You are on page 1of 3

In the Classroom

Entropy, Disorder, and Freezing


Brian B. Laird
Department of Chemistry, University of Kansas, Lawrence, KS 66045; *laird@pilsner.chem.ukans.edu

The entropy, S, is one of the most important concepts about because our intuition about this relationship has been
in the thermodynamics of chemical systems, but it is also the developed primarily at low density where packing is not
most difficult one to understand and to teach. Typically, in important. In the next section, the issues of packing, structural
introductory chemistry courses, entropy is introduced as a disorder, and entropy at high density are discussed.
measure of‘ “disorder” or “randomness”. Later, this view is
often justified by an appeal to the Boltzmann view of entropy Packing at High Density: Suitcases and Hard-Spheres
as a measure of the number of microstates, Ω, consistent with
a given macroscopic state; that is, A familiar macroscopic example of a system in which the
maximum entropy state at high density is a spatially ordered
S = k ln Ω (1) state, as opposed to a disordered one, is the suitcase-packing
where the constant of proportionality, k, is Boltzmann’s problem: Imagine packing a suitcase for a trip. If the trip is a
constant. The argument is then made that increasing the short one, relatively few items are required and, from experi-
“disorder” in a system corresponds to an increase in Ω, often ence, when the total volume of items to be packed is much
with appeals to illustrative examples from the macroscopic less than the volume of the suitcase we know that the easiest
world such as the probability of various poker hands and the way to pack is to randomly toss the items in and shut the
tendency of children’s rooms to become messy. suitcase. Putting the items to be packed in some “ordered”
The Boltzmann entropy can be interpreted within the arrangement requires extra work and the order would most
context of information theory as a measure of “lack of infor- probably be destroyed during transport. We put this into
mation”—the greater the number microstates consistent with statistical mechanical terms by saying that there are far more
a system in a given macroscopic state, the less information is ways in which the suitcase can be packed in a “disordered”
available about the precise microscopic state of that system arrangement than in an “ordered” one; that is, the disordered
at a given instant. (If there is only one microscopic state, say state has a higher entropy than the ordered one. This low-
for a perfect crystal at T = 0, then there is no lack of infor- density case is consistent with the standard paradigm.
mation about the precise microscopic state and the entropy Now, suppose instead that the trip is a long one and
is zero.) This is, of course, a statistical mechanical view. many items are required to be packed in the same suitcase.
Within the context of thermodynamics, one can similarly In this case the volume of the objects to be packed is on the
define entropy in terms of “lack of constraint” (1) (here a order of the total volume of the suitcase. This would be a
constraint is any restriction, internal or external, placed on high-density system. From experience, one knows that if the
the number or ranges of the independent macroscopic variables items to be packed are randomly tossed into the suitcase it is
that are necessary to describe the thermodynamic state of the impossible to shut the case. Such a configuration is then in-
system). One could, of course, define disorder as “lack of compatible with the volume constraints of the suitcase. On
information” or “lack of constraint”, but this is (i) unnecessary, the other hand, if the contents are packed in a neat and
given that more precise descriptors exist (as discussed above), ordered arrangement, the suitcase can be closed without
and (ii) undesirable, since the concept of disorder is so closely difficulty. We could say then that the number of ordered
associated with structural disorder that attempting to generalize “microstates” (i.e., arrangements of suitcase contents) that are
the concept can lead to significant misunderstanding in the consistent with the given “macroscopic” constraints (i.e., the
classroom, for both students and educators. Some of the prob- fixed volume of the suitcase) is greater than the number of
lems associated with the use of entropy as a measure of struc- corresponding disordered states; therefore, an ordered arrange-
tural disorder have been discussed earlier in these pages (2). ment in our (high-density) suitcase can be said to have a
The problems in the use of the term disorder to describe higher entropy, in the Boltzmann sense, than a structurally
entropy are perhaps most acute in the study of the freezing disordered one. The disorder-to-order transition as the suitcase
transition. Namely, there exist situations where, given a liquid and density is increased can be said to be purely entropy driven,
a crystal of the same material subject to the same thermodynamic since energetic concerns do not enter into the analogy. (The
conditions, the crystal phase has a higher entropy than the concept of purely entropy-driven transitions is not new and
(metastable) liquid. This is in direct contradiction to the usual has been discussed in the physics literature by Frenkel [3].)
assumption regarding entropy and the commonly held concept A classic molecular model that illustrates this effect is a
of disorder (i.e., structural or topological disorder) because system consisting of hard spheres of diameter σ. The interatomic
the crystal exhibits long-range order, which is absent in the interaction energy, v(r), of two such particles separated by a
fluid phase. So if this is true, then the usual assumption of distance r is given by
entropy and its relation to structural disorder would not only v(r) = ∞; r<σ
be irrelevant in determining entropy differences, it could even (2)
v(r) = 0 ; r ≥σ
give the wrong answer! Such situations can indeed arise at
high density when packing considerations dominate the This interaction is a cartoon of real interactions between atoms
thermodynamics, and the violation of the conventional and molecules; nevertheless, it is an important model system
wisdom about the relationship of entropy to disorder comes because it has been well established that the structure and

1388 Journal of Chemical Education • Vol. 76 No. 10 October 1999 • JChemEd.chem.wisc.edu


In the Classroom

dynamics of simple liquids at high density are determined relative to that of the fcc crystal as a function of reduced num-
almost entirely by the repulsive part of the interaction poten- ber density, ρ* = (N/V )σ 3, at fixed (arbitrary) temperature.
tial. (The effects of the attractive part of the interaction can (The data to construct this figure were taken from Table III
be accounted for by treating them as a small perturbation of ref 5.) From this figure, it is seen that, at high density, the
about the reference repulsive part, which, in turn, can be well entropy of the hard-sphere crystal becomes greater than the
approximated by a hard sphere of some effective diameter entropy of the fluid at the same density and temperature! This
[4 ].) It is also an important model in introductory and physical does not, however, lead to a negative latent heat of fusion,
chemistry courses because of its use in motivating the van because a Maxwell construction on the Helmholtz free energy
der Waals equation of state for real gases. A experimentally determined by Hoover and Ree yields (reduced) coexistence
realizable system for which a classical hard-sphere potential is a densities for the hard-sphere crystal and fluid of ρc* = 1.043
nearly exact model can be constructed using monodisperse and ρ*f = 0.945, respectively. The latent heat of this transition
polystyrene spheres; the phase behavior of such systems has would still be positive, since the entropy of the crystal at ρc*
been studied (5). is lower than the entropy of the fluid at ρ*, f but as Figure 1
Since the potential energy of any configuration in which illustrates, the lower entropy of the crystal at its coexistence
at least two atoms overlap (i.e., are separated by less than σ) density is not due to the crystal being more ordered, but
is infinite, only non-overlapping configurations will contribute instead simply due to the coexistence density of the fluid
and the potential energy will be zero; therefore, the total being lower than that of the crystal (in this case by about
internal energy will be purely kinetic. The contribution of 15%)—in much the same way that the entropy of an ideal
the kinetic energy to the Helmholtz free energy (the “ideal gas sample at high density is less than that of the same sample
gas” part) is independent of the specific interaction between at low density.
the atoms (at least classically) and is a function of only the The relevance of this model to the real world has been
temperature and bulk density. As a consequence, the difference demonstrated by Crawford and Daniels (8) in their experi-
in Helmholtz free energy, ∆A = ∆U – T∆S, between a hard- ments on the high-pressure phase diagram of argon. Using a
sphere crystal and a fluid at the same density and tempera- rough approximation for the temperature dependence of the
ture must equal {T∆S. If a transition, at fixed T, from fluid effective hard-sphere diameter for argon, they were able to
at low density to crystal at high density occurs, then at some reproduce their experimental values for the crystal and liquid
density ∆A must become negative—which can only occur if molar volumes of argon for a variety of temperatures with
∆S = Scrystal – Sfluid becomes positive; that is, at high density surprising accuracy, given the crude approximations involved.
the entropy of the crystal must exceed that of a fluid at equal Thus, one can conclude that, at least for simple systems such
temperature and density for a freezing transition to exist in as argon, the freezing transition is driven, as in hard spheres,
such a system. by the observed fact that the entropy of the fluid phase is
Alder and Wainwright’s demonstration of a freezing decreasing much faster as a function of density than that of
transition for hard spheres into a face-centered cubic crystal the crystalline phase. The attractive forces between the argon
(6 ) was one of the most important early results to come from atoms have a small role in determining the temperature
computer simulation. Later, a more detailed simulation by dependence of the solid–liquid phase transition, but it is the
Hoover and Ree (7) determined the phase diagram for such packing constraints associated with the repulsive part of the
a system. Figure 1 shows the entropy of the hard-sphere fluid potential that determine the existence of the transition. This
is in direct conflict with the usual mechanism of freezing
0.4 based on attractive forces that is given in most introductory
chemistry textbooks.
Such entropy-driven phase transitions are not restricted to
freezing. In 1949, Onsager (9) speculated that a collection of
microscopic hard rods would undergo an isotropic–nematic
[S (fluid) – S (crystal)] / k

0.2

phase transition at high density. In other words, at low


density the rods (which could be crude models for a liquid-
crystal forming molecule) would be randomly oriented in
0.0 space (isotropic phase), but at high density there would be a
preferred orientation relative to some laboratory-fixed axis
(nematic phase). Onsager’s speculation was later confirmed by
computer simulation (10). Real systems for which the hard-
-0.2 rod system is a good approximation are difficult to construct;
however, there are several examples of a related idealized
system, namely, charged hard-rods (e.g., the tobacco mosaic
virus [11]). Like hard spheres, these hard rods do not interact,
-0.4 so the existence of such a transition implies that at high
0.90 0.95 1.00 1.05
density the entropy of the nematic phase would be higher
Reduced Number Density than that of the isotropic one. In the spirit of our suitcase
Figure 1. The entropy in units of Boltzmann’s constant of the hard- analogy, the easiest way to arrange the matches in a full (or
sphere fluid relative to that of the fcc crystal as a function of re- nearly full) matchbox so that they all can fit inside is to
duced number density, ρ* = ( N/V ) σ3. Data are taken from Table order them orientationally—there are more ways of filling
III of ref 5. the matchbox with all the matches lined up in (more or less)

JChemEd.chem.wisc.edu • Vol. 76 No. 10 October 1999 • Journal of Chemical Education 1389


In the Classroom

the same direction than with randomly oriented matches. giving students a molecular-level understanding of entropy.
Again, at high density the spatially ordered phase is seen to First, as discussed in ref 2, using a subjective term like dis-
be the one with the higher entropy. order (which most students and educators assume refers to
structural disorder) to describe a real, quantitative property
An Easy Demonstration such as entropy leads to genuine confusion when comparing
two systems that are known to differ in entropy. An example
To illustrate the above concepts, as well as the phenom- would be two ideal gas samples with identical n, T, and P but
enon of freezing, in the classroom, the following demonstration with differing volumes, but for which an assessment of rela-
is simple and illustrative. This demonstration comes to me tive spatial disorder is not a well-defined operation. Second,
from A. D. J. Haymet, but it may have earlier origins. as illustrated by the freezing of a system of hard spheres, there
Materials Needed exist systems at high density, where packing considerations
1. 1 bag regular M&M’s or similar disc-shaped candy. dominate, in which the more ordered phase possesses a higher
(Ball bearings also work and have the advantage that entropy than a spatially disordered phase at the same density
they respond better to the shaking process, but they and temperature. The greater free volume and local mobility
are not as tasty.) afforded by the more efficient packing of the ordered phase
2. A flat-bottomed glass dish (preferably square) of a size overcomes the effect of increased positional ordering in de-
that fits on the overhead projector. It is a good idea to termining the entropy at high density. Such situations have
use a non-contaminated dish so that the candy can be relevance in the real world, since the structure and dynamics
eaten by the students later—always a popular side of a real, simple fluid are determined primarily from the
benefit to the demonstration! repulsive part of the interaction, which is well approximated
Procedure by a hard-sphere potential.
Place the glass dish on the overhead projector and place Acknowledgments
a few of the candies inside. Shake the dish gently from side
to side to simulate the effect of temperature. At this point I gratefully acknowledge support from the National Sci-
the view on the overhead screen will resemble a 2-D model ence Foundation under Grant CHE-950281 and from the
of a gas. Begin placing more and more candies into the dish Petroleum Research Fund under grant PRF#30970-AC. I also
while continuing the shaking motion. The gas density will thank the referees for insightful comments and constructive
begin to increase until it approaches liquid-like values (NOTE: criticism.
for a system of hard spheres or disks there is no true liquid-
to-vapor transition.) At this point the system will still appear Literature Cited
to be spatially disordered. However, as the density is further
1. Reiss, H. Methods of Thermodynamics; Dover: New York, 1996.
increased by continuing to add more candy, there comes a
2. Lowe, J. P. J. Chem. Educ. 1988, 65, 403–406.
point when clusters of hexagonal crystals appear that cannot 3. Frenkel, D. Phys. World 1993, 6, 24–25.
be dissipated by shaking. These clusters will grow until the 4. Hansen, J. P.; McDonald, I. R. Theory of Simple Liquids, 2nd ed.;
entire candy configuration is crystalline. It is important to Academic: New York, 1986.
note that this point is below the maximum packing density. 5. Poon, W.; Pusey, P.; Lekkerkerker, H. Phys. World 1996, 9, 27–34.
What one has done here is an analog simulation of the 6. Alder, B. J.; Wainwright, T. E. J. Chem. Phys. 1957, 27, 1208–1209
freezing of hard disks! 7. Hoover, W. G.; Ree, F. H. J. Chem. Phys. 1968, 49, 3609–3617.
8. Crawford, R. K.; Daniels, W. B. Phys. Rev. Lett. 1968, 21, 367–369.
Summary 9. Onsager, L. Ann. NY Acad. Sci. 1949, 51, 627–659.
10. Frenkel, D. Mol. Phys. 1987, 60, 1–20.
In this paper I have argued that the standard assumption 11. Fraden, S.; Maret, G.; Caspar, D. L. D.; Meyer, R. Phys. Rev. Lett.
of entropy and disorder is problematic from the standpoint of 1989, 63, 2068–2071.

1390 Journal of Chemical Education • Vol. 76 No. 10 October 1999 • JChemEd.chem.wisc.edu

You might also like