Electron Transpiration Cooling (ETC)

• Electron Transpiration Cooling (ETC) is a mechanism that selectively removes high-energy electrons to achieve efficient heat extraction in both cryogenic and hypersonic regimes.
• In cryogenic systems, superconducting tunnel contacts filter electrons above the gap, as demonstrated in PtSi-Si devices where the electron temperature drops markedly under optimal bias.
• In hypersonic applications, combined thermionic and photoemission processes, along with sheath and backflow effects, determine the net cooling performance and impose strict design constraints.

Electron transpiration cooling (ETC) denotes electron-mediated heat removal by preferentially extracting energetic electrons from a hot region. In the cited literature, the term spans two distinct physical realizations. In cryogenic semiconductor devices, electron transpiration is the energy-selective tunnelling of hot electrons into a superconductor, reducing the electron-gas temperature in a semiconductor island (Prest et al., 2014). In hypersonic aerothermodynamics, ETC refers to thermionic emission from a hot leading edge, with cooling determined by whether emitted electrons escape, are recollected, or redistribute heat over the surface (Ghosh et al., 11 Feb 2025, Boyer et al., 7 Aug 2025). Across both settings, ETC is governed by an energy filter, a return path for parasitic heat, and a global transport constraint that determines whether nominal emission translates into net cooling.

1. Definitions and physical regimes

In the low-temperature solid-state formulation, ETC is implemented with superconducting tunnel contacts that selectively remove electrons with energies above the Fermi level by more than the superconducting gap $\Delta$. Prest et al. describe this as electron cooling in silicon using platinum silicide as a superconductor contact, where hot electrons tunnel out of the semiconductor and the electron gas cools as a result (Prest et al., 2014).

In the hypersonic formulation, ETC is identified with thermionic cooling, also described as electron transpiration cooling, for leading edges that can reach temperatures exceeding $2000\,^\circ\mathrm{C}$. There, the emitted electron population is set by thermionic emission and, in some models, augmented by photoemission. Cooling is not determined by emission alone: reflected electrons, incident plasma electrons, ion effects, sheath structure, and downstream collection all enter the energy balance (Ghosh et al., 11 Feb 2025, Boyer et al., 7 Aug 2025).

A useful unifying interpretation is that ETC always depends on energy-selective charge removal from the emitting surface or electron gas, followed by a transport problem. In the superconductor–semiconductor case, the filter is the superconducting density of states. In hypersonic ETC, the filter is the work function plus sheath or virtual-cathode barrier. This suggests that ETC is less a single device class than a family of nonequilibrium electron-transport cooling mechanisms.

2. Cryogenic ETC in superconductor–semiconductor junctions

Prest et al. formulate cooling power per tunnel junction as

$$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$

where $R_N$ is the normal-state junction resistance, $V$ the bias per junction, and $f_N$ and $f_S$ the Fermi functions in the normal and superconducting electrodes at temperatures $T_e$ and $T_s$ respectively (Prest et al., 2014). Physically, electrons with energies above the Fermi level by more than $\Delta$ tunnel out, removing heat.

The superconducting density of states is written in Dynes form,

$2000\,^\circ\mathrm{C}$0

with $2000\,^\circ\mathrm{C}$1 the Dynes sub-gap leakage parameter and $2000\,^\circ\mathrm{C}$2 following the BCS temperature dependence, $2000\,^\circ\mathrm{C}$3, with $2000\,^\circ\mathrm{C}$4 at $2000\,^\circ\mathrm{C}$5 (Prest et al., 2014). The role of $2000\,^\circ\mathrm{C}$6 is central: sub-gap leakage enables unwanted low-energy tunnelling and directly degrades cooling.

The dominant parasitic load is electron–phonon coupling,

$2000\,^\circ\mathrm{C}$7

where $2000\,^\circ\mathrm{C}$8 is the material-specific coupling constant, $2000\,^\circ\mathrm{C}$9 the electron-gas volume, and $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$0 the lattice temperature (Prest et al., 2014). Because the dependence is $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$1, the lattice heat load falls rapidly at sub-kelvin temperatures, enabling net cooling. Heat balance for an S–Sm–S device is then written as

$$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$2

with $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$3 the residual Joule heating in series resistances (Prest et al., 2014).

The device realization uses a 10 nm PtSi thin film on silicon-on-insulator with a 140 nm buried oxide. The central island is implanted with As to $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$4 and has sheet resistance $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$5. Each PtSi–Si contact has area $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$6, and two junctions in series form an S–Sm–S cooler. The measured normal-state resistance is $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$7 per junction, with series island resistance $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$8. Aluminium pads overlap PtSi for four-point I–V measurements, eliminating contact-resistance artefacts (Prest et al., 2014).

3. Thin-film PtSi realization and measured cooling

The PtSi implementation is motivated by gap engineering. Bulk PtSi has $$\dot Q(V,T_e,T_s)=\frac{1}{e^2R_N}\int_{-\infty}^{\infty}(E-eV)\,n_S(E)\,[f_N(E-eV,T_e)-f_S(E,T_s)]\,dE,$$9, whereas a 10 nm film suppresses $R_N$0 to $R_N$1, corresponding to $R_N$2 (Prest et al., 2014). A smaller $R_N$3 reduces sub-gap leakage currents $R_N$4 and shifts the optimum cooling bias to lower voltages, improving performance at $R_N$5. Thin films also crystallize better, yielding smoother PtSi–Si interfaces and higher-quality tunnel barriers (Prest et al., 2014).

At $R_N$6, the measured I–V characteristics show strong suppression of sub-gap current near $R_N$7, while $R_N$8 exhibits the classic U-shape of a superconducting tunnel junction with rounded peaks at $R_N$9 (Prest et al., 2014). Fitting the isotherm model with fixed $V$0 gives $V$1 and $V$2. When the heat-balance model is incorporated, the bias-dependent electron temperature drops from $V$3 to $V$4 at $V$5 (Prest et al., 2014).

Peak cooling power at $V$6 is $V$7, or $V$8, and the coefficient of performance is described as typically a few percent at optimum bias (Prest et al., 2014). The lower $V$9 of PtSi allows operation down to $f_N$0, below the regime easily accessed with Al, for which $f_N$1 (Prest et al., 2014).

The limiting factors are explicit. Dynes sub-gap leakage of $f_N$2 is higher than in state-of-the-art NIS coolers, where $f_N$3–$f_N$4 is typical. Series resistance narrows the usable bias window and introduces Joule heating, while residual thermal coupling to the lattice and environmental heat leaks set a floor to the minimum achievable $f_N$5 (Prest et al., 2014). The same Schottky-barrier tunnel junctions can in principle operate as thermometers as well as refrigerators, and PtSi/Si coolers could be monolithically integrated with silicon qubits, single-electron pumps, or ultra-low-noise detectors (Prest et al., 2014).

4. Thermionic and photoemission ETC for hypersonic surfaces

For hypersonic leading edges, the fundamental thermionic-emission law is the Richardson–Dushman relation

$f_N$6

with local cooling power density estimated as

$f_N$7

when each escaping electron carries away an average energy on the order of $f_N$8 (Ghosh et al., 11 Feb 2025). Boyer and Fisher write the same ideal thermionic cooling contribution per unit area as

$f_N$9

with $f_S$0 the actual emission current, emphasizing that the useful cooling rate depends on the emission that survives sheath constraints rather than the Richardson–Dushman value alone (Boyer et al., 7 Aug 2025).

Ghosh and Fisher extend the formulation to photoemission using a modified Fowler–DuBridge description. The spectral electron emission density including photon energy $f_S$1 is written as $f_S$2 and, with a virtual-cathode potential included, as $f_S$3. The reflected current density per spectral bin is then

$f_S$4

where $f_S$5 is the normalized 3D random-energy probability density (Ghosh et al., 11 Feb 2025). In this framework, net cooling is the difference between emitted and reflected electron energy fluxes.

The sheath or virtual-cathode structure is prescribed in Cartesian, cylindrical, and spherical coordinate systems. The reported sheath depths and extents are $f_S$6 and $f_S$7 for Cartesian, $f_S$8 and $f_S$9 for cylindrical, and $T_e$0 and $T_e$1 for spherical geometry (Ghosh et al., 11 Feb 2025). Electron trajectories are launched from a random-energy distribution,

$T_e$2

and propagated under

$T_e$3

using a Runge–Kutta 5(4) ODE integrator (Ghosh et al., 11 Feb 2025).

The surface geometry in that study is a planar disk of radius $T_e$4, representing a local stagnation-point region. Two temperature profiles are examined: a step function with $T_e$5 for $T_e$6 and $T_e$7 for $T_e$8, and a Gaussian profile $T_e$9 (Ghosh et al., 11 Feb 2025). The characteristic gradient length is stated to be comparable to the electron lateral-travel distance, making ETC partly a heat-spreading mechanism rather than only a local heat-removal mechanism.

Photoemission contributes materially in the visible-light model with $T_s$0 and $T_s$1, where total emitted current can increase by $T_s$2–$T_s$3 over pure thermionic values at $T_s$4, and preliminary tests show $T_s$5 up to $T_s$6 larger when photoemission is included (Ghosh et al., 11 Feb 2025). A plausible implication is that ETC optimization for hypersonic surfaces may require co-design of temperature field, work function, and illumination spectrum.

5. Kinetic transport limits, backflow, and misleading cooling metrics

Zhang et al. model ETC-relevant transport with a one-dimensional-in-space, three-dimensional-in-velocity electrostatic PIC–MCC plasma diode spanning a full cathode–anode gap (Zhang et al., 8 May 2026). The domain length is $T_s$7 with $T_s$8 cells, $T_s$9, time step chosen so that $\Delta$0, and $\Delta$1 particles per cell (Zhang et al., 8 May 2026). Poisson’s equation is solved with $\Delta$2 and $\Delta$3; thermionic emission is injected at the cathode as a prescribed flux $\Delta$4 sampled from a half-Maxwellian at $\Delta$5 (Zhang et al., 8 May 2026).

The diagnostic structure distinguishes emitted flux, reflected flux, net transport, and anode collection:

$\Delta$6

Cathode-side cooling metrics are defined as

$\Delta$7

and

$\Delta$8

with the ion-recombination term written as $\Delta$9 and $2000\,^\circ\mathrm{C}$00 (Zhang et al., 8 May 2026).

The principal numerical result is a sharp transition from weak-backflow transport to backflow-limited transport. Below $2000\,^\circ\mathrm{C}$01, the backflow ratio remains below $2000\,^\circ\mathrm{C}$02 and $2000\,^\circ\mathrm{C}$03. At $2000\,^\circ\mathrm{C}$04, $2000\,^\circ\mathrm{C}$05, $2000\,^\circ\mathrm{C}$06, and $2000\,^\circ\mathrm{C}$07. Above $2000\,^\circ\mathrm{C}$08, further emission produces over-compensation: $2000\,^\circ\mathrm{C}$09 rises to $2000\,^\circ\mathrm{C}$10 while $2000\,^\circ\mathrm{C}$11 and $2000\,^\circ\mathrm{C}$12 decrease (Zhang et al., 8 May 2026).

The transition is associated with full-gap potential restructuring. Near $2000\,^\circ\mathrm{C}$13, the interior potential collapses and the rise is concentrated near the anode, weakening the near-cathode barrier and promoting backflow (Zhang et al., 8 May 2026). A purely local virtual-cathode estimate,

$2000\,^\circ\mathrm{C}$14

would give $2000\,^\circ\mathrm{C}$15 for the shallow local dip $2000\,^\circ\mathrm{C}$16, whereas the full-gap effective barrier

$2000\,^\circ\mathrm{C}$17

shows that the loss is dominated by global sheath and transport collapse rather than a single local dip (Zhang et al., 8 May 2026).

This result directly addresses a common misconception: stronger emission does not necessarily improve ETC. Zhang et al. show that cathode-side cooling metrics may continue to grow after the transition because they count energy removed at emission, even when emitted electrons fail to escape the full gap (Zhang et al., 8 May 2026). In that regime, boundary-emission diagnostics cease to be reliable proxies for useful ETC transport.

6. Passive ETC energetics, system-level constraints, and design window

Boyer and Fisher analyze passive ETC for a discretized $2000\,^\circ\mathrm{C}$18-radius tungsten leading edge with wall temperature near $2000\,^\circ\mathrm{C}$19 at stagnation, embedding a one-dimensional collisionless plasma sheath model into a leading-edge framework (Boyer et al., 7 Aug 2025). The sheath model uses

$2000\,^\circ\mathrm{C}$20

with wall and sheath-edge boundary conditions, flowfield electron flux

$2000\,^\circ\mathrm{C}$21

and ion flux

$2000\,^\circ\mathrm{C}$22

The effective work function is reduced by the Schottky effect:

$2000\,^\circ\mathrm{C}$23

The maximum floating-emitter current with cold ions is given as

$2000\,^\circ\mathrm{C}$24

and the actual emission current must be found iteratively with the sheath solution (Boyer et al., 7 Aug 2025).

The complete surface energy balance includes convective–kinetic heating, radiative heating, thermionic cooling, flowfield electron and ion collection heating, and Joule heating:

$2000\,^\circ\mathrm{C}$25

or, more explicitly,

$2000\,^\circ\mathrm{C}$26

This formulation makes the central systems result explicit: ETC cooling must be compared against the heating caused by recollected flowfield electrons and ions, not only against emission-side energy removal (Boyer et al., 7 Aug 2025).

The parametric study reveals a narrow operating window. Defining $2000\,^\circ\mathrm{C}$27 as the minimum seeding needed to just prevent a virtual cathode, corresponding to $2000\,^\circ\mathrm{C}$28, the Schottky-enhanced Richardson model predicts $2000\,^\circ\mathrm{C}$29 cooling at $2000\,^\circ\mathrm{C}$30, while the full ETC circuit yields only $2000\,^\circ\mathrm{C}$31 (Boyer et al., 7 Aug 2025). At $2000\,^\circ\mathrm{C}$32, Schottky theory increases cooling by $2000\,^\circ\mathrm{C}$33, yet the full circuit loses $2000\,^\circ\mathrm{C}$34 of cooling, to $2000\,^\circ\mathrm{C}$35. At $2000\,^\circ\mathrm{C}$36, the circuit reverses and heats the leading edge by about $2000\,^\circ\mathrm{C}$37 (Boyer et al., 7 Aug 2025).

That reversal is attributed to the floating potential at the stagnation point becoming less positive than downstream, so charge conservation drives electrons away from the stagnation region and deposits heat instead of extracting it (Boyer et al., 7 Aug 2025). Raising the work function from $2000\,^\circ\mathrm{C}$38 to $2000\,^\circ\mathrm{C}$39 at $2000\,^\circ\mathrm{C}$40 reduces emission until the ETC circuit produces net heating instead of cooling, and above $2000\,^\circ\mathrm{C}$41 heating grows rapidly (Boyer et al., 7 Aug 2025).

The same study identifies blunt radii and dielectric coatings as protective but ETC-suppressing. Blunter radii of $2000\,^\circ\mathrm{C}$42–$2000\,^\circ\mathrm{C}$43 behave more like floating surfaces, so ETC is essentially off. An insulating coating forces $2000\,^\circ\mathrm{C}$44 everywhere, reducing adverse overcompensation heating by up to $2000\,^\circ\mathrm{C}$45 but also blocking net cooling at stagnation (Boyer et al., 7 Aug 2025). Reducing wall resistivity by two orders of magnitude recovers $2000\,^\circ\mathrm{C}$46 of lost cooling, after which diminishing returns set in; collector lengths beyond $2000\,^\circ\mathrm{C}$47 likewise show diminishing returns, while even $2000\,^\circ\mathrm{C}$48 retains $2000\,^\circ\mathrm{C}$49 of nominal cooling (Boyer et al., 7 Aug 2025).

The design guidance emerging from these results is internally constrained rather than monotonic. Passive ETC should operate in a narrow window near exact or slight under-compensation, stated as $2000\,^\circ\mathrm{C}$50 for the examples studied; use low-$2000\,^\circ\mathrm{C}$51 materials or surface treatments in the range $2000\,^\circ\mathrm{C}$52–$2000\,^\circ\mathrm{C}$53; maintain high electrical conductivity without driving the system into severe overcompensation; and use sharp leading-edge radii near $2000\,^\circ\mathrm{C}$54 if ETC is to remain active (Boyer et al., 7 Aug 2025). This suggests that ETC for hypersonic vehicles is not limited by raw thermionic capability but by a coupled electrothermal circuit condition in which collection heating, sheath stability, and geometry can nullify or reverse nominal cooling.