Past exam of the quantum condensed matter course of the University of Cambridge / 2008

Official past-paper archive: www.sp.phy.cam.ac.uk/drp2/home. The direct PDF link has not yet been verified.

For complete wetting, the meniscus has curvature radius equal to the capillary radius $r=d/2$. The Young–Laplace equation gives the supporting pressure difference $\Delta p=2\gamma /r=4\gamma /d$. Balance it against the hydrostatic pressure $\rho gh$. Taking the sap's mass density to be water-like, $\rho \simeq 10^3\mathrm{kg}{\mathrm{m}}^{-3}$, givesThis is the height supported by the stipulated capillary-pressure model. The density assumption is needed because the surface tension and capillary diameter alone do not determine a height.

For the symmetric regular solution model, the dimensionless mixing free energy is $f/(k_{B}T)=\varphi \log \varphi +(1-\varphi )\log (1-\varphi )+\chi \varphi (1-\varphi )$. At the thermodynamic critical point, its second and third derivatives in composition vanish. The third derivative gives ${\varphi }_c=1/2$; the second then gives $4-2{\chi }_c=0$. ThereforeThis uses the equal-size, composition-symmetric mixture model.

The mean interface height fluctuation is zero; its root mean square roughness is obtained by summing the mode variances. Replacing the two-dimensional wavevector sum by its continuum density gives the thermal roughness of a tension-dominated interface:Take $q_{\min }=2\pi /L$ with $L=\sqrt{A}=10^{-2}\mathrm{m}$, and $q_{\max }=2\pi /d$ with $d=2\times 10^{-10}\mathrm{m}$. At $T=300\mathrm{K}$,The logarithm makes the answer only weakly sensitive to order-one choices of cutoff. Gravity has been omitted as stipulated; the fixed mean height removes the zero wavevector mode.

A liquid crystal flows like a liquid while retaining some molecular order. In a nematic liquid crystal, elongated molecules have long-range orientational order but no ordinary positional lattice. The nematic director $\mathbf{n}$ specifies the preferred axis, with $\mathbf{n}$ and $-\mathbf{n}$ equivalent for head-tail-symmetric molecules. The nematic order parameter $S=⟨(3{\cos }^2\theta -1)/2⟩$ distinguishes an isotropic distribution, $S=0$, from perfect parallel alignment, $S=1$. A smectic phase additionally has a periodic layered density modulation, while molecules remain mobile within layers.

The broken rotational symmetry allows birefringence and distinct elastic costs for nematic splay, nematic twist and nematic bend. Under crossed polarizers, rotating a thin aligned sample changes the transmitted intensity because its two optical polarization components acquire different phases. Heating through a nematic-to-isotropic transition removes that orientational birefringence. An electric field can rotate the director if the dielectric anisotropy of a nematic favors alignment, explaining the optical switching principle of a liquid-crystal cell. Molecular shape, temperature, concentration and surface anchoring determine which ordered phase appears.

The restoring force in ideal rubber elasticity is mainly entropic: stretching restricts the conformations of network strands. A Gaussian chain of $N_s$ statistical segments of length $b$ has extension free energy $3k_{B}T{R}^2/(2N_s{b}^2)$, so its small-extension spring constant is proportional to $T$. For an affine network with active-strand number density ${\nu }_s$, the shear modulus is approximately $G={\nu }_s{k}_{B}T$. Increasing the cross-link density normally raises $G$ and shortens the strands; polymer entanglement can add effective constraints. Dangling ends and loops do not contribute like elastically active bridges, while junction fluctuations reduce the affine-network estimate.

For incompressible uniaxial stretching by ratio $\lambda$, the nominal tensile stress is $G(\lambda -{\lambda }^{-2})$ in the ideal Gaussian-network model. Finite contour length causes stronger stiffening as strands approach full extension; swelling dilutes the active strands and changes their reference conformations. A direct demonstration is to hang a load from a stretched rubber strip and warm it gently: an entropy-dominated network tends to contract under that fixed load. This temperature dependence contrasts with a model based solely on stretching bond energies. The simple formula assumes equilibrium, adequate cross-linking and temperatures away from glass transitions or degradation.

DLVO theory combines the attractive Van der Waals force with repulsion between overlapping electrical double layers. A charged colloid attracts counterions; their electrostatic and osmotic response screens its field over the Debye length. Adding salt generally shortens this range and can lower the repulsive aggregation barrier, allowing particles to reach the strongly bound primary minimum. A shallow secondary minimum can instead give weak reversible association. Whether Brownian encounters cause rapid aggregation depends on barrier heights relative to $k_{B}T$, not just on the sign of the force at one separation.

Adsorbed or grafted polymer layers can provide steric stabilization: overlap costs chain entropy and increases the local osmotic pressure. Nonadsorbing smaller particles or polymers can create a depletion force, because bringing large surfaces together increases the volume accessible to the smaller species. Polymer bridging can be attractive when one chain binds more than one particle. A useful comparison experiment varies salt concentration while monitoring scattered light or aggregation, then repeats with a sterically protected suspension. Charge, solvent quality, polymer coverage and particle size determine which interaction dominates; none is a universal constant of all colloids.

For a microscopic swimmer, the Reynolds number $\mathrm{Re}=\rho UL/\eta$ is usually small. Taking $L=2\mu \mathrm{m}$, $U=20\mu \mathrm{m}{\mathrm{s}}^{-1}$ and water-like properties gives $\mathrm{Re}\simeq 4\times 10^{-5}$. The Stokes equations then balance viscous stress and pressure almost instantaneously, so a swimmer cannot coast appreciably when its propulsive stroke stops.

The scallop theorem rules out a reciprocal opening-and-closing stroke in an otherwise unchanged Newtonian environment: reversing the sequence of shapes reverses the motion, and changing the stroke speed does not prevent cancellation. Organisms therefore use nonreciprocal motion. A sperm-like bending wave travels along a flagellum; a bacterium can rotate a helical flagellar bundle. The unequal parallel and perpendicular drag of a slender filament converts those motions into a net thrust. A freely suspended swimmer is force-free and torque-free overall, so body drag balances thrust and the body counter-rotates against the rotating bundle. Tracking a body and its flagellum together can reveal these opposite rotations. The helical microswimmer with a spherical head makes that force and torque balance quantitative.

A linear viscoelastic fluid responds linearly to small imposed strain histories but retains memory of earlier deformation. For shear strain $\gamma (t)$ and a relaxed initial state, Boltzmann superposition givesHere $G(t)$ is the causal relaxation modulus, the stress following a unit step strain. Doubling a sufficiently small strain history doubles the stress, and responses to added histories add. Elasticity stores deformation energy; viscosity dissipates it. Linearity is an amplitude approximation, not an assertion that the material has no relaxation times.

The Maxwell model has a spring of modulus $G_0$ in series with a dashpot of viscosity $\eta$. Their stresses are equal and strains add, giving $\dot{\gamma }=\dot{\sigma }/G_0+\sigma /\eta$. After a step strain ${\gamma }_0$, the strain rate is zero for $t>0$, so $\dot{\sigma }=-\sigma /\tau$ with $\tau =\eta /G_0$. Initially the spring takes the step, $\sigma (0^{+})=G_0{\gamma }_0$. Hence

For the convention $\gamma =\mathrm{Re}({\gamma }_0{e}^{i\omega t})$, the complex viscoelastic modulus is $G^{\ast }=i\omega {\int }_0^{\infty }G(t)e^{-i\omega t}dt$. ThusThe storage modulus exceeds the loss modulus for $\omega \tau >1$, so the response is predominantly solid-like above the crossoverAt low frequency $G^{\ast }\simeq i\omega \eta$; at high frequency $G^{\ast }\to G_0$.

The creep compliance is the strain per unit step stress: if $\sigma (t)={\sigma }_{0}H(t)$, then $\gamma (t)={\sigma }_{0}J(t)$ for $t>0$. A Newtonian fluid obeys $\sigma =\eta \dot{\gamma }$, so under constant stress its initially zero strain grows as $\gamma ={\sigma }_{0}t/\eta$. Therefore

Passive microrheology uses thermal bead motion as a probe of mechanical compliance. The fluctuation-dissipation theorem relates equilibrium displacement fluctuations to the same linear response measured by a small applied force. The stated thermal stress scale has units of energy per bead volume; dividing the mean-square displacement by $a^2$ provides a dimensionless effective strain. With that prescribed analogy,The displacement is relative to the starting position, and the variance here is the sum over three spatial coordinates. For a Newtonian fluid, substitution gives $⟨\Delta r^2⟩=k_{B}Tt/(\pi \eta a)=6Dt$, exactly the Stokes–Einstein relation $D=k_{B}T/(6\pi \eta a)$.

The continuum interpretation assumes a dilute no-slip spherical probe, thermal equilibrium, linear response and a homogeneous incompressible medium on the bead scale. The stated stress estimate is an effective scale for the analogy; a literal instantaneous stress variance would also require a microscopic or temporal averaging prescription.

Under a step stress, the Maxwell spring extends immediately by ${\sigma }_0/G_0$, while the dashpot adds ${\sigma }_{0}t/\eta$. Hence $J(t)=G_0^{-1}+t/\eta$, and the microrheology of a Maxwell fluid relation givesThis idealized overdamped relation describes an elastic plateau followed by long-time diffusion. Its nonzero extrapolated intercept is not the actual displacement variance at exactly zero lag, which is zero. Inertia or fast solvent relaxation must resolve the initial rise; those processes were omitted from the ideal spring-dashpot model.

An entropic network has approximately one thermal energy of elastic cost per mesh volume, giving $G_0\sim k_{B}T/{\xi }^3$. At $T=300\mathrm{K}$, the supplied mesh size and viscosity giveThe intermediate-time plateau in the three-dimensional mean squared displacement isAfter the omitted fast initial transient, the log-log sketch is approximately flat for $t\ll \tau$, crosses over near $\tau$, and grows linearly for $t\gg \tau$. The figure shows the Maxwell compliance prediction over this range.

An ordinary 25–30 frame-per-second recording samples intervals of about $0.03$–$0.04\mathrm{s}$, comparable to $\tau$. It therefore has little access to a clear plateau at lags well below the crossover, even if its positional precision resolves a displacement of a few tenths of a micrometre. Higher frame rates and calibrated localization noise would improve the modulus measurement. The bead radius is only twice the mesh size, so treating the network as a homogeneous continuum is also an approximation.

Take the same segment length and a large common number of steps. The ideal walks have size proportional to $N^{1/2}$; excluded volume swells a three-dimensional chain to approximately $N^{0.59}$ and a two-dimensional chain to $N^{3/4}$. Projection changes an isotropic walk's prefactor, rather than its exponent. The projected three-dimensional ideal walk has a smaller planar size than an otherwise equal two-dimensional ideal walk. Therefore the asymptotic ordering isThe following subparts make the size and projection factors explicit. Comparing root-mean-square sizes gives the same ordering as comparing mean distance magnitudes here.

For the planar random walk with independent steps of length $b$, cross terms vanish and $⟨R^2⟩=N{b}^2$. Thus $R_{\mathrm{rms}}=b{N}^{1/2}$. At large $N$, the planar Gaussian end distribution gives $⟨R⟩\simeq b\sqrt{\pi N}/2$.

A planar self-avoiding walk cannot revisit its occupied sites. In the good-solvent scaling regime, the two-dimensional Flory approximation gives $R\sim b{N}^{3/4}$, larger at large $N$ than either ideal walk and the projected three-dimensional self-avoiding walk. This is a scaling statement, with microscopic prefactors dependent on the walk model.

For an isotropic three-dimensional ideal walk, each Cartesian component contributes $N{b}^2/3$ to the mean square. Keeping two components gives $⟨R_{\mathrm{proj}}^2⟩=2N{b}^2/3$. The planar projection of an isotropic random vector therefore has $R_{\mathrm{proj},\mathrm{rms}}=b\sqrt{2N/3}$, smaller than the unprojected planar ideal walk by $\sqrt{2/3}$. The corresponding two-dimensional Gaussian mean distance has the same ratio.

A three-dimensional self-avoiding walk has swollen size $R\sim b{N}^{\nu }$, with $\nu \simeq 0.59$; the three-dimensional Flory approximation gives the nearby estimate $3/5$. Isotropic projection gives $⟨R_{\mathrm{proj}}^2⟩=(2/3)⟨R^2⟩$, preserving this exponent. For the mean distance rather than its root mean square, the orientation average instead gives $⟨R_{\mathrm{proj}}⟩=(\pi /4)⟨R⟩$. Neither prefactor changes its position between the ideal $N^{1/2}$ walks and the planar swollen $N^{3/4}$ walk.

The Kuhn length $l_K$ is the segment length of an equivalent freely jointed chain chosen to match contour length $L$ and the unperturbed mean-square end-to-end distance: $⟨R^2⟩=L{l}_K$. Since $R\le L$, its maximum finite-chain value is $l_K=L$, attained by a straight rigid rod. Independent steps of length $b$ have $L=Nb$ and $⟨R^2⟩=N{b}^2$, soFor a long freely rotating chain with fixed angle $\alpha$ between successive forward bond vectors, orientational correlations decay as $c^s$ with $c=\cos \alpha$. Summing them gives the Kuhn length of a freely rotating chain:Positive persistence increases the statistical segment length, while negative bond correlations can reduce it. A chemical internal bond angle is supplementary to this forward-vector angle, so the cosine convention matters. Near the perfectly aligned limit the finite-chain bound $l_K\le L$ replaces the long-chain asymptotic formula.

Write $\ell$ for a segment length and $v_2$ for the excluded-area coefficient, so that the interaction term has the correct dimensions. For a coil of planar size $R$ and area of order $R^2$, the two-dimensional Flory approximation isThe first term is the entropic stretching cost and the second counts repulsive monomer contacts. Minimization gives $2R/(N{\ell }^2)-2v_2{N}^2/R^3=0$, henceFor $v_2$ of order ${\ell }^2$ this becomes $R\sim \ell N^{3/4}$. The parameter multiplying $N^2/A$ in a two-dimensional interaction energy must have dimensions of area; it cannot simultaneously denote an unsquared bond length.

The polymer overlap concentration is reached when the number of chains per coil area is of order one. If $n_c$ is the chain number per unit interfacial area, then $n_c^{\ast }{R}^2\sim 1$. Equivalently, the monomer area fraction is ${\varphi }^{\ast }\sim N{\ell }^2/R^2$, assuming an occupied monomer area of order ${\ell }^2$. Below this scale coils are mostly separate; above it they interpenetrate.

For the planar good-solvent coil, $R\sim \ell N^{3/4}$. The two-dimensional polymer overlap fraction is thereforeThis is a scaling estimate; the monomer geometry and choice of coil radius set order-one numerical factors.

Here the osmotic pressure is a two-dimensional surface pressure, with units of energy per area. In the dilute regime, independent whole chains provide the translational entropy, soIn the semidilute polymer solution, use a correlation blob containing $g$ monomers. Within it the chain still follows the planar good-solvent law $\xi \sim \ell g^{3/4}$. Its area fraction is $\varphi \sim g{\ell }^2/{\xi }^2\sim g^{-1/2}$, so $g\sim {\varphi }^{-2}$ and $\xi \sim \ell {\varphi }^{-3/2}$. There is about one independent thermal contribution per blob area. The semidilute osmotic pressure of a two-dimensional polymer solution is thusIt is independent of total chain length once each chain contains many blobs. The two estimates agree at ${\varphi }^{\ast }\sim N^{-1/2}$, where each gives $(k_{B}T/{\ell }^2)N^{-3/2}$.

For a single carrier type with unit Hall factor, the magnitude of the Hall voltage is $V_H=IB/(net)$, where $t$ is the slab thickness parallel to the magnetic field. ThereforeThe longitudinal electrical resistivity is $\rho =Rwt/L=1000(2\times 10^{-3})(0.5\times 10^{-3})/(5\times 10^{-3})=0.20\Omega \mathrm{m}$. Using $1/\rho =ne\mu$, the electron mobility magnitude isWithout the Hall polarity these data do not identify the carrier charge sign. The inference assumes one dominant carrier population; electron-hole compensation would require a multicarrier analysis.

Put $N=N_x{N}_y{N}_z$ primitive cells in a periodic box. In each reciprocal-space direction, the allowed Bloch wavevectors are spaced by $2\pi /L_i$, giving $N_i$ independent values in the first Brillouin zone. There are consequently $N$ wavevectors per band and, including two electron spin states at each wavevector,The total and per-atom counts are different; a factor $N$ cannot remain in the per-atom count for this monatomic lattice.

In independent-electron band theory, a partially occupied band permits low-energy changes of occupancy and gives a metal. Filled bands separated from empty bands by a finite band gap give a band insulator at zero temperature. An odd number of electrons per primitive cell cannot fill spin-degenerate bands, whereas an even number can; even filling alone does not ensure an insulator if bands overlap. Magnetic symmetry breaking and strong correlations, such as a Mott insulator, go beyond this simple counting criterion.

A plasmon is a quantum of a collective longitudinal oscillation of electron density against the positive background. A small charge displacement creates an electric restoring field; Newton's law then gives the plasma angular frequency. Assuming approximately one conduction electron per atom and spacing $d=3\text{A˚}$ gives $n\sim d^{-3}=3.7\times 10^{28}{\mathrm{m}}^{-3}$. HenceThe ordinary frequency is about $1.7\times 10^{15}\mathrm{Hz}$, with energy $\hslash {\omega }_p\sim 7\mathrm{eV}$. A different conduction-electron count per atom changes this estimate by its square root; dielectric and band-structure corrections also modify the simplest free-electron result.

The tight-binding approximation starts with localized atomic orbitals. Their overlap allows hopping, broadening each atomic energy into a band. For example, a single orbital on a simple cubic lattice with nearest-neighbor hopping gives $E(\mathbf{k})=E_0-2t[\cos (k_{x}a)+\cos (k_{y}a)+\cos (k_{z}a)]$. Small overlap means a narrow band; orbital symmetry and hopping signs determine its extrema and dispersion.

The nearly-free electron model starts instead with free-electron parabolas, folded into the Brillouin zone, and treats the periodic crystal potential as a perturbation. Near a Bragg plane, states differing by a reciprocal vector become degenerate and mix, opening a gap approximately $2|V_{\mathbf{G}}|$. Away from such degeneracies, the energy remains close to the free-electron dispersion. These are complementary limits of the same Bloch-band framework, not different counts of electronic states. Both can give metals or insulators depending on filling and band overlap, but their natural basis functions and useful expansion parameters differ.

Thomas-Fermi screening treats the carrier density locally as that of a uniform electron gas whose chemical potential is shifted by the slowly varying electrostatic potential. Linearizing gives $\delta n=(\partial n/\partial \mu )e\phi$ for electrons. The resulting screened Poisson equation has inverse screening lengthFor a degenerate gas, $\partial n/\partial \mu =g_{\mathrm{vol}}(E_F)$ including both spins, giving Thomas-Fermi screening length $k_s^{-1}$. A nondegenerate semiconductor instead has $\partial n/\partial \mu =n/(k_{B}T)$, giving the Debye length $\sqrt{\epsilon k_{B}T/(n{e}^2)}$. Electrons and holes contribute additively to the charge susceptibility when both are present; the lattice permittivity replaces the vacuum value.

The approximation requires a potential varying slowly compared with microscopic electronic length scales and a small perturbation for the linearized result. It captures smooth electrostatic screening, but not the full nonlocal response or Friedel oscillations. Low carrier density or carrier freeze-out can make a semiconductor's screening much weaker than a metal's.

The Hubbard model combines band motion and local repulsion:Hopping favors itinerant states, while positive $U$ penalizes simultaneous opposite-spin occupation on one site. In a uniform mean-field treatment, a spin imbalance reduces the interaction cost but raises the band-filling energy. The Stoner criterion for instability of the unpolarized metal is $U{g}_{\uparrow }(E_F)>1$ when $g_{\uparrow }$ is the density of states per site, per energy, for one spin. Using a total two-spin density of states changes the displayed numerical convention.

The ordered mean-field state has exchange-split spin bands and a spontaneous magnetization, with the kinetic and interaction costs balancing at equilibrium. This is a model of itinerant ferromagnetism, not a guarantee that every repulsive Hubbard system becomes ferromagnetic. Filling, lattice structure and competing order matter: near half filling on a bipartite lattice, antiferromagnetism and a Mott insulator can be favored instead. The approximation also omits important spin fluctuations and strong-correlation effects.

In a two-dimensional electron system under a perpendicular magnetic field, orbital energies form Landau levels with spacing $\hslash {\omega }_c$. Each resolved spin level has $eBA/h$ states in area $A$. At integer filling, the integer quantum Hall effect has Hall conductance ${\sigma }_{xy}=\nu e^2/h$ in magnitude and vanishing longitudinal conductance in the ideal plateau regime; inversion of the conductivity tensor gives Hall resistance $h/(\nu e^2)$.

A Hall-bar experiment cools the sample, passes a small current and records longitudinal and transverse voltages while sweeping the field. The Hall resistance develops plateaus while the longitudinal resistance develops minima. Disorder broadens the Landau levels and localizes many bulk states, allowing the chemical potential to move through a range of carrier density without changing the number of conducting edge channels. Extended states near a Landau-level center allow transitions between plateaus. Spin splitting determines which integer sequence is resolved. The robustness comes from the quantum state and its edge transport, rather than the Hall bar having a specially calibrated geometric width.

For photon absorption or emission, electron energy changes by $\pm \hslash \omega$ and crystal wavevector by $\pm \mathbf{q}$ modulo a reciprocal lattice vector. Optical photon momentum is normally tiny on the Brillouin zone scale, so interband optical transitions are approximately vertical on an energy-wavevector diagram. A direct-gap semiconductor can therefore absorb a photon between valence and conduction states near the same wavevector; an indirect transition usually needs an additional phonon or another source of momentum. Occupation and optical matrix elements impose further restrictions.

Within one ideal parabolic band of a clean metal, such a small photon momentum cannot normally supply the required finite energy difference. Intraband optical absorption then relies on momentum relaxation by impurities, phonons or boundaries, as in the finite-scattering-time Drude response. Interband transitions can still occur. The first panel shows the near-vertical direct transition.

For an acoustic phonon with ${\omega }_q\simeq v_{s}q$, energy and crystal momentum conservation require $E_{{\mathbf{k}}^{\prime }}-E_{\mathbf{k}}=\pm \hslash v_{s}q$ and ${\mathbf{k}}^{\prime }-\mathbf{k}=\pm \mathbf{q}+\mathbf{G}$. The sound speed is usually much smaller than the relevant electron velocity in a metal, so scattering can change the electron's direction substantially while changing its energy only slightly. It is approximately elastic for many transport calculations, though its finite energy transfer is essential for cooling.

The middle panel gives an allowed backscattering example with a small energy loss. In a degenerate metal, the Pauli exclusion principle blocks occupied final states, restricting low-temperature scattering to a thermal neighborhood of the Fermi surface. Phonon occupations also shrink on cooling. A normal process has $\mathbf{G}=0$; an Umklapp process transfers reciprocal-lattice momentum. In semiconductors, accessible valleys and their separation add further momentum constraints.

An optical phonon often has approximately constant energy $\hslash {\omega }_0$ over the relevant wavevector range. Its absorption or emission therefore moves the electron between energy surfaces separated by this finite amount, with the phonon supplying the associated crystal momentum modulo a reciprocal vector. The last panel shows emission between two points on a parabolic conduction band.

Emission from a band with a lower edge requires at least $\hslash {\omega }_0$ of available kinetic energy and an unoccupied final state. For nondegenerate hot carriers this produces a threshold and can lead to rapid energy loss in successive phonon emissions. Absorption is weighted by the Bose-Einstein distribution $n_B({\omega }_0)$, while emission has the factor $n_B+1$; absorption is strongly suppressed at $k_{B}T\ll \hslash {\omega }_0$. In a metal, Pauli blocking also prevents an electron near the Fermi energy from simply dropping into the occupied Fermi sea.

Let $\alpha ={\hslash }^2/(2m^{\ast })$, so the initial conduction-electron kinetic energy is $E_1=\alpha k_1^2$. In the normal momentum-conserving process with the final hole fixed at zero wavevector, the two final conduction electrons obey ${\mathbf{k}}_1={\mathbf{k}}_2+{\mathbf{k}}_3$. Their total kinetic energy satisfiesEnergy conservation for impact ionization is $E_1=E_g+\alpha (k_2^2+k_3^2)$. Consequently $E_1\ge E_g+E_1/2$, givingThe bound is attained when ${\mathbf{k}}_2={\mathbf{k}}_3={\mathbf{k}}_1/2$; opposite spins permit the two electrons to share that wavevector. Fixing the hole momentum is essential: a mobile final hole gives a different threshold. This is the threshold for impact ionization with a hole at zero momentum.

The two hot electrons subsequently redistribute energy by electron-electron collisions and transfer it to the lattice by electron-phonon scattering. Optical-phonon emission can provide large discrete losses while the electrons are hot; acoustic phonons dominate later small energy exchanges. Electron-electron collisions alone conserve the electron system's total energy and do not cool it into the lattice. The electron-hole pair eventually recombines radiatively or through nonradiative channels, such as defect-assisted recombination or Auger recombination. Recombination is distinct from the preceding intraband cooling.

Only states within an energy window of order $k_{B}T$ around the Fermi energy can change occupation in a degenerate metal. A Sommerfeld expansion gives electronic energy above its zero-temperature value proportional to $g(E_F)(k_{B}T{)}^2$, hence $C_e=\gamma T$. Three-dimensional linear-dispersion acoustic phonons have density of states proportional to ${\omega }^2$; integrating their Bose occupations gives thermal energy proportional to $T^4$ and the Debye heat capacity $C_{mph}=\beta T^3$. Thus $C/T=\gamma +\beta T^2$ at sufficiently low temperature in the normal state.

Reading the straight low-temperature fit gives approximately $\gamma \simeq 33\mathrm{mJ}\mathrm{mo}{\mathrm{l}}^{-1}{\mathrm{K}}^{-2}$ and $\beta \simeq 3.8\mathrm{mJ}\mathrm{mo}{\mathrm{l}}^{-1}{\mathrm{K}}^{-4}$. These estimates agree with the reported fit to this material in the original experiment.^[ref] Treating the graph values as approximate, the molar density of states isThe stipulated two electrons per formula unit give $n=2N_A/V_m\simeq 2.62\times 10^{27}{\mathrm{m}}^{-3}$. For a single spin-degenerate parabolic band, the molar density of states is $g(E_F)=3(2N_A)/(2E_F)=3N_A/E_F$. HenceThese are model-dependent effective parameters inferred from the measured heat capacity and assumed carrier count, rather than an independent determination of a unique band structure.

Using the Debye prefactor stated in the question givesThis prefactor counts three effective modes per formula unit. Consistently choosing the same count for the Debye wavevector gives $k_D=(6{\pi }^2{N}_A/V_m{)}^{1/3}\simeq 4.3\times 10^9{\mathrm{m}}^{-1}$. The corresponding average sound speed isThere is a normalization choice here: a formula unit has 31 atoms. With the conventional full atomic-mode count, the molar Debye coefficient contains an additional factor 31, and both ${\Theta }_D$ and $k_D$ increase by $31^{1/3}$. That convention gives ${\Theta }_D\simeq 251\mathrm{K}$ and $k_D\simeq 1.34\times 10^{10}{\mathrm{m}}^{-1}$, while leaving the inferred sound speed unchanged. The mode count in the Debye model must be used consistently; mixing the printed temperature with an atomic-count wavevector gives an erroneous sound speed.

The upward curvature above about $3\mathrm{K}$ indicates low-energy vibrations missing from a pure acoustic $T^3$ model. In this cage compound, the original analysis fits an additional localized mode with $\hslash \Omega /k_B\simeq 32\mathrm{K}$, assigned to a rattling mode of a barium guest atom.^[ref] At temperatures high compared with all relevant phonon energies, the full lattice approaches the Dulong-Petit law, $C_{mlattice}\to 3(31)R=93R\simeq 773\mathrm{J}\mathrm{mo}{\mathrm{l}}^{-1}{\mathrm{K}}^{-1}$. The reduced three-mode model alone approaches $3R$ and omits the other branches. An approximately linear electronic term can still be added while the carriers remain degenerate, but cannot be extrapolated beyond their Fermi temperature.