Christopher D. Barnet
General Sciences Corporation, Laurel MD 20707-2929
John Caldwell, Cindy C. Cunningham
York University, Toronto ON M3J 1P3 Canada
Xin-Min Hua
NRC and GSFC/NASA, Greenbelt MD 320771
at the center of Mars' disk was 6.25 
1
mbar and the Cameron band absorption indicates that the CO mixing ratio was
600 
300 ppm.
Keywords: Mars, carbon monoxide, Cameron band, ultraviolet, HST/GHRS
Previous measurements of carbon monoxide (CO) on Mars (Clancy et al., 1983) have shown that its mixing ratio in the Martian atmosphere is variable and that there may be seasonal fluctuations. Further, the altitude distribution of CO is not well constrained.
We have observed Mars with the HST/GHRS, in the spectral vicinity of the
Cameron 0-0 band near 2060Å. The CO mixing ratio derived from the present
study, 600 
300 ppm, is comparable to measurements by some others
(Clancy et al. 1990, and Encrenaz et al. 1991) which show CO
mixing ratios in the range of 550--800 ppm. However, it is considerably
smaller, by a factor of 3--5, than those of Kakar et al. (1977) and of
Good & Schloerb (1981).
Measurements of CO from a variety of studies performed during
different Martian seasons consistently give CO mixing ratios in the above
range. This is puzzling since the main atmospheric constituent, CO
, is
subject to substantial seasonal variations due to condensation (fall and
winter) and sublimation (spring) on the polar caps. Measured variations in
CO
should result in measurable variations in CO by a factor of 2--4
according to predictions from Mars general circulation models.
The photochemistry, transport properties and thermal structure of the atmosphere of Mars, and the temporal variations of those properties on daily and seasonal time scales, are important to understanding the basic questions about the Martian environment---its origin, evolution and stability. This paper, in which a previously observed species is observed by a new technique, addresses one of these areas---photochemistry---directly.
The photochemistry of Mars is currently uncertain in several respects. One is
the question of the suitability of homogeneous (gas phase) chemistry to
explain the abundance of trace species, including CO as well as H
O, O
and O
, in the atmosphere. Atreya & Gu (1994,1995) have suggested that
some kind of heterogeneous catalytic process, perhaps involving aerosols of
ice or dust, may be required to explain the recycling rate of CO
on Mars:
specifically, aerosol surface chemistry may be required to provide a sink for
some of the trace species which participate efficiently in the recycling of
CO
.
The present CO result, which is derived from ultraviolet spectroscopy, is independent from, but consistent with, other results cited above, all of which were made at much longer wavelengths, where the atmospheric radiative transfer is much different. In principle, our result has the potential to determine the temperature of the Martian atmosphere from rotation-vibration line structure, using radiative transfer calculations where the source function is entirely due to scattering in an optically thin medium, which is specifically independent of thermal radiation. In practice, however, uncertainties in transforming the measured Mars flux to albedos preclude such determination in the present data.
Like the other methods, ours is extremely insensitive to the vertical distribution of CO on Mars.
An ultraviolet spectrum of Mars was obtained by the G200M grating
of the Goddard High Resolution Spectrograph (GHRS) aboard the Hubble Space
Telescope (HST) on April 26, 1993, at 5:04 UT. The grating has a spectral
resolution of 
25000 and, for this
observation, was centered on the Cameron 0-0 CO band between 2060 and 2070Å.
The observing parameters for Mars at the time of the observation are
summarized in Table 1.
Table 1: Parameters for the April 26, 1993, 5:04 UT Mars Observation
In order to compare the data to radiative transfer atmospheric models, one
must convert flux to normalized intensity, 
. The first step
in this process is to divide the HST Mars spectrum by a solar spectrum.
We chose the 1983 solar measurements of Hall & Anderson (1991) because
theirs are the only ones in this wavelength region with sufficient spectral
resolution (0.1Å) not to compromise information in the HST spectrum.
Their data, obtained from a balloon near 40km altitude, were
extrapolated to zero optical depth by correcting for ozone, O
Herzberg
continuum absorption, and Rayleigh scattering.
Dividing the Mars spectrum by the solar spectrum requires that both
spectra be sampled at the same wavelengths. The Mars spectrum is sampled at
0.08Å, while the solar spectrum is sampled at 0.1Å.
We, therefore, used a cubic spline interpolation to calculate the solar flux at
the exact wavelengths of the HST data points. The two spectra are shown
together in Figure 1. In that figure, the data have been smoothed
with a 3-point running mean filter to improve the signal to noise ratio.
Figure: The Mars spectrum (darker line) and the Sun, with arbitrary
normalization. Both spectra have been smoothed, and Mars has been shifted by
Å with respect to the Sun, as discussed in the text. Near 2062Å,
where the CO Cameron band occurs, Mars is noticeably relatively less bright
than the Sun. Aliasing effects in solar Fraunhofer lines is apparent.
Before we plotted Figure 1, we had discovered an unexpected relative
wavelength shift between the planetary and solar data. To investigate the
apparent anomaly, we introduced a systematic variable wavelength shift,
, between the two spectra and minimized the square of the sum
of the variances in the continuum region, outside the region of the CO band,
where we expect real differences to occur in spectral detail. The minimum in
the sum of the square of the variances in the continuum region corresponds to
a wavelength shift of 
Å, which is included in
Figure 1.
The wavelength shift is not a Doppler effect. At the time of observation,
the geocentric velocity of Mars was 16.25kms
which
corresponds to a redshift in the Mars spectrum of +0.11Å (i.e., Mars
is receding from the Earth). This ``best fit'' shift is therefore larger
than the Mars radial velocity by a factor of 4, and in the opposite sense.
During the data reduction process, we learned that there has been a
systematic problem with the HST ground software system for doing spectroscopy
of moving targets. Specifically, the standard target specification process
for a planet to be observed with the HST includes not specifying the right
ascension and declination of the target in advance but allowing the scheduling
system to insert the correct position at the time of observation. That aspect
of planet tracking has worked very well, but it had not been realized that
there was no corresponding insertion of the correct position in the on-board
GHRS Doppler correction software. The latter effectively assumed that all
planetary targets were at 0
RA, 0
dec, instead of their correct
positions, and applied erroneous Doppler compensation to the data before
telemetry to the ground.
In view of this known defect in the data, which are otherwise clearly good, we have not further investigated the wavelength discrepancy between the planetary and solar spectra.
Two features are apparent in Figure 1. First, in the region between 2060 and 2065Å, where the CO Cameron band occurs, Mars is noticeably less bright than the Sun compared to the rest of the spectrum. This is the essential result of this paper, and is discussed in detail below.
It is also obvious in Figure 1 that there is significant aliasing between the two spectra in the solar Fraunhofer lines. This problem has been found in many HST and other planetary spectroscopy programs, and is due to the intrinsic difficulty in matching the characteristics of two different spectrographs. Since it is not possible to observe the Sun with the HST, it is necessary to use a solar spectrum obtained elsewhere. When the planetary spectrum is divided by the solar spectrum to obtain an albedo, the aliasing results in ``noise'' features that are frequently the limiting factor in determining the value of the data. See Figure 3 below, for example.
One potential solution to this problem is to use HST observations of solar substitutes, including planetary satellites. At the time of writing, there are plans to perform just such observations for calibration purposes. One problem is that, especially in the ultraviolet, the signal from satellites is sufficiently weak that there will be significant photon noise in the substitute spectrum. A potential solution to that problem is to use the brightest known planetary satellite which, unfortunately, is also a very difficult object for the HST to track. Efforts will be made to observe the Moon for this purpose, and it is hoped that future papers will benefit from them.
The final albedo spectrum, which is included in Figures 2--5 for comparison with atmospheric models, is
where 
is the heliocentric distance of Mars in AU; 
is
the solar flux at 1 AU; 
is the measured intensity of Mars; 
, is the wavelength shift required to bring the two spectra into
coincidence; and 
is the area of the GHRS aperture, 4 arcsec
.
Models of the geometric albedos of Mars were calculated to match
the HST observations. The atmospheric code is derived from a doubling and
adding algorithm developed at Princeton University by R. E. Danielson,
P. G. Wannier, and W. D. Cochran. For the new Mars observations, it was
necessary to modify the code to calculate the albedo at the observational
phase angle of 36.9
.
The albedo of Mars in the wavelength region of the CO Cameron band is
dominated by optically thin Rayleigh scattering from CO
molecules, with
reflection of sunlight from the surface also being significant
(Caldwell 1973). In this study, we simulate the Martian environment by two
vertically homogeneous layers. The first layer consists of uniformly mixed
CO
and CO. The second layer, which is modeled as a thick layer of grey,
isotropic scatterers, simulates the dark reflecting surface. The free
parameters are the abundance of CO
, the mixing ratio of CO, and the
surface reflectivity.
Since laboratory measurements of absorption cross section are extremely difficult at Martian pressures (typically 7mbar), a theoretical model was employed to simulate the Cameron 0-0 CO band. The theoretical basis of this model and comparisons to laboratory measurements at 185mbar and 215mbar are discussed by White et al. (1993) and by Caldwell et al. (1993). In summary, we are confident that the model successfully reproduces CO line positions and strengths, including the latter as a function of temperature.
The model was calculated at a spectral interval of 0.001Å, which ensures that each rotational line is represented by a minimum of 5 points. It was run for temperatures of 150, 200 and 250K, respectively, and at Martian pressures. It turned out that the data were too noisy to allow estimation of the temperature from the variation of line strengths within the band, so in our best-fitting model below, we used 200K. Individual line profiles were assumed to have Doppler broadening only.
It is computationally prohibitive to calculate the exact atmospheric model at every wavelength point (roughly 10000 per model) within the CO absorption range. Instead, we calculated the continuum model every 5Å\ and for each such subrange, we calculated the atmospheric model for 10 logarithmically-spaced values of absorption, corresponding to the local range of CO absorption. Effectively, the result of this calculation is a two-dimensional grid of albedo as a function of absorption coefficient and wavelength. The true albedo at each wavelength can be found accurately by interpolation within this grid. We verified a segment of our interpolation method with the ``tru'' model calculated at full resolution.
The high resolution atmospheric model was then binned to 0.08Å and both the data and model were smoothed with a five-point running mean for presentation. The treatment of the model is directly comparable to that for the data.
Because the continuum model contains two free parameters, there is
some uncertainty in the ``best'' model fit, which we discuss below. Our
preferred model has a CO
column abundance of 84 meter-amagats
(corresponding to a CO
surface pressure of 6.25mbar), a CO mixing ratio
of 600 ppm, and a grey surface albedo of 0.035. In Figure 2, we
show the model for these parameters, both at full resolution and with binning
and smoothing. Also shown in this figure are the GHRS Mars data.
Figure: The full resolution (i.e., 0.001Å resolution) model
is shown as a light line. These data, binned to 0.08Å and smoothed
with a 5 point running mean filter (solid dark curve), are compared with
the GHRS data (dashed curve), also smoothed with a 5 point running mean
filter. This is the ``best fit'' atmospheric model consisting of a
surface pressure of 6.25mbar, 600 ppm of CO, and a surface albedo of
0.2.
The surface albedo and abundance of CO
can both be adjusted to
fit the continuum level. However, increasing the ratio of CO
to surface
contribution will cause the slope of the continuum to increase toward shorter
wavelengths because the continuum scattering is optically thin and exhibits
the 
dependence of Rayleigh scattering. Conversely, to create a
``flatter'' albedo than shown would require a significantly greater
contribution of light from the surface.
In Figure 3, we compare the best fit model with an extreme model
having a single layer of CO
and no reflecting surface. We require 150
meter-amagats (surface pressure = 11.1mbar) to match the albedo of the Mars
spectrum in the case of no surface reflectivity, but it is clear that the
slope of the continuum is too large. In general, we find that models with
more than 
7mbar of CO
have too large a slope in the continuum.
Also shown in Figure 3 is a model with 84 meter-amagats without a surface, to
illustrate the albedo of the surface.
Figure: Three models compared to the GHRS Mars spectrum, which is shown as
a broken line. Three features which are probably due to aliasing effects
with solar features are indicated as ``solar''. The ``best fit'' model,
including non-zero surface reflectivity and 6.25mbar of CO
is shown as a
dark line; a model with no surface reflectivity and 11.1mbar of CO
is
shown as a lighter line; a model with 6.25mbar of CO
and no surface
albedo is included at the bottom of the figure.
We estimate the uncertainty in the CO
surface pressure to be
about 
1mbar. The CO mixing ratio is relatively insensitive to the
ratio of albedo concentration of CO
and surface since the atmospheric
scattering is optically thin in both the extreme models.
In Figure 4, we show two models where the abundance of CO is varied.
Both models have 6.25mbar of CO
and a surface albedo of 0.035. The
light line represents 50% more CO (900 ppm) than the ``best model'' (600
ppm) and the dark line represents 50% less CO (300 ppm). From these more
poorly fitting models, we estimate the error on the CO abundance to be 
300 ppm.
Figure: Two models are shown to illustrate the sensitivity of the
model to the abundance of CO. The dark line has 50% more CO and the
light line has 50% less CO than the ``best fit'' model.
Although the molecular band model shows some variation of absorption coefficient with temperature, the low signal-to-noise of the GHRS data and the binning to 0.08Å preclude determination of the temperature of the absorbing CO. We chose the 200K model for our analysis based on temperature soundings and average conditions on Mars.
In Figure 5, we show the temperature effect. Clearly, the greatest spectral difference between the models with different temperature occurs in the long wavelength part of the band, between 2065 and 2070Å, where several solar aliasing features occur. (See Figure 1, above.) The aliasing problem precludes distinguishing between the temperature models in the current reduction of the data.
Figure: Two models are shown to illustrate the sensitivity of the
model to the temperature of CO. The dark curve (deeper absorption) is
the absorption of CO at 250K and the lighter curve is at 150K.
We have demonstrated that measurement of the difficult CO 0-0 Cameron band is feasible with the HST. Future observations covering the different Mars seasons will be an important HST science objective in the coming years. With the advent of long-slit spectroscopy with the Space Telescope Imaging Spectrograph (STIS), the observations can be very efficient in covering different latitude regions on the planet. If a suitable solar substitute object, preferably the Moon, can be observed with the HST, the resulting albedo curves will be much less noisy than the one presented herein.
Accurate and comprehensive studies of the chemical properties of Mars with the HST will be an important part of its future planetary program.
Atreya, S. K., & Gu, Z. G. 1994, JGR, 99, 13133
Atreya, S. K., & Gu, Z. G. 1995, Adv. Space Res., 16, 57
Caldwell, J. 1973, Icarus, 18, 489
Caldwell, J., Cunningham, C. C., Barnet, C. D., Wu, C. Y. R., Chen, F. Z, & Judge, D. L. 1993, Fifth Int'l Conf. on Lab. Res. for Planetary Atmospheres
Clancy, R. T., Muhleman, D. O., & Berge G. L. 1990, JGR, 95B, 14543
Clancy, R. T., Muhleman, D. O., & Jakosky B. M. 1983 it Icarus, 55, 282
Encrenaz, T., Lellouch, E., Rosenquist, J., Drossart, P., Combes, M. & Billebaud, F. 1991, paper presented at the EGU Congress, Wiesbaden, Germany
Good, J. C. & Schloerb, F. P. 1981, Icarus, 47, 166
Hall, L. A. & Anderson, G. P. 1991, JGR, 96, 12927
Kakar, R. K., Waters, W. J., & Wilson, W. J. 1977 Science, 196, 1090
Lellouch, E., Encrenaz, Th., Phillips, T., Falgarone, E., & Billebaud, F. 1991 Planet. Space Sci., 39, 209
Lellouch, E., Paubert, G., & Encrenaz, Th. 1991, Planet. Space Sci., 39, 209
White, H. P., Hua, X.-M., Caldwell, J., Chen, F. Z., Judge, D. L., & Wu, C. Y. R. 1993 JGR, 98, 5491