Surface Ices and the Atmospheric Composition of Pluto
Tobias C. Owen; Ted L. Roush; Dale P. Cruikshank; James L. Elliot; Leslie A. Young; Catherine
de Bergh; Bernard Schmitt; Thomas R. Geballe; Robert H. Brown; Mary Jane Bartholomew
Science, New Series, Vol. 261, No. 5122. (Aug. 6, 1993), pp. 745-748.
Science is currently published by American Association for the Advancement of Science.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For
more information regarding JSTOR, please contact email@example.com.
Tue May 22 10:21:36 2007
5. CGS4 is a cryogenic grating spectrometercontain-
ing a two-dimensionalarray of lnSb photodiodes,
usable in the 1- to 5-pn wavelength band at re-
solving powers between 200 and 20,000. UKIRT is
operated by the Royal Observatory Edinburgh on
behalf of the United Kingdom Science and Engi-
neering Research Council. Triton spectra were
obtained on 29 June 1991 (comparisons stars BS
7340 and BS 6965), 18 October 1991 (comparison
star BS 7340),5 May 1992 (poor data) and 27 and
28 May 1992 (comparison stars BS 6998 and BS
7504) (presented here in Fig. 1). Additional data
showing the N , band at higher resolution (from
which the wavelengths given in the text are de-
rived) were obtained on 22, 23, and 24 September
1992.All dates in universal time.
6. B. Hapke, J. Geophys. Res. 86, 3039 (1981);
lcarus 59, 41 (1984); lcarus 67, 264 (1986).
7. T. L. Roush, J. B. Pollack, F. C. Witteborn, J. D.
Bregman, J. P. Simpson, lcarus 86, 355 (1990).
8. B. Schmitt, E. Quirico, E. Lellouch, Proc. Symp.
Titan, ESA SP-338, 383 (1992);J. R. Green, R. H.
Brown, D. P. Cruikshank, Bull. Am. Astron. Soc.
23, 1208 (1991); G. B. Hansen, Appl. Opt. 25,
2650 (1986); S. G. Warren, Bull. Am. Astron. Soc.
24, 978 (1992). In our modeling we derive the
radiance factor at zero degrees phase angle (i=
e = 0 degrees). We assume isotropic surface
scattering, a lunar-likeregolith parameterfor Hap-
ke's h-value, and we make no corrections for
macroscopic surface roughness. The continuum
is defined as a series of straight line segments
connecting local maxima such that data between
maximaare not cut by the continuum. This has the
effect of eliminating absolute albedo information.
However, preliminary calculations of geometric
albedos for these mixtures are entirely consistent
with the values derived from the telescopic data.
We shifted the optical constant values to agree
with the 0b~e~ati0nal
data (1992 smoothed data)
prior to calculation of the spectra.
9. B. Schmitt etal., Bull. Am. Astron. Soc. 22, 1121
(1990) (abstract). The CH, fraction has been
changed since publication of this abstract and is
now "between 0.05 and 0.5%." A more extensive
paper has been submitted to Icarus.
10. A. L. Broadfoot et a/., in (2);G. L. Tyler et al.,
Science 246, 1466 (1989).
11. L. Trafton, lcarus 58, 312 (1984)
12. The wavelength calibration of the CGS4 spec-
trometer is determined from emission-line lamps
and is accurate to 0.0001 pm (one sigma).
13. The presence of a-N, is unexpected on Triton
because the surface temperature (if it is in fact in
strict vapor pressure equilibrium with the atmo-
sphere at P , = 16 pbar) is 37.7 K. The uniform
temperature predicted by Trafton (11) further
argues against coexistence of surface exposures
of the a and p phases. In addition, although the
phase transition temperature in N , is raised when
the nitrogen is contaminated with CO or CH, [T. A.
Scott, Phys. Rep. 27 (no. 3), 89 (1976)], the
concentrations of the contaminantsimplied by our
spectra are too low to raise the transition temper-
ature at least the 2 K required. Nevertheless,
physica1,studies of solid N ,
CH, and near the 35.6 K phase transitiontemper-
ature show that the N , in close proximity to a CH,
molecule is in the a phase, while that further away
from a pontaminant molecule is in the p phase.
Thus, a mixed a-p spectrum may result.
14. J. Eluszkiewicz, J. Geophys. Res. 96, 19,219
15. T. C. Owen et a/., Int Astron. Union Circ. 5532
16. F. Herbert and B. R. Sandel, J. Geophys. Res. 96,
17. The shifting of the bands of CH, diluted in N , is
not related simply to the concentration. In partic-
ular, the band at 2.324-pm shifts slightly and then
splits as the CH, concentration is reduced; even-
tually the "pure" component diminishes as the
2.311-pm component dominates.
18. S. A. Sandford, L. J. Allamandola, A. G. G. M.
Tielens, G. J. Valero,Astrophys.J.329,498 (1988).
19. Transmission spectra of molecular mixtures of CO
in N , show that the matrix shift of the CO overtone
band at 2.351 pm is approximately -0.0006 pm,
or about 1 cm-'. This resultwas confirmed by D.
Hudgins at Ames Research Center.
20. G. N. Brown, Jr., and W. T. Ziegler, Adv Ciyog.,
Eng. 25, 662 (1979)
21. M. H. Stevens et al., Geophys. Res. Lett. 19,
669 (1992); V. A. Krasnopolsky et al., J. Geo-
phys. Res. 98, 3065 (1993).
22. We thank Douglas Hudgins for his laboratory work
on the N , + CO spectrum. We also thank the staff
of the United Kingdom InfraredTelescope Facility
for their support during the observations reported
12 January 1993; accepted 16 April 1993
Surface Ices and the Atmospheric
Composition of Pluto
Tobias C. Owen,* Ted L. Roush, Dale P. Cruikshank,
James L. Elliot, Leslie A. Young, Catherine de Bergh,
Bernard Schmitt, Thomas R. Geballe, Robert H. Brown,
Mary Jane Bartholomew
Observationsof the 1.4-to 2.4-micrometer spectrumof Pluto revealabsorptions of carbon
monoxide and nitrogen ices and confirm the presence of solid methane. Frozen nitrogen
is more abundant than the other two ices by a factor of about 50; gaseous nitrogen must
therefore bethe major atmosphericconstituent.The absence of carbondioxideabsorptions
is one of several differences between the spectra of Pluto and Triton in this region. Both
worlds carry information about the composition of the solar nebula and the processes by
which icy planetesimals formed.
Although Pluto is usually classified as a
planet, its closest relative in the solar sys-
tem appears to be Triton, Neptune's largest
satellite. Both of these obiects evidentlv
formed from the solar nebula at a distance
of -40 astronomical units (AU) from the
sun, where temperatures were <50 K. In
this respect, they may.be considered huge
icy planetesimals that somehow escaped
accretion by the giant planets. The lower
end of the mass distribution of such obiects
is represented by the common comets, ob-
jects 2060 Chiron, 5145 Pholus, 1992
QB1, and the great comet of 1729 (1).
These objects represent an especially prim-
itive stage in the transition from the grains
and gas of interstellar clouds to the la nets
and satellites of the solar system.
We have presented observations of the
near-infrared spectrum of Triton (2). We
used the same instrumental configuration to
study Pluto, without the benefit of Voyager
data to provide a context for our work. It is
the only planet not yet visited by space-
craft, but its recent occultation of a star (3)
and the mutual eclipses and occultations
exhibited by Pluto and its synchronously
orbiting satellite Charon (4) have helped to
define this distant system (5).
Using the cooled grating array spectrom-
eter with the United Kingdom Infrared
Telescope, we recorded Pluto's spectrum
from 1.4 to 2.4 pm (4160 to 7140 cm-l) at
a resolution of 350 on 27 and 28 May 1992
(UT) (6). Previous observations at lower
resolution had established that there is solid
CH, on Pluto's surface (7). Our spectra
conkrmed the presence of this ice, reveal-
ing the same series of strong CH4 bands
seen on T~~
ln addition, the (2,o) band
of CO at 1-35 km and the N, absorption at
2.15 um were added, both of which are also
presentin ~~spectrum (2) ( ~ i ~ .
Des~itethese general similarities, the
spectra of Pluto and Triton are not identi-
cal. we do not find the solid CO, absorp-
tions on Pluto that are so prominent on
Triton. In particular, the absence of the
strong triad bf CO, bands near 2.0 pm (2)
means that the amount of this ice on Pluto
must be less than one-third of the amount
that forms the spectrum of Triton. The
shapes of the CH4bands in the region from
2.0 to 2.4 pm are broader and deeper on
Pluto, whose spectrum exhibits an addi-
tional CH4 feature at 1.48 pm that is not
T. C. Owen, Institute for Astronomy, University of
Hawaii, 2680 Woodlawn Drive, HO~OIUIU,HI 96822.
T. L. Roush, Department of Geosciences, San Fran-
cisco State University,San Francisco, CA 94132, and
NASA Ames Research Center, Space Sciences Divi-
sion, Moffett Field, CA 94035-1000.
D. P. Cruikshank, NASA Ames Research Center,
Space Sciences Division, Moffett Field, CA 94035-
8 " " " .
J. L. Elliot and L. A. Young, Department of Earth,
Atmospheric and Planetary Sciences, Massachusetts
lnstitute of Technology, Cambridge, MA 02139.
C, de Bergh, Obse~atoirede Paris, 92195 Meudon
B. Schmitt, Laboratoire de Glaciologie et Geophy-
sique de I'Environnement,38402 St. Martin d'Heres,
T. R. Geballe, Joint Astronomy Centre, Hilo, HI 96720.
R. H. Brown,Jet PropulsionLaboratory, Pasadena,CA
M. J. Bartholomew, Sterling Software, Inc., NASA
Ames Research Center, Moffett Field, CA 94035-
*To whom correspondence should be addressed.
6 AUGUST 1993
present on Triton (Fig. 2). Another CH,
band of comparable strength in the labora-
tory (at 1.68 pm) does not appear in the
spectrum of either Pluto or Triton. This
difference in appearance of the CH, absorp-
tions must carry some information'about a
difference in temperature, composition, or
average grain size on the surfaces of Pluto
and Triton. Unfortunately, the laboratory
data are not yet adequate to determine this
The N, feature at 2.15 pm is distinctly
weaker on Pluto than on Triton; there is a
hint of an accompanying absorption at 2.16
pm (Fig. 1). The association of this feature
with N, is confirmed by the fact that its
strength has the same proportion to the
primary 2.15-pm N, absorption as in the
spectrumof Triton (2). Further study of this
band should help to place better constraints
on the poorly defined surface temperature of
Pluto (8) because the relative intensity of
the 2.16-pm feature is strongly tempera-
At the scale of the entrance slit of the
spectrometer, Pluto and its satellite Charon
are not resolved into individual obiects. so
the resulting spectrum is a composite of the
contributions of each. The spectrum of
Charon is known to exhibit absorption
from H,O ice (10, 11). At a given wave-
length, Charon's contribution to the com-
bined spectrum with Pluto is weighted by
the relative areas of the two bodies and by
the geometric albedos of their surfaces.
From the mutual eclipses and transits of
Pluto and Charon observed in the 1980s, we
ado~t the radii of Pluto and Charon as 1142
and596 km, respectively. The area ratio of
Pluto to Charon is then 3.67, with Charon's
contribution as 27% of the total (12).
We have modeled the Pluto spectrum
Fig. 1. Smoothed spectra o f Triton (heavy
trace) and Pluto (light trace) are shown in the
region from 2.1 to 2.4 km The dashed line in
the Pluto spectrum indicates the continuum in
the region o f N , absorption. Absorption from
ices o f N , (2.15km),CH, (2.2,2.32, and 2.38
km), and CO (2.35km) are apparent.
using the same basic approach we applied to
Triton (2, 13). The model consists of an
intimate mixture (a salt-and-pepper config-
uration) of N,, CH4, and CO ices to
represent Pluto, plus a spatially segregated
area of H,O ice representing Charon's con-
tribution to the spectrum (Fig. 2). The
H,O spectrum used for Charon's contribu-
tion was computed separately on the basis
of scattering theory with the use of the
Charon spectrum of Buie and colleagues
(I I) and optical constants of the pure ices
(14). Absorption bands of H20 ice in the
wavelength region of interest are broad (0.1
to 0.3 pm), and in the model they princi-
pally affect the shape of the continuum on
which the narrower bands of CH,, N,, and
CO are superimposed. In the present case,
the H20 ice that is included as a spatially
segregated component in the model spec-
trum most strongly affects the regions from
1.45 to 1.55 pm and from 2.10 to 2.30 pm.
The effect of the H20ice is not evident in
the observed spectrum of Pluto plus
Charon; our calculated spectrum is too low
at 1.5 pm and too high at the longer
As in the case of Triton, we find that N,
is the most abundant surface ice on Pluto.
The abundances of CH, and CO are greater
on Pluto than on Triton but are still small
relative to that of N,. As in the case of
Triton, Pluto's CH, bands are shifted from
their laboratory-measured wavelengths,
demonstrating that CH4 and N2 must be
mixed at the molecular level to form a solid
solution. A checkerboard or salt-and-pep-
per model for the surface distribution of the
two ices does not help to explain the
observed wavelengths of the CH4 bands.
Thus, the intimate mixture (salt-and-pep-
per) model we have used (Fig. 2) is not
correct but is the best we can do with
available laboratory data for the optical
constants of these ices (15). As noted in
our discussion of Triton's spectrum (2), the
shifting of the CH, bands is a complex
phenomenon related to the concentration
in the solvent (N,) that affects not only the
central wavelength but the shape of the
bands as well. Furthermore, the various
CH, bands in this spectral region behave
differently from one another (16). In the
case of our model of Pluto's spectrum, the
derived abundance of CH, (1.5%) is con-
sistent with the degree of shift of the bands.
This consistency means that CH4ice is more
abundant on the surface of Pluto than on
Triton, where it is -0.05% (2). The exact
abundance will be more reliably derived
from models that incorporate optical con-
stants of true molecular mixes of N, and
CH, when they become available. The pre-
sent uncertainty in the CH, abundance as
determined from our models is a factor of 2.
We modeled the CO in Pluto's spectrum
using only the (2,O) band at 2.352 pm,
although the weaker (3,O) band at 1.578
pm may also be marginally present on the
short wavelength slope of a CH4 band. As
in the case of Triton, we cannot establish
whether the CO is present as individual
grains or dissolved in a molecular mixture
with N, and CH,, because the wavelength
shift of CO that results from matrix effects
is below the resolution limit of our Pluto
spectrum. The similarities in the vapor
pressures of CO and N, suggest that these
ices are mixed at the molecular level, but
there is no independent information from
the planet's atmosphere or surface to help
resolve this question. If CO is present as
individual grains of ice, the particle sizes
cannot be extremely small because of the
quenching effect that small grains would
have on the strength of the N, band. In our
best fitting model, the abundance and grain
size of CO are 0.5% and 0.5 mm, respec-
tively, again with a factor of 2 uncertainty.
There are several conclusions we can
draw from this coupled set of observations
of these two frigid worlds. The first involves
Pluto's atmosphere. If we assume that CH4,
CO. and N, are the onlv volatiles on Pluto.
that the ices are in ideal solutions, and that
the atmospheric composition is determined
simply by vapor-ice equilibrium, then the
partial pressures are Pi = Xi.Vi, where Xi is
Fig. 2. A comparison o f a contin-
uum-adjusted spectrum o f Pluto
and a model spectrum based on
the indicated mixture o f ices over
the wavelength range o f 1.4to 2.4
km. This model is an intimate
mixture o f the components rather
than the expected molecular mix-
VOL. 261 6 AUGUST 1993
the mole fraction on the surface (0.026,
0.005, and 0.969 for CH4, CO, and N,,
respectively) and Vi is the vapor pressure of
the ith gas over its pure ice (17). Thus,
using the surface temperature, the observed
mixine ratios in the solid. and the known
vapor pressures (18) we can find the partial
pressure and the atmospheric mixing ratio qi
for each gas. Unfortunately, Pluto's surface
temperature is not well determined: pub-
lished values range from 31 K to 59 K,
depending on the observations used and the
different assumptions of the surface proper-
ties, especially the solar phase integral and
emissivity (8). We have constructed Table
1 for three representative temperatures. For
each temperature and volatile there is a
tabulated vapor pressure over pure ice, par-
tial pressure in the atmosphere, and mixing
ratio in the atmosphere.
We can tightly constrain the possibili-
ties for the surface temperature and pressure
by combining our inferences about the at-
mospheric composition with the constraints
imposed by the 1988 stellar occultation.
The data in Table 1 demonstrate that for all
temperatures, the atmosphere is >99% N,,
close to the 100% N, atmosphere consid-
ered by Elliot and Young (19). The absence
of near-surface haze would suggest a surface
radius of 1206 + 11 km, a surface temper-
ature of 35.3 + 0.4 K, and a surface
pressure of 3.3 + 0.8 pbar. This tempera-
ture is close to 35.62 K, the triple point of
vapor, alpha, and beta N,. The presence of
a near-surface haze would decrease the sur-
face radius, with a corresponding increase
in surface temperature and pressure.
Although our discovery of N, and CO
substantiates the suggestion by Yelle and
Lunine (20) that a gas heavier than CH4 is
present in Pluto's atmosphere, the amount
of CH, gas that we infer is less than the
0.1% apparently required by their model to
maintain the lower atmosphere at 106 K.
However. their model has alreadv survived
some observational tests. It predicts that
Table 1. Vapor pressures, partial pressures,
and mixing ratios for gases in Pluto's atmo-
the atmospheric temperature profile should
be isothermal with altitude, which has been
established from the Kuiper Airborne Ob-
servatory occultation data (19) for a zone
that begins -10 km above the surface (or
haze layer) and extends upward for several
scale heights (21). At those altitudes, un-
der the assumption of 100% N, (19), the
atmospheric temperature is 104 +- 21 K.
The agreement of this observationally de-
rived value with Yelle and Lunine's predic-
tion may mean that even less CH4 than
they proposed is needed to raise the tem-
perature or that there is another way of
increasing the CH4 vapor pressure in the
atmosphere than by sublimation from sur-
face frosts. An atmospheric heating source
other than CH, (such as aerosols) may also
be at work.
The role of CH, heating is significant to
an understanding of the difference between
the atmospheric models for Pluto and Tri-
ton, which results from the amount of CH,
in each atmosphere. For Triton, the
amount of CH4 (as determined by Voyager
observations) is too low for radiative Dro-
cesses to affect the energy balance signifi-
cantly (22). However, in the current model
for Pluto radiative absorption and reemis-
sion by CH4set the energy balance at 106 K
(20). A test of the CH, atmospheric abun-
dance that we have inferred from the sur-
face ice spectra will come from an observa-
tional study of CH, gas absorption in Plu-
As on Triton: ~ l u t o shows no evidence
of other ices; we found the same list of
absent ices on that planet. In one respect,
this absence is even more puzzling than for
Triton because Pluto is darker and exhibits
a 0.3-magnitude (-30%)
brightness as it rotates, indicating an inho-
mogeneous surface. If the general similarity
of Pluto's sDectrum to that of Triton is at all
diagnostic of surface conditions on these
two bodies, we expect thicker, more wide-
spread deposits on Pluto of dark material
similar to the isolated patches the Voyager
cameras revealed on Triton. This exoecta-
tion carries with it the idea that intermedi-
ate products in the chemical reactions lead-
ing from CH, and N, to this dark material
should be Dresent. Evidentlv their concen-
tration is extremely low.
The relatively high abundance of molec-
ular N, on Pluto and Triton supports the
widely held hypothesis that the missing
(that is, unobserved) 70% of the cosmic
abundance of N, expected in the interstel-
lar medium (ISM) is in the form of N, (24).
The low abundance of N, in Halley's Com-
et (2.5) could then be understood to result
from sublimation and desorption of an orig-
inal endowment of N, (26). The low rela-
tive abundance of CO on Pluto and Triton
poses a problem, however. From ISM abun-
dances, one expects the CON, ratio to be
-1. Evidently some combination of the
partial processing of CO to CH4 and CO,
during accretion [following pathways dem-
onstrated in the laboratory (27)] with sub-
sequent layering on the surface according to
vapor pressure led to the presently observed
state. From this view, the absence of ob-
servable CO, on Pluto, in contrast to Tri-
ton, resulted from the current difference in
the general circulations of the atmospheres
of these two bodies, corresponding to the
different inclinations of their rotational
axes. On Triton, the flight of N, from the
sunlit southern hemisphere could have ex-
posed underlying CO, to view. Observa-
tions over the next few decades could easilv
test this model if the circulation pattern on
Pluto changes sufficiently before the atmo-
sphere simply freezes out.
Finally, the amount of neon in the
atmospheres of these objects is important
because the cosmic abundance of this ele-
ment is about equal to that of N, (28) and
its vapor pressure is 5 x lo6 the pressure of
neon at 35 K. Neon should therefore dom-
inate these atmospheres, although on Tri-
ton this is clearly not the case (29). If neon
were so abundant, there would be a lower
mean molecular weight on Pluto from oc-
cultation observations than the value cor-
responding to 99% N, (Table 1) (30). If
Triton and Pluto formed at temDeratures
above 20 K, neon would not have been
trapped in the constituent water ice (31)
and would therefore not be expected in
their two atmospheres.
We now only have observations of one
hemisphere of this variegated planet. Fu-
ture investigations may reveal some spectral
signature that could help to identify the
dark material on both of these objects.
Laboratory studies of the CO, CH4, and N,
ices to develop better matches to the spec-
tra should provide better insight into sur-
1. The data are from W. K. Hartman, D. Tholen, K.
Meech, and D. P. Cruikshank [Icarus83, 1 (1990)l
for 2060 Chiron; €3. E. Mueller, D. J. Tholen, W. K.
Hartmann, and D. P. Cruikshank [ibid. 97, 150
(1992)l for 5145 Pholus; and D. Jewitt and J. Luu
[Nature 362, 730 (1993)] for 1992 QB1. These
three objects have diameters on the order of 200
km. The great comet of 1729 achieved naked-eye
visibility although it never came closer to the sun
than 4.0 AU [R.A. Lyttleton, The Cornetsand Their
Origin (Cambridge Univ. Press, Cambridge,
1953), p. 52; B. G. Marsden, Catalogue of
Cornetary Orbits (Minor Planet Center, Smithson-
ian Astrophysics Obse~atoly,Cambridge, MA,
1986), ed. 5, p. 10.
2. D. P. Cruikshank et al., Science 261, 742 (1993).
3. W. €3. Hubbard, D. M. Hunten, S. W. Dieter, K. M.
Hill, R. D. Watson, Nature 336, 452 (1988); J. L.
Elliot et al., lcarus 77, 148 (1989).
4. M. W. Buie and D. J. Tholen, lcarus 79, 23 (1989).
5. S. A. Stern,Annu. Rev. Astron. Astrophys. 30, 185
SCIENCE VOL. 261
6 AUGUST 1993
6. Observationswere as follows: 27 May 1992 (UT), Download full-text
556 to 7:34 and 8:07to 9:31 (UT), Charon phase
0.096;28 May 1992 (UT),6:25 to 8:08 and 9:00to
9:20 (UT), Charon phase 0.255, where Charon
phase 0 is near northern elongation and Charon
phase 0.25 is near eastern elongation. The obser-
vations on 27 May had better signal to noise and
received greater weight in the analysis.
7. D. P. Cruikshank, C. B. Pilcher, D. Morrison,
Science 194, 835 (1976).
8. W. J. Altenhoff et ab, Astron. Astrophys. 190, 15
(1988): H. H. Aumann and R. G. Walkter, Astron.
J. 94, 1088 (1987), M. V. Sykes, R. M. Cutri, L. A.
Lebofsky, R. P. Binzel, Science 237, 1336 (1987);
E. F. Tedesco et al., Nature 327, 127 (1987).
9. K. A. Tryka, R. H. Brown, V.Anicich, D. P. Cruik-
shank, T. C. Owen, Science 261, 751 (1993); W.
Grundy, B. Schmitt, E. Quirico, Icarus, in press.
10. R. L. Marcialis, G. H. Rieke, L. A. Lebofsky,
Science 237, 1349 (1987)
11. M. W. Buie et al., Nature 329, 522 (1987).
12. D. J. Tholen and M. W. Buie, Astron. J. 96, 1977
13. Our models were calculated with the use of scat-
tering theory developed by B. Hapke [J. Geo-
phys. Res. 86,3039 (1981); lcarus 59,41 (1984);
ibid. 67, 264 (1986)]. In all of our calculations we
derive the radiance factor at a phase angle of 0" (i
= e = 0"). We assume isotropic surface scatter-
ing, a lunar-like regolith parameter for Hapke's h
value, and make no corrections for macroscopic
surface roughness.The continuum is defined as a
series of straight line segments connecting local
maxima such that data between maxima are not
cut by the continuum. This definition eliminates
absolute albedo information. However, prelimi-
nary calculations of geometric albedos for these
mixtures are entirely consistent with the values
derived from the telescopic data. We shifted the
wavelengths of the optical constants to agree with
the wavelength values for our observed Pluto
spectrum before calculation of the spectra.
14. B. Schmitt, E. Quirico, E. Lellouch, Proceedingsof
the Symposium on Titan, Toulouse, France, 15 to
18 September 1991, publ. no. SP-338 (European
Space Agency, Paris, 1992), p. 383; J. R. Green,
R. H. Brown, D. P. Cruikshank, Bull. Am. Astron.
Soc. 23, 1208 (1991); G. B. Hansen, Appl. Opt.
25, 2650 (1986); S. G. Warren, ibid. 23, 1206
(1984); Bull. Am. Astron. Soc. 24, 978 (1992).
15. The noisiness of the data in the short-wavelength
wing of the 2.3-pm CH, bands (Fig.2) exemplifies
the inadequacy of the available optical constants.
This effect does not appear in our model for the
Triton spectrum (2) because there is less CH, in
Triton's surface ices than in those of Pluto.
16. The shifting of the bands of CH, diluted in N , is
related not only to the concentration. In particular,
the band at 2.34 bm shifts slightly and then splits
as the CH, concentration is reduced. ~ventuall~
the component of the band produced by pure
methane diminishes in intensity as the 2.311-pm
component resulting from solution in N ,
nates the spectrum.
17. L. Trafton, Astrophys. J. 359, 512 (1990). Eventu-
ally it should be possible to use the equation for
nonideal solid solutions of N ,
have the form P, = K.Y.V(,where K is a constant
2 1; no value for K is now available.
18. G. N. Brown, Jr., and W. T. Ziegler, Adv. Cyog.
Eng. 25, 662 (1980).
19. J. L. Elliot and L. A. Young, Astron. J. 103, 991
20. R. V. Yelle and J. I. Lunine, Nature 339, 228
21. Pluto's pressure scale height is 55.7 & 4.5 km
22. R. V. Yelle, J. I. Lunine, D. M. Hunten, lcarus 89,
23. J. L. Elliot et a/.,in preparation.
24. W. M, lrvine and R. F. Knacke, in Origin and
Evolutionof Planetaryand SatelliteAtmospheres,
S. K. Atreya, J. B. Pollack, M. S. Matthews, Eds.
(Univ. of Arizona Press, Tucson, 1989), pp. 3-34.
25. J. Geiss, Astron. Astrophys. 187, 189 (1987).
and CH,, which
Rev. Mod. Astron. 1, 1 (1988).
26. T. Owen, Astrophys. Space Sci., in press.
27. A. Bar-Nun and S. Chang, J. Geophys. Res. 88,
28. E. Anders and N. Grevesse, Geochim. Cosmo-
chim. Acta. 33, 197 (1989).
29. A. L. Broadfoot et a/., Science 246, 1459 (1989);
G. L. Tyler et a/, ibid., p. 1466.
30. The partial pressure of any additional species
such as neon would have contributed to the
occultation determination of the refractivity scale
height (19). This additional pressure would have
forced us to lower our estimate of the surface
temperature below that permitted by observations
in order to reduce the pressure of N , to accom-
modate the new species.
31. D. Laufer, E. Kochavi, A. Bar-Nun, Phys. Rev. B
36, 9219 (1987).
8 March 1993; accepted 2 June 1993
The Phase Composition of Triton's Polar Caps
N. S. Duxbury and R. H. Brown
Triton's polar caps are modeled as permanent nitrogen deposits hundreds of meters thick.
Complex temperature variations on Triton's surface induce reversible transitions between
the cubic and hexagonal phases of solid nitrogen, often with two coexisting propagating
transition fronts. Subsurface temperature distributions are calculated using a two-dimen-
sional thermal modelwith phase changes. The phase changes fracture the upper nitrogen
layer, increasing its reflectivity and thus offering an explanation for the surprisingly high
southern polar cap albedo (approximately0.8) seen duringthe Voyager 2 flyby. The model
has other implications for the phase transition phenomena on Triton, such as a plausible
mechanism for the origin of geyser-like plume vent areas and a mechanism of energy
transport toward them.
Since the discovery of N, on Triton (I),
Neptune's largest moon, and especially
since the Voyager flyby (2), there have
been several attempts to model the trans-
port of volatiles on Triton in response to its
complex seasonal cycle (3). It has usually
been assumed that the albedo distribution
on Triton is the result of the seasonal N,
transport (4-8), but so far no models have
successfully reproduced the observed albedo
An understanding of the mechanisms
driving the albedo distribution is somewhat
incidental to the auestion of the vertical
phase composition of N, ice deposits on
Triton. Ground-based s~ectralmeasure-
ments show that Triton's illuminated sur-
face is mostly covered with frozen N, at
least many centimeters deep (1). The mean
insolation on Triton is greatest at the equa-
tor and smallest at the poles (4, 5, 9). As a
result, any N, in excess of that which can
be sublimated and recondensed during one
of Triton's extreme seasons (3) is transport-
ed to permanent polar caps, which may be
several hundred meters thick and extend as
far toward the equator as 245' of latitude,
depending on Triton's total inventory of
surface N2 (8). The seasonal redistribution
of volatiles also causes global temperature
variations on time scales of a few tens of
years (4, 5, 1O), which can be as much as
15 to 20 K or as little as 2 to 4 K, depending
again upon the total surface inventory of
N,. Furthermore, the 38 K temperature of
M.S. 183-501, Jet Propulsion Laboratory, California
Institute of Technology, 4800 Oak Grove Drive, Pasa-
dena, CA 91109.
Triton's lower atmosphere is thought to be
representative of all the N, ice on Triton's
surface (4, 5) and is perilously close to the
temperature (35.61 K) of the a-p (cubic-
hexagonal) phase transition in solid N,.
The subsurface ice layer on Triton is
therefore likely to experience the passage of
multiple phase transition fronts as the glob-
al temperature oscillates above and below
35.61 K. Besides the absorption and liber-
ation of latent heat at the phase transition,
there is also a large change in volume over
a small range in temperature: Laboratory
measurements indicate that the density of
solid N, changes by 1 to 2% in a range of
about 1 K around 35.61 K (1I). The in-
duced stresses cause severe fracturing of the
crystalline solid when the transition is from
the p to the a phase (1 1). (To our knowl-
edge, experiments to determine whether
a-N, crystals shatter when the phase tran-
sition is approached from lower tempera-
tures have not been done; experiments (I I)
have dealt only with powdered a-N2 be-
cause large crystals are difficult to obtain.)
We assume that Triton is completely
differentiated (2, 8), with a silicate core of
radius -1000 km overlain by a water-ice
mantle about 350 km thick and a thin
veneer of solid N,, no more than 1 km
thick. We include in our heat transfer
model the effect of the reversible phase
transition from the denser cubic a phase to
the hexagonal P phase that occurs when
the temperature rises to 35.61 K (for the
14N2isotope at equilibrium vapor pressure),
at which latent heat of 55.62 callmol for
the 14N2 isotope (11) is absorbed. The
SCIENCE VOL. 261 ' 6 AUGUST 1993