WorksPhysics and Astronomy
Intraband vs. interband optical nonlinearities of transparent conducting oxides
Abstract
We show that the intraband thermo-optical response to ultrashort intense infra-red illumination of transparent conducting oxides is much stronger than the nonlinear response associated with resonant and non-resonant interband transitions in typical experimental setups. This finding rules out several speculative explanations of recent measurements.
Introduction
One of the ways to classify nonlinear optical effects is via their speed and strength [1, 2, 3, 4]. In particular, the abundant Kerr effect [5, 6] is known to be (nearly) instantaneous yet weak; this stems from its non-resonant nature. Indeed, it involves virtual transitions of valence electrons to the conduction band in solids. Wave mixing interactions have similar characteristics. On the other hand, resonant (i.e., real) transitions gives rise to much stronger nonlinear effects, but while they modify the permittivity instantaneously, their decay time is limited by the lifetime of the excited electron states. The latter class of nonlinear effects include free carrier generation due to valence to conduction band transitions, involving resonant single/multi photon absorption, thermal effects, stress, strain etc.. Thus, the scientific community is in an ongoing quest to find the nonlinear optical material that allows enjoying the best of the two worlds - a nonlinear optical effect which is both fast and strong.
On this background, there was a great deal of excitement when the first studies of transparent conducting oxides (TCOs) showed a unusually strong and fast nonlinear optical response [7, 8, 9]. Those materials are highly doped semiconductors; they are also occasionally referred to as low electron density metals [10] or -doped metals [11]. Originally, the strong nonlinear optical response of TCOs was associated with the proximity to the frequency at which the permittivity of the TCO changes sign from positive to negative; this is the well-known epsilon-near-zero (ENZ) point. In this spectral regime, even though the absolute change of the permittivity may not be much different from other spectral regimes, the relative permittivity change can, ideally, diverge. Indeed, it is the relative change which is responsible to the strength of the nonlinear response [12, 7, 13]. This observation motivated a wide range of nonlinear optical effects in TCOs, all showing indeed unusually strong effects [14, 15, 16, 17, 18, 19, 20, 21], see the following reviews for more information [22, 23, 14].
However, it was soon realized that the relative permittivity change can never diverge because of the small, yet non-zero imaginary part of the permittivity. Thus, the response has “only” a resonance at the ENZ point [24, 25]. Instead, the strong nonlinear effects seen in measurements were shown to originate from the non-parabolicity of the conduction band [26], and associated dependence of the plasma frequency [27] on the electron temperature1. Later work re-framed this dependence in the (more transparent) context of the momentum/energy dependence of the effective mass of the band structure of semiconductors [28]. The analysis in [28] pointed to the total absorbed energy as the key parameter affecting the temperature rise, hence, the permittivity change; hence, it also indicates its instantaneous absorptive nature.
This insight raised the need to determine the temperature. This originally was done using the Two Temperature Model (TTM) [29, 30, 31, 32], as done earlier in noble metals. However, the formulae used for the thermo-physical parameters in the TTM did not account for the unique features of the TCOs, namely, the low conduction electron density and the non-parabolicity. As a consequence, the TTM-based theory provided a qualitative match to measurements, but not a quantitative one [15, 33].
A remedy to this limitation was provided by Sarkar, Un et al. in [10], who derived the thermophysical parameters of TCOs rigorously from a Boltzmann model for the electron dynamics. This work showed that the electron heat capacity of TCOs is much smaller than for noble metals, and should include high-order corrections due to the elevated temperature, as in textbook formulae [34]. More importantly, it showed that a correct account for the interactions requires accounting also for momentum conservation (this was not necessary for noble metals). With these considerations, and with a Lindhard-based model for the permittivity [35], a quantitative match to experiments was obtained [33]. Remarkably, the latter work also showed that a thermal model (i.e., based on the TTM, which naturally emerges from the microscopic Boltzmann model for the electron distribution function [31, 10] by multiplying by the energy and summing over all energy states) yields almost the same predictions as the non-thermal model. It showed also that there is no need to include the non-thermal part appearing in the so-called extended two temperature model [36, 37] even for pulses only a few femtosecond long. All this happens because the time scale for induced thermalization is faster than even for noble metals [1, 10]. This model thus reveals a simple physical picture - the permittivity changes essentially instantaneously by absorption of light - it is fairly insensitive to whether the electron distribution is thermal or non-thermal; it then relaxes on the cooling rate, determined by the strength of interactions. This shows that the nonlinearity associated with conduction band electrons is essentially thermo-optic. Further studies based on the more advanced, fully-quantum coherent density matrix formulation [38], demonstrated the small role of the deviation from equilibrium also for pulses no more than a few femtosecond long.
At about the same time, a new gap opened between the TCO theory and measurements. In particular, in the quest for (the already demonstrated) time reversal of optical pulses [39, 40] and temporal photonic crystals at optical frequencies, Lustig et al. compressed their pump pulse to a single cycle and TW/ intensities, and measured the resulting changes to probe transmissivity and reflectivity from a thin ITO layer [41]. Surprisingly, they observed a probe limited permittivity rise and decrease in transmissivity, which nevertheless was replaced midway by a very slow decay; the reflectivity showed the usual fast rise-slow decay dynamics. Peculiarly, this feature was observed in a more recent experiment only with a much longer and much more intense pump pulse. Complex dynamics was observed also in [42], measured with an even longer probe pulse. None of these complex dynamical features was reproduced in the density-matrix based model of [38, 43], which included only the conduction electrons. Several recent works proposed that nonlinearities associated with interband transitions may be able to explain the complex dynamics observed in these experiments [44, 45, 42, 46, 47].
In this paper, based on the success of the thermal model to reproduce correctly the results of the more general non-thermal non-adiabatic coherent model [35, 38, 43], we couple the thermal model for the electron dynamics to a rigorous FDTD (Finite Difference Time Domain) model for the Maxwell equations in a thick TCO layer. Unlike the approach in [35], the current approach addresses both field and temperature non-uniformities. We also provide a simple way to account for the temperature dependence of the permittivity within this model. To this model we add the two leading interband nonlinearities - the Kerr effect and 3 photon absorption (since two-photon absorption is well below the gap), see Figure 1 for a schematic illustration of these various processes. We then show that these two effects are far weaker than the absorptive/thermo-optic nonlinearity, thus, implying on the unlikelihood of the processes suggested so far to explain the observed behaviour for short pulses.

Figure 1. Schematic of the electronic states and the optical processes considered in this work. The conduction band follows the Kane two-band dispersion, and the valence band is taken to be flat (dispersionless). Conduction electrons form a degenerate gas that fills the band up to the Fermi-level (blue shading, with a thermally broadened Fermi edge). All orange arrows correspond to the photon energy . The (thick) curved arrow marks (indirect) intraband absorption of a photon by an electron; the resulting non-thermal electron subsequently relaxes back towards the Fermi-level by (net) phonon emission (e-ph) and by carrier–carrier scattering (e-e), shown as black arrows connecting states on the dispersion curve. Filled circles denote electrons. The vertical (thin) arrows starting at the valence band represent the non-resonant single-photon virtual transition responsible for the Kerr effect, and three-photon absorption (3PA), in which three successive steps promote an electron into empty conduction-band states. Both processes start below the hatched dipole-forbidden region at the top of the valence band.
Model
Configuration. Given the recent interest in exploring fast features in reflection and transmission near the temporal overlap of the pump and probe pulses, as suggested by [41], we focus our analysis on an thin film under similar conditions as a case study. Within this context, we seek to test the hypotheses that interband interactions can give rise to these fast features by expanding the existing free-electron thermal model of the intraband nonlinearity. Specifically, we consider a 310 nm thick ITO layer on a thick layer of glass, see Figure 2. The (effective) plasma frequency and collision rate for the sample are set to reproduce the reported values in [41 Fig. 1D] by extracting the real and imaginary permittivity values from the figure and fitting this to a Drude dispersion model seeded with properties of TCOs with a comparable crossover permittivity, see Appendix § 5.§ 1. We set a spatio-temporal Fourier-limited Gaussian pump (probe) pulse with central wavelength nm (nm, i.e., close to the ENZ point in [41]). Their respective durations are set to a Full-Width-Half-Maximum of 6 and 44 fs, respectively. The incident pump intensity is set to 1 TW/cm, and the probe intensity is sufficiently low so that it does not affect the dynamics.

Figure 2. Schematic illustration of the thin film on glass where the change in reflection and change in transmission due to the excitation of the ITO is determined. We consider a case of near normal incidence for both the pump and probe beams.
Optical measurements of the interband absorption threshold [48, 49, 50, 51] as well as recent Density Functional theory (DFT) calculations of the band structure of ITO [52] reveal the threshold for interband transitions is eV; it consists of the actual 2.6 eV gap, the Fermi energy (estimated to be eV for the sample of [41], determined through the conduction band profile of see Appendix § 5.§ 1), and a range of 0.5 − 0.7 eV of valence bands which are effectively dipole forbidden (so effectively inaccessible to light); the latter are not taken into account in most recent works. Thus, at least 3 photons are required to exceed the interband transition threshold of the heavily doped ITO sample; this process is depicted in Figure 1. Higher-order multi (i.e., 4,5,...)-photon absorption events are neglected as scaling of N-photon absorption are generally progressively weaker [53]; this choice is justified a-posteriori. Note that despite the specific case taken here, the model proposed matches multiple other recent experiments [46, 54], is general and can be applied to other materials with appropriate adjustments for material and excitation parameters. Specifically, for smaller bandgap materials or higher pump energies, two-photon absorption may become relevant.
Model Description. We solve Maxwell’s equations in one spatial dimension and in time for a pump pulse and a delayed probe pulse normally incident on a TCO layer on a glass substrate, see Figure 2.
Specifically, we solve
and
where the overall electromagnetic fields are given by and ; and are the permittivity and permeability of the vacuum. An auxiliary differential equation is used to describe the electric current originating from the light-driven free (i.e., conduction band) electrons [55]; thus, the associated (linear) Drude (i.e., intraband) current is given by
where , is the Drude damping rate and is the plasma frequency, both effectively averaged over electron energies (hence also -dependent), see Appendix § 5.§ 2. This dependence constitutes (within the assumption of thermal equilibrium) the (slow) thermo-optic nonlinearity [56] (to be shown to be the dominant one). By Poynting’s Theorem, the absorbed electromagnetic field power density is given by
where is the nonlinear current induced by interband 3PA transitions, see also below.
Notably, the absorption (4) is a strongly oscillatory quantity (see [38]), which is nevertheless traditionally averaged out [57]. The (positive) average gives rise to the traditional heat source. Yet, the rapid (doubled carrier frequency) oscillations represent an observable feature; specifically, they were observed experimentally in glass [58]. They represent the fast free-carrier induced current discussed in [56]. This classical feature is nothing but the (nonlinear extension) of the quiver motion / Bloch oscillations [34] (see Figure 1), and was shown to resemble quantum coherent oscillation arising from quantum master equations [43, 38] or TD-DFT equations for electron dynamics [58]. The permittivity modulation associated with this oscillation cen be well approximated by a cubic (coherent) nonlinearity (and thus, incorrectly lumped onto the (by definition, incoherent) Kerr coefficient, see Appendix § 7).
The absorbed power serves as a heat source that gives rise to the photo-thermal effect - it causes the electron temperature to increase, and consequently, also the temperature of the phonons to increase due to collisions. The details of these processes are traditionally described by the Two Temperature Model (TTM) [29, 31, 10], i.e.,
Here, the electron heat capacity takes the standard free electron form [34], with high-order corrections necessary at high temperatures (see, e.g., [59 Eq. (25)])2; the electron temperature dependence of the – coupling relies on the analytic expression derived in [10]. is set to an empiric value of ; its value is secondary, as it is associated with nanosecond dynamics.
Note that for simplicity, we do not include diffusive heat transport [60, 37], as it is expected to be relatively slow due to the very short – collision times [10].3 For simplicity, we also neglect ballistic transport, which could be more significant, yet, not likely to affect the dynamics in more than a mild quantitative manner, see [61].
Integral expressions are given in Appendix § 5.§ 2 also for the effective damping rate and the plasma frequency appearing in Eq. (3); as said, these manifest the thermo-optic effect - the reverse coupling of the heat equations to the equation for the electric current, and from it to the electromagnetic fields. In earlier work [35], and (and their temperature dependence) were given implicitly within a Lindhard type integral over the (non-thermal) electron distribution. In Appendix § 5.§ 2 we extract them into a form which requires only knowing the non-parabolicity and Fermi-level. Notably, the dependence suggested in Appendix § 5.§ 2 generalizes the dependence of the effective mass on the non-parabolicity of the conduction band [28]. It also allows us to model the temperature-dependence of all these parameters up to several thousand degrees where , and thus, to circumvent inaccuracies occurring at such temperatures when simpler fits are used [62, 63, 64]. For the pump-probe delays studied in the current work, the dependence of the collision rate on the phonon temperature can be neglected.
The thermo-optic nonlinearity described so far merely implement existing knowledge in a manner that is spatially distributed (thus, improving the calculation of the spectroscopic properties) and is simpler to implement compared to early versions involving energy integrals over the electron population. We now move on beyond it, by including optical nonlinearities associated with interband transitions.
The first additional mechanism included in our model is the instantaneous Kerr effect (i.e., the effect of non-resonant excitation of valence band electrons, see Figure 1), which appears in the expressions for the effective permittivities:
Here, is the linear non-dispersive contribution to the permittivity due to interband transitions for frequencies far from the interband threshold and describes the instantaneous effects of self-phase modulation of the pump and cross-phase modulation of the probe.4 In Appendix § 6.§ 1, we provide a detailed estimate of the value of this coefficient, taking into account well established values known for similar doped semiconductors, and making sure not to confuse it with the intraband fast nonlinearity. We then exaggerate the value we estimated 10-fold, in order to emphasize our emerging conclusion, that the Kerr effect is much smaller than the thermo-optic nonlinearity.
A second additional nonlinear mechanism we include is (simultaneous) 3-photon absorption events which promote valence electrons to the conduction band (see Figure 1). The process of 3PA results in multiple nonlinear effects, each having a distinct contribution to the dynamics; these were not previously discussed in detail before, to our knowledge. First, it gives rise to a nonlinear (bound/valence electron) current, (appearing already in Eq. (1)) where is the vacuum impedance and is the coefficient of three photon absorption events. This form is phenomenologic, based on the Poynting theorem; it shortcuts the more elaborate inclusion of a proper Lorentz-like equation for the current associated with the valence electrons (e.g., [65, 66, 45]).
The 3PA effect further contributes to increasing the conduction electron density (and hence, the effective plasma frequency in Eq. (3) which is linearly proportional to it); this effect is manifested via an electric field-dependent contribution to the effective plasma frequency [67, 68] via the electron density. Specifically,
where is the pump central frequency. The decay of excess charge due to recombination typically occurs on the scale longer than the time scale of interest here ( fs for Shockley-Reed-Hall recombination [7, 15]); for that reason, the relaxation of the electron density is neglected.
Lastly, as already shown by the last term on the RHS of Eq. (4), 3PA contributes to the heating source an amount which is proportional to the excess energy deposited by the three absorbed photons above the interband transition threshold .
Note that as for the intraband absorption, we do not remove the fast oscillation arising from the powers of the electric field; this is yet another manifestation of the fast nonlinearity arising from the free electron response. As for the Kerr effect, in Appendix B.2, we estimate the value of the 3PA coefficient based on measured values for similar doped semiconductors. The value chosen, , is again, most likely an exaggerated value. Thus, overall, all physical parameters associated with the inraband response are computed rigorously from microscopic models [10] (and are known within reasonable sample-to-sample variation) whereas in the absence of verified DFT calculations for the nonlinear optical coefficients of ITO, the missing values are estimates based on the measured response of comparable wide bandgap semiconductors.
All quantities are computed on the same time grid and advanced using a 4th-order Runge-Kutta method, whereas the electric and magnetic field are staggered in space (as for the Yee grid); the currents are spatially-collocated with the electric field. We have verified that in the linear regime, the probe dynamics is in agreement with the reflection and transmission Fresnel theory (with the relative error being no more than a fraction of a percent).
The model described above exceeds the state-of-the-art for TCOs ([45, 46]) by relying on rigorous temperature dynamics (extracted from microscopic electronic modelling [10, 35]) and corresponding temperature-dependence of optical and thermal parameters as well as on their unique non-parabolicity and low electron density; this contrasts the phenomenological fit of the nonlinearity via a series of high-order powers of the electric field [66, 45, 46, 42]. Our model also adds interband MPA that to our understanding, was never accounted for before in a complete manner in TCO studies. Our model nevertheless does not include coherent nonlinearities giving rise to nonlinear wave mixing, and in particular, not the more sophisticated nonlinearities usually included in hydrodynamic models [65, 69, 70, 66, 45], as those too involve mostly coherent second-order nonlinearities, which are out of the scope of the current study. Similarly, as for [43, 38], our model does not account for the strong nonlocality in TCOs [66, 71], although it may become relevant as the film becomes thinner.
Furthermore, our model does not account for stimulated emission (which reduces the total absorption) or excited state absorption (i.e., non-simultaneous multi photon absorption events), because those effects seem to mostly cancel each other [43].
Finally, we do not include the Kerr nonlinearity of the substrate and focus our analysis on the thin ITO film. Isolation of the thin film response in experiment can be achieved under cross-polarized excitation between pump and probe, where the effect of the substrate is sufficiently small compared to the TCO that it can be neglected in most scenarios. However, under parallel polarization this may not always be true [72, 64] and care must be taken in analysis. Yet in this case, proper normalization can remove this effect for well-known substrates such as fused silica to isolate the thin film response.
Results
Overall, the pulses traverse the layer, with the pump being partially absorbed, so that heat, nonlinear currents and excess free carriers are being generated in the ITO layer, along with cross-phase modulation being induced upon the probe by the Kerr effect. The dynamic permittivity changes are primarily generated by the thermo-optic nonlinear response which themselves arise from the increase of the electron and phonon temperatures. Figure 3 shows the resulting spatially-averaged dynamics of the electron temperature and various material properties (over the thickness). One can see that the change builds up on the pump-duration time scale, but relaxes on a much longer time scale, dictated primarily by coupling; notably, there is no fast oscillatory feature. For our choice of parameters, the electron temperature rises by several thousand degrees, the effective plasma frequency drops by , and the the electron collision rate grows drastically.

Figure 3. Dynamics of spatially-averaged pump-induced change of (a) electron temperature , (b) effective plasma frequency (blue) and collision rate (red), and (c) real (blue) and imaginary (dashed red) part of the intraband part of the ITO layer permittivity , at the central probe frequency for incident intensity of 1 TW/cm; the interband effects of Kerr and 3PA are not included in this simulation.
As observed for longer pulses before [35], the real part of the permittivity undergoes order unity changes, and the increase of the imaginary part5 are only somewhat smaller
In Figure 4(a), we show the dependence of the spectroscopic quantities (, , ) on the pump-probe delay time. Upon illumination, these quantities change by . Note that since depends on the delay between the pump and probe, one can expect the peaks of and reflectivity not to be at the same position (as pointed out already in [45]).

Figure 4. (a) Transmissivity (red), Reflectivity (green) and Absorptivity (black) through an ITO layer with a pure thermo-optic nonlinearity (dots) illuminated by a pump pulse of maximal intensity of 1 TW/cm² as a function of the pump-probe delay. (b) and (c) show the change in these quantities due to a(n exaggerated) (b) Kerr (circles) and (c) 3PA (stars) nonlinearities, respectively.
We now turn to consider the secondary MPA and Kerr interband effects. We will see that they are negligible in comparison to the thermo-optic nonlinearity.
Kerr nonlinearity
When we include the effect of the Kerr nonlinearity in the ITO layer, a clear instantaneous change is observed on the background of the thermo-optic nonlinearity (Figure 4(b)). Following Appendix § 6.§ 1, we set so that the corresponding maximal index change () would be comparable to the thermo-optic nonlinearity (see Figure 3(c)). Despite that, we observe transmissivity (reflectivity) changes of only () with respect to the case with only a thermo-optic nonlinearity, i.e., -fold (-fold) smaller changes. Thus, our rigorous treatment shows that . This is a result of the transient nature of the permittivity change, which limits the interaction time that leads to reflection [3]; indeed, this result contrasts the CW illumination case for which [44]. Since the scaling of the Kerr-induced changes are linear with the incident intensity (not shown), the emerging conclusion is that in order to explain the experimental results of [41], the ITO Kerr coefficient should be more than an order of magnitude higher than the already 10-fold exaggerated value set in our simulations to play a meaningful role in and explain the rapid transient reported in [41].
Multi-photon absorption
Figure 4(c) shows the effect of 3PA on the transmissivity, reflectivity and absorptivity. Overall, for our generic choice of parameters, 3PA hardly affects these quantities - its effect is , i.e., even smaller than the Kerr effect. The main reason for the smallness of 3PA is the fact that for the experimental configuration we are reproducing [41] (which is representative of many other experimental systems [46, 54]), 3PA is at the spectral threshold for interband transitions.
An additional reason is more subtle and somewhat unexpected. We observe that the different effects caused by 3PA have opposing contributions to the permittivity, thus, partially canceling each other. In particular, Figure 5(a) shows that for 1 TW/cm incident intensity, while the nonlinear current leads to an increase of the transmittance, the generation of excess free carriers in the conduction band (see Eq. (8)) has a slightly stronger effect of opposite sign (heat generation also has a negative effect, but is much weaker); the effects on the absorptivity have opposite signs (Figure 5(b)), and the reflectivity is less sensitive to either (Figure 5(c)). This cancellation reduces the overall effect of 3PA by another order of magnitude. To add to the complexity, similar results for half the incident intensity reveal that the reflectivity and absorptivity show different trends (e.g., the differential absorption actually changes sign) and while the transmissivity seems to be qualitatively similar and quantitatively lower, it is only 4-fold and not 8-fold lower, as naively expected. This tells us that one cannot predict reliably what will be the magnitude of 3PA for much higher incident intensities. This aspect is therefore missed in studies that do not consider all three effects arising from the multi-photon absorption events [42].

Figure 5. Breakdown of the effects caused by 3PA for incident intensity of . The change of transmissivity due to the 3PA induced current (last term on the RHS of Eq. (1)) is shown by squares. The addition of the 3PA-induced heat generation (last term on the RHS of Eq. (4); + symbols) causes a slight reduction of the transmissivity change, and the addition of the excess charge carriers to the conduction band (see Eq. (8); stars, as in Figure 4(c)) inverts the sign of the change. Similar behaviour is seen for the absorptivity (except for opposite signs); the reflectivity is affected mildly. (d)-(f) shows the same data for incident intensity of .
While 3PA may be stronger for higher intensities and/or other TCOs which have a somewhat narrower gap, it is to be emphasized that, like the thermo-optic nonlinearity, all three effects arising from 3PA are cumulative over time (i.e., non-instantaneous), and hence, they are unlikely to yield a fast feature as seen in [41]. Furthermore, even if a fast feature does arise, 3PA actually pushes the material system further away from thermal equilibrium, thus making the realization of temporal photonic crystals even less likely.
Our results shed light on recent studies that associated fast features to MPA. Specifically, [42] claimed-offsetof that 2PA and 3PA may explain the rapid (and oscillatory) behaviour of the transmissivity and reflectivity changes induced by a few femtosecond pump pulse absorbed in a thin ITO layer. Notably, however, the model in that work did not consider the dominant thermo-optic nonlinearity, and instead, relied on the phenomenologic MPA effects within a 3 level system model (rather than within the continuous band characteristic of solids). Furthermore, the interband threshold was likely underestimated, leading to the unlikelihood that 2PA processes contribute meaningfully to the response even in the presence of an empty conduction band.
A related paper [46] used the model described in [45] to claim that the intense illumination of light with the ITO layer causes the generation of high (5th, ...) harmonics, which are then linearly absorbed due to single photon interband transitions. Here too, the thermo-optic nonlinearity is replaced by a somewhat phenomenological polynomial expansion of instantaneous nonlinearities. Our results on the smallness of the Kerr and MPA effect and the ratio between them challenges the likelihood of high-order harmonic process becoming significant as they rely on similar nonlinear susceptibilities which should not vary by orders magnitude, and raises the need for an explicit quantification of the HHG efficiency [16]. More generally, the temporally delayed response of the effect is key to the correct interpretation of the reported results. Specifically, it is observed experimentally that the contribution ascribed to the harmonic absorption occurs at times beyond the pulse width of the pump, yet the modulation of the permittivity due to harmonic generation and its absorption is expected to reach its maximum near the peak of the pump pulse. The delay in the observed feature suggests an effect with a longer rise time / cumulative effect, rather than harmonic absorption.
Finally, both these studies did not account for the nonlinear current nor to the heating associated with it, and hence, not to the mutual cancellations.
Conclusions
In this work, we advanced our understanding of the nonlinear optical response of TCOs to short and intense pulse illumination. Our work relies on the success of thermal models to reproduce the dynamics observed from the more complex non-thermal models. First, we provided a simplification of the electronic models for the dependence of the TCO permittivity on the electron and phonon temperatures developed in [10, 35, 38, 43]. This simplification enables easy implementation in future studies. Second, unlike these previous works, we implement the model for a layer geometry which supports nonuniform fields and temperatures.
Our work demonstrated the dominance of the intraband (thermal) nonlinearity of TCOs over the contribution of valence to conduction band virtual transitions (i.e., Kerr nonlinearity) and real transitions (due to multi photon absorption); the former give rise to a small (3rd order) instantaneous permittivity change, and the latter give rise to an even smaller (5th order) accumulation of a permittivity change that rises as rapidly as the pulse duration (hence, the intraband nonlinearity), but decays on a near-picosecond time scale. These findings are robust to variations of the optical and thermo-physical (heat capacities, coupling etc.) parameters, see discussion in Appendix § 5.§ 2. Thus, it seems that these effects cannot explain the rapid decrease of transmissivity observed in [41].
This finding has two main consequences. First, it shows that realization of temporal photonic crystals is unlikely in ITO (and TCOs in general) because of the dominant (and cumulative nature of the) thermo-optic nonlinearity - surpassing the damage threshold is avoidable only for very low repetition rates.
Second, this demonstrates the quantitative unlikelihood of several mechanistic explanations proposed for the fast spectroscopic features observed in [41] and recent works that associated dynamic changes measured from TCO layers to the interband effects.
Additional nonlinear effects associated with interband transitions (such as the AC Stark effect [74], depletion of the valence band population, Fermi smearing and band shifting [75], a change of the interband collision rate etc.) affect primarily the UV spectral range, i.e., close to the interband absorption edge. While the current work does not rule out the importance of these effects, they are not expected to be significant at the near IR frequencies associated with the ENZ point and their inclusion may be at the focus of future efforts. In this vein, one should emphasize that our approach for the interband nonlinearities should be made more accurate for higher frequencies, for the high harmonics generated in the system, and especially for pump frequencies reaching the interband absorption threshold [15]. For example, one should then adopt a full Lorentzian approach that is more suitable to treat short pulses, chirp effects, and resonant enhancement of the nonlinear response; this was doe in [66, 45, 46].
Our model opens the way to the consideration of other explanations of the yet unexplained fast features observed in [41], in particular, in more extreme conditions of pulse energies, as those studied in [64], both of which require a more advanced model that captures the very high temperatures developing in these systems. This advance will be needed in order to test the claim in [64], the fast feature seen in experiments was ascribed to impact ionization (aka carrier multiplication); a simple numerical implementation showed a good match to the experimental data. However, the inverse effect which was not accounted for, sometimes called Auger recombination can be easily shown to be of identical probability, if the electron system is in thermal equilibrium. While earlier work relied upon in the current study showed that this is a good approximation, those studies did not include the dynamics of the valence band population. Thus, the question of which of the two effects (sometimes referred to jointly as Auger scattering) dominate would require adding the valence band electrons to the non-thermal model [38]. While the current work provided a step forward in that direction, this task still lies beyond the current scope .
Determination of material parameters
Linear room temperature response
The linear properties of the ITO film were determined through fitting of the linear permittivity provided by [41]. Due to the narrow spectral range provided, data for ITO films available to the authors with a comparable permittivity crossover and loss were used to solidify parameters. Since we only require the long-wavelength response of the ITO film (beyond 1100nm), the interband transitions can safely be ignored and lumped into a single within the Drude model
Here, is the electron charge, is the vacuum permittivity, is the electron mass appearing in the Kane formula for the non-parabolic band structure.
Within this limit, we fit the standard Drude model to the ITO permittivity of [41] and find = 3.99, , . These values agree with ellipsometry results for similar ITO films. With the Drude properties known, we can then leverage the Kane-band model of the conduction band energy profile to uniquely determine the Fermi-level of the material for a given scattering rate. Taking:
with and , we gradually increase the Fermi energy and evaluate the carrier density and average effective mass using the definition from [76], comparing the resistivity and plasma frequency to the value extracted from the linear permittivity fit. The result identifies , and . The final value of resistivity matches that expected from the Drude-model fit of the permittivity, and the generated permittivity subsequently provides a good fit to the linear permittivity dispersion of the sample provided by [41].
Nonlinear high temperature response
In order to feed the simulations, we would now like to derive an analytic form for the thermo-optic response of the ITO as a function of the extracted values for the non-parabolicity and the Fermi energy , in a manner that reconciles the intraband Lindhard formula [35 Eq. (9)]6
with the Drude formula (9). Here,
is the density of electron states,
is a correction factor associated with the non-parabolicity, and
is the electron collision rate contribution to the permittivity. Here, we use the (standard) energy-dependent (Fermi liquid based) expression for collision rate, with given by [10 Eq. (19)] and being the Umklapp branching ratio; below, it is set to unity since its value is anyhow embedded in . For simplicity, we neglect the seemingly opposing negative correction obtained in [43], and the positive correction obtained in [10 Eq. (18)] due to the non-parabolicity. As for noble metals, the dependence on the phonon temperature is assumed linear due to its relatively modest and slow increase, but this dependence can be generalized if necessary.
Using the expression for in [10 Eq. (24)] and neglecting the deviation from thermal equilibrium, we insert the Fermi function
into the Lindhard integral (11). In the small-damping regime, , the integral term in this equation, responsible for the free electron response, is approximated by
Here, we defined the frequency-independent quantities
Matching to the small damping limit of the Drude formula (9), allows identifying the effective plasma frequency and collision rate as
These parameters are plotted in Figure 6(a) as a function of the electron temperature, using their room temperature values extracted in Appendix § 5.§ 1.

Figure 6. (a) Electron temperature dependence of the effective plasma frequency (blue) and collision rate (red) computed via Eqs. (16) (solid lines) or as in [77] (dashed lines). (b) Electron temperature dependence of the real (blue) and imaginary (red) parts of the Drude permittivity computed via the exact Lindhard integral (Eq. (11)) (dots) or via the approximations based on Eqs. (16) (solid lines) and based on the parameters in [77] (dashed lines).
Note that the temperature dependence of the (effective) plasma frequency is equivalent to a temperature-dependent effective mass used in some previous work [28, 1]. Also note the factor 2 that originates from the approximate denominator appearing in the (energy-space) Lindhard form (11).
Now, we use the values of , and extracted from the measurements at room temperature (see Appendix § 5.§ 1) to determine consistently the values for , , and .
In principle, one can now proceed by applying the Sommerfeld expansion to and . However, this expansion is divergent, so that it is bound to fail above a threshold electron temperature (for which is no longer smaller than ); in particular, since here is relatively low compared to a noble metal, the electron temperature rise is significant, and monotonically decreases with [35], the range of validity of the Sommerfeld expansion is only –K. Instead, we compute and numerically, and then interpolate the resulting expressions and feed them to the FDTD code. This approach would also eventually fail once becomes comparable to , as expected to happen at very high temperatures [77].
Figure 6(b) demonstrates the good match between the approximated calculations (Eqs. (16)) and the complete Lindhard expression (11). However, it also shows that the resulting expression for (, hence,) there is far greater sensitivity to the electron temperature compared to the calculation in [77]. The reason for that is the different proportions given to the and contributions to (13). This stems from the unknown Umklapp weight for TCOs, and the difficulty to separate these contributions to the permittivity. This unavoidable arbitrariness has a quantitative effect on the strength of the thermo-optic nonlinearity.
Estimate of magnitude of interband nonlinear coefficients
Estimate of the Kerr coefficient
There are several theoretical approaches to estimate the value of the Kerr coefficient. The bound electronic effect far from resonance can be derived from quantum mechanical perturbation theory [5, 1, 44] or through the well developed scaling theory for the nonlinear refractive index of semiconductors [78]. Here, we presume that the virtual transitions of valence band electrons in ITO are similar to those that occur in the undoped InO material at the same frequency. We consider the role of doping upon the interband process through its shift of the Fermi-level, but consider the electrons as heavy so that their contribution to the linear permittivity in the near-infrared is negligible. In this approximation, we leverage experimental data of lightly doped ZnO and ZnS [2, 79] - wide bandgap semiconductors with similar optical and electrical properties. In the wavelength range of 750nm to 950nm (comparable to that of the pump wavelength in our case), the average value of the intensity dependent refractive index is . Assuming negligible two-photon absorption, reasonable given that , this corresponds to value of [80]. We note that this value likely represents an overestimate of the for ITO as the bandgap of heavily doped ITO is approximately 1eV larger than undoped ZnO and ZnS, and the probe wavelength of 1225nm used in [41] is longer than the dataset - both of which are expected to result in a reduced value of intensity dependent index. Despite this, in subsequent calculations we exaggerate further the bound electron nonlinearity by 10-fold and take . This places its strength on par with the thermo-optic nonlinearity (see Appendix § 7) to more readily test its role in generating a fast response.
The above discussion involved valence to conduction band virtual transitions. In [44], Narimanov considered inter-conduction band virtual transitions. While such virtual transitions between occupied and unoccupied conduction-band states may contribute to the nonlinear optical response, their impact is expected to be limited in degenerately doped TCOs such as ITO. Away from conduction-band critical points, the relevant conduction bands are highly dispersive, resulting in a relatively small joint density of states (JDOS) and a broad distribution of transition energies. Consequently, any enhancement associated with inter-conduction band transitions is expected to be spectrally diffuse rather than resonant. Although localized regions of large JDOS could produce stronger effects, such features are not expected for the nearly linear conduction-band dispersion near the Fermi level in ITO. We therefore expect inter conduction band processes to provide, at most, a secondary contribution relative to other established nonlinear mechanisms.
Estimate of the 3PA coefficient
The value of the 3PA coefficient of ITO was estimated by averaging experimental measurements of from similar wide-bandgap semiconductors (ZnO, ZnS, [2, 79]) under similar experimental parameters, as well as by fitting this data with the multi-photon absorption model by Brandi and de Arajo [53] and then adjusting for the excitation parameters and bandgap of the ITO film. Averaging the measured 3PA coefficients near 800nm, we find ( [81], assuming is negligible) valid under an excitation regime where . However, for ITO with a bandgap of 4.5eV excited at 800nm, the band-edge normalized excitation energy is 0.34, which is significantly smaller than the experimental measurements of ZnO and ZnS. Due to the strong dependence upon the pump photon energy in this energy region, we estimate the scaled value for ITO to be () [53]. The average of the measured ZnO and ZnS data therefore represents a high-end value of 3PA while the scaled value represents a more realistic estimate of the expected 3PA the ITO case. Despite this, we employ only the former (high end) values, to account for the broad spectral bandwidth of the pump.
Connection to Fast Free Electron Nonlinearity
Here, we would like to provide a note on the connection between previous descriptions of the fast free electron nonlinearity [1] and the absorptive/thermo-optic nonlinearity included in § 2. From the perspective of [1], the absorptive/thermo-optic nonlinearity contains two simultaneous contributions - a slow nonlinearity that arises from free-carrier absorption of the pump energy, and a fast nonlinearity that arises from induced currents within the non-parabolic conduction band [82, 1]. The former has been called the intraband nonlinearity [15], and is hallmarked by the fact that the resulting modulation lasts for times beyond the pump temporal width, typically with relaxation times on the scale of 100’s fs. The latter is a near instantaneous nonlinearity of free-electron motion that exists only during the pump pulse. Although this distinction is useful to clarify terminology, in the approach outlined here, both contributions are captured by solely determining the induced current density through the analytical description of the momentum/energy-dependent effective mass.
A final point of clarification is that while the fast free-carrier nonlinearity is nearly instant, and also results in permittivity modulation, it is distinct from the bound electron Kerr effect and they should not be confused. These two effects can have different magnitudes and dependence upon incident intensity. As an example, the effective third-order susceptibility of the fast free-carrier nonlinearity in ITO was predicted to be [1], while the bound electron Kerr process herein is estimated to be .
References
- [1]J. B. Khurgin, M. Clerici, and N. Kinsey. Fast and slow nonlinearities in epsilon-near-zero materials. Laser Photon. Rev., 15:2000291, 2021.DOI
- [2]N. Vermeulen, D. Espinosa, A. Ball, J. Ballato, P. Boucaud, G. Boudebs, C. L. A. V. Campos, P. Dragic, A. S. L. Gomes, M. J. Huttunen, N. Kinsey, R. Mildren, D. Neshev, L. A .Padilha, M. Pu abd R. Segundo, E. Tokunaga, D. Turchinovich, J. Yan, K. Yvind, K. Dolgaleva, and E. W. Van Stryland. Post-2000 nonlinear optical materials and measurements: data tables and best practices. J. Phys. Photonics, 5:035001, 2023.
- [3]J. Khurgin. Nonlinear optics from the viewpoint of interaction time. Nature Photonics, 17:545–551, 2023.DOI
- [4]D. Fomra, A. Ball, S. Saha, J. Wu, Md. Sojib, A. Agrawal, H. J. Lezec, and Nathaniel Kinsey. Nonlinear optics at epsilon near zero: From origins to new materials. Appl. Phys. Rev., 11:011317, 2024.DOI
- [5]R. W. Boyd. Nonlinear Optics. Academic Press, San Diego, 2nd edition, 2003.DOI
- [6]G. Fibich. The nonlinear Schrödinger equation. Springer, 2015.
- [7]N. Kinsey, C. DeVault, J. Kim, M. Ferrera, V. M. Shalaev, and A. Boltasseva. Epsilon-near-zero al-doped zno for ultrafast switching at telecom wavelengths. Optica, 2:616–622, 2015.DOI
- [8]M. Z. Alam, I. De Leon, and R. W. Boyd. Large optical nonlinearity of indium tin oxide in its epsilon-near-zero region. Science, 352:795–797, 2016.DOI
- [9]L. Caspani, R. P. M. Kaipurath, M. Clerici, M. Ferrera, T. Roger, J. Kim, N. Kinsey, M. Pietrzyk, A. Di Falco, V. M. Shalaev, A. Boltasseva, and D. Faccio. Enhanced nonlinear refractive index in ϵ-near-zero materials. Phys. Rev. Lett., 116:233901, 2016.DOI
- [10]S. Sarkar, I. W. Un, and Y. Sivan. The electronic and thermal response of low electron density drude materials to ultrafast optical illumination. Phys. Rev. Applied, 19:014005, 2023.DOI
- [11]R. M. Córdova-Castro, M. Casavola, M. Van Schilfgaarde, A. V. Krasavin, M. A. Green, D. Richards, and A. V. Zayats. Anisotropic plasmonic cus nanocrystals as a natural electronic material with hyperbolic optical dispersion. ACS Nano, 13:6550–6560, 2019.
- [12]J. B. Khurgin, G. Sun, W. T. Chen, W.-Y. Tsai, and D. P. Tsai. Ultrafast thermal nonlinearity. Sci. Rep., 5:17899, 2015.
- [13]I. Gurwich and Y. Sivan. A metal nanosphere under intense continuous wave illumination - a unique case of non-perturbative nonlinear nanophotonics. Phys. Rev. E, 96:012212, 2017.DOI
- [14]A. Capretti, Y. Wang, N. Engheta, and L. Dal Negro. Enhanced third-harmonic generation in si-compatible epsilon-near-zero indium tin oxide nanolayers. Opt. Lett., 40:1500–1503, 2015.DOI
- [15]M. Clerici, N. Kinsey, C. DeVault, J. Kim, E. G. Carnemolla, L. Caspani, A. Shaltout, D. Faccio, V. Shalaev, A. Boltasseva, and M. Ferrera. Controlling hybrid nonlinearities in transparent conducting oxides via two-colour excitation. Nat. Commun., 8:1–7, 2017.DOI
- [16]Y. Yang, J. Lu, A. Manjavacas, T. S. Luk, H. Liu, K. Kelley, J.-P. Maria, E. L. Runnerstrom, M. B. Sinclair, S. Ghimire, and I. Brener. High-harmonic generation from an epsilon-near-zero material. Nat. Phys., 15:1022–1026, 2019.DOI
- [17]V. Bruno, C. DeVault, S. Vezzoli, Z. Kudyshev, T. Huq, S. Mignuzzi, A. Jacassi, S. Saha, Y. D. Shah, S.A. Maier, D.R.S. Cumming, A. Boltasseva, M. Ferrera, M. Clerici, D. Faccio, R. Sapienza, and V.M. Shalaev. Negative refraction in time-varying strongly coupled plasmonic-antenna-epsilon-near-zero systems. Phys. Rev. Lett., 124:043902, 2020.DOI
- [18]J. Bohn, T. S. Luk, C. Tollerton, S. Hutchings, I. Brener, S. Horsley, W. L. Barnes, and Euan Hendry. All-optical switching of an epsilon-near-zero plasmon resonance in indium tin oxide. Nat. Commun., 12:1017, 2021.DOI
- [19]E. Minerbi, S. Sideris, J. B. Khurgin, and T. Ellenbogen. The role of epsilon near zero and hot electrons in enhanced dynamic THz emission from nonlinear metasurfaces. Nano Lett., 22(15):6194–6199, 2022.DOI
- [20]R. Tirole, E. Galiffi, J. Dranczewski, T. Attavar, B. Tilmann, Y.-T. Wang, P. A. Huidobro, A. Alú, J. B. Pendry, S. A. Maier, S. Vezzoli, and R. Sapienza. Saturable time-varying mirror based on an epsilon-near-zero material. Phys. Rev. Applied, 18:054067, 2022.DOI
- [21]W. Jafray, F. Belli, E. G. Carnemolla, C. Dobas, M. Mackenzie, J. Travers, A. K. Kar, M. Clerici, C. DeVault, V. M. Shalaev, A. Boltasseva, and M. Ferrera. Near-zero-index ultra-fast pulse characterization. Nature Communications, 13:3536, 2022.
- [22]O. Reshef, E. Giese, M. Z. Alam, I. De Leon, J. Upham, and R. W. Boyd. Beyond the perturbative description of the nonlinear optical response of low-index materials. Opt. Lett., 42:3225–3228, 2017.DOI
- [23]N. Kinsey, C. DeVault, A. Boltasseva, and V. M. Shalaev. Near-zero-index materials for photonics. Nat. Rev. Mater., 4:742–760, 2019.DOI
- [24]R. W. Boyd. Material slow light and structural slow light: similarities and differences for nonlinear optics. Journal of the Optical Society of America B, 28:A38–A44, 2011.
- [25]J. B. Khurgin. Slow light in various media: a tutorial. Advances in Optics and Photonics, 2:287–318, 2010.DOI
- [26]Evan O. Kane. Band structure of indium antimonide. J. Phys. Chem. Solids, 1(4):249–261, 1957.DOI
- [27]M. Z. Alam, S. A. Schulz, J. Upham, I. De Leon, and R. W. Boyd. Large optical nonlinearity of nanoantennas coupled to an epsilon-near-zero material. Nat. Photonics, 12:79–83, 2018.DOI
- [28]R. Secondo, J. Khurgin, and N. Kinsey. Absorptive loss and band non-parabolicity as a physical origin of large nonlinearity in epsilon-near-zero materials. Opt. Mater. Express, 10:1545–1560, 2020.arxiv.org/abs/2004.00134
- [29]S. I. Anisimov, B. L. Kapeilovich, and T. I. Perelman. Electron emission from metal surfaces exposed to ultrashort laser pulses. Sov. Phys. JETP, 39:375–378, 1974.
- [30]C. K. Sun, F. Vallée, L. H. Acioli, E. P. Ippen, and J. G. Fujimoto. Femtosecond tunable measurement of electron thermalization in gold. Phys. Rev. B, 50:15337–15448, 1994.pubmed.ncbi.nlm.nih.gov/9975886
- [31]N. Del Fatti, C. Voisin, M. Achermann, S. Tzortzakis, D. Christofilos, and F. Vallée. Nonequilibrium electron dynamics in noble metals. Phys. Rev. B, 61:16956–16966, 2000.DOI
- [32]T. Stoll, P. Maioli, A. Crut, N. Del Fatti, and F. Vallée. Advances in femto-nano-optics: ultrafast nonlinearity of metal nanoparticles. Eur. Phys. J. B, 87:260, 2014.DOI
- [33]J. Baxter, A. Pérez-Casanova, L. Cortes-Herrera, A. Calá Lesina, I. De Leon, and L. Ramunno. Dynamic nanophotonics in epsilon-near-zero conductive oxide films and metasurfaces: A quantitative, nonlinear, computational model. Adv. Photonics Res., 4:2200280, 2023.
- [34]N. W. Ashcroft and N. D. Mermin. Solid State Physics. Hartcourt College, Fort Worth, 1976.
- [35]I. W. Un, S. Sarkar, and Y. Sivan. An electronic-based model of the optical nonlinearity of low electron density Drude materials. Phys. Rev. Applied, 19:044043, 2023.DOI
- [36]E. Carpene. Ultrafast laser irradiation of metals: Beyond the two-temperature model. Phys. Rev. B, 74:024301, 2006.DOI
- [37]Y. Sivan and M. Spector. Ultrafast dynamics of optically heat gratings in metals. ACS Photonics, 7:1271–1279, 2020.DOI
- [38]I. W. Un, S. Sarkar, and Y. Sivan. Quantum-optical theory of the few-femtosecond nonlinear optical response of drude metals with a nonparabolic conduction band. Phys. Rev. B, 113:165411, 2026.DOI
- [39]D. Marom, D. Panasenko, R. Rokitski, P.-C. Sun, and Y. Fainman. Time-reversal of ultrafast waveforms by wave mixing of spectrally decomposed waves. Opt. Lett., 25:132, 2000.pubmed.ncbi.nlm.nih.gov/18059806
- [40]O. Kuzucu, Y. Okawachi, R. Salem, M.A. Foster, A.C. Turner-Foster, M. Lipson, and A.L. Gaeta. Spectral phase conjugation via temporal imaging. Opt. Exp., 17:20605, 2009.DOI
- [41]E. Lustig, O. Segal, S. Saha, E. Bordo, S. N. Chowdhury, A. Fleischer Y. Sharabi, A. Boltasseva, O. Cohen, V. M. Shalaev, and M. Segev. Time-refraction optics with single cycle modulation. Nanophotonics, 12:2221–2230, 2023.DOI
- [42]M. G. Ozlu, C. Fruhling, I. Hoffman, J.-I. Choi, M. Ferrera, A. Boltasseva, and V. M. Shalaev. Nonlinearity reversal in epsilon-near-zero indium tin oxide driven by few-cycle light pulse. ArXiv, 2026.arxiv.org/abs/2606.11567
- [43]I. W. Un, S. Sarkar, and Y. Sivan. Origin, strength, and speed of the nonlinear optical response of transparent conducting oxides to single-cycle optical pulses. Nanophotonics, 15:70272, 2026.arxiv.org/abs/2608.11756
- [44]E. Narimanov. Ultrafast optical modulation by virtual interband transitions. ACS Photonics, 12:402–408, 2025.
- [45]M. Scalora, M. A. Vincenti, D. de Ceglia, N. Akozbek, M. Ferrera, C. Rizza, A. Alù, N. Litchinitser, C. Cojocaru, and J. Trull. Extreme electrodynamics in time-varying media. Phys. Rev. A, 112:013502, 2025.arxiv.org/abs/2502.11837
- [46]W. Jaffray, S. Stengel, A. Boltasseva, V. M. Shalaev, C. Rizza, D. de Ceglia, M. A. Vincenti, M. Scalora, M. Clerici, and M. Ferrera. Single-pump hybrid nonlinearities in transparent conductors. ArXiv, 2025.arxiv.org/abs/2605.22246
- [47]E. Narimanov and B. Shapiro. Carrier drift modulation and the hyperbolic time crystals. ArXiv, https://arxiv.org/abs/2601.00547, 2026.
- [48]I. Hamberg and C. G. Granqvist. Evaporated sn-doped in2o3 films: Basic optical properties and applications to energy-efficient windows. Journal of Applied Physics, 60(11):R123–R159, 1986.DOI
- [49]H. Kim, C. M. Gilmore, A. Piqué, J. S. Horwitz, H. Mattoussi, H. Murata, J. H. Kafafi, and D. B. Chrisey. Electrical, optical, and structural properties of indium-tin-oxide thin films for organic light-emitting devices. Journal of Applied Physics, 86:6451–6461, 1999.DOI
- [50]Iftikhar M. Ali and M. A. Al-Jenabi. Structural and optical properties of in2o3 and indium tin oxide thin films. Journal of University of Anbar for Pure Science, 11(1):1991–8941, 2017.
- [51]E. Klein, K. Hubert, O. Paul, and P. Ruther. Low-temperature plasma annealing of sputtered indium tin oxide for transparent and conductive thin-films on glass and polymer substrates. Thin Solid Films, 693, 2020.DOI
- [52]N. Kinsey, D. Fomra, A. Ball, H. Lezec, I.W. Un, Y. Sivan, and D. O. Demchenko. Interband transitions in degenerately doped indium-tin-oxide. in preparation.
- [53]H. S. Brandi and C. B. de Araujo. Multiphoton absorption coefficients in solids: a universal curve. Journal of Physics C: Solid State Physics, 16:5929–5936, 1983.DOI
- [54]A. C. Harwood, S. Z. J. Lim, T. V. Raziman, Y. Li, J. Stones, J. W. G. Tisch, S. A. R. Horsley, S. Vezzoli, J. B. Pendry, and R. Sapienza. Intraband and interband competition drives ultrafast modulations of indium tin oxide. ..., 2026.
- [55]S. Hagness and A. Taflove. Computational Electrodynamics: The Finite-Difference Time-Domain Method. Artech House, 3rd edition, 2000.
- [56]N. Kinsey and J. Khurgin. Nonlinear epsilon-near-zero materials explained: opinion. Opt. Mater. Express, 9:2793, 2019.DOI
- [57]J. D. Jackson. Classical electrodynamics. Wiley & Sons, New York, NY, 3rd edition, 1998.
- [58]A. Sommer, E. M. Bothschafter, S. A. Sato, C. Jakubeit, T. Latka, O. Razskazovskaya, H. Fattahi, M. Jobst, W. Schweinberger, V. Shirvanyan, V. S. Yakovlev, R. Kienberger, K. Yabana, N. Schultze, and F. Krausz. Attosecond nonlinear polarization and light–matter energy transfer in solids. Nature, 86:534, 2016.DOI
- [59]I. W. Un and Y. Sivan. Parametric study of temperature distribution in plasmon-assisted photocatalysis. Nanoscale, 12:17821–17832, 2020.DOI
- [60]A. Block, M. Liebel, R. Yu, M. Spector, Y. Sivan, J. García de Abajo, and N. F. van Hulst. Tracking ultrafast hot-electron diffusion in space and time by ultrafast thermomodulation microscopy. Science Advances, 5:eaav8965, 2019.DOI
- [61]J. Hohlfeld, H. J. Wellershoff, S. S. Gudde, J. Conrad, U. Jähnke, and E. Mathias. Electron and lattice dynamics following optical excitation of metals. Chem. Phys., 251:237–258, 2000.DOI
- [62]H. Wang, K. Du, C. Jiang, Z. Yang, L. Ren, W. Zhang, S. J. Chua, and Ting Mei. Extended drude model for intraband-transition-induced optical nonlinearity. Phys. Rev. Applied, 11:064062, 2019.DOI
- [63]J. I. Choi, V. Mkhitaryan, C. Fruhling, J. B. Khurgin, A. V. Kildishev, V. M. Shalaev, and A. Boltasseva. Pathway to optical-cycle dynamic photonics: Extreme electron temperatures in transparent conducting oxides. ArXiv, https://arxiv.org/abs/2512.24641, 2026.
- [64]Y. Li, S. Vezzoli, T. Klee, J. J. Broughton, R. Tirole, A. C. Harwood, H. Allegre, S. A. R. Horsley, R. Sapienza, J. B. Pendry, and J. W. G. Tisch. Ultrafast switching of optical properties in a doped semiconductor by intense femtosecond laser pulses. Light: Science & Applications, 15:336, 2026.DOI
- [65]M. Scalora, M.A. Vincenti, D. de Ceglia, V. Roppo, M. Centini, N. Akozbek, and M. J. Bloemer. Second- and third-harmonic generation in metal-based structures. Phys. Rev. A, 82:043828, 2010.DOI
- [66]M. Scalora, J. Trull, D. de Ceglia, M. A. Vincenti, N. Akozbek, Z. Coppens, L. Rodríguez-Sune, and C. Cojocaru. Electrodynamics of conductive oxides: Intensity-dependent anisotropy, reconstruction of the effective dielectric constant, and harmonic generation. Phys. Rev. A, 101:053828, 2020.DOI
- [67]Q. Lin, O. J. Painter, and G. P. Agrawal. Nonlinear optical phenomena in silicon waveguides: modeling and applications. Opt. Exp., 15:16604, 2007.pubmed.ncbi.nlm.nih.gov/19550949
- [68]Y. Sivan, G. Cistitis, E. Yuce, and A.P. Mosk. Femtosecond-scale switching based on excited free-carriers. Opt. Exp., 23:16416–16428, 2015.DOI
- [69]C. Ciracı, E. Poutrina, M. Scalora, and D. R. Smith. Origin of second-harmonic generation enhancement in optical split-ring resonators. Phys. Rev. B, 85:201403(R), 2012.DOI
- [70]E. Drobnyh and M. Sukharev. Plasmon enhanced second harmonic generation by periodic arrays of triangular nanoholes coupled to quantum emitters. The Journal of Chemical Physics, 152, 2025.
- [71]F. Thouin, D. M. Myers, A. Patri, B. Baloukas, L. Martinu, A. I. Fernández-Domínguez, and S. Kéna-Cohen. Field enhancement and nonlocal effects in epsilon-near-zero photonic gap antennas. ACS Nano, 9:7996 – 8004, 2025.arxiv.org/abs/2502.07801
- [72]Sepehr Benis, Natalia Munera, Sanaz Faryadras, Eric W. Van Stryland, and David J. Hagan. Extremely large nondegenerate nonlinear index and phase shift in epsilon-near-zero materials. Optical Materials Express, 12(10):3856–3871, 2022.
- [73]S. Gurung, S. Bej, Q. Dang, A. Sahoo, A. Anopchenko, Z. Yi, A. V. Sokolov, A. Marini, and H. W. H. Lee. Control of ultrafast hot electron dynamics in epsilon-near-zero conductive oxide thin films. Science Advances, 11:eadu8850, 2025.DOI
- [74]S. H. Autler and C. H. Townes. Stark effect in rapidly varying fields. Phys. Rev., 100:703, 1955.DOI
- [75]P. Winsemius, M. Guerrisi, and R. Rosei. Splitting of the interband absorption edge in Au:temperature dependence. Phys. Rev. B, 12:4570–4572, 1975.DOI
- [76]D. Fomra, K. Ding, V. Avrutin, U. Özgür, and N. Kinsey. Al:Zno as a platform for near-zero-index photonics: enhancing the doping efficiency of atomic layer deposition. Opt. Mater. Express, 10:3060, 2020.DOI
- [77]I. W. Un, N. Harcavi, and Y. Sivan. Optical nonlinearity of transparent conducting oxides - more metallic than realized. ACS Photonics, 12:1718–1721, 2025.DOI
- [78]M. Sheik-Bahae, D.C. Hutchings, D.J. Hagan, and E.W. Van Stryland. Dispersion of bound electronic nonlinear refraction in solids. IEEE J. Quantum Electron., 27:1296–1309, 1991.
- [79]J. He, Y. Qu, H. Li, J. Mi, and W. Ji. Three-photon absorption in zno and zns crystals. Opt. Exp., 13:9235, 2005.DOI
- [80]Raúl del Coso and Javier Solis. Relation between nonlinear refractive index and third-order susceptibility in absorbing media. Journal of the Optical Society of America B, 21(3):640–644, 2004.DOI
- [81]Richard L. Sutherland and Marcel Dekker. Handbook of Nonlinear Optics. 2nd edition, 2003.
- [82]Y. R. Shen. The Principles of Nonlinear Optics. Wiley-Interscience, 2002.