Residual Entropy

Cool carbon monoxide gas to a crystal near absolute zero. The Third Law predicts its entropy should vanish. Careful calorimetry says otherwise. Real CO crystals retain about 4.6 J/K/mol of entropy even as $T \to 0$ . This leftover entropy is not a failure of thermodynamics. It reveals something important. Molecules can freeze into a solid with their orientations still scrambled. They are unable to find the single ordered ground state. This happens fast enough to matter on laboratory timescales. The mismatch between calorimetric and spectroscopic entropies gave chemists something valuable. It offered a direct thermodynamic window into molecular-scale disorder. This article builds the statistical picture of residual entropy from the Boltzmann formula. You will see why linear, weakly polar molecules like CO trap orientational disorder. You will see why ordinary ice does too. This happens through the different mechanism of proton disorder. Two fully worked numerical examples follow. Alongside these are the standard traps students fall into. These arise when comparing "Third Law entropy" with "practical absolute entropy."

Key takeaways

The Third Law entropy of a perfect crystal is zero at $T=0$ , but many real crystals are not perfect at $T=0$ .
Residual entropy is calculated from $S = k_B \ln W$ , where $W$ counts frozen-in molecular arrangements.
CO crystals retain roughly 4.6 J/K/mol because head-tail orientation (C–O vs O–C) is frozen in almost at random.
Ordinary ice retains about 3.4 J/K/mol from proton disorder obeying the Bernal–Fowler ice rules.
Comparing calorimetric entropy (from $\int C_p/T\,dT$ ) with spectroscopic/statistical entropy exposes the residual term as a discrepancy.
Key idea: residual entropy is disorder frozen into a solid because kinetics, not thermodynamics, stops it from finding its unique ground state.

The core insight

When a crystal cools faster than its molecules can rearrange into the single lowest-energy orientation, a macroscopic number of near-degenerate arrangements get locked in at $T=0$ , and the Boltzmann formula $S=k_B\ln W$ assigns that frozen multiplicity a real, measurable, nonzero entropy that calorimetry detects as a shortfall relative to the Third Law prediction.

The physical picture

The Third Law states that a perfect crystalline substance has zero entropy at absolute zero, because there is exactly one accessible microstate: every atom in its unique lowest-energy position, $W=1$ , so $S = k_B \ln W = 0$ . This is an idealization. It assumes the system can actually reach that single, perfectly ordered ground state as $T \to 0$ . Many real substances cannot. Consider carbon monoxide, CO. The molecule has a very small dipole moment (about 0.11 D) and its two ends, carbon and oxygen, are close in size and electron density. In the crystal, neighboring CO molecules interact almost as strongly whether a molecule points "C–O" or "O–C." The energy difference between orientations is tiny, of order $10-100$ J/mol, comparable to thermal fluctuations at temperatures where the crystal is already forming. As the solid cools, molecules need cooperative rearrangement to sort themselves into the true, fully ordered antiferroelectric or ordered ground state. That rearrangement requires molecular rotation within a rigid lattice, an activated process that becomes exponentially slow as $T$ falls. Long before the ordering process can complete, the relaxation time exceeds any experimental timescale, and the orientational arrangement is "frozen in," random and disordered, all the way down to $T \to 0$ . Water ice presents a related but distinct mechanism. In ordinary hexagonal ice (ice Ih), each oxygen atom is tetrahedrally surrounded by four neighboring oxygens, linked by hydrogen bonds. The Bernal–Fowler ice rules state that each oxygen must have exactly two hydrogens close to it (as in an intact H₂O molecule) and one hydrogen on each of its four hydrogen bonds — so each bond has exactly one proton on it, sitting near one of the two oxygens it connects. This leaves multiple ways to arrange the protons on the fixed hydrogen-bond network, all with essentially the same energy. Reorganizing them requires cooperative proton transfer along hydrogen bonds, again a process too slow to complete before the crystal is far below its freezing point. The result is proton disorder, frozen in, producing residual entropy in ice as well. In both cases, the physical origin is the same: a large number of energetically near-equivalent microstates exist for the frozen orientational or positional degrees of freedom, and the system becomes kinetically trapped among them rather than relaxing to the unique thermodynamic ground state.

Figure

Two frozen microstates of CO in a linear chain, indistinguishable in energy

Figure

Calorimetric entropy versus statistical entropy for CO: the gap is the residual term

The core relations

S(0) = k_B \ln W_0

where: $W_0$ is the number of energetically indistinguishable microstates frozen in at $T=0$ , and $S(0)$ is the residual (Third-Law-violating) entropy per system.

S_{m,\text{res}} = R \ln w

where: $w$ is the number of accessible orientations per molecule (assumed independent), and $R = N_A k_B$ gives residual entropy per mole.

S_{m,\text{cal}}(T) = \int_0^{T} \frac{C_{p,m}}{T'}\,dT' + \sum_i \frac{\Delta H_{i}}{T_i}

where: the integral of heat capacity over temperature plus any phase-transition terms gives the "Third-Law" entropy assuming $S(0)=0$ ; comparing $S_{m,\text{cal}}$ to $S_{m,\text{spec}}$ (from spectroscopy/statistical mechanics) reveals the residual term as their difference.

S_{m,\text{res}} = S_{m,\text{spec}} - S_{m,\text{cal}}

where: $S_{m,\text{spec}}$ is computed independently from molecular partition functions using spectroscopic constants (bond lengths, vibrational frequencies, moments of inertia), assuming a perfectly ordered crystal at $T=0$ .

Derivation AResidual entropy of a two-orientation molecular crystal

Goal: derive the residual molar entropy of a crystal built from $N_A$ molecules, each able to freeze into one of two nearly degenerate orientations, and show how this reduces to the CO estimate.

Step 1. Define the microstate count. Suppose each molecule in the crystal can sit in one of $w$ orientations with (near) equal energy, and the choice for each molecule is statistically independent of its neighbors (a good first approximation when the orientational energy differences are much smaller than $k_BT$ at the temperature where the crystal forms). Then the number of distinguishable arrangements of $N$ molecules is

W = w^{N}.

Step 2. Apply the Boltzmann formula. Using $S = k_B \ln W$ ,

S = k_B \ln\left(w^{N}\right) = N k_B \ln w.

Step 3. Convert to a molar quantity. Set $N = N_A$ (one mole) and use $N_A k_B = R$ :

S_m = R \ln w.

Step 4. Specialize to CO. CO has two orientations per molecule (C–O and O–C pointing along the lattice axis), so $w=2$ :

S_m = R \ln 2 = (8.314\ \text{J K}^{-1}\text{mol}^{-1})(0.6931) = 5.76\ \text{J K}^{-1}\text{mol}^{-1}.

Step 5. Compare to experiment and interpret the difference. The measured residual entropy of CO is about 4.6 J/K/mol, somewhat below the fully random estimate of 5.76 J/K/mol. This tells us the two orientations are not perfectly equally probable — the small CO dipole and quadrupole interactions bias the crystal slightly toward one relative alignment of neighbors, reducing $W$ below the maximal $2^{N}$ but not eliminating the disorder. The random-orientation model is therefore an upper bound, and the shortfall between 5.76 and 4.6 J/K/mol is itself a measure of residual short-range order in the frozen glassy arrangement.

Boxed result:

\boxed{S_{m,\text{res}} = R\ln w \quad \left(\text{ideal, fully random limit}\right), \qquad S_{m,\text{res}}^{\text{CO, obs}} \approx 4.6\ \text{J K}^{-1}\text{mol}^{-1}}

Worked examples

Example 1 — Residual entropy of ice from the Pauling ice-rule count. Water ice is built from oxygen atoms arranged so every oxygen has four hydrogen-bonded neighbors, and every hydrogen bond carries exactly one proton, positioned close to one of its two oxygens (the Bernal–Fowler rules), such that every oxygen keeps exactly two "close" protons forming an intact H₂O unit. Step 1: Naive count ignoring the ice rules. Each of the $2N$ hydrogen bonds in a crystal of $N$ water molecules (each oxygen has 4 bonds, shared between 2 oxygens, giving $2N$ bonds total) can independently place its proton on either of its two possible sites. This gives an over-count

W_{\text{naive}} = 2^{2N}.

Step 2: Restrict to configurations satisfying the "two-near, two-far" rule per oxygen. For a given oxygen with 4 hydrogen bonds, there are $2^4 = 16$ ways to assign near/far positions to its four protons, but only $\binom{4}{2}=6$ of these put exactly two protons close to that oxygen (correct H₂O stoichiometry). So the fraction of configurations satisfying the rule at a single oxygen is

f = \frac{6}{16} = \frac{3}{8}.

Step 3: Apply this fraction independently at each of the $N$ oxygen sites (Pauling's approximation, treating the constraints as statistically independent, which is only approximate but very accurate).

W \approx W_{\text{naive}} \times f^{N} = 2^{2N}\left(\frac{3}{8}\right)^{N} = \left(4 \times \frac{3}{8}\right)^{N} = \left(\frac{3}{2}\right)^{N}.

Step 4: Compute the molar residual entropy.

S_m = k_B \ln W = N_A k_B \ln\left(\frac{3}{2}\right) = R\ln(1.5).

S_m = (8.314\ \text{J K}^{-1}\text{mol}^{-1})(0.4055) = 3.37\ \text{J K}^{-1}\text{mol}^{-1}.

Sanity check: the experimentally measured residual entropy of ordinary ice, from the gap between calorimetric third-law entropy and spectroscopic entropy, is $3.4-3.5\ \text{J K}^{-1}\text{mol}^{-1}$ — Pauling's simple combinatorial estimate of 3.37 J/K/mol matches to within about 2%, a celebrated success of elementary statistical reasoning that Linus Pauling published in 1935.

Example 2 — Extracting the CO residual entropy from calorimetric versus spectroscopic data. Suppose calorimetric measurements (heat capacity integration plus phase-transition contributions from $0$ K to the normal boiling point, corrected to 298 K and 1 bar with the ideal-gas entropy of vaporization) give

S_{m,\text{cal}}^{\circ}(\text{CO}, 298\,\text{K}) = 193.3\ \text{J K}^{-1}\text{mol}^{-1},

while an independent statistical-mechanical calculation from the rotational constant ( $B = 1.9313\ \text{cm}^{-1}$ ), vibrational frequency ( $\tilde\nu = 2143.3\ \text{cm}^{-1}$ ), and molar mass ( $28.01\ \text{g/mol}$ ) using the rigid-rotor/harmonic-oscillator partition functions gives

S_{m,\text{spec}}^{\circ}(\text{CO}, 298\,\text{K}) = 197.9\ \text{J K}^{-1}\text{mol}^{-1}.

Step 1: The spectroscopic value assumes a perfectly ordered crystal at $T=0$ (the Third-Law reference), so it is the "true" thermodynamic entropy relative to a genuine zero-entropy ground state. *Step 2: The calorimetric value implicitly also assumed $S(0)=0$ during the integration, so any real frozen-in disorder present in the actual crystal is simply missing from $S_{m,\text{cal}}$ — the integration never "sees" it because it does not show up as heat capacity in a normal calorimetric run (the associated relaxation is far too slow). Step 3: The residual entropy is the shortfall.*

S_{m,\text{res}} = S_{m,\text{spec}}^{\circ} - S_{m,\text{cal}}^{\circ} = 197.9 - 193.3 = 4.6\ \text{J K}^{-1}\text{mol}^{-1}.

Sanity check: this matches the accepted literature value for CO (4.6-5.0 J/K/mol depending on the source and the crystal preparation), and it sits below the fully randomized $R\ln2 = 5.76$ J/K/mol ceiling from Example set 1's Step 4, consistent with a small orientational bias — as expected for a molecule with a nonzero (if tiny) dipole moment.

Limiting cases & common traps

Situation	What happens	Why
Perfectly symmetric molecule with truly zero orientational energy difference and $w$ equally likely orientations	$S_{m,\text{res}} = R\ln w$ exactly	Boltzmann formula with equal a priori probabilities
Slow cooling / annealing near the ordering transition	Residual entropy decreases toward the true Third-Law value	More time allows cooperative reordering to reach lower- $W$ states
Strongly polar or strongly hydrogen-bonded solids (e.g., NaCl, most simple ionic crystals)	Residual entropy is essentially zero	Large energy difference between orientations makes the ordered state kinetically accessible
N₂O versus CO	N₂O also shows measurable residual entropy (linear, weakly asymmetric ends)	Same head-tail mechanism as CO, smaller magnitude since ends are more different
Comparing "standard entropy" tables from different sources	Values can differ by the residual term if one source uses spectroscopic and another calorimetric data	Always check whether $S^\circ$ was obtained by calorimetric integration (may hide residual entropy) or full statistical calculation (does not)

The chemist behind it

The clearest early demonstration of residual entropy came from comparing calorimetric entropies (built up from heat-capacity measurements down to a few kelvin) against entropies computed from molecular spectroscopy using statistical thermodynamics. When the two routes disagreed for CO, N₂O, H₂O ice, and similar substances, the gap was traced to frozen orientational disorder. Linus Pauling's 1935 combinatorial treatment of the ice-rule counting problem — the calculation reproduced in Example 1 — turned this qualitative idea into a sharp, testable quantitative prediction, and its close match to experiment remains one of the most quoted successes of simple statistical counting arguments in physical chemistry.

Origins · 1 minLinus Pauling and how this was discovered

Linus Pauling

1901-1994 · derived the combinatorial residual entropy of ice from the Bernal-Fowler proton-disorder rules in 1935, matching calorimetric measurements to within a few percent

Ask the tutor: How would the residual entropy calculation change for a molecule like N₂O, whose two ends are chemically different but still hard to distinguish crystallographically?

Does every molecular solid show residual entropy

No. Residual entropy appears only when a crystal has multiple energetically near-degenerate arrangements that cannot equilibrate to the unique ground state before experiments reach very low temperature. Simple ionic solids like NaCl have one strongly preferred arrangement, so residual entropy is essentially zero. Molecules with nearly indistinguishable ends (CO, N₂O, N₂-CO mixed sites) or networks with combinatorial bonding freedom (ice) are the classic cases. The key ingredient is a small energy gap between competing arrangements combined with slow reordering kinetics.

Is residual entropy the same as the entropy of a glass

They are closely related. Both arise from a system freezing into one of many nearly equienergetic configurations rather than reaching the true equilibrium ground state, and both are frozen in by kinetics rather than by energy considerations. Structural glasses (like amorphous SiO2) show a much larger residual entropy from disordered bond networks, while CO and ice show a smaller, more specific orientational or positional disorder on an otherwise regular crystal lattice. Conceptually, "residual entropy" is the general phenomenon; a glass is one dramatic example of it.

How is residual entropy actually measured experimentally

It is inferred, not measured directly. Chemists calculate the "calorimetric" entropy by integrating $C_p/T$ from near 0 K up to a reference temperature, adding any phase-transition entropy terms. Separately, they calculate the "spectroscopic" or statistical entropy from the molecule's rotational and vibrational constants using partition functions, which assumes a perfectly ordered crystal at 0 K. If the spectroscopic value exceeds the calorimetric value, the difference is assigned to residual entropy frozen into the real crystal.

Why does cooling rate affect the amount of residual entropy observed

Residual entropy is a kinetic freezing phenomenon, so slower cooling gives molecules more time to find lower-energy, more ordered arrangements before reorientation becomes too slow to proceed at all. Rapidly quenched samples trap more disorder and show entropy closer to the fully random $R\ln w$ limit. Very slowly annealed crystals can approach the true, lower Third-Law entropy, though many practical solids never fully reach it on laboratory timescales, even over years.

Does residual entropy contradict the idea that entropy is a state function

No. Entropy remains a state function; the issue is which state the system is actually in. A frozen-disordered crystal at $T=0$ is thermodynamically a different (metastable) state from the hypothetical perfectly ordered crystal at $T=0$ , and it legitimately has a different, higher entropy. The Third Law's zero-entropy statement applies specifically to the true equilibrium ground state, not to every solid that happens to be very cold.