Dry granular flows can generate surface features

resembling those seen in Martian gullies

Troy Shinbrot*†, N.-H. Duong‡, L. Kwan‡, and M. M. Alvarez§
Departments of *Biomedical Engineering and ‡Chemical and Biochemical Engineering, Rutgers University, Piscataway, NJ 08854; and §Instituto Tecnologico
y de Estudios Superiores de Monterrey, Monterrey, 52926, Mexico

Edited by Henry J. Melosh, University of Arizona, Tucson, AZ, and approved April 7, 2004 (received for review December 11, 2003)

Over the past decade or more, contradictory evidence of Martian

climate, indicating that surface temperatures seldom if ever ap-
proach the melting point of water at midlatitudes, and geomorphic
features, consistent with liquid flows at these same latitudes, have
proven difficult to reconcile. In this article, we demonstrate that
several features of liquid-erosional flows can be produced by dry
granular materials when individual particle settling is slower than
characteristic debris flow speeds. Since the gravitational acceler-
ation on Mars is about one-third that on Earth, and since particle
settling speeds scale with gravity, we propose that some (although
perhaps not all) Martian geomorphological features attributed to
liquid flows may in fact be associated with dry granular flows in the
presence of reduced gravity.

A number of recent studies have investigated evidence for

liquid water on Mars. Much of the available evidence has
concerned recent, or even contemporary (1), geomorphological
features, such as eroded channels and gullies that are common
signatures of terrestrial water flow (2–4). This evidence is
difficult to reconcile (5–7) with surface temperatures on Mars,
which seldom exceed ⫺50°C at latitudes and locations where
these features are often found (6, 8). This paradox persists
despite recent Mars Rover data supporting the case for ancient
surface water. Several authors have proposed possible resolu-
tions to this contradiction (9–12), notably a recent analysis
suggesting that some liquid-like flow features in Martian gullies
may instead be associated with dry granular flow (1). In this
article, we observe that gravity on Mars is 38% that on Earth
(13), and behaviors of dry grains at reduced gravitational
acceleration have never, to our knowledge, been catalogued. As
we will show, reduced gravity has the effect of prolonging
fluidization of particle flows by decreasing particle settling
speeds as compared with debris flow speeds. We present data
demonstrating that many features that have been attributed to
liquid flow in Martian gullies can indeed be reproduced in
terrestrial laboratory experiments designed to mimic reduced
gravity.
Our point of departure is the observation that the speed at
which small dry particles (14) settle in a gas can be approximated
by the Stokes settling relation:
共␳s ⫺ ␳g兲d2g
Vs ⫽ , [1]
18␮g
whereas the terminal speed that large debris flows can reach
while flowing through a gas is bounded by (15):
Fig. 1. Liquid-like flow patterns in dry grains under conditions of reduced
settling-to-flow speed ratio. (a) Features found in laboratory flows by using
lightweight, hollow beads immediately after an initial collapse, exhibiting
liquid-like behaviors such as rivulets near the steep upstream edge (black
arrow), narrowed incipient channel (white arrows), and reflected waves (em-
anating from white arrows). (b) Later development of waves from a separate
experiment. (c) Superficially similar wave-like dunes downhill from central
peaks (top of snapshot) of Hale crater, Mars.¶
stress that nothing is profound about the difference between
these relations, which simply reflect the fact that particles, being
small, settle at a low Reynold’s number, whereas debris flows,
being large, travel at a high Reynold’s number. Because of this
difference, however, it is possible that Martian gravity being
about one-third of Earth’s could permit particles on Mars to
settle more slowly, and therefore remain fluidized longer, than
on Earth. Indeed, arguably particle fluidization on Mars may be
further amplified both by wind (17) and by the fact that the
atmospheric density, ␳g, is two orders of magnitude smaller on
Mars than on Earth (18). This raises the possibility, analyzed in
subsequent sections, that grains on Mars may collectively flow
more rapidly than on Earth while at the same time they
individually settle more slowly. Since particle flows persist until
individual particles settle against one another (19), it follows that
reducing the ratio of settling to flow speeds, Rst ⫽ Vs兾Vt, can
produce distinctly liquid-like appearances (Fig. 1).
As a first example, in Fig. 1 a and b we show low-density,
hollow ceramic beads (PQ Corp., Augusta, GA), with bulk
density 1兾15 that of solid glass and diameters ranging from 4 to
90 ␮m. The beads are deposited into a simple acrylic box 40 cm

Vt ⫽ 冑 4D␳sg
3Cs␳g
. [2]
Here ␳g and ␮g are the gas density and viscosity, d, ␳s, and Cs are
characteristic particle diameter, density, and drag coefficient, g
is gravitational acceleration, and D gives the scale of the debris
flow (16). These relations are only correct to the lowest order;
however, they illustrate that, all other things being equal, particle
settling speeds go as g, whereas debris flow speeds go as 公g. We
This paper was submitted directly (Track II) to the PNAS office.
Abbreviation: NASA, National Aeronautics and Space Administration.
†To whom correspondence should be addressed. E-mail: shinbrot@soemail.rutgers.edu.
¶Malin, M. C., Edgett, K. S., Carr, M. H., Danielson, G. E., Davies, M. E., Hartmann, W. K.,
Ingersoll, A. P., James, P. B., Masursky, H., McEwen, A. S., et al. (2000) Autumn Afternoon
in Hale Crater, NASA’s Planetary Photojournal MOC2–257, November 17, 2000. Available
at www.msss.com兾mars㛭images兾moc兾nov㛭00㛭hale.
© 2004 by The National Academy of Sciences of the USA

8542– 8546 兩 PNAS 兩 June 8, 2004 兩 vol. 101 兩 no. 23 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308251101


Fig. 2. Alcoves. (a) Rounded hillocks separating flowing gullies appear
immediately after the initial collapse of a flat surface of a bed of hollow beads.
(b) Alcove features in Martian polar pits cited elsewhere (see footnote ‡‡ and
ref. 5) as evidence for surface water flow. (c) Laboratory alcoves with charac-
teristic ‘‘spur-and-gully’’ structure seen also in Martian landforms (e.g., Fig.
3b). (d) A filled alcove from Dao Vallis on Mars discussed elsewhere†† (8): this
alcove is filled with subsequent deposits, but has a similar outline to that
shown in c. (e) Upstream precipice of gullies in laboratory with cohesive glass
beads. To produce the jagged upstream feature shown, we perform the same
gradual tipping experiment as in a, but we use solid-glass particles sieved
between 45 and 90 ␮m that have been exposed to high humidity (90% RH for
several weeks), thus producing aggregates held together by liquid bridges
(25). ( f) Detail of upstream edge of channel shown in Fig. 3c. This snapshot
shows an eroded gap between upper container wall and granular bed (see
text): gap between arrows is ⬇3 mm wide.
on a side, and the bed of beads is then tipped until the beads
begin to flow. The low mass of the beads slows settling in air,
producing visible waves, sloshing, and reflection from bound-
aries that strongly resemble liquid flows (20) (movies of these
flows and additional data are published as supporting informa-
tion on the PNAS web site). Once the particles come to rest, this
liquid-like flow leaves signatures in the frozen state; thus, waves
that reflect from the narrowed region of the incipient channel
(Fig. 1a, white arrows) are visible, as are fine rivulets upstream
(black arrow). As time progresses, the downstream waves ad-
vance, and the upstream rivulets accentuate to form mountain-
ous features as shown in Fig. 1b (see also Figs. 2c and 3a). By
comparison, in Fig. 1c, we display wave-like (20) sand dunes in
the Hale crater on Mars. Like our experiments, these dunes
emanate downhill from steep peaks: based on estimates using
shadow length and time of day,储 the peak visible at the top of this
figure is ⬎150 m above the crater floor, which is visible at the
bottom of the figure. Similar fluid-like features are seen in
Newton Basin in Sirenum Terra,** again traveling downhill from
steep peaks.
Evidently, dry grains can flow in a liquid-like manner when
their settling speed is sufficiently reduced—in our experiments
by a simple reduction in grain density. We caution that reducing
储Thelargest of the central peaks has been estimated (www.msss.com兾mars㛭images兾moc兾
nov㛭00㛭hale) to be 630 feet from the crater floor; the peak shown in Fig. 1c casts a shadow
one quarter as long. This estimate neglects the height and inclination of the surface on
which the shadow falls, but sets a reasonable approximation for the height in the figure.
**Malin, M. C., Edgett, K. S., Carr, M. H, Danielson, G. E., Davies, M. E., Hartmann, W. K.,
Ingersoll, A. P., James, P. B., Masursky, H., McEwen, A. S., et al. (2002) Gullies in crater at
42.4°S, 158.2°W, NASA’s Planetary Photojournal MOC2–320, October 7, 2002. Available
at www.msss.com兾mars㛭images兾moc兾e7㛭e12㛭captioned㛭rel.
Shinbrot et al.
grain density does carry the potential to generate spurious
effects associated with complex air-particle interactions during
settling (21, 22), and as we discuss shortly, some of the effects
that we report are indeed significantly air-mediated. Although
no earth-bound experiment can fully capture Martian condi-
tions, we find aspects of numerous different morphologies that
have been attributed to liquid-borne flow in Martian gullies.
Morphological Comparisons
To compare laboratory and Martian morphologies, we note that
Martian geological features that have been identified as indi-
cating liquid surface water on Mars have been classified into
three categories (5): upstream features (e.g., alcoves), midstream
features (channels), and downstream features (aprons). In the
next three sections, we briefly compare features of our terrestrial
experiments in each of these flow regions in turn with Martian
analogues. After these qualitative comparisons, we analyze the
applicability of terrestrial, laboratory-scale, experiments for the
simulation of Martian, geological-scale, systems.
Upstream Features
In Fig. 2, we compare some upstream features of laboratory
experiments with Martian gully features. In Fig. 2a, we show a
laboratory pattern of gullies in which grains flow separated by
static hillocks. In the experiment shown, the hollow beads of Fig.
1 are deposited in a simple rectangular box that is tipped until
flow begins; the originally flat bed surface is visible at the top of
the figure. By comparison, in Fig. 2b, we show a photograph‡‡ of
Martian gullies (5), separated by superficially similar hillocks.
As time progresses in our experiments, the hillocks erode, giving
rise to a characteristic ‘‘spur-and-gully’’ morphology, shown
in Fig. 2c. Comparable spur-and-gully formations are seen in
innumerable locations both in terrestrial and in Martian geology.
For example, in Fig. 3b we see a portion of such a formation on
Mars, and in Fig. 2d, we show a Martian alcove that has been
filled with subsequent deposits, leaving an outline that suggests
an originally scalloped alcove similar in shape to that seen in
Fig. 2c.
Since a variety of particle types are doubtless present at the
Martian surface (14), we performed separate experiments using
round, angular, and weakly cohesive grains. We found that,
provided the grains remain sufficiently small (ⱕ60 ␮m), many of
the features that we describe for hollow beads were retained, but
channels became irregular and sinuous for angular and weakly
cohesive grains [termed Geldard type C in the powder literature
(23)]. Tipping a bed of these grains causes aggregates to dislodge
in chunks (24), leaving jagged upstream outcroppings, such as in
Fig. 2e (possibly comparable with Martian outcroppings as in
PHYSICS
Fig. 3d).
Midstream Features
As with upstream features, midstream channels can develop a
rich variety of patterns, depending on elapsed time (compare
Fig. 1 a and b), particle properties (Fig. 2 e vs. f ), and bed
preparation. The most common midstream feature seen is a
broad channel emanating from a spur-and-gully, as shown in Fig.
3a alongside a comparable Martian channel in Fig. 3b. Such
channels are often straight, but can be sinuous as in this example.
In this experiment as in those shown in Fig. 1, a bed of hollow
††Malin, M. C, Edgett, K. S, Davis, S. D, Caplinger, M. A, Jensen, E., Supulver, K. D, Sandoval,
J., Posiolova, L. & Zimdar, R. (2000) MOC image M03– 04950, Malin Space Science Systems
Mars Orbiter Camera Image Gallery, October 16, 2000. Available at www.msss.com兾
moc㛭gallery兾ab1㛭m04兾images兾M0304950.html.
‡‡Malin,M. C., Edgett, K. S., Carr, M. H., Danielson, G. E., Davies, M. E., Hartmann, W. K.,
Ingersoll, A. P., James, P. B., Masursky, H., McEwen, A. S., et al. (2000) Polar Pit Wall,
NASA’s Planetary Photojournal MOC2–237, June 22, 2000. Available at www.msss.com兾
mars㛭images兾moc兾june2000兾sp㛭pit.
PNAS 兩 June 8, 2004 兩 vol. 101 兩 no. 23 兩 8543
Fig. 3. Channels. (a) Hollow spheres after prolonged flow, exhibiting a
sinuous stream and multiply branched inlet patterns. (b) Mountainous gully in
Candor Chasma on Mars.§§ (c) Narrow channel under conditions of increased
bed cohesion described in text. (d) Channels in gullies described in previous
studies of Martian surface water (6).¶¶

beads was aerated (by recently avalanching the grains) and then
tipped rapidly to produce comparatively wide flows of beads
(1–2 cm or 500–1,000 bead diameters across).
To better understand how our laboratory granular flows carve
channels and to explore a sampling of the spectrum of structures
that can be produced by granular flows at reduced Rst, we
performed variants of our flow experiments. Notably, by com-
parison with earlier results, in Fig. 3c we prepared the bed
(details follow) so as to generate a well settled mass that was then
tipped more slowly, producing the long narrow channels shown
(about 20 diameters across) with steep upstream boundaries
(magnified in Fig. 2f ) and conical downstream debris deposits
(Fig. 4a). In detail, for Fig. 3c we first allowed the beads to settle
overnight; this preparation is necessary because the small size
and density of the beads implies that they will only very slowly
settle and expel air from interstices. If we instead tilt a bed of
beads that has been recently agitated, we obtain broad channels
as in Fig. 1b. In Fig. 3c, after overnight settling, we slowly stirred
the bed with a rod to break up aggregates, and then tipped the
box onto a corner and gently tapped it to move the beads into
the corner and to smooth the bed surface. Finally, we slowly
tilted the box until flow began. We repeat the experiment by
tapping and then tilting the box again. In Fig. 3d, we show
comparable Martian gully channels discussed in investigations of
the case for liquid water on Mars (1, 5).
Downstream Features
The final category of flow features that have been attributed to
surface water flow on Mars is downstream depositional patterns,
two of which are shown in Fig. 4. As we have mentioned, Fig. 4a
displays a detail of the downstream conical pile of material from
the channel of Fig. 3c. Alongside this pile, we show the conical
(1, 5) deposit from the Martian channel of Fig. 3d, suggesting
§§Malin, M. C, Edgett, K. S, Davis, S. D, Caplinger, M. A, Jensen, E., Supulver, K. D., Sandoval,
J., Posiolova, L. & Zimdar, R. (2000) MOC image M11– 02514 Malin Space Science Systems
Mars Orbiter Camera Image Gallery, October 16, 2000. Available at www.msss.com兾
moc㛭gallery兾m07㛭m12兾images兾M11兾M1102514.html.
¶¶Malin, M. C., Edgett, K. S, Davis, S. D, Caplinger, M. A, Jensen, E., Supulver, K. D, Sandoval,
J., Posiolova, L. & Zimdar, R. (2000) MOC image M03– 02290 Malin Space Science Systems
Mars Orbiter Camera Image Gallery, October 16, 2000. Available at www.msss.com兾
moc㛭gallery兾ab1㛭m04兾images兾M0302290.html.
Fig. 4. Deposits, aprons, and swales. (a) Detail of conical depositional pile
downstream of channel shown in Fig. 3c. (b) Detail of comparable deposi-
tional cone from Martian gully channel shown in Fig. 3d. (c) Depositional
apron of hollow beads, showing a fine swale that formed by erosion from a
fluidized stream. (d) Later evolution of downstream deposit, showing exten-
sive variegated erosion. (e) Comparative variegated aprons downstream of
channel described elsewhere (5).
that dry granular flows may account for simple depositional
structures. More complicated features are also seen downstream
of Martian gully streams, however; so in Fig. 4c we show a
portion of an apron containing hollow beads at the downstream
edge of a flowing region (the base surface of the tipped
container is visible at the bottom of the snapshot). As indicated
by the arrow, subsequent flows of beads cut a thin swale into the
apron; similar swales develop into striated channels as shown in
a later evolution of the hollow beads in Fig. 4d. Such multiply
striated regions are not common in our experiments, but they can
be found at the tail end of flowing regions as shown in the figure,
especially when the flow encounters an obstacle: here, the
vertical end wall of the container, which is located at the very
bottom of this snapshot. By comparison, swales cut into aprons
on Mars as identified by the arrow in Fig. 4e have been attributed
to water flow (5).
Analysis
Evidently, reducing Rst by changing particle properties has the
effect of producing remnant structures that are similar to some
geomorphic features believed to imply water flows in Martian
gullies. A unifying feature of all of the liquid-like flow patterns
that we have seen in laboratory dry grains is that, for the
considerations of the interplay between settling and flow speeds
that we have summarized to come into play, particles need to
become fluidized to start with. The conditions imposed by using
hollow and small beads permit the grains to remain fluidized, but
without a source of fluidized grains, the patterns that we have
shown would not form or endure.

8544 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308251101 Shinbrot et al.

 Table 1. Data for settling comparisons using Eqs. 1 and 2
Flow type ␳s ␳g d D g ␮g Cs Rst ⫽ Vs兾Vt

Terrestrial laboratory 0.15 1.3䡠10⫺3 0.002 1 980 1.8䡠10⫺4 100 0.005

Terrestrial debris 2 1.3䡠10⫺3 0.2 104 980 1.8䡠10⫺4 100 2.6
Martian debris 2 1.9䡠10⫺5 0.2 104 360 1.8䡠10⫺4 100 0.2

All units are cgs, and these data apply at above 5–10 Torr: below that value, air viscosity scales (15) with
pressure2兾3, and settling speeds become ballistic (see Analysis). Moreover, these are upper bounds. For example,
particle settling speeds on Earth never exceed debris flow speeds, and as emphasized in the text, these data are
only presented as scaling relations. Particle densities, ␳s, for hollow beads are reported by the manufacturer, and
for debris flows are measured in refs. 30 and 31. Gas densities, ␳g, and viscosities, ␮g, are a matter of record, see
for example, ref. 4. Hollow bead sizes, d, are measured in our laboratory from digitized microscope photographs,
and geological particle sizes are chosen to be representative of coarse sand. Debris sizes, D, are approximations
from Figs. 1 and 2 in our experiments and are from recorded data (e.g., ref. 31 and A. Mashiyama and K. Kurita,
unpublished work) for geological flows. The drag coefficient, Cd, is obtained from a recent computational study
(32) and represents an overestimate of drag across a geological surface.

In our experiments, we have found that fluidization of slowly
settling particles can appear spontaneously in one of two ways.
First, a region of steep slope may abruptly collapse, releasing
enough energy to fluidize grains downstream. Such a case is seen
in Fig. 2a, where downstream flows emanate from steep-walled
upstream alcoves after the initial collapse of the bed. Fluid-like
flows are readily observed to be generated after a region of the
upstream bed collapses, producing an avalanche of rapidly
flowing grains. Similarly, we find liquid-like flows typically
originate in steep mountainous regions (1). An example of this
is shown in Fig. 3a, where we show results of an experiment with
hollow beads in which jagged surfaces, such as those shown
previously in Fig. 1b, are permitted to mature through prolonged
flow into sinuous and multiply branched channels. Alongside this
snapshot from our laboratory, in Fig. 3b we display an actual
mountainous region that supplies a Martian gully channel (5, 6).
The scenario of dry granular mass flow initiated by an upstream
collapse at a steep wall is consistent with available Mars data; in
Martian gullies, it has been remarked that a universal feature
associated with gully morphologies that indicate water flow is
steep upstream walls (1); moreover, collapses that generate
significant mass flows have been seen in recent Mars Express
photographs (www.esa.int兾export兾externals兾images兾016-
090204㛭2-0037㛭02-6p.jpg). In our experiments [both the single
collapse events (Fig. 2a) and prolonged flow (Figs. 2c and 3a)],
the fluidized granular cascades that we observe leave upstream
edges that are invariably steep and multifaceted.
A second mechanism that spontaneously generates fluidiza-
tion in our experiments appears when fluidized flow initiates at
an already steep surface such as the container wall, where
unobstructed air velocities are presumably highest relative to the
geological substrates seem certain to be more robust than our
simple compacted bed, and so the processes, whether water- or
airborne, that could erode them are bound to be enormously
slower and more incremental than in our model laboratory
system. It is therefore surprising that Figs. 1–4 indicate that some
geomorphological features associated with liquid flows may be
duplicated on the laboratory scale with dry grains by suitably
reducing Rst. Similar scaling arguments have been applied to
extrapolate laboratory fluid experiments to geophysical scales
(26) and to relate laboratory granular experiments to extrater-
restrial craters (27). Even more germane, scaling using a ratio
between settling and debris flow speeds has been reported in
terrestrial particle-laden oceanic and volcanic flows (28, 29). In
those systems, direct comparisons between laboratory and geo-
logical scales have been made, including identification of a
transition between liquid-like, suspended, and solidified, sedi-
mented, phases that is seen when settling and flow speeds
become comparable.
To assess the extent of applicability of our terrestrial exper-
iments using reduced–density grains to Martian geological
flows, in Table 1 we display values that appear in Eqs. 1 and 2
alongside the ratio Rst ⫽ Vs兾Vt for our laboratory experiments
using (i) hollow beads, (ii) some terrestrial debris flows, and (iii)
some Martian debris flows. We reiterate that Eqs. 1 and 2 are
intended only as scaling relations, and the values provided in the
table are only representative approximations. Nevertheless, the
data illustrate that both our laboratory beads and Martian
geological grains can settle more slowly than they flow (i.e.,
Rst ⬍ 1), whereas terrestrial geological grains typically do not
(Rst ⬎1).

substrate material. Such a situation is seen in the rivulets in Fig.
1a. Air flow near the container wall visibly erodes material, as
evidenced by the fact that a substantial groove is produced
between the wall and the granular bed: this groove is highlighted
(arrows) in Fig. 2f, which presents an enlarged view of the
upstream edge of the channel of Fig. 3c. The groove is smoothly
polished and is observed to enlarge steadily over time through
erosion.
Evidently, either in flow initiated by a collapse or in steady
flow originating near a smooth boundary, the resulting upstream
structures exhibit steep walls similar to those that have been
remarked (1) to be a signature of Martian gully flows. From our
investigations of low Rst materials, it appears that steep upstream
walls may be far from coincidental to Martian gullies, but rather
may serve as a source of energized particles without which
fluidization could not begin.
Scaling
Conditions and scales are enormously different between labo-
ratory and Martian geological dimensions. Moreover, Martian
PHYSICS
In conclusion, several features in Martian gullies that are
suggestive of liquid water flow have been revealed in past and
ongoing Mars missions. We have not attempted to comprehen-
sively duplicate all of these features in the laboratory; however,
from our limited studies, it appears that some features that have
been attributed to liquid flow may in fact also occur in dry grains
under conditions that prolong fluidization, as can be anticipated
under reduced gravity. In particular, when we reduce the settling
ratio, Rst, by using low-density beads, we have found (i) that
grains can flow, slosh, and reflect from boundaries in a liquid-
like manner, (ii) that flowing grains can erode elongated chan-
nels, and (iii) that remnant structures left by these flows can
include hillocks, steeply banked alcoves, sinuous streams, mul-
tiply branched inlets, depositional cones, and variegated aprons
that resemble those reported in Martian gullies. The duplication
of these features under dry conditions does not imply that liquid
water does or does not exist in Martian gullies, nor does it speak
to growing evidence of subsurface water, either liquid or solid (4,
33–36). In addition, in light of the magnificent images and data
recently returned by National Aeronautics and Space Admin-

Shinbrot et al. PNAS 兩 June 8, 2004 兩 vol. 101 兩 no. 23 兩 8545

istration’s (NASA’s) Mars Rovers and the European Space
Agency’s Mars Express, some of which has provided significant
new evidence of the former presence of liquid water on Mars, we
hasten to reiterate that our results deal specifically with gully
features known to be geologically recent, for which the discor-
dance between geological and climatic evidence is especially
problematic (1, 17). For these features, our results imply that
surface landforms that are diagnostic of liquid water under
terrestrial conditions may not by themselves be diagnostic on
Mars (17).
Interestingly, laboratory investigations of granular f lows at
reduced pressure performed elsewhere (37, 38) indicate that
granular f lows are significantly inf luenced by their surround-
ing atmosphere until pressures fall below 5–10 Torr; at much
below this pressure, grains travel nearly ballistically and do not
settle in a Stokes-like manner (i.e., as in Eq. 1), whereas, above
this value, grains remain f luidized as described. By compari-
1. Treiman, A. H. (2003) J. Geophys. Res. 108, 2002JE001900.
2. Malin, M. C. & Carr, M. H. (1999) Nature 397, 589–591.
3. Segura, T., Toon, O. B., Colaprete, A. & Zahnle, K. (2002) Science 298,
1977–1980.
4. Baker, V. R., Carr, M. H., Gulick, V. C., Williams, C. R. & Marley, M. S. (1992)
in Mars, eds. Kieffer, H. H., Jakosky, B. M., Snyder, C. W. & Matthews, M. S.
(Univ. Arizona Press, Tucson, AZ), pp. 493–554.
5. Malin, M. C. & Edgett, K. S. (2000) Science 288, 2330–2335.
6. Costard, F., Forget, F., Mangold, N. & Peulvast, J. P. (2001) Science 295,
110–113.
7. Malin, M. C. & Edgett, K. S. (2002) J. Geophys. Res. 106, 23429–23570.
8. Christensen, P. R. (2003) Nature 422, 45–48.
9. Hoffman, N. (2000) Science 290, 711–712.
10. Saunders, R. S. & Zurek, R. W. (2000) Science 290, 712 (lett.).
11. Doran, P. T. & Forman, S. L. (2000) Science 290, 711–712 (lett.).
12. Knauth, L. P, Klonowski, S. & Burt, D. (2000) Science 290, 711 (lett.).
13. Kieffer, H. H., Jakosky, B. M. & Snyder, C. W. (1992) in Mars, eds. Kieffer,
H. H., Jakosky, B. M., Snyder, C. W. & Matthews, M. S. (Univ. Arizona Press,
Tucson, AZ), pp. 1–33.
14. Christensen, P. R., Bandfield, J. L., Bell, J. F., III, Gorelick, N., Hamilton, V. E.,
Ivanov, A., Jakosky, B. M., Kieffer, H. H., Lane, M. D., Malin, M. C., et al.
(2003) Science 300, 2056–2060.
15. Bird, R. B., Stewart, W. E. & Lightfoot, E. N. (2001) Transport Phenomena
(Wiley, New York), 2nd Ed., pp. 61, 179.
16. Tritton, D. J. (1988) Physical Fluid Dynamics (Oxford Univ. Press, Oxford), pp.
33–34.
17. Leovy, C. B. (2003) Nature 424, 1008–1009.
18. Tillman, J. E. (1988) J. Geophys. Res. 93, 9433–9451.
19. Jaeger, H. M., Nagel, S. R. & Behringer, R. P. (1996) Rev. Mod. Phys. 68,
1259–1273.
20. Adomeit, P. & Renz, U. (2000) Int. J. Multiphase Flow 26, 1183–1208.
21. Pasquero, C., Provenzale, A. & Spiegel, E. A. (2003) Phys. Rev. Lett. 91, 054502.
8546 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308251101
son, atmospheric pressures on Mars, although variable, aver-
age close to 8 Torr (39, 40). Thus, it appears that Mars’
atmosphere serendipitously lies very close to the transition
pressure at which the behaviors we describe begin to come into
play (17). We speculate that the tendency for liquid-like f low
features to be seen on colder, poleward, sides of midlatitude
craters and not on warmer, opposing, sides (5, 7) may in fact
be a ref lection of reduced atmospheric densities and pressures
on the warmer sides rather than being due to enhanced liquid
formation on the colder sides. Definitive testing of this spec-
ulation and of the presence of liquid water itself must await
future explorations focused on distinguishing between true
liquid-based f lows and granular surrogates on Mars.
We thank Carlos Caicedo and Cyrus Chi for dedicated technical
assistance and Conway Leovy, Allan Treiman, and Jerry Gollub for
insightful remarks. This work was supported by the National Science
Foundation, Division of Chemical and Transport Systems.
22. Tee, S. Y., Mucha, P. J., Cipelletti, L., Manley, S., Brenner, M. P., Segre, P. N.&
Weitz, D. A. (2002) Phys. Rev. Lett. 89, 054501.
23. Kunii, D. & Levenspiel, O. (1991) Fluidization Engineering (Butterworth–
Heineman, Oxford, U.K.).
24. Genovese, F. C., Watson, P. K., Castellanos, A. & Ramos, A. (1997) in Powders
and Grains 97, eds. Behringer, R. P. & Jenkins, J. (Balkema, Rotterdam), pp.
151–154.
25. Tardos, G. I. & Gupta, R. (1996) Powder Technol. 87, 175–180.
26. Sommeria, J., Meyers, S. D. & Swinney, H. L. (1988) Nature 331, 374–376.
27. Walsh, A. M., Holloway, K. E., Habdas, P. & de Bruyn, J. R. (2003) Phys. Rev.
Lett. 91, 104301.
28. Gerald G. J., Ernst, R., Sparks, S. J., Carey, S. N. & Bursik, M. I. (1996) J.
Geophys. Res. 101, 95JB01900
29. Fierstein, J., Houghton, B. F., Wilson, C. J. N. & Hildreth, W. (1997) J.
Volcanol. Geotherm. Res. 76, 215–227.
30. Iverson, R. M. (1997) Rev. Geophys. 35, 245–296.
31. Denlinger, R. P. & Iverson, R. M. (2001) J. Geophys. Res. 106, 553–566.
32. Chehata, D., Zenit, R. & Wassgren, C. R. (2003) Phys. Fluids 15, 1622–1631.
33. Titus, T. N., Kieffer, H. H. & Christensen, P. R. (2003) Science 299, 1048–1051.
34. Kranoppolsky, V. A. & Feldman, P. D. (2001) Science 294, 1914–1917.
35. Mitrofanov, I., Anfimov, D., Kozyrev, A., Litvak, M., Sanin, A., Tretyakov, V.,
Krylov, A., Shvetsov, V., Boynton, W., Shinohara, C., et al. (2002) Science 297,
78–81.
36. Mustard, J. F., Cooper, C. D. & Rifkin, M. K. (2001) Nature 412, 411–414.
37. Pak, H. K., van Doorn, E. & Behringer, R. P. (1995) Phys. Rev. Lett. 74,
4643–4646.
38. Evesque, P. (1990) J. Phys. France 51, 697.
39. Tillman, J. E. (1988) J. Geophys. Res. 93, 9433–9451.
40. Zurek, R. W., Barnes, J. R., Haberle, R. M., Pollack, J. B., Tillman, J. E. &
Leovy, C. B. (1992) in Mars, eds. Kieffer, H. H., Jakosky, B. M., Snyder, C. W.
& Matthews, M. S. (Univ. Arizona Press, Tucson, AZ), pp. 493–555.
Shinbrot et al.