betaFeedback

WorksPhysics and Astronomy

Revealing laser-induced modification of weak ferromagnetism through a spin-reorientation phase transition in iron borate Fe3_3BO6_6

Abstract

In the iron oxide Fe3_3BO6_6 weak ferromagnetism stems from competing contributions from two-site Dzyaloshinskii-Moriya interactions (DMI) and from on-site anisotropies at the two 4c4c and 8d8d types of nonequivalent Fe3+^{3+} sites. As a result, spin reorientation (SR) transition Γ2Γ_2 →\rightarrow Γ4Γ_4 in Fe3_3BO6_6 uniquely comprises 90∘^\circ spin switching and a 35% change of weak ferromagnetic moment. We explore ultrafast kinetics of such a transition as a proxy for weak ferromagnetism contributions modified by femtosecond laser pulses. Using magneto-optical pump-probe technique we reveal that the SR transition is triggered by an abrupt decrease of the weak ferromagnetic moment in the Γ2Γ_2 phase, and relate it to the laser-induced modification of the DMI and anisotropy contributions to the spin canting. By tuning the pump pulse phonon energy between different optical absorption ranges, we show that laser-driven impact on spin canting possesses a nonthermal character and is achieved through the charge-transfer transitions 2p→3d2p\rightarrow3d within Fe3+^{3+}O62−^{2-}_6 octahedra rather than by localized 3d→3d3d\rightarrow3d transitions at Fe3+^{3+}. The decreased weak ferromagnetic moment controls the threshold and the latency of the SR transition, as well as a final state. The latter is achieved via the precessional switching to the Γ4Γ_4 phase activated by the laser-induced heating. Final state is found to be a coexistence of Γ4Γ_4 and Γ2Γ_2 phase with modified spin canting persisting even at hundreds of picoseconds.

Introduction

The kinetics of phase transitions reveals how materials transform from one state to another, determining the pathways, timescales, and intermediate states that emerge during such transformations. The ultrafast laser techniques has enabled the exploration of phase transitions kinetics far from equilibrium, where the system is driven on timescales comparable to or shorter than the relaxation times of its degrees of freedom [1, 2, 3, 4, 5, 6, 7]. Laser excitation can induce transient, hidden, and metastable phases inaccessible under equilibrium conditions, and thus alter the very pathways by which a system transforms. Phase transitions in magnetic materials, with their coupled spin, lattice, and electronic subsystems, provide a particularly interesting and challenging task for such studies [8, 9, 10, 11, 12].

Spin-reorientation (SR) transitions being the abrupt switching of the magnetic moments from one crystallographic axis to another represent a particularly interesting case for studies under ultrafast laser excitation. Indeed, they rely on a delicate balance between several interactions stabilizing a particular spin-orientational states [13]. Diverse microscopic sources of magnetic anisotropy provide such a balance in various crystals and structures. These sources include include spin-orbital coupling giving rise to a conventional single-ion as well as two-ion [14] contributions, exchange coupling exploited for a unidirectional anisotropy [15], two-ion spin-orbital coupling yielding Dzyaloshinskii-Moriya interaction (DMI) [16, 17], as well as many-ion magnetic-dipolar interactions playing a crucial role in high-magnetization materials and nanostructures. Changing either of these interactions may steer the material through SR transition. Thus, studies of SR transitions kinetics in both laser-driven and equilibrium regimes are indispensable for a profound understanding of various contributions to magnetic anisotropy. Examples include ultrafast laser heating as a universal approach to both intrinsic [12] and engineered [18] SR transitions; revealing a link between finite timescale of a SR transition and electron-lattice relaxation in rare-earth orthoferrites [19], exploiting non-thermal pathways to control anisotropy via laser-driven phonons [20] and THz electric fields [21], unraveling spin fluctuations in a vicinity of SR transitions [22], and many others.

In antiferromagnets, interactions favoring spin canting, i.e. weak ferromagnetism [23], play a significant role in emerging of SR transitions, as this phenomenon is genuinely linked to orientational spin configurations [24]. A textbook case is a SR transitions (Morin transition) in the hematite that also is a switching between antiferromagnetic and weak ferromagnetic states [25]. Since laser-driven change and control of weak ferromagnetism is a subject of ongoing research nowadays [26, 27, 28, 29], corresponding SR transitions could provide valuable insights into this problem. In particular, materials with an intrinsic link between weak ferromagnetism and SR transition should facilitate resolving an absolute sign of laser-induced spin canting, as well as distinguishing between changes of antisymmetric and symmetric exchange couplings and anisotropy, which competition gives rise to weak ferromagnetism.

A canted antiferromagnet iron borate Fe3_3BO6_6 is a model material for such studies. This compound takes a unique place among weak ferromagnets, with the spin canting having a twofold origin, namely, conventional two-site DMI and single-ion anisotropies at nonequivalent crystallographic sites of magnetic ions. In equilibrium, the first-order SR transition in the iron borate involves a 90° reorientation of the magnetization, accompanied by a 35% change of the latter within a narrow temperature interval of ∼1 K around TSR=415T_{\mathrm{SR}} = 415 K. This, on the one hand, raises a question regarding the leading parameter governing phase transition. On the other hand, it offers an opportunity to address the laser-driven changes of competing contributions to the spin canting.

In this Article, we report on the laser-driven kinetics of the SR transition in Fe3_3BO6_6. By performing the time-resolved magneto-optical pump-probe experiments, we demonstrate that a decisive role in driving the laser-induced SR transition is played by a reduction of the magnetization, achieved through modification of the spin canting. The SR transition is found to proceed via a hidden phase whose magnetization value corresponds to neither of the equilibrium states. We establish how the net magnetization in this hidden phase either hinders or enables complete 90° reorientation facilitated by laser-induced heating. As for the laser-induced spin canting change, we reveal that it relies on laser-driven charge transfer transitions promoting electrons from 2p O2−^{2-} to 3d Fe3+^{3+} states thus altering both DMI and single-ion contributions non-thermally. This process can be realized either by resonantly pumping the change-transfer transition or by pumping below the absorption edge and employing the two-photon absorption. These results complement and extend the findings reported in Ref. [30].

This paper is organized as follows. In Sec. II we discuss the main properties of Fe3_3BO6_6, origin of the weak ferromagnetism and characteristics of the equilibrium SR transition, as well as briefly overview its optical properties relevant to the present study. Section III introduces the samples and their characterization and outlines the experimental strategy to study the kinetics of laser-induced transition. In Sec. IV we discuss the experimental results and reveal features of laser-induced SR transition in Fe3_3BO6_6 by varying laser pulse fluence and photon energy, sample temperature, and applied magnetic field. In Sec. V we discuss signatures, non-thermal origin and possible mechanism underlying the laser-induced spin canting change in Fe3_3BO6_6, analyze how it impacts the SR transition boundaries, and outline the complete scenario of the switching. Section VI presents the concluding remarks and outlook.

Magnetic and optical properties of Fe3_3BO6_6

Magnetic structure and spin-reorientation transition in Fe3_3BO6_6

The iron borate Fe3_3BO6_6 belongs to an orthorhombic space group PnmaPnma (crystallographic point group mmmmmm) in which twelve Fe3+^{3+} ions in the unit cell occupy two non-equivalent 8d and 4c distorted octahedral positions [32]. Unit cell parameters are a=10.05a = 10.05, b=8.55b = 8.55 and c=4.47c = 4.47 Å [33]. Magnetic moments of the iron ions within both positions order at the same temperature TN=508T_N = 508 K. Within each position they are coupled antiferromagnetically and experience canting leading to a weak ferromagnetism, as illustrated in Figure 1(a).

Crystal structure, absorption spectrum, and magnetic energy profiles of Fe3BO6

Figure 1. (a) Crystal structure of Fe3_3BO6_6, spins arrangement, net magnetization M and antiferromagnetic vector L below and above the SR transition. (b) Schematic absorption spectrum in a range 1-3 eV for a light polarized along the c-axis [31]. Vertical arrows indicate the excitation by the pump pulses with photon energy in a range of low and moderate absorption governed by the localized d−dd-d transitions (yellow band) and in the range of strong absorption due to CT transitions (green band). (c) Magnetic energy profiles Φ(φ)\Phi(\varphi) at different equilibrium temperatures T0T_0 and corresponding phases: Γ2\Gamma_2 (blue background), Γ2+Γ4\Gamma_2+\Gamma_4 (gradient background), and Γ4\Gamma_4 (yellow background). Also shown are the conditions for the equilibrium phase boundaries Eq. (7), where K1a(c)K_1^{a(c)} depends on equilibrium ma(c)m^{a(c)} according to Eqs. (6), (4). (d) Evolution of the magnetic energy profile at different laser excitation fluences FF. In contrast to the equilibrium case, phase boundaries are defined by K1aK_1^a with mam^a reduced by laser-driven change of the spin canting.

Weak ferromagnetism in Fe3_3BO6_6 comprises a peculiar case when two mechanisms of the phenomenon, namely single-ion magnetic anisotropy and two-ion DMI, coexist and are of comparable strength. We consider a two-sublattice model with the two magnetizations M1,2=M0m1,2\mathbf{M}_{1,2}=M_0\mathbf{m}_{1,2}, where m1,2\mathbf{m}_{1,2} are unit vectors. Each magnetic sublattice consists of collinear magnetic moments at 8dd and 4cc positions in a proportion 2:1. Then ferromagnetic and antiferromagnetic (Néel) vectors are M=M0m\mathbf{M}=M_0\mathbf{m} and L=M0l\mathbf{L}=M_0\mathbf{l}, respectively Figure 1(a), with

m=0.5(m1+m2)l=0.5(m1−m2).(1)\mathbf{m} = 0.5(\mathbf{m}_1+\mathbf{m}_2) \\ \mathbf{l} = 0.5(\mathbf{m}_1-\mathbf{m}_2). \tag*{(1)}

Phenomenologically, we write the thermodynamic invariant following [34]:

Φ=E2m2−damalc+dcmcla+a12lc2+c12la2+a24lc4+c24la4+f2lc2la2,(2)\Phi= \frac{E}{2}m^2-d_a m_a l_c+d_c m_c l_a +\frac{a_1}{2}l_c^2+\frac{c_1}{2}l_a^2+\frac{a_2}{4}l_c^4+\frac{c_2}{4}l_a^4+\frac{f}{2}l_c^2l_a^2, \tag*{(2)}

with ma(c)m_{a(c)} and la(c)l_{a(c)} being the projections of m\mathbf{m} and l\mathbf{l} on the a(c)a(c)-axis. Here, E/2>0E/2>0 is an effective exchange parameter describing antiferromagnetic coupling between the sublattices, a1a_1, c1c_1, a2a_2, c2c_2, ff are the bilinear and biquadratic magnetocrystalline anisotropy parameters. A pair of the parameters dad_a and dcd_c describes weak ferromagnetism, as the term −damalc+dcmcla-d_a m_a l_c+d_c m_c l_a allows the non-zero net moment along either aa- or cc-axis, respectively.

The physical meaning of the distinct parameters dad_a and dcd_c is evident if one rewrites the relevant term in Eq. (2) in the form with sublattice magnetizations [16, 35]:

Dub⋅[m1×m2]+d(m1am1c−m2am2c).(3)D\mathbf{u}_b\cdot[\mathbf{m}_1\times\mathbf{m}_2]+d\left(m_{1a}m_{1c}-m_{2a}m_{2c}\right). \tag*{(3)}

Then da=D+dd_a=D+d, dc=D−dd_c=D-d. Here DubD\mathbf{u}_b is the Dzyaloshinskii-Moriya vector, which is aligned with the bb-axis in the considered compound. The corresponding term describes the weak ferromagnetism originating from the DMI, with the sublattices canted away from the strictly antiparallel alignment by an angle ∝DE−1\propto DE^{-1}. The second term is essentially the single-ion-like contribution to the weak ferromagnetism enabled by different biaxial anisotropy at 4bb and 8dd sites. Indeed, such a term favors m1\mathbf{m}_1 and m2\mathbf{m}_2 to orient orthogonally to each other in the acac-plane. Competing with the antiferromagnetic exchange coupling, this leads to the canting of sublattices by ∝dE−1\propto dE^{-1}.

Significance of the both contributions to the weak ferromagnetism in Fe3_3BO6_6 is evident from its peculiar SR phase transition. The transition Γ2→Γ4\Gamma_2\to\Gamma_4 consists in an abrupt rotation of m(l)\mathbf{m}(\mathbf{l}) in the acac-plane from a(c)a(c)-axis to c(a)c(a)-axis. Importantly, the weak ferromagnetic moment changes both its orientation and magnitude upon the SR transition. Indeed, from Eqs. (2),(3) it follows that the two contributions to the weak ferromagnetism either add to or compete with each other depending on the spin alignment within the acac-plane, since only the single-ion-like term changes its sign upon the 90° spin reorientation in this plane. This yields the orientational dependence of the total weak ferromagnetic moment:

ma=daE−c2=D+dE−c2,mc=dcE−a2=D−dE−a2,(4)m^a=\frac{d_a}{E-c_2}=\frac{D+d}{E-c_2}, \\ m^c=\frac{d_c}{E-a_2}=\frac{D-d}{E-a_2}, \tag*{(4)}

where ma(c)m^{a(c)} denotes the magnitude of m\mathbf{m} when it is aligned along the a(c)a(c)-axis [36, 37]. It is established that the SR transition in Fe3_3BO6_6 occurs at TSR=415T_{\mathrm{SR}}=415 K [38, 39]. The weak magnetic moment reorientation by 90° from the aa to the cc-axis is accompanied by a reduction in its magnitude to MSc=0.6 MSaM_S^c=0.6\,M_S^a [39], where MSc(a)=M0mc(a)M_S^{c(a)}=M_0m^{c(a)}. The pronounced difference in the magnetization in the Γ2\Gamma_2 and Γ4\Gamma_4 phases signifies that da≠dcd_a\ne d_c and the single-ion-like contribution to weak ferromagnetism (dd) is comparable to the one related to the DMI (DD). A rough estimate using Eq. (4) and neglecting a difference in magnetic anisotropy parameters a2,c2a_2,c_2 gives d≈0.25Dd\approx0.25D in the vicinity of the SR transition.

Another distinct feature of the SR transition in Fe3_3BO6_6 is that it is of the first order and has a very narrow temperature hysteresis in a very narrow temperature interval of 0.5–3.0 K [38, 39]. To highlight the relevance of the weak ferromagnetism to this feature of the SR transition we use the conventional description of the transition based on an expansion of the thermodynamic potential [3]:

Φ=K1cos⁡2φ+K2cos⁡4φ,(5)\Phi=K_1\cos^2\varphi+K_2\cos^4\varphi, \tag*{(5)}

where K1K_1 and K2K_2 are effective anisotropy parameters, φ\varphi is an angle between the weak ferromagnetic moment m\mathbf{m} and the aa-axis Figure 1(a). The effective anisotropy constants K1K_1, K2K_2 in Eq. (5) can be expressed via the parameters entering Eq. (2):

K1=dcm−dam−a12+c12−a22+f2=−2dm−a12+c12−a22+f2;K2=a24+c24−f2,(6)K_1=d_c m-d_a m-\frac{a_1}{2}+\frac{c_1}{2}-\frac{a_2}{2}+\frac{f}{2} \\ =-2dm-\frac{a_1}{2}+\frac{c_1}{2}-\frac{a_2}{2}+\frac{f}{2}; \\ K_2=\frac{a_2}{4}+\frac{c_2}{4}-\frac{f}{2}, \tag*{(6)}

where conditions l2+m2=1l^2+m^2=1 and m≪1m\ll1 were used, and only the terms linear on mm are preserved. The first order character of the SR transition in Fe3_3BO6_6 sets the constrain K2<0K_2<0. The phase is given by the following relations

K1<0:Γ2K1>0, K1+2K2<0:Γ2+Γ4K1+2K2>0:Γ4(7)K_1<0:\qquad\Gamma_2 \\ K_1>0,\ K_1+2K_2<0:\qquad\Gamma_2+\Gamma_4 \\ K_1+2K_2>0:\qquad\Gamma_4 \tag*{(7)}

Corresponding energy profiles are shown in Figure 1(c). The boundaries of the phase transition and phase coexistence are determined by both parameters responsible for the weak ferromagnetism in addition to the magnetocrystalline anisotropy. Narrowness of the transition corresponds to the two conditions for the phase boundaries satisfied at nearly the same temperature, K1(ma)=0K_1(m^a) = 0 and K1(mc)+2K2=0K_1(m^c) + 2K_2 = 0, that is enabled by ma≠mcm^a \ne m^c and a particular ratio between the weak ferromagnetism parameters DD and dd.

Whereas the DMI term in Eq. (3) has been extensively studied over the decades and its origin from exchange-spin-orbital coupling is established [16, 17], the single-ion-like term was investigated in detail only in a few antiferromagnets. This is despite the fact that the seminal works that introduced the phenomenon of weak ferromagnetism [40, 41, 42, 43] already addressed the contribution of the anisotropy. Weak ferromagnetism arising exclusively from the single-ion-like contribution was found experimentally in e.g. NiF2_2 [44] and some other materials [45], and was described theoretically in [36, 41, 43]. Recently, owing to an altermagnetism of NiF2_2 [46] the theory behind this contribution was revisited to emphasize a role of the nonequivalent crystallographic positions of magnetic Ni2+^{2+} ions [46]. It is argued that the corresponding gg-tensor anisotropy yields significant orbital contribution to the weak magnetic moment [46]. Although the microscopic theory of the weak ferromagnetism in Fe3_3BO6_6 has not yet been addressed, the nonequivalent 8dd and 4cc positions occupied by Fe3+^{3+} ions [Figure 1(a)] appear to be essential for a sizable magnitude of dd.

Non-negligible single-ion contribution to the weak ferromagnetism as compared to the DMI distinguishes Fe3_3BO6_6 from other Fe3+^{3+}-based weak ferromagnets such as rare-earth (RE) orthoferrites REFeO3_3, hematite α\alpha-Fe2_2O3_3, and another iron borate FeBO3_3, where DMI is the leading interaction responsible for the spin canting. In α\alpha-Fe2_2O3_3 and FeBO3_3, d≡0d \equiv0 as the single-ion-like contribution is not allowed by their symmetry [40]. The SR (Morin) transition in α\alpha-Fe2_2O3_3 involves switching between phases with symmetry-allowed and forbidden weak ferromagnetism [25]. Although orthoferrites are described by the thermodynamic potential [Eq. (2)], it was found that d≪Dd \ll D and the corresponding term can be neglected [34, 47]. Consequently, SR transitions of a Γ2→Γ4\Gamma_2 \to\Gamma_4 type observed in many orthoferrites result from peculiarities of the magnetocrystalline anisotropy related to the presence of RE ions, and preserve the magnitude of a weak magnetic moment [13].

Such a distinct feature of the equilibrium SR transition in Fe3_3BO6_6 makes it a peculiar case for the study under ultrafast laser excitation. Indeed, revealing transition kinetics should allow to distinguish and analyze stages through which the spin reorientation and change of a weak ferromagnetic moment occur, and to identify the leading factor in the whole transition. Next, it should enable us to reveal and discern laser-induced changes of the two competing weak ferromagnetism parameters DD and dd. On the other hand, high temperature of the transition and its narrowness suggest that laser excitation offers a way to effectively switch the antiferromagnetic vector of this material by 90∘90^\circ.

Optical properties of Fe3_3BO6_6

Fe3_3BO6_6 is a dielectric material with band gap above 2.5 eV [31]. Below-bandgap absorption spectrum consists of 3d3d-3d3d transitions at Fe3+^{3+} sites: 6A1→4T1^6A_1 \to^4T_1 at 1.5 eV and 6A1→4T2^6A_1 \to^4T_2 at 1.9 eV [48]. Third band 6A1→4A1^6A_1 \to^4A_1, 4E^4E at around 2.5 eV is at the tail of a electric-dipole-allowed bands responsible for fundamental absorption band gap [31]. Pronounced distortions of the oxygen octahedra surrounding Fe3+^{3+} at the both positions lead to increased oscillator strength of the d−dd-d transitions as compared to some other iron oxides. Joint theoretical and experimental study of fundamental absorption bands in Fe3_3BO6_6 and a number of related iron oxides was reported in [31]. It was shown that the absorption around 3.0 eV is dominated by electric-dipole-allowed 2p−3d2p - 3d charge-transfer (CT) transitions t2u(π)→t2gt_{2u}(\pi) \to t_{2g} in FeO6_6 octahedra [31]. In both optical ranges, Fe3_3BO6_6 demonstrates noticeable trichroism [31, 48] closely related to its crystal structure.

In this work, we analyze the magnetization dynamics driven by laser pulses with central photon energies below and in the d−dd-d and CT transitions ranges. Figure 1(b) schematically reproduces relevant absorption spectra in the range of 6A1→4T1^6A_1 \to^4T_1 transition from Ref. [48] and of the fundamental absorption around 3.0 eV from Ref. [31] for light polarized along the cc-axis, that is in accordance with the pump pulse polarization as described in Sec. III C.

Sample and experimental details

Samples

For a detailed investigation of the kinetics of the SR transition, measurements were conducted on the two single-crystalline samples with different crystallographic orientations [Figure 2(a, Figure 2b)]. The samples were cut from bulk single crystals grown by the flux method and obtained from different sources. In a (100)-cut sample, referred to as a Sample A, the magnetization M\mathbf{M} is aligned along the sample normal below the SR transition temperature TSRT_{\mathrm{SR}}. Upon heating above TSRT_{\mathrm{SR}}, M\mathbf{M} undergoes a reorientation into the sample plane, aligning along the in-plane cc-axis. In the (110)-cut Sample AB, the magnetization makes an angle β=45∘\beta= 45^\circ with the sample normal below TSRT_{\mathrm{SR}}, and is aligned within the sample plane above TSRT_{\mathrm{SR}}.

Experimental geometry, MOKE hysteresis loops, temperature dependence of static MOKE, BLS spectrum, and qFMR frequency

Figure 2. (a,b) Experimental geometry for magneto-optical detection of the SR transition for the (a) Sample A and (b) Sample AB. Magnetic field HH is applied close to the cc-axis and makes a small angle (not shown) with the sample plane. (c,d) Equilibrium magneto-optical hysteresis loops at temperatures below (T=385T = 385 K, blue curves) and above (T=420T = 420 K, red curves) TSRT_{\mathrm{SR}} for the (c) Sample A and (d) Sample AB. (e) Temperature dependence of static MOKE θSp\theta_{\mathrm{S}}^{\mathrm{p}} and θSl\theta_{\mathrm{S}}^{\mathrm{l}} proportional to MSaM_{\mathrm{S}}^{a} and MScM_{\mathrm{S}}^{c} below and above TSRT_{\mathrm{SR}}, respectively for the Sample A. Blue and red symbols show the data obtained upon cooling down and heating up the sample, respectively. Switching of the magnetization by 90∘90^\circ and the change of the intersublattice angle at SR transition are schematically shown in insets. (f) BLS spectrum of Fe3_3BO6_6 at μ0H=200\mu_0H = 200 mT and two temperatures 418 and 465 K. Solid lines are Gaussian fits. (g) Temperature dependence of the qFMRq_{\mathrm{FMR}} frequency obtained from BLS (circles) and from literature data [49] (solid line).

Magneto-optical characterization

For characterization of a magnetic state of the samples we used magneto-optical magnetometry based on a combination of polar (pMOKE) and longitudinal (lMOKE) Kerr effects. Probe polarization rotation was measured with 170-fs pulses at the photon energy of 2.4 eV with the magnetic field applied close to the sample plane and to the cc-axis, as shown in Figure 2(a, Figure 2b). Small (about 5∘5^\circ) angle the field makes with the sample plane provides a nonzero projection of the field on the sample’s easy axis below TSRT_{\mathrm{SR}}. Probe angle of incidence of 45∘45^\circ ensures that the polarization rotation is sensitive to the in-plane magnetization above TSRT_{\mathrm{SR}} through lMOKE and to the out-of-plane projection of the magnetization below TSRT_{\mathrm{SR}} through pMOKE. Probe is polarized along the cc-axis.

Figures 2(c, d) show equilibrium MOKE hysteresis loops θS(μ0H)\theta_{\mathrm{S}}(\mu_0H) below and above TSRT_{\mathrm{SR}} measured in both samples. The loop shape at T0=385T_0 = 385 K indicates that the strength of the out-of-plane projection of the field is sufficient to saturate the samples along their easy axes in the Γ2\Gamma_2 phase. In-plane field strength, in turn, is insufficient to deviate the magnetization from the aa-axis. Thus, at T0<TSRT_0 < T_{\mathrm{SR}}, θS\theta_{\mathrm{S}} originates from the pMOKE and is proportional to the out-of-plane magnetization, θSp=VpMSa\theta_{\mathrm{S}}^{\mathrm{p}} = V^{\mathrm{p}}M_{\mathrm{S}}^{a} in the Sample A, and θSp=cos⁡β⋅VpMSa\theta_{\mathrm{S}}^{\mathrm{p}} = \cos\beta\cdot V^{\mathrm{p}}M_{\mathrm{S}}^{a} in the Sample AB. Above TSRT_{\mathrm{SR}}, θS\theta_{\mathrm{S}} stems from the lMOKE and is proportional to the in-plane magnetization, θSl=VlMSc\theta_{\mathrm{S}}^{\mathrm{l}} = V^{\mathrm{l}}M_{\mathrm{S}}^{c} in both samples. Taking into account that MSc=0.6MSaM_{\mathrm{S}}^{c} = 0.6M_{\mathrm{S}}^{a}, we obtain ratios between the longitudinal VlV^{\mathrm{l}} and the polar VpV^{\mathrm{p}} magneto-optical coefficients, [Vl/Vp]A=0.64[V^{\mathrm{l}}/V^{\mathrm{p}}]_{\mathrm{A}} = 0.64 for the Sample A, and [Vl/Vp]AB=0.3[V^{\mathrm{l}}/V^{\mathrm{p}}]_{\mathrm{AB}} = 0.3 for the Sample AB. We note that VlV^{\mathrm{l}} and VpV^{\mathrm{p}} and their ratio in two samples are different, since they are governed by different components of the linear magneto-optical and optical susceptibilities due to different crystallographic orientations of the samples and the pronounced trichroism of Fe3_3BO6_6 Sec. II B.

Figure 2(e) shows the temperature dependence of the static MOKE signal at saturation measured for the Sample A. The equilibrium SR transition occurs at TSR=403T_{\mathrm{SR}} = 403 K, marked by the abrupt change of θS\theta_{\mathrm{S}} as a result of the difference between the saturation magnetizations in the Γ2\Gamma_2 and Γ4\Gamma_4 phases. The two curves obtained upon cooling down (blue symbols) and heating up (red symbols) coincide within experimental resolution limited by the temperature sweep step of 5 K. This confirms that the first-order SR transition in the sample exhibits a narrow thermal hysteresis in agreement with literature [38, 39]. Thus, from the equilibrium characterization it follows that, upon excitation of Fe3_3BO6_6 by a laser pulse at T0<TSRT_0 < T_{\mathrm{SR}}, a change of MOKE magnitude at saturation along with the decrease of effective coercivity would be two signatures of the SR transition. We note that the relatively large effective coercivity at T<TSRT < T_{\mathrm{SR}} is a result of a small projection of the field on the corresponding easy axis.

Since the magnetization reorientation during the laser-driven SR transitions often occurs via precessional motion [12], the quasi-ferromagnetic resonance mode (qFMR) was measured using the Brillouin light scattering (BLS) technique in the back-scattering geometry. A single-mode laser with a wavelength of 532 nm was focused on the sample surface into a spot with a diameter of 25 μ\mum using an objective lens. Frequency resolution was achieved using a multi-pass tandem Fabry-Perot interferometer with a Free Spectral Range (FSR) of 100 GHz and a contrast of 101210^{12}, which allowed us to reliably resolve the inelastic magnetic scattering peaks from the strong elastic background. The experiments were performed in a back-scattering geometry at an incidence angle of 30∘30^\circ. To select the magnetic scattering cross-section and suppress the elastic background, a cross-polarization configuration was employed. All experiments were carried out for the Sample A in the temperature range of TBLS=418−485T_{\mathrm{BLS}} = 418 - 485 K in an external magnetic field of μ0H=200\mu_0 H = 200 mT applied in the sample plane along the cc-axis. Figure Figure 2(f) shows the exemplary BLS spectra obtained at two temperatures 418 and 465 K. By fitting these spectra with a Gaussian function to determine the central frequency, the temperature dependence of the FMR frequency was obtained [Fig. 2(g)]. Comparing with the literature data [49], one can see qualitatively similar behavior of the temperature dependence of the frequency, although it differs quantitatively. This is due to the difference in the experimental methods and the presence of an external magnetic field in BLS experiments. Below, the measured qFMRq_{\mathrm{FMR}} frequency is used as a reference to characterize the laser-induced magnetization dynamics as well as the phase of the material following the laser excitation.

Pump-probe experiment

The laser-induced SR transition in Fe3_3BO6_6 was investigated using a femtosecond two-color magneto-optical pump-probe setup with tunable pump photon energy. Laser pulses (duration: 170 fs, repetition rate: 5 kHz) were generated by a Yb3+^{3+}:KGd(WO4_4)2_2 regenerative amplifier PHAROS (Light Conversion) with a central photon energy E=1.2E = 1.2 eV. The pulse was split into pump and probe pulses. Along with the 1.2 eV excitation, additional pump pulse in the photon-energy range of E=1.3−1.85E = 1.3 - 1.85 eV was generated using an optical parametric amplifier (ORPHEUS, Light Conversion). Pump pulses in this spectral range were focused at normal incidence into a spot of 60 μ\mum diameter. For the pump pulses with photon energy E=3.0E = 3.0 eV, the β\beta-barium borate (BBO) crystal was used to obtain the second harmonic from the optical parametric amplifier output at 1.5 eV. The 3.0 eV pump pulses were focused into a spot of 165 μ\mum diameter. Another BBO crystal was used to frequency-double the 1.2 eV pulses, generating probe pulses at 2.4 eV, which were the focused into a 35 μ\mum diameter spot at an incidence angle of 45∘45^\circ as shown in Figure 3(a, Figure 3b). Incidence angle of probe pulses was kept the same as in the static measurements (Sec. IIIB) to enable comparison between the static and dynamic magnetization changes upon SR transition. The pump fluence FF was varied in a range of 0.005 - 0.6 J⋅\cdotcm−2^{-2}, exceeding the probe fluence by up to a factor of 50.

Experimental geometry and pump-probe signals for Samples A and AB

Figure 3. Experimental geometry for pump-probe experiments for (a) Sample A and (b) Sample AB. Laser-induced probe polarization rotation Δθ\Delta\theta as a function of time delay tt measured in the magnetic field μ0H=±200\mu_0H = \pm200 mT at T0=385T_0 = 385 K and F=0.5F = 0.5 J⋅\cdotcm−2^{-2} and E=1.2E = 1.2 eV for (c) Sample A and (d) Sample AB. (e) Zoomed in pump-probe signal in the time range after excitation for the Sample A and the corresponding (f) first and (d) second derivatives of the experimental signals (symbols) and of their fits (lines). Vertical red lines designate the time delays range where the corresponding derivative is considered to be zero.

Laser-induced change of magneto-optical rotation of the probe polarization Δθ\Delta\theta was measured as a function of the pump-probe time delay tt for negative and positive sign of an external magnetic field. To extract the field-

sign dependent contribution, the signals were evaluated as

Δθ(t)=0.5[Δθ(H+;t)−Δθ(H−;t)].(8)\Delta\theta(t) = 0.5 \left[\Delta\theta(H_{+};t) - \Delta\theta(H_{-};t)\right]. \tag*{(8)}

Typical signals Δθ\Delta\theta measured in the external field μ0H±=±200\mu_0H_{\pm} = \pm200 mT at the initial temperature T0=385T_0 = 385 K and F=0.1F = 0.1 J⋅\cdotcm−2^{-2} and E=1.2E = 1.2 eV for the Samples A and AB are shown in Figure 3(c, Figure 3d). The dynamics exhibit the same characteristic behavior in both samples, indicating the same nature of the observed phenomena.

Transient signal Δθ\Delta\theta for both samples demonstrates a peculiar behavior. A rapid decrease within ≈3\approx3 ps after excitation is followed by a latency period τL\tau_{\mathrm{L}} during which the signal remains constant at a value ΔθL\Delta\theta_{\mathrm{L}}. After the latency period, at t=tSRt = t_{\mathrm{SR}} pronounced decaying oscillations start resulting in a signal settling at a final level ΔθF\Delta\theta_{\mathrm{F}} at t>100t > 100 ps. To provide quantitative description of such behavior, we fitted Δθ\Delta\theta transients using a piecewise function with the exponential evolution 0→ΔθL0 \rightarrow\Delta\theta_{\mathrm{L}} within 0<t<tSR0 < t < t_{\mathrm{SR}}, and the two damped cosine-functions superimposed by another exponential evolution ΔθL→ΔθF\Delta\theta_{\mathrm{L}} \rightarrow\Delta\theta_{\mathrm{F}} at t>tSRt > t_{\mathrm{SR}}:

{ΔθL[exp⁡(−tτ1)−1],for t<tSR;ΔθF[exp⁡(−t−tSRτ2)−1]+∑i=1,2Θiexp⁡(−tτoi)cos⁡[2πfit+ϕi],for t>tSR,(9)\begin{cases} \Delta\theta_L\left[\exp\left(-\frac{t}{\tau_1}\right)-1\right], & \text{for } t < t_{\mathrm{SR}};\\ \Delta\theta_F\left[\exp\left(-\frac{t-t_{\mathrm{SR}}}{\tau_2}\right)-1\right]+\sum_{i=1,2}\Theta_i\exp\left(-\frac{t}{\tau_{oi}}\right)\cos\left[2\pi f_i t+\phi_i\right], & \text{for } t > t_{\mathrm{SR}}, \tag*{(9)} \end{cases}

where ΔθL,F\Delta\theta_{L,F} are the signals during the latency and at the final state; τ1,2\tau_{1,2} are characteristic times of two exponential functions. In Figure 3(e) a signal zoomed in delay time range right after excitation is shown for the Sample A. To determine τL\tau_L, the first d(Δθ)/dtd(\Delta\theta)/dt and the second d2(Δθ)/dt2d^2(\Delta\theta)/dt^2 derivatives of the signals and of their fits, were obtained as demonstrated in Figure 3(f, Figure 3g). We extracted τL\tau_L as a time range where d(Δθ)/dt=d2(Δθ)/dt2=0d(\Delta\theta)/dt = d^2(\Delta\theta)/dt^2 = 0. A consistent fitting procedure was applied to all experimental data, ensuring a uniform analysis across the entire set of data. This approach allows for a reliable comparison of the extracted parameters and confirms that the underlying physical mechanism is common to both samples.

Results

Magnetic field dependence

We start by examining the evolution of the magnetic state underlying such magneto-optical transient at fluence F=0.5F = 0.5 J·cm−2^{-2} and temperature T0=385T_0 = 385 K [Fig. 3(c)] by analyzing how it is affected by an applied magnetic field. As the dynamics Δθ(t)\Delta\theta(t) is found to be similar in the two samples, here we here focus at the trends observed in the Sample A. We unveil the magnetic configurations during the latency preceding the precession and at the final state by measuring MOKE hysteresis loops at the time delays t=3t = 3 and 200 ps [Fig. 4(a, b)], respectively. Additionally, we verify that at negative time delay t=−10t = -10 ps (dashed lines in Figure 4(a,Figure 4b)), the hysteresis loop is the same as that measured under equilibrium conditions [Fig. 2(c)] at the same T0T_0. This confirms that the system relaxes back to its initial state between two subsequent pump pulses. During the latency τL\tau_L, the coercivity is unchanged, showing that the magnetization remains aligned along the aa-axis [Fig. 4(a)]. However, pMOKE at saturation is reduced to ΔθL\Delta\theta_L. Thus, magnetization MaM^a is reduced at this stage by ΔMa\Delta M^a, and ΔθL=VpΔMa\Delta\theta_L = V^p\Delta M^a. Figure 4(d) zooms in the corresponding delay time range for two values of the magnetic field. The latency τL\tau_L is found to decrease with the external magnetic field [Fig. 4(e)] indicating that its origin is related to the behaviour of the magnetic system after the laser excitation. In turn, the signal magnitude during the latency ΔθL\Delta\theta_L is independent of the field as shown in Figure 4(f).

MOKE hysteresis loops and field dependences for Sample A

Figure 4. MOKE hysteresis loops for Sample A at the time delays (a) t=3t = 3 ps and (b) 200 ps after laser excitation (symbols), and at t=−10t = -10 ps (dashed line). Solid lines are the guides to an eye. (c) Field dependences of the precession frequencies f1f_1 (open symbols) and f2f_2 (closed symbols) obtained from Δθ(t,F)\Delta\theta(t,F) at T0=385T_0 = 385 K, and of the qFMR frequency (line) calculated at T0=430T_0 = 430 K for Sample A. Star shows the qFMR frequency at TBLS=430T_{\mathrm{BLS}} = 430 K from BLS measurements Figure 2(g). (d) Zoomed in transient signal during 2 ps after the excitation for the Sample A obtained at H=40H = 40 and 240 mT. Magnetic fields dependences of (e) the latency τL\tau_L, (f) ΔθL\Delta\theta_L during the latency τL\tau_L (red symbols) and ΔθF\Delta\theta_F at t>100t > 100 ps (blue symbols).

At the final state, t=200t = 200 ps, the signal appears to be a superposition of the two types of hysteresis loops, characteristic of the Γ2\Gamma_2 and Γ4\Gamma_4 phases [Fig. 4(b)]. This reveals that the SR transition indeed occurs at the given pump fluence, but with only a fraction NSWN^{\mathrm{SW}} of the material is switched to the Γ4\Gamma_4 phase. Thus, the signal ΔθL\Delta\theta_L at t=200t = 200 ps is the sum of the pMOKE from the unswitched fraction and that of the MOKE from the switched fraction. As shown in Figure 4(f), ΔθL\Delta\theta_L weakly grows with the field.

The final state is achieved at t>100t > 100 ps via the magnetization precession [Fig. 3(b)]. The corresponding oscillatory signal comprises two contributions [Fig. 4(c)]. The first one possesses a higher amplitude and a lower frequency f1f_1. The second contribution possesses a lower amplitude and a higher frequency f2f_2. To compare f1f_1 and f2f_2 with the qFMR frequency in the Γ4\Gamma_4 phase, we note that the external magnetic field HH is applied along the cc-axis being the easy axis in this phase. Then the frequency can be calculated using the expression f(H)=γ(H+HA)f(H) = \gamma(H+H_A) [50], where γ=28\gamma= 28 GHz·T−1^{-1} is the gyromagnetic ratio, and HA=2K1(Msc)−1H_A = 2K_1(M_s^c)^{-1} is the effective anisotropy field assuming a contribution from biquadratic anisotropy terms entering K2K_2 being small. For calculations, Msc=1900M_s^c = 1900 A·m−1^{-1} was taken from [51], K1=1.6⋅103K_1 = 1.6 \cdot10^3 J·m−3^{-3} was obtained from the qFMR data at T0=430T_0 = 430 K in zero external field [49]. The field dependence of f2f_2 [Fig. 4(c)] agrees well with the calculated values of the qFMR frequency in the Γ4\Gamma_4 phase at 430 K, indicating that it is a precession in the Γ4\Gamma_4 phase and it possesses a linear regime. In contrast, we associate the field-dependent frequency f1f_1 with the precession in a non-perturbative transient regime, and discuss it in details in Sec. V D. Thus, field dependences of the transient signals reveal the three stages of the magnetic state evolution: initial change of magnetization without reorientation, precessional switching, and quasi-steady state corresponding to coexistence of Γ2\Gamma_2 and Γ4\Gamma_4 phases.

Fluence dependence

For laser-induced phase transitions, fluence dependences of their parameters are a key fingerprint [18, 20]. Therefore, we examined Δθ(t)\Delta\theta(t) traces measured for the Sample A at different FF, fixed initial temperature T0=385T_0 = 385 K and the pump photon energy of E=1.2E = 1.2 eV Figure 5(a). Evidently, ΔθL,F\Delta\theta_{\mathrm{L,F}}, τL\tau_{\mathrm{L}}, and the precession parameters experience both quantitative and qualitative changes with pump pulse fluence.

Pump-probe signals and fluence dependences

Figure 5. (a) Pump-probe signals Δθ(t)\Delta\theta(t) (symbols) at μ0H=200\mu_0 H = 200 mT, T0=385T_0 = 385 K at various pump fluences, and their fits (lines). (b) Fluence dependences of ΔθL\Delta\theta_{\mathrm{L}} during the latency τL\tau_{\mathrm{L}} (red symbols) and ΔθF\Delta\theta_{\mathrm{F}} at t>100t > 100 ps (blue symbols). Solid line is a decrease of magnetization due to laser-induced heating evaluated based on the effective temperature TFT_{\mathrm{F}}. Fluence dependences of (c) the fraction NSWN^{\mathrm{SW}} of the material switched to the Γ4\Gamma_4 phase, (d) the experimental and calculated amplitude Θ1\Theta_1 (e) the precession frequency f2f_2 (circles) collated with the temperature dependence of the qFMR frequency obtained from BLS (stars), (f) the latency τL\tau_{\mathrm{L}}.

As demonstrated above in Figure 4(b), at t>100t > 100 ps, the magneto-optical response comprises a signal from the switched fraction of sample manifested via pMOKE, and from the non-switched fraction, manifested via lMOKE for this geometry. It can be evaluated as

θF=NSWVlMc+(1−NSW)cos⁡β⋅Vp(Ma+ΔMa),(10)\theta_{\mathrm{F}} = N^{\mathrm{SW}} V^{\mathrm{l}} M^{\mathrm{c}} + \left(1-N^{\mathrm{SW}}\right)\cos\beta\cdot V^{\mathrm{p}}\left(M^{\mathrm{a}}+\Delta M^{\mathrm{a}}\right), \tag*{(10)}

where Vl⋅Mc=θSlV^{\mathrm{l}}\cdot M^{\mathrm{c}} = \theta_{\mathrm{S}}^{\mathrm{l}}, and cos⁡β⋅Vp⋅(Ma+ΔMa)=θSp+ΔθL\cos\beta\cdot V^{\mathrm{p}}\cdot\left(M^{\mathrm{a}}+\Delta M^{\mathrm{a}}\right) = \theta_{\mathrm{S}}^{\mathrm{p}}+\Delta\theta_{\mathrm{L}}, β=0\beta= 0 for the Sample A. In the pump-probe experiment one measures difference between the magneto-optical signals in the laser-induced state θF\theta_{\mathrm{F}}, and in the initial equilibrium state θSp\theta_{\mathrm{S}}^{\mathrm{p}}:

ΔθF=θF−θSp=θSp−NSWθSl−(1−NSW)(θSp+ΔθL)=NSW(θSl−θSp−ΔθL)+ΔθL.(11)\begin{aligned} \Delta\theta_{\mathrm{F}} = \theta_{\mathrm{F}}-\theta_{\mathrm{S}}^{\mathrm{p}} = \theta_{\mathrm{S}}^{\mathrm{p}}-N^{\mathrm{SW}}\theta_{\mathrm{S}}^{\mathrm{l}}-\left(1-N^{\mathrm{SW}}\right)\left(\theta_{\mathrm{S}}^{\mathrm{p}}+\Delta\theta_{\mathrm{L}}\right) \\ &= N^{\mathrm{SW}}\left(\theta_{\mathrm{S}}^{\mathrm{l}}-\theta_{\mathrm{S}}^{\mathrm{p}}-\Delta\theta_{\mathrm{L}}\right)+\Delta\theta_{\mathrm{L}}. \tag*{(11)} \end{aligned}

From this relation and the experimental data of ΔθL\Delta\theta_{\mathrm{L}} and ΔθF\Delta\theta_{\mathrm{F}} (Figure 5(b)), we extracted the dependence of the switched fraction NSWN^{\mathrm{SW}} on the pump fluence FF. As seen in Figure 5(c) there is a threshold fluence Fth≈0.25F_{\mathrm{th}}\approx0.25 J cm−2^{-2} at which NSWN^{\mathrm{SW}} starts to increase from zero, that is characteristic for laser-induced phase transitions. Importantly, only at the highest fluence F=0.6F = 0.6 J cm−2^{-2}, the SR transition is completed, i.e. NSW=1N^{\mathrm{SW}}=1, and only the Γ4\Gamma_4 phase is present.

We also examine if the precession observed at intermediate time delays is indeed related to the SR transition. According to the suggested interpretation, the amplitude Θ1\Theta_1 should be related to the magnetization length achieved during the latency, Ma+ΔMaM^{\mathrm{a}}+\Delta M^{\mathrm{a}} multiplied by the switched fraction NSWN^{\mathrm{SW}}:

Θ1∝NSWVp(Ma+ΔMa)=NSW(θSp+ΔθL).(12)\Theta_1 \propto N^{\mathrm{SW}}V^{\mathrm{p}}\left(M^{\mathrm{a}}+\Delta M^{\mathrm{a}}\right)=N^{\mathrm{SW}}\left(\theta_{\mathrm{S}}^{\mathrm{p}}+\Delta\theta_{\mathrm{L}}\right). \tag*{(12)}

In Figure 5(d), the experimentally obtained fluence dependence of the amplitude Θ1\Theta_1 is compared with that calculated using the relation above. As expected, precession occurs at F>FthF>F_{\mathrm{th}}, and the experimental and calculated fluence dependences of its amplitude exhibit qualitatively similar behavior. Quantitative difference stems from the fact that the original fit function Eq. (9) used to obtain Θ1\Theta_1 does not account for a gradual increase of the lMOKE contribution during the precession due to growing in-plane projection of magnetization. We note that taking this process into account in a fit function would overburden the latter without yielding significant physical details.

While the fluence dependence of the precession amplitude confirms the precessional character of the switching, the precession frequency after SR transition offers a means to characterize the laser-induced phase [18, 20, 28, 52]. The frequency is given by the transient magnetic parameters MSM_{\mathrm{S}}, K1,2K_{1,2} of the laser-induced phase, and therefore we can use the frequency f2f_2 of the precession in a linear regime to characterize this phase. In Figure 5(e), we compare two dependences, namely the increase of f2f_2 with pump fluence and the temperature dependence of the equilibrium qFMR frequency obtained from BLS measurements. The comparison reveals that the effective temperature TF=TBLST_{\mathrm{F}} = T_{\mathrm{BLS}} reached by the material after excitation for any pump fluence F≥FthF \geq F_{\mathrm{th}} exceeds TSRT_{\mathrm{SR}}. On the one hand, this observation confirms the occurrence of a laser-driven SR transition. On the other hand, such laser-induced heating would be expected to drive a complete SR transition (i.e., NSW=1N^{\mathrm{SW}} = 1) already at FthF_{\mathrm{th}}, in accordance with equilibrium studies [38, 39]. This observation reveals that laser-induced heating alone is insufficient to achieve a complete SR transition.

Apparent contradiction between the characteristics of the precession and the final state draws ones attention to the first stage of the laser-induced magnetic dynamics characterized by the parameters ΔθL\Delta\theta_{\mathrm{L}} and τL\tau_{\mathrm{L}}, which fluence dependences are presented in Figure 5(b,Figure 5f). Importantly, the latency time τL\tau_{\mathrm{L}} decreases monotonically with increasing fluence and vanishes at fluences when NSWN^{\mathrm{SW}} reaches 1, suggesting close connection between these two parameters. The initial magnetization decrease ΔθL\Delta\theta_{\mathrm{L}} observed during this stage shows a non-linear behavior. This decrease is found to be much steeper than the one expected for a thermally-induced decrease of magnetization [see solid line in Figure 5(b)]. The latter is evaluated based on the effective temperature TF(F)T_{\mathrm{F}}(F) as

θp(TF)−θsp(T0)=θp(0)(1−TFTN)ζ−θsp(T0).(13)\theta^{p}(T_{\mathrm{F}})-\theta_{\mathrm{s}}^{p}(T_{0})=\theta^{p}(0)\left(1-\frac{T_{\mathrm{F}}}{T_{\mathrm{N}}}\right)^{\zeta}-\theta_{\mathrm{s}}^{p}(T_{0}). \tag*{(13)}

Here θp(0)≈54\theta^{p}(0) \approx54 mdeg is pMOKE at low temperature obtained from the extrapolation of the θp(T)\theta^{p}(T) dependence [Figure 2(e)], ζ=0.365\zeta= 0.365 is the critical exponent of the magnetization of the 3D Heisenberg ferromagnet.

Temperature dependence

A combined analysis of the evolution of the signals and their parameters with initial sample temperature is performed to further confirm that the SR transition is observed and to evaluate the role of laser-induced heating. Figure 6(a) shows the pump-probe signals measured at various initial temperatures at fixed fluence F=0.5F = 0.5 J⋅\cdotcm−2^{-2} for Sample A. All signatures of the laser-induced SR transition are observed only in the range of 335335 K <T0<TSR< T_{0} < T_{\mathrm{SR}}, when the sample is in the Γ2\Gamma_{2} phase initially and the fluence is sufficient to induce the SR transition. At lower or higher fluence the lowest limit shifts to higher or lower temperature, respectively, as was verified at the fluence value F=0.35F = 0.35 J⋅\cdotcm−2^{-2}.

Pump-probe signals and temperature dependences of the magnetic parameters

Figure 6. (a) Pump-probe signals Δθ(t)\Delta\theta(t) (symbols) measured at F=0.5F = 0.5 J⋅\cdotcm−2^{-2} at various initial temperatures at μ0H=200\mu_{0}H = 200 mT, and their fits (lines). (b) Temperature dependence of ΔθL\Delta\theta_{\mathrm{L}} (red symbols) during the latency τL\tau_{\mathrm{L}} and ΔθF\Delta\theta_{\mathrm{F}} at t>100t > 100 ps (blue symbols). Solid line is a decrease of magnetization due to laser-induced heating evaluated based on the effective temperature TFT_{\mathrm{F}} [Eq. (13)]. Temperature dependence of (c) the switched fraction NSWN^{\mathrm{SW}}, (d) the experimental (symbols) and calculated (line) amplitude Θ1\Theta_{1}, (e) the precession frequency f2f_{2} (circles) collated with the temperature dependence of the qFMR frequency obtained from BLS (stars), and (f) the latency τL\tau_{\mathrm{L}}.

The temperature dependences of the parameters ΔθL\Delta\theta_{\mathrm{L}} and ΔθF\Delta\theta_{\mathrm{F}}, describing the magnetic state during initial stage τL\tau_{\mathrm{L}} and at the final time t>100t > 100 ps, respectively, are plotted in Figure 6(b). Using these values and Eq. (11), we calculated the switched fraction NSWN^{\mathrm{SW}} [Figure 6(c)]. Next, using Eq. (12), values NSWN^{\mathrm{SW}}, τL\tau_{\mathrm{L}} and θsp\theta_{\mathrm{s}}^{p} we calculate the amplitude Θ1\Theta_{1} [Figure 6(d)]. The precession frequency during the SR transition is used to estimate the final temperature TFT_{\mathrm{F}} by collating two dependences, namely the laser-induced precession frequency f2f_{2}

and the equilibrium qFMR frequency obtained from BLS [Figure 6(e)]. As in the case of the fluence dependence, laser-induced heating above TSRT_{\mathrm{SR}} appears to be insufficient for reaching the complete switching, NSW=1N^{\mathrm{SW}} = 1. The latency τL\tau_{\mathrm{L}} [Figure 6(f)] shows a decrease with temperature, while ΔθL\Delta\theta_{\mathrm{L}} grows with the temperature only weakly, and its magnitude exceeds the one estimated using Eq. (13) from a thermally-induced demagnetization [see solid line in Figure 6(b)].

At T0>TSRT_{0} > T_{\mathrm{SR}} the transient Δθ(t)\Delta\theta(t) does not show any signatures of the laser-driven SR transition at any studied fluence, as shown in Figure 6(a) for T0=405T_{0} = 405 K. The transient signal consists of clinical decrease Δθ\Delta\theta and only its magnitude grows with fluence.

When the initial temperature T0T_{0} is approaching but not exceeding TSRT_{\mathrm{SR}}, two additional effects emerge, both characteristic of the laser-driven SR transitions [see data for T0=396T_{0} = 396 K in Figure 6(a)]. First, at negative time delays, the signal Δθ\Delta\theta is not equal to zero. At positive delays, the transient signal at this temperature demonstrates additional fluctuations. Note that at T0>TSRT_{0} > T_{\mathrm{SR}} both additional effects disappear. We ascribe such behavior to the fact that, at T0T_{0} being in the vicinity of TSRT_{\mathrm{SR}}, the excited system returns to different initial states after each pump pulse. Then the system is excited to a distinct state by a subsequent pulse. This is due to the fact that the first-order phase transitions are characterized by a sharp change in the thermodynamic properties of the system when passing through the transition temperature, when the system stability is lost and fluctuations are greatly increased [22]. Thus, the closer the initial sample temperature to the transition, the larger is the contribution to the transition from increasing fluctuations.

Pump photon energy dependence

Further insight into the laser-induced transition can be gained based on the changes of transient signals with the pump pulse photon energy. Three spectral ranges are chosen based on optical absorption spectrum [Fig. 1(b)]. At E=1.2E = 1.2 eV, there is no resonant absorption, and the absorption coefficient α=280\alpha= 280 cm−1^{-1} is the lowest. In the range 1.3−1.851.3-1.85 eV, we scan through the d−dd-d transition, with α\alpha taking moderate values between 450 and 1050 cm−1^{-1}. At 3.03.0 eV, there is a fundamental absorption due to charge-transfer (CT) transitions 2p(O2−)→3d(Fe3+)2p(\mathrm{O}^{2-})\to3d(\mathrm{Fe}^{3+}), with α=1.87×105\alpha= 1.87 \times10^5 cm−1^{-1}.

Figure 7(a) shows the Δθ(t)\Delta\theta(t) traces measured at different pump photon energy EE at the fixed initial temperature T0=385T_0 = 385 K for the Sample B. For this plot, pump fluences were chosen such that ΔθL\Delta\theta_L has the same value of -3 mdeg at the high-energy end of intermediate spectral range (E=1.85E = 1.85 eV) and at E=3.0E = 3.0 eV. For E=1.3−1.85E = 1.3-1.85 eV and 3.03.0 eV the fluence was of F=0.160F = 0.160 J·cm−2^{-2} and 0.0120.012 J·cm−2^{-2}, respectively. Figures 7(b-c) show the parameters ΔθL,F\Delta\theta_{L,F}, as well as NSWN^{\mathrm{SW}} as functions of the pump photon energy EE. All three parameters exhibit similar dependences in the ranges of low and moderate absorption, with weak growth as EE increases and nearly constant values in the range 1.5−1.851.5-1.85 eV. Noticeably, at E=3.0E = 3.0 eV this correlation between ΔθL\Delta\theta_L, ΔθF\Delta\theta_F, and NSWN^{\mathrm{SW}} is broken down, with ΔθF\Delta\theta_F and NSWN^{\mathrm{SW}} reaching much larger values at the same value of ΔθL\Delta\theta_L. This result further substantiates the assumption that there is no simple connection between the laser-induced heating related to absorption α\alpha and the behavior of ΔθL\Delta\theta_L, ΔθF\Delta\theta_F, and NSWN^{\mathrm{SW}}. Finally, fluence dependences ΔθL\Delta\theta_L observed at different pump photon energies and shown in Figure 7(d-Figure 7f) appear to have different character depending whether the pump pulse is in the range of the strong fundamental absorption or not. Only in the former case [Fig. 7(f)] this dependence is linear, while in other cases it follows ∝−F2\propto-F^2 [Fig. 7(d-e)].

Pump-probe signals and pump photon energy dependences

Figure 7. (a) Pump-probe signals Δθ(t)\Delta\theta(t) (symbols) measured at initial temperature T0=385T_0=385 K and at μ0H=200\mu_0H =200 mT for pump photon energy E=1.3−1.85E = 1.3-1.85 eV at F=0.16F = 0.16 J·cm−2^{-2} and for pump photon energy E=3.0E = 3.0 eV at F=0.012F = 0.012 J·cm−2^{-2}. Pump photon energy dependences of (b) ΔθL\Delta\theta_L (red symbols) and ΔθF\Delta\theta_F (blue symbols), and (c) NSWN^{\mathrm{SW}}. Pump fluence was of F=0.012F = 0.012 J·cm−2^{-2} at E=3.0E = 3.0 eV and F=0.16F = 0.16 J·cm−2^{-2} at E=1.2E = 1.2, 1.3−1.851.3-1.85 eV. Experimental dependence (symbols) of ΔθL/ΘSop\Delta\theta_L/\Theta_S^{\mathrm{op}} on pump fluence for Sample AB at T0=385T_0=385 K for pump photon energies (d) E=1.2E = 1.2 eV, (e) 1.51.5 and 1.751.75 eV, and (f) 3.03.0 eV, and their fits (lines) using functions −F2-F^2 (d,e) and −F-F (f). In (d) the same dependence is shown for the Sample A as well.

Overall, experimental results presented in this Section IV provide clear evidence of a nontrivial laser-induced SR transition in Fe3_3BO6_6. The transition occurs via precession and is only partial in a certain range of fluences and initial sample temperatures, so that a fraction of the material NSWN^{\mathrm{SW}} transforms into the phase Γ4\Gamma_4 while the rest remains in the Γ2\Gamma_2 phase. But the fluence and temperature dependences of NSWN^{\mathrm{SW}} cannot be explained based on a laser-induced heating only. The switching is preceded by a net magnetic moment decrease ΔMa∝ΔθL\Delta M^a \propto\Delta\theta_L that proceeds without rotation and persists during a time interval τL\tau_L. ΔMa\Delta M^a does not correlate with laser-induced heating, and shows different fluence dependences depending on the pump photon energy. The latency time τL\tau_L itself decreases with fluence, temperature, and the applied magnetic field. All this observations suggest that the first stage of the process represents a bottleneck for the complete transition.

DISCUSSION

Laser-induced change of the parameters responsible for the spin canting

We first discuss the physical origin of the first stage of the magnetization evolution. As discussed above, during the time interval τL\tau_L the signal ΔθL\Delta\theta_L is proportional to a decrease in the magnetization ΔMa\Delta M^a, without its rotation by 90∘90^\circ. This decrease cannot be explained by laserinduced heating leading to demagnetization, as demonstrated in Figure 5(b) and Figure 6(b), where the calculated decrease of magnetization (solid lines) does not reproduce the experimental data. We therefore suggest a mechanism for the observed reduction of the net magnetic moment due to a change in the intersublattice canting.

The decrease of the canting between the sublattices can result from either decrease of the two- and single-site contributions to the weak ferromagnetism, DD and dd, or from the increase of the exchange parameter EE [see Eq. (4)]. Distinguishing between these two processes in a weak ferromagnet is challenging [26]. However, in the present case one could argue that the laser-induced change of exchange coupling is not feasible. Indeed, the laser-induced change ΔMSa/MSa=ΔθL/θSp\Delta M_{\mathrm{S}}^{a}/M_{\mathrm{S}}^{a} = \Delta\theta_{\mathrm{L}}/\theta_{\mathrm{S}}^{p} achieved at large fluences is of 50 % [Figure 7(d-Figure 7f),Figure 8(a)], requiring a corresponding increase of the intersublattice exchange coupling by nearly a factor of two. Thus, we consider the change of DD and dd as the most feasible underlying process.

Figure 8. (a) The ratio of the change ΔθL\Delta\theta_{\mathrm{L}} to the θSp,l\theta_{\mathrm{S}}^{p,l}, plotted as a function of pump fluence for the Sample A, measured at μ0H=200\mu_{0}H = 200 mT and E=1.2E = 1.2 eV for T0=385T_{0} = 385 K (blue symbols) and T0=420T_{0} = 420 K (red symbols) (b) Relative change of the parameters DD and dd as a function of pump fluence as evaluated based on ΔθL(F)\Delta\theta_{\mathrm{L}}(F) using Eq. (14). Lines are the fits using ∝−F2\propto-F^{2} function.

Experimental dependences ΔθL(F)\Delta\theta_{\mathrm{L}}(F) below and above TSRT_{\mathrm{SR}}, and the ratio of the equilibrium parameters d≈0.25Dd \approx0.25D (see Sec. II A), enable us to evaluate the relative changes of both parameters DD and dd. Indeed, from Eqs. (4) we obtain

ΔθLθSp(l)=ΔD±ΔdD±d(14)\frac{\Delta\theta_{\mathrm{L}}}{\theta_{\mathrm{S}}^{p(l)}}=\frac{\Delta D \pm\Delta d}{D \pm d} \tag*{(14)}

where ±\pm and superscript p(l)p(l) correspond to the case below (above) TSRT_{\mathrm{SR}}. Figure 8(a,Figure 8b) shows corresponding fluence dependences obtained at T0=385T_{0} = 385 and 420420 K for the Sample A and with the pump at the photon energy E=1.2E = 1.2 eV. Both parameters DD and dd decrease with fluence with their relative changes reaching ∼40\sim40 % and ∼30\sim30 %, respectively. However, the equilibrium value of DD is approximately four times larger than that of dd [Sec. II A]. Thus, despite the comparable relative changes of the two parameters, the laser-induced variation of DD provides the leading contribution to the observed change of the angle between antiferromagnetic sublattices.

Both, ΔD\Delta D and Δd\Delta d exhibit quadratic dependence on the laser fluence for photons at E=1.2E = 1.2 eV [Figure 8(b)]. Analogous conclusion can be drawn from the data obtained at the pump photon energies in the range of intermediate absorption [Figure 7(e)]. At E=3.0E = 3.0 eV, ΔD\Delta D, Δd\Delta d exhibit linear dependences based on the corresponding dependence of ΔθL(F)\Delta\theta_{\mathrm{L}}(F) [Figure 7(e)]. This unambiguously reveals that the CT transition from 2p2p states of O2−\mathrm{O}^{2-} to 3d3d states of Fe3+\mathrm{Fe}^{3+}, namely t2g(π)→t2gt_{2g}(\pi) \rightarrow t_{2g} centered at 3.03.0 eV [31], is required to enable the change of these parameters. Indeed, at E=1.2E = 1.2; 1.3−1.851.3-1.85 eV such transition requires a two-photon absorption [Figure 2(b)], which is in a full agreement with the fluence dependence ∝F2\propto F^{2}. At E=3.0E = 3.0 eV the pump pulses are resonant with the CT transitions [Figure 1(b)], and the one-photon absorption takes place. Also, pump fluences required to reach the same values of ΔD\Delta D, Δd\Delta d and ΔMa∝ΔθL\Delta M^{a} \propto\Delta\theta_{\mathrm{L}} are an order of magnitude lower for the direct excitation at E=3.0E = 3.0 eV as compared to the two-photon absorption.

We observe no distinct features in the spectral dependence of ΔθL\Delta\theta_{\mathrm{L}} [Figure 7(b)] that could be attributed to electronic 3d−3d3d-3d transitions centered at around 1.51.5 eV. Their possible contribution to the magnetic response appears to be strongly suppressed compared to the dominant two-photon absorption process, and therefore does not produce a discernible feature in the measured signals. On the one hand, d−dd-d transitions change the electrons distribution within 3d3d orbitals of Fe3+\mathrm{Fe}^{3+} affecting its effective orbital momentum, orbitals overlap with O2−\mathrm{O}^{2-}. This can change both the DMI parameter DD, as well as the single-ion parameter dd. On the other hand, CT transitions also change the number of electrons on 3d3d orbitals of Fe3+\mathrm{Fe}^{3+} thus impacting DMI and single-ion anisotropy though altering overlap with oxygen orbitals, effective orbital momentum, and magnetic moment. Our results of spectral and fluence dependences of ΔθL\Delta\theta_{\mathrm{L}} clearly distinguishes between the two pathways to alter DD and dd and shows the leading role of the CT transition at least in the regime of strong excitation.

It is worth noting that the fluences used in our experiments in the low and moderate absorption are of order of fractions of J⋅cm−2\mathrm{J}\cdot\mathrm{cm}^{-2}, that is considerably higher than the typical fluences used in most of experiments on laser-driven change of spin canting. Using such fluences is enabled by a high initial temperature in the experiments leading to a large heat capacity and, as a result, relatively low laser-induced heating. We argue that high fluences facilitated reaching large values of laser-induced modulation of DD and dd. It also allowed dominating contribution from the two-photon absorption [53, 54] and CT-based mechanism across the whole studied range of the pump pulse photon energies

The evolution of the spin canting triggered by the change of D,dD,d unfolds on a picosecond timescale. Δθ\Delta\theta reaches ΔθL\Delta\theta_{\mathrm{L}} within 3-6 ps depending on the pump fluence and initial temperature [Figure 5(a) and 6(a)]. This timescale agrees with the softening of the quasi-antiferromagnetic resonance (qAFMR) mode, which decreases from ≈300\approx300 GHz at T0=385T_0 = 385 K [49]. Given that such a softening can be induced by a reduction of da=D+dd_a = D + d, we suggest that the spin canting evolves via a single cycle of strongly damped qAFMR [26]. This implies that the characteristic time for the decrease of DD, dd is comparable to or shorter than 3–6 ps. Δθ\Delta\theta transients obtained at F<FthF < F_{\mathrm{th}} at T0<TSRT_0 < T_{\mathrm{SR}} [Fig. 5(a)] and at all fluences used in our study at T0>TSRT_0 > T_{\mathrm{SR}} [Fig. 6(a)] suggests that the laser-induced change of DD, dd persists on a time scale of beyond 100 ps. Such long timescale is justified by the fact that the CT transitions are responsible and the corresponding excited state may have a long relaxation times. Notably, such a long timescale of ΔD\Delta D, Δd\Delta d is in agreement with recent results on transient change of the DMI parameter in a weak ferromagnet ErFeO3_3 [28], obtained by studying qAFMR frequency under excitations by laser pulses with E=1.55E = 1.55 eV and a fluence below that used in our experiments.

Phase boundaries for the laser-induced SR transition

Associating the initial stage of the magnetization evolution to the laser-induced spin canting enables us to comprehend the apparent bottle-neck for the transient SR transition in Fe3_3BO6_6, i.e. laser-induced heating being insufficient for complete transition. We recall that at equilibrium, SR transition involves a change of the weak ferromagnetic moment due to a change of the canting of the sublattice magnetizations [Eq. (4)]. Based on the phenomenological approach outlined in Sec. II A, we now define the SR boundaries [Eq. (7)] in the case of the laser-induced transition. As established in our experiments, right after the excitation the magnetization remains aligned along the aa-axis. Thus, during the latency, the energy profile is determined by the magnitude of Ma=M0maM^a = M_0m^a. As changes of M0M_0 are supposed to be small, we focus on the effect of mam^a. To trigger the precession, the minimum corresponding to the Γ4\Gamma_4 should emerge [Fig. 1(d)]. In particular, such local minimum require K1(ma)≥0K_1(m^a) \ge0, that is analogous to the condition at equilibrium. Once this condition is satisfied via laser-induced change of parameters entering K1K_1, partial switching to Γ4\Gamma_4 phase is possible via precession, and 0≤NSW≤10 \le N^{\mathrm{SW}} \le1. However, the complete switching requires K1(ma)+2K2(ma)≥0K_1(m^a) + 2K_2(m^a) \ge0, that is different from the equilibrium condition. As a result, the effective temperature TFT_F reached after the excitation does not serve as a proper merit for the laser-induced SR transition, as evident from the dependence of NSWN^{\mathrm{SW}} [Fig. 5(c)] and from the estimated heating TFT_F [Fig. 5(e)]. Condition K1(ma)+2K2(ma)≥0K_1(m^a) + 2K_2(m^a) \ge0 is now controlled by the spin canting achieved via the change of DD, dd.

To relate the boundary conditions to the experimental values ΔθL∝ΔMa\Delta\theta_L \propto\Delta M^a, we introduce a magnetization mismatch parameter ΔM\Delta_M characterizing a difference between the magnetization MScM_S^c in the equilibrium Γ4\Gamma_4 phase and [MSa+ΔMa(F)][M_S^a + \Delta M^a(F)] at t≲tSRt \lesssim t_{\mathrm{SR}}. This difference is multiplied by VlV^l that enables evaluation of ΔM\Delta_M based on the measured equilibrium and transient MOKE values. Thus, we obtain:

ΔM=θSl−VlVP⋅cos⁡β[θSp+ΔθL].(15)\Delta_M = \theta_S^l - \frac{V^l}{V_P \cdot\cos\beta}\left[\theta_S^p + \Delta\theta_L\right]. \tag*{(15)}

This fluence dependence is plotted in Figure 9(a) for the Sample A and the pump photon energy E=1.2E = 1.2 eV. We chose this set of data as the one with minimal laser-induced heating due to the lowest absorption [Fig. 1(b)]. The magnetization mismatch ΔMth\Delta_M^{\mathrm{th}} at FthF_{\mathrm{th}} is found to be equal to the discontinuity of the equilibrium magneto-optical signal at TSRT_{\mathrm{SR}} [Fig. 2(e)] expressed according to Eq. (15). The complete laser-induced SR transition (NSW=1N^{\mathrm{SW}} = 1) is realized when ΔM=0\Delta_M = 0, i.e. [MSa+ΔMa(F)][M_S^a + \Delta M^a(F)] matches MScM_S^c. Concomitantly, the latency τL\tau_L reduces to zero [Fig. 5(e)]. This fully agrees with the defined above boundaries for the laser-induced transition.

Fluence, temperature, and pump photon energy dependences of $\Delta M$ and switched fraction $N^{\mathrm{SW}}$, phase diagram, barrier dependence, and schematic energy profiles

Figure 9. (a) Fluence, (b) temperature, and (c) pump photon energy dependences of ΔM\Delta M during the latency τL\tau_{\mathrm{L}} and of the switched fraction NSWN^{\mathrm{SW}}. Data in panels (a) and (b) are obtained at E=1.2E=1.2 eV at (a) T0=385T_0=385 K and (b) F=0.5F=0.5 mJ·cm−2^{-2}. Data in panel (c) are obtained at T0=385T_0=385 K, for the same fluences as in Figure 7(b). (d) Phase diagram t−ΔMt-\Delta M of the laser-induced SR transition based on the fluence dependence at T0=385T_0=385 K with the color-coded switched fraction NSWN^{\mathrm{SW}}. Open and closed symbols mark the start and the end (tSRt_{\mathrm{SR}}) of the latency. (e) 2D plot of NSWN^{\mathrm{SW}} vs. barrier ΔΦ\Delta\Phi and the final temperature TFT_{\mathrm{F}} as obtained from the fluence dependences at T0=385T_0=385 (circles) and 365 K (triangles) and the temperature dependences at F=0.5F=0.5 (diamonds) and 0.35 J·cm−1^{-1} (pentagons). (f) Schematic illustration of evolution of the energy profile Φ(φ)\Phi(\varphi) following laser excitation and the corresponding values of magnetization ma(c)m^{a(c)}.

Switching scenario and the role of laser-induced heating

The analysis demonstrates that the change of spin canting by a laser pulse with the fluence above FthF_{\mathrm{th}} prepares the state with the two-minima potential, corresponding to K1>0K_1 > 0 and K1+2K2<0K_1 + 2K_2 < 0 [Fig. 1(d)]. For the switching of the magnetization from the aa-axis (φ=0\varphi= 0) to the cc-axis (φ=90∘\varphi= 90^\circ), the barrier between the minima should overcome. The height of the barrier ΔΦ\Delta\Phi derived from Eq. (5) is:

ΔΦ=−K124K2−K1−K2−μ0M0mH1+K12K2(16)\Delta\Phi= -\frac{K_1^2}{4K_2} - K_1 - K_2 - \mu_0M_0mH\sqrt{1+\frac{K_1}{2K_2}} \tag*{(16)}

and is controlled mainly by ΔM∝Ma\Delta_M \propto M^a via K1K_1 and by an applied magnetic field. The barrier height spans from −K2−μ0M0H-K_2 - \mu_0M_0H to 00 as K1K_1 takes a values corresponding to the two phase boundaries. Following the Néel-Brown model for a macrospin switching [55], the ratio between ΔΦ\Delta\Phi and the thermal energy kBTFk_BT_F defines the probability NSW=1−exp⁡(−t/τL)N^{\mathrm{SW}} = 1-\exp(-t/\tau_L) and a characteristic time required for switching between the two minima. Here we take the latency τL\tau_L as a characteristic time and express it as

τL∝exp⁡[ΔΦkBTF].(17)\tau_L \propto\exp\left[\frac{\Delta\Phi}{k_BT_F}\right]. \tag*{(17)}

In particular, the growth rate of NSWN^{\mathrm{SW}} with increase of ΔM\Delta_M should change when the range of the corresponding TFT_F is changed. This is realized in an experiment performed at various initial temperatures T0T_0 and fixed fluence F=0.5F = 0.5 J⋅\cdotcm−1^{-1} for the Sample A [Fig. 9(b)]. Since ΔθL\Delta\theta_L is of non-thermal origin, its value is defined by the fluence and only weakly depends on T0T_0 [Fig. 6(b)]. Correspondingly, the magnetization mismatch ΔM\Delta_M is weakly altered upon increase of T0T_0 [Fig. 9(b)], while TFT_{\mathrm{F}} shows a pronounced linear growth with T0T_0. As a result NSWN^{\mathrm{SW}} shows a pronounced increase with T0T_0 and complete switching is realized at nonzero magnetization mismatch [Fig. 9(b)]. This thermally activated switching is further confirmed by the data obtained at different pump photon energies E=3.0E=3.0 eV. Figure 9(c) sums up the values of ΔM\Delta M and NSWN^{\mathrm{SW}} obtained at T0=385T_0=385 K at different pump photon energies with fluences chosen in a way that the values of ΔM\Delta M are the same at E=3.0E=3.0 eV and 1.85 eV. Evidently, at E=3.0E=3.0 eV, strong absorption and higher TFT_{\mathrm{F}} lead to larger NSWN^{\mathrm{SW}} at the same ΔM\Delta M, as compared to the results obtained at medium or low absorption.

To show the agreement between the temperature and fluence dependences, we use Eq. (17) that accounts for the variations of τL\tau_{\mathrm{L}} with fluence and temperature. The sets of τL\tau_{\mathrm{L}} and TFT_{\mathrm{F}}, along with NSWN^{\mathrm{SW}}, were obtained from the fluence dependences measured at T0=365T_0=365 and 385 K and from the temperature dependences at F=0.35F=0.35 and 0.5 J·cm−2^{-2} for the Sample A. Then ΔΦ\Delta\Phi was evaluated based on Eq. (17). In Figure 9(e) we plot resulting dependence of NSWN^{\mathrm{SW}} on Φ\Phi and TFT_{\mathrm{F}} showing agreement between different data sets.

Suggested switching scenario is also supported by the decrease of τL\tau_{\mathrm{L}} with an applied magnetic field [Fig. 5(e)]. Indeed, as the field is applied along the cc-axis, it reduces the barrier between the two minima [Eq. (16)] leading to a decrease of τL\tau_{\mathrm{L}} and and increase of ΔθF\Delta\theta_{\mathrm{F}} [Fig. 5(f)], i.e. an increase of the switched fraction NSW∝ΔθFN^{\mathrm{SW}}\propto\Delta\theta_{\mathrm{F}}.

Coexistence of 2 and 4 phases

The precession of magnetization that takes place during the switching demonstrates a peculiar character, with beatings, that are most evident when the pump photon energy is of 3.0 eV [Fig. 7(a)]. Such a feature was reported earlier for, e.g., the THz-pulse-driven switching in orthoferrite TmFeO3_3 via the 2nd order SR transition [21] and was attributed to the competing contributions from magnetic precession in non-perturbative and perturbative regimes. While the switching mechanism considered here is different and the SR transition is of the 1st order, the observed temporal behavior can also be understood along similar lines. Indeed, two frequencies indicate that there are at least two magnetization trajectories corresponding to different potentials. In our experimental geometry for the Sample A, magnetization precession is detected only around the potential minimum with Φ=π/2\Phi=\pi/2 [Fig. 1(d)]. Sample AB potentially allows detection of the precession around the initial potential minimum with Φ=0\Phi=0. However, a close resemblance between the transients obtained under the same conditions for the two samples [Fig. 3(a,b)] suggests that it is sufficient to consider only the magnetization precession around Φ=π/2\Phi=\pi/2.

As was established in Sec.IVB, the frequency f2f_2 agrees well with the qFMR frequency at elevated temperature in the Γ4\Gamma_4 phase. Frequency f1f_1 is lower and signifies the precession in the shallower minimum. Based on this, we conjecture that following the thermally activated switching over the barrier, two precession trajectories are available, as schematically shown in Figure 9(f). The first one corresponds to the local minimum at Φ=π/2\Phi=\pi/2 in the potential supporting Γ2+Γ4\Gamma_2+\Gamma_4. The precession frequency is f1f_1. It is defined by D+ΔDD+\Delta D, d+Δdd+\Delta d. Accordingly, net magnetization is ∝[D+ΔD−(d+Δd)]\propto[D+\Delta D-(d+\Delta d)], and is lower than the equilibrium magnetization in Γ4\Gamma_4 phase, as depicted in lower panels in Figure 9(f). The second trajectory corresponds to the global minimum in the potential supporting only Γ4\Gamma_4 phase, with equilibrium qFMR frequency f2f_2.

Notably, the lower-frequency precession emerges at somewhat lower fluences than the higher-frequency one. This suggests that concomitant observation of the two precession trajectories is enabled by the inhomogeneous spatial pump excitation profile. The fact that the beating is much more pronounced under the excitation with the pump pulses with E=3.0E = 3.0 eV can be a result of different excitation pathways. At this pump photon energy, both the change of DD, dd and the heating are due to the one-photon absorption processes. As a result, both ΔD\Delta D, Δd\Delta d and of TFT_{\mathrm{F}} possess similar spatial distribution. At other pump photon energies, ΔD\Delta D, Δd\Delta d relies of the two-photon absorption, and their spatial profile in narrower as compared to the profile of TFT_{\mathrm{F}}.

These features of the precessional switching show that laser pulse excitation brings Fe3_3BO6_6 from the initial Γ2\Gamma_2 phase to a state with coexisting Γ2′\Gamma_2' and Γ4′\Gamma_4' phases. Magnetization mam^a in the Γ2′\Gamma_2' phase does not correspond to the that in the equilibrium Γ2\Gamma_2 phase. Analogously, mcm^c in the Γ4′\Gamma_4' phase differs from the equilibrium value as evident from the presence of the precession at f1f_1. Interestingly, these phases are not realized at equilibrium and can be considered as hidden ones. At the final state, following the damping of the precession, the values of ΔθL\Delta\theta_{\mathrm{L}} yield that there is Γ2′+Γ4\Gamma_2' + \Gamma_4 phase conxistance [right column in Figure 9(f)].

Conclusion

In conclusions, we have unraveled kinetics of the laser-induced spin-reorientation phase transition Γ2→Γ4\Gamma_2 \to\Gamma_4 in the weak ferromagnet Fe3_3BO6_6. The results demonstrate that the laser-induced spin reorientation in Fe3_3BO6_6 cannot be described as a purely heating-driven phase transition. Instead, we suggest and analyze the scenario when the laser excitation first modifies the weak ferromagnetic state through a transient reduction of the canting angle with the dominant contribution originating from a change of the DMI parameter, as well as the single-ion contribution. This modification of these both contributions creates a transient magnetic state with a two-minimum energy potential. The reorientation proceeds through thermally-assisted crossing of the barrier separating the two orientational magnetic states Γ2\Gamma_2 and Γ4\Gamma_4. The demonstrated separation between the ultrafast preparation of the magnetic state and the subsequent switching provides a framework for understanding and controlling nonequilibrium spin-reorientation dynamics. The dependence of the transient magnetic state on photon energy and fluence, together with the control of the switching probability and latency by temperature and magnetic field, provides several independent parameters for manipulating the reorientation kinetics.

Our findings regarding the laser-induced change of spin canting add an important contribution to a puzzle of laser-driven change of competing DMI and isotropic exchange parameters in iron oxides exhibiting weak ferromagnetism [26, 27, 28, 29, 56], when two underlying mechanisms have been put forward based on changing the electronic configuration of Fe3+^{3+} ions either through the CT transitions or localized d−dd-d transitions. We demonstrate that, in the regime of strong laser-induced excitation, it is the CT transition that leads to reduced DMI and single-ion-like contributions to weak ferromagnetism by ∼40\sim40 %. This regime is realized by combining a high base temperature of the sample and the excitation in the spectral range of low or moderate absorption. Such conditions are readily available in a number of weak ferromagnetic iron oxides possessing ordering temperatures well above room temperature. Finally, we note that a recent study of the related orthoferrite DyFeO3_3 suggested a laser-induced suppression of the intersublattice isotropic exchange interaction by nearly 90 % driven by CT transitions [29], that clearly is in contrast to our findings. The effect was confined to a near surface region and probed through magnon spectrum. Thus, joint consideration of recent reports raises several interesting questions, such as whether mechanisms of laser-induced DMI and exchange perturbations strongly differ even for related iron oxides, whether surface and bulk magnetic states evolve distinctly under laser excitation, and whether different probing approaches demonstrate sensitivity to different processes.

Acknowledgments

We thank L. A. Shelukhin for useful discussions. We thank L. N. Bezmaterynkh for providing the Sample AB. The work of A.V.K. and A.M.K was supported by the Russian Science Foundation (grant No. 26-12-00401). The work of S.N.B. was supported by the Belarusian Republican Foundation for Fundamental Research (grant No. 25KI-070).

References

References

  1. [1]S. Koshihara, T. Ishikawa, Y. Okimoto, K. Onda, R. Fukaya, M. Hada, Y. Hayashi, S. Ishihara, and T. Luty, Challenges for developing photo-induced phase transition (PIPT) systems: From classical (incoherent) to quantum (coherent) control of PIPT dynamics, Physics Reports 942, 1 (2022).
  2. [2]D. Fausti, R. I. Tobey, N. Dean, S. Kaiser, A. Dienst, M. C. Hoffmann, S. Pyon, T. Takayama, H. Takagi, and A. Cavalleri, Light-induced superconductivity in a stripe-ordered cuprate, Science 331, 189 (2011), https://www.science.org/doi/pdf/10.1126/science.1197294.DOI
  3. [3]S. de Jong, R. Kukreja, C. Trabant, N. Pontius, C. F. Chang, T. Kachel, M. Beye, F. Sorgenfrei, C. H. Back, B. Bräuer, et al., Speed limit of the insulator–metal transition in magnetite, Nature Materials 12, 882 (2013).
  4. [4]D. Wegkamp and J. Stähler, Ultrafast dynamics during the photoinduced phase transition in vo2, Progress in Surface Science 90, 464 (2015).DOI
  5. [5]G. Li, R. Medapalli, J. H. Mentink, R. V. Mikhaylovskiy, T. Blank, S. K. K. Patel, A. K. Zvezdin, T. Rasing, E. E. Fullerton, and A. V. Kimel, Ultrafast kinetics of the antiferromagnetic-ferromagnetic phase transition in FeRh, Nature Communications 13, 2998 (2022).DOI
  6. [6]J. A. de Jong, I. Razdolski, A. M. Kalashnikova, R. V. Pisarev, A. M. Balbashov, A. Kirilyuk, T. Rasing, and A. V. Kimel, Coherent control of the route of an ultrafast magnetic phase transition via low-amplitude spin precession, Physical Review Letters 108, 157601 (2012).pubmed.ncbi.nlm.nih.gov/22587282
  7. [7]T. Kurihara, H. Watanabe, M. Nakajima, S. Karube, K. Oto, Y. Otani, and T. Suemoto, Macroscopic magnetization control by symmetry breaking of photoinduced spin reorientation with intense terahertz magnetic near field, Phys. Rev. Lett. 120, 107202 (2018).DOI
  8. [8]H. Ichikawa, S. Nozawa, T. Sato, A. Tomita, K. Ichiyanagi, M. Chollet, L. Guerin, N. Dean, A. Cavalleri, S.-i. Adachi, et al., Transient photoinduced ‘hidden’ phase in a manganite, Nature Materials 10, 101 (2011).DOI
  9. [9]J. Li, M. Turaev, M. Matsubara, K. Kliemt, C. Krellner, S. Pal, M. Fiebig, and J. Kroha, Discovery of an intrinsic non-hermitian phase transition in a bulk condensed-matter system, Science 393, 1152 (2026).DOI
  10. [10]S. O. Mariager, F. Pressacco, G. Ingold, A. Caviezel, E. Möhr-Vorob’eva, P. Beaud, S. L. Johnson, C. J. Milne, E. Mancini, S. Moyerman, et al., Structural and magnetic dynamics of a laser induced phase transition in FeRh, Physical Review Letters 108, 087201 (2012).pubmed.ncbi.nlm.nih.gov/22463562
  11. [11]I. Radu, K. Vahaplar, C. Stamm, T. Kachel, N. Pontius, H. A. Dürr, T. A. Ostler, J. Barker, R. F. L. Evans, R. W. Chantrell, et al., Transient ferromagnetic-like state mediating ultrafast reversal of antiferromagnetically coupled spins, Nature 472, 205 (2011).DOI
  12. [12]A. V. Kimel, A. Kirilyuk, A. Tsvekov, R. V. Pisarev, and T. Rasing, Laser-induced ultrafast spin reorientation in the antiferromagnet TmFeO3, Nature 429, 850 (2004).pubmed.ncbi.nlm.nih.gov/15215858
  13. [13]K. P. Belov, A. K. Zvezdin, A. M. Kadomtseva, and R. Z. Levitin, Spin-reorientation transitions in rare-earth magnets, Soviet Physics Uspekhi 19, 574 (1976).
  14. [14]O. N. Myasov, U. Nowak, K. Y. Guslienko, and R. W. Chantrell, Temperature-dependent magnetic properties of FePt: Effective spin hamiltonian model, Europhysics Letters 69, 805 (2005).
  15. [15]R. L. Stamps, Mechanisms for exchange bias, Journal of Physics D: Applied Physics 33, R247 (2000).
  16. [16]A. Moskvin, Dzyaloshinskii-Moriya coupling in 3d insulators, Condensed Matter 4, 84 (2019).DOI
  17. [17]V. V. Mazurenko, Y. O. Kvashnin, A. I. Lichtenstein, and M. I. Katsnelson, A Dzyaloshinskii-Moriya interaction guide to magnetics micro-world, J. Exp. Theor. Phys. 132, 506 (2021).DOI
  18. [18]L. A. Shelukhin, R. R. Gareev, V. Zbarsky, J. Walowski, M. Münzenberg, N. A. Pertsev, and A. M. Kalashnikova, Spin reorientation transition in CoFeB/MgO/CoFeB tunnel junction enabled by ultrafast laser-induced suppression of perpendicular magnetic anisotropy, Nanoscale 14, 8153 (2022).DOI
  19. [19]J. A. de Jong, A. V. Kimel, R. V. Pisarev, A. Kirilyuk, and T. Rasing, Laser-induced ultrafast spin dynamics in ErFeO3, Physical Review B—Condensed Matter and Materials Physics 84, 104421 (2011).
  20. [20]D. Afanasiev, J. R. Hortensius, B. A. Ivanov, A. Sasani, E. Bousquet, Y. M. Blanter, R. V. Mikhaylovskiy, A. V. Kimel, and A. D. Caviglia, Ultrafast control of magnetic interactions via light-driven phonons, Nature Materials 20, 607 (2021).DOI
  21. [21]S. Schlauderer, C. Lange, S. Baierl, T. Ebnet, C. P. Schmid, D. C. Valovcin, A. K. Zvezdin, A. V. Kimel, R. V. Mikhaylovskiy, and R. Huber, Temporal and spectral fingerprints of ultrafast all-coherent spin switching, Nature 569, 383 (2019).
  22. [22]M. A. Weiss, A. Herbst, J. Schlegel, T. Dannegger, M. Evers, A. Donges, M. Nakajima, A. Leitenstorfer, S. Goennenwein, U. Nowak, et al., Discovery of ultrafast spontaneous spin switching in an antiferromagnet by femtosecond noise correlation spectroscopy, Nature Communications 14, 7651 (2023).DOI
  23. [23]I. E. Dzyaloshinsky, A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics, Journal of Physics and Chemistry of Solids 4, 241 (1958).DOI
  24. [24]T. Moriya, Anisotropic superexchange interaction and weak ferromagnetism, Physical Review 120, 91 (1960).DOI
  25. [25]F. J. Morin, Magnetic susceptibility of α-Fe2O3 and α-Fe2O3 with added titanium, Physical Review 78, 819 (1950).DOI
  26. [26]R. V. Mikhaylovskiy, E. Hendry, A. Secchi, J. H. Mentink, M. Eckstein, A. Wu, R. V. Pisarev, V. V. Kruglyak, M. I. Katsnelson, T. Rasing, et al., Ultrafast optical modification of exchange interactions in iron oxides, Nature Communications 6, 8190 (2015).DOI
  27. [27]R. V. Mikhaylovskiy, T. J. Huisman, V. A. Gavrichkov, S. I. Polukeev, S. G. Ovchinnikov, D. Afanasiev, R. V. Pisarev, T. Rasing, and A. V. Kimel, Resonant pumping of d- d crystal field electronic transitions as a mechanism of ultrafast optical control of the exchange interactions in iron oxides, Physical Review Letters 125, 157201 (2020).DOI
  28. [28]P. M. P. van Kuppevelt and A. V. Kimel, Laser-induced dynamics of the magnon spectrum probed by double-pump THz spectroscopy, Applied Physics Letters 129, 132405 (2026).
  29. [29]V. Radovskaia, R. Andrei, J. R. Hortensius, R. V. Mikhaylovskiy, R. Citro, S. Chattopadhyay, M. Na, B. A. Ivanov, E. Demler, A. V. Kimel, et al., Photoengineering the magnon spectrum in an insulating antiferromagnet, Nature Physics , 1 (2026).DOI
  30. [30]A. V. Kuzikova, N. A. Liubachko, S. N. Barilo, A. V. Sadovnikov, R. V. Pisarev, and A. M. Kalashnikova, Switching of an antiferromagnet controlled by spin canting in a laser-induced hidden phase, Physical Review Letters 135, 266705 (2025).DOI
  31. [31]R. V. Pisarev, A. S. Moskvin, A. M. Kalashnikova, and T. Rasing, Charge transfer transitions in multiferroic BiFeO3 and related ferrite insulators, Physical Review B Condensed Matter and Materials Physics 79, 235128 (2009).DOI
  32. [32]C. Voigt and D. Bonnenberg, Temperature dependence of the magnetic susceptibility of Fe3BO6, Physica B+C 80, 439 (1975).DOI
  33. [33]J. White, A. Miller, and R. Nielsen, Fe3bo6, a borate isostructural with the mineral norbergite, Acta Crystallographica 19, 1060 (1965).
  34. [34]V. I. Ozhogin, V. G. Shapiro, K. G. Gurtovoi, E. A. Galst’yan, and A. Y. Chervonenkis, Statics and linear dynamics of orthoferrites-phase transition of the ”one-and-one-half kind”, Soviet Physics JETP 35, 1162 (1972).
  35. [35]E. A. Turov, A. B. Kolchanov, V. V. Men’shenin, I. F. Mirsaev, and V. V. Nikolaev, Symmetry and physical properties of antiferromagnets (FizMatLit, Moscow, 2001 (in Russian)).
  36. [36]S. N. Martynov, Single-ion mechanism of weak ferromagnetism and a spin-flop transition in one-and two-position antiferromagnets, JETP Letters 108, 196 (2018).DOI
  37. [37]E. A. Turov et al., Physical properties of magnetically ordered crystals, (No Title) (1965).
  38. [38]A. S. Kamzin and L. A. Grigor’ev, Mössbauer investigation of the surface and volume spin-reorientation phase transition in Fe₃BO₆, JETP Lett 57, 552 (1993).
  39. [39]L. T. Tsymbal, Y. B. Bazaliy, L. N. Bezmaternykh, A. Slawska-Waniewska, S. V. Vasiliev, N. Nedelko, A. I. Linnik, A. N. Cherkasov, Y. I. Nepoсhatykh, V. Y. Dmitrenko, et al., Orientation phase transition in Fe₃BO₆: Experimental determination of the order of the transition, Physical Review B 74, 134429 (2006).
  40. [40]I. E. Dzyaloshinsky, A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics, Journal of Physics and Chemistry of Solids 4, 241 (1958).DOI
  41. [41]I. E. Dzialoshinskii, The magnetic structure of fluorides and the transition metals, Sov. Phys. - JETP 6, 1120 (1958).
  42. [42]T. Moriya, Anisotropic superexchange interaction and weak ferromagnetism, Physical Review 120, 91 (1960).DOI
  43. [43]T. Moriya, Theory of magnetism of NiF₂, Physical Review 117, 635 (1960).
  44. [44]L. M. Matarrese and J. W. Stout, Magnetic anisotropy of NiF₂, Physical Review 94, 1792 (1954).
  45. [45]G. A. Petrakovski, M. A. Popov, A. D. Balaev, K. A. Sablina, O. A. Bayukov, D. A. Velikanov, A. M. Vorotynov, A. F. Bovina, A. D. Vasil’ev, and M. Boehm, Magnetic properties of Pb₂Fe₂Ge₂O₉ single crystals, Physics of the Solid State 51, 1853 (2009).DOI
  46. [46]D. Jo, D. Go, Y. Mokrousov, P. M. Oppenеer, S.-W. Cheong, and H.-W. Lee, Weak ferromagnetism in alter-magnets from alternating g-tensor anisotropy, Physical Review Letters 134, 196703 (2025).DOI
  47. [47]D. Treves, Magnetic studies of some orthoferrites, Physical Review 125, 1843 (1962).DOI
  48. [48]B. Andlauer and R. Diehl, Optical absorption of Fe₃BO₆ in the range of the transitions ⁶A₁ → ⁴T₁, ⁴T₂, Physica B+C 89, 50 (1977).DOI
  49. [49]V. E. Arutyunyan, K. N. Kocharyan, and R. M. Martirosyan, Magnetic resonance in the orthorhombic antiferromagnet Fe₃BO₆, Soviet Physics JETP 69, 783 (1989).
  50. [50]A. G. Gurevich and G. A. Melkov, Magnetization oscillations and waves (CRC press, 2020).DOI
  51. [51]R. Wolfe, R. D. Pierce, M. Eibschütz, and J. W. Nielsen, Magnetization and Mössbauer effect in single crystal Fe₃BO₆, Solid State Communications 7, 949 (1969).DOI
  52. [52]A. V. Kuzikova, L. A. Shelukhin, F. M. Maksimov, A. I. Chernov, R. V. Pisarev, and A. M. Kalashnikova, Laser-driven first-order spin reorientation and Verwey phase transitions in magnetite Fe₃O₄ beyond the range of thermodynamic equilibrium, Physical Review B 107, 024413 (2023).
  53. [53]S. P. Zeusсhner, J.-E. Pudell, A. von Reppert, M. Deb, E. Popova, N. Keller, M. Rössle, M. Herzog, and M. Bargheer, Measurement of transient strain induced by two-photon excitation, Physical Review Research 2, 022013 (2020).DOI
  54. [54]I. Chaban, R. Deska, G. Privaулt, E. Trzop, M. Lorenc, S. E. Kooi, K. A. Nelson, M. Samoc, K. Matczyszyn, and T. Pezeril, Nonlinear optical absorption in nanoscale films revealed through ultrafast acoustics, Nano Letters 22, 4362 (2022).DOI
  55. [55]W. F. Brown Jr, Thermal fluctuations of a single-domain particle, Physical Review 130, 1677 (1963).
  56. [56]V. A. Gavrichkov, S. I. Polukeev, and S. G. Ovchinnikov, Contribution from optically excited many-electron states to the superexchange interaction in Mott-Hubbard insulators, Physical Review B 95, 144424 (2017).

Paper details

vlnc:W_8s7xj2bf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents