A perspective on two-dimensional van der Waals opto-spin-caloritronics

Impressive advances in the synthesis of atomically thin semiconductors are tapping MX₂ (M = Mo, W; X = S, Se, Te) TMD monolayers as the “building blocks” of quantum optoelectronics and computing devices.^24–26 The 1H-phase semiconducting TMD monolayers have a hexagonal crystal structure that consists of a plane of metal atoms sandwiched between two planes of chalcogen atoms.²⁷ Because of a break in inversion symmetry and strong spin–orbit coupling, TMD monolayers exhibit large spin and valley polarization, placing them on the short list as potential candidates for applications in spintronics, especially as a component in a heterostructure with other materials.^2,28 In an exciting intersection between the fields of spin caloritronics and 2D materials, Dastgeer et al. reported a large enhancement in V_SSE (∼1000%) in a Pt/Ni₈₁Fe₁₉ heterostructure by inserting a thin layer of WS₂ in between Pt and Ni₈₁Fe₁₉ [Figs. 2(a)–2(c)].²⁹ Herein, the SSE signal varied sensitively with WS₂-thickness and, in this work, the tri-layer WS₂ was found to yield the highest V_SSE [Fig. 2(c)] while the monolayer (ML) WS₂ did not improve the SSE signal of the Pt/WS₂/Ni₈₁Fe₁₉ heterostructure. In any case, the origins of the enhanced SSE and its dependence on WS₂ thickness in the Pt/WS₂/Ni₈₁Fe₁₉ heterostructures are unclear. However, Lee et al. recently showed a giant enhancement of V_LSSE (∼900%) in Pt/ML-WSe₂/YIG and attributed it to the large spin–orbit coupling and improved spin mixing conductance of ML WSe₂ [Figs. 2(d)–2(f)].³⁰ The authors also observed a strong dependence of V_LSSE on the surface coverage of the WSe₂ layer, where the largest LSSE was obtained for ∼15% surface coverage [Fig. 2(f)]. A similar trend was reported for Pt/MoS₂/YIG by the same group; however, MoS₂ was in the form of a multilayer film.³¹ In these cases, the origin of the enhanced spin mixing conductance of Pt/TMD/FM (FM = Ni₈₁Fe₁₉, YIG; TMD = WS₂, WSe₂, MoS₂) remains unclear as the magnetic characteristics of the TMD layer itself (thought to be defective ferromagnets) were not analyzed in detail. Furthermore, the fact that the FM layer (Ni₈₁Fe₁₉ or YIG) could magnetize the TMD layer (WS₂, WSe₂, MoS₂) by magnetic proximity contributes to this uncertainty.^32,33 Indeed, in their latest study, Lee et al. revealed that the induced magnetization of the WSe₂ ML on YIG could enhance the spin mixing conductance and consequently the LSSE signal in Pt/ML-WSe₂/YIG.³⁴ However, this proximity-induced ferromagnetism was not decoupled from the defective or defect-induced ferromagnetism of the WSe₂ ML itself, nor was it manipulated by external stimuli, such as light. In this context, we suggest the introduction of optically tunable FM order within a TMD monolayer, while preserving its semiconducting nature, as a promising strategy to enhance the spin mixing conductance resulting in an optically tunable LSSE in HM/ML-TMD/FM. The successful formation of the HM/magnetic TMD monolayer (HM = Pt or PtSe₂; TMD = V-doped WS₂ or V-doped WSe₂) interfaces will also provide unique opportunities to exploit and control spin transport in low dimensional magnetic systems. In thermoelectrics, low dimensionality is an essential concept for improving thermoelectric conversion performance because a change in the density of states (DOS) can modulate the Seebeck coefficient.³⁵ However, low-dimensional effects have never been observed in spin caloritronics, and such magnetic 2D-TMDs may be a promising platform to investigate such effects. The use of low dimensionality is promising not only for thermo-spin effects, such as SSE and the spin Peltier effect (SPE), but also for magneto-thermoelectric effects, such as the anomalous Nernst/Ettingshausen effects and the anisotropic magneto-Seebeck/Peltier effects, which are directly affected by the electronic structure. Various spin-caloritronic phenomena have been summarized in this review.³⁶ 

These perspectives may become possible as single layers of semiconducting TMDs (WSe₂ and WS₂) have recently been shown to exhibit strong and tunable room-temperature ferromagnetism [Fig. 3(a)] through controlled doping of vanadium atoms, which was further validated by DFT calculations.^10–12 Indeed, increasing vanadium concentration increases the saturation magnetization of V-WS₂ and V-WSe₂ and the bottom-up fabrication of these essentially 2D dilute magnetic semiconductors has allowed the highest doping level ever attained for atomically thin vanadium-doped TMDs.^10,12 Further investigation of this unique class of 2D magnets led to the discovery of thermally induced spin flipping (TISF) in V-WSe₂ monolayers.¹² Interestingly, the TISF phenomenon can be achieved at low magnetic fields (less than 100 mT) and can be manipulated by the vanadium concentration within the V-WSe₂ monolayers. These findings constitute a major breakthrough toward the use of 2D dilute magnetic semiconductors in spin-caloritronic devices with high efficiency spin-to-charge conversion.^30,32 Furthermore, this distinct class of 2D magnets, capable of sustaining magnetic ordering at room temperature, also supports emerging quantum computing concepts, such as spin logic switches, spin transistors, and valleytronics.^24,32,37

A recent study by Jimenez et al. demonstrated light-mediated room temperature ferromagnetism within V-doped TMD monolayers.³⁸ Interestingly, the magnetization of the V-WS₂ ML was tuned by varying light intensity at low power (less than 5 mW/cm²). In this work, the authors attributed absorbed photons from light illumination to an imbalance in the carrier population that led to changes in the magnetic moment. DFT calculations confirm that the magnetic moment of a V-WS₂ monolayer can be altered by hole concentration from optical doping [Figs. 3(b) and 3(c)]. One of the most promising advantages of using light is that it enables local, flexible, and reversible control of magnetization and/or magnetic states. While the previous study³⁸ focused on uniform illumination, it may also be interesting to exploit the local control of spin-caloritronic phenomena and functionalities. The basic concept has been described in the work³⁹ in which the magnetization direction/distribution and resultant heat currents are controlled through all-optical switching. Beyond this demonstration, in magnetic TMD-based systems, the magnetic states and resultant spin-caloritronic phenomena can be actively controlled. All this reveals an innovative turning point: “optically controlled spin-caloritronics or opto-spin-caloritronics” based on 2D-TMD magnetic semiconductors [Fig. 3(d)]. Going beyond conventional spin-caloritronics, this avenue of research combines 2D magnetic TMDs with spin-caloritronics in a field that harnesses “light as the new heat.”

The idea of using an ultrafast optical pulse to trigger or switch magnetization has recently attracted much attention due to its potential applications in ultrafast all-optical switching, ultrafast data storage and sensing technology, and terahertz spintronics.^40,41 The first experiment on ultrafast light control of magnetization was realized back in 1996.⁴² Since then, many breakthrough research studies in this fascinating direction have been published.^41,43 Light is also an important source of pure spin current generation via the spin-photo-voltaic effect,⁴⁴ light-induced impulsive magnetic moments,⁴⁵ and, of course, the SSE.⁴⁶ In the latter mechanism, an ultrafast laser pulse introduces a temperature gradient into the system, for example, in Pt/YIG,⁴⁶ resulting in a dynamical LSSE within 100 fs. Besides the ability to detect ultrafast spin dynamics, femtosecond laser pulses can also generate sizable temperature differences in a few hundred K across interfaces. A study by Seifert et al.⁴⁶ used THz spectroscopy to detect SSE, where THz emissions originated from transient charge currents converted from spin currents via ISHE. This study⁴⁶ demonstrates that THz spectroscopy is a suitable tool to investigate dynamics of spin-to-charge-current conversion on ultrafast timescales as discussed in a recent perspective.⁴⁷ However, THz detection is limited to sub- to few-THz frequencies, which hinders a wider range of dynamic probing. Furthermore, the dynamics of heat generated by the laser pulse and heat dissipation over a short timescale cannot be probed by THz spectroscopy. Therefore, complementary techniques are needed to probe non-equilibrium dynamics of spins and carriers in the SSE, which can be accomplished using the time-resolved magneto-optical Kerr effect (TR-MOKE) in reflection configuration,^48,49 time-resolved Faraday rotation (TRFR),⁵⁰ or time-resolved second harmonic generation (TR-SHG) techniques.⁵¹ The two former techniques can detect spin precession frequency (spin-wave motion) and spin accumulation of the involved materials over a short timescale. Such studies may yield rich information on the dynamics of spin-wave generation, spin current transport through the atomically thin quantum material interface (non-magnetic or magnetic 2D-TMD), and spin-charge conversion.

In addition to a longitudinal spin current, theory predicts the coexistence of a transversal spin Seebeck currents within a few picoseconds.⁵² In a system, such as Pt/TMD/YIG, the transversal spin current from YIG could couple to a spin-polarized valley of the TMD layer, which can be harvested before vanishing. The use of coherent light to generate spin currents (carried by spin waves or magnons) is also an intriguing research direction as it could generate coherent spin waves on ultrafast timescales. The transport of coherent spin waves differs from the diffusive transport driven by a temperature gradient, and the former approach appears to be a more effective transport and is of potential interest for quantum communications and computing applications. To fully exploit ultrafast SSE through 2D-TMD interfaces, it is essential to understand how this effect is probed by ultrafast spectroscopies.

Due to the small ISHE voltage detectable by a conventional SSE measurement, it is very challenging to disentangle the SSE voltage from undesirable voltage sources, such as the conventional Seebeck and Nernst effects.^53,54 All-optical pump-probe noncontact methodologies with pico- or sub-picosecond resolutions, as mentioned above, would be possible alternatives to overcome such difficulties. In an all-optical SSE experiment, the pump pulse acts as the thermal source that generates hot electrons within the pulse duration when incident on the HM layer, which is typically on the order of 10–100 fs. These hot electrons subsequently cool down via electron–electron scattering (sub-picosecond) and electron–phonon interaction (few to tens of picoseconds).⁴³ SSE is then either detected by THz emission⁴⁶ [Figs. 4(a)–4(c)] or Kerr rotation⁴⁸ [Figs. 4(d)–4(f)]. In the former case, THz radiation originates from the transient charge current that was converted from the SSE spin current via ISHE. For example, in this study,⁴⁶ THz radiation originating from SSE is proved by (i) a THz electric field, which is linearly polarized and oriented perpendicular to M [Fig. 4(a)], (ii) the sign of the THz signal is reversed when the magnetization M is reversed [Fig. 4(b)], and (iii) the signal does not depend on the pump polarization. It is worth noting from the last point that optical spin injection or demagnetization depends on the polarization stage of the optical beam (circular or linear) and the independence of the pump polarization indicates that the pump only acts as a heat source. This implies that contributions from the anomalous Nernst effect, the photo-spin-voltaic effect, and optically induced magnetization can be safely neglected. In these measurements, THz emission is generated from a transient charge current converted from the spin current via ISHE, which requires the use of a heavy metal with strong spin–orbit coupling, such as Pt, whereas in TR-MOKE measurements of SSE, Cu is used owing to its long spin diffusion length and long spin-relaxation time as seen in Figs. 4(d)–4(g).^48,55 On the other hand, the Kerr rotation signal is not based on the charge current via ISHE but instead originates from spin accumulation at the surface of the metal layer, which strongly depends on the spin diffusion length, l_s, and hence, the species and thickness of the HM. Indeed, the spin diffusion length of Pt is much shorter (∼7–10 nm) than that of Cu (∼350–500 nm) at room temperature. Furthermore, Cu supports large electron temperature excursions during laser pulse heating because of the weak electron–phonon interaction in this metal allowing most of the spins to diffuse toward and accumulate at the metal surface when Cu is used, which is suitable for MOKE signal detection.^55–57 Besides the Cu metal layer, Au and Ag can also be used in SSE experiments utilizing TR-MOKE to detect spin accumulation, although the diffusion lengths of these metals are smaller compared to Cu.⁵⁵ 

As mentioned earlier, light can be used as an additional knob to control room temperature magnetism in V-doped TMD monolayers (e.g., V-WS₂, V-WSe₂),³⁸ which opens up opportunities to control ultrafast SSE by light. The origin of the light-controlled magnetism in that study³⁸ was assigned to an excess of holes due to the capture of electrons by V-doped ions. However, the carrier and spin dynamics over short timescales have not yet been explored in this system. Ultrafast magneto-optical studies are sorely needed to probe charge and spin dynamics that can elucidate the origin of light-induced magnetism as the spin properties of TMDs may affect the SSE. Therefore, studies on the spin dynamics of the TMD layer itself will provide important insight into the spin transport and relaxation phenomena that form the dynamics of spin-charge conversion in SSE. Semiconducting TMD monolayers exhibit intriguing spin-valley polarization properties owing to their strong spin–orbit-coupling and broken central symmetry. The unique spin-valley coupling in TMDs presents opportunities to use valleys and light to manipulate spins, such as valley-specific optical selection rules depicted in Fig. 5(a).⁵⁸ Upon the photoexcitation of left-handed or right-handed circularly polarized light (LCP or RCP) near the first excitonic resonance, certain spin-polarized electrons can be excited into a corresponding valley. For instance, in a MoS₂ monolayer, RCP excitation with photon energy near the A exciton transition promotes spin-up electrons in the conduction band of the K valley, and vice versa. It is worth noting that MoS₂ is a non-magnetic material.

Kerr rotation spectroscopy has successfully monitored the spin population, coherence, and relaxation under circularly polarized light excitations in n-type MoS₂,^59,60 WS₂,^60,61 and p-type WSe₂^62–64 monolayers with long spin lifetimes ranging from a few to hundreds of nanoseconds at low temperature, which is much longer than that of an exciton lifetime. The origin of long spin lifetimes was attributed to resident electrons and holes^59,60,62,63 and long-lived spin valley-polarized dark trions.^61,64 When increasing temperature, optically induced Kerr rotation signals and spin relaxation times were significantly reduced and almost disappeared at temperatures above 40 K for the case of MoS₂.^59,60 As presented in Fig. 5(b), the sign of the optically induced rotation switched when the opposite handedness of circularly polarized light was used for excitation, which indicates that opposite spin carriers were optically injected into different valleys of the MoS₂ monolayer. Yang et al. further resolved the spin coherence with a Larmor precession frequency that is linear with an applied transverse external magnetic field as seen by the solid curves in Fig. 5(b).⁵⁹ Interestingly, under an external magnetic field, the Kerr rotation signal dropped within 100 ps because of the out-of-plane spin–orbit field on the spin precession as well as rapid intervalley scattering. The small coherent spin precession that persists for longer times is attributed to localized or trapped electrons that were not subject to fast intervalley scattering. The above-mentioned experiments were mostly conducted on n- or p-doped semiconducting monolayers. Meanwhile, the contribution of excitons into spin properties remains unexplored largely due to strong exciton binding energies and their short lifetimes. Strong exciton binding energies could also hamper spin transport and spin-charge conversion in SSE. The introduction of magnetic dopants (e.g., V-doped TMDs) is particularly interesting because it offers tunable magnetism and increases screening, which results in a lower exciton binding energy as well as excessive charges. Spin injection into a TMD semiconducting monolayer from the above studies was achieved by photo-excitation using circularly polarized light. Manipulation of magnetism can also be obtained with ultrafast heating by light. Very recently, light-induced demagnetization has been observed in a few layers of a ferromagnetic semi-insulating Cr₂Ge₂Te₆ nanoflake using time-resolved Faraday rotation in Fig. 5(c). In this study,⁵⁰ an intensive laser pulse induces spin temperature in a short timescale leading to a demagnetization; hence, a change in the magnitude and the direction of the magnetization from its equilibrium direction occurred. Two lifetime constants for recovering the magnetization were found, which were 400 ps and 8 ns and attributed to energy relaxation processes by optical phonons and acoustic photons, respectively.

Unambiguously, TR-MOKE and THz spectroscopies are capable of measuring the ultrafast SSE in bulk systems with timescales from picosecond to nanosecond. These techniques are particularly suited for the investigation of the SSE in heterostructured systems that incorporate a 2D-TMD buffer layer in between HM and FM layers as they can independently investigate physical processes involved in each material [see Fig. 6(a)]. Since the first observation of the SSE in 2008, we have witnessed a surge in studies of this exotic effect over a wide range of materials and structures. Despite the large body of SSE studies in bulk materials, there remains a lack of ultrafast SSE studies in the above-mentioned 2D/3D heterostructures. No doubt this is because of the difficulty behind quantifying the SSE coefficient from TR-MOKE and THz spectroscopies as a rigorous calibration method would be required to quantify the effect. Nonetheless, we propose below that the second harmonic generation technique is capable of measuring and quantifying the spin Seebeck coefficient. Being a second-order nonlinearity, SHG is highly sensitive to a central symmetry breaking. In the presence of a perturbation, such as a DC field, SHG is strongly enhanced via a four-wave-mixing process, and this is the so-called electric field induced second harmonic generation (EFISH) as seen in Fig. 6(b).^65,66 Therefore, it is an excellent probe for detecting SSE voltage. Then, to quantify the amount of generated voltage, the SHG intensity is measured as a function of an externally applied field/voltage in the absence of the pump pulses.

Most of the above-mentioned methodologies to detect SSE are indirect measurements that either measure charge current or voltage in the HM layer via ISHE or by spin accumulation. Direct detection of a pure spin current is quite challenging because there is no net charge current, nor a net magnetization involved. It has been predicted that a longitudinal spin current has peculiar symmetry properties in which the motion of electrons along the spin current direction is a chiral phenomenon.⁶⁷ These exotic properties result in a chiral optical nonlinearity that induces second harmonic generation if circularly polarized light is incident and, therefore, it is a well-suited technique to detect pure spin current. Indeed, soon after this theoretical prediction, an experimental work verified SHG induced by pure spin currents in a GaAs sample.⁵¹ Besides the TR-SHG created by electric field-induced SHG, an observation by Werake et al.⁵¹ suggests that SHG is a useful tool to probe pure spin current directly from the SSE, and details of this have yet to be explored in the context of opto-spin-caloritronics.

Finally, we recall the roadmap for 2D van der Waals magnet-based spin-caloritronics and opto-spin-caloritronics (Fig. 1). It is worth mentioning that the large SSE has recently been predicted to occur in intrinsic 2D ferromagnetic systems, such as in monolayers of CrI₃⁶⁸ and CrPbTe₃,⁶⁹ as well as their van der Waals heterostructure CrI₃/NiCl₂.⁷⁰ An extremely large SSE value of 1450 μV K⁻¹ can be achieved for monolayer CrI₃ and 1320 μV K⁻¹ for monolayer CrPbTe₃ at low temperatures (<100 K). In contrast, much smaller SSE values (∼18 nV K⁻¹) have been experimentally reported for CrGeTe₃/Pt and CrSiTe₃/Pt.⁷¹ A similar trend has also been experimentally observed for CrBr₃/Pt.⁷² Further studies are, therefore, needed to resolve the significant deviation between the theoretically predicted and experimentally obtained SSE values. Since these intrinsic 2D magnets are air-sensitive, careful attention must be paid to the effect of oxidation on the magnetization and, hence, the spin to charge conversion efficiency in these heterostructures. On the other hand, 2D-TMD semiconductors (e.g., WSe monolayer) have been shown to reduce the conductivity mismatch and boost the SSE in HM/FM (Pt/YIG) bilayers.³⁰ The findings of room-temperature ferromagnetism in 2D-TMD semiconductors through magnetic doping (e.g., V-WS₂ and V-WSe₂ monolayers)^10–12 have the potential to transform the fields of van der Waals spintronics and spin-caloritronics, as well as to establish the subfield of “opto-spin-caloritronics” in which the SSE voltage can be controlled by varying intensity of light, which has been demonstrated to mediate the magnetization of a magnetic TMD monolayer.³⁸ Exploring ultrafast magnetism and ultrafast opto-spin-caloritronics in these TMD monolayers and heterostructures will become a very exciting research area. The rich thermal- and opto-mediated magnetism showcased so far in this unique class of 2D dilute magnetic semiconductors sets the stage for further experimental and theoretical research that should be performed in the near future to draw out all of these exciting possibilities.

M.H.P. acknowledges support from the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering under Award No. DE-FG02-07ER46438 and the University of South Florida Nexus Initiative (UNI) under Award number R15301. M.T.T. acknowledges support from the USF New Researcher Grant No. 18325. L.M.W. acknowledges support from the U.S. Department of Energy under Grant No. DE-FG02–06ER46297.

The authors have no conflicts of interest to disclose.

The data that support the findings of this study are available within the article.