Electrical Generation of Surface Phonon Polaritons

Christopher R. Gubbin and Simone De Liberato∗

School of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, United Kingdom
(Dated: April 25, 2023)
Efficient electrical generation of mid-infrared light is challenging because of the dearth of materi-
als with natural dipole-active electronic transitions in this spectral region. One approach to solve
this problem is through quantum-engineering of the electron dispersion to create artificial transi-
tions, as in quantum cascade devices. In this work we propose an alternative method to generate
mid-infrared light, utilizing the coupling between longitudinal and transverse degrees of freedom
due to the nonlocal optical response of nanoscopic polar dielectric crystals. Polar crystals support
arXiv:2206.03812v3 [cond-mat.mes-hall] 24 Apr 2023

sub-diffraction photonic modes in the mid-infrared. They also support longitudinal phonons, which
couple efficiently with electrical currents through the Fröhlich interaction. As we have shown in
previous theoretical and experimental works, these two degrees of freedom can hybridize forming
longitudinal-transverse polaritons. Here we theoretically demonstrate that longitudinal-transverse
polaritons can be efficiently generated by electrical currents, leading to resonant narrowband pho-
tonic emission. This approach can therefore be utilised to electrically generate far-field mid-infrared
photons in the absence of dipole-active electronic transitions, potentially underpinning a novel gen-
eration of mid-infrared optoelectronic devices.

INTRODUCTION operating there have to rely on artificial electronic tran-

sitions, created through bandgap engineering [29–31].
Surface phonon polaritons (SPhPs) are hybrid light- Recent works in surface phonon polariton physics have
matter excitations formed from the coupling of free- suggested a solution to this problem.
photons to optical phonons in a polar dielectric crys-
While longitudinal and transverse excitations are or-
tal [1, 2]. They are highly promising for mid-infrared
thogonal and thus non-interacting in bulk linear polar
nanophotonics because, like plasmons in the visible,
crystals, in systems with broken translational symme-
they are morphologically dependent sub-diffraction exci-
try they are able to hybridise through common bound-
tations. This means they are largely tuneable [3–8] and
ary conditions on the electrical and mechanical fields.
have great potential in any application which benefits
This effect was recently demonstrated in a study of hy-
from a strong electric field such as sensing [9], nonlin-
bridization between transverse SPhPs localised in 4H-SiC
ear optics [10–12] or near field imaging [13, 14]. Novel
nanopillar resonators and zone-folded longitudinal opti-
properties of hyperbolic SPhPs [15–21] allow for great
cal (LO) phonons [32]. The resulting excitations were
flexibility in the design of mid-infrared SPhP based de-
termed longitudinal-transverse polaritons (LTPs): a ref-
vices compared to systems operating in the visible [22].
erence to the hybrid nature of their electric field and to
Moreover SPhP resonances are derived from the lattice
the resulting anti-crossing in the modal dispersion around
phonons, and do not require a high electron density allow-
the LO phonon frequency. Our theory is a version with
ing them to have significantly narrower linewidths than
retardation of Ridley’s one [33–35], in which the role of
mid-infrared plasmonic alternatives [23].
interface polaritons is played by SPhPs and the role of
In the field of optoelectronics [24] a key proposed ap-
hybridons by LTPs. The inclusion of retardation in our
plication of SPhPs is in mid-infrared thermal emis-
theory allows us to describe nanostructured systems in
sion. Through Kirchhoff’s law their intrinsically narrow
which light can be outcoupled by arrays of defects [36].
linewidths allow for the design of narrowband thermal
Importantly, the longitudinal polarization field of the LO
emitters with directional and spectrally tuneable emis-
component can couple directly to charge currents tran-
sion [25–28]. Despite these impressive results thermal
siting through the lattice. The electrical generation of
emitters run on spontaneous emission from the thermally
localised phonons in polar nanolayers has been studied
oscillating charges of the polar lattice and cannot achieve
previously in the context of both optical [35, 37–39] and
substantial temporal coherence or lasing. Moreover, the
acoustic phonons [40]. It is known to lead to measur-
only way to enhance emission is by increasing the de-
able effects such as velocity saturation in semiconductor
vice temperature which leads to large short wavelength
devices [41, 42]. It has also previously been predicted
emission and diminished efficiency in the target spectral
that coherent generation of localised LO phonons could
range.
be possible in GaN quantum wells [43, 44].
Resonant electrical injection would solve these problems
In this Paper we explore the idea that the unique hybrid
but materials do not typically have strong interband
nature of LTPs can allow them to be utilised as intercon-
transitions in the mid-infrared and optoelectronic devices
nects between electronic and photonic degrees of freedom
in optoelectronic devices. We develop a transparent the-
ory of LTP-driven Fröhlich scattering in a planar device,
∗ s.de-liberato@soton.ac.uk calculating the efficiency of SPhP emission relative to the
 2

a) LO Phonon SPhP the Reststrahlen region between the lattice longitudinal

k and transverse optical phonon frequencies. In this spec-
tral window a polar crystal is characterized by a negative
dielectric function and responds like a metal to external
P
P electromagnetic fields. In local optical theories the po-
larization field of SPhPs is almost everywhere divergence
free, ∇·P = 0, excepting the interfaces of the polar mate-
Longitudinal Transverse rial. In this Paper we refer to SPhPs as transverse excita-
tions, signifying that their polarization field and complex
b) wavevectors are orthogonal in each medium (k·P = 0), as
schematically shown in Fig. 1a. SPhPs are purely trans-
Ω Ω Ω ...= verse because local optics parameterizes the transverse
SPhP LO SPhP LO LTP dielectric function of the polar lattice with only the zone-
center k = 0 phonon frequencies, neglecting dispersion of,
and energy transport in, the optical phonon branches.
c) As we demonstrated in a recent series of publications,
e−
Fröhlich Interaction which allowed us to quantitatively reproduce previously
unexplained experimental data [45–49], accounting for
the optical phonon dispersion leads to a nonlocal the-
Dressed Fröhlich Interaction LO
e− ory of polar optics analogous to that in plasmonic sys-
tems [50–54] in which LO phonons and SPhPs interact
e− e− through shared electro-mechanical boundary conditions.
In large resonators the effect is an increased non-radiative
Ω ... = damping [45] as propagative phonons leach energy from
LO SPhP LTP the SPhP, but in nanoscale systems the interaction can
e− e− become coherent. In this regime energy passes multiple
times between LO phonons and SPhPs before dissipat-
FIG. 1: a) Comparison of the polarization field P for a ing so the two modes can no longer be considered dis-
longitudinal optical (LO) phonon and a transverse tinct. Instead they are best understood as a hybrid exci-
surface phonon polariton (SPhP). In the former the tation, termed longitudinal-transverse polariton (LTP),
polarization field is parallel to the wavevector k, while illustrated in Fig. 1b, whose electric field is a linear mix-
in the latter it is orthogonal to it. b) Illustration of ture of each component. The crossover point between
strong coupling between LO and SPhP modes. Energy these regimes occurs where the thickness of the polar
is passed coherently at frequency Ω between the two medium approaches the longitudinal phonon propagation
degrees of freedom, leading to the formation of a hybrid length in the material and a discrete LO phonon spec-
LTP mode. c) Diagrammatic sketch of how the coupling trum emerges. At this limit an appreciable fraction of
between LO phonons and SPhPs can dress the Fröhlich LO phonons excited at the edge of the polar crystal are
interaction between electrons and LO phonons, leading able to transit the layer before decaying non-radiatively,
to a resonant interaction between electrons and LTPs. allowing their energy to recycle into SPhPs.

Longitudinal phonons can be excited by electrons passing

thermal case. We demonstrate that electrically excited
emission is possible, suggesting that LTP interconnects
could be widely useful in optoelectronic devices operating
across the mid-infrared.
I. OVERVIEW
Polar crystals support short-wavelength LO phonons.
These are dispersive, meaning their frequency is a func-
tion of wavevector ωL (k). Longitudinal phonons are
characterized by a curl-free polarization field, ∇ × P = 0,
which in bulk implies P is parallel to the propagation
vector k, so k × P = 0, as illustrated in Fig. 1a. Op-
tical phonons at the surface of a polar crystal can also
couple to photons, forming sub-diffraction optical modes
termed surface phonon polaritons (SPhPs), which exist in
through a polar crystal. Charged electrons perturb the
ions of the crystal lattice from their equilibrium positions.
This emission of LO phonons, named after Fröhlich, is
one of the leading dissipation channels for Ohmic losses
and is illustrated by the upper diagram in Fig. 1c. In
a system supporting LTPs the emitted LO phonons can
convert energy into SPhPs, and following the scheme in
Fig. 1b this leads to a dressed Fröhlich interaction in
which electrons resonantly emit LTPs, illustrated in the
lower diagram Fig. 1c. As SPhPs can couple to the exter-
nal environment through nanophotonic elements, emit-
ting radiation in the far-field, this mechanism is able to
act as an interconnect between electronic and photonic
sub-systems, allowing injected electrical currents to radi-
ate photons in the far-field.
In this Paper we study the simple example of an LTP
interconnect consisting of a thin polar film of thickness

E∥
εc
V d ε modal frequency < ωqSP is indicated by the green dash-
εc
FIG. 2: Sketch of the device studied in this Paper.
SPhPs supported by a polar layer of thickness d and
dielectric function  in a cladding semiconductor with
dielectric constant c are excited by application of a
voltage across the film. The SPhPs are then emitted at
a rate proportional to their photonic fraction.
d embedded in a bulk material with dielectric constant
c , sketched in Fig. 2 [47]. For simplicity we calculate
the modes of our device considering infinite lateral ex-
tent. We consider the polar layer to be cubic silicon
carbide (3C-SiC), characterized in the local limit by the
Lorentzian dielectric function
ωL2 − ω (ω + iγ)
 (ω) = ∞ 2 − ω (ω + iγ) , (1)
ωT
where ωL = 973cm−1 is the zone-center LO phonon fre-
quency, ωT = 796cm−1 is the zone-center TO phonon
frequency, γ = 4cm−1 is the non-radiative damping rate,
and ∞ = 6.52 is the high-frequency dielectric constant
[55]. The cladding is taken as silicon (Si), characterized
by the dielectric constant c = 11.71.
The polar layer supports a SPhP mode at each inter-
face which are uncoupled in a thick film. When d ap-
proaches the skin-depth of the SPhP they begin to hy-
bridize forming two polariton branches, one with fre-
quency below the uncoupled mode frequency, and one
with frequency above [56]. In this work we concern our-
selves with the high-frequency SPhP whose frequency is
close to the zone-center longitudinal phonon frequency
[57]. In Appendix A 1 the complex frequency of this mode
is derived in the local-response-approximation, where the
dispersion of the underlying phonon modes of the lattice
is disregarded. In Fig. 3a we plot the local density of
states which describes the SPhP dispersion
 field X and the normal component of the lattice stress
SP 1 = ωqSP
ρq (ω) = 
 , (2)
π ω − < ω SP 2 + = ω SP 2
q q
in which ωqSP is the SPhP frequency as a function of in-
plane wavevector q for a layer thickness d = 2nm. At
small wavevectors the mode is in the epsilon-near-zero
(ENZ) regime [58, 59] where the SPhP phase shift over
the film thickness is negligible, and the frequency lies
near the zone-center LO phonon frequency ωL , indicated
3
by the black dashed line in Fig. 3a. At large wavevectors
this is no longer the case, the SPhPs on each interface
of the polar layer begin to decouple and the mode red
shifts toward the asymptotic SPhP frequency of the bi-
layer which it reaches when q is sufficiently large that
the SPhPs on each interface no longer overlap. The real
dotted line. Note that for simplicity in Fig. 3 and for
the remainder of this Paper we ignore the lower energy
SPhP mode. This excitation, whose small-wavevector
dispersion begins at ωT approaches the same asymptotic
frequency from the low-wavenumber side in the large
wavevector limit.
A thin film whose thickness approaches the LO phonon
propagation length also acts as a Fabry-Pérot resonator
for LO phonons, leading to the appearance of a discrete
spectrum of longitudinal modes with quantized modal
frequencies
q
L
ωq,n = ωL2 − βL2 (|q|2 + ξn2 ), (3)
in which n ∈ N labels the discrete phonon branch,
βL = 15.3 × 105 cm s−1 is the LO phonon velocity in the
low-wavevector quadratic regime [45, 60] and the quan-
tized out-of-plane wavevector of the nth localised phonon
mode is given by
(2n − 1)π
ξn = . (4)
d
The factor 2n − 1 here ensures only LO phonons with
equal parity to the high-frequency SPhP branch are con-
sidered, as in the symmetric structure under consider-
ation only these modes are coupled. The remaining
phonon modes couple to the low-frequency SPhP and are
not considered in this Paper.
In the full nonlocal problem the transverse and longitudi-
nal electromagnetic fields in the waveguide must be cal-
culated simultaneously. As the standard Maxwell bound-
ary conditions on the transverse components of the elec-
tric and magnetic fields leave an under-specified prob-
lem this requires the application of additional boundary
conditions. We have previously derived the appropriate
additional conditions on the mechanical fields in the po-
lar layer considering continuity of the nonlocal Poynting
vector, which accounts for energy carried in both the elec-
tromagnetic and mechanical fields, across each interface
[45]. This imposes continuity of the lattice displacement
tensor τ̄ . Note that in this work we ignore TO phonon
modes in the polar material, an approximation we pre-
viously demonstrated was valid in SiC because of the
large LO-TO splitting [47]. In this limit the problem
is over-specified and it is not possible to utilize all the
additional boundary conditions, we choose to apply the
Pekar-Ridley condition [61] X = 0 at the interfaces of the
polar film. Under these approximations it is possible to
derive the analytic expression for the LO-SPhP coupling
 4

a) 1000 b) 1000
j=u 1.5

Density of States (×10−1 cm)

ωL
n=1 n=1
Wavenumber (cm−1 )

Wavenumber (cm−1 )
950 950 1
j=m

900
n=2 n = 20.5
j=l

850 850 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
In-plane Wavevector (×108 m−1 ) In-plane Wavevector (×108 m−1 )

FIG. 3: Density of states and modal dispersions for a 3C-SiC film of thickness 2nm characterised by Eq. 1,
sandwiched in a Si cladding (c = 11.71). In panel a), representing the prediction of a local theory, the black dashed
line shows the bulk LO phonon frequency ωL and the dot-dashed green line the local SPhP dispersion
ωqSP (Appendix A 1). In panel b), representing the prediction of a nonlocal theory, the black dashed lines shows the
L LTP
localised LO phonon frequencies ωq,n and the dot-dashed green lines the LTP frequencies ωq,j . In both panels blue
solid lines mark edge of the light-cone in the cladding.

frequency [48] The nonlocal physics which leads to LTP formation

2βL2 ωL2 − ωqSP 2
|Ωq,n |2 =
d2 ωqSP ωq,n
L
The LO-SPhP coupling is mediated by interactions
between the two excitations at the surface of the polar
layer, meaning it is highly sensitive to the time the LO
phonon takes to propagate across the film. For this
reason it is diminished in thicker films, and for modes
close to the zone-center LO frequency where the phonon
group velocity is small.
For a 2nm 3C-SiC film we observe two localised LO
modes in the spectral window under study, illustrated
by the the n = 1 and n = 2 dashed black lines in
Fig. 3b. The LTP modes are linear superpositions of
the underlying SPhP and LO phonon modes coupled
at frequencies given by Eq. 5, whose eigenfrequencies
can be derived explicitly by Hopfield diagonalization
[62] as discussed in Appendix A 2. The complex LTP
LTP
frequencies are given by ωq,j where j labels the discrete
polariton branches. The total number of LTP branches
is equal to the number of bare modes: one greater than
the number of localised LO phonons. We find three
modes: an almost dispersionless upper polariton ω > ωL
(j = u), a middle branch (j = m) sandwiched between
the localised LO frequencies, and a lower polariton
branch (j = l) which tends to the the ENZ modal
frequency at large in-plane wavevectors. The multimode
analogue of Eq. 2 is plotted in the colormap Fig. 3b,
LTP
while real modal frequencies < ωq,j are shown with
dash-dotted green lines and marked by the relative value
of j.
. (5)
can be understood considering the dispersion of the
LO phonon branch in the lattice. Away from the
zero-wavevector Γ point the band red-shifts. At each
frequency the LO phonon mode has a unique wavelength,
calculable from the dispersion relation, and when an
integer multiple of this wavelength fits within the polar
film a Fabry-Pérot mode forms. These discrete modes
couple to the SPhP, leading to the altered dispersion in
Fig. 3b. The wavevectors of the three localised modes
participating in the coupling satisfy the resonance
condition in Eq. 4.
The process of LTP-driven electrically excited emission
sketched in Fig. 1 has many moving parts which must be
taken into account to correctly model the conversion of
electrical to photonic energy. Electrons are driven by ex-
ternally applied voltages and can radiate LTPs which can
then either decay radiatively emitting photons into the
far-field or non-radiatively. To arrive at a fundamental
and intelligible understanding of the potential of LTP in-
terconnects for the design of mid-infrared optoelectronic
devices we keep the full microscopic complexity of the
LTP physics and make important simplifications to the
description of the couplings to the external environment:
I We do not utilize a self-consistent out-of-equilibrium
many-body theory of the electrical transport [59].
Instead we consider an electron gas at thermal equi-
librium with a well defined temperature Te described
by a Maxwell-Boltzmann distribution
8π 3 e−ε~k /kB Te
f~k (Te ) = , (6)
V Nc (Te )
 5
h ∗ i3/2
where Nc (Te ) = 4η m2π~ k B Te
is the density of temperature Te and the LTP population. The emission
2
(+) and absorption (−) coefficients are defined in Ap-
states in the conduction band, η is the degeneracy pendix D. When the lattice is at equilibrium with the
factor in the conduction band, V is the electronic electron gas the LTP population remains stationary, im-
quantisation volume, ~k is the three-dimension elec- plying that ΓLTP TH
q,j (Nq,j (T ), T ) = 0. We do not mark ex-
tron wavevector and ε~k is the electron energy. In Ap- plicitly the dependence upon the electron density ne , but
pendix C we consider the balance between electron both Γ+ −
q,j (Te ) and Γq,j (Te ) are linearly proportional to it.
energy and drift velocity in the thermalized electron In Eq. 7 we also introduced a finite lattice temperature Tl
gas, demonstrating that both the drift velocity and to account for thermal LTP occupation, with equilibrium
applied electric field between the device contacts can thermal distribution
be uniquely determined using only the temperature
" ! #−1
Te and carrier density ne of the gas. In all calcu- ~ω LTP
TH q,j
lations presented in this Paper the electron density Nq,j (T ) = exp −1 . (9)
is taken to be ne = 1018 cm3 . This density is cho- kB T
sen low enough not to significantly perturb the SPhP
resonances [60] and to allow us to consider the gas The population in the steady-state Ṅq,j = 0 is therefore
to be non-degenerate, removing the need to consider given by
Pauli blocking in the outgoing electronic states (see
Appendix C). ss
Nq,j (Tl ) + Γ+
LTP TH
γq,j q,j (Te )
Nq,j (Te , Tl ) = LTP + Γ− (T ) − Γ+ (T )
. (10)
II LTPs in the system are below-light-line excitations. γq,j q,j e q,j e
An outcoupling mechanism must be introduced in or-
der to allow them to radiate in the far-field. We will We also introduce a related quantity, the spectral steady-
not evaluate a specific extraction grating or prism ge- state population, which allows to take into account the
ometry, as here our aim is to provide a general eval- finite linewidth of the LTP resonances. This relates to
uation of the efficiency of the underlying emission the branch-resolved population through
mechanism. A good estimate of the level at which ss ss
an LTP is able to couple to the far-field can be ob- Nq,j (ω, Te , Tl ) = Nq,j (Te , Tl ) ρLTP
q,j (ω) , (11)
tained by considering how photon-like the mode is.
The effective photon population is given by the LTP in which ρLTP
q,j (ω) is the density of LTP states at wavevec-
population times its photonic fraction, equal to the tor q in branch j at frequency ω. This density is modeled
ratio of the group velocity to the speed-of-light in as a Lorentzian and is defined in Appendix A2.
vacuum Note that our rate-equation does not account for Joule
LTP
heating of the phonon reservoir. Although in a realis-
1 dωq,j tic device the temperature of the electronic sub-system
ξq,j = . (6)
c dq Te and of the lattice Tl would evolve toward an equi-
librium over time, describing this process is beyond the
In this model polaritons with vanishing group ve-
scope of this Paper as the dynamics would be device- and
locity, in the flat-band region of the LTP disper-
operation-dependent.
sion, have a vanishing coefficient ξq,j as they are es-
sentially just localised LO phonon modes decoupled
from the SPhP. Those in regions near anti-crossings
achieve instead a finite ξq,j from their SPhP compo- II. RESULTS
nent and for suitably small values of the wavevector
they can be extracted using standard and relatively LTP-driven electrically excited emission is the result
efficient gratings or prisms. of two chained processes: incoupling from electronic cur-
rent to LTP modes and subsequently outcoupling to the
We are now able to formulate a rate equation for the far-field. The former depends on the density of initial
LTP populations Nq,j in the presence of both thermal and final states within the electron gas, and is enhanced
and electrical pumping at moderately large wavevectors where the polariton is
LTP
 TH  phonon-like but the transferred wavevector is not suf-
Ṅq,j = γq,j Nq,j (Tl ) − Nq,j + ΓLTP
q,j (Nq,j , Te ), (7) ficient to prohibit phase-matching. This differs from a
LTP thermal excitation which only varies according to the
where γq,j , defined in Appendix A 2 is the branch- and
modal frequency and the corresponding Bose-Einstein
momentum-resolved LTP damping and TH
population Nq,j (Tl ) at the temperature of the crystal
+ − lattice.
ΓLTP
q,j (Nq,j , Te ) = Γq,j (Te )(1 + Nq,j ) − Γq,j (Te )Nq,j ,
(8) We are now in a position to estimate the efficiency of
the LPT interconnect, by comparing it with the same
is the term describing LTP emission and reabsorption device operated as a thermal emitter. To simulate ther-
by the electron gas, depending on both the electronic mal emission we consider the system at the equilibrium
 6

Density of LTPs ×10−2 cm Density of Photons ×10−6 cm
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

a) 1,000
Wavenumber (cm−1 ) b)

950

900

Thermal (Td = 600K) Thermal (Td = 600K)

850
c) 1,000 d)
Wavenumber (cm−1 )

950

900

Electrical (Tl = 0K, Te = 1000K) Electrical (Tl = 0K, Te = 1000K)

850
e) 1,000 f)
Wavenumber (cm−1 )

950

900

Electrical (Tl = 300K, Te = 1000K) Electrical (Tl = 300K, Te = 1000K)

850
g) 1,000 h)
Wavenumber (cm−1 )

950

900

Electrical (Tl = 600K, Te = 1000K) Electrical (Tl = 600K, Te = 1000K)

850
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
In-plane Wavevector (×108 m−1 ) In-plane Wavevector (×108 m−1 )

FIG. 4: Comparison of LTP and photonic steady-state populations generated by thermal and electrical processes. In
the first column we plot the LTP density, while in the second column we plot the photonic density, obtained by
multiplying the LTP density by the photon fraction ξq,j from Eq. 6. The first row (a,b) shows data obtained
considering purely thermal emission, obtained setting both the lattice and the electronic temperatures to the same
equilibrium device temperature Tl = Te = 600K. The other three rows show instead data obtained considering
electronic temperature Te = 1000K and lattice temperatures equal to Tl = 0K (c,d), Tl = 300K (e,f), and Tl = 600K
(g,h). Other parameters as in Fig. 3. In all panels the blue-solid line indicates the light-cone edge in the cladding,
and the dot-dashed green lines indicate the real LTP frequencies. Note that the color in the figure has been
saturated to allow the thermal and the much larger electrical populations to be plotted on the same color scale.

device temperature Tl = Te = 600K, a temperature typ-
ical for SiC-based thermal emitters [28]. For the elec-
trically excited emission we consider instead a fixed elec-
tronic temperature Te = 1000K, a temperature chosen to
have electrons propagating near their saturation velocity
(Appendix C), and three different lattice temperatures:
Tl = 0, 300, and 600K. The results corresponding to the
first value represent the intensity of purely electrically
excited emission, allowing us to estimate its intrinsic ef-
ficiency when compared to the purely thermal process.
The results corresponding to the second and third val-
ues provide an estimate for the populations expected in
actual devices operating at room temperature and high
temperature respectively.
In the left column of Fig. 4 we plot the steady-state
LTP populations per unit frequency from Eq. 11 in the
case of purely thermal (a), purely electrical (Tl = 0K)
(c), room temperature (Tl = 300K) (g), and high
temperature (Tl = 600K) (h) operation.
In the thermal case all regions of the dispersion are
similarly populated: because the Reststrahlen region
TH
is relatively narrow the thermal occupancy Nq,j (Tl )
varies slowly over the plotted region. For the purely
electrical incoupling (c) LTPs are instead far more
likely to be generated in the spectral region near to the
zone-center LO phonon frequency ωL . Electrical, finite
temperature results (g,h) are intermediate between the
previous two: while the dominant emission remains
the electrical one near to the LO frequency, additional
thermal radiation leads to an increased emission into the
lower polariton branch. In the right column of Fig. 4 we
plot instead the steady state photon population for the
same parameters, that is we multiply the results in the
left column by the corresponding photon fraction from
Eq. 6. We can see how large photonic components are
generated in the smaller wavevector regions, specifically
in high-dispersion regions below the polariton stopbands.
Note that the photonic populations in this figure go
to zero near to the light-line, in the region where the
underlying LTP group velocity is near-zero.
Regardless of the extraction mechanism used to
emit LTPs to the far-field, both electrical and ther-
mal emission in the same mode would be extracted
with the same efficiency. Moreover, the wavevector
region q ≈ 107 , well resolved in Fig. 4, only requires
standard micrometer-sized gratings for outcoupling.
This figure thus provides the main result of this paper,
demonstrating that the LTP interconnect we propose
can emit more than standard narrow-band SPhP-based
thermal emitters [28], while greatly reducing the energy
dissipated outside the narrow region of interest.
In order to proceed with the quantitative comparison
between electrical and thermal emission, we consider the
7
figure of merit
ss
(1) Nq,j (Te , Tl )
ηq,j (Te , Tl , Td ) = ss , (12)
Nq,j (Td , Td )
which is the in-plane momentum q and branch j resolved
ratio between electrically excited emission with electronic
temperature Te and lattice temperature Tl , and purely
thermal emission at the equilibrum device temperature
(1)
Td . In Fig. 5 we plot the branch resolved values of ηq,j
as a function of electronic temperature Te , with the in-
plane wavevector fixed to q = 107 m−1 , for a 2nm active
region, equilibrium device temperature Td = 600K and
Tl = 0K (a), Tl = 300K (b), and Tl = 600K (c). As the
(1)
electronic temperature increases, the magnitude of ηq,j
increases and the electrically excited emission becomes
substantially larger than the thermal one.
As already seen in Fig. 4 the majority of electrical gen-
eration happens into the upper-polariton branch (purple
triangles) and middle-polariton branch (red diamonds),
which are spectrally close to the zone-center LO phonon
frequency ωL , thus causing little dissipation outside a
specific and narrow frequency region and improving the
energy efficiency of the process when compared to the
broadband thermal dissipation. At finite values of Tl
(Fig. 5b,c) the figure of merit for the middle- and upper-
polaritons is slightly enhanced because of the additional
thermal emission. In the lower polariton branch (green
squares) the electrically excited emission is instead much
smaller due to the strong spectral separation between ωL
and the polariton frequency [48], and at finite Tl it is
largely dominated by the thermal component.
The second figure of merit we explore is the internal
quantum efficiency, that is the number of photons emit-
ted by a single electron propagating through a single layer
given by
X ΓLTP ss
q,j (Nq,j (Te , Tl ), Te )ξq,j
η (2) (Te ) = , (13)
q,j
ν (Te , Tl ) ne A
in which ν (Te , Tl ) is the drift velocity, related to both the
temperature of the lattice Tl and of the electron gas Te
as described in Appendix. C, ne is the electronic density,
and A is the device in-plane area, which cancels from
(2)
the final result. In Fig. 6a we plot ηq,j as a function of
electronic temperature for a 2nm active layer operating
at Tl = 0K (green squares), Tl = 300K (red diamonds),
and Tl = 600K (purple triangles). When the electronic
temperature grows, more LTPs are generated in the sys-
tem, even if the growth is sublinear due to the increase
of electronic group velocity ν(Te , Tl ) with the electronic
temperature, as electrons spend less time in the active
region. At finite Tl the η (2) is suppressed near the equi-
librum value Te = Tl , where the sum in Eq. 13 vanishes.
The asymptotic behaviour Te  Tl is unaffected by the
finite temperature of the lattice. In Fig. 6b we plot the
same quantity as a function of the active layer thickness
at fixed electronic temperature Te = 1000K for each lat-
tice temperature, showing a saturation of the efficiency
 8

a) a) 2
Lower Polariton

Tl = 0K

Figure of Merit η (2) ×10−6

6 Middle Polariton
Upper Polariton 1.5
Figure of Merit η (1)

4 Td 1

0.5 Tl = 0K
2
Tl = 300K
Tl = 600K
0
0 500 1,000 1,500 2,000
0
0 500 1,000 1,500 2,000 Electronic Temperature (K)
b) b)
Lower Polariton

Tl = 300K 2

Figure of Merit η (2) ×10−6

6 Middle Polariton
Upper Polariton
Figure of Merit η (1)

1.5

4 Td
1

0.5 Tl = 0K
2
Tl = 300K
Tl = 600K
0
1 2 3 4 5
0 Thickness (nm)
0 500 1,000 1,500 2,000
c)
FIG. 6: Figure of merit η (2) . In panel (a) we plot it as a
Lower Polariton Tl = 600K function of the electronic temperature Te , for an active
6 Middle Polariton region of thickness 2nm. In panel (b) we plot the same
Upper Polariton quantity as a function of the thickness for a constant
Figure of Merit η (1)

electronic temperature Te = 1000K. Results are plotted

for lattice temperature Tl = 0K (green squares),
4 Td Tl = 300K (red diamonds), and Tl = 600K (purple
triangles). Dashed lines in panel (a) mark the three
lattice temperatures, at which the respective electrical
LTP generation vanishes.
2

for thicker layers in which nonlocal effects decrease as

0 shown in Eq. 5. The maximal quantum efficiency is of the
0 500 1,000 1,500 2,000 order of 10−6 per nanometric layer. Superlattice struc-
tures can be used to increase the total emission efficiency,
(1)
FIG. 5: Figure of merit ηq,j from Eq. 12, as a function while being much more robust to imperfections and eas-
of the electronic temperature Te , evaluated at fixed ier to fabricate that those required for quantum cascade
in-plane wavevector q = 107 m−1 , for active region of devices.
thickness 2nm, device temperature Td = 600K, and
lattice temperatures Tl = 0K (a), Tl = 300K (b), and
Tl = 600K (c). For each temperature the figure of merit
III. CONCLUSIONS AND PERSPECTIVES
is plotted for each polariton branch, with reference to
the dispersion in Fig. 3 these correspond to the lower
(green squares), middle (red diamonds) and upper In this work we have investigated a novel method
(purple triangles) polariton branches. of mid-infrared light-generation, exploiting the intrin-
 9

sic optical nonlocality of polar nano-devices. Utilis- [48]

ing longitudinal-transverse polaritons (LTPs) as inter-  h i
 αc  q
− i|q| αc (z+d) iq·r
z < −d,
connects between the microscopic degrees of freedom  αhc |q| iαc e


, e
of a polar lattice and the far-field, we have demon- ~ u,q q i|q| −α(z+d) iq·r
E = − |q| , α e e −d < z < 0,
strated a sizeably improved narrow-band emission effi- 
 h i
ciency is achievable when comparing to equivalent ther- 
− α c  q , i|q| −αc z−αd iq·r
αc |q| αc e e z > 0,
mal mid-infrared emitters. We illustrated this using a  h i
simple, high-symmetry system in order to provide an un- 
 αc  q
, − i|q| eαc (z+d)−αd eiq·r z < −d,

hαc |q| iαc
biased assessment of the fundamental process of electrical ~ l,q q i|q|
LTP generation. An optimised device could utilize more E = |q| , − α eαz eiq·r −d < z < 0,

 h i
complex geometries, as a channel waveguide or nano-  αc  q i|q| −αc z iq·r
 − αc |q| , αc e e z > 0,
resonator to spatially restrict LTP excitation, and to ma-
(A2)
nipulate generated LTPs, but our aim in this paper is to
demonstrate the fundamental efficiency of a novel emis- respectively. Denoting by ~z = [0, 1] the unit vector along
sion channel and we leave device optimization for future z, the magnetic field in each case can be found utilizing
works. We hope our results will open the way toward an the Maxwell-Faraday equation as
experimental observation of LTP-driven electrically ex- 
cited emission and from there to a novel generation of    α (z+d)
e c z < −d,
mid-infrared optoelectronic devices. ~ u,q = iω q iq·r −α(z+d)
H ~z × , 0 e e −d < z < 0,
c2 α |q| 
e−αc z−αd z > 0,

   α (z+d)−αd
FUNDING e c z < −d,
~ iω q iq·r αz
Hl,q = 2 ~z × ,0 e e −d < z < 0,
c α |q| 
e−αc z
S.D.L. is supported by a Royal Society Research fel- z > 0,
lowship and the Philip Leverhulme prize. The authors (A3)
acknowledge support from the Royal Society Grant No.
RGF\EA\181001 and the Leverhulme Trust Grant No. which have been constructed for continuity at the film
RPG-2019-174. interfaces z = 0, −d. The modes of the waveguide are
linear superpositions of those on each of it’s interfaces
with electric field
Appendix A: Local and Nonlocal Spectra ~ q = UE
~ u,q + LE
~ l,q ,
E (A4)

1. Local Response of Thin Polar Films where U, L are coefficients describing the contribution of
the SPhPs at either interface. After applying boundary
In this work we consider a trilayer waveguide, consist- conditions on the tangential component of the electric
~ q · r we recover the following dispersion relation
field E
ing of a polar film of thickness d sandwiched between
semi-infinite positive dielectric cladding layers, occupy-  αc     
( αc − 1)e−αd 1+ α c
U 0
ing the region −d < z < 0. The system is translationally αc 
αc
= .
invariant in the xy plane. Modes of the system are char- αc + 1 (αc
αc − 1)e
−αd L 0
acterized by a 2D in-plane wavevector q, and out-of-plane (A5)
wavevectors given by This equation is satisfied when the determinant of the
matrix is zero, which leads to the transcendental disper-
p
α = |q|2 − ω 2 /c2 , sion relation
p
αc = |q|2 − c ω 2 /c2 , (A1) 1 − r2 e−2αd = 0, (A6)

respectively in the film and the cladding, where  (c ) is in which r is the Fresnel reflection coefficient
the dielectric function in the film (cladding) and c is the αc − αc 
r= . (A7)
speed of light in vacuum. In all the Paper in-plane 2D αc + αc 
vectors are bold, and three dimensional vectors are indi-
cated either listing the in- and out-of-plane components After manipulation Eq. A6 can be recast in the form [48]
in squared parenthesis or, when notationally clear, with  
αc c α
an arrow symbol. The full 3D wavevector of the mode in 2 − tanh (αd) + = 0. (A8)
c α αc
the layers is thus written as [q, α], and similarly we write
the full 3D coordinate [r, z]. Although in the thin film limit Eq. A8 can simplify dra-
Each interface between the polar film and cladding matically [58] in this work we cannot utilise this ap-
supports a surface phonon polariton with electric field proximation as it breaks down at the larger wavevec-
components at the upper and lower interfaces given by tors under study. The SPhP frequency ωqSP utilised in
 10

LTP
the main body is found by numerical solution of Eq. A8. polaritons (LTPs) [48] with frequencies ωq,j
This yields a complex modal frequency as a function of X
LTP ˆ† ˆ
wavevector q, which leads to the loss rate ĤLTP =~ ωq,j dq,j dq,j , (A14)
 q,j
γ
γqM = = ωqSP = , (A9)
2 described by creation operators
X
in the wavevector region under study [58], a consequence dˆq,j = αq,j âq + βq,j,n b̂q,n , (A15)
of strong confinement in the polar layer. Note that we n
neglect the frequency and polarization dependence of the
phononic loss rate. in which j indexes the polariton branch, and the Hopfield
coefficients αq,j , βq,j,n preserve the unitary nature of the
transformation
2. Nonlocal Response of Thin Polar Films X
|αq,j |2 + |βq,j,n |2 = 1. (A16)
n
The nonlocal optics of the system can be derived semi-
analytically considering the dispersion of LO phonons This ensures that the resulting LTP modes are also
in the polar film [48]. In the typical quadratic model bosonic, whose creation and annihilation operators sat-
of phonon dispersion [45] the dispersive LO phonon fre- isfy the commutation relation
quency takes the form h i
  dˆq,j , dˆ†q0 ,j 0 = δq,q0 δj,j 0 , (A17)
ωq~L 2 = ωL2 − βL2 |q|2 + qz2 , (A10)
and j indexes the different LTP branches.
where ωL is the frequency at zero-wavevector, ~q = [q, qz ] The homogeneous lineshape of the j th LTP branch can
is the 3D wavevector and βL is a characteristic veloc- be written as a Lorentzian
ity. In a planar layer enforcing Pekar boundary condi-
LTP
tions, as when the Reststrahlen bands in the layer and 1 γq,j
ρLTP
q,j (ω) = , (A18)
the cladding do not overlap, this leads to discrete, lo- LTP )2 + (γ LTP )2
π (ω − ωq,j q,j
calised LO phonon modes with frequencies
q where the homogeneous LTP linewidth can be written
L
ωq,n = ωL2 − βL2 (ξn2 + |q|2 ), (A11) from Eq. A15 as a weighed sum of the different involved
scattering channels as
where ξn is the quantized out-of-plane wavevector defined LTP


γq,j = γ + |αq,j |2 γqM − γ . (A19)
in Eq. 4 in the main text.
The SPhP mode considered in Appendix A 1 and these
localised LO phonons interact through shared electro- 3. Quantization of the LO Mode
mechanical boundary conditions [45]. In the trilayer un-
der study we previously derived the analytical form of
In this Paper we aim to describe interactions between
this coupling Ωn,q , given by Eq. 5 [48]. Using this result
electrons and localised LO phonon modes through the
the interacting system can be described in the rotating
Fröhlich interaction. To do this we require the Hamilto-
wave approximation by the quadratic Hamiltonian
nian modeling the interaction of the two systems. Given
X X an electronic charge density operator ρ̂ we can write
ĤLTP =~ ωqSP â†q âq + L
ωq,n b̂†q,n b̂q,n Z h i
q n
ĤInt = − d2 r dz ρ̂ (r, z) φ̂ (r, z) + ĵ (r, z) · Â (r, z) ,
 
+ Ωq,n âq b̂†q,n + â†q b̂q,n , (A12) (A20)
where φ̂ is the electric potential arising from LO phonon
in which â†q (b̂†q,n ) are the annihilation operators for the modes, ĵ is the electric current density operator and Â
is the electromagnetic vector potential. The first term
SPhP (nth LO phonon) mode, obeying bosonic commu- of this Hamiltonian describes the interaction between an
tation rules electronic charge and the induced LTP charge in the lat-
h i tice, while the second describes interaction between elec-
âq , â†q0 = δq,q0 ,
tronic currents and the electromagnetic tranverse fields.
h i
The former one is dominant for low-frequency currents
b̂q,n , b̂†q0 ,n0 = δq,q0 δn,n0 . (A13)
and in this Paper we will consider only this component,
using the Fröhlich Hamiltonian
The quadratic bosonic Hamiltonian in Eq. A12 can be Z
diagonalized using a multimode Hopfield-Bogoliubov di-
agonalization into a spectrum of longitudinal-transverse ĤFrö = − d2 r dz ρ̂ (r, z) φ̂ (r, z) . (A21)
 11

In a local model the discontinuity in the out-of-plane elec- Appendix B: Electron Gas
tric field at the boundaries of the polar layer leads to a
finite induced charge which, due to the transverse nature Electrons interact with LO phonons through the
of the electric field, is unable to spread into the interior Fröhlich Hamiltonian in Eq. A21. Introducing the
of the polar layer. In our nonlocal model instead, charge fermionic annihilation operators ĉ~k with 3D momentum
induced at the edge of the polar layer is able to spread ~k = [k, k], describing electrons in the conduction band,
into the interior, manifesting as a source of LO phonons.
the free electron Hamiltonian is given by
This charging is entirely encoded in the LO phonon fields
and their corresponding scalar potential. X
The electric potential at fixed in-plane wavevector q in Ĥ = ε~k ĉ~† ĉ~k , (B1)
k
branch n with out-of-plane wavevector ξn given by Eq. 4 ~
k
is proportional to
where the free-electron energy is ε~k = ~2 |~k|2 /2m∗ and m∗
iq·r
uq,n (r, z) = sin [ξn (z + d/2)] e . (A22) is the effective electron mass in the cladding region. We
consider a Si cladding region and we will thus consider
To parameterize the Hamiltonian Eq. A21 it is necessary the conductivity effective mass m∗ = 0.26me , where me
to quantize the phonon field. This can be carried out is the free-electron mass. The electronic creation and
from the corresponding electric field annihilation operators obey Fermionic anti-commutator
~ q,n (r, z) = −Bq,n ∇uq,n (r, z) , rules
E (A23)
{ĉ~k , ĉ~† 0 } = δ~k,k~0 , (B2)
where Bq,n is a constant to be determined by the quan- k
tization procedure. Utilizing the results of Refs. [49, 63],
where we developed a quantum theory on local and non- and have free plane-wave wavefunctions of form
local polar media, 1 ik·r ikz
Z  2  ψ~k (r, z) = √ e e , (B3)
L2 2 LA
20 ∞ ωq,n ωL − ωT
3
d r ~ q,n |2
 |E
L
~ωq,n L 2 − ω 2 + β 2 (|q|2 + ξ 2 ) 2
ωq,n T L n where L is the electronic quantization length along z and
L

A the sample surface. By inverting Eq. A15 to write
= sgn ωq,n ,
(A24) the LO phonon operators in terms of the LTPs we can
rewrite Eq. A21 as
L
where ωq,n is the mode frequency, given in this case be
Eq. 3. For the Fabry-Pérot modes under study, consid- (2π)2 X
ĤFrö = ~κq,k,k0 ,j δ (k0 − k − q)
ering that ξn = nπ/d we can find A
~
k,q,~
k0 ,j
Z h i
 
~ q,n |2 = |Bq,n |2 Ad ω 2 − ω 2 ,
d3 r|E (A25) × ĉ~† ĉ~k0 dˆ†q,j + ĉ~† 0 ĉ~k dˆq,j ,
k k
2 L q,n
2βL (B4)
with A the device in-plane area. After substituting
Eq. A25 back into Eq. A24 we recover the quantization in which ~k = [k, k] is the incoming electronic wavevector
constant and ~k 0 = [k0 , k 0 ] is the outgoing one. Integrating over
s s the in-plane component of the out-going wavevector k0
L
~ωq,n βL2 we put the Fröhlich Hamiltonian in the form
Bq,n = ρ 2 L2
, (A26)
0 q,n Ad ωL − ωq,n X h i
ĤFrö = ~κq,k,k0 ,j ĉ~† ĉ~k+~q dˆ†q,j + ĉ~† ĉ~k dˆq,j ,
L k k+~
q
in which, compared to the local case ωq,n = ωL , we ac- ~
k,k0 ,q,j
quired a dispersive Fröhlich constant (B5)
L2
ωq,n clearly describing the dressed Fröhlich process in Fig. 2c,
ρq,n = ∞ 2 . (A27)
ωL2 − ωT where the electron-LTP coupling coefficient has form
Finally, we can write the second-quantized form of the eX
LO phonon potential κq,k,k0 ,j = Bq,n βq,j,n Ξk,k0 ,n , (B6)
~ n
X h i
φ̂ (r, z) = Bq,n uq,n (r, z) b̂q,n + b̂†−q,n , (A28) with the electron-phonon overlap integral given by
q,n
Z 0
1 0
which parameterizes the LO phonon contribution to Ξk,k0 ,n = dz ei(k−k )z sin [ξn (z + d/2)] . (B7)
Eq. A21. L −d
 12

Note that in systems containing free charges the di- the drift velocity, and the temperature of the electron
electric function differs from Eq. 1, and is instead given gas.
by We are able to write the balance equations between elec-
tronic energy ε and drift velocity ν as [65]
ωL2 − ω (ω + iγ) X nj e2 1
 (ω) = ∞ 2 − ∗ , ∂ε ~ωL
ωT − ω (ω + iγ) 0 mj ω (ω + iΓ) = −eEν − , (C1a)
j=e,h
∂t τε (Te , Tl )
(B8)
where Γ is a characteristic free carrier scattering rate, ∂ν eE ν
=− ∗ − , (C1b)
nj is the carrier density, m∗j is the carrier conductivity ∂t m τm (Te , Tl )
effective mass, and j refers to electrons e, or holes h. In
SiC the relatively high frequencies of the Reststrahlen where E is the scalar electric field, τε (Te , Tl ) (τm (Te , Tl ))
band mean that even at large charge densities the shift in is the energy (momentum) relaxation time dependent on
dielectric function is minimal [64]. Deviations are most both the electron and lattice temperatures. In the steady
severe near to ωL where the Lorentz dielectric function state these equations can be directly solved
vanishes. Note that the predominant effect is a small ~ωL
blue shift in SPhP frequencies. This would actually be eEν = − , (C2)
beneficial for the effect we aim to describe, as the SPhP τε (Te , Tl )
mode would move toward ωL where electron-phonon eE ν
∗
=− , (C3)
coupling is strongest. m τm (Te , Tl )
allowing the electric field or drift velocity to be eliminated

τm (Te , Tl )
Appendix C: Thermalized Electron Model m∗ ν 2 = ~ωL , (C4a)
τε (Te , Tl )
Electrons interacting with the polar nanolayer consid- e2 E 2 ~ωL
= , (C4b)
ered in this Paper are assumed to propagate under the m∗ τm (Te , Tl ) τε (Te , Tl )
influence of an applied electric field polarized perpendic-
ular to the nanolayer. An average electron propagates Although it is not strictly possible to define a relaxation
at the drift velocity, which is limited through momentum time for the polar interaction, a consequence of the in-
relaxation with the crystal lattice. The lattice also pro- elastic and non velocity-randomising nature of the inter-
vides energy relaxation channels, which reduce the mean action [66], it is possible to approximate an energy and
electronic energy and prevent the electron gas from over- momentum relaxation time utilizing Fermi’s golden rule.
heating. We describe the electron gas as thermalised at Following Conwell [67], the probability per unit time of a
a certain electronic temperature Te . In the following we carrier scattering out of state with momentum ~k = [k, k]
calculate closed relationships between the applied field, with exchanged momentum ~q = [q, q] is given by

    
1 2π X ~ ~ 2 ~ ~ 2
= |hk + ~q, NL − 1|ĤFrö |k, NL i| δ ε~k+~q − ε~k − ~ωL + |hk − ~q, NL + 1|ĤFrö |k, NL i| δ ε~k−~q − ε~k + ~ωL ,
τ~k (Tl ) ~
q
~
(C5)

in which the first term is the transition probability from where EF is the bulk Fröhlich matrix element
phonon absorption, the second is from phonon emission.
Here NL (Tl ) is the equilibrium phonon population in  
m∗ e2 ~ωL 1 1
the LO phonon modes which are assumed to be non- eEF = − , (C7)
4π~2 0 ∞ st
dispersive. If we assume spherical bands we can compute
the inverse scattering time directly as
in which ∞ (st ) is the high-frequency (static) dielectric
constant of the lattice and 0 is the permittivity of free-
 r 
1 2eEF −1 ε~k space.
=p ∗ NL (Tl ) sinh The inverse scattering time τ~−1 (Tl ) is not a relaxation
τ~k (Tl ) 2m ε~k ~ωL k
r  time, and as discussed above one cannot be strictly de-
−1 ε~k fined for the polar interaction. We can however define the
+ (NL (Tl ) + 1) sinh − 1 , (C6)
~ωL rate of energy change for an electron with momentum ~k
 13

due to the polar interaction, given by Boltzmann distribution at electronic temperature Te

8π 3 e−ε~k /kB Te
f~k (Te ) = , (C9)
dε~k ~ωL LA Nc (Te )
= , (C8)
dt PI τ~k (Tl ) h ∗ i3/2
where Nc (Te ) = 2η m2π~ k B Te
2 is the density of states
in the conduction band, ne is the electronic density and
Pseudo-relaxation times can then be calculated by taking η is the degeneracy factor in the conduction band. Inte-
the distributional average of this rate of energy change. grating over the distribution we find the energy pseudo-
Assuming the electron gas is described by a Maxwell- relaxation time

  r  
1 1 dε~k 2 eg(Tl )−g(Te )/2 − eg(Te )/2 p g (Te )
≈ =− eEF g (Te )K0 , (C10)
τε (Te , Tl ) ~ωL dt PI πm∗ ~ωL eg(Tl ) − 1 2
where the angular brackets are an average calculated over the distribution in Eq. C9, the Ki are modified Bessel
functions, and we defined the dimensionless quantity g (T ) = ~ωL /kB T .
To find the momentum pseudo-relaxation time we start instead by calculating the change of electronic wavevector in
the transport direction ~z for a carrier with momentum ~k = [k, k], given by
    
dk 2π X
= k|h~k + ~q, NL − 1|HFrö |~k, NL i|2 δ ε~k+~q − ε~k − ~ωL − k|h~k − ~q, NL + 1|HFrö |~k, NL i|2 δ ε~k−~q − ε~k + ~ωL .
dt ~
q
~
(C11)
As for the previous case we can integrate analytically, yielding the result
 s r  s s !
dk eEF k ~ωL ~ωL −1 ε~k ~ωL ~ωL ~ωL
=− NL 1+ − sinh +(NL +1) 1− + sinh−1 1− .
dt ~ |~k| ε~k ε~k ~ωL ε~k ε~k ε~k
(C12)
Analogously to the energy relaxation rate we can find the pseudo-relaxation time by integrating over the electronic
distribution. To do this we need to assume the distribution function is drifted, as the gas propagates at drift velocity ν
along ~z. The zero-order spherically symmetric term does not contribute, the first-order perturbation of the distribution
is
(1) ~ν 8π 3 e−ε~k /kB Te
f~ (Te ) = k , (C13)
k kB Te LA Nc (Te )
and integrating over the distribution yields the momentum pseudo-relaxation time
  3/2
1 1 d (~k) 2eEF NL (Tl ) g (Te )
≈ ∗ =− √ √ ∗ eg(Te )/2
τm (Te , Tl ) m ν dt PI 3 π 2m ~ωL
       
g(Tl )−g(Te ) g (Te ) g(Tl )−g(Te ) g (Te )
× e + 1 K1 + e − 1 K0 . (C14)
2 2
We thus arrive at our results for the drift velocity and the field intensity
 g(T )−g(T )   
g(Te )
~ωL τm (Te , Tl ) 3~ωL e l e
− 1 K 0 2
ν2 = ∗ =√      ,
m τε (Te , Tl ) ∗
2m g (Te )
g(Te )
g(Te )
e g(Tl )−g(Te ) + 1 K1 + e g(T l )−g(T e ) − 1 K0
2 2

(C15a)
 2 ∗ 2 2 h i  
E (Te , Tl ) m ~ωL 2NL (Tl ) g (Te ) g(Te ) g(Tl )−g(Te ) g (Te )
= 2 2 = e e − 1 K0
EF e EF τm (Te , Tl ) τε (Te , Tl ) 3π 2
       
g (Te ) g (Te )
× eg(Tl )−g(Te ) + 1 K1 + eg(Tl )−g(Te ) − 1 K0 . (C15b)
2 2

Appendix D: Calculation of Electrical Injection by the distribution in Eq. C9. As described in Ap-

In this Appendix we calculate the LTP generation rate

by a thermalized, non-degenerate electron gas, described
 14

pendix C this electronic temperature can be related to LTP of in-plane wavevector q in branch j, while exchang-
the electronic drift velocity. ing out-of-plane momentum k 0 is given by
The rate at which electrons with wavevector ~k emit an

Γ∓ ~ = 2π~ne AL|κq,k,k0 ,j |2 δ(εk,k − εk∓q,k0 ∓ ~ωq,j

LTP
)f~k (Te ), (D1)
q,k,k0 ,j

where the ∓ refers to emission (+) and absorption (−) of a LTP respectively. Integrating over the incident electronic
wavevector ~k we derive the total rate at which all electrons emit LTPs in branch j with in-plane wavevector q
X Z Z Z Z Z Z
∓ ∓ AL2 0 2 ∓ 2 ne 0
Γq,j = Γ ~ 0 = dk dk d k Γ ~ 0 = ~AL dk dk d2 k|κq,k,k0 ,j |2
q,k,k ,j (2π)4 q,k,k ,j Nc
k0 ,~
k
ε~
−k k
LTP
×e B Te δ(εk,k − εk∓q,k0 ∓ ~ωq,j ). (D2)

We account for linewidth broadening of the LTP modes by using, instead of the Dirac’s delta, the Lorentzian density
of states from Eq. A18.

[1] Hillenbrand, R., Taubner, T. & Keilmann, F. Phonon-
enhanced light–matter interaction at the nanometre
scale. Nature 418, 159–162 (2002).
[2] Caldwell, J. D. et al. Low-loss, infrared and tera-
hertz nanophotonics using surface phonon polaritons.
Nanophotonics 4, 44–68 (2015).
[3] Sumikura, H. et al. Highly confined and switchable mid-
infrared surface phonon polariton resonances of planar
circular cavities with a phase change material. Nano Let-
ters 19, 2549–2554 (2019).
[4] Spann, B. T. et al. Photoinduced tunability of the rest-
strahlen band in 4H−SiC. Physical Review B 93, 085205
(2016).
[5] Ellis, C. T. et al. Aspect-ratio driven evolution of high-
order resonant modes and near-field distributions in lo-
calized surface phonon polariton nanostructures. Scien-
tific Reports 6, 32959 (2016).
[6] Gubbin, C. R., Maier, S. A. & De Liberato, S. Theoreti-
cal investigation of phonon polaritons in SiC micropillar
resonators. Physical Review B 95, 035313 (2017).
[7] Dubrovkin, A. M. et al. Resonant nanostructures
for highly confined and ultra-sensitive surface phonon-
polaritons. Nature Communications 11, 1863 (2020).
[8] Wang, T., Li, P., Hauer, B., Chigrin, D. N. & Taubner,
T. Optical properties of single infrared resonant circular
microcavities for surface phonon polaritons. Nano Letters
13, 5051–5055 (2013).
[9] Berte, R. et al. Sub-nanometer thin oxide film sensing
with localized surface phonon polaritons. ACS Photonics
5, 2807–2815 (2018).
[10] Gubbin, C. R. & De Liberato, S. Theory of nonlinear
polaritonics: χ(2) scattering on a β − SiC surface. ACS
Photonics 4, 1381–1388 (2017).
[11] Razdolski, I. et al. Second harmonic generation from
strongly coupled localized and propagating phonon-
polariton modes. Physical Review B 98, 125425 (2018).
[12] Kitade, S., Yamada, A., Morichika, I., Yabana, K. &
Ashihara, S. Nonlinear shift in phonon-polariton disper-
sion on a SiC surface. ACS Photonics 8, 152–157 (2021).
[13] Taubner, T., Korobkin, D., Urzhumov, Y., Shvets, G.
& Hillenbrand, R. Near-field microscopy through a sic
superlens. Science 313, 1595–1595 (2006).
[14] Kiessling, R. et al. Surface phonon polariton resonance
imaging using long-wave infrared-visible sum-frequency
generation microscopy. ACS Photonics 6, 3017–3023
(2019).
[15] Caldwell, J. D. et al. Sub-diffractional volume-confined
polaritons in the natural hyperbolic material hexagonal
boron nitride. Nature Communications 5, 5221 (2014).
[16] Dai, S. et al. Tunable phonon polaritons in atomically
thin van der waals crystals of boron nitride. Science 343,
1125–1129 (2014).
[17] Li, P. et al. Hyperbolic phonon-polaritons in boron ni-
tride for near-field optical imaging and focusing. Nature
Communications 6, 7507 (2015).
[18] Woessner, A. et al. Highly confined low-loss plasmons in
graphene–boron nitride heterostructures. Nature Mate-
rials 14, 421–425 (2015).
[19] Basov, D. N., Fogler, M. M. & de Abajo, F. J. G. Polari-
tons in van der waals materials. Science 354, aag1992
(2016).
[20] Ma, W. et al. In-plane anisotropic and ultra-low-loss
polaritons in a natural van der waals crystal. Nature
562, 557–562 (2018).
[21] Autore, M. et al. Boron nitride nanoresonators for
phonon-enhanced molecular vibrational spectroscopy at
the strong coupling limit. Light: Science & Applications
7, 17172–17172 (2018).
[22] He, M. et al. Anisotropy and modal hybridization in
infrared nanophotonics using low-symmetry materials.
ACS Photonics 9, 1078–1095 (2022).
[23] Taliercio, T. & Biagioni, P. Semiconductor infrared plas-

monics. Nanophotonics 8, 949–990 (2019).
[24] Gubbin, C. R., De Liberato, S. & Folland, T. G. Surface
phonon polaritons for infrared optoelectronics. Journal
of Applied Physics 131, 030901 (2022).
[25] Greffet, J.-J. et al. Coherent emission of light by thermal
sources. Nature 416, 61–64 (2002).
[26] Schuller, J., Taubner, T. & Brongersma, M. Optical an-
tenna thermal emitters. Nature Photonics 3, 658–661
(2009).
[27] Arnold, C. et al. Coherent thermal infrared emission by
two-dimensional silicon carbide gratings. Phys. Rev. B
86, 035316 (2012).
[28] Lu, G. et al. Engineering the spectral and spatial dis-
persion of thermal emission via polariton–phonon strong
coupling. Nano Letters 21, 1831–1838 (2021).
[29] Faist, J. et al. Quantum cascade laser. Science 264,
553–556 (1994).
[30] Faist, J. et al. Distributed feedback quantum cascade
lasers. Applied Physics Letters 70, 2670–2672 (1997).
[31] Ohtani, K. et al. An electrically pumped phonon-
polariton laser. Science Advances 5 (2019).
[32] Gubbin, C. R. et al. Hybrid longitudinal-transverse
phonon polaritons. Nature Communications 10, 1682
(2019).
[33] Ridley, B. K. Continuum theory of optical phonon hy-
brids and their interaction with electrons in a quantum
well. In Quantum Well and Superlattice Physics IV, vol.
1675, 492–498 (International Society for Optics and Pho-
tonics, 1992).
[34] Ridley, B. K. Electron-hybridon interaction in a quantum
well. Phys. Rev. B 47, 4592–4602 (1993).
[35] Ridley, B. K. Electrons and phonons in semiconductor
multilayers, vol. 5 (Cambridge University Press, 2009).
[36] Gubbin, C. R., Martini, F., Politi, A., Maier, S. A. &
De Liberato, S. Strong and coherent coupling between
localized and propagating phonon polaritons. Physical
review letters 116, 246402 (2016).
[37] Constantinou, N. C. & Ridley, B. K. Guided and interface
lo phonons in cylindrical gaas/alx ga1−x as quantum wires.
Phys. Rev. B 41, 10627–10631 (1990).
[38] Riddoch, F. A. & Ridley, B. K. On the scattering of
electrons by polar optical phonons in quasi-2d quantum
wells. Journal of Physics C: Solid State Physics 16,
6971–6982 (1983).
[39] Ridley, B. K. Hybrid phonons in semiconductor nanos-
tructures, vol. 1 (Oxford University Press, 2017).
[40] Wendler, L. & Grigoryan, V. Acoustic interface waves
in sandwich structures. Surface Science 206, 203–224
(1988).
[41] Ridley, B. K., Schaff, W. J. & Eastman, L. F. Hot-
phonon-induced velocity saturation in gan. Journal of
Applied Physics 96, 1499–1502 (2004).
[42] Khurgin, J. B., Bajaj, S. & Rajan, S. Amplified spon-
taneous emission of phonons as a likely mechanism for
density-dependent velocity saturation in GaN transistors.
Applied Physics Express 9, 094101 (2016).
[43] Komirenko, S. M., Kim, K. W., Kochelap, V. A., Fe-
dorov, I. & Stroscio, M. A. Coherent optical phonon
generation by the electric current in quantum wells. Ap-
plied Physics Letters 77, 4178–4180 (2000).
[44] Komirenko, S. M., Kim, K. W., Kochelap, V. A., Fe-
dorov, I. & Stroscio, M. A. Generation of coherent con-
fined lo phonons under the drift of two-dimensional elec-
trons. Phys. Rev. B 63, 165308 (2001).
15
[45] Gubbin, C. R. & De Liberato, S. Optical Nonlocality in
Polar Dielectrics. Phys. Rev. X 10, 021027 (2020).
[46] Gubbin, C. R. & De Liberato, S. Nonlocal scattering
matrix description of anisotropic polar heterostructures.
Phys. Rev. B 102, 235301 (2020).
[47] Gubbin, C. R. & De Liberato, S. Impact of phonon nonlo-
cality on nanogap and nanolayer polar resonators. Phys.
Rev. B 102, 201302 (2020).
[48] Gubbin, C. R. & De Liberato, S. Quantum theory of
longitudinal-transverse polaritons in nonlocal thin films.
Phys. Rev. Applied 17, 014037 (2022).
[49] Gubbin, C. R. & De Liberato, S. Polaritonic quantization
in nonlocal polar materials. The Journal of Chemical
Physics 156, 024111 (2022).
[50] Mortensen, N. A., Raza, S., Wubs, M., Søndergaard,
T. & Bozhevolnyi, S. I. A generalized non-local opti-
cal response theory for plasmonic nanostructures. Nature
Communications 5, 3809 (2014).
[51] Garcia de Abajo, F. J. Nonlocal effects in the plasmons
of strongly interacting nanoparticles, dimers, and waveg-
uides. The Journal of Physical Chemistry C 112, 17983–
17987 (2008).
[52] David, C. & Garcı́a de Abajo, F. J. Spatial nonlocality in
the optical response of metal nanoparticles. The Journal
of Physical Chemistry C 115, 19470–19475 (2011).
[53] Raza, S., Toscano, G., Jauho, A.-P., Wubs, M. &
Mortensen, N. A. Unusual resonances in nanoplasmonic
structures due to nonlocal response. Physical Review B
84, 121412 (2011).
[54] Ciraci, C. et al. Probing the ultimate limits of plasmonic
enhancement. Science 337, 1072–1074 (2012).
[55] Moore, W. J., Holm, R. T., Yang, M. J. & Freitas, J. A.
Infrared dielectric constant of cubic sic. Journal of Ap-
plied Physics 78, 7255–7258 (1995).
[56] Maier, S. A. Plasmonics: Fundamentals and Applications
(Springer US, 2007).
[57] Passler, N. C. et al. Strong coupling of epsilon-near-
zero phonon polaritons in polar dielectric heterostruc-
tures. Nano letters 18, 4285–4292 (2018).
[58] Campione, S., Brener, I. & Marquier, F. Theory of
epsilon-near-zero modes in ultrathin films. Phys. Rev.
B 91, 121408 (2015).
[59] Khurgin, J. B., Clerici, M. & Kinsey, N. Fast and slow
nonlinearities in epsilon-near-zero materials. Laser &
Photonics Reviews 15, 2000291 (2021).
[60] Karch, K., Pavone, P., Windl, W., Schütt, O. & Strauch,
D. Ab initio calculation of structural and lattice-
dynamical properties of silicon carbide. Phys. Rev. B
50, 17054–17063 (1994).
[61] Pekar, S. I. Theory of electromagnetic waves in a crystal
with excitons. Journal of Physics and Chemistry of Solids
5, 11–22 (1958).
[62] Hopfield, J. J. Theory of the contribution of excitons to
the complex dielectric constant of crystals. Phys. Rev.
112, 1555–1567 (1958).
[63] Gubbin, C. R., Maier, S. A. & De Liberato, S. Real-space
hopfield diagonalization of inhomogeneous dispersive me-
dia. Phys. Rev. B 94, 205301 (2016).
[64] Dunkelberger, A. D. et al. Active tuning of surface
phonon polariton resonances via carrier photoinjection.
Nature Photonics 12, 50–56 (2018).
[65] Khurgin, J., Ding, Y. J. & Jena, D. Hot phonon effect
on electron velocity saturation in GaN: A second look.
Applied Physics Letters 91, 252104 (2007). Publisher:
 16

American Institute of Physics. 2010).

[66] Jacoboni, C. Theory of Electron Transport in Semi- [67] Conwell, E. High field transport in semiconductors, vol. 9
conductors: A Pathway from Elementary Physics to of Solid state physics : advances in research and applica-
Nonequilibrium Green Functions (Springer, Germany, tions (Academic Press, 1969), 1st edn.