Terahertz Spectral Signatures of Ultrafast Spin Transports in Ferromagnetic Heusler Alloy

Although the Co‐based Heusler compounds are predicted to be half‐metals, their sub‐picosecond demagnetization dynamics upon laser excitation show a transition‐metal‐like behavior. Any possible role of ultrafast nonlocal spin transport on the ultrafast demagnetization of half‐metallic Heusler compounds has only been inferred indirectly by time‐resolved magneto‐optical Kerr effect. Here, an ultrafast optically driven spin current traveling from Co2FeSi half metal into an adjacent Pt layer by using terahertz (THz) emission time‐domain spectroscopy (TDS) is demonstrated. By varying the magnetic field and excitation symmetry, the sizable THz generation from Co2FeSi/Pt stack arises from optically induced spin transport across the Co2FeSi/Pt interface in conjunction with spin‐to‐charge current conversion, via the inverse spin Hall effect (ISHE). The chemical ordering, interface roughness and magnetization of the samples become vital to engineer the THz emission properties. The amplitude of THz emission is comparable to that of CoFeB/Pt stack, thereby indicating Co2FeSi as an efficient ultrafast light‐driven spin current injector. Finally, by varying the pumping energy, the contributions are distinguished to the ultrafast spin current from thermalization and optical excitation of majority spins in Heusler alloy. This work illustrates THz emission TDS as a powerful tool to investigate the ultrafast spintronic properties of Heusler alloy/normal‐metal bilayers.


Introduction
Half-metallic Heusler compounds are particularly promising candidates for efficiently generating pure spin currents for spintronic devices. [1,2] The halfmetallic Heusler alloys have a unique electronic structure, where the majority spin-channel has a metallic character while the minority spin-channel has a bandgap and ideally exhibit 100% spinpolarization at the Fermi level. [3,4] As such, the electrons pass within the halfmetallic Heusler alloys are polarized either "spin up" or "spin down." The combination of high Curie temperature and small magnetic damping constant makes the cobalt-based Heusler alloys excellent candidates for spintronic device applications. [5] For example, Co 2 FeSi in its L2 1 phase has been demonstrated to be a promising material for spin lightemitting diodes with more than 50% spin-injection efficiency. [6] For over two decades, ultrafast laser excitation has been used to quench the magnetic state of materials. [7,8] Understanding the mechanism governing the de-and remagnetization is important for all-optical magnetization switching, which is based on the complex energy transfer between subsystems of electron, spin, and lattice. [9][10][11] Accordingly, femtosecond (fs) laser pulses have been used to drive a diverse range of half-metals into a nonequilibrium states and monitor the response of the spin dynamics using time-resolved magneto-optical Kerr effect (TRMOKE). [12][13][14][15] The half-metals CrO 2 (P = 96%), Fe 3 O 4 (P = −80%), and LaSrMnO 3 (P = 96%) exhibit very slow demagnetization in hundreds of picoseconds. [16,17] Note that the demagnetization time ( M ) has been found to be related to the degree of spin polarization (P) by M ∝ (1 − P) −1 . [16,18] However, as opposed to CrO 2 , Fe 3 O 4 , and LaSrMnO 3 , the laser-induced demagnetizations of Co-based Heusler alloys Co 2 MnSi ( M ≈ 0.34 − 0.38 ps), Co 2 MnAl ( M ≈ 0.16 ps), Co 2 FeSi ( M ≈ 0.25 − 0.28 ps), and Co 2 FeAl ( M ≈ 0.2 ps) are much faster. [19][20][21][22] These ultrafast demagnetization time scales are similar with that of elemental 3d ferromagnets, such as Co (≈0.1 ps), Fe (≈0.10-0.30 ps), and Ni (≈0.60-0.80 ps). [23,24] The nontrivial ultrafast demagnetization in half-metals Heusler alloys was initially interpreted by the antisite Co defects enhance the phase-space for spin-flip scattering. [16] However, even for perfect half-metallicity in such Heusler compounds, the demagnetization is still in the sub-picosecond timescale, which raises the questions regarding the nature of ultrafast spin dynamics in Cobased Heusler alloys.
Steil et al. proposed a model based on hole-mediated spin-flip scattering below the Fermi energy to explain the optical-excited local demagnetization. [19,22] Recently, a near-instantaneous optically induced spin transfer effect (OISTR) [25,26] was observed from one element to another in Heusler alloys on the extremely fast (<10 fs) time scales. [27][28][29] In fact, after the photoexcitation, the laser-induced nonlocal super-diffusive of spin transports in the ferromagnetic (FM)/heavy metal (HM) nanometer multilayers, due to different scattering lengths (or times) for majority versus minority spins. [30][31][32] This ultrafast spin transport can also lead to a decrease of the magnetization. [33][34][35][36] More recently, the emission of terahertz (THz) radiation in Co 2 FeAl/Pt and Co 2 Fe 0.4 Mn 0.6 Si/HM bilayers by 800 nm fs laser pulses was interpreted by the photoexcited spin current super-diffusing from Heusler alloy to HM layer. [37,38] However, the spin transport was argued not of relevance in Co 2 FeSi and Co 2 MnSi, due to the absent of highly excited electrons in the majority density of states in Co 2 FeSi and Co 2 MnSi with the pumping photon energy of 1.55 eV. [19] Despite the increasing experimental progress, a comprehensive understanding of the laser-induced spin dynamics in Heusler alloys remains one of the most challenging problems governing potential applications of the Co-based Heusler alloys in novel opto-spintronic devices and THz emitters. First, the main experimental technique employed so far was TRMOKE in the visible range, which is known to uncover the transient demagnetization of the sample within only the probed surface region. However, almost Heusler samples investigated by TRMOKE are, in fact, capped with a HM layer, [39][40][41] which is commonly used to protect the Heusler alloys against oxidation. Thus, TRMOKE measurement lacks a convincing demonstration of ultrafast spin and charge transports after fs laser excitation, [42,43] which also cannot be obtained by conventional electric circuits with electrodes. In addition, Heusler alloy behaves differently from the elemental magnets, the majority spin channel in Co 2 FeSi is metallic, while the minority spin channel is semiconducting. [44][45][46][47][48] Therefore, another challenge is a clear experimental distinguishing between contributions to the ultrafast spin dynamics from majority and minority spins in Heusler alloy. [49] In this study, we address these challenges using ultrafast THz emission TDS, which gives direct access to the ultrafast demagnetization, [50,51] and spin-to-charge current conversion in FM based heterostructures, such as FM/NM, [52][53][54][55] FM/2D hybrid metal halides, [56] antiferromagnets/NM, [57] FM/topological insulator, [58][59][60] FM/MoS 2 [61] and semimetals/NM, [62,63] on the (sub-) picosecond timescales in a contact-free and rapid fashion. Following optical excitation of Co 2 FeSi/Pt, we observe sizable emission of broadband THz wave generation. We not only control the photoexcitation geometry and the photon polarization, but also engineer the spin polarization of Co 2 FeSi/Pt heterostructures. All experimental results demonstrate a significant photoinduced nonlocal spin-to-charge current conversion after fs photoexcitation. By changing the photonenergies (E) of pump pulses, we consequently distinguish between the laser-driven spin-transports from majority and minority spins. Our results reconcile the debate between previous experimental studies and shed light on the fundamental laser induced spin dynamics in Heusler/HM heterostructure. Furthermore, our results lay the foundation of the spintronic platform for device applications involving rapid data storage, broadband THz generation and spectroscopy. Figure 1a depicts the schematic illustration of laser-induced THz emission from Co 2 FeSi/Pt heterostructures. We excite the sample with a laser pulse (pulse duration of 120 fs, repetition rate of 1 kHz) with tunable wavelengths from 550 to 1550 nm, consisting of a Ti: sapphire regenerative amplifier operated with an optical parametric amplifier. The sample was placed in a static in-plane magnetic field of H = ± 200 mT (along ±x axis). The THz emission from the sample was measured in the farfield using free-space electro-optic sampling (see the "Experimental Section" for details). Samples used in this study were Co 2 FeSi(3)/Pt(3), which were fabricated on single-crystalline MgO (001) or SiO 2 substrates by dc magnetron sputtering at various substrate temperatures of T s . The values in parentheses are the layer thicknesses (see the "Experimental Section" for details). As shown in Figure 1b, we choose Pt as the HM layer because it is known to have a high spin-to-charge conversion efficiency. The X-ray diffraction (XRD) peaks of (002) and (004) indicate that Co 2 FeSi film exhibits a mixed L2 1 and B 2 phase (see Note S1, Supporting Information). The thickness of Co 2 FeSi layer was chosen sufficiently small to minimize the laser-induced temperature gradients. The thickness of Co 2 FeSi/Pt bilayer of 6 nm is smaller than the attenuation length of the pump pulse (≈20 nm). In addition, two reference samples CoFeB(2)/Pt (2) and Fe(2)/Pt(2) were deposited on MgO (001) substrate.

Figure 2a
shows typical THz emission waveforms from MgO/Co 2 FeSi/Pt with T s = 400°C, under opposite applied magnetic field. The wavelength of the pump beam is 800 nm and the pump fluence is 1.4 mJ cm −2 . One can see two main results from the data in Figure 2a. First, the emitted THz traces reverse the sign upon reversal of the applied magnetic field, demonstrating the magnetic nature of THz emission from Co 2 FeSi/Pt bilayer. Second, the polarity of THz signal is oppositely switched upon change the sign of excitation geometry in the laboratory frame, from pumping the Co 2 FeSi/Pt side (+n) to the MgO substrate side (−n), as the external magnetic field was kept constant. The THz emission from Co 2 FeSi/Pt is not, like 25 nm thick Co 2 MnSn capped with 2 nm MgO, [48] which is dominated by the ultrafast magnetic dipole. For magnetic dipole emission, E M THz (t) ∝ 2 M(t) t 2 , the THz waveform exhibits the same polarity for ±n configurations. Figure 2b shows the corresponding fast Fourier transform results of the temporal waveforms. The bandwidth of THz emission for −n case is ≈0.2-3.5 THz, which is broader than ≈0.2-2.5 THz for +n case. This is due to the absorption and dispersion of the THz pulses within the MgO substrate. Figure 2c shows the THz peak amplitude of odd component in the magnetization, E THz = (E +H − E −H )/2 , which is spin-dependent. It increases with increasing the pump fluences and can be fitted by a saturation formula, E THz (F p ) = B × F p /(F p + F sat ), where B characterizes the factors in terms of the THz radiation, F p is the pump fluence and F sat is the saturation fluence. The fitted saturation fluence F sat is around 3.7 mJ cm −2 . As the pump fluence is larger than 0.25 mJ cm −2 , we observe a spin-independent, even component of E THz = (E +H + E −H )/2 , which is an order of magnitude smaller than the odd component.
To further clarify the physical origin of the THz emission, we performed the THz generation from the Co 2 FeSi/Pt excited by both linear-and circular-polarized pump pulses in a reflection geometry ( Figure 2d). As shown in Figure 2e,f, it is found that both the polarization-independent and polarization-dependent THz radiations are observed for opposite orientation of the inplane magnetization. Figure 2e shows the linearly polarization induced modulation of THz peak amplitude up to ≈9.0% (+H) and ∼15.0% (−H) of the polarization-independent THz emission. Figure 2f shows the QWP angle-dependent modulation of THz peak amplitude up to ≈9.0% (+H) and ≈12.0% (−H) of the averaged THz signal. These small modulations and even component of E THz = (E +H + E −H )/2 can be largely attributed to the nonlinear optical rectification effect from both structural-and magnetization-induced centrosymmetric breaking, [64,65] which is consistent with the observation in CoFeMnSi/Pd. [65] To summarize, our observations exhibit that both magnetic-dipole radiation and nonlinear optical effect do not make a noticeable contribution to the THz emission from Co 2 FeSi/Pt bilayer. All the above experimental features confirm the spin-to-charge conversion as the dominating THz emission origin, as shown in Figure 1b.
where j c is the in-plane transient charge current, j s is the out-of-plane spin current, Pt is the spin Hall angle for Pt layer and M denotes the magnetization in the FM layer.

THz Generation Characterizations from Co 2 FeSi/Pt
To gain further insight into the behavior of THz generation from Co 2 FeSi/Pt, in the following, we make three observations. First, Figure 3a plots raw THz waveforms emitted from Co 2 FeSi/Pt/MgO, Co 2 FeSi/Pt/SiO 2 , and CoFeB/Pt/MgO. As shown in Figure 3b, the THz peak amplitude of Co 2 FeSi/Pt grown on MgO substrate is a factor of ≈1.84 larger than that fabricated on SiO 2 substrate, for identical optical excitation. From previous works, the Co 2 FeSi thin film grown on MgO substrate is more easily fabricated with fully ordered phase L2 1 . [66,67] In contrast, the Co 2 FeSi thin film fabricated on SiO 2 substrate commonly exhibits a disordered phase A2. [68] This observation is consistent with the spin light-emitting diode measurement, which shows a disorder-induced reversal of spin polarization. [69] Thus, we interpret our experimental result such that the total spin polarization includes contributions from different Co 2 FeSi phases exhibiting partial compensation each other. Note that the THz peak amplitude of Co 2 FeSi/Pt/MgO reaches 80% of CoFeB/Pt/MgO, thereby identifying half-metal Heusler as a quite efficient ultrafast light-driven spin current injector. In addition, the observed band width of the THz emission for all samples are similar, regardless of the degree of chemical ordering, which is well consistent with the ultrafast spin-to-charge current conversion within the Pt layer.
Second, Figure 3c shows the THz emissions from Co 2 FeSi/Pt/MgO grown with various substrate temperatures of T s = 400, 450, 500, 550, and 600°C. The THz waveforms have approximately the identical shape for all samples. While, the THz peak amplitude decreases with increasing of T s (red bars in Figure 3d). In addition, we measured the decay rates of photoexcited spin precession of Co 2 FeSi/Pt, using the TRMOKE (see Note S2, Supporting Information). The blue bars in Figure 3d show the T s dependent damping parameter eff , which is around 0.04 for T s ≤ 450°C eff exhibits a rising at T s > 450°C, which is not inherent to the ordered Co 2 FeSi structure, but primarily due to the extrinsic mechanism. [41] It should be mentioned that the Pt layer was prepared after the Co 2 FeSi thin layer was naturally cooled down to room temperature. Therefore, less atomic intermixing occurs during the heating process, which has minor impact on the THz emission. [70,71] The increase of eff versus T s has be attributed to the enhanced two-magnon scattering occurred at the grain-grain boundaries arising from the growth of Co 2 FeSi grains. [41,72] Although the T s dependent peak amplitude of THz emission exhibits an opposite trend of eff , they originate from the same physical interpretation. The less interface roughness can be obtained in Co 2 FeSi/Pt with lower T s . The lower the interfacial spin memory loss, the longer the momentum scattering time contributes to the increase of the effective electron propagation lengths, and gives rise to the increased THz amplitude.
Third, as the spin current depends on the magnetization of Co 2 FeSi layer, we expect a temperature dependence of the THz emission. The blue curve of Figure 3f shows the bulk magnetization of the Co 2 FeSi/Pt sample versus the ambient temperature T 0 , as measured by superconducting quantum interference device (see the Experimental Section). The magnetization increases with decreasing T 0 . We cryogenically cool the Co 2 FeSi/Pt to investigate the T 0 dependence of the THz emission signals, as shown in Figure 3e. The circles in Figure 3f display the peak values of THz emission normalized to the signal measured at 5 K (maximum THz signal) as a function of T 0 . The THz emission, and thus the ultrafast THz spin current in- crease monotonically with decreasing T 0 , with a similar trend of the magnetization of Co 2 FeSi/Pt. This similarity provides another evidence that the THz generation arises from the ultrafast spin-to-charge conversion. Note that the efficient spin-tocharge conversion by using ultrafast THz emission spectroscopy is consistent with the steady state ferromagnetic resonance measurements. [73][74][75]

Driving Force of Ultrafast Spin Dynamics
So far, the ultrafast spin transport and ultrafast demagnetization in Stoner-type ferromagnetic metal layer is believed to originate from a generalized spin voltage, F↑ − F↓ , which describes the equilibration of the chemical potential F↑ and F↓ of majority-and minority-spin electrons in the FM layer. [76,77] Note that Co 2 FeSi possesses different electronic band structures than Fe near the Fermi energy, the metal and the semiconduc-tor coexist, as shown in Figure 1c. Thus, the light-induced spin currents in Co 2 FeSi differ either from semiconductors [78][79][80] or metals. [52][53][54][55] The photonenergy dependent THz emission are further performed to identify the photo-induced thermalization and the optical transitions from majority/minority spins to the out-ofequilibrium spin current injection of Co 2 FeSi/Pt heterostructure. Figure 4a,c shows the representative THz waveforms from Co 2 FeSi/Pt and Fe/Pt measured from 0.8 to 2.48 eV, under the same experimental condition (excitation fluence is 0.5 mJ cm −2 ). Figure 4b,d shows the corresponding Fourier-transform spectra. The THz amplitude with maximum intensity at around 0.58 THz are observed for both Co 2 FeSi/Pt and Fe/Pt samples. First, our experimental results indicate that Fe/Pt delivers almost identical THz pulses with various pumping photon energies (red squares in Figure 4e). [81,82] While, for Co 2 FeSi/Pt, the THz peak amplitude keeps almost constant value within the photoenergy range from 0.8 to 1.08 eV. It is important to see that the THz peak am- plitude of Co 2 FeSi/Pt increases as the photon energies is larger than 1.08 eV.
In a semiclassical approach, for the case of metallic channel, by applying both spin voltage P ↑ and gradient of the electronic temperature A ↑ between the metallic side of Co 2 FeSi and Pt layer, a majority spin current can be modeled by where is the occupied numbers of majority (minority) spin electrons in the Co 2 FeSi layer.ñ is the occupied numbers of the Pt layer. We use "∼" to denote the quantity related to the Pt layer. There are two ways to generate P ↑ and A ↑ . One is the optical transition, the other is the thermalization induced by pump pulse. Thus, P ↑ = P ↑ op. + P ↑ th. and A ↑ = A ↑ op. + A ↑ th. . In Equation ( For the semiconductor side, the minority spin current is given by where ) As shown in Figure 4f-h, the pump pulses with various photon energy E induce different transitions in Co 2 FeSi, cross over the bandgap of semiconductor channel (E 1 ≈ 0.8 eV), and further the optical transition threshold of metallic channel (E 2 ≈ 2.2 eV). First, in the case of E < E 1 < E 2 , the semiconductor channel is not excited. The optical transition of the metal channel is also turned off. As shown in Figure 4f, ⏟⏞⏞⏞⏞⏞⏞⏞⏞ ⏟⏞⏞⏞⏞⏞⏞⏞⏞ ⏟ thermalization . The j s, th.
contributes the THz emission with pump wavelength around 1550 nm. Second, when the photon energy is larger than E 1 , but smaller than E 2 , the optical transitions of semiconductor . Because the contributions from electrons and holes are largely compensated each other, the j s, minor. of minority optical transition is negligible for THz generation, which is well consistent with our observation that THz amplitude does not change within the pump wavelength range from 1550 to 1150 nm. Third, under a higher-energy pumping, E > E 2 , a further contribution of metallic side is added within the net spin current, as shown in Figure 4h. j s = j s, th. − j s, minor. + j s, major. = ⏟⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏟ majority optical transition . Our experimental result indicates that the j s, major. of majority optical transition would additionally contribute about 27.8% to the j s , which increase the THz emission. In our analytical model, A is a higher order with respect to P in the moment expansion. Thus, the temperature gradient A ↑ th. is less significant than the spin voltage P ↑ th. (see Note S3, Supporting Information). We conclude that the generation of spin current under laser pumping is mainly attributed to the thermalization of majority spin driven by spin voltage. Furthermore, in contrast to the elemental ferromagnets, the optical transitions from majority-spins contributes more significant than the minority-spins, to the ultrafast spin current.

Conclusion
To summarize, we have experimentally demonstrated that the signature of ultrashort laser driven spin-dependent transport in Heusler Co 2 FeSi interfaced with a metallic Pt layer is imprinted on the THz emission. Our measurements reconcile the debate between previous experimental studies on ultrafast demagnetization in Heusler compounds. THz emission spectroscopy reveals a sizable ultrafast spin injection in the Co 2 FeSi/Pt with a spin-to-charge-current conversion efficiency of the same order as the CoFeB/Pt reference system. Furthermore, we reveal that THz spin currents in Co 2 FeSi/Pt can be engineered by the chemical ordering, interface roughness, magnetization, and photoenergy. Our results will allow for not only a better understanding of the fundamental spin dynamics to ultrafast optical stimuli in Heusler bilayer, but also a promising THz spintronic emitter.

Experimental Section
Sample Fabrication: Co 2 FeSi(3)/Pt(3) bilayers were fabricated on single-crystalline MgO (001) or SiO 2 substrates by dc magnetron sputtering under a base pressure better than 3.0 × 10 −8 Torr. The deposition rate was fixed as 0.10 Å s −1 for the Co 2 FeSi layer and 0.52 Å s −1 for the Pt layer, respectively. The Co 2 FeSi layer was sputtered from an alloy target with a stoichiometric composition of Co 50 Fe 25 Si 25 under an Ar pressure of 4 mTorr. The Co 2 FeSi/Pt samples were deposited at various substrate temperatures of T s = 400, 450, 500, 550, and 600°C, which were further in situ annealed at the same temperature for 10 min prior to the Pt layer deposition. And then, the Pt layer was sputtered after the Co 2 FeSi layer was naturally cooled to room temperature. The optical absorptance spec-trum of the sample was measured in Vis-NIR spectral range (Shimadzu, UV3600) (see Note S4, Supporting Information).
THz Emission Set-Up: The fundamental laser pulse with wavelength at 800 nm is generated by a Ti:sapphire amplifier. The pulse repetition rate was 1 kHz. The pulse width was 120 fs. The fundamental laser pulse is divided into two beams, one is wavelength-tuned through OPA as pump light, and the other beam is used as probe beam to detect THz signal emitted by the sample. The sample was placed in a static in-plane magnetic field of H = ±200 mT by permanent Neodymium magnet, that is enough to saturate the magnetization. The pump fluence is controlled by a gradient neutral density filter. A half-wave plate (HWP) or a quarter-wave pleat (QWP) was used to change the polarization of the incident laser beam. The THz emission spectroscopy can be performed in both transmission and reflection configurations. The generated THz transient signal in the far-field was focused on a 1 mm-thick <110> ZnTe crystal and detected by contact-free electro-optic sampling (EOS). The samples can be placed in a continuous flow helium cryostat with the temperature varying from 5 to 300 K.
Low-Temperature Magnetization Measurements: The magnetization of Co 2 FeSi/Pt at low temperature was measured by the Quantum Design Superconducting Quantum Interference Device (MPMS3). The area of testing sample was about 1 mm 2 , fixed on the quartz rod by low-temperature glue. The magnetization along the surface of the sample was measured. First, the thin-film sample was premagnetized by applying a 2000 Oe magnetic field along the surface of the sample. Then, the magnetization of the sample was measured by 50 Oe probe magnetic field after removing the premagnetization magnetic field.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.