Motions and Relaxations of Confined Liquids

proven to be one of the most b&ng aspects of liquids, in spite of

When a liquid is confined in a narrow gap (as near a cell long-standing interest (15, 16).
membrane, in a lubricated contact between solids, or in a Rapid advances are now being made with the advent of new
porous medium), new dynamic behavior emerges. The experimental tools. The author's intention in writing a review at this
effective shear viscosity is enhanced compared to the bulk, time is to bring out the excitement of a field that is in the throes of
relaxation times are prolonged, and nonlinear responses rapid development.
set in at lower shear rates. These effects are more promi-
nent, the thinner the liquid film. They appear to be the
manifestation of collective motions. The flow of liquids Flow of Confined Liquids
under extreme confinement cannot be understood simply
by intuitive extrapolation of bulk properties. Practical Basic principles. Neglecting surface effects for the moment, recall
consequences are possible in areas from tribology and that liquids can be distinguished from solids as sustaining no
materials processing to membrane physics. deformation at equilibrium. In responding to force, a liquid eases
deformation by flowing at some rate. Shear deformation, in which
one surface slides tangent to another, is especially simple to inter-
pret. At sufficiently low shear rates (shear rate equals velocity divided
FIG. by filrn thickness), flow obeys Newton's law,

C
ONSIDER THE ARMCHAIR EXPERIMENT DEPICTED IN
1. Take a drop of liquid, put it between a ball and a
table-and let the ball fall. Of course the liquid squirts out,
initially rapidly, then slower and slower as the liquid thickness where o is the shear stress (stress is the force divided by area), dyldt
becomes less than the radius of the ball. This problem was solved is the shear rate ( j ) ,and q is the viscosity, which has dimensions of
more than 100 years ago in a classic analysis of Reynolds (1). mass length-' time-'. A conventional unit is the poise (cgs units).
Experimentally, Israelachvili and co-workers s h ~ w e dthat the film The range of liquid viscosities is.enormous. At room temperature
eventually stabilizes at a finite thickness of a few molecular diameters the viscosities of water, honey, and road tar are approximately 0.01,
(2, 3 ) . The liquid film supports the weight of the ball! An extraor- 10, and 10'' poise, respectively. The viscosity of a liquid may be
dinarily large pressure is needed to squeeze out the final few layers thought of qualitatively as its resistance to flow.
of liquid between two solid surfaces. Why? These notions become more subtle when the liquid film is so thin
Liquids tend to organize in strata parallel to a solid boundary, as to be anisotropsurfaces. An example may be seen in Fig. 1; after
provided that the boundary is smooth compared to the molecular the ball has fallen, +e ic in the directions normal to and parallel to
size. A density profile for a liquid of spherical particles is shown
schematically in Fig. 2. The mean local density ,ofliquid is plotted
against the distance between two solid boundaries. The graph is
analogous to the well-known radial distribution function that Fig. 1. Hypothetical experiment
showing that a liquid can s u p p r t a
describes the order around an average molec~llein an isotropic liquid normal force. A liquid droplet is
(4). Close to the boundaries, the liquid density is vanishingly low Liquid placed between a ball and a flat
because no particle can be located precisely there, but the local surface. The graph, in which liquid
density is correspondingly large at a distance of one particle radius thickness is plotted schematically
removed, while again it reaches a minimum one further particle against time after the ball has begun
to fall, shows that the film thickness
radius away, and so forth. This density wave typically propagates remains finite (a few molecular di-
outward for a distance of three to ten particle dimensions. In the mensions), even at equilibrium.
situation depicted in Fig. 1,the liquid supports the weight of the ball Time
when the interfacial regions depicted in Fig. 2 b e p to overlap.
This much is well understood in principle (5, 6).
-
But what is the organization in directions parallel to the surfaces?
Because surfaces extend laterally over macroscopic distances, there
Fig. 2. Schematic diagram of the
are opportunities for interesting degrees of order or disorder in the local density of a liquid of spherical
direction tangent to the surfaces. This fundamental difference from particles, plotted against the dis- $ g
the bulk liquid, which has only recently been investigated (7-14), is tance between two solid bound- 3
discussed below. aries. The tendency to order in lay- -
Apart from questions of structure, what are the dynamics of indicated err parallel to the boundaries is
by decaying oscillations
liquids in intimate contact with a solid boundary? This question has with a period of about a particle
Distance
diameter. This differs from the ra-
The author is in the Materials Research Laboratory and Depamnent of Materials dial distribution function of a bulk liquid in the sense that interesting degrees
Science and Engineering, University of Illinois, Urbana, IL 61801. of order and disorder parallel to the boundary are also possible.

1374 SCLENCE, VOL. 253

 the ultrathin liquid supports an equilibrium normal force. It acts as Fig. 3. Schematic diagram illustra- *
a solid in this direction. tion periodic shear deformation.
The word "shear' signifies sIiding of
What of sliding of one surface tangent to another? Long ago, it one body parallel to another. The
was recognized that it takes more force to squeeze a liquid out of a arrow denotes the shear direction.
\F%\
\ \ I \ I
confined space than it does to shear the film (17). When dealing with
an ultrathin film, analysis shows that if local viscosity coefficients
could be measured, they would vary with distance across a thin
liquid film just as the local density does (18-20). A laboratory shear
experiment averages over the width of the film. Nonetheless it is
meanin@ to define an effective viscosity based on Eq. 1.However, together, fluid drains smoothly until, at a thickness of -50 4
when a liquid film is sufficiently thin, its response to a tangential equilibrium normal forces of alternate attraction and repulsion set in
force is that of a solid. Then one measures rigidity or yield stress. as successive liquid layers are pushed out (3, 32). These are
What determines the effective viscosity of a film that is so thin? manifestations of the inhomogeneous density profile illustrated in
What makes ultrathin liquids turn solid when they are sufficiently Fig. 2. The period of oscillation of the forces is the width of a
thin? Why is solidification suppressed by continuous motion? These methylene segment, -4 A, showing that molecules are aligned
are key problems in this area. preferentially parallel to the boundaries. An interesting consequence
Measuring thepow of ultrathin liquidjlms. The experimental approach of the oscillatory profile of force versus distance is that the same
is to confine a liquid between two parallel plates of single crystals whose normal force is satisfied by several values of the film thickness;
area is vast compared to the thickness between them. This is an ideal therefore, the profile measured in any particular experiment is
condition in which to study flow behavior because it allows one to ultrasensitive to experimental details, such as the rate at which the
produce liquid filmswhose thickness is defined down to a resolution of liquid droplet is squeezed down to a particular distance (24, 29). In
an angstrom-less than the thickness of a single molecule! Fig. 1, there are multiple solutions for the film thickness at which a
The various instruments that have been devised to do this (21-25) ball would stop falling.
are extensions of the surface forces apparatus (26, 27). The surface Linear response. Shear viscosity depends on film thickness. In the
forces apparatus has been described in detail (26, 27). In brief, liquid region of inhomogeneous density in Fig. 1, the effective viscosity is
is confined between two crystals (usually of muscovite mica or thin larger than in the bulk and grows with diminishing thickness. In
films of other materials coated onto muscovite mica) oriented as Fig. 4, the limiting effective viscosity at low shear rate calculated
cylinders at right angles to one another. The distance between the with Eq. 1, q o e is , plotted against f ilm thickness. In this experiment,
surfaces can be controlled from thousands of angstroms down to the crossed cylinders first flattened to form parallel plates at film
molecular dimensions. thickness of -40 A, and therefore this is the first point at which
The innovation to measure shear flow is based on having parallel shear measurements were made. The effective viscosity was already
plate (rather than crossed cylinder) geometry. As the smoothly greater than in the bulk. Drainage to lesser thickness was caused by
curved solid sheets are brought close together, the liquid in the final applying additional net normal pressure. The four data points at the
few molecular layers resists being squeezed out. A soft glue placed least thickness (26 a 1 A) all refer to a film of approximately six
under the mica sheets flattens instead, resulting in a circular area of segmental widths. The adjustments of thickness by 1 to 2 A with
closest approach on each sheet. Optical interferometry of the surface increasing net normal pressure, less than the segmental width, reflect
separation and of the surface contour confirms that the two flattened compressibility of the liquid. The effective viscosity seems to di-
zones are parallel and provides the contact diameter. The diameter of verge.
the flattened area (on the order of lo5 A) is vast compared with the Nonlinear response. The simple flow properties of dodecane in a
thickness of the confined liquid. beaker turn complex when dodecane is confined. The essentially
With this approach, my laboratory has studied viscous dissipation nonlinear shear response under confinement is illustrated in Fig. 5.
and elasticity of confined liquids by using periodic sinusoidal The data refer to a film of thickness 27 A at a modest net normal
oscillations over a range of amplitudes and frequencies (24, 25, pressure (0.12 MPa; - 1.2 a m ) . The maximum viscous force during
28-33). The geometry of the shear deformation is illustrated in Fig. a cycle of oscillation),,C f is plotted against the maximum velocity
3. Other workers have used a different apparatus to study the (urn,).
friction encountered during sliding at a constant speed (23, 34-38). Two immediate conclusions follow from the data in Fig. 5. First,
Early measurements involved dry sliding (21, 22, 39). The technical f,, grew in direct proportion t? v,, at sufficiently low values, but
difficulties of these methods have been discussed elsewhere (40). beyond a modest velocity (400 A s-') it grew more slowly. Second,
the data show that the reducing variable by which to analyze the
viscous force was indeed velocity, not frequency (43) as at smaller
Effective Viscosity of a Simple, Confined strains. This follows because when amplitude and frequency were
varied separately, the viscous force depended only on h e i r product.
Molecular Liquid The characteristic length of this flow problem may safely be taken
Every liquid that has been investigated [a variety of nonpolar
liquids of various sizes, shapes, and flexibilities (24, 28-33)] shows
the following qualitative effects, but findings for dodecane are
emphasized in this article for illustration.
Dodecane, CH,(CH,),oCH,, is a simple flexible chain molecule
12 carbons long. Its bulk shear viscosity at room temperature is 0.01
poise, independent of shear rate up to shear rates > 1010 s-'. The
chain length is -18 A, and the thickness of the chain is -4 A. The Fig. 4. Effective limiting viscosity at
low shear rate, plotted against film 0'
static structure of chains of similar length, confined between parallel thickness, for confined dodecane
plates, has been simulated by computer (41, 42). films at 28°C. Arrows indicate direc- '20 30 40 50
As mica cylinders separated by a dodecane droplet are pushed tion of increasing compression. Thickness (A)

ARTICLES 1375
Fig. 5. Maximum vis- by viscous dissipation, straightforward calculation shows that it was
cous force plotted versus 5 efficiently dissipated because the area of the films was so large
maxlmum velocity dur-
relative to the thickness. Even cursory inspection of raw data such as
ing a cycle of oscillation
for dodecane films of in Fig. 5 shows these effects; they do not depend on analyzing the
thickness 2.7 nm and net 3 data in terms of an effective viscosity. While the characteristic length
normal pressure 0.12 bv which to normalize the velocity might be somewhat less than the
MPa at 28°C. Open cir- total film thickness [one or nvo layers of fluid might be pinned to
cles: amplitude varied
from0.!9tol8Onn,at i each surface during the experiment (13, 16, 44)], such fine-tuning of
1.3 Hz. Filled circles: the analysis would not change the relative numbers analyzed above,
frequency varied from nor their orders of magnitude. The consistent measurements ob-
0.02 to 52 Hz at an am- 0 100 200 300 400 tained in repeated experiments with different liquids would indicate
plitude of 4 0 nm. Dot-
ted line extrapolates the vmax(nm s-')
that the results are general and must be examined at face value.
zone of linear response
(32).
Discussion
to be the thickness of the liquid film (other conceivable length scales Measurements at large deformation. It is at first astonishing to
have similar magnitude; see below). For the same set of data shown scientists who study the flow of liquids that the data depend on
in Fig. 5, Fig. 6 shows the effective viscosity (v,,) calculated with velocity rather than on the excitation frequency. One is not accus-
Eq. 1, plotted against mavimum effective shear rate (j,,) on log-log tomed to this (43). However, the strains customarily investigated
scales in view of the large changes in magnitude. At low shear rate, (strain is shear amplitude divided by film thickness) are on the order
qeKwas constant. As the shear rate increased, extensive decrease of of unity or less. The strains reported here are enormously larger-15
q,, occurred. The rate of this shear thinning follou~edan empirical for the constant amplitude experiments in Fig. 5.
power law, vCE- jCtf-". The power a, always slightly more than In fact, a tendency for frequencv and amplitude shear data to
two-thirds, was not quite constant; at a given film thickness it converge at large strains was already observed long ago in studies of
increased modestly with increasing net normal pressure (32), tend- bulk flow (45). The matter has been relatively neglected since then.
ing toward unity for the thinnest films (33). Ferry has suggested as a rough qualitative interpretation that an
The effective viscosity always exceeded the viscosity of bulk experiment performed with sinusoidal oscillations of huge ampli-
dodecane, but decayed by a factor greater than 100 as the shear rate tude relative to the film thickness is really more like a steady-flow
was raised. Measurements of film thickness failed to detect any experiment, with shear rate averaged over a cycle (46). In other
changes with shear ( ~ A); 1 this puts an upper limit of 4% on words, most of the plot in Fig. 5 is essentially a non-Newtonian flow
possible shear-induced changes in the mean liquid density. Measure- plot.
ments were reversible when the velocity was raised and lowered. Enhanced viscosity and slow relaxation. Physically, a nonlinear
This pattern of flow behavior is also characteristic of other simple ~iscositysets in when the experiment distorts the structure of the
nonpolar liquids. For illustration, also shown in Fig. 6 is an fluid. This happens when the experimental time scale is faster than a
experiment with OMCTS, octamethylcyclotetrasiloxane, a silicone characteristic time of Brownian motion. In Fig. 6, the onset of
oil of different chemical shape as well as chemical composition. For nonlinear response is readily estimated by extrapolating the plateau
some years, OMCTS, [Si(CH,),-0-I,, a compact-shaped (but and power-law zones until they cross. The conclusion that the
flexible) ring molecule with diameter -9 A has been a reference nonlinear response sets in atye, > 20 spl for dodecane implies that
liquid in studies of liquid microstructure (2, 5). The stronger the longest system relaxation time was -1120 = 0.05 s-more than
intensity of interaction with the surface and the different molecular 10' times slower than in the bulk. This is a central theme: relaxation
packing probably account for the different quantitative response as is much slower than in the bulk liquid.
compared to dodecane; a detailed discussion is not warranted at this Further insight into the molecular origin of slow relaxation comes
time. For the present we emphasize the qualitative agreement. The when one calculates, from the rate of increase in q,, with increasing
effective viscosity was controlled by the velocity or normalized net normal pressure, the activation volume (AT/,,,) for flow (32).
velocity of the experiment and again showed extensive shear thin- The finding is that AV,,, is tremendously larger than for flow in the
ning. Thus this pattern of behavior held even for a molecule that is bulk. In the zone of linear viscous response, the activation volume
not a linear chain. corresponds to the volume of -200 molecules; this stands in sharp
What of conceivable artifacts? The observation of reversibility contrast to difision in the bulk liquid, where AV,,, amounts to the
rules out shear-induced chemical degradation. As for heat generated volume of only a single segment of the molecule (47). However,
under confinement, AV',,,, decreases with increasing ye, in the zone
of nonlinear response (32). These findings indicate that the unit
Fig. 6. Double logarith-
event in shear flow was collective and that shear thinning involved
mic graph of effective the breaking up of some structure. The activation volume in the
viscosity as a function of zone of linear response is an estimate of the correlated volume that
strain rate. Circles: exists even in the absence of flow.
dodecane film specified Explanations are still conjectural. If collective motions are respon-
in Fig. 4. Triangles:
OMCTS film of thick- sible for the high viscosity and slow relaxation, this might still be
ness 2.7 nrn and net nor- consistent with rapid motion of individual molecules. Indeed, the
mal pressure 0.14 IMP^. existing computer simulations of confined liquids indicate that the
Open symbols: ampli- difhsion of spherical molecules is only moderately slowed down by
tude varied at constant
frequency. Filled sym- confinement (see critical discussion below). The problem then
bols: frequency varied at becomes to understand the origin of collective motions and how
constant amplitude ( 3 2 ) . they might be broken up by externally driven flow.

SCIENCE. VOL. 253

 If this were a strictly two-dimensional (2-D) fluid, then long-time ment with laboratory experiments has been asserted (13).
tails of velocity correlations resulting from hydrodynamics could However, the solidification phenomenon is general. It is observed
come into play, leading to divergence of the 2-D shear viscosity at for every liquid investigated (23, 24, 28-30, 33-38). In particular,
low frequency (4, 48) and its logarithmic decay with increasing shear control experiments show that the ability of material to form bulk
rate (49). However, the deformation in these experiments was not crystals is not necessary. A glass-formingliquid (atactic polyphenyl-
applied within a 2-D plane. Moreover, long-time hydrodynamic methylsiloxane; glass transition temperature -20°C) also solidifies
tails are not expected to be prominent at high density (50) as in the (28).
\ ,

present experiments. Alternatively, long-lived orientational correla- Finally, the yield stress depends strongly on experimental history
tions (51) or density fluctuations (52) could arise because a state of (28, 29). The film thickness at solidification depends on the rate at
two dimensions is approached. In fact, the finding that veff - which the liquid drop is thinned (29). The stress grows over
jeff-2'3 is predicted by a cluster model of shear thinning (52). A remarkably long times-minutes to hours, depending on the liquid
provocative recent scanning tunneling microscopy (STM) study of (28, 29). This strengthening can be seen in an experiment with
alkanes at the liquid-graphite surface suggests that microcrystalline OMCTS (Fig. 7). The yield stress on first measurement was -3.5
domains actually form in the monolayer immediately at the surface MPa. but this value nearlv, tripled over a 10-min interval. The
i

(53). Further investigations are needed to clarify these striking sluggish increases in the yield stress imply that whatever the
effects. A problem in assessing 2-D explanations is that rigorous structure may be of the solidified state, apparently it is MI of defects.
predictions for systems that are close to-but not quite-in two The following picture has been proposed tentatively (28, 29). As
dimensions remain to be worked out. discussed above, collective motions appear to be responsible for the
high viscosity and slow relaxation under confinement. Increasing
confinement may slow down the relaxation to the point that, finally,
flow must be mechanically activated over the %me scale of an
Solidification experiment.
When it is sufficiently thin, a confined liquid solidifies, in the sense This idea of mechanical activation is consistent with the magni-
that it will not allow shear until a critical shear stress (yield stress) is tude of the yield stress; interpreted as an energy density, the Geld
exceeded. This occurs especially for films less than four molecular stress is in the range of 0.3 to 1 kT per molecule (28) (k is
dimensions thick (23, 24, 28-31, 34-38). It is tempting to attribute Boltzmann's constant and T is the absolute temperature). If one
the phenomenon simply to strong adsorption. However, even when accepts the rule of thumb that the yield stress is onthe order of 0.01
the overall response is liquid (four to ten molecular dimensions),the to 0.1 times the shear modulus, the estimated shear modulus
response can have an elastic component in addition to a viscous resembles that of bulk glasses. Loss of fluidity may reflect vitrifica-
component (32, 33). This suggests that the transition to solidifica- tion imposed by the liquid's confinement.
tion, with decreasing film thickness, could be continuous.
Perhaps the parallels to what happens when gases condense onto
solids are not widely enough appreciated. Conaensation of noble Roles of Surface Composition and Structure
gases on graphite is well studied (54). Theoretical understanding of
the origin of the solid and liquid monolayer phases that form- Until recently the alarming possibility existed that all of these
commensurate or incommensurate with the underlying solid-is findings might be specific to mica as the solid boundary. There were
well developed (54). The present case is decidedly more complicat- good reasons to use mica, which forms a transparent single crystal
ed, of course, not only by the effects of externally driven motion, but free of steps over a large area, but the question remained of what
also by the complicated molecules and substrates that are of concern. would happen if the surface lacked long-range periodicity and
In future studies, ideas in these two fields of investigation, which adsorption were weaker.
arose independently, hopefully may stimulate one'another. Recently methods were developed to blanket mica with a securely
Molecular interpretations of the solidification phenomenon are attached, self-assembled organic monolayer (55, 56). The surface
controversial and may remain so until the molecular packing is composition, composed of chemically reacted octadecylsilane
measured directly. Such experiments would be difficult for reasons (OTE) chains, is an array of closely packed methyl groups.' The
discussed below. However, the experimental situation is simpler methyl groups are amorphous over distances greater than 100 i% in
than might be feared. One might imagine that the state of an films of this type (57); liquids could not crystallize epitaxially onto
ultrathin film might depend on everything conceivable+specially these monolayers. In addition, the surface energy is only 22 mJ m-2
the direction of shear relative to crystalline lattices of the solid (56), considerably less than that of freshly cleaved mica (200 to 400
boundaries and the relative orientation of these two crystalline mJ mP2); liquids are expected to adsorb more weakly than onto
lattices. The search for universal behavior would be a lost cause. mica. Thus, substantially weaker coupling to the surface is expected
Indeed, this may be so for the very thinnest films (34), but than in the case of mica.
fortunately not when the film thickness is larger.
What is presently understood about solidification can be briefly
summarized. First, for the very thinnest films, the yield stress
increases with diminishing film thickness (23). Such yield stress
behavior is a familiar fact of life, static friction. However, solidifi-
Fig. 7. Yield stress of
cation is suppressed by continuous shear (23, 24, 28-38), as occurs octamethylcyclotetrasi-
in dynamic friction. loxane (OMCTS) plot-
Second, computer simulations show that a liquid of spherical ted against elapsed time
particles tends to crystallize epitaxially at a solid surface. The during repetitive cycles
of oscillation at 1.3 Hz
crystalline layers, typically one or two particle diameters thick (7, and at 23°C. Film thick-
10-14), can be torn apart and kept disordered by shear (13, 14). The ness was 9 A, and net 0
inference is that the transition from static to dynamic friction with normal pressure was 0.8 0 10 20 30
increasing shear force could reflect shear-induced melting. Agree- MPa (28). Time (min)

20 SEPTEMBER 1991 ARTICLES 1377

 Fundamentally the same shear viscosity was observed for dode- zene) adsorbs to porous glass (65). However these differences
cane confined between OTE layers (Fig. 8) as obtained with mica. disappear for a nonpolar liquid or if the glass is treated to be
The thickness of dodecane is 32 A.
Once again, effective viscosity, nonpolar (65). In fact, reorientation of supercooled oxygen in
q , , is plotted against maximum effective shear rate, j,, on porous glass is actually faster in small pores (22 A diameter) than in
logarithmic axes. Despite quantitative differences from the situation larger pores (66, 67). Greater restrictions on thermodynamic fluc-
with mica, in both cases qeEwas much greater than in the bulk and tuations in the smaller pores were invoked to explain this paradox-
decreased by a factor greater than 100 as j,, increased. These ical finding (66, 67).
experiments show that the behavior described above resulted essen- Very recently, evidence was found (68) that difision in ultra-
tially from geometrical confinement. small pores is of the form expected in a 2-D geometry (4, 69, 70).
Analysis of deuteron relaxation rates measured by NMR showed
that dipole-dipole correlations persist to considerably longer times
Other Probes of Surface Dynamics A
in small pores (down to 34 diameter) thah they do in the bulk
(68). With increasing pore size, the transition from 2-D to 3-D
The difficulty in studying the solid-liquid interface is that it is relaxation was also inferred (68, 71). Extensions of these studies
buried between two bulk phases and that there is so little material may allow one to set criteria for when a particular physical system
there to study. It takes only -1 x lop7 g cmp2 of organic liquid to is truly 2-D.
form a monolayer. Therefore, experiments with the sensitivity to The other major approach at present is simulation by computer.
measure dynamics are few. Phenomena can be simulated that would be impossible to measure
When two crossed cylinders of mica are squeezed together, an experimentally-for example, the trajectories and positions of indi-
effective viscosity can be inferred from the rate at which the fluid vidual molecules as a function of their positions in the liquid film.
drains out (58-60). Results can be interpreted as if one or two layers Indeed, most of what is known explicitly about equilibrium liquid
of fluid molecules are immobilized on each surface on the time scale structure at the solid-liquid interface comes from simulations (6-14).
of the experiment (58, 61), which is consistent with the findings Results have begun to emerge regarding difision and flow of
described above. The spectacular enhancement of viscosity described spherical particles (liquid argon) in planar confined geometries. The
above is not observed, however (58, 59). The essential physical diffision rates parallel to the surfaces are found to be retarded
differences may be that the film thickness in those experiments was relative to the bulk, but only weakly (18-20, 44, 72, 73). As noted
many times larger than in the shear experiments described above. If above, this could be reconciled with experiments if the relaxation
so, the different findings emphasize the extreme consequences for time that controls the difision of individual particles is not the same
dynamics when fluid is confined within a narrow gap. as 'controls the viscous response.
In another experimental approach, liquids have been confined Computations about dynamics are so time-consuming, however,
within porous glasses in order to increase the surface area moni- that the longest relaxation time that can be studied with ptesent
tored by the experiment. In this curved geofietry, the melting computers is only -1 x lopz s. In order to exceed the rate of
temperature and the glass transition temperature are lower than in thermal motion, the lowestrate of externally driven flow that can be
the bulk, not higher as in a planar geometry; the reasons are still studied is -10" spl. Closer comparison with experiments should
not completely understood, especially with regard to the shift in be possible when the systems that can be modeled are large enough
the glass transition temperature (62, 63). In view of these differ- to look for the collective relaxation effects suggested above.
ences in molecular packing, how liquid dynamics measured in a
porous medium relates to liquid dynamics in a planar geometry is
not yet clear conceptually.
Rapid developments are expected based on spectroscopic meth-
Future Directions
ods, which have been used so far by only a few workers. Nuclear One of the most baffling aspects of liquids has long been the
magnetic resonance (NMR) measurements of hydrocarbon mobility solid-liquid interface, despite its practical importance in areas such as
in zeolites show that the rates of conformational changes are friction and lubrication, flow of colloidal particles past one another,
significantly slower than in the bulk (64). Picosecond measurements liquids in porous geological formations, the wall-stick boundary
of reorientation times support this, showing moderately slower condition of classical continuum hydrodynamics, and many materi-
relaxation times (by a factor of 3) when a polar molecule (nitroben- als processing operations. In all of these situations the order
imposed by the surface contends with the usual isotropy of the
liquid, and the thermodynamic state depends not only in the usual
Fig. 8. An experiment 5 fashion on temperature and pressure, but also on the strength of
with dodecane confined surface-fluid interactions.
between two self-assem- One reason for slow progress is simply that liquids in the bulk
bled monolayers of con- ,4 - were traditionally studied by chemical physicists, while liquids at
densed OTE (see text).
surfaces were studied by engineers. There was little interaction
The alkyl chains of the
monolayers were termi-
nated by methyl groups.
- g
f 3-
F
o 0 0; o os between these communities and their common problems tended to
be neglected. Another reason is the paucity of experimental methods
Changes in the effective "f %tI with which to probe this interface that is buried between two
viscosity are plotted, on
double logarithmic axes,
* - (1 Hz)
*a? - condensed phases. This is starting to change with the advent of new
as a function of shear experimental tools. The preceding account has necessarily been
rate at 28°C. Open cir- 1 limited to a few examples. The field is developing rapidly. Presently
cles: amplitude varied at 0 1 2 3 it is poised at the stage where a growing body of data is begging for
constant frequency (1.3 definitive explanation.
Hz). Filled circles: fre- Log ?,fi (S-')
quency varied at con- The patterns described in this article appear to be general, but the
stant strain (strain is the rationofamplitude to film thickness, in this case, 15 magnitudes likely depend on particular properties of the liquid (the
(33). Film thickness was 32 A, and net normal pressure was 0.18 MPa. molecular size, shape, polarity, and chemical composition), as well
SCIENCE, VOL. 253
 as on the strength of surface-fluid attraction. This has been consid- 22. J. N. Israelachvili and D. Tabor, Wear 24, 386 (1973).
23. J. N. Israelachvili, P. M. McGuiggan, A. M. Homola, Science 240, 189 (1988).
ered based on general statistical mechanical considerations (6), but 24. J. Van Alsten and S. Granick, Phys. Rev. Lett. 61, 2570 (1988).
system-specific chemical effects are still poorly understood. 25. J. Peachey, J. Van Alsten, S. Granick, Rev. Sci. Instrum. 62, 463 (1991).
This interdisciplinary area should be greatly stimulated as more 26. D. Tabor and R. H. S. Winterton, Proc. R. Soc. A 312, 435 (1969).
27. J. N. Israelachvili and G. E. A d a m s , / Chem. Soc. Faraday Trans. II 74, 975
theorists seek to explain the data that has accumulated. Contact is (1978).
waiting to be made with modern ideas of liquid-state physics, 28. J. Van Alsten and S. Granick, Langmuir 6, 214 (1990).
wetting phenomena, and systems trapped far from equilibrium. 29. , Macromolecules 23, 4856 (1990).
30. , Tribol. Trans. 33, 436 (1990).
Other progress is likely with new experimental techniques. The 31. G. A. Carson, H.-W. Hu, S. Granick, ibid., in press.
potential of spectroscopic and diffraction investigations has been 32. H.-W. Hu, G. A. Carson, S. Granick, Phys. Rev. Lett. 66, 2758 (1991).
33. G. A. Carson, H.-W. Hu, S. Granick, unpublished experiments.
emphasized. New experimental approaches are needed to resolve the 34. M. L. Gee, P. M. McGuiggan, J. N. Israelachvili, A. M. Homola, J. Chem. Phys.
questions raised above of shear ordering and disordering. 93, 1895 (1989).
The approaches emphasized in this article also apply to more 35. A. M. Homola, J. N. Israelachvili, M. L. Gee, P. M. McGuiggan, / . Tribol. I l l ,
675 (1989).
complicated liquids. For example, liquids of highly intertwined 36. P. M. McGuiggan, J. N. Israelachvili, M. L. Gee, A. M. Homola, Mater. Res. Soc.
polymer chains exhibit more prominent elasticity—while still re- Symp. Proc. 140, 79 (1989).
maining liquid—than they do in the bulk (33). Fluid mixtures 37. A. M. Homola, J. N. Israelachvili, P. M. McGuiggan, M. L. Gee, Wear 136, 65
(1990).
[containing, for example, polymers (74), block copolymers, deter- 38. A. M. Homola, H. V. Nguyen, G. Hadziioannou, J. Chem. Phys. 94, 2346
gents, and proteins] and aqueous media are also obvious next steps. (1991).
39. J. Van Alsten and S. Granick, Tribol. Trans. 32, 246 (1989).
The question of biological systems is open^ A tantalizing hint of 40. See appendix of H.-W. Hu and S. Granick, Macromolecules 2 3 , 6 1 3 (1990); also see
connections to be made with membrane physics is suggested by the appendix of (25).
experiments, discussed above (Fig. 8), in which the solid boundary 41. M. Vacatello, D. Y. Yoon, B. C. Laskowski, J. Chem. Phys. 93, 779 (1990).
42. M. W. Ribarsky and U. Landman, ibid., in press.
consisted of hydrocarbon chains. Of course, only the simplest 43. J. D. Ferry, Viscoelastic Properties of Polymers (Wiley, New York, ed. 3, 1980).
nonpolar liquids have been studied to date. In broaching the 44. J. Koplik, J. R. Banavar, J. F. Willemsen, Phys. Rev. Lett. 60, 1282 (1988).
behavior of water, especially of water close to proteins, nucleic acids, 45. W. Philippoff, Trans. Soc. Rheol. 10, 317 (1966).
46. J. D. Ferry, private communication.
and membranes, experiments in which solid organic surfaces of 47. T. Vardag, F. Bachl, S. Wappman, H.-D. Ludemann, Ber. Bunsenges. Phys. Chem.
tailored chemical composition are used may result in direct biolog- 94, 336 (1990).
48. B. J. Alder and T. E. Wainwright, Phys. Rev. A 1, 18 (1970).
ical implications. 49. G. P. Morriss and D. J. Evans, ibid. 32, 2425 (1985).
50. A. J. C. Ladd, W. E. Alley, B. J. Alder, J. Stat. Phys. 48, 1147 (1987).
REFERENCES AND NOTES 51. For a review, see D. R. Nelson, Phase Transitions 7, 1 (1983).
52. J. Douglas, private communication.
1. O. Reynolds, Trans. Roy. Soc. 177 (Part 1) 1886, 157 (1886). 53. J. P. Rabe and S. Buchholz, Phys. Rev. Lett. 66, 2096 (1991).
2. R. G. Horn and J. N. Israelachvili, J. Chem. Phys. 75, 1400 (1981). 54. For a review, see K. J. Strandburg, Rev. Mod. Phys. 60, 161 (1988).
3. H. K. Christenson, D. W. R. Gruen, R. G. Horn, J. N.. Israelachvili, ibid. 87,1834 55. G. Carson and S. Granick, J. Mater. Sci. 5, 1745 (1990).
(1987), and references therein. 56. C. Kessel and S. Granick, Langmuir 7, 532 (1991).
4. For a review, see J.-P. Hansen and I. R. McDonald, Theory of Simple Liquids 57. C. E. D. Chidsey, G.-Y. Liu, P. Rowntree, G. Scoles, J. Chem. Phys. 9 1 , 4421
(Academic Press, New York, ed. 2, 1986). (1989).
5. For a review of experiments, see H. K. Christenson, J. Dispersion Sci. Technol. 9, 58. D. Y. C. Chan and R. G. Horn, ibid. 83, 5311 (1985).
171 (1988). 59. J. N. Israelachvili, J. Colloid'interface Sci. 110, 263 (1986).
6. For references to calculations and simulations of particles between two walls, see R. 60. J. Van Alsten, S. Granick, J. N. Israelachvili, ibid. 125, 739 (1988).
Kjellander and S. Sarman, Mol. Phys. 70, 215 (1990). 61. R. G. Horn, S. J. Hirz, G. Hadziiouannou, C. W. Frank, J. M. Catala, J. Chem.
7. J. E. Lane andT. H. Spurling,^4ws*.y. Chem. 29, 2103 (1976); Chem. Phys. Lett. Phys. 90, 6767 (1989), and references therein.
67, 107 (1979). 62. C. L. Jackson and G. B. McKenna, ibid. 93, 9002 (1990).
8. F. F. Abraham, J. Chem. Phys. 68, 3713 (1979). 63. , J. Non-Cryst. Solids 1 3 1 - 1 3 3 , 221 (1991).
9. I. K. Snook and W. van Megen, ibid. 72, 2907 (1980). 64. B. G. Silbernagel, A. R. Garcia, J. M. Newsam, R. Hulme, J. Phys. Chem. 93,
10. M. Schoen, D. J. Diestler, J. H. Cushman, ibid. 87, 5464 (1987). 6506 (1989).
11. C. L. Rhykerd, Jr., M. Schoen, D. J. Diestler, J. H. Cushman, Nature 330, 461 65. J. Warnock, D. D. Awschalom, M. W. Shafer, Phys. Rev. B 34, 475 (1986).
(1987). .' ' 66. ,Phys. Rev. Lett. 57, 1753 (1986).
12. M. Schoen, C. L. Rhykerd, D. J. Diestler, J. H. Cushman, Science 245, 1223 67. D. D. Awschalom and J. Warnock, Phys. Rev. B 35, 6779 (1987).
(1989). 68. M. Mackowiak, G. Liu, J. Jonas, J. Chem. Phys. 93, 2154 (1990).
13. P. A. Thompson and M. O. Robbins, Phys. Rev. A 4 1 , 6830 (1991); Science 250, 69. J. Tabony and J. P. Korb, Mol. Phys. 56, 1281 (1985).
792 (1990). 70. J. P. Korb, M. Winterhalter, H. M. McConnell, J. Chem. Phys. 80,1059 (1984).
14. M. Lupkowski and F. van Swol, J. Chem. Phys. 95, 1995 (1991). 71. G. Liu, M. Mackowiak, Y. Li, J. Jonas, Chem. Phys. 149, 165 (1990).
15. J. C. Maxwell, Philos. Trans. R. Soc. 157, 49 (1867). 72. P. A. Thompson and M. O. Robbins, Phys. Rev. Lett. 63, 766 (1989).
16. S. Goldstein, Modern Developments in Fluid Dynamics (Clarendon Press, Oxford, 73. I. Bitsanis, S. A. Somers, H. T. Davis, M. Tirrell, J. Chem. Phys. 93, 3427 (1990).
1938), vol. 2, pp. 676-680. 74. J. Klein, D. Perahia, S. Warburg, Nature 352, 143 (1991).
17. Committee for the Study of Viscosity of the Academy of Sciences at Amsterdam, 75. It is a privilege to acknowledge the contributions of outstanding co-workers: J.
Second Report of Viscosity and Plasticity (Nordemann, New York, 1938). Van Alsten, G. Carson, H.-W. Hu, and J. Peachey. I am grateful to J. Douglas, J.
18. I. Bitsanis, J. J. Magda, M. Tirrell, H. T. Davis, J. Chem. Phys. 87, 1733 (1987). D. Ferry, N. M. Kostic, K. Schweitzer, and F. van Swol for thoughtful criticisms
19. T. K. Vanderlick and H. T. Davis, ibid., p. 1791. of a draft of this manuscript. Supported by the U.S. National Science Foundation
20. I. Bitsanis, T. K. Vanderlick, M. Tirrell, H. T. Davis, ibid. 89, 3152 (1988). through the Tribology Program and through the Materials Research Laboratory at
21. A. I. Bailey and J. S. Courtney-Pratt, Proc. R. Soc. A 227, S00 (1955). the University of Illinois.

20 SEPTEMBER 1991 ARTICLES 1379