CU26), all of the objects in our survey have been observed on 2-4 nights. For the fainter objects, we averaged a small number (typi- Two distinct populations of cally three) of 300-s images; taken together in time, this yields an effectiveexposure time of typically 900 s. Because of the motion of Kuiper-belt objects KBOs, two combined images were created for each group. One group had the individual images registered on the stars, and the s. C. Tegler* & W. Romanishint other group had the images registered on the KBO before combin- *Department of Physics& Astronomy, Northern Arizona University, Flagstaff, ing (averaging) the images. Bycombining the images in this way,the Arizona 86011, USA motion of the KBO from image to image did not contribute to any t Department of Physics& Astronomy, University of Oklahoma, Norman, smearing of the KBO image in the combined image. Oklahoma 73019, USA To maximize the signal-to-noise ratio of the measurements in the 300- or 900-s images, we have applied a well known technique of The discoveryof the first memberof the Kuiper belt1-a formerly stellar photometry, as follows.KBOsare so faint that the uncertainty hypothetical ancient reservoir of objects located beyond Nep- in their instrumental magnitudes is dominated by sky noise. Under tune's orbit-started a revolution in our understanding of the such conditions, it is common in stellar photometry to use a rather outer SolarSystem:thereis no longer asharpedgeat Pluto'sorbit. small aperture when making the instrumental magnitude measure- About 60 Kuiper-belt objects,intermediate in sizebetweencomets ment, in order to minimize the sky noise; a correction is then made and planets, are now known2 to exist on stable circular orbits to the 'total' instrumental magnitude using observations in small around the Sun,andno doubt manymore objectsawaitdiscovery. and large apertures of a brighter comparison star in the same field. But owing to the recentdiscoveryand intrinsic faintnessof these We use this method to get instrumental magnitudes for the KBOs. objects,little hasbeendoneto exploretheir physicalandchemical For each 300-s or 900-s image, we measured the KBO with a small properties3-S.Here we report the results of a two-year survey of aperture (typically 1.8arcsec diameter; 3 pixel radius); we also the broad-bandoptical coloursof aboutone-quarterof the known measured a brighter comparison star with the same aperture, and Kuiper-belt objects.We find that their colours indicate the pre- again with a larger aperture (typically 9.0 arcsec diameter; 15 pixel senceof two distinct populations: one consistsof objectswhose radius) that contained all the light from the comparison star. We surfacecoloursareonly slightly redderthan the colour of the Sun, then applied the magnitude correction from the small and large while the other consistsof the reddestobjectsknown in the Solar apertures of the comparison-star measurement to the KBO System. measurement, yielding the 'total' instrumental magnitude for the Our surveyusesHarris B (450nm), V (550nm) and R (650nm) KBO. For averaged 900-s images, the magnitude corrections were 5-inch X 5-inch glass filters in front of a 1,200 X 800 pixel CCD (charge coupled device) camera at the [/9 Cassegrain focus of the 1.0 Steward Observatory 2.3-m telescope. The CCD has outstanding I quantum efficiency,particularly in the blue where it exceeds 95%. . KBOs 0 Centaurs This high quantum efficiencygreatly aided our blue photometry of 0.8 these faint objects. We used the CCD in a 2 X 2 binning mode, yielding 600 X 400 pixel images, covering 3 X 2 arcmin on the sky i at 0.30 arcsec per pixel. Here we will use "pixel" to refer to the binned 0.3-arcsec pixels. The typical full-width at half-maximum of the stellar point-spread function is 1.5arcsec. Observations of Kuiper-belt objects (KBOs) present a formidable ::r challenge for a photometrist. Their faintness requires us to image I small areas of the sky to faint brightness levels; at such levels there 0.2 are numerous faint stars and galaxies in the images. The Earth's revolution about the Sun results in the motion of KBOs relative to 0.0 the "fixed" background of stars and galaxies in the images 0.4 0.6 0.8 1.0 1.2 1.4 (~3 arcsech-1). Hence, KBOs sometimes pass near background B-V stars and galaxies, and the stellar and galactic light sometimes contaminates the light from KBOs. Finally, some KBOs show brightness variations that are probably the result of the rotation Figure 1 Colour-colour plot of the KBOs (filled squares) and Centaurs (open and irregular shape of the KBOs or of albedo variations over their circles) in our survey. The colour of the Sun is represented by the standard solar surfaces. Observations of these faint, moving, variable objects symbol (0) at B - V = 0.67 and V -R = 0.36. The average colours of C and D require great care. class asteroids'6 are represented by bold letters. KBOs and Centaurs fall into two We have developed an observational and image processing discrete populations: a solar to slightly red population similar to C and D class procedure customized for photometry of KBOS5-7.The faintness asteroids. and an extraordinary red population. We note that the V - R colours for and motion of these objects pushes the optimum exposure time in three additional objects, 1996 RQ20, 1994 JR1 and 7066 Nessus, fall into the two opposite directions. On the one hand, the faintness of the objects populations; see Table 1.The error bars are the uncertainty in the mean, "/~n, pushes us towards exposure times that are aslong aspossible, so that where" is the standard deviation of n 300-s or 900-s exposures. Photon noise, sky we can maximize the signal-to-noise ratio of the data. On the other noise and "unseen" background objects just below our detection limit are hand, the motion ofKBOs pushes us towards short exposure times, sources of random error that contribute to the size of the error bars. We estimate otherwise the light from a KBO is spread over many pixels, thereby that we can detect background objects as faint as 5-10% the brightness of a KBO degrading the signal-to-noise ratio. We have found that taking or Centaur (for most of the objects in our survey). Therefore, "unseen" back- individual 300-s exposures works best. Specifically, we obtain ground objects should affect our photometry by less than 5-10% (0.05-0.10 mag). between two and 20 images per object, per filter, per night, Uns¥en background objects should affect the colours by far less, because the depending on the brightness of the object, and then compute the objects affect the B, V and R'filters in the same sense, and to tend to cancel out in mean and the uncertainty in the mean of the instrumental magni- the colours. As the scatter in our measurements is ~0.05 mag for most objects, tudes. we conclude that the effect of "unseen" background objects on our measure- With two exceptions (the bright objects 5145 Pholus and 1997 ments is negligible «0.05 mag).

NATURE IVOL 39215 MARCH 1998 49 letters to nature determined from the images registered on the stars, and the KBO The only other colour survey with as many objects as ours shows a measurements were made from the images registered on the KBOs. diversity of colours3, rather than two colour populations. We believe The aperture correction is highly dependent on the seeing. Over two that the difference is due to the fact that our photometry is years we have been subject to a variety of seeing conditions and significantly more accurate. Our uncertainties in both the B - V hence the aperture corrections range between ~0.5 and 1.0mag. and V - R measurements are less than, or equal to, 0.05mag for Typically, the correction is ~o. 7 mag. A comparison of several most KBOs (see Table 1); in contrast, the uncertainties in the other bright field stars on a given image indicates that the uncertainty colour surver are typically between 0.1 and 0.2 mag for most KBO in the aperture correction for an image is ~0.01 mag. Aperture measurements in that sample (see Table 5 in ref. 3). As the photometry was done with the PHOT task in the IRAFDIGIPHOT populations in Fig. 1 are separated by 0.3-0.4 mag, uncertainties packagel7.For allphotometry, the skyvalue was found in an annulus of 0.05 mag are essential to detect the discrete populations. The from 20 to 30 pixel radius. The sky value was the peak of a gaussian accuracy of our measurements is certainly due to the fact that (1) we fit to the histogram of pixel values in the sky annulus. If necessary, obtain on average a dozen 300-s images per filter, per object, and (2) the sky annulus was first 'cleaned' of objects which might bias the we measure the brightnesses of KBOs using the central, highest sky measurement by replacing them with a patch of nearby sky. signal-to-noise pixels in the KBO images, and scale to the total Because we typically spent many hours per night looking at a single brightness of the KBOs using bright comparison stars in the same object, we are able to examine the paths of the objects for faint fields. galaxies and stars that would corrupt our photometry. We discarded What is the mechanism responsible for the bimodal colour any KBO images that passed in front of faint background stars or distribution? The lack of any correlation between the colours of galaxies. the objects in Table 1 and their inclination angle, eccentricity, The KBO instrumental magnitudes were corrected to zero air- perihelion distance, semi-major axis, or absolute magnitude, mass using nightly extinction coefficients. We observed three or makes it difficult to point to a specific mechanism in a conclusive four standardstars,twoto threetimesper night,overa rangeof ~ 1 way.What is clear is that the surface evolution ofKBOs is a complex airmass for the derivation of our extinction coefficients. Our problem. Laboratory simulations show that the impact of solar coefficients are consistent with standard values for Kitt Peak. Finally, ultraviolet photons, solar-wind and solar-flare particles, and cosmic KBO instrumental magnitudes were placed on the Kron-Cousins rays on icy surfaces in the outer Solar System alters the chemical system using transformation equations derived from observations composition of those surfaces. For example, data from laboratory of standard stars in the M67, PG0231 + 051, PG1633 + 099 and simulations indicate that bright, hydrocarbon-rich ices initially PG2213- 006 fields9,1O.As a check on our colour transformation redden with radiation exposure, but then become grey with increas- equations and the V zero-point equation, we treated a few ing radiation dosel1,13.In addition, the impact ofparticles ranging in additional standard stars as objects and derived V magnitudes size from micrometeorites to kilometre-sized bodies results in and colours for comparison with published values. The comparison cratering which further alters the chemical composition of the suggests our equations are accurate to the 0.01mag level. surfaces. The results of our survey are unexpected; they are summarized in It is also possible that coma formation may modify the surfaces of Table 1 and in the colour-colour plot of Fig. 1. The reddest objects Centaurs and KBOs. For example, the Centaurs 2060 Chiron and are towards the upper right -hand corner of the figure, and the bluest 5145 Pholus have nearly identical perihelion distances, but the latter objects are towards the lower left-hand corner. Colours for KBOs has an extraordinarily red colour and shows no evidence of coma and Centaurs (recent refugees from the Kuiper beltll on outer planet activity, and the former has a grey (solar) colour and does show crossing orbits) are represented by filled squares and open circles, coma activityl4.The coma activity of 2060 Chiron is not correlated respectively. Clearly, there is evidence for two discrete colour with perihelion distancel5. Hence, something other than solar populations. The population in the upper-right corner of Fig. 1 is heating alone is driving the coma formation of 2060 Chion; perhaps extraordinarily red; the bright Centaur 5145 Pholus is a member of other Centaurs and KBOs will be found to have coma activity. this population. The population in the lower-left corner of Fig. 1 Further observational, theoretical and laboratory work is necessary contains objects with neutral (solar) colours to slightly red colours to explain conclusively the unexpected bimodal colour distribution (similar to C, P and D class asteroids) and includes the bright of ,KBOand Centaur surfaces. 0 Centaur 2060 Chiron. We point out that three objects (1996 RQ20, 1994JR1 and 7066 Nessus) in Table 1 do not appear in Fig. 1,which Received 2 December 1997; accepted 20 January 1998. is becausewehavenot yet measuredtheir B - V colours.Wenote 1. Jewitt, D. & Luu, J. Discovery of the candidate Kuiper belt object 1992 QB1. Nature 362, 730-732 (1993). - the V R colours of these three objects fall in the two populations. 2. Marsden, B. G. List afTransneptunian Objects (1997); available at http://cfa-www.harvard.edu/cfa/ps/ The Sun is plotted for reference and is represented by the standard listsfTNOs.html solar symbol (0) at B - V = 0.67 and V - R = 0.36. 3. Luu, J. & Jewitt, D. Color diversity among the centaurs and kuiper belt objects. Astron. J. ll2, 2310- 2318 (1996). 4. Brown, W. & Luu, J. X. CCD photometry of the centaur 1995 GO. Icarus 126, 218-224 (1997). 5. Romanishin, W., Tegler, S. c., Levine, J. & Butler, N. BVR photometry of centaur objects 1995 GO, Table 1 Colours of KBOsand Centaurs 1993 HA2, and 5145 Pholus. Astron. J. ll3, 1893-1898 (1997). 6. Tegler, s. C. & Romanishin, W. The extraordinary colors of trans-neptunian objects 1994 TB and 1993 Object Class V-R a/vn B-V a/vn Ref. Sc. Icarus 126, 212-217 (1997)...... 7. Tegler, S. C. et al. Photometry of the trans-neptunian object 1993 Sc. Astran. J. ll4, 1230-1233 1997C829 KBO 0.61 0.04 1.08 0.07 (1997). 1996T866 KBO 0.76 0.05 1.02 0.05 8. Weintraub, D. A., Tegler, S. C. & Romanishin, W Visible and near infrared photometry of the centaur 1996T066 KBO 0.70 0.07 1.16 0.10 objects 1995 GO and 5145 Pholus. Icarus 128, 456-463 (1997). 1996TP66 KBO 0.68 0.03 1.17 0.05 9. Landolt, A. U. UBVRI photometric standard stars in the magnituderange 11.5 < V < 16.0 around the 1996T066 KBO 0.38 0.03 0.74 0.04 celestial equator. Astran. J. 104, 340-371 (1992). 1996TL66 KBO 0.35 0.01 075 0.02 10. Richmond, M. W Automated photometry at Leuschner observatory. Int. Amateur-Professional 1996 R020 KBO 044 0.05 Photoelectr. Phatom. Common. 55, 21-30 (1994). 1995 HM5 KBO 0.41 0.12 0.60 0.15 11. Stern, A. & Campins, H. Chironand the centaurs: escapees from the Kuiper belt. Nature 382, 507-510 1994TB KBO 0.68 0.06 1.10 0.15 6 (1996). 1994JR1 KBO 0.36 0.08 12. Andronico, G., Baratta, G. A., Spinella, F. & Strazzulla, G. Optical evolution of laboratory produced 19938C KBO 0.70 0.04 1.27 0.11 6 organics: applications to Phoebe, Iapetus, outer belt asteroids, and cometary nuclei. Astran. Astrophys. 1997CU26 Cen. 0.48 0.01 0.77 0.05 184,333-336 (1987). 1995GO Cen. 0.47 0.04 0.75 0.04 5 13. Thompson, R. W, Murray, B. G. j. P. T., Khare, B. N. & Sagan, C. Coloration and darkening of 1995 DW2 Cen. 0.43 0.05 0.73 0.08 methane clathrate and other ices by charged particle irradiation: applications to the outer solar 7066 Nessus Cen. 0.77 0.05 5 system. J. Geophys. Res. 92, 14933-14947 (1987). 5145Pholus Cen. 0.78 0.04 1.19 0.10 5 14. Bus, S. T.,Schleicher, Bowell, E. & A'Hearn, M. F. Detection of CN emission ITom 2060 Chiron. Science ...... 251,774-777 (1991).

50 NATURE IVOL 39215 MARCH 1998 letters to nature

15. Bus, S. j., A'Hearn, M. E, Bowell, E. & Stern, S. A. 2060 Chiron: evidence for activity near aphelion. Icarus (submitted). 16. Degewij, j. & Van Houten, C. j. in Asteroids (ed. Gehrels, T.) 417-435 (Univ. Arizona Press, Tucson, 1979). 17. Tody, D. in Astronomical Data Analysis Software and Systems II (eds Hanisch, R. j., Brissenden, R. j. V. & Barnes, j.) 173 (ASP Con!. Ser. 52, Tucson, 1993).

Acknowledgements. We thank the NASA Origins of Solar Systems program for support of this work, and the Steward Observatory Telescope Allocation Committee for consistent allocation of telescope time.

Correspondence and requests for materials should be addressed to S.C.T. (e-mail address: [email protected]).

NATURE IVOL 39215 MARCH 1998 51