Lattice vibrations, electrons and heat transport

Heat has microscopic carriers

In many insulating crystals, lattice vibrations carry heat. Electrons can contribute strongly in metals. Defects, interfaces and thermal motion scatter carriers; their roles depend on temperature and length scale.

A crystal can transfer energy from a warmer region to a cooler one while carrying no electrical current. That does not identify its heat carrier. An insulating crystal can use collective atomic vibrations; in a metal, electrons can transport energy even when their opposite charge flows cancel. First identify what moves and what is being counted.

Heat capacity describes a change in stored thermal energy with temperature. Conductivity describes a transport response to a spatial temperature gradient. A large population of excited motions is not by itself a large heat flux. At uniform equilibrium temperature, energy-bearing motions occur in opposite directions without a net heat flow. Heat capacity and thermal conductivity develops that storage-versus-transport distinction and the units used here.

A lattice transmits energy through coupled motion

Atoms in a solid interact through their bonding environment. If one region is displaced, forces change on its neighbours; their subsequent motion changes forces farther away. Energy can propagate while each atom remains near its own mean site. The first figure uses a schematic spring-coupled row to distinguish local motion from an energy-flow direction. It is not a literal picture of every bond or a calculated trajectory.

A vibration mode is a coordinated pattern of motion of many atoms. In quantum mechanics its excitation energy changes in discrete amounts; a phonon names one such quantum of a lattice-vibration mode. A phonon is not an atom and is not a little material grain travelling through the crystal. Lattice energy transfer does not require a net flow of ions from one end to the other, or metallic electrical conduction. In a disordered solid a description in terms of long-lived crystal modes can become less adequate.

Scattering changes how a directional population transfers energy. A mass impurity or a changed bond stiffness disturbs a vibration pattern. A grain boundary can transmit some energy and reflect or redistribute some. Electrons can exchange energy and momentum with vibrations and defects. A scattering event need not destroy energy; it can redistribute it between directions, frequencies or carriers. Not every redistribution relaxes net heat flow equally.

In the ideal harmonic approximation, restoring forces are proportional to displacement and the normal modes do not interact through that approximation alone. Real bonding is anharmonic: higher-order force terms allow interactions among vibrations. Their available populations and scattering pathways change with temperature. At sufficiently low temperatures, adding populated vibrations can raise heat transport even while scattering changes; in other regimes stronger heat-relaxing scattering lowers it. No universal monotonic conductivity-versus-temperature curve follows from the phrase “more vibration”.

A mean free path is a characteristic distance associated with a specified scattering or relaxation process, not a vibration wavelength. Its value can vary widely among modes. Confinement and low-dimensional structures develops a deliberately limited single-path kinetic example and its boundary assumptions. Here the question is different: what do thermal and electrical observations identify when several energy routes and interfaces contribute?

Opposite electron counts can carry unequal energy

Electrons transport electrical charge and can also transport energy. In a metal near local equilibrium, the distribution of occupied and accessible electronic states matters; an applied temperature difference need not give the same net response as an electric field. Charge carriers and electronic energy bands explains occupation, mobile carriers and scattering. Heat and charge are different weighted accounts of that distribution.

Use an original crossing record at one specified plane, with rightward taken as positive. Let Ṅ_R and Ṅ_L be the numbers of electrons crossing rightward and leftward per second. The dot denotes a rate, not a concentration per volume. With electron charge −e, the signed electrical current is I = −e(Ṅ_R − Ṅ_L), where e is the positive charge magnitude. Equal opposite counts give zero current.

Supply Ṅ_R = Ṅ_L = 5.00 × 10¹⁵ s⁻¹, mean energy E_R = 0.0300 eV for rightward crossings and E_L = 0.0100 eV for leftward crossings, using one common energy reference. The net energy rate through the chosen area is Q̇ = Ṅ_R E_R − Ṅ_L E_L. Using the explicitly rounded conversion 1 eV = 1.60 × 10⁻¹⁹ J for this example gives Q̇ = 5.00 × 10¹⁵ × 0.0200 × 1.60 × 10⁻¹⁹ = 1.60 × 10⁻⁵ W, or 16.0 µW, to the right. Meanwhile I = 0. Because there is no net particle transfer, a common change of energy reference cancels from this rate; it is the heat flow in this account. This is a rate through an area, not heat flux per area.

The record illustrates different weights, not a measured metallic distribution or a ballistic-device prediction. Its two energies are supplied averages of crossings, not the energies of all electrons at two temperatures. In a real open electrical circuit, charge can rearrange and establish a thermoelectric field that enforces zero net current. “Open circuit” does not mean electrons stop moving, and it does not forbid electron-carried heat. If a net particle flow were present, chemical/electrochemical energy transport would also need to be separated.

A thermal/electrical link is an additional model

For a simple matched bulk comparison, suppose electrons and lattice vibrations can each be represented by an effective local conductivity under the same small temperature gradient. Define κ = κ_e + κ_lat, where κ_e is the electronic contribution under zero net electrical current, and κ_lat is the lattice contribution within this two-channel model. They need not be independent microscopic systems: electron–phonon interaction can influence both. Assume the same local temperature, fixed material state and direction, and no additional radiative, magnetic, ionic or convective contribution.

In suitable ordinary metallic regimes, a Wiedemann–Franz approximation writes κ_e = LσT. Here σ is electrical conductivity in S/m, T is absolute temperature in kelvin and L is a Lorenz coefficient. For the first fictional calculation, stipulate L = 2.44 × 10⁻⁸ W Ω K⁻², close to the conventional Sommerfeld value. The units give (W Ω K⁻²) × (Ω⁻¹ m⁻¹) × K = W/(m K). In the conventional degenerate-metal regime, thermal excitation affects a narrow range near the occupied–empty electronic-state boundary. A coefficient near the conventional value can be useful when scattering and its energy dependence let heat and charge relax in the corresponding way. Strong energy dependence, inelastic energy exchange or correlated electronic motion can defeat that approximation. The relation concerns the electronic contribution, not total κ; calling a substance metallic does not alone validate that L.

At a common T = 300 K, fictional state A has σ_A = 1.00 × 10⁷ S/m and κ_A = 100 W/(m K). State B has σ_B = 8.00 × 10⁶ S/m and κ_B = 90.0 W/(m K). These are exact supplied inputs for comparable specimens, directions, geometry and operating conditions; no real material or measurement uncertainty is represented. The stipulated L gives κ_e,A = 73.2 and κ_e,B = 58.56 W/(m K). Subtraction gives κ_lat,A = 26.8 and κ_lat,B = 31.44 W/(m K). A lower total κ can coexist with a larger inferred lattice contribution because the electronic part fell more.

Now replace the single value by a supplied admissible model range, 2.20 × 10⁻⁸ ≤ L ≤ 2.60 × 10⁻⁸ W Ω K⁻². This range is not a confidence interval. If the same unknown L applies to both states, the inferred difference is Δκ_lat = κ_lat,B − κ_lat,A = (90 − 100) − 300L(8.00 × 10⁶ − 1.00 × 10⁷) = −10 + 6.00 × 10⁸ L. The range then gives +3.20 to +5.60 W/(m K), so its sign remains positive within that shared-L assumption.

If L_A and L_B can instead vary independently over that range, A’s residual is 22.0–34.0 and B’s is 27.6–37.2 W/(m K). Their possible difference spans 27.6 − 34.0 = −6.40 to 37.2 − 22.0 = +15.2 W/(m K). The sign is no longer identified. A common model parameter and two independent parameters are different physical assumptions, even when their individual ranges look identical. A residual is an inferred contribution under those assumptions, not a direct phonon measurement or proof of which defect changed.

A boundary drop is not a change of interior conductivity

Energy crossing an interface must couple to the modes or carriers available on the other side. Bonding, structure and transmission therefore matter. It is not generally correct to treat every boundary as a transparent continuation of bulk material or to predict its conductance from a defect count alone. For a local steady interface model, define an area-specific thermal boundary resistance r_b by ΔT_b = q r_b. Here q is heat rate per area in W/m², r_b is in m² K/W and ΔT_b is the positive hot-side minus cold-side temperature drop.

The two temperatures in this boundary model are the continuum temperatures extrapolated from the adjacent, locally equilibrated interiors. The jump is an ideal representation of a thin interfacial region, not a claim that one atom has two temperatures. Direction-dependent incoming carrier populations and an ultrafast electron population need not have those same temperatures. Consistent temperature definitions belong to the model.

Use a separate fictional insulating crystal assembly with two equal pieces and one interface halfway through its total length H. Every interior has κ_bulk = 100 W/(m K), and the interface has supplied r_b = 1.00 × 10⁻⁶ m² K/W. The cross-sectional area is A = 1.00 mm² = 10⁻⁶ m², and the steady heat rate is Q̇ = 0.100 W, so q = Q̇/A = 1.00 × 10⁵ W/m². Exclude other contact drops, side losses, heat generation and storage. Assume each piece is long enough for a local, constant-κ diffusive description at this state.

Within the two interiors the combined drop is qH/κ_bulk. Add the one interfacial drop q r_b: ΔT_total = q(H/κ_bulk + r_b). At H = 100 µm the interior and boundary drops are each 0.100 K, giving 0.200 K total. At H = 1000 µm the interior drop is 1.00 K but the boundary drop is still 0.100 K, giving 1.10 K total. The interface did not become more resistive in the shorter assembly; it simply occupies more of the total resistance account.

Define a whole-assembly equivalent conductivity κ_eq = qH/ΔT_total = H/(H/κ_bulk + r_b). It is 50.0 W/(m K) for 100 µm and 90.9091 W/(m K) for 1000 µm. These are legitimate specified assembly responses, not different intrinsic κ_bulk values. A length-dependent result can therefore arise while each interior remains diffusive; it is not by itself proof of ballistic transport or quantum confinement. The actual temperature profile and location of the drop help separate those explanations.

Simple links have restricted regimes

Electrical and thermal conductivities can be related in some metallic regimes because electrons contribute to both. That does not supply a universal relationship for polymers, ceramics or composites. At small scales, boundary effects can make bulk Fourier descriptions incomplete.

Fourier’s local law q_x = −κ(dT/dx) assumes an appropriate local temperature and transport regime for the specified direction. Carrier relaxation distances must be small enough relative to the scale of the temperature variation, and observation times long enough relative to the relaxation being averaged. The valid interface model just used applies Fourier behaviour inside each piece and supplies a separate boundary relation. A very short path, a rapid pulse or unequal electron and lattice temperatures can need a different description; replacing every such case by a smaller bulk κ does not explain it.

A 2017 study of vanadium dioxide nanobeams near their metal–insulator transition reported an electronic heat response inconsistent with the conventional Lorenz estimate. Importantly, the lattice contribution was not simply assumed away: its evaluation combined first-principles calculations with X-ray scattering information, and thermal and electrical transport were compared along the same beam direction. The conclusion depended on that wider evidence and model. It is an example of testing the link, not authority to assign the same failure or coefficient to an unrelated specimen.

Keep the observations separate from the decomposition. A changed κ and σ can reflect different carrier populations, scattering, structure, direction, interfaces or measurement paths. Independent structural, temperature-dependent and carrier-sensitive evidence can challenge those alternatives. From measured transport to device behaviour follows contacts and operating state; Thermal diffusivity and response times separates a steady conductivity from a body’s response time.

Worked examples

Try the reasoning

1. Can grain boundaries affect heat transport?

   Show the answer and reasoning

   Yes.

   They can scatter carriers or add interfacial resistance.

2. Does smaller size always lower conductivity by the same factor?

   Show the answer and reasoning

   No.

   Carrier paths, interfaces and geometry set the trend.

3. A separate crossing record has equal rates 2.00 × 10¹⁵ electrons/s in both directions, but mean energies 0.0100 eV rightward and 0.0400 eV leftward. No reservoir temperatures are assigned. Find the signed current and energy rate with the same rounded conversion.

   Show the answer and reasoning

   Net current is zero and Q̇ = −9.60 µW: the record carries energy to the left.

   Equal numbers cancel charge current. The energy difference is −0.0300 eV, so 2.00 × 10¹⁵ × (−0.0300) × 1.60 × 10⁻¹⁹ = −9.60 × 10⁻⁶ W. Direction follows the weighted crossings, not the fact that the particles are electrons. This is not a claim of heat flowing from an assigned colder reservoir to a hotter one.

4. A new fictional matched bulk record at 400 K has σ = 5.00 × 10⁶ S/m and total κ = 70.0 W/(m K). If L = 2.44 × 10⁻⁸ W Ω K⁻² and the same two-channel model are explicitly supplied, find κ_e and κ_lat. Is σ alone enough if L is unspecified?

   Show the answer and reasoning

   κ_e = 48.8 and κ_lat = 21.2 W/(m K). Without an applicable L or other independent information, σ alone does not separate them.

   Use absolute temperature: LσT = 2.44 × 10⁻⁸ × 5.00 × 10⁶ × 400. Subtract that result from 70.0. The calculation uses stipulated inputs at 400 K; it does not transfer A’s 300 K material state or infer a temperature trend from unrelated specimens.

5. Another supplied 300 K record has σ = 1.00 × 10⁷ S/m but total κ = 50.0 W/(m K). The fixed-L calculation returns κ_lat = −23.2 W/(m K). Has a negative lattice conductivity been measured? What upper L would even permit a nonnegative residual?

   Show the answer and reasoning

   No. The assumed comparison is inconsistent. Nonnegative κ_lat would require L ≤ 1.66667 × 10⁻⁸ W Ω K⁻² for these exact inputs.

   The nominal electronic estimate is 73.2, already larger than the total. Requiring LσT ≤ κ gives L ≤ 50/(10⁷ × 300). Even the earlier 2.20–2.60 × 10⁻⁸ range fails. Recheck the applicable electronic model, material state, thermal/electrical paths, measurements and other contributions; this arithmetic alone does not prove a specific unconventional mechanism.

6. Retain A/B at 300 K but narrow the supplied admissible Lorenz range to 2.30–2.50 × 10⁻⁸ W Ω K⁻². Find the possible lattice change if one L is shared; then test whether independent L_A/L_B identify its sign.

   Show the answer and reasoning

   Shared L gives +3.80 to +5.00 W/(m K). Independent choices give −1.00 to +9.80 W/(m K), so their sign is still unsettled.

   The shared result is −10 + 6.00 × 10⁸L. Independently, A’s residual is 25.0–31.0 and B’s 30.0–34.8; subtract opposite interval endpoints to get −1.00 and +9.80. These are conditional ranges, not confidence levels, and the narrower marginal range still does not justify common transfer.

7. Keep the 100 µm one-interface assembly and its κ_bulk and r_b, but double its area to 2.00 mm² while keeping Q̇ = 0.100 W. Find q, total drop and κ_eq. Did r_b double or halve?

   Show the answer and reasoning

   q = 5.00 × 10⁴ W/m², ΔT_total = 0.100 K and κ_eq = 50.0 W/(m K). The supplied area-specific r_b is unchanged.

   Both interior and boundary drops halve with q. Their sum is 0.0500 + 0.0500 K. The whole boundary resistance in K/W equals r_b/A and therefore halves, but its area-specific parameter does not. The effective conductivity is unchanged because q and total drop scale together.

8. A rapid excitation leaves the electron population much hotter than the lattice during the observation interval. May the steady common-temperature A/B decomposition be used unchanged just because σ and a total energy input are available?

   Show the answer and reasoning

   No. The shared local temperature and steady transport conditions have not been established.

   Electron–lattice energy exchange, spatial temperature fields, observation time and the meaning of the measured response need a suitable nonequilibrium model. A transient energy input is not by itself a conductivity, and a fitted scalar value cannot repair missing physical assumptions.

What to carry forward

• Name heat carriers and their scattering mechanisms.
• Use bulk transport laws within an appropriate scale.
• Lattice energy can propagate while atoms oscillate locally; electron-carried heat can remain when net electrical current is zero.
• Conductivity depends on occupied modes, directional transport and relaxation, not simply the presence of more vibrations.
• A Lorenz-based residual depends on the material regime, matched observations and whether a model parameter transfers between states.
• Bulk gradients, interfacial drops and whole-assembly equivalent conductivity refer to different parts of the same energy account.