WorksPhysics and Astronomy
Extrinsic Bulk Photovoltaic Effect from Asymmetric Defect Scattering
Abstract
The extrinsic bulk photovoltaic effect from asymmetric electron-defect scattering can make an important contribution to the total photocurrent and is highly diagnostic of defect type. We present an ab initio Boltzmann framework combining optical excitation with electron-phonon and symmetric and skew electron-defect scattering, and we apply it to monolayer GeS. Because the defects that generate the current also relax the anisotropy that sustains it, the response peaks at a finite defect fraction. Within the scattering-limited model, cooling to ~K shifts the peak position toward lower defect fractions and yields a maximum enhancement of more than six orders of magnitude at fixed defect fraction. At selected low-temperature defect fractions, the total extrinsic current can exceed the intrinsic shift current by more than an order of magnitude, so that the defect fraction itself becomes a tunable parameter.
Introduction
In noncentrosymmetric, single-phase materials, incident light can generate a direct photocurrent in the bulk without a junction [1, 2, 3, 4, 5]. This bulk photovoltaic effect (BPVE) is attributed to several primary mechanisms [6, 7, 8, 9]. First-principles calculations have largely focused on intrinsic mechanisms associated with photoexcitation [10, 11, 12], including the shift current and injection current [13, 14, 15], as well as the asymmetric photoexcitation processes including the ballistic current [16, 17]. More recently, intrinsic contributions from carrier relaxation and recombination [18, 19] have also been modeled from first principles [20, 21]. However, extrinsic mechanisms such as asymmetric defect scattering, while identified in early kinetic theories of the BPVE [3] and indicated by experiments to strongly reshape photoexcited carrier distributions [22], have not, to our knowledge, been evaluated quantitatively from first principles in real materials. Consequently, the magnitude and controllability of extrinsic photocurrents remain largely unknown.
In this work, we develop a first-principles Boltzmann transport equation (BTE) framework to compute the leading extrinsic BPVE generated by electron-defect skew scattering. By combining ab initio optical excitation, electron-phonon scattering, and both symmetric and asymmetric electron-defect scattering within an iterative BTE solver, we evaluate the steady photocurrent beyond the relaxation-time approximation. Furthermore, we derive a minimal analytical model that reveals the competition between skew conversion and defect-induced relaxation, predicting a crossover from linear to inverse scaling of photocurrent with defect fraction and identifying a finite optimal defect fraction. For monolayer GeS containing either of two representative point defects, we find that this optimal defect fraction is strongly temperature-dependent and that the extrinsic current can significantly exceed the intrinsic shift current in the cold, dilute regime.
Theory
In a crystal lacking inversion symmetry, electron-defect scattering can contain an antisymmetric component. Denoting by the transition rate from the Bloch state to , in general,
Inversion asymmetry permits this contribution, subject to the remaining crystal symmetries [3, 23, 24].
Within the leading skew-scattering approximation [23, 24, 25], we expand the single-defect transition matrix through second order in the defect perturbation potential ,
where
is the electron-defect matrix element, is the Bloch-state energy, and is the retarded electronic Green’s function.
Retaining only the first-order term in Eq. (2) yields the symmetric defect-scattering kernel
We define the asymmetric, or skew, component as
Upon substitution of the -matrix expansion, the principal-value contribution cancels by Hermiticity of , leaving the on-shell contribution
This rate satisfies
Here, “skew scattering” denotes the leading antisymmetric rate, Eq. (6), cubic in the defect potential [23, 24, 25]. For dilute, uncorrelated defects, the single-defect kernels in Eqs. (4) and (6) yield rates linear in the defect fraction [24, 26]. Together with intrinsic electron-phonon scattering, these processes define the linear collision operator [23, 27]
where , , and denote the inelastic electron-phonon, elastic symmetric defect, and elastic skew defect collision operators, respectively [23, 26, 28].
Under continuous-wave illumination in the low-injection limit, we treat the conduction-band electron and valence-band hole sectors independently. We neglect Pauli blocking and retain the collision term to first order in these dilute carrier occupations. For either carrier sector, the photoinduced distribution obeys [27, 29]
where denotes the corresponding photoexcited electron or hole distribution, is the ab initio optical generation rate, and
is the out-minus-in collision term built from the electron-phonon rates and Eqs. (4) and (6) [27, 30].
Because carrier recombination is omitted, scattering alone cannot remove the photoexcited carriers [27, 30]. Summing Eq. (9) over the states within either carrier sector gives
where
are the photoexcited electron or hole population and its corresponding generation rate. The electron and hole populations grow together under interband excitation, while the total number of physical electrons remains conserved. As shown next, the part of the distribution produced by skew scattering, which carries the current, can reach a steady state even while the total carrier population in both sectors is still growing.
To separate the growing population from the current-carrying component [31], we write
where is the population-conserving zero mode satisfying , is its time-dependent amplitude, and has zero net carrier number. The secular population growth is confined to , while relaxes to a finite steady state .
If is even under time reversal, the odd band velocity ensures that
Under these conditions, the growing population carries no current and determines the steady photocurrent [31, 32].
Two experimentally accessible parameters govern the skew photocurrent: the defect fraction per sulfur site and the temperature . To describe their effects, we reduce the BTE to an angular-harmonic model [31, 33, 34]. On each energy shell, we expand the distribution in the polar angle of :
where is the angular-harmonic index and is the corresponding harmonic amplitude.
Within this minimal isotropic model, linearly polarized excitation generates an isotropic population and an optical alignment. Neither carries current directly, as the photocurrent resides in the odd sector [31]. Moreover,
meaning that the isotropic population cannot generate an component [31]. The skew photocurrent is therefore produced by conversion of the current-free alignment into the current-carrying component.
Let denote the relaxation time of angular harmonic [33, 34], where
with being the phonon-induced relaxation rate and representing the elastic defect-relaxation rate per unit defect fraction [31]. As mentioned above, the skew-scattering rate is likewise linear in . Denoting the component of the optical generation rate by , the optical alignment becomes
Acting on the alignment, skew scattering generates a source for the current-carrying harmonic, where is the skew-conversion coefficient. Balancing this source against relaxation gives the photocurrent density [31]
At fixed illumination intensity, the skew conductivity shares this dependence on and . Therefore, increasing the defect fraction strengthens the skew conversion but simultaneously diminishes the current-carrying and alignment lifetimes. In the phonon-dominated dilute regime, , Eq. (18) reduces to . Conversely, once both relevant harmonics become defect-dominated, , the current obeys , making the response self-limiting. If the crossover fractions and are well separated, an intermediate shoulder or plateau emerges when only one harmonic is defect-limited.
Maximizing the magnitude of Eq. (18) yields a unique optimal defect fraction for the modeled carrier channel,
Thus, is the geometric mean of the current-carrying and alignment crossover fractions. Comparable scales produce a sharp turnover; well-separated scales broaden the maximum.
Cooling reduces the phonon relaxation rates , shifting the crossover fractions and to lower defect fractions. The cooling enhancement therefore depends strongly on defect fraction: it persists while phonons control the relevant lifetimes but is suppressed once defect scattering dominates. Spontaneous phonon emission can still contribute to at , so that electron-phonon scattering need not vanish [28, 29].
The first-principles BTE retains the full angular structure: anisotropic energy contours and momentum-dependent optical and scattering matrix elements in GeS can mix harmonics beyond the and sources, allowing extrema or sign reversals absent in Eq. (18) [31].
Computational details.—We perform density functional theory (DFT) calculations using Quantum ESPRESSO [35] with the Perdew-Burke-Ernzerhof (PBE) functional [36] and norm-conserving pseudopotentials. Monolayer GeS has in-plane lattice constants and and a out-of-plane cell parameter. We use a wave-function cutoff and an Monkhorst-Pack electronic wave-vector mesh [37]. The resulting PBE band structure features an indirect gap of and a minimum direct gap of at the Brillouin-zone center .
We compute phonons and electron-phonon matrix elements using density functional perturbation theory (DFPT) [38] on an phonon wave-vector mesh. These are interpolated using maximally localized Wannier functions [39, 40] constructed from the four highest valence and three lowest conduction bands, which contain all vertical transitions within the photon-energy range considered here.
The electron-defect matrix elements are calculated with Perturbo [29]. The defect perturbation is obtained from the difference between the Kohn-Sham potentials of defect-containing and pristine supercells [26, 41]. We study the sulfur vacancy and isovalent selenium substitution , with substantially different leading-order scattering strengths. For dilute, uncorrelated defects, we obtain the dependence by rescaling the single-defect symmetric and skew rates.
We compute from interband momentum matrix elements using modified Wannier90 [42] and Fermi’s golden rule for continuous-wave light polarized along at –.
We solve Eq. (9) on a electronic wave-vector mesh via fixed-point iteration accelerated with Anderson mixing [43]. This retains momentum-dependent in-scattering in beyond the relaxation-time approximation. We solve electrons and holes independently, separating the particle-number zero mode from the steady-state current-carrying distribution.
We incorporate temperature explicitly through the phonon occupations in the microscopic electron-phonon scattering rates, reporting results at different temperatures. The calculation retains spontaneous phonon-emission processes allowed for the photoexcited carriers. For direct comparison, the intrinsic shift-current conductivity of monolayer GeS is computed under identical conditions [5, 44].
Results.—For symmetry with the polar axis along [44], the mirror symmetry forbids an in-plane current along under -polarized illumination, while a current along is allowed. We therefore consider the response and define
where is the photocurrent per unit width and is the optical electric field amplitude. We denote by and report it in . This component aligns with the largest intrinsic shift-current response considered for monolayer GeS [44].
We apply the framework across defect fractions –. Because the analytical model treats one carrier sector at a time, we first compute the electron [conduction-band (CB)] and hole [valence-band (VB)] contributions, and , independently. Unless noted, results are presented at .
Figure 1 illustrates the defect-fraction dependence of the steady-state skew current response at . For , both carrier channels exhibit the single-turnover structure predicted by Eq. (18): the response initially scales linearly with and is subsequently suppressed as defect scattering determines the relevant lifetimes. We identify peak positions near for electrons and for holes.
Figure 1. Extrinsic skew conductivity of monolayer GeS at and as a function of defect fraction per sulfur site, for V (red circles) and Se (blue squares). Filled and open symbols denote the electron and hole channels, respectively. The dashed line marks the intrinsic shift-current conductivity at the same photon energy. The symmetric-logarithmic ordinate displays the zero crossing of the Se hole response near .
For , the weaker defect scattering shifts the crossover toward larger , resulting in a response that remains largely linear within the computed dilute range. The hole channel exhibits a sign reversal, indicating additional microscopic skew-conversion pathways captured by the full BTE that are absent from the single-harmonic analytical model.
The cooling enhancement depends strongly on defect fraction [Fig. 2]. The enhancement factor
ranges from order unity at the defect-dominated edge to over at low defect fractions. In the phonon-dominated regime, where , the suppression of thermal phonons strongly amplifies the response.
Each factor in Eq. (18) turns from constant to linear at , and the 0 K thresholds lie below the 300 K ones, so the logarithmic slope of the enhancement steepens from 0 to to as the two 0 K thresholds are crossed and relaxes back as the 300 K thresholds follow. The segment slopes in Figure 2 reflect this trend: Se steepens from to (electrons) and to (holes) between and , whereas V, whose thresholds are roughly an order of magnitude lower, already shows (electrons) and (holes) in the first decade and flattens above .
Figure 2. Cooling enhancement factor at as a function of defect fraction . Symbols are as in Figure 1. This enhancement ranges from order unity in the defect-dominated part of the dilute regime to over at low defect fractions. Labels indicate logarithmic slopes between adjacent calculated points.
The measurable skew conductivity is the signed sum
As shown in Figure 3(a), cooling shifts the peak position toward lower defect fractions. For V, the electron-channel peak shifts from at 300 K to at 0 K, consistent with the trend predicted by Eq. (19).
Figure 3. Low-temperature response of monolayer GeS. (a) Skew conductivity at K and eV as a function of defect fraction , for V and Se. Filled and open symbols denote the electron and hole channels, respectively. Dashed lines mark the intrinsic shift-current conductivity at the same photon energy. (b) Total extrinsic response at selected defect fractions held fixed across five photon energies: the K response of V at and of Se at , the K V response at (gray, dotted), and the intrinsic shift-current spectrum (solid gray). The selected fractions lie near the electron-channel peaks in the 3.5 eV defect-fraction scans and are not reoptimized at each photon energy. The dip near 2.2 eV marks the sign change of the shift-current response.
The absolute magnitude of this effect is substantial [Fig. 3(b)]. At 300 K at , the total extrinsic response provides a non-negligible contribution, reaching up to approximately 15% of the intrinsic shift-current magnitude across the calculated spectrum. At 0 K, the extrinsic contribution can substantially exceed the intrinsic shift current. At the selected defect fractions in Figure 3(b), the Se and V responses exceed the shift-current magnitude by factors of 7–30 and 18–77, respectively. These large total responses occur at finite defect fractions near the electron-channel peaks, rather than in the perfectly pristine limit.
At the lowest temperatures and most dilute defect fractions, carrier relaxation becomes slow enough that recombination may provide an additional limiting time scale. In this regime, the present results should be interpreted as the scattering-limited response of the asymmetric-defect mechanism [31]. Adding carrier-recombination terms and occupation-dependent generation and scattering to the BTE would account for carrier loss and saturation, which can limit the enhancement and shift the optimal defect fraction.
**Conclusion.—**We extend first-principles BPVE calculations to photocurrents generated by asymmetric defect scattering, treating optical generation, phonon relaxation, and symmetric and skew defect scattering within a common Boltzmann framework. In monolayer GeS, competition between skew conversion and defect-induced relaxation produces a finite optimal defect fraction that shifts to lower values upon cooling. The low-temperature extrinsic response can exceed the intrinsic shift current. At fixed photon energy and illumination intensity, the minimal model predicts a single-channel photocurrent that rises linearly with defect fraction and decays inversely with it beyond the maximum. Measurements across defect fraction and temperature could test this competition. The contrasting dependence on defect fraction of the two defect types, including the sign reversal of the Se hole contribution at 300 K, also suggests that photocurrent measurements could help distinguish defect types, provided that electron, hole, and intrinsic contributions can be resolved.
Acknowledgments
This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award No. DE-SC0024942. Computational support was provided by the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy, Office of Science User Facility located at Lawrence Berkeley National Laboratory, operated under Contract No. DE-AC02-05CH11231.
References
Section S1 gives sufficient conditions for a finite steady photocurrent in the Boltzmann equation without recombination. This result is used in the main text after Eq. (13). Section S2 derives the angular-harmonic model of Eqs. (14)–(19) and states its assumptions. Notation follows the main text. is the rate from to , , and is the defect fraction per sulfur site. Equations of the main text are cited by their number there. Equations of this document carry the prefix S.
Steady photocurrent without recombination
The recombination-free Boltzmann equation conserves carrier number under collisions, while optical generation continually adds carriers. The total population therefore grows with time. The current can still approach a steady value if this growth is confined to collision-invariant modes that carry zero current. We give sufficient conditions below.
Setting
Consider one dilute carrier sector on a finite, time-reversal-paired grid, with state indices including bands where needed. The master equation is
Here is an arbitrary vector on the discrete carrier-state grid, with components , and is the total out-scattering rate from state . The source is the optical generation rate, and is the photoinduced occupation. Self-transitions are omitted, and the total transition rates must satisfy . Although the skew contribution can have either sign, the truncated Born expansion must remain in a regime where the total rates are nonnegative. This dilute-carrier equation neglects Pauli blocking and assumes an equilibrium phonon bath [S1–S3].
The following three properties of the collision operator give a simple sufficient criterion:
(i) (particle conservation), as follows by summing Eq. (22), where is the vector with every component equal to one.
(ii) The directed transition graph has a single closed communicating class . This is a set of states that can reach one another through scattering, with every allowed transition from the set staying inside it. All other states are transient: under collisions alone, carriers starting outside eventually enter it with probability one.
(iii) Its stationary zero mode is even under time reversal, .
Spectrum of the collision operator
The nonnegative transition rates and particle conservation make a finite continuous-time Markov generator. Under property (ii), we apply the Perron–Frobenius theorem within the closed class . It gives a unique zero mode within that class. The population on the transient states decays into under collisions alone. Together, these results give one zero eigenvalue of , with algebraic multiplicity one. The normalized zero mode is positive within and zero outside it. Every other eigenvalue has positive real part, so all remaining modes decay [32].
The spectral argument uses particle conservation and one closed class rather than pairwise detailed balance [32]. At finite temperature, the Boltzmann distribution is stationary under and by detailed balance [29, 28], and under by Eq. (45). It depends on only through and is therefore even. Here is the carrier energy, is the temperature, and is Boltzmann’s constant.
At strictly 0 K, the equilibrium Bose occupation of the phonon modes vanishes, removing phonon absorption. Emission lowers carrier energy, while elastic defects preserve it. Scattering can therefore connect high-energy states to lower-energy states through one-way paths. Suppose every retained state can reach the same terminal band-edge shell. Also suppose elastic scattering connects all states within that shell, including any distinct valleys. Here “terminal” means that every allowed scattering transition from the shell stays within it. The shell is then the unique closed class , and the other states are transient. This establishes (ii), with supported on the shell.
For a complete, time-reversal-paired shell, a uniform population is stationary under elastic symmetric scattering. It is also stationary under skew scattering by Eq. (45). With absorption absent and no allowed lower-energy final state, phonons leave this population unchanged. The population is even under time reversal, which establishes (iii) in this setting.
Decomposition of the distribution
Under property (ii), normalize and define
Here is the identity operator on the carrier-state space. The projector extracts the component along the zero mode, while extracts a deviation with zero net carrier number. For a nonuniform , is an oblique projector because is not parallel to . Particle conservation and give
In particular, maps into itself. The notation denotes the restriction of to this invariant subspace. Its action on any is simply . Since the only zero mode is and , no nonzero vector in lies in the kernel of . Thus is invertible on the finite-dimensional space .
We now solve Eq. (22) for a time-independent optical source . Using the notation of Eq. (12) in the main text, write
Summing the kinetic equation over all states gives
The zero-mode contribution is therefore . Applying to the same kinetic equation gives the remaining evolution,
Its stationary solution is . Subtracting this solution yields a homogeneous equation,
Hence
Combining these expressions for and gives the explicit solution of the decomposition in Eq. (12) of the main text,
Here the last two terms are embedded in , and the last term decays as established in Sec. S1.2.
Finiteness of the current
The number-current functional is . Time reversal gives , so property (iii) implies
Thus the growing term in Eq. (24) contributes no current and
This establishes the finite-current statement used after Eq. (13) of the main text, including at 0 K when the stated conditions hold. The proof uses particle conservation, one closed class, and an even zero mode. Although skew scattering can convert a generic even distribution into an odd component, collisions leave unchanged because it is a zero mode of the full operator .
Electron and hole sectors
Without recombination or interband scattering, the electron and hole collision operators form separate blocks. Generation adds equal numbers of electrons and holes, and the preceding argument applies to each dilute carrier block when it satisfies properties (i)–(iii). For each sector , denoting conduction electrons and holes, respectively, let be its stationary non-growing distribution. The superscript labels the corresponding sector’s operator, projector, and source. The physical charge current is
Here is the positive elementary charge, and and are the electron and hole band velocities. The hole momentum is opposite to that of the missing valence electron. Equivalently, using valence-electron deviations and gives the usual electron-language charge current. The electron and hole contributions generally do not cancel, so the main text reports their signed sum as the total response.
Regime of validity
Real carriers cannot accumulate indefinitely. Both the stationary anisotropic deviation and the growing population must remain dilute and unsaturated [30, 27]. Let be the decay time of the current-carrying modes and the recombination time. Let denote the population available before band filling or depletion invalidates the linear source. A measurable plateau requires the time window
A large elastic out-scattering rate does not by itself guarantee small occupations, because energy relaxation can be much slower and allow carriers to accumulate in long-lived states. Within the linear regime, implies , where is the illumination intensity.
Angular-harmonic model
Collision operator, angular harmonics, and current
Under continuous-wave illumination in the low-injection limit, the photoinduced electron or hole occupation , measured relative to the equilibrium occupation , obeys the linear dilute-carrier Boltzmann equation of the main text,
Each collision operator has the out-minus-in form
where is the rate from to . Band indices are suppressed, and the model treats one carrier channel at a time. We separate ordinary relaxation from elastic skew scattering,
The superscript is reserved for the combined ordinary operator . The elastic symmetric defect kernel and operator are denoted by and throughout. The antisymmetry of follows from its definition in the main text.
Defect scattering is elastic, so and connect only states on the same constant-energy contour. Phonon scattering connects different energy shells. The model keeps it only as a loss rate from a given shell, as described in Sec. S2.3. The steady state can then be solved shell by shell, and the current is obtained by summing over shells.
Angular harmonics.
For the reduced model, take an effective circular energy contour with a uniform angular measure, parameterized by , avoiding the nonuniform measure and additional pocket indices that general contours may require. We expand
This angular-harmonic expansion is the two-dimensional analogue of a Legendre expansion and uses a complete basis on the chosen circle, which approximates the actual constant-energy contour. Since is real, its coefficients satisfy .
The angle is measured from the axis about the valley center. For the conduction band the center is . For the valence band the shell consists of the two time-reversal-related valleys at on –Y. Since exchanges these valleys, the parity statements below refer to their combined distribution on an effective time-reversal-paired shell. The reduced model represents this operation by .
The index counts the lobes of the angular pattern. is isotropic, is a one-sided displacement, and is a two-lobed alignment. Under , which represents in the effective paired-shell description, even- patterns remain unchanged, whereas odd- patterns change sign.
Current.
The band velocity is odd in and, in the isotropic approximation used below, takes the form , containing only harmonics. Here is the speed on the chosen circular energy shell. Projecting the distribution onto this velocity and using harmonic orthogonality gives
Within this isotropic velocity approximation only carries current, whereas anisotropic velocities can also project onto higher odd harmonics. An alignment is anisotropic, but it has equal weight at and and so carries no current.
Symmetry selection rules
For elastic scattering by nonmagnetic static defects, the following symmetry relations follow from time reversal and exchange [25 S7]. Expand the defect kernel on the effective shell as , where are harmonic indices and are the corresponding Fourier coefficients. The first symmetry is microscopic reversibility. In a nonmagnetic crystal, time reversal gives , which reads on the shell. In terms of the coefficients,
The second symmetry is the exchange property. The elastic symmetric defect kernel has coefficients , and the skew part obeys . Combining with Eq. (34),
The in-scattering term acting on samples coefficients (or their exchanged counterparts). The out-scattering term satisfies the same parity restriction. Equation (35) is therefore a parity rule for the elastic defect kernel. Symmetric defect scattering connects harmonics of equal parity, with even. Skew scattering connects harmonics of opposite parity, with odd. The symmetric operator relaxes the even and odd sectors separately and can never create current from an even distribution. The leading phonon relaxation channel in an equilibrium time-reversal-invariant bath also preserves global parity. Within the population-current mechanisms retained here, skew scattering couples the even source to the odd current sector. Thus preserves global momentum parity.
Two consequences are used below. First, a rotationally invariant kernel, , has , where
is its Fourier coefficient and is the Kronecker delta. Thus is always even. By Eq. (35), a rotationally invariant skew kernel is forbidden. The skew kernel must carry the orientation of the crystal, here the polar axis of . This is what allows it to convert an alignment into a directed current.
Second, the mirror of acts as and maps . A mirror-symmetric distribution has , so is purely imaginary. By Eq. (33), while . This is the angular-basis form of the statement in the main text that is the only in-plane component under -polarized light.
Angular-harmonic lifetimes
We first consider the elastic symmetric defect kernel. In the isotropic approximation it depends only on the scattering angle , so and . This is the angular-relaxation construction used in transport theory [33, 34]. The skew kernel remains anisotropic, as required by Sec. S2.2. The symmetric defect part of Eq. (30) becomes a convolution. Its eigenfunctions are the angular harmonics,
The part integrates to zero because is even. Each harmonic relaxes with its own rate, weighted by . For this is the usual momentum-relaxation rate, which is insensitive to small-angle scattering. For it is the alignment-relaxation rate, which is most sensitive to scattering near . For narrowly forward scattering, the approximation makes the defect relaxation rate proportional to , so the and defect-limited lifetimes differ by a finite factor rather than being parametrically separated in this limit.
Since independent phonon and defect transition kernels add, we represent phonon scattering in the reduced model by a diagonal effective decay rate , including loss to other shells that is absent from the elastic convolution in Eq. (36). The total relaxation rate is then
Here , where is the single-defect kernel of Eq. (4) of the main text. Equation (36) therefore gives the defect rate . Linearity in holds for dilute, uncorrelated defects, without multiple-scattering interference [26, 41]. This is also the regime in which the single-defect rates were rescaled in the numerics. is the phonon-limited rate of harmonic . It contains the inelastic loss from the shell, which is common to all . It also includes angular relaxation from quasi-elastic phonon scattering within the shell. The model places the temperature dependence in and assumes that cooling reduces these rates for a fixed carrier distribution. The defect rates and the skew coefficients introduced below are taken independent of and of . This means that the defect potential and its screening are those of the fixed, undoped host.
For the weight vanishes. Elastic scattering therefore does not relax the isotropic occupation of a shell, and exactly. The value of on a given shell is set by energy relaxation and generation. This mode plays no role in the current, as shown in Sec. S2.5, so its value is not needed.
Optical source
For single-photon electric-dipole absorption, the lowest-order population source is [13]
with . For time-reversed partner states, time reversal gives and, up to Bloch-state phases, . The sign and the conjugation disappear in the modulus squared. For a real polarization vector , which is linearly polarized light, this gives . The source is globally even (with partner bands summed when needed). In the effective paired-shell representation its expansion contains only even harmonics,
The same conclusion holds in the length gauge, where and . Here is the electron mass, is the interband transition frequency, and is the interband position matrix element. No odd population is injected directly at this order. This excludes the corresponding direct injection current for linearly polarized light, but not the intrinsic shift current, which is a distinct optical response [13, 14]. Circular polarization or phonon-assisted ballistic generation can add odd source terms [16] and are not considered here.
As an illustrative circular-contour source, assume that the transition dipole rotates with , . For ,
With the complex angular-harmonic convention of Eq. (39), this source has , or a real cosine amplitude of , with this normalization set by the illustrative dipole above. The dependence of the true matrix elements and the anisotropy of the contours add higher even harmonics to Eq. (40). These are neglected in the model and discussed in § 2.4.6.
Skew conversion
Inserting the antisymmetry into Eq. (30) merges the two terms,
Elastic symmetric scattering acts on occupation differences. The skew term acts on their sum and converts angular components. When its row sums vanish as shown below, its real matrix is antisymmetric and contributes no quadratic dissipation on the elastic shell, while still changing individual occupations [25].
The isotropic population generates no current. For a constant , Eq. (41) reduces to with . Antisymmetry alone gives only , which is particle conservation. It does not give for each . The parity rule (Eq. (35)) likewise permits . The vanishing follows instead from the explicit form of the skew rate, Eq. (6) of the main text,
Because , exchanging any two arguments of the triple product turns it into its complex conjugate,
Likewise , which is the antisymmetry used throughout. Summing Eq. (42) over gives
The two intermediate states now enter on an equal footing. The summation domain and the product of delta functions are symmetric under , while is antisymmetric by Eq. (43). The sum therefore vanishes in pairs, and for every . Hence
for any energy-only , at the retained third order for a static Hermitian defect potential and complete, consistently weighted intermediate-state sums, without assuming isotropy. This is the elastic probability-sum constraint underlying the skew collision integral [25]. Thus skew scattering leaves the isotropic population unchanged but can act on the anisotropic part of the distribution.
Conversion of the alignment. The parity rule allows the retained alignment to couple to , but the coupling strength is material dependent and may vanish. We define it through the projection of the skew collision operator below. In the mirror-symmetric alignment sector, and . Together with the reality of the distribution and source, these relations make and real, so we use them directly as in the main text. For the current projection, define , so . For the mirror-symmetric alignment sector, define the real coefficient by
The subscript 1 selects the Fourier coefficient, and includes both components. Its sign depends on the defect potential, the source orientation, and the current convention. In the dilute single-defect Born expansion, the skew rate is linear in defect fraction and third order in the potential, whereas the leading symmetric rate is second order [25, 26].
The first-order skew approximation requires the norm of on the relaxing subspace to be small. Under this condition, the odd response is first order in skew strength and its feedback on the even alignment is second order. Higher odd harmonics do not contribute to the isotropic-velocity current at this order. The alignment and source are the same amplitudes used in Eqs. (17)–(18) of the main text, with fixed current-normalization factors absorbed into the proportionality constant.
Closed-form response and relation to the full calculation
Balancing each harmonic in Eq. (29) at steady state, and using Eqs. (37), (39), (45) and (46), gives
With Eq. (33), and after dividing by the fixed illumination intensity,
This is Eq. (18) of the main text. The defect fraction appears once in the numerator, through the conversion, and twice in the denominator, through the relaxation of the alignment that is converted and of the current that results.
With , , and held fixed, the finite optimal fraction in Eq. (19) of the main text requires positive and and a nonzero . At strictly 0 K, spontaneous emission can vanish at a band minimum [28], in which case the low-friction limit must be re-examined under the steady-current conditions of Sec. S1.
Relation to the full calculation. The model uses a circular effective shell, isotropic ordinary relaxation, a truncated source, diagonal effective phonon losses, and weak-skew conversion. The full calculation retains the actual bands, angular structure, and in-scattering rather than these shell-model reductions. The isotropic approximation makes diagonal in the angular basis, Eq. (36). Replacing the phonon operator by diagonal effective losses then gives one per harmonic, Eq. (37). Another approximation is the truncation of the even sector to , Eq. (40). The underlying global momentum-parity rules, the energy-only cancellation (S24), source evenness, and the mirror constraint on current direction survive under their stated physical assumptions. The full solution solves Eq. (29) with the anisotropic , the full skew operator, and all bands, and retains the complete in-scattering structure. The model is used to interpret the single-channel scaling, namely the unit-slope rise, the plateau or peak, the optimal fraction in Eq. (19) of the main text, and the cooling enhancement in Figure 2. Absolute magnitudes and the signed sum of the electron and hole channels are taken from the numerics.
Structure beyond the model can arise from the neglected even harmonics. The anisotropic matrix elements of GeS generate , and anisotropic symmetric scattering mixes them with . Each even harmonic is converted to by with its own coefficient , defined in the same mirror-symmetric sector and phase convention as by
Here is real in this sector. Each alignment is quenched at its own threshold . Two such pathways with conversion coefficients of opposite sign and different thresholds can produce a sign reversal if their relative weights exchange dominance as varies. This is one mechanism consistent with the behavior of the Se hole channel at 300 K in the main text.
Supplemental References
References
- [1]R. von Baltz and W. Kraut, Theory of the bulk photovoltaic effect in pure crystals, Phys. Rev. B 23, 5590 (1981).DOI
- [2]B. I. Sturman and V. M. Fridkin, The Photovoltaic and Photorefractive Effects in Noncentrosymmetric Materials, Ferroelectricity and Related Phenomena, Vol. 8 (Gordon and Breach Science Publishers, New York, 1992).
- [3]V. I. Belinicher and B. I. Sturman, The photogalvanic effect in media lacking a center of symmetry, Sov. Phys. Usp. 23, 199 (1980).
- [4]L. Z. Tan, F. Zheng, S. M. Young, F. Wang, S. Liu, and A. M. Rappe, Shift current bulk photovoltaic effect in polar materials—hybrid and oxide perovskites and beyond, npj Comput. Mater. 2, 10026 (2016).
- [5]S. M. Young, F. Zheng, and A. M. Rappe, First-Principles Calculation of the Bulk Photovoltaic Effect in Bismuth Ferrite, Phys. Rev. Lett. 109, 236601 (2012).
- [6]Z. Dai and A. M. Rappe, Recent progress in the theory of bulk photovoltaic effect, Chem. Phys. Rev. 4, 011303 (2023).arxiv.org/abs/2206.00602
- [7]B. I. Sturman, Ballistic and shift currents in the bulk photovoltaic effect theory, Phys. Usp. 63, 407 (2020).
- [8]A. M. Burger, R. Agarwal, A. Aprelev, E. Schruba, A. Gutierrez-Perez, V. M. Fridkin, and J. E. Spanier, Direct observation of shift and ballistic photovoltaic currents, Sci. Adv. 5, eaau5588 (2019).DOI
- [9]A. M. Burger, L. Gao, R. Agarwal, A. Aprelev, J. E. Spanier, A. M. Rappe, and V. M. Fridkin, Shift photovoltaic current and magnetically induced bulk photocurrent in piezoelectric selenite crystals, Phys. Rev. B 102, 081113(R) (2020).DOI
- [10]F. Wang and A. M. Rappe, First-principles calculation of the bulk photovoltaic effect in KNbO₃ and (K,Ba)(Ni,Mn)O₃−δ, Phys. Rev. B 91, 165124 (2015).DOI
- [11]A. M. Schankler, L. Gao, and A. M. Rappe, Large Bulk PiezoPhotovoltaic Effect of Monolayer 2H-MoS₂, J. Phys. Chem. Lett. 12, 1244 (2021).DOI
- [12]L. Z. Tan and A. M. Rappe, Enhancement of the Bulk Photovoltaic Effect in Topological Insulators, Phys. Rev. Lett. 116, 237402 (2016).DOI
- [13]J. E. Sipe and A. I. Shkrebtii, Second-order optical response in semiconductors, Phys. Rev. B 61, 5337 (2000).DOI
- [14]S. M. Young and A. M. Rappe, First Principles Calculation of the Shift Current Photovoltaic Effect in Ferroelectrics, Phys. Rev. Lett. 109, 116601 (2012).pubmed.ncbi.nlm.nih.gov/23005660
- [15]R. Fei, L. Z. Tan, and A. M. Rappe, Shift-current bulk photovoltaic effect influenced by quasiparticle and exciton, Phys. Rev. B 101, 045104 (2020).DOI
- [16]Z. Dai, A. M. Schankler, L. Gao, L. Z. Tan, and A. M. Rappe, Phonon-Assisted Ballistic Current from First- Principles Calculations, Phys. Rev. Lett. 126, 177403 (2021).DOI
- [17]Z. Dai and A. M. Rappe, First-principles calculation of ballistic current from electron-hole interaction, Phys. Rev. B 104, 235203 (2021).DOI
- [18]V. I. Belinicher, E. L. Ivchenko, and B. I. Sturman, Kinetic theory of the displacement photovoltaic effect in piezoelectrics, Sov. Phys. JETP 56, 359 (1982).
- [19]P. Zhu and A. Alexandradinata, Anomalous shift and optical vorticity in the steady photovoltaic current, Phys. Rev. B 110, 115108 (2024).
- [20]A. M. Schankler, Y. Bao, and A. M. Rappe, First-Principles Calculation of the Shift Current Bulk Photovoltaic Effect from Carrier Recombination, Phys. Rev. Lett. 136, 106904 (2026).DOI
- [21]J. Yu, A. Grieder, J. Simoni, R. Ullah, Z. Bai, R. Sundararaman, A. Alexandradinata, and Y. Ping, Photogalvanic currents from first-principles real-time density-matrix dynamics (2026), arXiv:2601.01059 [cond-mat.mtrl-sci].arxiv.org/abs/2601.01059
- [22]S. Ganguly, S. Varghese, A. M. Schankler, X. Xu, K. Morita, M. Viret, A. M. Rappe, G. Catalan, and K.-J. Tielrooij, Direct spatiotemporal imaging of carriers reveals a long-lived bulk photovoltaic mechanism (2025), arXiv:2509.01729 [cond-mat.mtrl-sci].arxiv.org/abs/2509.01729
- [23]E. Deyo, L. E. Golub, E. L. Ivchenko, and B. Spivak, Semiclassical theory of the photovoltaic effect in non-centrosymmetric systems (2009), arXiv:0904.1917 [cond-mat.mes-hall].arxiv.org/abs/0904.1917
- [24]P. Sukhachov, P. Zhu, E. Banyas, L. Z. Tan, and A. Alexandradinata, Vortex-enhanced photovoltaic current in disordered topological materials (2026), arXiv:2606.28509 [cond-mat.str-el].arxiv.org/abs/2606.28509
- [25]H. Isobe, S.-Y. Xu, and L. Fu, High-frequency rectification via chiral Bloch electrons, Sci. Adv. 6, eaay2497 (2020).
- [26]I.-T. Lu, J.-J. Zhou, and M. Bernardi, Efficient ab initio calculations of electron-defect scattering and defect-limited carrier mobility, Phys. Rev. Materials 3, 033804 (2019).DOI
- [27]M. Bernardi, First-principles dynamics of electrons and phonons, Eur. Phys. J. B 89, 239 (2016).DOI
- [28]F. Giustino, Electron-phonon interactions from first principles, Rev. Mod. Phys. 89, 015003 (2017).DOI
- [29]J.-J. Zhou, J. Park, I.-T. Lu, I. Maliyov, X. Tong, and M. Bernardi, Perturbo: A software package for ab initio electron–phonon interactions, charge transport and ultrafast dynamics, Comput. Phys. Commun. 264, 107970 (2021).DOI
- [30]C. Jacoboni, Theory of Electron Transport in Semiconductors: A Pathway from Elementary Physics to Nonequilibrium Green Functions, Springer Series in Solid-State Sciences, Vol. 165 (Springer, Berlin, Heidelberg, 2010).
- [31]See Supplemental Material for the steady-state conditions and the angular-harmonic model, which includes Refs. [S1–S14].
- [32]J. R. Norris, Markov Chains (Cambridge University Press, Cambridge, 1997).
- [33]N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders College Publishing, 1976) chap. 16.
- [34]S. V. Koniakhin, Ratchet effect in graphene with trigonal clusters, Eur. Phys. J. B 87, 216 (2014).DOI
- [35]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials, J. Phys.: Condens. Matter 21, 395502 (2009).
- [36]J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett. 77, 3865 (1996).pubmed.ncbi.nlm.nih.gov/10062328
- [37]H. J. Monkhorst and J. D. Pack, Special points for Brillouin-zone integrations, Phys. Rev. B 13, 5188 (1976).DOI
- [38]S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Phonons and related crystal properties from density-functional perturbation theory, Rev. Mod. Phys. 73, 515 (2001).DOI
- [39]N. Marzari and D. Vanderbilt, Maximally localized generalized Wannier functions for composite energy bands, Phys. Rev. B 56, 12847 (1997).DOI
- [40]F. Giustino, M. L. Cohen, and S. G. Louie, Electron-phonon interaction using Wannier functions, Phys. Rev. B 76, 165108 (2007).
- [41]I.-T. Lu, J. Park, J.-J. Zhou, and M. Bernardi, Ab initio electron-defect interactions using Wannier functions, npj Comput. Mater. 6, 17 (2020).DOI
- [42]G. Pizzi, V. Vitale, R. Arita, S. Blügel, F. Freimuth, G. Géranton, M. Gibertini, D. Gresch, C. Johnson, T. Koretsune, J. Ibáñez Azpiroz, H. Lee, J.-M. Lihm, D. Marchand, A. Marrazzo, Y. Mokrousov, J. I. Mustafa, Y. Nohara, Y. Nomura, L. Paulatto, S. Poncé, T. Ponweiser, J. Qiao, F. Thöle, S. S. Tsirkin, M. Wierzbowska, N. Marzari, D. Vanderbilt, I. Souza, A. A. Mostofi, and J. R. Yates, Wannier90 as a community code: new features and applications, J. Phys.: Condens. Matter 32, 165902 (2020).
- [43]D. G. Anderson, Iterative Procedures for Nonlinear Integral Equations, J. ACM 12, 547 (1965).DOI
- [44]T. Rangel, B. M. Fregoso, B. S. Mendoza, T. Morimoto, J. E. Moore, and J. B. Neaton, Large Bulk Photovoltaic Effect and Spontaneous Polarization of Single-Layer Monochalcogenides, Phys. Rev. Lett. 119, 067402 (2017).DOI


