Light-driven electrodynamics and demagnetization in FenGeTe2 (n = 3, 5) thin films

Introduction

The extensive study of Fe_nGeTe₂ (n = 3, 5) (FGT) materials is driven by their two-dimensional van der Waals (VdW) nature and the high-temperature ferromagnetic properties. Van der Waals magnets may sustain novel phases of matter, like the Quantum anomalous Hall Effect^1,2,3 or the spin liquid state⁴. Moreover, their magnetic properties can be tuned through different degrees of freedom like thickness^5,6, atomic composition^7,8, or gating bias^9,10. In contrast to the firstly discovered 2D magnets like CrI₃^11,12,13 and Cr₂Ge₂Te₆¹⁴, FGT compounds possess higher critical temperatures, up to room temperature^15,16. For this reason, FGT is suitable for integration into van der Waals magnetic heterostructures, paving the way for high-temperature spintronic research and applications based on these devices^{17,18,19,20,21,22}.

FGT has a layered centrosymmetric crystal structure that depends on the number n = 3, 5 of Fe atoms in the unit cell^15,16, in which Fe and Ge sublattices are sandwiched between two Te layers along the c axis. This number is critical in determining the magnetic and electrodynamical properties of the FGT compounds. For instance, when n = 3, Fe₃GeTe₂ (F3GT) belongs to the P63/mmc space group, and the Fe I atoms arrange in a hexagonal lattice containing only Fe atoms (see Fig. 1a). Fe II and Ge atoms are instead covalently bonded in an adjacent layer^23,24. Fe₅GeTe₂ (F5GT) crystallizes in the rhombohedral (R3bar{m}) space group, with a larger unit-cell due to the inclusion of excess Fe atoms¹⁵. The additional Fe atoms strengthen the magnetization due to the creation of a 3D exchange network similar to a 3D Heisenberg model²⁵, enhancing the exchange interaction and resulting in a larger Curie temperature (T_c ~ 300 K)^15,26,27 with respect to F3GT (T_c ~ 230K). The 3D nature of intralayer magnetic interactions hinders the perpendicular magnetic anisotropy (PMA), which is absent in bulk F5GT^15,27,28, whereas F3GT exhibits a single hard magnetic phase with a near square-shaped magnetic loop, large coercivity^29,30, and a large PMA increasing monotonously with decreasing temperatures^9,19. As revealed by other studies, both materials appear to host additional magnetic phases like anti-ferromagnetism^31,32 (F3GT) and ferrimagnetism²⁸ (F5GT), and complex magnetic structures like skyrmions^33,34 for F3GT, and butterfly-shaped MR³⁵ and high-density meron textures³⁶ for F5GT.

Light-driven electrodynamics and demagnetization in FenGeTe2 (n = 3, 5) thin films — **Fig. 1: Crystal structure and Optical pump-THz probe measurements of Fe_nGeTe₂ (n = 3, 5) films.**

From an electrodynamical perspective, both systems are strongly correlated ferromagnetic metals, where quantum fluctuation effects are crucial for a correct description of the ground state³⁷. For instance, Kondo physics at low temperatures has been clearly found in F3GT and recently similar effects have been observed in F5GT³⁸. Deviations from the itinerant Stoner model are more prominent in F3GT^39,40, where local magnetic moments play a crucial role at low temperatures, establishing a Kondo-Ising scenario^41,42, mediated by an incoherent-coherent crossover (T ~100 K) where the 3d conduction electrons hybridize with the local moments^43,44. The mass increase of the conduction electrons leads to electrodynamical changes such as a strong increase in the Sommerfeld coefficient^8,37 and a transfer of optical spectral weight toward lower frequencies⁴⁵. Besides the Kondo physics, FGT also exhibits strong electron-phonon (e-ph) coupling^15,46 and large anomalous Hall effect, which are linked to the presence of topological nodal states^45,47,48 and an enhanced skew scattering⁴⁹, respectively for F3GT and F5GT.

Given the plethora of magnetic and electrodynamical features of FGT compounds, their growth and characterization as thin films are yearned for technological applications. We already showed in refs. ^22,27 that growth by molecular beam epitaxy yields high-quality F3GT and F5GT films with robust magnetism in the ultrathin limit. In this work, instead, we make use of optical pump-THz probe spectroscopy (OPTPS) to study the out-of-equilibrium dynamics of conductive electrons in F3GT and F5GT thin films, addressing the light-induced changes that appear in the material’s phase after being illuminated by a femtosecond optical pulse, on a time scale ranging from sub-ps to hundreds of picoseconds. We observe stark differences in the THz photoconductivity between the magnetic and non-magnetic phases, suggesting a strong dependence of the electronic transport on the magnetic ordering. Given the magnetic properties and differences between F3GT and F5GT, we discuss the light-induced THz response in terms of electrodynamical and demagnetization effects. Moreover, the time variation of the photoconductivity and its dependence on the pump properties, like fluence and polarization, shine a light on the mechanisms that describe the magnetization evolution in FGT, as induced by Stoner excitations, Elliot-Yafet spin flips, or magnon formation, and their correlation to the THz conductivity through electron-spin interactions.

Experimental results and discussion

Optical properties

FGT films were prepared by molecular beam epitaxy (MBE) in an ultra-high vacuum deposition system²⁷. Four films were prepared onto 500 μm thick Al₂O₃ (0001) substrates (see methods): two F3GT films of thickness 15 monolayers (ML) and 3 ML (1 ML ~ 0.81 nm), and two F5GT films of thickness 15 ML and 3 ML (1 ML ~0.97 nm). Magnetic characterization of the films with magneto-optical Kerr effect is provided in the Supplementary Information (SI). The samples’ optical properties were initially measured through the study of their optical reflectance (R) across a broad spectral range, from THz (40 cm⁻¹) to visible (25,000 cm⁻¹)(see methods). The R data are shown in Supplementary Fig. 1a. To analyze the behavior of the FGT films, it is necessary to consider the optical response of the substrate. To do so, the Al₂O₃ substrate data were analyzed through a Kramers-Kronig (KK) constrained fitting to extract its complex optical refraction index⁵⁰ (the resulting optical indices for Al₂O₃ are shown in Supplementary Fig. 1b). The optical data are then modeled in terms of an air/film/substrate trilayer⁵¹ and fitted through the RefFit program⁵⁰ (more information can be found in the section “Multi-layer Stacking Model” of the SI). The real part of the optical conductivity is shown in Fig. 1b (the imaginary part in Supplementary Fig. 1d). The conductivity data suggests a metallic behavior for the FGT films, as already observed in F3GT single crystals⁴⁵. The flat behavior at low frequencies can be described by a broad Drude term, whose plasma frequency is higher for the F5GT compound. In contrast to bulk F3GT, where the Drude plasma frequency is ω_p ~ 800 cm⁻¹⁴⁵, in thin films the Drude contribution extends toward higher frequencies, smoothly mixing with the activation of interband transitions (see Supplementary Table 1 in the SI). The DC conductivity extracted from the optical data is comparable to the resistivity measurements obtained using a van der Pauw 4-probes geometry (see Supplementary Table 2).

Room temperature optical pump-THz probe spectroscopy

The control of films’ conductivity through light on very short time scales (below 1 ns) can cast light over the electronic transient physical processes at irradiation, namely, charge transport and the subsequent interaction of electrons with lattice and spin degrees of freedom. After a short optical pulse, electrons are excited from a valence band state to a conduction band state, and their evolution in time can be sampled with a second probe pulse in the THz range.

To study the temporal dynamics of the FGT charge carriers in a sub-ps time scale, an ultrashort optical pulse (30 fs) with wavelength centered at 800 nm has been used. Upon changing the pumping wavelength, from 400 nm to 10 μm, no modification to the out-of-equilibrium evolution of the FGT films was observed. Therefore, only the 800 nm data are reported in this work to avoid redundancy. Fig. 1c represents the optical setup used for the pump-probe experiment. Here, two delay lines are used to tune respectively the time delay between the pump and THz signal on the sample, and the THz and probe signal on an electro-optical detector (see methods). After the pump arrives, the photoinduced current is probed by a THz pulse in a transmission configuration. More specifically, the transmitted THz electric field change ΔE_t = E_t − E₀ is measured as a function of the pump-probe delay time t between the optical and THz signal, where E_t and E₀ denote the peak value of the THz electric field at a delay time t and without the pump, respectively. In the thin film approximation, the photoinduced relative transmission change ΔE_t/E₀ is proportional to the THz negative photoconductivity (PC) −Δσ_1;t/σ_1;0, where σ₁ is the real part of the optical conductivity. No external magnetic field has been applied to the samples during the optical measurements.

Figure 1 d shows ΔE_t/E₀ = − Δσ_1;t/σ_1;0 for the four FGT films at 300 K when pumped with a 5 mJ/cm⁻² linearly polarized pulse at 800 nm. The time axis is separated into two parts to highlight the temporal behavior in a short and long time window. Different behaviors for the photoinduced response can be identified depending on the sample compound and thickness. The thinner (3 ML) F3GT and F5GT films show an almost equal trend in time, with only a change in magnitude probably due to the difference in their THz conductivity (see Fig. 1b). For the thicker films (15 ML), instead, the evolution in time shows an opposite trend: the conductivity of the F3GT film increases (decrease in THz transmission) similarly to the 3 ML films while the conductivity of the F5GT film decreases, with a very slow decay to equilibrium after an initial upward trend reaching its peak at ~20 ps.

During the pump illumination, the energy is transferred from the photons to the charge carriers (the electrons) of the system. This rapid process, where an out-of-equilibrium regime is created, is sequentially followed by a fast thermalization of the electron bath to a higher temperature through electron-electron scattering. This process may take up to 1 ps, depending on the material⁵², and is readily observed in Fig. 1d for the 3 ML films of F3GT and F5GT, and for the 15 ML film of F3GT, as a negative change of the THz transmission trace (positive change in conductivity). After this first fast process, the electron bath dissipates its energy toward the lattice, forcing a decay to equilibrium that can be as slow as a few nanoseconds, depending on the material thermal conductivity, thickness, and substrate interaction⁵³. Moreover, during the energy transfer toward the lattice, acoustic phonons are created through the electron-phonon coupling (thermoelastic effect)⁵⁴. The presence of a strain pulse of coherent acoustic phonons traveling inside the material is observed as oscillations in the pump-probe dynamics, which can take the shape of sinusoidal waves like the one observed in Fig. 1d for the 15 ML F3GT (~1 THz) after 3 ps. Similar but less intense oscillations are observed in the 3 ML films, after a fast change in the pump-probe trace at nearly 3 ps. This latter effect is not fully understood. Previous results revealed the emergence of light-driven room-temperature ferromagnetism in thin F3GT⁵⁵, induced by the enhancement of the T_c through the Stoner criterion after a redistribution of the electronic states. As will be discussed later, the induced magnetization could contribute to the THz conductivity giving rise to the feature at 3 ps (see Fig. 1d). The rise and decay processes of the out-of-equilibrium dynamics described above can be modeled with a sum of exponential terms convoluted with an activation term that describes the rising edge of the photoinduced transient. The fitting model takes the form⁵⁶

$$Delta {E}_{t}/{E}_{0}=sum _{i}{a}_{i}exp (-t/{tau }_{i})left(1-{rm{erf}}left(frac{-4t{tau }_{i}+{s}^{2}}{2sqrt{2}s{tau }_{i}}right)right)$$

(1)

where a_i are the amplitudes of the different decay components, τ_i are the respective relaxation times, ({rm{erf}}) is the error function, and s is the effective rise time. Assuming two decay components (τ_fast and τ_slow) and an additional oscillatory term (cos (omega t+varphi )), the best-fitting curve for the experimental data in Fig. 1d of the 15 ML F3GT film is shown as a pink discontinuous curve overlaid to the data. The fitting parameters are reported as an inset in the same Figure. An extended Figure, up to 30 ps, is shown in Supplementary Fig. 4b. Two scattering times (τ_fast = 1.7 ± 0.2 ps, τ_slow = 300 ± 50 ps) are ascribed to the electron and lattice temperature decays, respectively, and the rising edge s = 0.55 ± 0.13 ps to the electronic temperature increase. The extracted oscillation frequency is ω = 0.9 ± 0.1 THz.

Until now the analysis has focused on F3GT (3 and 15 ML) and F5GT (3 ML), which share the paramagnetic phase at 300 K (see Supplementary Fig. 3 for the magnetization curves). In contrast, the 15 ML F5GT film possesses a higher Curie temperature (T_c ~ 300 K) and is instead in a ferromagnetic phase. This difference in magnetic order is readily appreciated by the opposite trend in the pump-probe trace as shown in Fig. 1d. The sample conductivity does not increase after illumination, but rather decreases following a two-step process: a fast change lasting less than 1 ps, and a slow evolution to its maximum value at ~ 20 ps. This trend cannot be described by Eq. (1), but rather by a sequence of two exponential functions, usually used to describe the two-step laser-induced demagnetization processes indicative of type-II demagnetization, similar to rare-earth ferromagnets Gd and Tb^57,58, as well as the VdW magnet Cr₂Ge₂Te₆⁵⁹. By defining τ_e and τ_m as the characteristic timescales for the fast and slow processes, the best fit gives τ_e = 0.7 ± 0.1 ps, which is a typical value for the electron-phonon relaxation in ferromagnetic transition metals⁶⁰, and τ_m = 6 ± 1 ps, ascribed to the demagnetization timescale. The fitting curve is shown as a discontinuous light-blue curve overlaid to the data in Fig. 1d. The interaction of femtosecond laser pulses with magnetic thin films induces a wide range of out-of-equilibrium phenomena, leading to an ultrafast demagnetization, and subsequent magnetization dynamics. However, these effects are usually observed through magneto-optical Kerr effect (MOKE) spectroscopy, whose results have been proven to be proportional to the magnetization even outside equilibrium⁶¹. The fact that it is possible to observe the magnetization dynamics with such clarity through THz spectroscopy suggests a direct relationship between the magnetization state and the electrons’ conductivity in F5GT. This effect is even more striking when considering the absence of an external magnetic field during the experiment, meaning that it relies only on the spontaneous magnetization of the sample. The ultrafast demagnetization hypothesis can also be supported by the observation of THz generation. Demagnetization effects are usually accompanied by a THz emission process at the pump arrival due to the oscillation of magnetic dipoles⁶². Such an effect, shown in Fig. 1e in terms of the THz amplitude vs time delay, is observed at the pump arrival for F5GT when the probe THz beam is blocked.

Light-induced demagnetization effects in magnetic films are driven by two types of effects: a reduction of the magnetic moment carried by each atom, and a random tilt of the magnetic moment of each atom, resulting in a cancellation of the average magnetization⁶³. However, to date, there is no consensus regarding the main source of demagnetization, and these processes are material-dependent⁶³. The reduction of the magnetic moment is due to the creation of Stoner excitations, resulting from a decrease of the exchange splitting⁶⁴, or Elliot-Yafet spin-flips⁶⁵, while the tilt is a consequence of the creation of magnons by a direct energy transfer through the electron-spin or lattice-spin coupling. The observation of rapid (sub-picosecond) demagnetization behavior in the THz photoconductivity of F5GT suggests that the carriers play a central role in driving demagnetization, aligning with the concept of Stoner excitations. This also implies that the magnetic order itself directly influences the low-energy electrodynamics, probably through a sensitive change in the electron scattering. This process is qualitatively different from what is found in standard ferromagnetic films^63,66,67, where the demagnetization is mainly driven by the production of magnons through a thermal transfer from the excited mobile electrons (positive photoconductivity after pump). A possible direct interaction of the photon field with the spins (an “all-optical switching”⁶⁸) is rejected on the basis of its helicity dependence. Indeed, no changes in the pump-probe trace are observed when pumping the system with a circularly polarized light.

Low temperature measurements

To further prove the ultrafast demagnetization hypothesis, the pump-probe measurements were repeated at 80 K, below the F3GT Curie temperature (~230 K). The results are shown in Fig. 2a. The key difference with respect to the 300 K data is the F3GT (15 ML) response, which changes sign and shows a process similar to F5GT (15 ML) but with a magnitude nearly 5 times higher. This change appears near the Curie temperature, as can be seen in the temperature-dependent traces shown in Fig. 2b and the fixed probe delay values in Fig. 2c, suggesting a direct contribution of the magnetic order on the THz conductivity also for F3GT. The thinner films, instead, although ferromagnetic at 80 K (see Supplementary Fig. 3), keep showing the same behavior as in the paramagnetic phase. This difference can be attributed to the weaker spontaneous magnetization associated with the thinner (3 ML) films²⁷. As anticipated in the previous section, the light-induced enhancement of the T_c through the Stoner criterion can produce a net spontaneous magnetization in these films⁵⁵, affecting the low-energy electrodynamics as already proved in the 15 ML samples, and addressing the 3 ps features observed at both 300 K and 80 K.

**Fig. 2: Optical pump-THz probe measurements of Fe_nGeTe₂ (n = 3, 5) films at different temperatures and fluences.**

The pump-probe trace of F3GT (15 ML) is quantitatively and qualitatively different from the F5GT (15 ML). The latter shows only a magnitude change after decreasing the temperature, as expected for a demagnetization process in standard ferromagnets⁶⁹. The F3GT trace, instead, shows a rapid increase followed by a slower (nearly 3 ps) increase to a flat equilibrium modified by slow fluctuations like the ones observed for the 300 K trace, which have been associated with the propagation of acoustic phonon modes. The magnetic properties of F3GT are different from those of F5GT, with F3GT having a strong magnetic anisotropy along the c-axis. This suggests that spontaneous magnetization can be higher in F3GT, accounting for the increase in the magnitude of the THz conductivity change. In F3GT, magnon formation can be a driver for demagnetization. Recent pump-probe MOKE experiments of F3GT flakes⁷⁰ have shown how the demagnetization effect can be associated with an angular momentum transfer between the electron, phonon, and spin subsystems. However, the pump-probe MOKE traces differ from the THz conductivity here reported. Indeed, for the F3GT electrodynamics, additional complexity is given by the presence of a second phase transition taking place in F3GT at 100 K, which is an incoherent-coherent crossover leading to a Kondo scenario⁴² that can be observed also in the pump-probe trace (see Fig. 2c) as a modulation near 100 K. In this phase, the itinerant electrons couple to the localized magnetic moments increasing their effective mass and leading to electrodynamical changes^8,45.

Further insights on the F3GT THz photoconductivity contributions are obtained by measuring the pump-probe trace as a function of the pump fluence. Fig. 2d and e show the results for the 15 ML F3GT. As a function of fluence, the pump-probe spectrum in Fig. 2d shows a rapid transition from a peaked trace to a smooth rising one, suggesting the co-existence of both light-induced hot carriers and demagnetization effects, accounting for the trace peak and the slow rise, respectively. The data highlights a transition from a hot-carriers-dominated effect to a demagnetization-dominated effect, going from low to high fluences. This transition is highlighted in Fig. 2e, showing how the 0 ps trace (peak position in Fig. 2d) decreases at higher fluences.

A possible explanation for the F3GT trace behavior comes from the Kondo ordering and thermal energy transfer, as described by a 3-temperature model (3TM). At 80 K, F3GT conduction electrons are affected by the Kondo effect, increasing their effective mass and mobility due to the elastic scattering by a lattice of local magnetic moments⁷¹. After illumination, the electrons are driven out of this coherent state due to their rising temperature, decreasing their conductivity. This effect is readily seen in Fig. 2c, where the THz transmission increases rapidly below 100 K. Other possible explanations for the negative photoconductivity observed might come from the dynamical formation of quasiparticles like polarons. However, for FGT, no such formation can be highlighted. Analogously, Stoner excitations are excluded since the observed effect has a clear temperature activation that is far away from the Curie temperature, while instead matching the Kondo-Ising scenario (~100 K).

The hot electrons that form after illumination alone also induce a demagnetization in the magnetized layer⁶³. At high fluences, this demagnetization can be attributed to a transfer of heat from the hot electrons to the lattice, causing the system to cross the Curie temperature and exit the ferromagnetic phase. At low fluences, instead, demagnetization effects are still present as suggested by MOKE experiments⁷⁰, and can be linked to the increase in the magnon bath’s temperature, which reduces the F3GT average magnetization. Longitudinal demagnetization effects like Stoner excitations are still possible, but it has been shown that in metallic systems their effect is secondary with respect to magnons⁶³ and a dominant magnon+Elliot-Yafet demagnetization picture has been proved for F3GT⁷⁰. Given the magnon demagnetization picture, already proved for F3GT, 3TM simulations, comprising the electron, magnon, and lattice thermal baths, were performed to understand the changes in the pump-probe trace as a function of fluence, as described in the SI. The resulting temperatures are shown in Supplementary Fig. 5a–c, highlighting the thermal transfer from electrons to magnons at low fluences, and the heating above T_c of the lattice at high fluences. The higher the fluence, the faster the system exits the magnetic state. This effect explains the disappearance of the peak from the THz trace above 1.7 mJ/cm² (see Fig. 2d), being substituted by the phase change dynamics at early times. The phase change effect is also in accordance with the experimental data at longer times, since the relaxation to equilibrium becomes very slow, of the order of nanoseconds, as shown in Fig. 2a and the inset of Fig. 2e for a 5 mJ/cm² fluence.

Finally, Fig. 3a and b show the results for the 3 and 15 ML F5GT, while the time fixed-changes are shown in Fig. 3c. The F5GT samples show only quantitative changes as a function of fluence, where the response of the 15 ML film saturates already at 5 mJ/cm², while the one of the 3 ML film seems to increase linearly. The latter observation is in accordance with the increased production of free carriers after pumping, while the former shows an increase of τ_m with fluence, similar to other FGT crystals⁷⁰, where the electron-phonon scattering event leads to a spin-flip (Elliot-Yafet mechanism) due to the presence of spin-orbit coupling. Indeed, at high enough fluences, the slow dynamics for the 15 ML F5GT crystal observed both at 80 K and 300 K can also be associated with the lattice heating after illumination, whose temperature reaches the Curie temperature. This induces a phase transition to the paramagnetic state, increasing the lifetime of the out-of-equilibrium state up to the ns scale.

**Fig. 3: Optical pump-THz probe fluence-dependent measurements of Fe₅GeTe₂ films.**

Discussion

The presented experiments demonstrated that illumination of FGT films with ultrashort optical pumps causes significant changes in the THz photoconductivity related to the demagnetization effects and production of hot carriers. Depending on the thickness and chemical composition (n = 3, 5) of Fe_nGeTe₂ thin films, the pump-probe behavior changes as a result of the variance in the magnetic properties and electrodynamical ordering between the different films. In particular, light-induced demagnetization effects in Fe₅GeTe₂ are found in THz transmission measurements, which are linked to a change of the longitudinal magnetization on a sub-ps time-scale through a decrease of the exchange splitting, and followed by a slower demagnetization trend induced by Elliot-Yafet spin-flips and the ferromagnetic-to-paramagnetic phase transition at high pump fluences. Substantial differences are obtained for the pump-probe trace of Fe₃GeTe₂, as caused by the production of hot carriers and their thermal exchange to magnons and the underlying lattice. Finally, temperature-dependent measurements demonstrated the sensitivity of optical pump-THz probe measurements to the magnetic and Kondo lattice ordering in Fe₃GeTe₂. The present results concern the interaction of light with the small spontaneous magnetization in FGT films. Future experiments under an applied magnetic field could bring further insights into the mechanisms of light-induced THz conductivity changes. The dependence of the electrodynamic properties on the magnetization state observed in this work indicates FGT compounds as candidate materials for the study of proximity effects in heterostructures and the production of tunable electro-optical and spintronic devices.

Methods

Sample preparation

The FGT films were grown by MBE on Al₂O₃(0001) substrates by co-evaporation of high-purity Fe, Ge, and Te, as reported in ref. ²⁷. The composition of the films (F3GT or F5GT) was adjusted by setting the Fe:Ge flux ratio to the desired stoichiometry, while keeping an over-stoichiometric Te flux. The substrate temperature during growth was 300 °C for F3GT and 350 °C for F5GT. After the deposition, the films were annealed at 550 °C under Te flux in order to improve their crystallinity, then capped at room temperature with a 3-nm-thick Al film that was naturally oxidized in air. θ/2θ X-ray diffraction scans shown in SI confirm that both F3GT and F5GT films are single-phase, with the expected c lattice parameter.

Optical characterization

Reflectance measurements at room temperature (RT) were taken through a Vertex 70v FTIR broadband interferometer, covering the spectral range from THz (40 cm⁻¹) to mid-infrared (MIR) (7000 cm⁻¹), and through a Jasco V-770 spectrometer, extending the spectral data from 5000 cm⁻¹ up to the visible (VIS) region (20,000 cm⁻¹). Optical pump-terahertz probe measurements were developed in a transmission geometry through the generation of THz pulses from a ZnTe (110) crystal (0.5 mm thick), whose electric field is detected in a time-domain experimental configuration through an electro-optical sampling in a second ZnTe (0.3 mm thick). All measurements were performed in vacuum. The signal covers a range of frequencies going from 0.6 THz to ~2.5 THz. It is generated by a 800 nm pulse (7 mJ) with a time duration of 35 fs, as obtained by a Ti:sapphire amplifier (Coherent^ⓒ Verdi G-series), with a repetition rate of 1 kHz. The signals used to pump the films along with the THz are obtained from the 800 nm itself which can be used directly or sent to an optical parametric amplificator (OPA) (TOPASprime from Light Conversion^ⓒ) for the generation of wavelengths from 1200 nm and 2400 nm wavelengths. A consecutive difference frequency generator (DFG from Light Conversion^ⓒ) takes in input the OPA output and generates wavelengths ranging from 2600 to 16,000 nm. A lens is used to focus the pump pulse onto the sample position. A 400 nm pump has also been obtained from the second harmonic generation of the 800 nm signal into a BBO nonlinear crystal. The THz detection is achieved through an electro-optical scheme based on the Pockels effect induced by the THz radiation itself in a ZnTe crystal. The change in birefringence is probed through an 800 nm pulse overlapping in time and space with the THz pulse in the detection crystal. The polarization change is then measured via a balanced photodiode.