bound d-molectHes (variable elfectlve mass for tht electrons). •200
i / *atoms / (screening length) Fusion rates versus the relative velocity of the Fusion rates for p-d and d-d reactions in a isotopes, for reacting protons, deuterons, and uniform electron gas, as a function of the
tritons (p,d,t) in an electron gas with'qfl-1.27 23 -3 (corresponding to solid Pd). Light curve: p-d temperature T, for 'VlO cm and for seve- fusion rate without screening. ral values of q...
Jones et al
NJ
-100 to FUSION RATE: FUSION RATES IN NON-HOMOGENEOUS ELECTRON GAS
* Inhomogeneities
> isotopes Local density approximation
> electrons
* Thomas Fermi for confined atomic systems
WS cell r, (~ 2.88 a.u. for Pd) to 00
Isotope • metal atom interaction: • WS cell
V(r) = &¥(£ Average fusion rate / pair:
Tx* f»A fj (a) Radial profiles of: TF potential V, electron den- Velocity-dependent fusion rate (v), for a thermal sity ne, and screening parameter qQ, for a TF model distribution of hydrogen isotopes in the atomic cell of Pd atoms, confined within a Wigner-Seitz cell of Pd, for various local effective temperatures T ,, of radius rQ=2.88 a.u. calculated according to the present model, for p-d (b)Spatial profiles of hydrogen isotopes. and d-d reactions.
~o.l?
N>
i
to1 v/v,h CONCLUSIONS:
T*»U ru.l.n t«.. (in .-»> {ro „ . 21 * analytical model of screening in a scattering for process
r««tion .c
10 * cross-over of fusion rates
N5 N3 O * screening important below ~ 200 eV
dramatic T - dependence
* cold fusion rates very low
* model of interest in other situations ? 221
CONTRIBUTION OF RIPPLON EXCITATION TO INTERACTION
ENERGIES OF PARTICIPATES
• 6+ 1 Rafael Garcia-Molina , Nuria Barberan* and Alberto Cras-Marti
• Departamento de Física. Aplicad*, Facultad de Ciencias, Universidad de Murcia. E-30071 Murcia, Spain. § School of Physics. University of East Anglia. Norwich NR4 7TJ. United Kingdom. t Department de Física Aplicada. Facultat de Ciéncies. Universitat d'Alacant, E-03080 Alacant. Spain.
Abstract Ripplons are quantized surface-tension waves in liquids. We have studied the contribution of virtual ripplon excitation: (0 to the image potential felt by a charge located in front of a liquid surface, and (ii) to the van der Waals attraction between two liquid surfaces. These interactions have been investigated as a function of the distance between the interacting systems and for planar or spherical particulate geometry. The polar or non polar nature of the medium is also considered. The range of validity of the formulae derived is discussed.
t Permanent «ddrast: Departamant* da Estructura y Carntltuyanus <»• •• MtUrla, Faculta* do Física, Unlvareléa* d< Baralefie. Diagonal 6*7. E-OMZt Barcelona. Spain. 222
The image potential felt by a charge placed in front or a liquid surface piays a central role in many physical problems. For interacting systems consisting of neutral particles the van der Waals potential describes the interaction. Interactions between charged or neutral particles and liquids are of current interest in research areas of physics, such as: - condensed matter physics: formation of two-dimensional electron Wigner lattices on liquid surfaces, helium atoms scattering from liquid helium, - atmospheric physics: charge distribution in clouds. - biophysics: charges and neutral atoms in the presence of biological membranes. - aerosol physics: aggregation, coagulation and deposition of liquid drops. therefore it appears interesting to improve our knowledge of the interaction potentials previously mentioned in order to provide a better description of the phenomena where these interactions are basic ingredients. Our aim in this work is to calculate the contribution of virtual ripplon excitation: (i) to the image potential felt by a point charge located in front of a planar or spherical liquid surface (Figs, la and lb), and (ii) to the van der Waals attraction between two liquid surfaces (Fig. lc). By ripplons we mean the quanta of capillary waves at the surface of the liquid, whose dispersion relation for a planar liquid surface is (Landau and Lifshitz 1959] V * *q * Sq3 and that corresponding to a spherical liquid surface of radius R is [Landau and Lifshitz 19591 z m-im+2)» "< '—^ where g is the gravity acceleration, a- is the liquid surface tension and d is the mass density of the liquid. 223
In this work we have considered the liquid deviation from flatness (<; in Figs. 1) as a small perturbation, therefore only terms up to first order in the surface shape oscillations corresponding to the rippions have been retained. In brief, the procedure used in our study is the following. First we have calculated the interaction potential entering in the corresponding problem when a rough surface is considered; from this interaction potential, the Hamiltonian that describes the system is obtained. Secondly the ripplon field is introduced in the resulting expressions according to the second quantization formalism. And finally the self-energy due to the presence of the ripplon field is given by the shift in the zero-point energy of the system. A detailed account of this procedure may be found in Gras-Marti and Ritchie (198S1. Barberan et at. (19891 and Gras-Marti et at. (19891.
Two methods may be used to obtain the interaction potential: the first one consists in solving the Poisson equation with the appropriate boundary conditions at the liquid surface. This method includes the full interaction between a charge and the liquid within a dielectric response model and we call this method the full dielectric calculation. The second one gives the interaction potential by adding the interactions taking place among the interacting particles through the whole liquid volume; this constitute* the pairwise summation method and it is a good approximation for nonpolar liquids: it does not take into account screening effects.
In Figs. 2 and 3 we show the self energy corresponding to a static electron in front of a planar and a spherical liquid surface, respectively, as a function of the charge liquid-surface distance. Two liquids have been considered: helium (nonpolar) and water (polar).
In Fig. 4 we show the stopping power for an electron traveling parallel and outside to a planar liquid surface. Fig. 4a corresponds to helium and Fig. 4b to water.
Finally, in Fig. 5 we show the ratio between the rippton contribution to the interaction energy and the van der Waals potential energy corresponding to the interaction between two scmiinfinite liquid media whose surfaces are separated by an average distance C. A 224
more detailed description of thic research is given in Gras-Marti and Ritchie (19SS). Barberan *t at. (1969) and Cras-Marti et at. (19691.
The following conclusions are derived from this work: In the case of an electron in front of a liquid surface, (i) the ripplon contribution to the interaction energy is . a perturbation to the classical image potential, but for polar liquids and short distances th» applicability of our first order approximation treatment is uncertain; in the case of a small spherical water drop (R • 5 A) the ripplon contribution to the potential energy of a bound electron is about half the binding energy calculated without considering that contribution {Ballester and Antoniewicz 1986). (ii) The full dielectric calculation and the pairwise summation method give similar results for nonpolar liquids in the case of planar surfaces but not for spherical surfaces. This seems to indicate that there are screening effects associated with the spherical geometry which are not included in the pairwise calculation and that these screening effects seem to be far more important than the effect of the permanent dipole moment of polar liquids, (iii) The ripplon contribution to the stopping power is negligible for high electron energies (v » i a.u.i.
In the case of two nonpolar liquids separated by an average distance t. the ripplon contribution to the potential energy seems to be significant for distances (AS a.u.
Acknowledgements
Financial support from the DCICyT (project number PS88-0066). and the program of Acetones Integradas with Germany and Italy is acknowledged. 225
Reference*
• Ballester, J. L. and Antoniewicz, P. R., 1966, J. Chem. Phys. §5, 5204.
Barberan, N., Carcia-Molina, R., and Graa-Marti, A.. 19f9, Pltys. Rev. B 12, 10.
Cras-Marti. A., Barberan, N., and Carcia-Molina, R.. 1989, "Rlpplon contribution to tht Interaction between two liquid swfMces", preprint.
Gras-Marti, A. and Ritchie. R. H.. 1985, Phys. Rev. B 2L 2649.
Landau. L. D. and Lifshitz. E. M.. 1959. Quid Mechanics (Pergamon. Oxford), pp. 238 and 240. 226
(a) rta. 1 vacuum
•k.
diXtctric liquid Jlfx, y)
surface displacenar
(e) vacuum 227
(at aallaa m. a to*
elMalaal laafa M—till
' KT*-
fall 41al««trlc
20 (0 M 100 z(eut)
Ik) wut
^>k elaaalaal iaaaa aa«a««tal KT1 ' «r*
tall olalctrlc —I—Itlaa
20 M N «0 228
«. s la)
—*
iOO *.*.. (ki
»•»
-
^r
(•I 229
(a) Mi« wt*. t
10*
W*
W*
-? 20 40 CO M WO j (M
10* *r*
10*
20 40 CO CO WO 230
|0» , «•"- no. s
I*'
uri.
0 5 10 IS ZO 231
Nuria Barberan1, Manuel Barranco2, Francesca Garcias2, Jesus Navarro3 and Llorenc Serra2 1 Departaaent d'Estructura i Constituent* de la Materia, Univeraitat de Barcelona, E-0S02I Barcelona, Spain. 2 DepartaMnt de rlcica, Universitat de sea Zllea Baleara, E-07071 ?I1M de Mallorca, Spain. 3 Departamnt da Flaica Atoaica, Molecular i Nuclear and iric, Univeraitat de Valencia - CSZC, £-46100 Valencia, Spain.
1. Introduction The increasing interest in the properties of microstructures during the last few years motivated by several problems, like the development of new reduced components in microelectronics or the fundamental mechanism of catalysis in many important chemical reactions, has generated a big amount of related literature [1]. Among the available experimental techniques, the scanning transmission electron microscope (STEM) allows the study of the response properties of targets with high resolution [2]. For small clusters, the increasing surface to volume ratio has suggested that the response would be dominated by surface type excitations. However, recent electron energy loss spectroscopy experiments on metallic spheres [3] have shown that, contrarily to what was expected, significant bulk scattering occurs even on less than 50 A diameter spheres. Bulk collective excitations on metallic spheres have been theoretically described using a non-local dielectric function [4], a hydrodynamical model (5,6] and a selfconsistent time-dependent local density approximation [7]. Here, we carry out a Random-Phase Approximation (RPA) sum rules (SR) calculation of the response (strength) function to multipole q-dependent operators of the type JL (
electron cloud of the unperturbed cluster is described either in the Kohn-Sham (KS) or an Improved Thomas-Fermi-Weizsacker (ITFW) plus a local density approximation (Dirac+Wigner) for the exchange and correlation effects.
2. RPX SUB rulas Sum rules m^ are moments of the strength function
S(E) - 2 6(E - En) f|2 (1) n where the sum (integral in the case of continuum spectrum) extends over all the excited states of the system. Q is the external field acting on the system; En, |n> and |$> are the excitation energies, the excited states and the ground state (gs) of the system, respectively. By definition,
k mk - J EE* S(E) dE - Li En* !i2 (2) n The mean energy K and the variance a2 read * * mi/inn m O* m2/mQ — (mj/m(j)* (3) Among these moments, the ones with k—1, 1 and 3 play an important role in the application of SR to the study of collective resonance states of the system. First of all, they can be obtained with RPA precision from Kohn-Sham calculations, essentially involving gs expectation values. Secondly, they can be used to estimate C and 2 1/2 ! £ B S E3 2 2 2 a ^ (E3 -Ei )^ (4) In this calculation we have used the following electron energy density functional:
E{n(r)) - J t(n) dr - 3/4 (3/x)1'3 J (n(r>)4/3 dr
- a J n(r)/
r*(r) n - 2^ \+i\2 and T(n) -2* l^^l2 are the electron and 233
kinetic energy densities, ra(n) - [3/4acn(r) ji/3 iS the local radius per electron, Vj(r) is the jellium Coulomb potential, R the radius of the jellium sphere, N the number of monovalent atoms and a-0.44 and b-7.8. We have used atomic units (au).
3. Results A. rinit* Spheres We have applied the SR method to Na spheres for which we have taken ra«4 au. Fig. 1 shows for Nag2 and L«l the contribution to m3(RPA) of the different terms entering its definition, that is, the kinetic term (T), exchange (ex) and correlation (cor) tenu and the direct (e-e) and jellium electron (j-e) Coulomb tens. Each contribution has been normalized dividing it by the total 013. Curve labeled C is the sum of the Coulomb e-e and j-e terms. This figure shows that for small q-values, m3 is dominated by the Couloab terms, and by the kinetic term for large q-values. Both contributions equalize at a value of q close to q,. - «fc>/vF • 0.45 1/2 au where OJp - (4«n) - 0.217 au is the plasma frequency and vr - 234
(3JC2 n)1'3 - 0.48 au is the Fermi velocity. For larger L-values, the relative importance or the kinetic term increases for small q-values, in agreement with the results obtained for surface excitations in ref. 9. The local Coulomb exchange and correlation contributions are negative, the later contribution being always negligible.
This cut-off qc is a lower bound estimate of the Landau damping onset momentum. The transition from a collective to a single electron-hole excitation is related to the increasing contribution of the kinetic energy to m3 for large values of q and independently, for large values of L.
051
0.1
In Fig. 2 we represent the ITFW Ex and E3 energies corresponding to Na92 L«0. One can observe that the difference between El and E3 increases when q increases; the electron-hole (non collective) contribution becomes the main excitation mechanism for values of q £ qc. Apart from L-0, all the other modes (see figure 3 for L-5) 235
present a region at small values of q for which £x and E3 are rather q-independent, tending to the q-0 surface mode value. This means first, that the excitation is mainly of surface type and second, that the energy of the «lngl« (one for each L-value) surface mode obtained in ref. 9 for the step electron density model: 2 2 2 2 W - Op L/(2L+1) • 2/3 (2L+1)(L-l) BF /R (6) m 2 a where fir (3/5) I/ v», hjLS_IAO__ej(£xJLJt!dflfiSAdSflJa__£fiC£fiCtii£t0_-£fi£B After this flat region (which is absent in the L-0 mode), the response starts being dominated by bulk modes and both energies increase rapidly. For q > qc the spectrum is dominated by electron-hole excitations. The energy and momentum are so large that the excited electron is quasifree and its dispersion relation is - q2/2. To give an idea of how these results depend on the ITFff approximation, we show in Fig. 3 the TF and KS results for E3 L-5. One can see that the agreement between both calculations is excellent. 236
It is also worth noting in Fig. 2 the structures along the El curve which are also present in the E3 curve, although less marked. The maxima of these structures correspond to the q-quantized bulk modes in the hydrodynamical approximation. Indeed, for a constant gs electron density these modes are excited at discrete values of q such that [5]
(Actually, eq. 7 is exact only for L«0 modes; for L»l it is a good approximation) . In Fig. 2 we have indicated by arrows the first four roots of eq. 7 corresponding to R«18.06 which is the jellium
radius for Na92. B. Infinite systras
If we calculate the m-^ and m3 RPA SR for a step electron density n of radius R and take the q and/or R —» «» limit, we get the following asymptotic expression for the excitation energy:
2 2 2 cor 2 4 CO *m3/m1 - where the term -vF/3x is the local exchange contribution, and cor £2 (n) - -2/9 a r3(n) [b+ 2 r3(n)]/[b + rs rs«4, the kinetic, exchange and correlation energy contributions are 0.138, -0.051 and -0.004 au. Thus, altogether exchange and correlation are a 40 % of the q2 kinetic energy contribution. For
Al 4. Concluding raurks We have studied the RPA response of metallic spheres to q-
and L-dependent operators lz,lqr) YL0 [12]. One of the main advantages of our method is that only electron gs quantities are needed to apply it. This allows one to use models of different complexity for describing the gs electron density, like the crude step density model, or Thomas-Fermi and Kohn-Sham models. Consequently, we can compare them and establish their influence on the quantitative results. In particular, we conclude that the ITFW gs electron densities yield results in good agreement with the full KS-RPA calculations. This is of great importance for the applicability of the method, since for spheres with N 2 200, selfconsistent KS-RPA calculations are not technically feasible due to the increasing washing out of the electron shell structure of the sphere. We have separately analyzed the q and L dependence of the collective excitation energies. We conclude that surface mode frequencies C0L are q-independent and that for each angular momentum value different from zero there is only one surface mode. The volume modes are q-quantized. For values of q larger than qc, i.e., short excitations wavelengths, no collective modes are possible and the excitation energy tends to that of a quasifree electron. We have obtained a simple RPA dispersion relation for the infinite electron gas which includes exchange and correlation effects, generalizing the dispersion relation obtained by Lundqvist in the plasmon pole approximation [11]. Finally, we conclude that the q-quantization is the origin of the oscillations in the scattering spectrum found by Batson [3] (see Fig. 7 of this ref.). In his experiment, he finds an oscillatory behaviour of the scattering amplitude as a function of the radius of the sphere. It seems clear from our results that for a fixed momentum q, the scattering will be resonant for all the radii R such that qR corresponds to a peak in the response function. This already stems from the step density hydrodynamical model. Here, we have shown that more realistic models also predict the same behaviour. 238
* Supported by the CAICYT (Spain), grants Nos. PB85-0072-C02-00, AE87-0027 and PB84-0388-C04-03 flj.- T. Halicioglu and C. W. Bauschlicher Jr., Rep. Prog- Phys. 51, 883 (1988) [2].- J. M. Cowley, Surf. Sci. HI, 1598 (1982) [3].- P. E. Batson, Surf. Sci. X5£, 720 (1985) [4].- R. Fuchs and F. Claro, Phys. Rev. B 25., 3722 (1987) [5].- N. Barberan and J. Bausells, Phys. Rev. B 21, 6354 (1985) [6].- R. Ruppin, Phys. Rev. B H, 2871 (1975) [7].- W. Ekardt, Phys. Rev. B 2&, 4483 (1987) [8].- LI. Serra, F. Garcias, M. Barranco, J. Navarro, C. Balbas and A. Maftanes, Phys. Rev. B 21, 8247 (1989) [9].- LI. Serra, F. Garcias, M. Barranco, J. Navarro, C. Balbas, A. Rubio and A. Mafianes, J. Phys. C (1989), to be published. [10J.- O. Bohigas, A. M. Lane and J. Martorell, Phys. Rep. 51, 267 (1979) tllj.- B. I. Lundqvist, Phys. Stat. Sol. 21, 273 (1969) [12].- LI. Serra, F. Garcias, M. Barranco, N. Barberan and J. Navarro, submitted to Phys. Rev. B (July 1989) 239
Wake-Riding Electrons Emitted By Antiprotons Traversing Solid Targets J. Burgddrfer, .7. Wang, and J. Muller Department of Physics University of Tennessee, Knoxville, TN 37996-1200 and Oak Ridge National Laboratory, Oak Ridge, TN 37831-6377
1. Introduction The dielectric response of the medium to a swift ion induces collective charge-density fluctuations which result in an oscillatory polarization potential trailing the ion ("wake"). The concept of such a "wake" dates back to Bohr1. The first quantitative treatment of the dynamical screening potential around an ion was pioneered by Neufeld and Ritchie2. Meanwhile, a large number of investigations implementing approximations at various levels
3 of sophistication have been performed . A typical example of the wake potential V.mmkt is shown in Fig. 1 for Slt+ in Aluminum at v, =* la.u. calculated in plasmon-pole approximation to the dielectric function. The plasmon-pole approximation is one of the simplest approximation which still accounts qualitatively for most of the features of the wake. Clearly, more subtle effects like bow waves and other dispersion effects are neglected. We will restrict ourselves in the following to the plasmon-pole approximation including a phenomenological damping. We also neglect in the following effects of the self-wake which can affect the shape of the wake potential4-*.
In 1974, Neelavathi, Ritchie and Brandt* put forward the intriguing proposal that the additional minima of the wake potential trailing the Coulomb wdl should support bound states. Electrons trapped in these wells have been named "wake-riding" electrons. These states would correspond to a new type of bound states which is non-atomic but nevertheless strongly correlated with the atomic (or ionic) projectile. A simple variataooal calculation using Gaussian orbitals showed that the ground state is bound over a wide range of parameters*''. Wake-riding states are, in fact, only transient quasi-bound sti 240
pUu)
w+ Fig. 1 Wake potential of a S projectile with * speed of vp = la.u.
in an Al- foil (n, = 0.0088,7 = 0.03) in the frame of the projectile. The
projectile moves in positive z-direction.
which decay into the low-energy Coulomb continuum of the projectile upon exit of the leading charged particle from the solid. Consequently, wake-riding electrons are expected to form a peak in the forward electron spectrum with lab velocities new v, s: v,. The ejection of isotachic electrons in forward direction by ions penetrating solids is therefore a signature of wake-riding electrons. Unfortunately, the real work! is much more complex
Ejection of electrons in forward direction in general, and in the velocity regime vt ~ 5P in particular, is induced by several competing mechanisms which can easily overshadow the wake-riding electrons. Most prominent are the "cusp* electrons which correspond to the low energy continuum states in the field of a positively charged ion formed by direct excitation or electron capture10. According to Wigner's threshold law for an attractive
Coulomb field the cross section for elastic scattering approaches constant at threshold in the frame of the projectile11, 241
da a
when lit, are multipole moments (anisotropy coefficients) and 0 is the polar angle of emis- sion. Both direct excitation of bound states of the projectile as well as electron capture from the valence or core levels of the target can populate those final states in the eon- tinuum. In fact, any smooth excitation function across threshold leads to (1). Upon a kinematic transformation this corresponds to a singular doubly differential cross ssction in the lab frame,
The enhancement of the cross section near vt £s vp according to (2) renders the observation of wake-riding electrons very difficult if not impo—ible.In addition, multiple scattering of binary encounter electrons in the solid provides an additional source of dectrons populating
continuum states with vt ~ vr. Recently, the experimental study of wake-riding electrons accompanying aatiproton transmission through carbon foils using the Low Energy Antiproton Ring facility at CERN has been proposed13. First experiments are presently underway. We have performad a theoretical study to explore the possible existence of a peak ofwake-riding" electrons accompanying antiprotons". Two important features to be discussod in detail below make their observation in the forward-electron spectrum for antiprotoas more likely than in the spectrum for positively charged ions: (a) The well-known cuspGke tnhanoameat in the forward spectrum of positively charged particles is absent, thereby facilitating the observation of wake-riding electrons which appear in the same region of the spectrum, and (b) the wake-riding states are localized a factor of - 3 closer to an antiproton than to a proton of the same speed. Electron capture probabilities into wake-riding states are therefore dramatically enhanced. 242
The threshold law (2) appears in the projectile centered continuum states as a final- state normalization factor
^) IT
where v = v, - vf is the velocity vector in the rest frame of the projectile with charge Zf.
An attractive final-state interaction between the electron and the ion (Zp > 0) leads to a
cuspiike enhancement ~ 2irZp/v of the cross section near v« « vp, while for Z, < 0 a pro-
nounced dip ("anticusp") ~ (2x\Zr\/v)txp(-2x\Zr\/v) occurs". The repulsive final-state interaction strongly inhibits forward scattering with small relative velocities v, thereby "burning'' a hole into the forward spectrum and suppressing ECC. Clearly, ELC cannot occur because an antiproton (p) does not support bound states. The presence of this "hole" for antiprotons may aiFord the opportunity to study the existence of wake-riding electrons.
2. Bound states in the wake
We have calculated the bound-state spectrum in the wake for antiprotons at vr = 6 a.u.. For an estimate of the yield of wake-riding electrons it is crucial to take into account contributions from excited states in the wake. For excited states the anisotropic harmonic- oscillator approximation underlying calculations for the ground state breaks down. Figs. 2 a and b shows the potential along the beam (£) axis (i.e., p = Q) and perpendicular to the beam axis (z = zmin) where zmtn is the coordinate of the center of the first wake trough. A crucial point for the likelihood to observe wake-riding electrons is the fact that
Zmin 2 (2 + s9n(Zr))^k€/4 (4) where A = 2xvr/ur is the wavelength of the wake oscillations and us, is the plasma fre- quency. Therefore, zmtm is approximately a factor 3 closer to an antiproton than to a proton. This increases the anharmonicity of the potential as clearly visible in Fig. 2a. The 243
-.1
-20 4) zfeu)
Fig. 2a Wake potential near an antipcotoo with v, = 6 a.u. in carboa along the beam axis (p = 0).
0h
-.1
-.2
10 20
Fig. 2b M s perpendicular to the beam iu(t* *„*, « -10.5 «.«.) 244
shape of the potential becomes important in the tail of the wavefunction which, in turn, enters the capture cross section for wake-riding states. In calculating approximate wave- functions we have retained the notion of approximate separability of the potential while incorporating anharznonic effects by writing13.
8.41 8.3- /^c~K . /—NJ~2 x=a.2- //7\ \ •-/ \ ' 8.1 - a a. CD -B.I- \\ / y\— -a.2- -B.3- -B.4- -8.5- -8.6: -8.7- -18 -28 -38 -4B
Fig. 3 Wavefunction Xi(z) in the wake of an antiproton in carbon at
vf = 6a.u.
where S(z) and T(p) are the potential curves taken from the cuts through the potential surface (Fig. 2). The wavefunction can be written in factorized form as
,«m# (6)
The resulting one-dimensional Schrddinger equations can be easily solved numerically. The solutions for the three lowest-lying states *,(r)(0 < * < 2) are shown in Fig. 3. Their binding energies (including the ground state energy in the p motion which closely rcwmblei a harmonic oscillator) are ««, =£ -0.1, ej S -0.03, and «j S 0.04. The second excited state is already a resonance in the continuum describing an electron bouncing forth and back 245
between the adjacent humps of the wake potential. Resonances in the continuum can contribute, however, when their lifetime due to autoionization is larger than the lifetime due to collisional destruction. The latter affects true bound states as well and can be taken into account within the framework of the transport theory discussed below. The numerically determined wavefunctions Xi(z) were fitted to a basis expansion of harmonic oscillator eigenstates. The latter are particularly convenient for calculating capture cross sections. It should be noticed that the accuracy of the resulting wavefunction is limited for two reasons: For highly excited states near the ionization (in our case, X1.2J the separability assumption(5) is no longer valid. We have investigated the energy spectrum in the near threshold regime of the wake using semiclassical methods16 and find significant deviations from a separable behavior characterized by ordered sequences of energy levels. Secondly, even though the fit to the harmonic oscillations basis converges rapidly in the Li norm (i.e. on the average) it can be quite inaccurate in the tail of the wavefunction which contributes substantially to the wake cross section.
3. Electron capture into the wake
The transition amplitude for electron capture into wake-riding states is given in second- order Born (2) approximation by
*./ = <*..*.|V, + VfCoV,|«,->, (7)
where Go is the free-particle Green's function and Vrt are the (effective) interaction poten- tials of the projectile and target. At high speeds, capture from the K shell of the target (in
the following carbon) dominates (i.e., $,- = $1,). Accordingly, Vt can be taken as the bare Coulomb potential of the carbon nucleus with an effective value Z, = 5.7, where we have taken into account screening effects by the passive la electron. The projectile potential Vp contains both the bare Coulomb potential and the dynamical screening potential. However, since capture requires a large momentum transfer in a hard collision and the dynamical screening potential is "soft,"1 only the Coulomb part is important. The evaluation of the 246
17 capture cross section ac using Eq (7) and standard techniques reveals the remarkable result that the first-order Born term (Bl) is negligibly small compared to the second-order Born term. This is a simple consequence of the fact that in the Bl approximation capture is mediated by high-momentum components of the initial- and final-state wavefunctiona. However, the "soft" wake potential exponentially suppresses high-momentum components •n *w«*c leading to an exceedingly small cross section11. The dominant contribution it therefore provided by the second-order Born term in Eq.{7) which closely resembles the well-known Thomas double-scattering mechanism19 for ion-atom collisions. Here an elec- tron is first scattered off the projectile by = 60°, followed by a second deflection at the target by about 60°, such that the electron finally propagates in approximately the forward
direction at zero speed relative to the projectile. The fact that Vp is repulsive for antipro- tons rather than attractive for positively charged ions is immaterial since pure Coulomb
scattering is invariant under charge conjugation (Zr -* -Zr) and resulting differences in the phase factor in the B2 term cancel because the Bl term is negligible.
The velocity dependence of 2t 3 At v, ~ 6a.u. the cross section per carbon atom is of the order of ae » \0~ em . Using the numerical wavefunctions (Eq.6) instead of the harmonic oscillator approximation and including the first excited state increases the cross section by a factor ~ 1.3 at vr = 6a.u. Inclusion of the second excited state results in a increase by a factor 3. Since, on one hand, excited states contribute significantly to the total cross section while, on the other hand, they are rather inaccurate, there is considerably uncertainty as to the resulting emission spectrum. Their influence on the emission spectrum will be illustrated below.
Considering the high solid target density and the fact that additional contributions from outer target shells have been neglected the experimental observation of a peak of wake- riding electrons should be within reach. For protons, on the other hand, the cross section is several orders of magnitude (> 6) smaller in the velocity range under consideration 247
because of the rapid decay of spatial overlap. This very likely accounts for the fact that wake-riding electrons have not yet been found10.
Fig. 4 Cross sections as a function of the velocity i> of p used as input for the classical transport simulation. —, cross section for capture of a carbon /("-shell electron into the ground state (harmonic-oscillator approximation) of the first binding well of the wake dp,---, cross section for electron ejection with laboratory velocities t>, > y, in » binary collision with p is first Born approximation. 4. Transport of wake-riding electrons
The wake-riding electrons, as well as electrons generated in binary-encounter events, suffer multiple scattering before exiting the foil. The determination of the observable emission spectrum requires the study of the electron transport in the presence of the field of the nearby projectile. We employ a microscopic Langevin equation,30 248
^ (8)
describing classical trajectories of an electron under the influence of the field of the pro- jectile as modified by the wake field. The electrons are subject to random forces F(t) representing stochastic collisions inside the solid. The complete solution of the transport problem is given by a Monte Carlo sampling of an ensemble of initial conditions for the phase-space coordinates which are propagated according to Eq. 8. The phase-space distribution of initial conditions consists of both binary-encounter (BE) electrons and wake-bound electrons. For the simulation of the initial velocity distri- bution of BE electrons, a first-order Born approximation for ionization has been employed.
Only energetic electrons with vc > 0.8vr have been included in the transport calculation.
The integrated BE cross section a BE for energetic electrons ve £ vr (Fig. 4) is several
orders of magnitude larger than vc. In order to improve the statistics we have calculated the stochastic evolution of wake-bound electrons separately and added their distribution function, weighted by their relative cross section, to the binary distribution. The initial classical distribution of the wake-bound electrons is determined by the spatial probabil-
3 ity density distribution |4watt(r)| restricted to the classically allowed region and by a uniform distribution over all negative energies larger than the value of the wake minimum. In order to relate the dynamical evolution in the bulk to the post-foil experimental observation, modifications due to the penetration of the exit surface must be taken into account. The sudden breakdown of the dynamical screening near the projectile at the sur- face leads to a redistribution of the final-state population. This has particularly dramatic effects for wake-riding electrons that are in the close proximity of a repulsive Coulomb field of the.antiproton. Figure 5 shows the velocity distribution of wake electrons after the sud- den switch-on of the Coulomb field which can be interpreted as a Coulomb half scattering at the antiproton near the exit surface. The initial distribution prior to the break down of screening was an isotropic velocity distribution of the classical wake-bound states centered 249
about the projectile (v, = vr = 6 a.u.). The defocussisg by Coulomb scattering is clearly visible.
Fig. 5 Velocity distribution of electrons originally bound to the wake inside
the foil after (half) scattering in a pure Coulomb potential oip{vf = 6a.u.) upon exit from the solid.
The effect of transport on the peak of wake-riding electron is illustrated in Fig. 6. Here we display the shape of the resulting peak of wake-riding electron* with and without modification by transport effects. The peak without multiple scattering corresponds to the singly-differential spectrum derived from the two-dimensional distribution of Fig. S integrated over the forward cone with cone half-angle 6 = 5*. Note that the peak position is shifted to lower velocities due to Coulomb defocussing, or more precisely, due to Coulomb half scattering. The shift in velocity is approximately given by
where Ee is the Coulomb energy near the wake minimum
1 (10) 250
This is the amount of energy (or velocity) gained by the election in a Franck-Condon type transition from the wake potential curve to the Coulomb potential curve at the eat surface when the dynamical screening breaks down. The steady-state (or "equilibrium") diitribution resulting from steady-state production and decay by multiple scattering of '.V'alee-ridini g electrons as described by the Langevin equation (8) does not display a peak
o.w
o/a
I 0.08 0.04
0.00
Fig. 6 Shape of the peak of wake-riding electrons emitted by an antiproton
in carbon (vr s 6a.u.) into a cone with —«ni»nifr of Q - s*_ riding electrons without multiple scattering, - - - steady-state solution of the transport equation (8) for the distribution under the influence of multiple scattering. shift but an enhanced tail of low-energy electron due to slowing down and a tincrtsai of intensity near the peak. Note that the distributions of fig. 6 do not include the batkgnwad contribution due to multiply scattered binr T electrons. lairing the latter into ~*~nit the resulting distribution at electron velocities near the projectile velocity (3 < », < 7) » showa in Figs. ? and 8 for v, » 6c*. and cone hatf-angks 8§ « 5* and 2&*. In Fig. 7 251
we have ?—'">*'* that only the ground state and the first excited state contribute to the capture while in Fig. 8 we have included the second excited state in the continuum as well. In the absence of a usual cusp, three features are clearly -risible: a steep rise at the upper end of the spectrum which signifies the mrnants of the anticusp valley in the single-collision spectrum, a background due to multiply scattered binary electrons inside
the valley, and a broad peak due to emission of wake-riding electrons near vt s: 5.6 on top of it. Since the wake-riding electrons give rise to a well-localised peak while the binary spectrum shows locally little angular dependence, the peak due to wake-riding electrons should become the dominant feature for sufficiently small 6s. The contribution of slowed- down binary electrons to the customary cusp peaks has also been found to be small31 for positively charged ions.
0.046
0.036 -'
0.02S
0.0«
Fig. 7 Normalised convoy electron spectrum for antiprotons {t> = 6a.u.) emitted into a forward cone of half-angle 6s * 5#(—) nd 2.5* ( ), only ground and first excited state (x«,i) included. It should be stressed that the accuracy of the present calculation is limited due to the uncertainty in both the calculation of o, as illustrated in Figs. 7 and 8 and the treatment of the transport of binary electrons. Since the volume of velocity space of the observed 252
0045
0 035
to
i 0025
0015
Fig. 8 As Fig. 7, but including the second excited state \2 (see »«**)•
spectrum is small compared to the volume of velocity space of all initial conditions for bi- nary electrons which can contribute to the spectrum after multiple scattering, even modest statistical accuracy requires a large number of trajectories. We used a total of 1.8 x 10* initial conditions which resulted in 2s 600 events in the forward spectrum for a cone angle 0 s 5# and which required smoothing using large bin sizes {At- > 0.40o.u.)- The validity of the second Born approximation at only moderately high velocities (v, £ 6a.u.) may be questionable. Furthermore, we have observed that the cross section depends sensitively on the shape of the wavefunctions in the exponential tail which, in turn, may be affected by the separable form of the wake-bound state and by the plasmon pole approximation to a free-electron-gas model employed in the present calculation.
In summary, while the observability of wake-riding electrons is not yet unambiguously established, our calculations show that chances for success are considerably higher for antiprotons that for protons. One of us (J.B.) should liks to thank Pedro Echenique, Rufus Ritchie, and Yasu Ya- mazaki for stimulating discussions. This was work was supported in part by the National 253
Science Foundation and by the U. S. Department of Energy under Contract No. DE AC05-84OR21400 with Martin Marietta Energy Systems Inc.
References
1. N. Bohr, K. Dan. Vidensk. Mat.-Fys. Medd. 18, No. 8 (1948).
2. J. Neufeld and R.H. Ritchie, Phys. Rev. 98,1632 (1955), 99,1125 (1955).
3. P.M. Echenique, R.H. Ritchie, and W. Brandt, Phyt. Rev. B.20, 2567 (1979).
4. P.M. Echenique, W. Brandt, and R. Ritchie, Phys. Rev. B33, 43 (1986).
5. J. Lindhard, private communication, (1989).
6. N.V. Neelavathi, R.H. Ritchie, and W. Brandt, Phys. Rev. Lett. 33, 370, 640 £ (1974). W. Brandt and R. Ritchie, Phys. Lett 62A, 374 (1977).
7. M. Day, Phys. Rev. Lett.44, 752 (1980).
8. A. Mazarro, P. Echenique, and R. Ritchie, Phys. Rev. B31,4655 (1985).
9. A. Rivacoba and P. Echenique, Phys. Rev. B36, 2271 (1987).
10. M. Breinig et al., Phys. Rev. A25, 3015 (1982).
11. J. Burgdorfer, Pbya. Rev. A33, 1578 (1986).
12. Y. Yamazald et al., CERN Report No. CERN-PSCC 87-39 (unpublished), p. 108; CERN Report No. CERN PSCC 87 40 (unpublished), p. 108, and private commu- nication.
13. J. Burgdorfer, J. Wang, and J. Muller, Phys. Rev. Lett. 62,1599 (1989).
14. C. Garibotti and J. Miraglia, Phys. Rev. A21, 572 (1980); M. Brauner and J. Briggs, J. Phys. B19, L325 (1986).
15. J. Muller, Diploma thesis, Universitat Frankfurt, 1989 (unpublished).
16. J. Muller, J. Burgdorfer, and D. Noid, to be published. 254
17. J. Briggs and L. Dube, J. Phys. B13, 771 (2980).
18. Y. Yamaialri and N, Oda, Nud. Znstr. Meth. 194,415 (1982); P. Echenique and R. Ritchie, Phys. Lett. Ill A, 310 (1985).
19. R. Shakeahaft and L. Spruch, Rer. Mod. Phys. 51, 369 (1979).
20. J. Burgdocfer, in XVIICPEAG New York, Book of invited papers, Ed. A. Dalsr garno, (American Institute of Physics, 1989), is press; J. Burgdorfer, Transport Theory for Convoy Electrons and Rydberg Electrons in Solids, in "Lecture Notes in Physics", Vol. 294, p. 344, Springer-VerUg, Berlin (1988).
21. H. Schroder, Z. Phys. DT, 65 (1987). 255
NOH ranuiMTXvx SZEAVXOOIS ZH HCITATHW AR> IOHISATIOM or ATOMS IT FAST CHAISES MlXXCXZfl
K. GAT1T and A. SALXV Laboratoire das Collisions Atoalquea, Universite Bordeaux I. 351, cours da la Liberation, 334OS Talanca Cadax, Franca
ABSTRACT
Non parturbaciva bahavioura in collisiona at rathar high impact valocicias ara axhibicad by means of apodal theoretical treatments. In tba casas of excitation and loniaaelon procaaaaa, it la shown that saturation effaces nay appear in contradiction with tha pradiceiona of the firat Born approximation .
1 - INTRODUCTION
Tha elaceronic stopping power for faat particles of charge ZL travailing in aattsr, is generally thought to ba proportional te zj. Indeed, the description of elementary procaaaaa arlaing in atomic colli- siona at high inpact velocities, ia often made by means of perturbation treatcents like the first Born approximation (BI) for excitation and ioni* sation [1] or the Continuum Distorted Wave (CStf) theory for charge transfer [2-4] . For tha first two processes, BI predicts that total crosa sections are proportional to Z* . However, non-perturbative bahavioura cf chaaa processes have bean pointed out recently at moderately high impact velocities.
An cxperiaantal study and a special theoretlcel analyais of tha excitation of tha levels Furthermore, the investigation of tha simultaneous capture and ionisation by impact of various bara ions (H*. Ha2*, 11s*. 0**), impinging at 1 MeV/nucleon on helium, showed tha standard BI to ba inadequate to describe the ionisation process at small impact parameters [8]. It In shown that the ionisation probability is also satureceo. in tha later case. A similar behaviour la also exhibited far double ionisation of helium by bara ions [9].
Whether or not tha predlctiona for tha stopping power might b* changed in view of these new results is in open question. 256
2 - T^E SATOBATIOH OF THE ELECTBOMIC EXCITATION OF HIGHLY H1A&CSP gfg rnT.T.TftTWC UITH ATOMS OF CAS The model, that has been adopted to investigate the excitation process, is strongly influenced by the actual collisions under considers* cion. The abovementioned excitation of the ion Fe2** impinging *t 400 HeV in various dilute gases is a typical instance. Therefore, the conditions of such collisions will be examined in order to introduce a simplified Modal. Since one looks after the excitation of the impinging ions Fe2**(ls2), the letters are called the "targets* in whet follows. In the same way, the atoms of the gas are named the "projectiles". A 400 HeV laboratory energy of iron ions corresponds to an impact velocity of 16.91 a.u.. To a first approximation, the electrons of the projectiles may be considered as free electrons impinging on targets Fe1** with an Impact velocity of 16.91 a.u.. Since a velocity of about 22 a.u. is required for free electrons to excite the Lyac helium-like line of Fe2t* , projectile electrons are unlikely to excite it directly. Indeed, high momentum components may be found in the orbital momentum distribution of projectile electrons. However, high aomen- turn electrons are located close to the projectile nucleus. Since impact paraaeters ouch less than the radius of the K-shell In the projectile contribute appreciably to the excitation, the nucleus of the projectile appears much aore likely to excite directly Fe2** than electrons bound to it. Thus, a reasonable model for direct excitation considers this latter to be due mainly to the Coulomb interactions between the target Fe2**and the projectile nucleus. Multiple electron processes, such a» simultaneous ioni- sation of Fe2** and charge transfer to Fe2** might also be invoked. However, they have been estimated not to be a dominant process [6]. Furthermore, in a collision where the projectile is a neutral atom, the process of electron transfer from the target into the projectile is strongly inhibited. Hence, the present model ignores the coupling between excitation and electron capture. Finally, a highly charged helium-like ion may be considered, to a good approximation, *» an ion with two hydrogen- like orbltals, which are defined with a screened nuclear charge.
Thus, our model for direct excitation consists in a projectile made of a bare nucleus of charge Zp Impinging on a target made of a hydrogen- like ion. the nuclear charge of which is Zj. The projectile may excite or ionise the target but is prevented from capturing the electron. The theoretical investigation of the excitation process starts from the fractional form of the Schwinger variational principle derived in the eikonal approximation (5,10}. In this treatment, the amplitude to excite the target from the state lct> to the state 10> is.
( 4£ IVI a ) ( e ivl < ) (1) v - v c; v 257
where u£ / and i|£ ^ «ra cha standard scattering wava funcions, V Is tha targec projeccile ir.caracclon and Gj is cha target propagator. Tha expression (1) is stationary with raspaet to slight variations of j T,,- I( 31 Vli )(0'')n( J IVlB) (2) l.J
DJf-( J | V - VC; V | l) (3)
vhara li> and I j > balong to B, and Bj raspactivaly. Tha axprassion (2} of T3a is stationary whan B, and S^ ara enlarged. In tha last iapleaentation of cha fomalisa. Cha sacond Bora-like Matrix elements ^ j |v G* V I 1 ^ have baan avaluatad through an axpansion of Gj on th« whole discraca sptccrua of cha targec [7,11,13] . Furthermore, cha convergence co cha stationary result has baan tasted through a stretching of tha sizes of A, and Bj . To achieve chis goal, A, and Bj have been chosen in tha complete sec of trsvlling eiganscatas of tha target. One has considered the excita- tion of che levels (Is, 21) and (Is. It) of fa2** by impact of various nuclei [7,11] and che excitation of tha levels 21 and 3JS of hydrogen atom* by proton iapaccs [11,13], where t - 0.1. Sines we are dealing with target excitation froa the ground state, the states ID and IJ> are restricted to a subset which contains the lowest target states, including la) and 10) [7.11.131.
Thus two kinds of avitrix elaaencs have eo be calculated : (1} First Born-like eleaents
(j I V I i) - BJ^1 (Z,,. Zy, v ) (4)
(ii) Second Born-like elements
where v is the iapact velocity. The actual perturbing potential is obtained by omitting the long range Couloab Interaction V between the projectile and the whole target i.a. :
V -i {6) 258
whara R is the incernuc', .ar distance while Z, and ZT are the charges of the projectile and of the target nucleus respectively. It is easy to show that V may be ignored in the deteraination of the transition probability J3]. thus, the potential responsible for the transition is :
V - Z, [i - -~-T-\ (7) (R IR - xl
where x is the target nucleus-electron distance.Then, for each pair (i.j). the following scaling law may be established for the k-th Born-like matrix element :
3 if* (7,. 2T, v) - Z% if** (I, ZT. v) (B)
Thus, the matrix elements D; ( defined in the expression (3) may be written:
Dn - Z,, Bjl (1. ZT. v) - Z\ RJP (I. ZT. v) (9)
while one has :
(j I V 1 i) - Zp Bj?" (1, 2T, v) (10)
When Zp is high enough, the matrix D is doninated by the second 3orn teras 2 and D'' becoaes proportional to Z^ . Thus. T(ja tends to a finite value
independent of Zp. It is known as the "s*cur*cion effect" of excitation cross sections when the charge of the projectile is increased [5,10].
For, che excitation of Fe2**(lsz) to the levels (Is, nl) where n - 2, 3 and € - 0, 1, the largest basis set is made of the 5 orbitals Is. ns , np0 , np., and np., .
However three stages of approximation have been considered for and j*p^. in order to the test the convergence.
\ (i) < > and are replaced by la> and 10) respectively i:i expression (1). The procedure is called the Schwinger-Born approximation (SB).
(ii) |u£^ and W'g// are expanded on a two-state basis set made of la) and l@>. The procedure is refered to as Schwinger 2-2 (S22).
(iii) Finally, Ar 1 I
Figure 1 - Excitation of the level (ls.2p) of Fez**(ls2). impinging at an energy of -CO MeV, in che laboratory fraae. on various atons of nuclear charges Z^. The total excitation cross sections per electron axe. indicated in co2. They are plotted as functions of Zp. Experimental data [5,6] : £ population cross sections ; • direct excitation cross sectioncti s obtained after substraction of the cascade contribution (10 X for He and N, 20 X for Ar) and an estimation of the double process contribution (30 Z for N, 50 Z for Ar). These contributions have not been estiaated far Xr [6]. Theoretical results : — • — BI ; — • » — BII ; — + — SB : S22 ; S55 ; —x—FSS5 .'5,6j (see text) ; A Reading's coupled state calculations 112].
Ar _. L. '5 us a » a is 2,
Figure 2 - sane legend as figure 1, but for the excitation of the level (ls,3p) 260
The coca! cross section for che excitation of Fe24* by various aeons (by various nuclei in our model) arc indicated on figures 1-2. Also plotted are che first (Bl) and second (BII) Bern approximations, as well as che first S55 results (FS55) obtained by Brendl* and Gayet [5,6.10] with a Halted expansion of che propagator fl*.. The present SSS results are close Co che FSSS ones. Details about numerical and analytical calculations may be found elsewhere [10,11]. Ic Is interesting to notice that S22 and S55 are rather close to each ocher. Except for che excitation of the level 3p, SB lies far from S55. Furcheraore, its behaviour is often different from the ones of S22 and S55 (7,11]. The new SSS results agree well with experimental data for the exci- tation of che levels 2p and 3p. A discrepancy with che results of Reading's coupled state calculations [12] appears for the excitation to 2p by argon impact. Ic is more pronounced with our new results than with the previous FS55 ones. However, SSS values for the excitation to 3p agree well with Reading's ones.
Ic is worth noting chat the saturation of total cross sections that is observed experiaentally is the one predicted by our varisxlonal treatment. Thus, we have siiown, chat non perturbative behaviours may occur ac rather high iapact velocities for processes «s simple as the excitation process. In any case, El lies far above both experimental and SSS theore- tical results. Values provided by BIX are even worse except for low Z,(Zp< 3) where perturbation conditions are fulfilled. In the lacer case, all theories are in good agreement. Thus, the effort required to make a BIZ prediction appears not Co be worth che trouble. 3 - SIMULTANEOUS CAPTURE AND IOWISATIOH OF HELIUM Recent coincidence experiments (14,15,16} have been performed to study the reactions : Iq* + H«s(ls2) — i<"--"* + He* (11)
— !<«•"•+ He** + e' (12) where Iq* is an ion of charge q. Since the helium target is in a singlet state, che application of an Independent Electron Model (IEM) is particularly simple. Let us denote by P£(P) and Pi(P) the single electron probabilities of capture and ionisation, respectively, in a collision with iapact parameter p.- In the present case, to be consistent with the indepen- dent electron model, Pe>j(P) has been calculated using a Harcree-Fock descripcion of che initial He(ls z) state ; further, one assumes that the second electron is not affected while the first one is ejected or captured. The single capture cross sections without or with simultaneous ionisation are 261
J^ 2Pe(P)(l • P, (P))
<*. , " 2» £ PdP 2Pe (p> P, (p) (14)
respectively, if w« n«gl«cc any other channel (e.g. double capture).
Tha avaluaclon of Pfi(p) haa baan made using the Continuum Distorted Wave fCDUA) codes of Belkic et «1 [17] to calculate the amplitudes of single capture. Then, the Hankel tranafonu of the latter ones penlt to evaluate Pe(P) [18].
Capcure into excited states nf, which plays an appreciable role in most cases considered here, is accounted for through tha n^3 lav (see e.g. [3]) :
3 Pe(P) a 2- pe (nf.P) +pe(n0.p) nj £. nf (15) nf-l nf- nj+1
where pc(nf,p) is the probability of capture into the final state nf. In 3 any case, ve have checked that i^ is large enough to ensure an n'f behaviour of pe(nf-,p) . For instance, a value of Kg as large as ten is s required for 0 * projectiles. The accuracy in tha calculations of pe(nf,P) has been verified by an integration over P in order to get tha total capture cross sections. Then, the latter is compared wieh that given by tha standard CDV2 code of Belkic et al [17]. The accuracy is found to be better Chan 1 Z in each case.
The ionisation probability has baan calculated using two different approaches : che Born approximation and the Multlpole Expansion Defined on One Centre (MEDOC), a procedure introduced by Cheshire and Sullivan [19]. In che first Sorn calculations, the Initial (bound) and final (continuum) states of the ejected electron ar« known from a numerical integration of the Schrddlnger equation for an electron in the field of a potential VT(r), where r(r,9,
For large projectile charges 2 , P, (P) as calculated with the Born approximation, aay be larger than one at small Impact parameters. The perturbation approach is clearly unsuitable to this case since the Sorn series is equivalent to a series in powers of Z,. Therefore, us made use of che MEDOC non-perturbative numerical method, which consists in expanding che one-electron wavefunction in partial waves around tha targee :
* (r.c) - £* V5(8.
The tiM-dependent Schrodlnger equation describing the aoclon of Che electron in both the potential VT (r) and che daw-dependant field of the projectile can be put in the fora of a set of coupled differencial equations in both r and t of Che type :
- Z R&,,,(r.t) Lt,m,iz.z) (17) { 2 dt2 flr ) I'm'
whero R'?>( are aatrix oleaent* of che potential between spherical harao- nics Y]J and Yj! . Thes« aquations are integrated by a finite difference procedure, in a code Bade by Salln (unpublished yet). Two points should be stressed. (I) The aethod cannot be expected to be adequate when the capture probabi- lity is large. This does not occur in Che cases considered here. (II) The limitation, therefore, is on the number of partial wrm* included in the expansion (not on the strenghc of the potential). In the present case we have included € - 0,1. A discussion on the accuracy of this approximation is given below together with the analysis of the results. The nost Important fact is illustrated in figure 3 : che capture probability decreases very rapidly for all che cases considered here when P > 0.5 au. This is in sharp contrast with the ionisation probability which decreases very slowly with impact parameter due to the dominant contribu- tion of the dlpole tera in the ionisaclon amplitude (At - 1). To evaluate che total ionisation cross section in the saae energy range, one aust inte- gr&re up to iapacc paraaecers larger than 10 au. Furthermore, che monopol* eera (A£ - 0), which plays a weak role in che total cross section, is doal- nant at saall iapact paraaecers. Therefore, no conclusion on siaultaneoua capture and ionisation can be drawn by extrapolating qualitative ideas on total ioaisation processes.
A second important fact is chac the saturation effects aentioned above are auch stronger for saall lapact paraaeters (figure 3). For 0** iapacc, che Born approximation to P,(p) gives probabilities larger than one at saall iapact paraaeters. Thus, deviations froa che Born approximation are auch aore visible in siaultaneous capture and ionisation processes than in single ionisacion collisions. Hence, aeasureaencs on double active- electron processes in heavy-ion-atoa collisions at high iapacc velocities are very welcoae.
a Our results for <*t\/ t are coapared with experimental data in figure 4 and in cable 2. Besides che use of che independent-electron uodel. their aain limitation lies in the introduction of t « 0, 1 only in the MEDOC calculations. We can evaluate the accuracy of this approximation by looking at Che convergence of the Born expansion with I. If we include only I - 0, 1 in che Born approximation, then che error in P,(P) is of the order of 10 Z. Of course ve cannot completely trust this argument for 0** iapact since the saturation effect is so strong that che Born results are unrealistic. However, even in this case, che contribution of t values larger Chan one decreases as p decreases. 263
peroMttr lou)
Figure 3 - Probabilities of capture Pc(p; and ionisation P,(p) of one electron of helium by s 1 0 * impact at 0.7 MEV amu , , P£ (p) normalised to 1 at P - 0.05 au where its value is 0.298 ; ---, P,(P) calculated by the MEDOC procedure (see text) with t - 0,1 ; —••—P,(P) calculated by the first Bora approximation with I • 0, ...,4.
1 •
OS 1.0 lapoct Mtr«y
Figure 4 - Ratio of the cross section of siaultaneous capture and ionisation to the single capture cross section cr. (/cr= as a function of the iapact energy in MeV asm'' for various projectiles (H*, He2*, Li3*, 08*). Theoretical predictions [20] : evaluation of capture by che CDW approximation and of ionisation by the Born approximation C5) and/or by the MEDOC treataent (•). Experiaental data : 0. [14] ; A. [15] : o, [l«]. 264
Table 1 - Cross sections for single capture and for capture with lonisation of heliua by ion impact (in ca2)+ab stand* for a x 106
CDW+Born CDU+MEDOC Experiment
Energy Ion (MeV anu'1
H* 0.4 1.9"'9 1.2-2° 6.3-2 1.9-1' 9.7'21 5.0-2 7.r21 3.2-2 0.75 1.2-20 4.1-22 3.5-2 1.0 3.O"21 7.9-23 2.7-2 3.0-2 1.5 7.3-2* 1.8"2
He2* 0.4 2.1"'8 7.1"19 3.3"1 2.3'18 4.9"19 2.V1 3.r'» 7-1? 1.4-' 0.75 1.8-'9 2.9-2° x.64-1 1.8'19 2.6-2° 1.42"1 1.1-' 1.0 5.O-20 6.1'21 1.21'1 1.0*' 1.5 7.7"21 6.1"22 7.9-2 8.5-2
Li3* 0.4 3.9•18 5.7-18 1.3 6.3'18 2.7'18 4.3'1 7.7'18 3.0''8 3.9'1
8< 0 0.75 5.I'18 6.9'18 1.37 1.1 1.0 1.8'18 1.9'18 1.05 0.9 1.5 3.7"19 2.8"19 7.7'1 0.5
Table 2 - Ionization cross sections of heliua by impact of ions with charge 2 Zp at 1.4 Mev/aau in en )
Single ionization Double ionizacion
Born MEDOC Born MEDCC+ Experiment Born MEOOC Experiment (t-O.l) 1 1.6'17 1.78"18 5.7-2° 15 3.6"15 1.3'ts 3.96"1' 1.69'15 1.79-15 2.9-15 3.0-16 2. 20 6.4"15 2.03'15 6.96-'* 2.7"" 2.60'15 9.12-15 4.8-'• 36 2.07-'* 3.9'" 2.26*15 6.16-15 5.72-15 9.6"1* 1.4-;5 1.7" 13 44 3.1"1* 4.8-'5 3.37-'s 8.17"15 7.21-15 2.r 2.0"15 265
Tha agreement bacwaan our calculations and experiment is vary good. Strong deviations from cha Born approximation for P, can be obsarvad. It is particularly striking in tha casa of Li3* impact. Evan for H* impact,we gat a 25 X diffaranca at 400 IcaV, although for tha sasw anargy, tha total ioni- sation cross saction Is in full agreeaent with tha Born approximation [14]. We nota that our calculations with tha MEOOC approach for H* and Ha* at 0.4 MeV amu'1 show a discrapancy of 35 Z with tha experimental data. Excopt In thasa last casas, our rasults show that tha experimental data in coapa> tibia with an indapandant alactron model if an accurate «valuation is made within this modal. Consequently, capcura with ionlsation can ba usad to proba tha small impact parameter ionlsation probability. It is worth noting that Salin [21] showad that siailar conclusions can ba drawn in studying singla and douhla ionlsation hallua by multi- chargad ion impact at 1.4 MsV aau*' . His fraaaworlc is axactly tha saaa as cha ona quotad abova, i.a. : I EM and either tha Born approximation or tha HEDOC traatnant with < - 0,1 to gat tha lonisatlon probability. Tha singla (S) and double (D) lonisation cross sections aay ba written as : too " 2 JJoPd P 2 PfCp) [1 'Pi(p) 1 cl8)
This work is particularly interesting since it is shown that saturation effects (due to non perturbativa contributions) can show up even for single ionisation. Salin's rasults are quoted in table 2. Because of the limitations of the MEDOC procedure already Mentioned , ona cannot expect MEDOC predictions to be reliable when capture processes are 5*por- tant. In the present application, CDWA calculations indicate that it might be the case for Zp£ 36 [21]. As co contributions of partial waves t>2. they have been evaluated in the Born approximation for single ionlsation. For Zpi 36, they appaar to be comparable to MEDOC results for £-0,1 (see table 2). Although Salin's results are not fully reliable for Zpi 36, they compare favourably to experimental data in any case. The difference betwaan MEDOC and Born predictions for double ionisation is impressive, indicating that perturbative treatments are irrelevant to double (and even single) ionisation when the energy of the projectile is not high enough. Thus, an independent electron model can be a powerful tool to investigate non perturbative behaviours of the lonisation process in simultaneous cap cure and ionisacion as well *s in double ionisation by fast ion impacts. COMCLPSIOH Ue have shown that non perturbativa behaviours show up even at high impact velocities. Therefore, care mist be exercised when dealing with high energy collisions. Indeed, the widely used Z| behaviour of excitation and ionisation processes cannot always be trusted. 266
[ 1] I. Laatavan-Vaiaaa, D. Hannacare end R. Gayat, J. Phya. Francs, 4f (1988) 1528 [ 2] R. Gayat. J. Phya. B 96 (1972) 483 [ 3] Dr. Balktc, R. Cayat and A. Salin, Phya. Rap. < (1979) 279 { 4] Dz. Balklc, R. Gayat and A. Saiin. Coaput. Pbyx. Coaaun. 32 (1984) 385 [5] B. Brandla. R. Gayat, J.P. Rozat and K. tfohrar. Phys. Rav. Latt. 24 (1985) 2007 [ 6] K.tfohrar, A . Chatloui, J.F. Roxat, A. Jolly, F. Faraandez, C.Staphan. B. Brandla and R. Gayat, J. Phya. B 19 (1986) 1997 [7] R. Cayat and M. Eouaaoud, (fuel. Inatrua. and Math. B 42 (1989) 515 [ 8) R. Gayat and A. Salin, J. Fhya. B 20 (1987) L571 [ 9] A. Salln, Fhya. Rav. A 3« (198?) 5471 [10] B. Brandla, Thaaa da Trolsiaaa Cyela. Wnivaraita Bordaaux 1. (1984), unpubllshad (availabla on raquaat ; taxt in franch) [11] M. Bouaaoud, Thaaa d'Etat, Univarsica Bordaaux I, (1988), unpublished (availabla on raquaat ; taxe in franch) [12] J.F. Raading, 1983, privata coaounlcaticn [13] H. Bouaaoud and R Gayac, Third Europaan ConCaranca on Atoale and Holacular Physics, Book of Abatract p 491, Bordaaux 1989, A. Salln Editor [U] H.B. Shah and H.B. Gilbody, J. Fhya. B 18 (1983) 899 115] H. Knudtan, L.K. Andaraan, P. Hvalplund, J. Soranaan and D. Ciric, J. Fhya. B 20 (1987) L253 [16] J.A. Tanix, M.W. Clark, R. Prica, S.M. Farguson and R.E. Olson, Mucl. Inatrua. Hath. B 23 (1987) 167 [17] Oz. Balkic, R. Gayat and A. Salin, Coaput. Fhya. Coaaun. 21 (1981) 153 ; Couvut. Phys. Coaaun. 22. (1983) 193 ; Coaput. Fhya. Coaaun. 22 (1984) 385 [18] R. HcCavroil and A. Salin. J. Phya. B 1 (1968) 163-71 ; J. Fhya. B U (1978) L693 [19] J.M. Chashira and B.C. Sullivan. Phya. Rav. J£2 C1967) 4 [20] R. Gayat and A. Salin, J. Fhya. B 20 (1987) LS71 [21] A. Salin. Fhya. Rav. A ?a (1987) 5471 267
Target thickness dependent convoy electron production of aligned molecular HeH*-ions'
J.Kemnler*, K.O.Groeneveld, O.Heil, K.Kroneberger, H.Bothard Institut fur Kernphysik dor J.W.Goethe Universitat D-6000 Frankfurt am Main, Germany
Many experiments have shown that the collective excitation of valence elec- trons effects the alignment [1] and the stopping power [2] of fast diatomic molecular ions penetrating thin solids. Collective excitations also seem to play an important role for the total electron emission induced by fast ate ic and molecular ions [3]. For a specific part of the overall electron distribution the convoy electrons, accompanying the projectile ion with
equal speed and direction (v.*vp), production mechanism had been proposed which attach their origin to the wake-potential [4],[5]. Because of the long range effects caused by the wake-structure molecular projectile ions are here a sensitive tool to investigate the few Angstrom environment of a fast moving ion in a solid. When the molecular ions enters the foil the binding electrons are stripped of and both fragments are mutually repelled by the Coulomb explosion: Measuring a physical signal as a function of target thickness means here to observe it for different lnternuclear distances r*. The wake-potential is most pronounced in the direction parallel to the beam. Therefore an experiment deals with the problem that the observed effects are smeared out because of the randomly oriented axes of the mole- cular ions. Until now, no simple method is known to deliver molecular projectile ions with the lnternuclear axis aligned to the beam direction, so a selection of the corresponding events after the target is necessary. The detection of the Coulomb exploding fragments with isotropic orienta- tions requires a complex experimental setup [6]. In a recent experuMnt there the convoy electron yield Yc{-*) for ffc* penetrating a thin carbon foil has been iwasured as a function of orientation of the die luster axis. It has been found that the yield Yc{•*=&) tor fragment ions aligned to the
This work has been funded by the German Federal Minister for Re- search and Technology (OffT) under the contract Number 060F173/2 Ti 476, DFG/Bonn. present address: IPN Lyon-1, Universite Claude-Bernard, F-69622 Lyon, France 268
EXPERIMENT He HeH -> C - FOIL
rw PPArm O DIAPHRAGM AS HV TARGET ELECTRON TAC SPECTROMETER
TIME
1: Experunentai set up for the detection of aligned fragments beam direction is smaller than for fragment ions with their axis perpendi- cular to the beam ( Yc(^=O°) < Yc(-*=90°) ). A result which does not fit into the proposed mechanism for wake-riding of convoy electrons. But also the dependence of the stopping power for N2* icns compared to the stopping for the atonic constituents cannot not yet be explained with the wake mechanism [7]. One important problem here is the effective charge or better the strength of the coupling between the fast projectile ion and the valence electrons in the solid. To get a better understanding for the contribution of the wake potential to convoy electron production it is therefore better to choose a sampler collision system. Because a complex device as described in ref. [6] was not available we used a simpler experimental set up which was able to detect the fraction w;nch is aligned with respect to the beam direction. The molecular ion beam penetrates a thin carbon target vhich is placed m the focus point of a magnetic sector spectrometer fig. 1). The emitted electrons are deflected by 90° and registered with a channeitron, whereas the Coulcmb exploding molecular fragments remain undisturbed. After a 269
MeV/u) -> 8O -. HeH* (O.3 c - FOIL
«O ' f'"\ At 5.8/
4O • • leading 1trailing H* CO 2O Z 3 O u 40 5O eo 7O TIME Fig. 2: Time spectrum of a coincidence between convoy electrons and H* fragments from HeH* (0.3 MeV/u). Trailing and leading H*-ions are clearly distinguishable. flight path f of 3.41 m the internuciear separation for HeH* of 0.3 MeV/u reaches 41 im so that only correlated fragments inclined by 5° can pass the diaphragm. With an electric field the fragments of different charge state and mass are separated before the reach the position sensitive parallel plate avalanche detector (FPAD). The separation between leading and trai- ling H+-ions is achieved with a time of flight technique which requires a start signal delivered here by convoy electrons produced during the pene- tration of the molecular ion through the solid. Because uncorrelated frag- ments, i.e. single He- and H-fragment which are strongly inclined with respect to the beam axis, can also pass the diaphragm with a second coinci- dence circuit it is ensured that only those events are counted when both, the He- and the H-ion reach the FPAD. It has been found that this condition is satisfied only by 5 % of all ions entering the detector. In a feasibility experiment [8] we studied the target thickness dependant convoy electron yield for aligned molecular projectile ions and compared it 270
1.75 HeH (O.3 MeV/u) -> C - FOIL
• ALIGNED 0 > 1.5O • O RANDOM I u 1.25
(He H ; * u 1.OO > 0 Ra t 5 1O 15 2O TARGET THICKNESS ex Cpg/cm23
Fig. 3: Ratio of the molecular convoy electron yield to the sun of the yields for the atomic constituents for aligned and randomly oriented mole- cular fragments. with the yield registered with randomly oriented molecular ions. In fig. 2 a time spectrun which was started with convoy electrons and stopped with the H* fragments from HeH* {0.3 MeV/u) on carbon foil is shown. Although the time resolution of FWH = 4.7 ns was not very accurate the fraction of trailing and leading H* ions is clearly distinguishable. In table 1 the ration of both fractions Htr«ii/i««d is plotted as a func- tion of specific target thickness ex. This ratio increases with increasing target thickness signifying that loosing the correlation to the He-ion is less probable for trailing fragments. This result can be interpreted as an effect of the wake force behind the fast He-ion. In contrast to these findings the yield of convoy electron shows for alig- ned and randan oriented molecular axis the same target thickness dependence and therefore presented no hints for a wake potential influenced ccnvoy electron production. However, although the convoy electron yield is inde- pendent from the orientation of the dicluster axis, it is still a function of the inteniuclear separatico rx which demonstrates the influence of the correlated action of the charges of He and H. Concluding, we presented a simple method for the detection of aligned dia- tomic molecular ions penetrating thin foils. Our first experimental results gave hints that the molecular enhancement for the convoy electron produc- tion of molecular ions is not swply a direct effect of the wake potential, but more contributed to the correlated action of the fast projectile ions m the solid.
Tab. 1: Ratio of trailing and leading H* ions Htraii/iaad for HeH* (0.3 Hev/u) in coincidence with convoy electrons for different specific target thicknesses ex. Also the non coincident charge state fractions F(q) for He* and He** are given.
Target thickness Htrai 1/lacd charge state fraction [ng/cm2 3 He* He** [%]
5 :t 2 2.05 ± 0.21 32 68 6.5 ii 1.5 2.35 ± 0.25 26 74 8 ii 2 3.05 ± 0.32 22 78 20 it 2 3.4 ± 0.35 22 78 atomic value [9] 16 84
References
[1] Z.Vager, D.S.Genmell, B.J.Zabransky, Phys. Rev. A 14 (1976) 638 [2] J.W.Tape, W.M.Gibson, J.Remillieux, R.Lauberti, H.F.Wegner, Itel. Instr. Meth. 132 (1976) 75 [3] K.Kroneberger, A.Clouvas, G.Scnlussler, P.Koscnar, J.Keanler, H.Rothard, C.Biedermann, O.Heil, M.Burkhard, K.O.Groeneveld, NUcl. Instr. Meth. B29 (1988) 621 [4] V.N.Neelavathie, R.H.Ritchie, W.Brandt, Phys. Rev. Lett. 33 (1974) 302 [5] V.H.Ponce, W.Meckbach, Can. At. Mol. Phys. 10 (1981) 231 [6] Z.Vaager, E.Kanter, Nucl. Instr. Meth. B33 (1988) 98 [7] M.Steuer. D.S.Geranell, E.Kanter, E.Johnson, B.Zabransky, IEEE KS-30 No. 2 (1983) 1069 [8] J.Kemnler, Dissertation 1988, Universitat Frankfurt am Main, Germany [9] J.B. Marion, F.C. Young, Nuclear Research Analysis North Holland Publ. Corp. (1968) Amsterdam 273
RADIATIVE ELECTRON CAPTURE BY CHANNELED IONS
J. M. PITARKE
Fisika Teorikoa Saila, Zicntzi Fakultatea, Euskal Herriko'Jnibertsitatea, 644 Posta kutxatila
48080 Bilbo, Basque Country, Spain
and
R.H. RITCHIE
Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge. TN 37831-6123 USA
and
The University of Tennessee, Department of Physics, Knoxville, TN 37996 USA
Abstract
Considerable experimental data have been accumulated relative to the emission of photons accompanying electron capture by swift, highly stripped atoms penetrating crystalline matter under channeling conditions. Recent data suggest that the photon energies may be less than that expected from simple considerations of transitions from the valence band of the solid to hydrogenic states on the moving ion. We have studied theoretically the impact parameter dependence of the radiative electron capture (REC) process, the effect of the ion's wake and the effect of capture from inner shells of the solid on the photon emission probability, using a statistical approach. Numerical comparisons of our results with experiment are made. 274
1. Introduction
Many processes can give rise to X-ray production when energetic heavy ions interact with
solids. Radiative electron capture (REC) is said to occur when an ion captures an electron into
one of its shells and emits a photon1.
This process contributes only in a very small way to the total production of radiation;
however, it becomes more important and amenable to interpretation when the bombarding ions
are channeled through the solid, since in these conditions the ions never approach closer than
0.1 or 0.2A to target atoms, so that the yield of characteristic X-rays, one of the most
important contributions to the total production of radiation, is very much smaller than that
observed when the beam is incident on the crystal in a random direction. Furthermore, it has
been shown that stripped, incident channeled ions have "frozen" charge states, and although
nonradiative electron capture is the dominant mechanism for charge transfer, in ion channeling
the inhibition of close collisions makes the radiative process relatively more important2.
The REC process has the possibility of becoming a significant tool in the study of the
momentum distribution of target electrons, so that some experimental data3"7 have been
accumulated. Indeed, recent data7 suggest that the photon energies may be less than that expected from simple considerations of transitions from the valence band of the solid to hydrogenic states on the moving ion.
The purpose of this paper is to extend previous5-8-9 theoretical approaches, and to study theoretically the impact parameter dependence of the REC process and the effect of capture from inner shells of the solid.
2.- Theory
We consider a solid, through which a swift, heavy, stripped ion of charge Zj and velocity v passes and captures an electron accompanied by the emission of a single photon. The 275
nonrelativistic Hamiltonian of the system may be written as
H=H0+Hrad+H< • <»
where H^ and HTad are the hamiltonians of the electron-ion system and the free radiation field, respectively, and
H1 = - - A . V (2) c is the interaction hamiltonian between them, with
C \ ik.r / + \ .-. e e (3) -) k, ("fc^.J-
kX ^ V ak x and ek x are, respectively, the annihilation operator and the polarization vector of a photon with wave vector k and polarization index X(K= 1,2), which in free space is perpendicular to the direction of propagation, coj^k is the frequency of a photon of wave vector k, and V is the normalization volume of the radiation field. The movement of the ion relative to the center of mass of the system has been neglected, as well as a term in H' which is quadratic in the potential vector. Unless otherwise stated, atomic units arc used throughout
The transition probability for a target electron to be captured into a k-shell orbital of the moving ion, emitting a photon, may be written in first order perturbation theory as follows:
iov t [dte (4) where < H'fj ( t) > represents the time-dependent matrix element of the matter-radiation interaction:
< H'fi (t) > = < 01 afc < 10 > , (5) where -iv.r Here (pj (r, t) is the wave function of the electron in the initial state, and JO > is the vacuum 276
state of the radiation field. In eq. (6) b represents the impact parameter. Eg represents the binding energy of the captured electron in the K orbital:
1 , 7t Z, CO E^=2Zr-^r ' (7) and
E, = jv2 , (8) with v the velocity of the ion. The effect of the ion's wake has been considered including in eq. (7) the potential of the polarized medium at the site of the ion10. We should mention, however, that for typical REC experiments this potential is approximately 15eV, in such a way that this correction turns out to be not greater than a 1 or 2% of the total REC energy. REC processes can proceed by capture of free, as well as bound electrons. In particular, with the aim of understanding how the emission probability corresponding to capture of bound electrons depends on the impact parameter, one can write, on very general grounds, the initial wave function of the target electron as follows:
• io>,i (10)
p = q + k +v , (11)
cof= cof - a , (12) and d£2 is the infinitesimal element of solid angle in the direction of emission. The spectrum obtained in this way has been found to be very much wider than that for capture from the 277
electron gas; moreover, it has been found to be wider as the impact parameter increases, as
is obvious from figure 1, where the REC peak widths obtained with eq. (10) are plotted
against the impact parameter when photons are emitted parallel to the beam direction.
In order to study in detail the effect of capture from inner shells of the solid on the photon
emission probability, a knowledge of the many-electron wave function of the solid is required;
however, in order to simplify we adopt a statistical approximation5, assuming that the
response of an infinitesimal element of volume in the solid at position r, where the total
density is n(r), is the same as that of an electron gas at that density. Thus, we define the local
Fermi energy to be 2/3
— . (13) 2
when capture occurs from the vicinity of r.
Assuming in this approximation that the wave functions of electrons in a given part of
the solid are plane waves:
where qj represents the initial momentum of the electron and qF is the Fermi momentum
(qF=1.92/rs), the REC probability per unit energy and solid angle of the emitted radiation is"
'(zj+qf)4 where q'i = qj - k - v . (16) and 2 ky + 17
The integral over qt in (15) covers a Fermi sphere of electrons, so that we obtain, after some 278
algebra, the following expression:
.2 ,,2 d3V ^"k f\ f2 . 3AZ+B' n (18) = TT" dq) dq q — , J dco. dQ 2 JI c v Q f* when
and zero otherwise. In these expressions,
A =ii sin 9 - (1 -}j.2)1/2cos 9 cos ip
B = (l-| (20)
1- - (H cos 6+ (1 -p2)1/2 sinG coscp)+ —^ c 2c2
(21)
(22)
(23) and
v \ I--cos 9 J , (24) with 9 being the angle formed by the direction of emission and the beam direction. To obtain
(18) summation over the directions of ek in a plane perpendicular to the given direction of the emitted radiation has been performed.
If we define the local Fermi energy to be given by (13) we may calculate, in our approximation, the REC probability per unit energy and solid angle when capture occurs from the vicinity of r, once the total elctronic density there is known.
We have employed the electronic density values shown in fig. 2 for silicon, as computed 279
from a relativistic Hartree-Fock program12 using the Wigner-Seitz boundary condition, and we have found that the peak energy of the emitted radiation decreases strongly when electrons are captured from inner shells ot the solid, as is obvious from fig. 3, where REC peak energies for 160 MeV S16+ stripped ions channeled through silicon are plotted against r, that is, the position where capture occurs, when photons are emitted at 46,5° to the incident beam.
Notice that the REC peak energy of photons resulting from capture of target electrons at
rws, the Wigner-Seitz radius, is almost equal to
EREC= E^E, (25)
as may be inferred from considerations of transitions from the valence band of the solid to
hidrogenic states on the moving ion, but it deviates when capture occurs from regions of
relatively high electron density. This variation may explain the REC "deficit" observed in
recent data, as we will see below.
Finally, if we assume that the bombarding channeled ions sample on straight-line trajectories all portions of the WS sphere corresponding to impact parameters greater than a given bmjn, the statistical average of the REC probabilities per unit energy and solid angle may be written as
d\ . 3 <-^i->=3 fr(r.bmin)JA.[EF(D]> (26) HO HO J J HIYI HO where A,
is given by (18), and EF (r )by (13), as discussed above. On the other hand, if one assume that each electron undergoing capture from the solid experiences the full Coulomb field of the projectile, neglecting screening by other electrons of the solid, essentially the same dependence of the REC probabilities on projectile velocity is predicted5-13. 280
3. Results Figure 4 exhibits plots of the REC probabilities per unit energy and unit solid angle for a
target electron to be captured accompanied by emission of radiation at 8 as obtained from eq.
16+ (26) for 160 MeV S (8 = 46.5°) ions channeled through silicon with bmin=0.3A . Figure 5 shows the dependence of the peak energy positions so obtained on the minimum impact
parameter, together with the REC energies calculated from (25), and the experimentally
measured REC energies 7.
Notice that the agreement with experiment is good when the minimun impact parameter is
0.3 A, showing that capture from regions of relatively high electron density in the channel
explains the experimentally observed REC deficit. The origin of the REC deficit is obvious,
too, from figure 6, v/here measured and calculated REC peak positions for bare sulfur ions channeled through silicon are shown as a function of incident ion energy, for different values of the minimun impact parameter.
This work has been sponsored by the Universidad del Pais Vasco / Euskal Herriko
Unibertsitatea, by the office of Health and Environmental Research, U.S. Department of
Energy, under contract DE- AC05-84QR21400 with Martin Marietta Energy Systems Inc., and by the US-Japan Cooperative Science Program of the National Science Foundation, Joint
Research Project No. 87-1631 l/MPCR-168.
Note: The authors are preparing a revised and more accurate version of the theory described above in which the REC probability is calculated us.'ng wave functions that describe the electron and the ion in relative and centcr-of-mass coordinates.
References
1. H.W.Schnopper, H.D. Betz, J.P. Dclvaille, K. Kalata, A. R. Sohual, K. Kaiata, A. R.
Sohual, K.W. Jones swd M.E. Wesner, Phys. Rew. Lett. 29, 898 (1972). 281
2. B.R. Appleton, R.H. Ritchie, J.A. Biggerstaff, T.S. Noggle, S. Datz, C.D.Moak and
H.Verbeek, J. Nucl. Mater. 63, 513 (1976).
3. H.W. Schnopper, H.D. Betz and J.P. Delvaille, in Atomic Collisions in Solids, edited
by S. Datz, B.R. Appleton and CD. Moak (Plenum, New York, 1975) Vol. II, p. 481.
4. B.R. Appleton, T.S. Noggle, CD. Moak, J.A. Biggerstaff, S. Datz, H.F. Krause, and
M.D. Brown, ibd., Ref.l.p. 499.
5. B.R. Appleton, R.H. Ritchie, J.A. Biggerstaff, T.S. Noggle, S. Datz, CD. Moak. H.
Verbeek, and V.N. Neelavathi, Phys. Rev. B 19, 4347, 1979.
6. Andriamonje S., Chevallier M., Cohen C, Dural J., Gailiard M.J., Genre R., Hage Ali
M., Kirsch R., L'Hoir A., Mazuy B., Mory J., Moulin J., Poizat J.C, Remillieux J.,
Schamaus D. and Toulemonde M., Phys. Rev. Lett., 59, 2271 (1987).
7. CR. Vane, S. Datz, P. Dittner, J. Giese, J. Gomez del Campo, NJones, H. Krause,
P.D. Miker, H. Schone and M. Schulz. Proceedings of the 11th Werner Brandt
Workshop on Penetration Phenomena of Charged Panicles in Matter, Oak Ridge,
Tennessee, April 14-15, 1988
8. J.S. Briggs and K. Dettman , Phys. Rev. Lett., 33, 1123 (1974).
9. J.E. Miraglia, R. Gayet and A. Salin, Europhys. Lett., 6, 397 (1988)
10. R.H. Ritchie, W. Brandt and P.M. Echenique, Phys.Rev. B 14, 4808 (1976), P.M.
Echenique , R.H. Ritchie and W. Brandt, Phys. Rev. B 20, 2567 (1979).
11. This result has been obtained in the laboratory frame, and it differs from the first Born
result found in the cm. system. The reason for this was not understood at the time of the
conference, but, subsequendy, this question has been resolved and we have found that the
correct result is only obtained in the cm. frame. This error resulted in a bigger REC
deficit; the new results will be shown in a future work.
12. T.C. Tucker, L.D. Roberts, C.W. Nestor, T.A. Carlson and F.B. Malik., Phys. Rev.
178, 998 (1969).
13. H.A. Bethe and E.E. Salpeter, Quantum Mechanics of One-and Two-Electron Atoms
(Academic, New York, 1957), p. 320. 282
Figure Captions
Fig. 1 REC lincwidths calculated from eq. (10), as a function of impact parameter.
Fig. 2 Relativistic Hartree-Fock calculations of the total electronic density in the
Wigner-Seitz sphere, for silicon12.
Fig. 3 REC positions for 160 MeV bare sulfur ions channeled through silicon, as
obtained from eq. (18), as a function of the distance r from the captured target
electron to the nearest atom in the crystal.
Fig. 4 REC probability per unit energy and solid angle, as obtained from eq. (26)
when photons are emitted at 0 = 46,5° with respect to the beam direction, for
160 MeV S16+ ions channeled through silicon.
Fig. 5 Calculated REC peak positions for 160 MeV bare sulfur channeled through Si,
with 6 = 46.5°, as a function of minimun impact parameter. Also shown are the
REC energies calculated from eq. (25) and the experimentaly measured ones of
ref. 7.
Fig. 6 Measured7 and calculated REC peak positions for bare sulfur ions channeled
through silicon, as a function of the incident energy of the ions, for b^s 0.1
and 0.3 A. Also shown are the REC energies obtained from eq. (25). 283
4 6 impact parameter (a.u.)
Fig. I 284
7 100-
1
Fig. 2 285
REC energy, as obtained from eq. (25)
Kj.3 286
6 7 pbotoa energy (keV) 287
7,0-
6,8-
REC energy, is obtained from eq. (25)
16,6- measured REC energy
8.6.4- u
6,2-
6.0" 1 2 3 minimum Impact parameter (a.u.)
Fij.5 288
100 120 140 160 180 220 EI(MeV)
Kg. 6 289
Direct Methods in Structure Determination by LEED
JB Pendry The Blackett Laboratory Imperial College London SW7 2BZ UK Abstract Most techniques for determination of surface structure interpret the raw experimental data by trial and error methods. For complex structures this is a very time consuming process because of the exponential growth of possible structures with degrees of freedom. Direct Methods sidestep this difficulty. They have been studied in the context of Xray diffraction and it is the heavy atom technique of Xray crystallography that inspires our approach to the surface problem. The idea is that we know the bulk structure lying behind the surface, and we can define a "reference surface" as near to the true surface structure as possible. It is subsequently treated as the "heavy atom" of the surface problem and all phases referred to the reference surface phases. The method works well for some simple examples and future applications of greater sophistication are discussed. 290
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED 1. Introduction Surface crystallography is now firmly established as a
discipline in its own right, with several hundred structures
to its credit [1]. A variety of techniques contribute to the success of the subject: Rutherford back scattering, surface
Xray diffraction, and low energy electron diffraction
(LEED). All have so far resisted attempts to use Direct Methods. LEED is arguably the most successful of the
techniques [2,3,4,5] with the majority of structures determined by this method. It owes this power to the very high density of information that the diffraction process produces: the coordinates of several independent surface atoms can be found using the information contained in diffraction data acquired over a range of a few hundred eV. Indeed it is the very success of the technique that is forcing our attention back to the Direct Methods [6,7]. Suppose that two atoms are adsorbed on a surface and we have no knowledge of their positions or relative configurations. Each atom has 3 coordinates, and we may wish to try 10 values for each coordinate. That works out to be 10 trials, each of which may involve a considerable computational effort. Even if each trial were to take only 1 second of computer time, the complete task would last for nearly two weeks of solid computing. In general for a system with N coordinates to be found and M trials per coordinate, we have, number of trials = M^. {1} 291
J. B. PENDRY There is a non polynomial dependence on the parameters, hence the problem falls into the category of NP problems which are legendary for defeating the attempts of computers to solve them.
In practice we can do better than this because surface scientists perform all sorts of ancillary experiments on
their surfaces: electron energy loss spectroscopy will, for example, tell us a lot about the bonding, and will certainly indicate whether the two atoms in the example above are
bound into a molecule. The vast number of unrestricted trials possible illustrates just how vital this ancillary information is. Even so it is all too possible to take some data on a surface structure, to run a small subset of trials and be lucky enough to obtain reasonable agreement with experiment, and yet be badly in error because an even better tut untried structure lurks in another part of parameter
space. The problem is especially serious because it discriminates against discovery of a really startling structure: anything too improbable will be dismissed by the operator as wasting valuable searching time and will remain undiscovered. Thus our capacity to add to our understanding is limited by our method of data analysis. New things are discriminated against by our inherent conservatism. The history of surface structure determination is littered with instances of false determinations. We need not feel too bad about that: even our distinguished colleagues in radio astronomy have the same problem. At one period the 292
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED number of radio sources actually diminished with time as scientists got better at interpreting the data. But we have no room for complacency. Direct Methods are not a universal solution, but for a class of problems to which they are suited, they offer a way forward that is relatively free cf the pitfalls of the trial and error procedure.
Direct Methods have been studied for many years in the context of Xray diffraction. The simple nature of Xray diffraction theory means that a Fourier transform of experimental diffracted amplitudes will give the atomic coordinates directly. Only the intensities are known so the problem reduces to one of finding the phases. This can be done either by using sophisticated theorems which limit the volume of phase space which needs to be searched, or more traditionally by the heavy atom method: a heavy atom is substituted into a complex unit cell and dominates the
A diffracted amplitudes, oG- Subsequently coordinates of all other atoms are to be referred to the position of the heavy atom so that A _ is chosen to be real. Since the other atoms OG produce a relatively small change in diffracted amplitude.
IAG'2 = 'AoGi2 + 2 'AOG' '6AQJ COS where 8 is the phase of 6AQ. Hence, apart from the usual ambiguity about centro- symmetric crystals, we know the amplitudes and can Fourier transform to find the coordinates. 293
J. B. PENDRY It is this heavy atom technique which inspires our approach to the surface problem, The idea is that we know the bulk structure lying behind the surface, and we can define a "reference surface" based on the same bulk structure, and as near to the true surface as our chemical intuition and other information enables us to guess. This reference surface is subsequently treated as the "heavy atom" of the surface problem and all phases referred to the reference surface phases.
What use are the phases? LEED is recognised as a multiple scattering problem and Fourier transformation, even of LEED amplitudes, generally gives a jumble of nonsense. Here we introduce the second important concept in our approach [8,9,10]: provided that the reference surface is reasonably close to the real structure the diffracted amplitudes will change by an amount that is first order in the atomic displacements. Suppose that we have a surface in which one of the top layer atoms in the unit cell is displaced by an amount 5r from the reference surface. The changes in amplitudes and the displacements will be connected by a tensor,
5A(kp) = Z. T(kp,j) 6rr (3) so that the change of intensity is, A(k 2 + s T(k j) 6r 2 l pH = IVV j P' ji 'W|a + M(lcp'j) 6rj where, 294
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED
M(kp,j) = A*(kp) Sj T(kp,j) + AQ(kp) E. T*2. Realisations of Direct Methods So far we have been deliberately vague on how a Direct
Method might be implemented, because there are several realisations that can be made, all having the same underlying theory, but which look very different in their details, here are some possible implementations, starting with the most simple. linear expansion of 6t in the atomic coordinates Starting from equation {11) we expand 6t(6r.) to first order
in 6t(6rj) = 6ryV ti'r^), (6) and hence.
» Z. M(kp,x..} 6Xj + M(kp,Vj) 6yj + Mlk^Zj) 6zj# (7) where,
M(k ,x.) = exp[i(K-K').r^]
+ AQ(kp) * exp[-i(K-K').r^].}. (8)
We have a matrix equation that can be inverted, provided that we have enough data points. In this very simple 295
J. B. PENDRY
realisation the coordinates are given directly by the
inversion process.
This most simple realisation shows how the method could work. It has been implemented for systems with very small displacements [6], but suffers from the major disadvantage that equation (6) is a very bad approximation for all but the smallest values of 6r.. We stress that this is not a fundamental limitation on Direct Methods as a whole which have much wider validity in other implementations power series expansion of 6t in the atomic coordinates The next logical step is to expand 6t(6r.) to higher orders in &Zy 6t(6r.) = Al 6r. + A2 idz^)2 + A3 (6r.)3 + ... , (9) where we have used shorthand notation for the powers of a vector, A3 (6r,)3 = + A3(x,x,y) fix. 8x. 6y. + .. (10) This power series can be derived analytically from the translation operators given in the appendix of ref [2]. It converges for all values of 6r.. To take a simple example: suppose that we are sure of the x and y coordinates of an adsorbate atom, then we need only make an expansion in powers of 6z,
n 6t(6z) = Sn An (6z) . (11) Substituting into (11) gives.
" |Ao(ltp)|2 * Sn=0 M(kp'n) where, 296
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED M(k ,n) = (A*(k ) expfiJK-K1)-r j P ^ ir ir ir J + Ao(kpJ * expt-i(K-K').rj].}. (13) and by inverting M we can obtain all powers of 62. At first sight this may seem to be giving redundant information: after all if we know 6z, we also have (6z) etc. But consider a system in which not all the atoms have exactly the same height due to thermal disorder or some other cause. Then we must take an average of equation (12) to give, lA'kp»la - IVV*2 = Sn=0 M'Vn) <(fiz)n>' ll4) where <(6z)n> = Jp(6z) (6z)n d6z. (15) and P(6z) is the probability distribution of 62. Thus we are in reality determining the moments of the distribution of heights, from which P(6z) itself can be inferred. If the system is an ordered one, <(6z)n> = (6z)n, (16) and we can use this to check that the system is indeed ordered. If not then we get some additional interesting information. Fourier expansion of 6t If the reference structure is periodic parallel to the surface then the change in diffracted intensity wili have the same periodicity in the parallel location of, say, an adsorbate atom and hence if we consider only displacements parallel to the surface, 6r , we can write. 297
J. B. PENDRY 2 5A< + A ( 6A ( 3 iA exp[i(K-K').rp].
+ Ao(kp) * exp[-i(K-K').rp].}
£g M(kp,g) exp(ig.6rp), (17) where g is a reciprocal lattice vector of the reference structure. In the most general case the adsorbate atom will be vibrating about within the unit cell and will be described by a probability distribution, P(6r ), and the measured changes in intensity will reflect an average over this distribution,
|A(kp)|* - |Ao(kp)|> = 2g M(kp,g) , (18) where,
P(6r } = J p exp(ig.6rp) d*6rp. (19) In other words we obtain the Fourier transform of the probability distribution of the adsorbate atoms. How many Fourier components do we need? That depends on how smooth the distribution is. If the adsorbate atom is freely migrating over the surface perhaps we will only need 5 components. On the other hand if the temperature is low and the adsorbate atoms are locked into, say, the hollow sites then P(6r ) is a delta function with many Fourier components, and in this instance it would be better to make a power series expansion of 6t about the hollow site. The expansion of 5t in the z- direction is probably always best done as a power series: atoms rarely have large excursions in the z- direction and the power series will almost always be the most compact expansion. 298
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED more ambitious schemes
The methodology outlined above requires a knowledge of the relative intensities of the beams, and preferable of their absolute intensities. LEED users have got used to the idea that absolute intensities do not matter and Direct LEED would be an even more attractive proposition if they were not required. Certainly it is true that it is possible to extract the same information by conventional analysis from, for example, the Y- function [11] for the hk beam, defined by,
1 2 Yhk(E) = L" /(L" + V^) (20)
L d ln(Ink)/dE 121)
where VQi is the imaginary part of the electron self energy. Direct Methods can be made to work with these more complex functions of the intensity, but certain complications develop. Consider first the case of a single adsorbate atom per unit cell whose position we are trying to find. For simplicity assume that only the z- coordinate is unknown. Using the same trick as before we can write the Y-function as a power series in the height, z, of the adsorbate atom,
n Yhk(Efz) = En M{E,h,k;n) z (22) which again has the form of an invertible matrix equation. Immediately we can see one problem: the definition of Y contains poles as a function of L, and therefore the power series expansion will have a finite radius of convergence, in contrast to the power series expansion for the 299
J. B. PENDRY
intensities which was absolutely convergent. Another more subtle problem occurs when we have a distribution of atoms over several values of z. We write,
n I(z) = En an (23) where is the average value of zn. Now expand Y as a power series in I,
n< n n Y(z) = En,bn,I = Sn.bn.(En an is the same as zn and the series simplifies again.
More complications arise when there is more than one adsorbate atom whose position we are trying to find. In this instance the power series expansion for Y becomes quite a mess involving terms of the form, Y(z) = ... + c(nl,nl\n2,n2') n1' n2' + ... 125) Once we move away from linear functions of the change in diffracted amplitudes, we find cross terms between the two coordinates appearing, and the helpful factorisation of the amplitudes into separate terms for each atom disappears. None of this prevents us trying to implement the Direct Method using the Y- function, but it does mean that terms in the series proliferate, and the structure is more difficult to extract. For a structure with just a few unknown 300
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED
coordinates there is no reason why the Y- function should
not be used.
3. Limitations to Direct Methods
Firstly we do have to assume that the displacement from the reference structure produce small changes in scattering. In practice we have found through experience of the tensor LEED method that displacements of up to 0.4A can be tolerated without producing large errors. It is possible that some sort of iterative procedure could be used for larger displacements whereby an inaccurate estimate of the r•'s was used to define a new reference surface closer to the true structure. We are always free to check the structure we obtain by a conventional dynamical calculation and direct comparison with the data. The great virtue of the Direct Method is that it goes straight to an estimate of the structural coordinates without having to plod around parameter space.
Secondly our method relies on differences in intensity between experimental spectra and those for a theoretical reference surface. Theory is good at predicting peak positions in LEED, but bad at absolute intensities. Ideally we should make two experiments: one on a known reference structure so that the differences could be found purely from experiment. Dynamical inputs such as the Debye Waller factor, inelastic damping, etc, have non- linear effects on the calculated intensities and some thought has to be given 301
J. B. PENDRY
for each system to how the effect of intensity errors can be
minimised.
Despite these limitations we are cautiously optimistic that Direct Methods have a valuable role to play in the future of surface crystallography, and that they will greatly extend the range of structures that can be studied.
Acknowledgements: I wish to thank my collaborators, Klaus Heinz, Klaus Miiller, Wolfgang Oed, and Philip Rous, for the contributions they have made to the ideas reported in this manuscript.
References [1] JM MacLaren, JB Pendry, PJ Rous, DK Saldin, GA
Somorjai, MA Van Hove, DD Vvedensky "Surface Crystallographic Information Service" (Reidel,
Dordrecht, 1987), [2] JB Pendry "Low Energy Electron Diffraction" (Academic, London, 1974). [ 3 ] MA Van Hove amd SY Tong "Surface Crystallography by LEED" (Springer, Berlin, 1979). [4] MA Van Hove, WH Weinberg,and CM Chan, "Low Energy Electron Diffraction" (Springer, Berlin, 1986). [5] K Heinz and K Miiller in "Structural Studies of
Surfaces" (Springer, Berlin, 1982). [6] JB Pendry, K Heinz, and W Oed, Phys Rev Lett, 61 (1988) 2953. 302
DIRECT METHODS IN SURFACE STRUCTURE DETERMINATION BY LEED [7] PJ Rous, MA Van Hove, and GA Somorjai, submitted to
Surf Sci. (1989). [8] PJ Rous, JB Pendry, DK Daldin, K Heinz, K. MUller, and N Bickel, Phys Rev Lett 57 (1986) 2951.
[9] PJ Rous and JB Pendry Comp Phys Ccmm 54 (1989) 137. [10] PJ Rous and JB Pendry Comp Phys Comm 54 (1989) 157. [11] JB Pendry, J Phys C 13 (1980) 937. 303
Non-local Exchange-Correlation Potential at a Metal Surface from Many-Body Perturbation Theory
A. G. Eguiluz Department of Physics, Montana State University, Bozeman, MT 59717, USA 304
Abstract
We report a first-principles evaluation of the exchange-correlation poten-
tial (Vxc) at a metal surface. An integral equation relating (V^£) and the
non-local electron self-energy (£xc) is solved numerically for a free-electron
metal surface, with use of a static approximation for Src. The strongly- inhomogeneous nature of the electron density profile at the surface is treated exactly, i.e., without invoking the usual local density approximation. Our
result for Vzc has the correct image-like asymptotic behavior; it derives im- plicitly from a non-local exchange-correlation energy functional. We study the effect of non-locality on the position of the effective image plane (z0) from an analysis of the image tail of Vxe and also from linear-response the- ory. The difference in the values of z0 obtained by both methods for low metallic densities is attributed to electron overlap effects. We als discuss the effects of Vzc on image-potential induced surface states. 305
Density functional theory (DFT) has proved to be an extremely suc- cessful scheme for electronic structure calculations, in which all many-body effects can be described through a local exchange-correlation (XC) potential Vxc-1 In the great majority of the applications the local density approxi- mation (LDA) is made, in which the XC potential is evaluated in a locally uniform electron gas of density n(x).1 The LDA has proved to be quite successful for many applications, such as calculations of total energies and work functions.2 However, recent variational calculations, which do not re- sort to the limit of slow density variations, yield surface energies which are substantially larger than the LDA values.3 Another drawback of the LDA is that it gives an XC potential which decays exponentially outside the surface, wheras the correct asymptotic behavior of Vxc is that it should be imagelike. Thus, an LDA based calculation failes to describe image- like surface states, which have been observed at various metal surfaces. Former attempts to overcome this shortcoming within a weighted density approximation* were successful, to the extend that an empirical form of the XC hole was used, which revealed an artificial value of the image plane position.5 Here we address the nonlocality of the XC energy functional from a more fundamental point of view. We present results of an a prior: evaluation of 6 Vxc(x) at a metal surface, in which the nonlocal electronic Green's func- tion is calculated directly, thus incorporating the strongly inhomogeneous nature of the electron density near the surface at the outset. In addition, we establish a relation between VXc and the electron density that enables 306
us to carry out linear response calculations with the same ease as LDA- based calculations. The jellium model for free-electron metals is used, in which the ionic charges are approximated by a homogeneous background.3 Our procedure is an adaption to the metal surface problem of work carried out earlier in the study of the fundamental band gap of semiconductors.7 It involves an interplay between many-body perturbation theory and density functional theory. The starting point of our calculation is the following exact integral equa- tion for 8
3 fd x' Vxc(x')fdEgo(x,x' | E) g(x',x | E) 3 3 = fd xx fd x2 fdE ffo(x,xi | E) £Xc(xi,X2 | E) g(x2,x j E), (1) where Exc is the XC part of the electronic self-energy operator, g is the
exact one-electron Green's function, and g0 is the corresponding Green's function for the DFT hamiltonian.6 This equation must be solved itera- tively, since the Green's function go is implicitely determined by Vxc- As a reasonable first approximation for SA'C, which includes the impor- tant physical effect of it's nonlocality, we use the static Coulomb-hole plus 9 screened exchange (COHSEX) approximation. The approximation g = gQ is made throughout. Eq. (1) then turns into a one-dimensional integral equation, depending on the electronic polarizability x° f°r zero wave vec- tor parallel to the surface,10 which can be solved by standard iteration: Starting from Vxc in LDA, the wave functions and energy eigenvalues ob- tained self-consistently with Vxc ars used in the computation of go, X*°\ 307
and SA'CI and the integral equation is solved, yielding a new potential Vxc- The procedure is repeated until convergence is achieved. Results for an Aluminum surface (r,=2.07) are given in the figure. Both the converged solution and it's LDA-COHSEX counterpart are shown. The crucial difference between the "local" (i.e. LDA) and nonlocal results for Vxc(z) is that the latter potential can be approximated by an imagelike potential sufficiently far in front of the surface. This represents an impor- tant improvement for studying, e.g., image-potential bound surface states or tunnel barriers in a scanning-tunneling microscope. Of great importance in surface physics is the effective image plane po- sition z0. It can be defined either by the asymptotic behavior of the XC potential (see figure caption), or by linear response (i.e., by calculating the density induced by a weak static electric field applied normal to the sur- face). The centroid of the induced density equals the linear-response value s of z0. Here we address the question wether these two values are the same, in other words, are there electron-overlap effects built into the equation that controls the image tail? We determine a value for z0 both from the image tail and in linear response. The latter is done by solving the DFT equations in the presence of a dc electric field, using the XC potential for the undisturbed density n.6 For different bulk metallic densities, the results are as follows: From the image tail of Vxc(z), we found that z0 (measured from the jellium edge) ranges from 1.19 a.u. for r, = 2.07 (Al) to -0.11 a.u. for r, = 5.63 (Cs). For the nonlocal linear response calculation, the results for z0 axe 1.16 a.u. for 308
r, = 2.07 and 0.68 a.u. for rt = 5.63, whereas for the local linear response calculation, the corresponding values are 1.25 a.u. and 0.795 a.u., respec- tively. The values agree basically for high-density metals (r, = 2.07), but differ substantially for low metallic densities. This interesting result can be understood from the fact that the work function decreases as r, increases. A smaller value of the work function leads to a more pronounced tailing out of the electrons into the vacuum. One then expects that any electron-overlap
effects built into the image tail of Vxc would become more significant as rt increases, in agreement with our observation. That overlap effects exist at all is a consequence of the fact that the asymptotic limit is reached rather quickly, as seen in the figure. This behavior is in disagreement with pre- vious studies by Ossicini cf al.,A who found, that the asymptotic limit is reached slowly, due to an artefact of the empirical choise of the XC hole.5 However, in agreement with these studies4 we have that the nonlocality leads to a small (as 10% ) reduction of the linear-response value of ZQ. The results of the present paper, based on the use of the COHSEX approximation for Exc, establish only a qualitative test for Vxc- Quanti- tatively, our LDA linear-response values of zo differ substantially from the result obtained in LDA using Wigner's interpolation for XC,2 i.e., 1.55 a.u. for rt = 2.07 and 1.17 a.u. for r, = 5.63. Improved results are expected from an evaluation of Lxc in the "GW approximation,9 which is currently in progress. 309
References
1 P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964); W. Kohn and L.J. Sham, ibid. 140, A1133 (1965).
2 N.D. Lang and W. Kohn, Phys. Rev. B 7,354 (1973).
3 E. Krotschek and W Kohn, Phys. Rev. Lett. 57, 862 (1986).
* S. Ossicini, CM. Bertoni, and P. Gies, Europhys. Lett. 1, 661 (1986); S. Ossicini, F. Finocchi, and CM. Bertoni, Surf. Sci. 189/190, 776 (1987) and Vacuum (this volume).
s E. Chacon and P. Tarazona, Phys. Rev. B 37, 4020 (1988).
6 A.G. Eguiluz and W. Hanke, Phys. Rev. 39, 10433 (1989).
7 L.J. Sham and M. Schliiter, Phys. Rev. Lett. 51, 1888 (1983); R.W. Godby, M. Schliiter, and L.J. Sham, ibid. 56, 2415 (1986); Phys. Rev. B 35, 4170 (1987).
8 L.J. Sham, Phys. Rev. B 32, 3876 (1985).
9 L. Hedin and S. Lundqvist, in Solid State Physics, edited by H. Ehrenreich, F. Seitz, and D. Turnbull (Academic, New York, 1969), Vol. 23, p. 1.
10 A.G. Eguiluz, Phys. Rev, B 31, 3303 (1985); Phys. Scr. 36, 651 (1987). 310
Fig. 1. Nonlocal XC potential(solid line) for r, = 2.07 (Aluminium). Also shown is the XC potential in LDA (dashed line) and the classical image potential V{m — — 4.*_ . (dashed-dotted line), z is measured from the jellium edge in units of Ap = 4.05A, z0 = 0.63k. 01- ) A0H3N3 313
Multiphonon Energy Exchange in the Collision of an Atom with a Surface
J . R. Mansor. Department of Physics and Astronomy Clerason University Clemson, South Carolina 29634 U. S. A.
and
R. H. Ritchie Department of Physics University of Tennessee Knoxville, Tennessee 37996
U. S. A.
and
Health and Safety Research Division
Oak Ridge National Laboratory Oak Ridge, Tennessee 37831 U. S. A.
Abstract
We consider the collision of an atom with a surface under conditions where the energy transfer involves multiple phonons. Such conditions occur with either relatively high incident energies or hot surfaces. Several simple models for the energy exchange to the surface by a projectile with translational energy in the eV range show that such a particle should lose a large fraction of its energy upon collision. A more elaborate model, based on quantum mechanical transition matrix theory, agrees well with new experimental data for the scattering of low energy helium atoms with alkali halide surfaces. 314
I. Introduceion There is considerable interest in the problen of energy exchange when an incident, low-energy atom or Molecule is reflected from a solid surface. ' Such studies have provided invaluable information on surface phonons as veil as knowledge of the particle-surface interaction. Recently, there have been some interesting experiments in which the energy exchange with the surface is measured to be a large fraction of the incident translational energy. •* Typically, such conditions occur when the incident aton or molecule has energy in the electron volt range , or when the surface temperature is large compared to its Debye temperature. The primary candidates for energy exchange mechanisms in such collisions are electron-hole pair creation and phonon excitation. It has been understood for some time that electron-hole pair creation is a process with very low probability in the case of helium atoms as projectiles. More recent examinations confirm this conclusion and extend it to a much larger class of incident atoms and molecules. In particular, for rare gas atom collisions, direct excitation of electron-hole pairs appears unlikely . Thus recent observations of electron-hole pair creation by heavy rare gas atoms incident on semiconductor surfaces may be due to secondary effects, such as local heating ** a result of energy exchange on collision. The object of this work is to examine the phonon mechanism for the energy transfer. We first consider two straightforward mechanisms, the direct mechanical excitation of acoustic phonons, and excitation of surface optical phonons via the interaction with the polarization field. In the case of moderately energetic rare gas atoms both of these mechanisms give a significant transfer of energy to the surface, in qualitative agreement «rith experiment. Ue then move on to a more elaborate model for the energy exchange based on a quantum mechanical transition matrix formalism. This model is used to calculate the temperature dependence of the diffuse inelastic scattered intensity and its distribution in energy. When compared with the detailed energy-resolved experiments of Safron et al. for the scattering of helium by alkali halide surfaces, good quantitative agreement is achieved. An important result of this agreement is that measurements of the diffuse inelastic background signal show directly the numbers of phonons exchanged as a function of incident energy and surface temperature. 315
II. Phonon Excitation Models We will take a fully quantum mechanical approach to the problem. There are two major reasons for using quantum rather than classical mechanics. First, there is substantial evidence that quantum effects such as diffraction are important in the surface scattering of heavy atoms7' , and secondly, a classical approach would preclude obtaining information on microscopic details of the interaction such as numbers and sizes of phonon quanta exchanged. The total probability per collision for exchanging a phonon with the surface is given by the generalized golden rule 2* 2 Rt 2 <|Tfi| «(Ef - Ei)> , (1)
where the sum is over all final states of the particle and crystal, the brackets signify an average over the initial states of the crystal, and the incident current is ji - *k^z/m with m the particle mass and k^z the component of incident wave vector normal to the surface. The transition matrix is related to the interaction potential V through the operator relation
T - V + VGT (2) where G is the Green function of the particle. In atom-surface scattering the dominant contribution to the inelastic transfers come from vertical motions of the surface atoms , and it is a reasonable approximation to consider a one dimensional potential depending only on the dimension z perpendicular to the surface. Expanding the potential as a series in the normal displacements u gives
V(z,u) - V(z) + V'(z)u + V"( )/ 1 o2 f f i -y- E|V'fi | | dt e , (4) 316
where <^ and - - — p(u) T\(u')eluZ . 'lj
where H is the mass of a surface atom, u^ is the maxiaum crystal frequency. n(w) is the Bose-Einstein function, and p(u) is the frequency distribution function. Ue will, for actual calculations, further approximate p(u) by a Debye distribution p(w) - 3u /u^ . For the interaction potential we will take an infinitely repulsive hard wall, a model which has been widely used and often gives excellent results in atom-surface scattering. The matrix elements are well known and are given by
fi fz iz * Combining Eqs. (6) and (5) with (4) and carrying out the trivial integration involving Che f-function leads to 4«kiz rf p(«> f R< dw jk. n(u) + k_ |n k with k+ - iz ~ 2«u/rt. If the incident normal energy is sufficiently high
we have k+ — k^z and Eq. (7) becoaes siaply
(n<«) +1/2) .
At low temperatures only phonon creation processes occur and writing the
energy of normal motion as E^ - ft kj2 /2a we have
t i«I)M . (9) while at high temperatures n(u) » k^T/ftu; and the probability of energy exchange , is
/ftwDM) (iksT/*wD) . . (10) 317
Of more immediate interest to che quescion ac hand is che total energy exchange. This is readily obtained from Eq. (7) by weighting the integrand wich flu. The result, valid for either low or high temperatures is
E - SmE^M . (11)
This staces that che fractional energy exchange for a parcicle incident normally on the surface is 8m/M, clearly indicating ChaC a heavy, energetic rare gas atom should lose a substantial fraction, If not virtually all of its energy upon collision. In spice of the fact that the results expressed in Eq. (11) agree qualitatively with the results measured in Ref. (3), there is ample evidence to suspect that it is inadequate. In its simplest terms, Eq. (11) gives the results for the Born approximation with the exchange of only a single phonon. A more complete calculation would include multiple phonon contributions coming from higher orders in the Born series, and would show that each contribution in attenuated by a Debye-Waller factor. The multiphonon corrections and the Debye-Waller factor tend to cancel each other as we show in the more detailed calculation in the section below. In fact, we note in passing chat this cancellation is complete in the special case of a collision wich a surface having only a single vibration frequency, i. e., for Che inelastic scattering of a poinc parcicle wich an Einstein oscillator. In this case, if R is che probability of single quantum exchange in the Born approximation, then the probability PN of N quantum exchange is given by the Poisson distribution PN - exp(-R)R /N!. The cotal energy exchanged is obtained upon multiplying PN by Nfiw and then summing over all N. The result is E - rtuR, which is identical to che Born approximation result. This simple result, although not a justificacion of che use of Eq. (11), does indicate how higher order effects tend to cancel. Before leaving this section it is of interest to consider another contribution Co che phonon energy loss, che exchange of surface optical phonons, which as our calculations indicate, can also be a significant effect. Surface optical phonons in insulacors and semiconductors can be excited by the polarization field of the incident particle and surface, the sane field which has been shown Co be very inefficient for the excitation of electron-hole pairs. The mutual polarization requires transitions of the incident atom to virtual excited states, hence the first nonzero contribution comes from the second order Born approximation for the cransition matrix <2), 318
Tfi - i fe " ii • Ei • Ei where the polarization potential is* [. ] iK-E -K|z| V(X) - S TJJ i . J4(K) e e (&£ + aK) (13) with
Ji(K> - (iKx. iKy, -K)
2 2 2 I" . - jrZe ftu>s/L K (15) where us the surface phonon frequency. In this case we take the matrix elements with respect to a plane wave basis set. Rather than calculating Rj directly it is more convenient to first calculate the self-energy for inelastic transitions , 2^(z), and then the inelastic transition probability becomes
- fdz Zi(z) (16) zo where zQ is the classical turning point of the repulsive part of the potential V(z) of Eq. (3). We have carried out a number of calculations for a variety of incident rare gas atons interacting with semiconductor surfaces. For incident energies in tht; eV range this mechanism gives energy losses which are a substantial fraction of the incident energy. The conclusions of this section are that two different models describing phonon exchange of the incoming particle with the surface, direct mechanical excitation of acoustical modes and excitation of surface optical modes, both lead to the same qualitative result. A moderately fast incident atom will lose a large fraction of its energy upon collision in a. very short time. In the next section we develop the theory in a manner that allows us to examine in detail the multiphcnon nature of the energy exchange. 319
III. Multiphonon Exchange For a more detailed treatment of multiphonon scattering we go back to Eq.(l) and apply the Van Hove transformation directly, vhich when expressed In terns of the transition rate from particle state i to state £ is
wfi - —* dtL e , (17) where the time dependence of the T-operators is given by the crystal hamiltonian in the interaction picture
iH t/ft -iH t/fi T(t) - e T e ' (18)
Without actually calculating the transition matrix, for example by assuming a model potential and using Eq.(2), we can gain a great deal of insight by noting that the T-matrix elements of interest in scattering theory are given by the coefficients of the wave function asymptotically far from the surface.
(W) T im ffii B fz . - — 2 2 S e e |nf> . * E kfz (nf} kf2
In Eq.(19) the ket |n£> is a state of the noninteracting crystal and the suns over final states are carried out consistent with overall energy conservation between particle and crystal. If the crystal is rigidiy displaced through a displacement Jl parallel to the surface and uz perpendicular, the effect on the transition natrix is
iK-U i(kfz izz Tfi -» Tfi e e " (20)
Ue are considering here the question of energy transfers involving exchange of many phonons and in this case the overwhelming majority of the phonons will be of small energy and hence of large wavelength. The displacement of a long wavelength phonon will approximate a rigid surface displacement, and this will be even more true if the wavelength is large compared with the coherence width of the particle wave packet. Making the assumption that the only tine dependence of the transition operator is in the displacement appearing in Eq.(20) the transition rate (17) takes the form 320
w T dt e fi " S fi where we use the notation k. - ( E, kj2 + k^2 ) and u - ( U,uz ). Vithin the harmonic approxination Che average over crystal states can be readily carried out1A to give
wfi--2 lTfir •"* dt<
where e'2W is the of the form of a standard Debye-Waller factor with U - <(k-Ji) >/2. Equation (22) is the essential result. Although we have used a very simplified approach to arrive at (22) it can be obtained in a much more rigorous way by decomposing the transition matrix into contributions from surface unit cells. Then a generalized form of £q.(22) capable of describing all the diffraction beams is obtained after making that single assumption, that the unit cell is not deformed by the phonon. This assumption is consistent with that stated above, and implies that the results should be valid for long wavelength modes. The important features of Eq.(22) are made clearer if we rewrite it as
2W(k T) wfi - F(fc) e- -' S(k,,T) . (23) where T is the surface temperature. This is in the classic form of a fora factor (essentially the square of Tj^) multiplied by the Debye-Waller factor and a structure factor, where the structure factor is the Fourier transform appearing in (22)
f i - fIdt e e (24)
The structure factor contains all the dynamics of phonon transfer. In particular, if it can be measured independently of the Debye-Waller factor as a function of temperature, it gives immediately the number of phonons involved in the scattering process. To see this we note that the displacement function appearing in the exponential of (24) is proportional to T for temperatures large compared with the Debye temperature. If this 321
exponential is expanded in terms of powers of the displacement correlation function, the result is a structure factor which is a power series ir. T
— d IK, L) dn ~ 9i 1 ~ 341 T ... ~ Jml T . • . • l^J j V aL £ II The n term in the series is the contribution arising from the transfer of n phonons, hence the dominant terms in the series at a given temperature tell immediately which numbers of phonon transfers are important. Note that unitarity arguments (conservation of particle flux) would imply that S(k.,T) should approach exponential behavior at large T in order to cancel the decreasing exponential behavior of the Debye-Waller factor. This behavior is precisely the general case of the cancellation of the multiphonon corrections by the Debye-Waller factor discussed in the previous section. In the next section we derive explicit expressions for a simple phonon model.
IV. Comparison with Experiment Recently, a series of experiments has been carried out measuring inelastic transitions in He-surface scattering which is ideally suited to the theoretical development presented here. The experiments involved helium beams scattered from alkali halid* crystals at energies from 50 to 100 meV and surface temperatures in the range 100-700 K. The detector was placed at the specular position and the scattered intensity was energy resolved by time-of- flight measurements. This enabled a separation of the elastically reflected specular peak from the diffuse inelastic background. As a function of increasing temperature the specular peak was observed to decrease exponentially as expected according to the Debye-Waller factor, while at the same tine a diffuse inelastic "foot" rose out of the background underneath the elastic peak. We can apply the theory of the preceding section using a Debye phonon spectrum as in Sec. II above. We look only at the high-temperature limit and consider only the vibrational displacements normal to the surface. Eq.(5) is used to evaluate <\iz > from which we obtain the exponent of the Debye-Waller factor as 3 o kBT 2 > (26)
The one parameter in this theory is up or the Debye temperature 6^ and this can be evaluated by comparing (26) with the measured value for the thermal attenuation of the specular elastic peak. 322
Using Eq.{22> we can calculate Che intensity of Che observed diffuse inelastic foot. The simplest quantity to check is che diffuse inelastic intensity at near zero energy exchange, or the height of the foot at the position of the specular peak in the time-of-flight spectrum. In this case the Debye-Waller factor and the form factor are the sair.e for both the elastic and diffuse, inelastic intensities, thus their ratio is given by the structure factor of Eq.(24) evaluated at (f - <£- In order to evaluate (24) vn must first determine the displacement correlation function in the exponential. Again caking normal vibration displacement and the Debye frequency distribution this is readily evaluated with the aid of (5;
2 <}c-a(O) k.-u(t)> —> (ki2 + kfz)
3kRT n k + k 2 " < iz fz> 3 (27)
With the results of (27) it is possible to find closed form expressions for the Fourier transform of (24). However it was found simpler to evaluate it numerically. The results of these calculations are shown in Fig.(l) for the case of 44meV heliua acorns incident on a NaCl crystal surface. The filled circles are the experimental points for che Debye-Waller thermal attenuation of the specular beam, and the line drawn through them is our calculation with a Debye temperature of 33OK. This choice compares favorably with the bulk Debye temperature of 321K. • The remaining experimental points are the diffuse inelastic intensity close to the elastic peak, the crosses are measured on the side in which the He gains energy and the squares are on the energy-loss side. The straight line drawn through them is a fit to the data having a slope of the calculation from (24) using the same Debye temperature. In the high- temperature region the agreement is quite good. Calculating the actual shape of the diffuse, inelastic foot would require a choice of an interaction potential model in order to determine the form factor appearing in (22). However, without complete knowledge of the form factor we should be able to estimate Che temperature dependence of the full width at half maximum (FWHM) of the diffuse foot. In this case we note that the actual quantity measured is the differential reflection coefficient, which is che transition rate multiplied by che available volurae in phase space and divided by the incident particle flux: 323
2 oR m |kf| w 3 3 fi • dEfdfif (2>r*) kiz
The form factor in (28) acts as a cut-off factor at large energies, but we make the assumption that it does not vary strongly in the region of the half maximum. In Fig.(2) the filled circles are the measured points for the FWHM as a function of temperature. The solid curve is the calculation using (26) in which we have used the form factor as a free parameter to scale the vertical axis. This does not allow us to make a quantitative comparison, but the qualitative trend is correct. The FWHM increases nonlinearly with temperature which indicates that multiphonon processes dominate at high T. Finally, we can address the question of how nany phonons are involved in the scattering process at a given temperature. This is given in Fig.(1) where it is seen that the structure factor is closely approximated by an increasing exponential at high T, S(T) « e . The n tern in the power series of Eq.(25) is then (cT)n/n! and the largest terms of the series are those for which cT •» n. Taking the above cited value of c - O.OO6O8K , we see that at T - 400K the typical number of phonons involved in a single scattering event is about two or three, while at 700K it is more of the order of four or five.
V. Conclusions We have considered the energy exchange process in single collisions of atoms and molecules with solid surfaces. We find that two separate mechanisms of phonon transfer, mechanical excitation of acoustical modes and excitation of surface optical modes by mutual polarization, can give rise to energy exchange that is comparable to the incident kinetic energy of the particle. The implication is that such collisions can deposit significant amounts of energy in a small region of the surface in a very short time. This helps support the idea that electron-hole pairs, which are known to be created in such collisions, are caused by secondary processes related to the local heating. In order to obtain a more detailed picture of the nature of the phonon- mediated energy exchange process, we have developed a theory of nultiphonon, inelastic atom-surface scattering which is valid for long wavelength vibrational modes. Predictions of this theory for the intensity and width of the diffuse inelastic signal in atom-surface collisions agrees well with recent experiments with He beams scattering off of alkali halide surfaces. These results give important information on the nature of the collision, such 324
as the number of phonons exchanged, and they also show promise of boing useful In analyzing experimental data in which it is desired to separate out the elastic and single quantun contributions from the total scattered signal. Acknowledgments. One of us would like to thank J. Skofronick anc* S. Safron for useful discussions. This research was supported by the Office of Health and Environmental Research, U. S. Department of Energy, under Contract No. DEACOS- 84OR21600 lith Martin Marietta Energy Systems, Inc., and by the U. S.-Spain Joint Committee for Scientific and Technological Cooperation.
Figure Captions 1. Elastic and diffuse inelastic scattered intensities in the specular direction for He reflected by NaCl as a function of surface temperature. The He energy is 44aeV. The filled circles are the elastic peak intensities and the line drawn through the points is the theoretical calculation. The squares are the diffuse inelastic intensity on the creation side of the elastic peak, while the crosses arc from the annihilation side . The straight line is a fit to experiment, and the theoretical prediction is the solid curve displaced below. 2. The full width at half maximum of the inelastic foot measured in the time-of-flight spectrum for the system of Fig.(1), The points are the experimental data and the solid curve is the theory of Eq.(28). 325
References
1. V. Bortolani and A. C. Levi, Rivista del Nuovo Cimento, 2.1 (1986). 2. V. Celli, in Many Body Processes at Surfaces, D. Langreth and H. Suhl, eds. (Academic Press, New York, 1984) p. 315. 3. A. Amirav and M. Cardillo. Phys. Rev. Lett. 52, 2289 (1986). 4. S. A. Safron, W. P. Brug, G. Chern, J. Duan, and J. G. Skofronicic, J. Vac. Sci. and Techn., to be published. 5. 0. Gunnarsson and K. Schonhasuer, Phys. Rev. B££, 2514 (1982). 6. J. Annett and P. Echenique, Phys. Rev. £3_£, 8986 (1987). 7. E. K. Schweitzer and C. T. Rettner. Phys. Rev. Lett. £2. 3085 (1989). 8. R. R. Hancox, Phys. Rev. 42., 864 (1932). 9. G. A. Armand and J. R. Manson, Phys. Rev. Lett. £3_, 1112 (1984). 10. J. H. Weare, J. Chem. Phys. £1, 2900 (1974). 11. J. R. Manson and G. Armand, J. Vac. Sci. Technol. AS, 448 (1987). 12. G. Boato and P. Cantini, Advances in Electronics and Electron Physics £Q, 95 (1983). 13. J. R. Manson, R. H. Ritchie and T. L. Ferrell, Phys. Rev. B21, 1080 (1984). 14. A. A. Maradudin, E. W. Montroll and G. H. Weiss, Theory of Lattice Dynamics in the Harmonic Approximation (Academic Press, Hew York, 1963). 15. J. Skofronick, S. Safron, G. Chern and J. R. Manson, to be published. 16. J. T. Lewis et al., Phys. Rev. 1£1, 877 (1967). 17. The Debye temperature would be expected to be smaller for the surface region, as is found in the case of electron scattering. However, the Debye temperature as measured by He scattering is usually found to be about that of the bulk. The reason for this is usually ascribed to the fact that the He atom, being a large particle, interacts with more than one surface atom. This point is discussed in Ref.(l). Intensity (Arbitrary Unlti) <3 00 0
\
s \
O \ \ s C* c M Ml K D \ n m 3 \ \ O / a s t n \/ tt rt M K K \ o A r* \ / \ V
/ \ 80 0 \ s
60
«
* 40 e 1 -* — * • * 20
800 200 400 600
T (K) Surface Temperature 329
Positronium Formation and Photoemission Spectroscopy for Surface
Akira Ishii Faculty of General Education, Tottori University Koyama, Tottori 680, Japan
Theoretical feasibility study of positronium formation spectroscopy is presented as a complementary spectroscopy to the UV photoemission. The- ory is very similar to photoemission if we recognized that Ps corresponds to a photon, the incident positron corresponds to a time-reversed LEED state. Calculation for Al(100) and Cu(100) showed that positronium formation is quite sensitive to the surface electronic structure. The time-reversed process of positronium formation is also possible to use as a spectroscopy. 330
1. Introduction The attempt to get more complete surface electronic structure is quite im- portant, because it affects results of photoemission analysis. As well-known, determination of band structure of solid is one of the most important prob- lem in physical science. However, especially for valence band electron, the- oretically reproduced photoemission spectrum depends strongly on model of surface barrier. Thus, we should determine the surface barrier to get rig- orous band structure from experiment. Nevertheless, at the moment, it has been believed to be very difficult to obtain information of surface electronic structures around the surface barrier separately from the bulk ingredients. Today, UV photoemission would be the best method for it. However, it is well-known that UV photoemission spectrum is consisted from photoelec- trons which comes from at least several atomic layers in the surface region. Therefore, photoemission spectrum is dominated by features which origi- nate from the bulk band structures. Identification and elimination of them to isolate surface served features are usually very difficult, because no one knows what is the expected solution. Especially for clean surface, it is very difficult to distinguish surface features among bulk features. Since positronium (Ps) is formed only outside of surface for metal and semiconductor, Ps spectrum from surface has been expected to be a very good probe for the surface electronic structure of the topmost layer[L-16j. Although ion neutralization has been also expected to be a good probe for electronic structures of the topmost layer, Ps formation spec'.roscopy (PsFS) is definitely better. The reason is that, by using PsFS, we can measure momentum distribution of electron from the top to the bottom of the band. Recently, STM (scanning tunneling microscope) would be considered as a good probe not only for surface atomic structure but for surface electronic structure. However, since STM is dominated from structure of using tip, the density of state of the tip should be removed from STM spectrumfi?]. Because we could not control atomic structure of the tip in detail, PsFS is better to get detailed information of surface electronic state. However, previous works on PsFS are not enough to get the advantage of this spectroscopy. Though there are some calculation on Ps spectrum formed from thermalized positron, no calculation on Ps spectrum formed from energetic positron beam incident on solid surface. Similarly, though there are a lot of experimental work on thermal positron, very few works 331
are on energetic Ps production by using positron beam[lj. Definitely, there are a lot of advantages on energetic positron beam method. First, we can observe electronic structures not only just below the Fermi surface but also of a whole band. Second, we can easily remove smearing of the spectrum due to the level broadening of Ps near surface[13]. Recently, some intense positron beam lines have begun to work in U.S., Germany and Japanfl]. Therefore, we can expect experimental support on the present work.
2. Theory In this letter, we present the first calculation of Ps energy spectrum by using multiple scattering theory presented by Ishii and Pendry[14], where we have included the separate ingredients known to be important in Ps formation: band structure, surface effects, matrix element, diffraction of the incoming positron wave and higher order corrections. Since the theory and program we used is quite similar to photoemission calculation[19,20], we can compare calculated Ps spectrum directly with the photoemission spectrum. In the recent Green's function formalism of PsFS[14], Ps momentum distribution, /(p), can be written as follows;
(1)
= j J J J dr+dT-dM>+dl>-