Stars in the Cellar: StarsCLASSES in LOST the AND Cellar: FOUND BY JAMES B. KALER

38 September 2000 Sky & Telescope For more than a century, the familiar spectral sequence OBAFGKM has stood firm. It is now being extended, testimony to both modern technology and the sequence’s amazing adaptability to the discovery of new stellar types.

the students thundered. Early Days I was teaching one of The modern tale of the stellar sequence my favorite subjects, begins back in the quiet time of 1890s “stellarM spectra,!” to a rather large upper-level Harvard, but it had a long prelude. The introductory class. The task birth of the sequence took place literally was to get them to learn the spectral se- in a ray of sunshine. quence — the basic stellar categories — Break sunlight with a prism or rain- without resorting to tired mnemonics drop and it splashes into an array of col- about kissing fine girls. In a fit of creativity ors from red through violet. Stretch this (or so it seemed), I burst into the class- spectrum carefully, with great color sep- room, and throwing my arms in the air aration, and out pop myriad fine (and like a college cheerleader, I called out: some not-so-fine) gaps in the colors. “Give me an O!” A dribble of “O”s fol- These “spectrum lines,” discovered in 1802 lowed. “Give me a B!” produced somewhat by British chemist William H. Wollaston, better results. As we worked our way are superimposed on the solar spectrum through A, F, G, and K the students got by the actions of atoms, their electron- ROBERT L. HURT into the act and called out louder. By the stripped ions, and by simple molecules The same sunlight that brightens our days end, M, we could have been in the stadi- in the outer layers of the Sun’s bright and warms our landscapes spreads into a um cheering the Illini on to victory. Spec- surface. Each element, ion, or molecule rainbow of colors called a spectrum. Send tral-type M stars, I told the students, lie at has its own unique pattern of light and that light through a fine prism or slit and the end of the sequence, populating the dark lines that form when the electrons grating and lines and gaps appear, creat- bottom of the temperature-pile of stars. bound to atoms absorb radiation at par- ing a sort of solar “fingerprint” (below) in- For a hundred years they’ve been consid- ticular wavelengths. Chemists create these dicating the elements in the Sun’s atmos- ered the coolest stars known. patterns in the lab by igniting pure sub- phere that absorb portions of the When something has endured for a cen- stances and recording the resulting pat- sunlight. Spectroscopy — the study of the tury, you assume it will never change — terns through spectroscopes. Comparison spectra of light-emitting and -reflecting much like a historic building. But if the ed- of laboratory spectra with the spectrum objects in the universe — reveals an ob- ifice is strong, you can surely add to it, and of sunlight allowed solar lines to be iden- ject’s temperature, composition, and ve- — remarkably — that is what has hap- tified. They turned out to be the lines for locity. Astron-omers routinely take spec- pened to the famed spectral sequence. No the common elements of Earth: hydro- tra in all wavelengths of light, including longer is it OBAFGKM (with side branch- gen, carbon, oxygen, iron, and so on. the infrared, where dwarf stars are most es R, N, and S). Now it is OBAFGKM From an absorption line’s strength readily detectable. Above: Cool, low-mass with newcomers L and T. Really! (the amount of energy it extracts from dwarfs range from true stars to Jupiter- What happened? Technology happened. the spectrum) and an application of like objects, and each has between 80 and New observational and analytical tools atomic theory, we can determine the 10 Jupiter masses. This artist’s conception came into use and, as a result, new class- composition of the Sun’s light-emitting of the T dwarf Gliese 570D shows a giant es were found. The story of the expand- surface. Measured by the number of storm, which may cause subtle variations ing spectral sequence is one of fresh dis- atoms, it turns out to be 92 percent hy- in its spectrum as the spot rotates in and covery,with profound implications for drogen. The rest is mostly helium and a

out of view. JERRY SCHAD understanding the galaxy, the stars, and tiny smattering of everything else, led by even the planets. oxygen, carbon, neon, and nitrogen. Take away the light stuff, the hydrogen and helium, and the “everything else” is ap- portioned about as it is in the Earth’s crust, powerful evidence that we and the Sun were both born at about the same time from the same dusty cloud of inter- stellar gas. Our familiar landscapes are Na Hα made from the solar distillate. MAURICE GAVIN

Sky & Telescope September 2000 39 Now go to the stars. What a surprise they gave to the pioneers in spec- troscopy. The first stellar spectra, ob- served in 1817 by the German physicist Joseph Fraunhofer, did not look any- thing like that of the Sun. By the 1860s, stellar rainbows were being studied in bulk by William Huggins in England and Angelo Secchi of the Roman College Ob- servatory. In spite of Huggins’s discovery that the Earth, Sun, and stars contain the same chemicals, the observational evi- dence stood firm: a great many stars did not have Sun-like patterns. Although the Sun is mostly hydrogen, its spectrum is actually dominated by absorptions from sodium and ionized calcium. However, hydrogen absorptions quite overwhelm the spectra of , Altair, and . In P. SABINO MAFFEO/VATICAN OBSERVATORY other stars — Betelgeuse and its reddish Above: Under the direction of Edward C. Pick- kin, for example — hydrogen is effective- ering and (standing), a ly absent, while we see complex bands team of 15 women worked at Harvard Col- produced by such molecules as titanium lege Observatory in the late 1880s classifying oxide. Why the differences? Although all stars according to their spectra. They pored stars are made of the same basic stuff, over the spectra of star fields (like the one of their spectra make it seem as if their ac- Eta Carinae shown here), sorting and classify- tual chemical compositions vary all over ing the stars. Their work formed the basis for the place. How best to understand what’s the Henry Draper Catalogue as well as the sys- going on? tem of star classification in use today. The first step, as in any science, is to classify. But how? In 1863 Secchi invent- Huggins for the honor of making the ed a system that would act as the proto- first permanent recording. (Huggins was type for future developments. It divided the first to photograph a nebula’s spec-

stars into five groups based on similar HARVARD COLLEGE OBSERVATORY trum.) Upon Draper’s death, his widow line patterns and colors. Roman numer- gave his recorded spectra, his telescope, als I through V identified the following and a memorial fund to Harvard. Picker- classes, respectively: blue-white stars with ing took up Draper’s work and built a simple hydrogen spectra (Vega, Sirius), spectrographic telescope of novel design. stars with more complex spectra like the Draper and Huggins had used a dispers- Sun ( and Arcturus), orange- ing prism at the telescope’s focus to view red stars with more complex bands of one spectrum at a time. Pickering, how- lines (Betelgeuse), red stars with different ever, placed the prism in front of the kinds of complex bands (such as 19 Pis- lens, so he could image spectra of all the cium), and finally those containing both stars in a field at once, allowing rapid emissions (bright lines) and absorptions classification by his assistant, Williamina ( and ). Fleming. With all those spectra available, a sim- Classifying at Harvard ple scheme that could handle all the de- Secchi’s scheme was oversimplified, since tail was needed. Beginning in 1890 Pick- stars within any one class could be quite HARVARD COLLEGE OBSERVATORY ering and Fleming expanded Secchi’s different from one another. Help came Top: Angelo Secchi, a Jesuit priest, was also a groups with Roman letters A through O, when Henry Draper, a physician and am- spectroscopist. His early stellar-spectrum based primarily on the strengths of the ateur astronomer in New York, and Ed- classification scheme was used throughout observed hydrogen lines (P and Q were ward C. Pickering, professor of astrono- the late 19th century and was the basis for used for those that did not fit). Further my at Harvard University, focused their the expanded stellar classes developed by observation showed that some classes were attention on the problem of sorting New York astronomer Henry Draper (center) erroneously assigned, unneeded, or could stars. In 1872 Draper photographed and Edward C. Pickering (bottom), who was be merged with others. Two other assis- Vega’s stellar spectrum, just beating out director of the Harvard College Observatory. tants, and Annie Jump

40 September 2000 Sky & Telescope Left: An objective-prism spectrogram of the Eta Carinae region of the southern Milky Way. This 140-minute exposure was made on May 13, 1893, in Arequipa, Peru, using Harvard College Observatory’s 8-inch Bache refractor (below) and is the first plate ever classified by . Nearly every major spectral class is represented.

orange (K), and orange-red (M). Star color is an indicator of stellar tempera- ture. The spectral sequence is the result of temperatures that range from a high of around 50,000° Kelvin for the hottest O stars to about 2,000° K at the end of HARVARD COLLEGE OBSERVATORY M. The sequence responds not to chemical change but to temperature-de- HARVARD COLLEGE OBSERVATORY pendent ionization and ef- disappears. O stars display ficiency of absorption (see absorptions of ionized he- page 42). lium, B stars of neutral he- A new dimension — lu- lium. From A through M minosity — was added in various metal lines (those 1943 with An Atlas of Stel- OWEN GINGERICH from elements heavier lar Spectra by William W. than helium) and their ions change their Morgan, Paul C. Keenan, and Edith Kell- strengths, ionized calcium becoming man (the “MKK” atlas). Stars on the strong through class G then weakening as “main sequence” (or “dwarfs,” ordinary neutral calcium and sodium begin to stars like the Sun) have a wide range of vi- dominate. The M stars (Secchi’s type III) sual luminosity: from a million times that show off the molecules in their atmos- of the Sun for O stars, down to a mil- pheres, chiefly titanium oxide (TiO). Sec- lionth of the Sun’s brightness for stars chi IV, class N,does not fit. Its stars are near M9. To these add the evolved dying red like those of class M, but they harbor stars — those cool giants and supergiants carbon lines rather than TiO. that have ballooned to huge proportions; HARVARD COLLEGE OBSERVATORY The great memorial to Draper and his some of them have diameters approaching Early spectra show the distinctive lines that work was the Henry Draper Catalogue, the size of Saturn’s orbit. Their large size astronomers use to determine the abun- published between 1918 and 1924 by leads to low density and subtle spectral dances of specific chemicals in a star’s outer Cannon and Pickering. It contains Can- changes. All of these differences in stars, atmosphere.This set of spectra from the Henry non’s remarkable solo classification of ranging from main sequence through sub- Draper Catalogue and classified by Annie 225,300 stars, which she later extended giant, giant, bright giant, and supergiant Jump Cannon illustrates the correspondence to 359,082. In spite of all the new de- are labeled V to I (not related to Secchi’s between the Harvard and Secchi schemes. velopments and advances in astronomy, Roman numerals) in the MKK. the original Draper catalog is still con- Cannon, saw that absorption lines other sulted — a testimony to its greatness. The Infrared Cellar than hydrogen fit better from one class Moreover, the most commonly used Astronomy students have always learned to another if B were placed before A and name for a telescopic star brighter than their OBAFGKM and sometimes their O classified before B.The result was the 10th or 11th magnitude is still its “HD RNS. And for years, that was the end of astronomer’s basic spectral alphabet: the number.” it, with no star cooler than M or N. For first three of Secchi’s groups morphing The key to understanding the Draper purposes of our discussion of stars be- into OBAFGKM; his deep red fourth sequence is color. The colors of spectrally yond M, however, the dying, carbon-rich group became class N.Cannon further ordered stars vary smoothly, from blue stars — the R, N, and S types — can be refined the system by decimalizing the (class O) through white (A), yellow (G), removed from consideration. We’ll stick classes, making B9 merge into A0 and so on through M. The classes are distinctive and precise. Astronomy students have always learned Hydrogen first strengthens from O to A, then weakens steadily toward M,where it OBAFGKM and RNS. Now they also have L and T.

Sky & Telescope September 2000 41 Temperature (degrees Kelvin) Temperature (degrees Kelvin) 10,000° 4,500° 3,000° 2,000° 1,500° 4,000° 3,000° 2,000° 1,400° 1,000°

0 OB 2 A Red giants 4 F M 4 K dwarfs A The H-R diagram G I N 6 8 S K E Q M dwarfs U 8 E 12 N C 10 E

Absolute visual magnitude White M dwarfs 16 dwarfs 12 L dwarfs Absolute infrared magnitude 14 20 L dwarfs

16 T dwarfs NEILL REID, UNIVERSITY OF PENNSYLVANIA 021354 6 0 1 2 354 6 Visual color index Infrared color index In the early 1900s two astronomers — Ejnar Hertzsprung in Denmark and Henry Norris Russell in the United States — independently discov- ered the relationship between a star’s temperature and absolute magnitude (a measure of its intrinsic brightness). When stars’ absolute magni- tudes are plotted against their temperatures (or, equivalently, their colors), most of the points in the resulting “H-R diagram” lie along a smooth curve called the main sequence. The two H-R diagrams here — one for visible-light data (left) and one for infrared (right) — show the entire main sequence from the hottest, brightest objects to the coolest, dimmest ones. (On the horizontal axis, color index refers to the difference be- tween an object’s magnitude measured at two different wavelengths.) T dwarfs are so cool and dim they can’t be observed visually. Their tem- peratures start around 1,000° Kelvin but may be as low as 750° Kelvin. It’s All in the Temperature ow can the spectra of stars all made of the same stuff be so bly ionized form takes over, weakening the lines of the singly ionized form. very different? Why are there so many spectral classes? Temper- Molecules behave similarly.They are fragile things that can exist only at low ature, which controls the state of the atoms in a gas, is the temperature, so we see titanium oxide in M stars and hydrides in the L sourceH of the differences. A spectrum is created within a star’s thin, par- stars. Increase the temperature and collisions blow the molecules apart. tially transparent outer layer, from which radiation escapes into space. Only the hardiest ones survive even solar-surface temperatures. The continuous background of color comes from the star’s warmer Superimposed on the ionization process is electron excitation. Think depths, while the cooler, upper gases superimpose the absorptions. of the atom as a stairway. Each atom or ion has a different set of steps, These dark gaps are produced when electrons attached to atoms col- each with different risers. Electrons can be on any of the steps but not lect individual photons of light and jump to higher-energy orbits. Since in between. It takes energy to climb the stairs. As a rule, electrons (like the orbits for any given atom are structured — that is, they involve spe- people) love their lowest-energy states. They can be kicked upstairs by cific energies — the electrons can capture only photons having specific collisions but immediately fall back down. But while on one of the energies, which in turn correspond to specific wavelengths. steps, they can get hit by another photon and jump even higher. The But to absorb a photon, an atom and its electrons must be in the act destroys the photon and produces an absorption line. right state. If they are not, absorptions will not appear, no matter how The visible hydrogen lines come from electrons that are temporarily much of a substance is present. The major factor is ionization. Remove on their second steps. If the temperature is too low, hydrogen sits sulki- an electron from an atom to create an ion with positive charge and you ly on the first step and refuses to accept any optical photons.The result: change the spectrum completely. no hydrogen absorption in cool class M or L, even though the stars are Collisions among the atoms in a gas are primarily responsible for setting almost totally hydrogen. As temperatures climb, more hydrogen atoms the ionization level. At low temperature (even in the Sun) all the hydrogen move onto step two. The absorptions intensify, right through class A is neutral. But around 10,000° Kelvin, where class A meets class B, hydrogen until the beast of ionization begins to eat them all up. Since each kind begins to ionize, and the hydrogen lines must weaken. In cool stars, like of atom has its own electronic structure and stairway, each behaves dif- class M, we see neutral calcium, but increase the temperature into K and ferently. Neutral helium does not appear until class B, and ionized heli- then G, and the calcium ionizes to produce a different spectrum. At higher um requires the great heat of class O. temperatures yet, the calcium is stripped of another electron and the dou- The final result is OBAFGKMLT, all from the same chemical composition.

42 September 2000 Sky & Telescope to main-sequence dwarfs. These stars recycled from unused Draper classes, L8, the last subclass (leaving room for a generate their energy by converting hy- could be confused with other kinds of potential L9), the hydrides are weakened drogen to helium deep in their cores. At objects. “D,” for instance, is part of and the potassium absorption becomes the cool end are M9.5 stars, with masses white-dwarf nomenclature, “Q” is a qua- incredibly strong and broad, dominating around 8 percent that of the Sun. Below sar designation, and so on. Only four the near-infrared spectrum. Comparison the 8 percent limit, the internal tempera- letters remained: H, L, T, and Y. A key with line-formation and solid-precipita- tures are too low for full hydrogen fusion characteristic of the classic sequence tion theory suggests a temperature range to be sustained. (class C excepted) is that the letters do from around 2,000° K at L0 to as low as Theoreticians, however, have long pre- not imply spectral characteristics. Thus, 1,300° K at L8. Total luminosities (as op- dicted that there should be lower-mass “H” might be construed as standing posed to those seen by eye) are compara- bodies, or “substars.” They would not be for “hydrides.” “L” seemed like a logical bly low, ranging downward from 0.0003 running the full fusion reactions seen in choice, and thus the sequence expanded times that of the Sun. hotter stars but would glow dimly from to OBAFGKML. We learn in astronomy class that stars heat generated by gravitational contrac- Spectral observations are difficult be- are supposed to be entirely gaseous. tion and the fusion of deuterium (a cause these red stars are faint. But, as However, in stars cooler than around L2, heavy form of hydrogen). Called “brown with the other classes, L contains a range the temperature becomes so low that ti- dwarfs,” they were avidly sought, but of properties, so it too must be subdivid- tanium actually precipitates into a solid none were confirmed until a few years ed. In cooler M stars, vanadium oxide form, into a mineral called perovskite ago when Gliese 229B (a dim companion (VO) is very strong, and its absorptions (CaTiO3). At various other cool tempera- to the red dwarf Gliese 229A) was dis- reach a maximum at L0. Proceeding tures VO precipitates as a solid, as does covered. Its mass is not yet directly down the sequence, VO disappears at L4, lithium chloride. These and other chemi- known, but 229B has a temperature so and TiO is hardly there at L7. cals produce a grainy stellar fog (and the low that it displays methane absorptions, As the temperature falls further, ab- term “rock star” takes on a whole new a substance no real star can have. How- sorptions of the alkali metals potassium, meaning). ever, one brown dwarf, plus a large num- rubidium, and cesium strengthen. Near Assessment of the content of class L is ber of candidates, hardly compares with the vast number of real M dwarfs out Early Stellar Classifications there. Substars are so cool that to find them Secchi Draper (Harvard) in any number astronomers had to break I Strong hydrogen lines A Strong, broad hydrogen lines from traditional optical methods and observe in the infrared, where low-tem- B Like class A but with the addition of the “Orion lines” perature bodies radiate most of their found in many stars in Orion, later determined to be neutral light. There are several search programs hydrogen lines for such substars, including the Deep C Doubled hydrogen lines Near Infrared Survey of the Southern Sky (DENIS) being done by the Paris In- D Emission lines present stitute of Astrophysics, and the Two Mi- II Numerous metallic lines E Fraunhofer “H” and “K” and the Hβ lines are seen cron All Sky Survey (2MASS) by the University of Massachusetts and Cal- F Similar to class E but with all the hydrogen lines present tech’s Infrared Processing and Analysis G The same as F but with additional lines Center. These have led the way, turning up hosts of faint, dim red “stars” that lie H The same as F but with a drop in intensity in the blue part of off the end of the classic sequence. Their the spectrum spectra look nothing like class M.In- I Like H but with additional lines stead of oxides, these stars are dominat- ed by hydrides (metal atoms with hydro- K Bands visible in spectrum gen attached) and by the raw metals L Peculiar variations of K themselves. III Prominent bands of lines, M Secchi’s third type each of which gets darker Enter L he For the first time in 110 years, as- toward the blue tronomers needed another letter to de- IV Deep red stars N Secchi’s fourth type scribe ordinary stellar dwarfs. Picking a V Bright spectrum lines letter sounds trivial until you try. R and O Spectra with mainly bright lines (Wolf-Rayet stars) (V was later restricted to S had already been added; R and N were P Planetary nebulae O stars by E. C. Pickering) then combined into “C” (for carbon). Most remaining letters, including those Q All other spectra (changed to designate novae in 1922)

Sky & Telescope September 2000 43 Potassium (K) TiO CrH L5Dwarf K I CrH FeH Cs I H2O FeH

M8 V

K L0 V

K I Cs I H2O FeH T Dwarf K L2 V

K L4 V Right: Spectra of L and T dwarfs show characteristic absorption fea- tures. L dwarfs have oxide molecules such as titanium oxide (TiO) — K

Relative intensity L6 V which also appear in M dwarfs. These oxides are weak in T dwarf spec-

tra. Cesium (Cs I) is seen throughout L and T, and iron hydride (FeH) is K strong, while chromium hydride (CrH) weakens. Potassium broadens L8 V

markedly through L. Water (H2O) appears in T dwarfs. Above: These T dwarf synthetic spectra simulate the traditional objective-prism view of the T dwarf FeH absorption bands of an L5 and a T dwarf between 400 and 1,000 Cs I H2O J. DAVY KIRKPATRICK / 2MASS nanometers. Below: The Low Resolution Imaging Spectrometer (LRIS) 600 700 800 900 1,000 on the Keck telescope was used to gather spectra of L and T dwarfs. Wavelength (nanometers)

stars the term “T dwarf” is now catching ly comets Earthward. And this is only on, along with “L dwarf.”Observations one way that these new additions to the are difficult, and there are insufficient spectral sequence might reach out and data to warrant subdivision of this class. touch us. But at the end, here are OBAFGKMLT. Where will these little bodies reach their end? We have no idea of the lower So What? limit to substellar masses. Brown dwarfs It is not enough just to know of the exis- are, like ordinary stars, presumably made tence of the L and T dwarfs. How do “in place” by collapsing interstellar these new classes affect what we know of clouds. Planets, on the other hand, are stars and about the galaxy? The data are built up from solids in dusty disks that sparse, but that has never stopped as- circulate around new stars. Perhaps the tronomers from drawing conclusions, masses of brown dwarfs and planets can sometimes even correct ones. actually overlap each other! There may, For example, we now may be able to therefore, still be new spectral classes to put to rest a long-standing speculation develop as observations descend the tem- difficult. The coolest stars should be thor- about the composition of the mysterious perature scale. If so, the old Harvard oughly mixed by convection. Above the dark matter that dominates the galaxy. M classification system, the century-old cre- “real star” mass limit, the internal temper- dwarfs are numerous and constitute 70 ation of Pickering, Fleming, Cannon, ature is so high that an old low-mass star’s percent of all stars along the main se- and Maury, can keep up with it as we natural lithium is destroyed by nuclear re- quence from M to O.The L and T star head into the new millennium. Keep actions. Thus, unless it is very young, a count, sparse as it is, suggests that there your eye on letter “Y.” star must be a brown dwarf if it has lithi- are twice as many brown dwarfs as there Back to my cheering astronomy class. um absorptions. About a third of the L are “real stars.” Nevertheless, their com- My then 13-year-old daughter, who had stars qualify as brown dwarfs. The low lu- bined mass is so low that it makes no a day off, was visiting to see Dad at minosities and temperatures of the L significant impact on the dark-matter work.After the cheers were over, a dwarfs, which indicate very low mass, con- problem. So we must look elsewhere for young man sitting next to her asked, firm that estimate. This suggests that the dark matter. “How would you like to have to live with class contains a mixture of real stars and The sheer number and faintness of that guy?” Imagine how annoyed he’d brown dwarfs. brown dwarfs raise the question of what been if I’d given them L and T to cheer As rich and useful as it is, class L still is really the nearest “star.” A brown about. cannot do all the work. Methane-rich dwarf could easily be hiding closer than Gliese 229B has a temperature of only Proxima Centauri, an M5 dwarf long James B. Kaler is professor of astronomy at 1,000° K. touted — at 4 light-years — as being the the University of Illinois. His latest book is Cos- The Sloan Digital Sky Survey and closest star to us. Knowing is more than mic Clouds, written for Scientific American Li- 2MASS have found many more such an academic exercise. A dim L or T star brary. He outlines a “star of the week” on his stars, recognizable by their odd colors might even now be stirring up the Oort Web site at http://www.astro.uiuc.edu/~kaler/ and horribly complex spectra. For these comet cloud and sending a rain of dead- sow/sow.html.

44 September 2000 Sky & Telescope