betaFeedback

WorksPhysics and Astronomy

Terahertz spin conductance probes of coherent and incoherent spin tunneling through MgO tunnel junctions

Abstract

We study femtosecond spin currents through MgO tunneling barriers in CoFeB(2 nm)|MgO(dd)|Pt(2 nm) stacks by terahertz emission spectroscopy. To obtain transport information independent of extrinsic experimental factors, we determine the complex-valued spin conductance G~d(ω)\tilde{G}_d (ω) of the MgO layer (thickness d= 0-6 Å over a wide frequency range (ω/2π=(ω/2π= 0.5-8 THz). In the time (t)(t) domain,Gd(t) G_d (t) has an instantaneous and delayed component that point to (i) spin transport through Pt pinholes in MgO, (ii) coherent spin tunneling and (iii) incoherent resonant spin tunneling mediated by defect states in MgO. A remarkable signature of (iii) is its relaxation time that grows monotonically with dd to as much as 270 fs at d=6d= 6 Å, in full agreement with an analytical model. Our results indicate that terahertz spin conductance spectroscopy will yield new and relevant insights into ultrafast spin transport for a wide range of materials.

Motivation. Transport of electron spin angular momentum is of central importance for future spintronic devices. To keep pace with other information carriers such as electrons in field-effect transistors [1] and photons in optical fibers [2], spintronic transport needs to be pushed to terahertz (THz) bandwidth and, thus, femtosecond time scales. Interesting applications of ultrafast spin transport include emission of broadband THz pulses [3] and generation of spin torque [4] for ultrafast magnetization switching.

To better understand and optimize THz spin transport, we need tools to adequately characterize it. An important quantity is the spin conductance G=IS/ΔμSG = I_{\mathrm{S}}/\Delta\mu_{\mathrm{S}}, which quantifies how much spin current ISI_{\mathrm{S}} is obtained when a spin-voltage drop ΔμS\Delta\mu_{\mathrm{S}} is applied across a conductor Figure 1(a). We focus on longitudinal spin transport, which can be described by populations of spin-up and spin-down electron states. At frequencies ≪1\ll1 THz, spin transport measurements typically rely on contacts [5, 6, 7]. At THz frequencies, however, measurement procedures of GG still need to be developed.

Figure 1. Spin conductance spectroscopy of a layer XX. (a) A F∣X∣HF|X|H stack is excited by a femtosecond laser pulse. The resulting transient spin voltage μs(t)\mu_s(t) in the ferromagnetic layer FF (here CoFeB) drives a spin current jSj_S through XX (here MgO). The spin current arriving in the heavy metal layer HH (here Pt) is converted into an in-plane charge current jCj_C that is measured by detecting the THz pulse it emits. The dynamics of μS(t)\mu_S(t) is determined using a sample with d=0d=0, which exhibits a spectrally flat interface spin conductance grefg^{\mathrm{ref}}. (b) Considered spin-transport channels through MgO: flow through conducting pinholes (PH), coherent tunneling (CT) and incoherent resonant tunneling (IRT). For IRT, γFXe−z/λ\gamma_{FX}e^{-z/\lambda} is the probability of an electron to tunnel from F=CoFeBF=\mathrm{CoFeB} to a defect at position zz in X=MgOX=\mathrm{MgO}, whereas γXHe−(d−z)/λ\gamma_{XH}e^{-(d-z)/\lambda} refers to the tunneling probability from zz to H=PtH=\mathrm{Pt}.

In this work, we introduce an approach to measure the spin conductance of a thin film XX between a ferromagnetic metal layer FF and a heavy-metal layer HH Figure 1(a). Using THz emission spectroscopy [8], we obtain the complex-valued spin conductance G~(ω)\widetilde{G}(\omega) at frequencies ω/2π=0.5\omega/2\pi= 0.5-8 THz. In the time domain, G(t)G(t) vs time tt can simply be understood as the spin current that would be obtained for a δ(t)\delta(t)-like spin-voltage pulse. Our procedure is demonstrated for the ubiquitous tunnel-barrier material X=MgOX = \mathrm{MgO} and reveals dynamic signatures of coherent and incoherent spin tunneling. We expect that THz spin conductance spectroscopy will provide important insights into ultrafast spin transport in a wide range of materials.

THz spin-conductance spectroscopy. The general idea of THz spin-conductance spectroscopy is shown in Figure 1. It relies on a F∣X∣HF|X|H stack, where the ferromagnetic layer FF (here CoFeB) serves as spin-current source, XX is the layer under investigation, and the heavy-metal layer HH (here Pt) acts as detector. First, a femtosecond laser pulse induces an ultrafast generalized spin voltage μS(t)\mu_{\mathrm{S}}(t) in F=CoFeBF = \mathrm{CoFeB} [9, 10] that drives a spin current through X=MgOX = \mathrm{MgO}. Second, the spin current jS(t)j_{\mathrm{S}}(t) arriving in the F=PtF = \mathrm{Pt} detection layer is converted into an in-plane charge current jC(t)∝jS(t)j_{\mathrm{C}}(t) \propto j_{\mathrm{S}}(t) by the inverse spin Hall effect (ISHE). Third, jC(t)j_{\mathrm{C}}(t) gives rise to the emission of a measurable ultrashort THz electromagnetic pulse with electric field E(t)E(t) [3, 11].

In the frequency domain, the spin conductance of MgO is given by an Ohm-type law, j~S(ω)=G~d(ω)μ~S(ω)\widetilde{j}_{\mathrm{S}}(\omega) = \widetilde{G}_{d}(\omega)\widetilde{\mu}_{\mathrm{S}}(\omega), where dd is the thickness of XX. The spin voltage in Pt is neglected owing to the short spin lifetime in Pt [12]. To determine μ~S(ω)\widetilde{\mu}_{\mathrm{S}}(\omega), we conduct a reference measurement on a sample without interlayer (i.e., d=0d = 0), where j~Sref(ω)=G~ref(ω)μ~S(ω)\widetilde{j}_{\mathrm{S}}^{\mathrm{ref}}(\omega) = \widetilde{G}^{\mathrm{ref}}(\omega)\widetilde{\mu}_{\mathrm{S}}(\omega) with known G~ref\widetilde{G}^{\mathrm{ref}}.

In our experiment, we measure electrooptic signals S(t)S(t) and Sref(t)S^{\mathrm{ref}}(t) rather than the emitted THz electric field E(t)E(t). However, SS and EE are related by an instrument response function that cancels in the frequency domain when the reference measurement is considered [13]. As derived in Supplemental Note 1, the XX spin conductance G~d\widetilde{G}_{d} normalized to the frequency-independent reference conductance G~ref\widetilde{G}^{\mathrm{ref}} is fully determined by the observables S~\widetilde{S} and S~ref\widetilde{S}^{\mathrm{ref}} through

G~d(ω)G~ref=j~S(ω)j~Sref(ω)=S~(ω)S~ref(ω).(1)\frac{\widetilde{G}_{d}(\omega)}{\widetilde{G}^{\mathrm{ref}}} = \frac{\widetilde{j}_{\mathrm{S}}(\omega)}{\widetilde{j}_{\mathrm{S}}^{\mathrm{ref}}(\omega)} = \frac{\widetilde{S}(\omega)}{\widetilde{S}^{\mathrm{ref}}(\omega)}. \tag*{(1)}

In the last step of Eq. (1), we took advantage of the facts that (i) the measured optical absorbance and THz impedance of all the samples of our experiment are the same and, thus, cancel, and that (ii) the spin conductance G~ref(ω)=gref\widetilde{G}^{\mathrm{ref}}(\omega) = g^{\mathrm{ref}} of the F∣F|Pt interface is constant over the frequency interval considered here (see Supplemental Note 2).

Experimental details. As spin conductor XX, the nonmagnetic insulator MgO is chosen. It is extensively used as a tunnel barrier in magnetic tunnel junctions, which provide large tunneling magnetoresistance for applications in nonvolatile memory devices [14, 15, 16, 17].

The samples are sub|TaN(1.5 nm)|CoFeB(2 nm)|MgO(dd)|Pt(2 nm)|TaN(1.5 nm) stacks with various MgO thicknesses of 0≤d≤150 \le d \le15 Å. The sample with d=0d = 0 is the reference sample. All layers are grown on an MgO substrate (sub) by DC magnetron sputtering in ultrahigh vacuum at a base pressure of 4×10−34 \times10^{-3} mbar, except the MgO layer, which is grown in the same vacuum chamber by radio-frequency sputtering using an off-axis gun tilted 90∘90^\circ from the substrate plane [18]. Atomic force microscopy shows a roughness <2<2 Å of all layers [18].

In the THz emission experiments Figure 1(a), the FF magnetization M\mathbf{M} of F=CoFeBF = \mathrm{CoFeB} is saturated by an external magnetic field of about 10 mT. The sample is excited by a train of linearly polarized ultrashort laser pulses (central wavelength 790 nm, nominal pulse duration 10 fs, pulse energy 2 nJ, repetition rate 80 MHz) from a Ti:Sapphire laser oscillator, focused to a spot of 30 µm full width of half maximum (FWHM) of intensity. The emitted THz pulse is detected by electro-optic sampling in a GaP(110) crystal (thickness 250 µm) using linearly polarized probe pulses (0.6 nJ) from the same laser. This procedure yields the THz signal S(t)S(t) as a function of delay tt between probe and THz pulse [10].

THz signals. Figure 2(a) and Fig. S1 show THz emission signals S(t)S(t) odd in M\mathbf{M} from CoFeB|MgO|Pt stacks for various MgO thicknesses dd. Signal components even in M\mathbf{M} are minor.

Figure 2. THz-emission raw data from CoFeB(2 nm)|MgO(dd)|Pt(2 nm). (a) Time-domain electro-optic THz signals Sd(t)S_d(t) odd in the CoFeB magnetization M\mathbf{M} for various MgO thicknesses of d=0d = 0, 3.0, 4.5 and 6.0 Å. The signal for d=0d = 0 is the reference signal Sref=S∣d=0S^{\mathrm{ref}} = S|_{d=0}. (b) Fourier amplitude spectra normalized to their maximum of the signals shown in panel (a).

The signal for d=0d = 0 is known to be dominated by ultrafast spin transport from CoFeB into Pt with G~d=0(ω)=gcontrefω\widetilde{G}_{d=0}(\omega) = g^{\mathrm{ref}}_{\mathrm{con}t}\omega [9, 10, 11, 19]. When dd is increased from 0 Å to 6.0 Å, the signal amplitude decreases by a factor of 60 Figure 2(a). One can show that, in this range, the THz signal is dominated by spin transport from FF through MgO into HH and the ISHE in HH, consistent with the signal origin for d=0d = 0 (see Supplemental Material).

Interestingly, with increasing dd, the signal amplitude not only drops, but the initially sharp temporal features become smoother Figure 2(a). This trend suggests the emergence of slower components in the spin-current dynamics jS(t)j_{\mathrm{S}}(t). It is confirmed by the normalized Fourier spectra Figure 2(b), whose width decreases with increasing dd and whose maximum shifts to lower frequencies.

Frequency-domain spin conductance. To gain more intrinsic insight into the spin transport through MgO from the signals of Figure 2, we make use of Eq. (1) to determine the normalized spin conductance G~d(ω)/gref\widetilde{G}_{d}(\omega)/g^{\mathrm{ref}} as a function of frequency ω/2π=0.5\omega/2\pi= 0.5-8 THz. The modulus ∣G~d(ω)/gref∣\lvert\widetilde{G}_{d}(\omega)/g^{\mathrm{ref}}\rvert is displayed in Figure 3(a). It consists of a frequency-independent component and an additional contribution below 3 THz. As expected from Figure 2(a), we observe a drastic overall amplitude reduction with increasing dd. At the same time, spectral weight is shifted to frequencies below 3 THz. The spectral phase arg⁡G~d(ω)\arg\widetilde{G}_{d}(\omega) Figure 3(b) of the spin conductance and its slope vs ω\omega increase with dd, indicating an increasingly non-instantaneous response.

Figure 3. Frequency-domain THz spin conductance G~d(ω)\widetilde{G}_d(\omega) of MgO layers with various thicknesses dd. (a) Fourier amplitude ∣G~d(ω)/gref∣|\widetilde{G}_d(\omega)/g^{\mathrm{ref}}| vs frequency ω/2π\omega/2\pi, where gref=G~d=0(ω)g^{\mathrm{ref}} = \widetilde{G}_{d=0}(\omega) is spectrally flat. The dashed line shows the Fourier transformation δ~exp(ω)\widetilde{\delta}_{\mathrm{exp}}(\omega) of the experimental δ\delta-function. (b) Spectral phase arg⁡G~d(ω)\arg\widetilde{G}_d(\omega).

Time-domain spin conductance. To obtain the time-domain spin conductance Gd(t)G_{d}(t), we inversely Fourier-transform G~d(ω)\widetilde{G}_{d}(\omega). The resulting normalized Gd(t)/grefG_{d}(t)/g^{\mathrm{ref}} is shown in Figure 4(a) for all MgO thicknesses.

Figure 4. THz spin transport through MgO layers. (a) Measured time-domain THz spin conductance of MgO films for various thicknesses dd (thin solid lines). The spin conductance Gd(t)G_d(t) can be understood as the current that is driven through the MgO layer by a δ(t)\delta(t)-like spin voltage in the CoFeB layer. Due to the finite time resolution, the δ\delta-function is broadened to the fictitious impulsive spin voltage δexp(t)\delta_{\mathrm{exp}}(t) (dashed black line). Accordingly, the extracted conductivity curves are also broadened and display Gd∗δexpG_d \ast\delta_{\mathrm{exp}}. Thick solid lines are fits based on Eq. (3). (b) Amplitude Ad≈fdPHA_d \approx f_d^{\mathrm{PH}} of the instantaneous, i.e., δexp(t)\delta_{\mathrm{exp}}(t)-like component of Gd(t)G_d(t) vs MgO thickness dd (blue circles). (c) Amplitude Bd/(1−Ad)≈gd,0IRT/grefB_d/(1-A_d) \approx g_{d,0}^{\mathrm{IRT}}/g^{\mathrm{ref}} of the delayed current component vs MgO thickness (blue circles). The red solid line is a fit based on Eq. (5) with λ=(1.2±0.1) A˚\lambda=(1.2 \pm0.1)\,\text{\AA}. (d) Decay time τd\tau_d of the IRT current component vs MgO thickness (blue circles). The red solid line is a fit based on Eq. (5) with Γ−1=(50±5) fs\Gamma^{-1}=(50 \pm5)\,\mathrm{fs}.

Note that Gd(t)G_{d}(t) can be considered as the spin current through MgO that is induced by an impulsive spin voltage. Therefore, Gd(t)G_{d}(t) for d=0d = 0 would ideally equal a δ(t)\delta(t)-type signal. However, as the bandwidth of our setup is finite, the extracted Gd(t)G_{d}(t) is the response to a broadened δ\delta-like pulse δexp⁡(t)\delta_{\exp}(t). To achieve a unipolar δexp⁡(t)\delta_{\exp}(t), which is as short as possible and free of oscillations, we apply a suitable window function δ~exp⁡(ω)\widetilde{\delta}_{\exp}(\omega) to G~d(ω)/gref\widetilde{G}_{d}(\omega)/g^{\mathrm{ref}} prior to Fourier transformation [20]. Our δ~exp⁡(ω)\widetilde{\delta}_{\exp}(\omega) [dashed line in Figure 3(a)] approaches zero at 12 THz. The resulting δ(t)\delta(t)-type spin-voltage pulse δexp⁡(t)\delta_{\exp}(t) has a FWHM of 90 fs [dashed line in Figure 4(a)] and defines the time resolution of Gd(t)G_{d}(t).

For d>0d > 0, the extracted Gd(t)G_{d}(t) traces [Fig. 4(a)] consist of an initial instantaneous response with a shape similar to δexp⁡(t)\delta_{\exp}(t) plus a subsequently decaying component, whose relative weight increases with dd. This result is fully consistent with the spectra of Figure 3(a), where the broad background and feature below 3 THz correspond to the instantaneous and slower time-domain response [Fig. 4(a)].

Interpretation. Thin MgO films have been studied extensively in the past. For d<10d < 10 Å, structural imperfections are reported, in particular oxygen vacancies [21, 22, 23, 24, 25, 26] and pinholes connecting FF and HH [27, 28]. Accordingly, we consider 3 possible contributions to the total spin current through MgO in CoFeB|MgO|Pt stacks [Figure 1(b)]: (i) spin transport through Pt- or CoFeB-filled pinholes (PH) in the MgO film [28, 29], (ii) coherent off-resonant electron tunneling (CT) through the MgO tunnel barrier [30, 31, 32] and (iii) incoherent resonant tunneling (IRT) through intermediate defect states in the vicinity of the Fermi energy of the CoFeB and Pt layer [23, 26, 33, 34, 35]. As the currents (i)-(iii) add up independently [Figure 1(b)], the spin conductance is the sum

Gd=GdPH+GdCT+GdIRT.(2)G_d = G_d^{\mathrm{PH}} + G_d^{\mathrm{CT}} + G_d^{\mathrm{IRT}}. \tag*{(2)}

While the PH and CT processes are instantaneous [10, 36] on the time scale of our experimental resolution of 90 fs [dashed black line in Figure 4(a)], the IRT mechanism may require more time to proceed.

We briefly discuss each mechanism in more detail. Regarding GdPHG_d^{\mathrm{PH}}, one expects that, for d<6d < 6 Å, MgO grows in islands [27] or exhibits pinholes [28]. The pinholes are filled with Pt of the subsequently grown Pt layer and, thus, provide a conductive channel between the CoFeB and Pt layer [Figure 1(b)]. GdPHG_d^{\mathrm{PH}} depends on the in-plane areal fraction fdPHf_d^{\mathrm{PH}} of pinholes according to GdPH∝greffdPHG_d^{\mathrm{PH}} \propto g^{\mathrm{ref}} f_d^{\mathrm{PH}}. Thus, for fdPH=1f_d^{\mathrm{PH}} = 1, the reference situation (d=0d = 0) is recovered. Because the PH contribution is instantaneous analogous to the reference sample, we obtain GdPH(t)=fdPHgrefδ(t)G_d^{\mathrm{PH}}(t) = f_d^{\mathrm{PH}} g^{\mathrm{ref}} \delta(t).

CT [Figure 1(b)] through the entire MgO barrier is instantaneous on the scale of our time resolution [36]. Therefore, GdCT(t)=gdCT(1−fdPH)δ(t)G_d^{\mathrm{CT}}(t) = g_d^{\mathrm{CT}} (1 - f_d^{\mathrm{PH}}) \delta(t), where the coefficient gdCTg_d^{\mathrm{CT}} is the amplitude of the impulsive spin current in the absence of pinholes.

Regarding IRT, oxygen vacancies are known to provide localized electronic states within the MgO band gap and, thus, open up a resonant transport channel [21, 24, 25]. In the simplest IRT realization, an electron tunnels from FF into a MgO vacancy and, subsequently, into HH [Figure 1(b)], similar to resonant tunneling in quantum wells [37, 38, 39, 40]. One can quantify the resonant tunneling as GdIRT(t)=gdIRT(t)(1−fdPH)G_d^{\mathrm{IRT}}(t) = g_d^{\mathrm{IRT}}(t)(1 - f_d^{\mathrm{PH}}), where gdIRT(t)g_d^{\mathrm{IRT}}(t) is the IRT-related spin conductance of an MgO barrier without pinholes. Based on Figure 1(b), we model gdIRT(t)g_d^{\mathrm{IRT}}(t) by a single-sided exponential decay gd,0IRTΘ(t)e−t/τdg_{d,0}^{\mathrm{IRT}}\Theta(t)e^{-t/\tau_d}, where Θ(t)\Theta(t) is the Heaviside step function and τd\tau_d can be considered as the characteristic time of IRT.

With these specifications, Eq. (2) turns into

Gd(t)gref=Adδ(t)+BdΘ(t)e−t/τd,(3)\frac{G_d(t)}{g^{\mathrm{ref}}} = A_d \delta(t) + B_d \Theta(t)e^{-t/\tau_d}, \tag*{(3)}

where

Ad=fdPH+(1−fdPH)gdCTgref,Bd=(1−fdPH)gd,0IRTgref.(4)A_d = f_d^{\mathrm{PH}} + (1 - f_d^{\mathrm{PH}})\frac{g_d^{\mathrm{CT}}}{g^{\mathrm{ref}}}, \qquad B_d = (1 - f_d^{\mathrm{PH}})\frac{g_{d,0}^{\mathrm{IRT}}}{g^{\mathrm{ref}}}. \tag*{(4)}

We fit the convolution of GdG_d [Eq. (3)] and our time resolution δexp\delta_{\mathrm{exp}} to the time-domain data of Figure 4(a), where τd\tau_d, AdA_d, BdB_d and possible deviations t0t_0 from time zero [i.e., t→t−t0t \to t - t_0 in Eq. (3)] due to substrate thickness variations [41] are fit parameters (see Supplemental Material).

The model fits [Figure 4(a)] describe our data excellently. The resulting amplitude AdA_d of the δ(t)\delta(t)-like contribution [Figure 4(b)] decreases monotonically with MgO thickness dd. One can show that Ad≈fdPHA_d \approx f_d^{\mathrm{PH}} (see Supplemental Material). This assignment and Figure 4(b) are consistent with previous work in which fdPHf_d^{\mathrm{PH}} was found to decrease with increasing MgO thickness dd on a scale of 2 Å [28]. According to calculations [18], MgO layers close to the CoFeB and Pt interface are slightly metallic. They have a total thickness of about 4 Å and can be understood to have a large PH fraction. This effect may explain the pronounced drop of AdA_d from d=3.0d = 3.0 Å to 4.0 Å [Figure 4(b)].

Using Ad≈fdPHA_d \approx f_d^{\mathrm{PH}}, we can now determine gd,0IRT/gref≈Bd/(1−Ad)g_{d,0}^{\mathrm{IRT}}/g^{\mathrm{ref}} \approx B_d/(1 - A_d), i.e., the peak amplitude of the IRT contribution [Eq. (3)], which is found to decrease strongly with increasing dd [Figure 4(c)]. Remarkably, the characteristic IRT time τd\tau_d is found to grow with dd [Figure 4(d)], consistent with the increase of the slope of arg⁡G~d(ω)\arg\widetilde{G}_d(\omega) vs ω\omega [Figure 3(b)].

Model of dynamic IRT. Qualitatively, we suggest the following dynamic scenario for the IRT conductance gdIRT(t)=gd,0IRTΘ(t)e−t/τdg_d^{\mathrm{IRT}}(t) = g_{d,0}^{\mathrm{IRT}}\Theta(t)e^{-t/\tau_d}. At time t=0t = 0, a δ(t)\delta(t)-like spin voltage in CoFeB drives instantaneous tunneling of spin-polarized electrons from CoFeB to MgO defect states. Simultaneously, spin-unpolarized electrons from Pt are transferred to maintain local charge neutrality.

The subsequent tunneling events from occupied defect states to CoFeB or Pt are stochastic. Their rate ∂tNσ\partial_tN^\sigma is proportional to the number NσN^\sigma of spins σ=↑\sigma=\uparrow or ↓\downarrow occupying the defect state. Accordingly, we obtain a (i) simple temporal exponential decay of NσN^\sigma and, thus, the σ\sigma-current from defects to Pt. (ii) The characteristic time of this process increases with increasing dd because the tunneling from defects to Pt becomes less likely as dd and, thus, the average distance of defect and Pt grow. Finally, as two tunneling events over distance dd are involved, (iii) the spin current from CoFeB to Pt decreases with increasing dd. The implications (i), (ii), (iii) of our model are fully consistent with the experimental results of Figure 4(a), Figure 4(c) and Figure 4(d), respectively.

More quantitatively, a rate-equation treatment and the assumption of vanishing out-of-plane charge current (see Supplemental Material) reproduce the relationship gdIRT(t)=gd,0IRTΘ(t)e−t/τdg_d^{\mathrm{IRT}}(t)=g_{d,0}^{\mathrm{IRT}}\Theta(t)e^{-t/\tau_d} with

gd,0IRT∝de−d/λ,τd−1=Γ1−e−d/λd/λ.(5)g_{d,0}^{\mathrm{IRT}}\propto d e^{-d/\lambda}, \qquad\tau_d^{-1}=\Gamma\frac{1-e^{-d/\lambda}}{d/\lambda}. \tag*{(5)}

Here, Γ−1\Gamma^{-1} quantifies the spin lifetime in a defect state for an infinitely thin MgO layer, and λ\lambda is the spin decay length in a MgO barrier. Eq. (5) provides good fits to the data of Figure 4(b), (Figure 4c) and yields Γ−1=(50±5)\Gamma^{-1}=(50\pm5) fs and λ=(1.2±0.1)\lambda=(1.2\pm0.1) Å in good agreement with theoretical predictions [31, 42].

Discussion. Measurement of the THz spin conductance relies on the measurement of just two THz emission signals, without having to know any instrument response functions, in contrast to extraction of the charge source current. This approach can be extended to, in principle, any layer XX other than MgO, if the following conditions are fulfilled: (1) The signal exclusively arises from the spin current jSj_S arriving in HH. (2) jSj_S solely originates from the spin voltage μS\mu_S. (3) The presence of the layers XX and H∣XH|X does not change the μS\mu_S dynamics of FF.

For our CoFeB|MgO(dd)|Pt system, (1) is fulfilled for the reference sample (d=0d=0), but also for d≠0d\ne0 because the ISHE of Pt dominates spin-to-charge conversion of the metal stack (see Supplemental Material). (2) is fulfilled for the reference sample [10] and, thus, for d≠0d\ne0, too. (3) is fulfilled because the presence of XX in F∣HF|H stacks was shown to leave the spin-voltage dynamics unchanged [10]. If XX has a large optical and THz refractive index, the modified pump-propagation and current-to-THz-field conversion need to be accounted for (see Supplemental Material).

In conclusion, we demonstrate THz conductance spectroscopy for the example of MgO barriers. In the time domain, we find ultrafast signatures of IRT, i.e., an exponentially decaying spin current with a relaxation time that increases with the MgO layer thickness. This behavior arises because the tunneling probability decreases with increasing thickness of the tunneling barrier. We anticipate that our method can be used to measure the spin conductance of a large set of materials, ranging from simple metals [43] to complex materials such as antiferromagnets [44, 45, 46], potentially also involving a transverse spin-current component.

References

  1. [1]J. A. del Alamo, Nanometre-scale electronics with III–V compound semiconductors, Nature 479, 317 (2011).DOI
  2. [2]D. Hillerkuss, R. Schmogrow, T. Schellinger, M. Jordan, M. Winter, G. Huber, T. Vallaitis, R. Bonk, P. Kleinow, F. Frey et al., 26 Tbit s−1 line-rate super-channel transmission utilizing all-optical fast Fourier transform processing, Nature Photonics 5, 364 (2011).DOI
  3. [3]T. Kampfrath, M. Battiato, P. Maldonado, G. Eilers, J. Nötzold, S. Mährlein, V. Zbarsky, F. Freimuth, Y. Mokrousov, S. Blügel et al., Terahertz spin current pulses controlled by magnetic heterostructures, Nature Nanotechnology 8, 256 (2013).
  4. [4]A. Kirilyuk, A. V. Kimel and T. Rasing, Ultrafast optical manipulation of magnetic order, Reviews of Modern Physics 82, 2731 (2010).DOI
  5. [5]G. E. W. Bauer, E. Saitoh and B. J. van Wees, Spin caloritronics, Nature Materials 11, 391 (2012).
  6. [6]A. Manchon, J. Železný, I. M. Miron, T. Jungwirth, J. Sinova, A. Thiaville, K. Garello and P. Gambardella, Current-induced spin-orbit torques in ferromagnetic and antiferromagnetic systems, Reviews of Modern Physics 91, 035004 (2019).DOI
  7. [7]J. Sinova, S. O. Valenzuela, J. Wunderlich, C. H. Back and T. Jungwirth, Spin Hall effects, Reviews of Modern Physics 87, 1213 (2015).DOI
  8. [8]T. S. Seifert, L. Cheng, Z. Wei, T. Kampfrath and J. Qi, Spintronic sources of ultrashort terahertz electromagnetic pulses, Applied Physics Letters 120, 180401 (2022).
  9. [9]G. Bierhance, A. Markou, O. Gueckstock, R. Rouzegar, Y. Behovits, A. L. Chekhov, M. Wolf, T. S. Seifert, C. Felser and T. Kampfrath, Spin-voltage-driven efficient terahertz spin currents from the magnetic Weyl semimetals Co2MnGa and Co2MnAl, Applied Physics Letters 120, 082401 (2022).
  10. [10]R. Rouzegar, L. Brandt, L. Nádvornik, D. A. Reiss, A. L. Chekhov, O. Gueckstock, C. In, M. Wolf, T. S. Seifert, P. W. Brouwer et al., Laser-induced terahertz spin transport in magnetic nanostructures arises from the same force as ultrafast demagnetization, Physical Review B 106, 144427 (2022).
  11. [11]T. Seifert, S. Jaiswal, U. Martens, J. Hannegan, L. Braun, P. Maldonado, F. Freimuth, A. Kronenberg, J. Henrizi, I. Radu et al., Efficient metallic spintronic emitters of ultrabroadband terahertz radiation, Nature Photonics 10, 483 (2016).DOI
  12. [12]R. Freeman, A. Zholud, Z. Dun, H. Zhou and S. Urazhdin, Evidence for Dyakonov-Perel-like Spin Relaxation in Pt, Physical Review Letters 120, 067204 (2018).DOI
  13. [13]T. S. Seifert, S. Jaiswal, J. Barker, S. T. Weber, I. Razdolski, J. Cramer, O. Gueckstock, S. F. Maehrlein, L. Nadvornik, S. Watanabe et al., Femtosecond formation dynamics of the spin Seebeck effect revealed by terahertz spectroscopy, Nature Communications 9, 2899 (2018).
  14. [14]S. Yuasa and D. D. Djayaprawira, Giant tunnel magnetoresistance in magnetic tunnel junctions with a crystalline MgO(001) barrier, Journal of Physics D: Applied Physics 40, R337 (2007).
  15. [15]D. Apalkov, B. Dieny and J. M. Slaughter, Magnetoresistive random access memory, Proceedings of the IEEE 104, 1796 (2016).DOI
  16. [16]L. Wang, H. Cheng, P. Li, Y. L. W. Van Hees, Y. Liu, K. Cao, R. Lavrijsen, X. Lin, B. Koopmans and W. Zhao, Processed optospintronics tunnel junctions, Proceedings of the National Academy of Sciences 119 (2022).
  17. [17]Z. Jin, J. Li, W. Zhang, C. Guo, C. Wan, X. Han, Z. Cheng, C. Zhang, A. V. Balakin, A. P. Shkurinov et al., Magnetic Modulation of Terahertz Waves via Spin-Polarized Electron Tunneling Based on Magnetic Tunnel Junctions, Physical Review Applied 14, 014032 (2020).DOI
  18. [18]M. A. Wahada, E. Şaşioğlu, W. Hoppe, X. Zhou, H. Deniz, R. Rouzegar, T. Kampfrath, I. Mertig, S. S. P. Parkin and G. Woltersdorf, Atomic Scale Control of Spin Current Transmission at Interfaces, Nano Letters 22, 3539 (2022).
  19. [19]O. Gueckstock, L. Nádvorník, M. Gradhånd, T. S. Seifert, G. Bierhance, R. Rouzegar, M. Wolf, M. Vafaee, J. Cramer, M. A. Syskaki et al., Terahertz Spin-to-Charge Conversion by Interfacial Skew Scattering in Metallic Bilayers, Advanced Materials 33, 2006281 (2021).
  20. [20]A. L. Chekhov, Y. Behovits, J. J. F. Heitz, C. Denker, D. A. Reiss, M. Wolf, M. Weinelt, P. W. Brouwer, M. Münzenberg and T. Kampfrath, Ultrafast Demagnetization of Iron Induced by Optical versus Terahertz Pulses, Physical Review X 11, 041055 (2021).DOI
  21. [21]P. G. Mather, J. C. Read and R. A. Buhrman, Disorder, defects, and band gaps in ultrathin (001) MgO tunnel barrier layers, Physical Review B 73, 205412 (2006).DOI
  22. [22]J. J. Cha, J. C. Read, R. A. Buhrman and D. A. Muller, Spatially resolved electron energy-loss spectroscopy of electron-beam grown and sputtered CoFeB/MgO/CoFeB magnetic tunnel junctions, Applied Physics Letters 91, 062516 (2007).DOI
  23. [23]J. P. Velev, K. D. Belashchenko, S. S. Jaswal and E. Y. Tsymbal, Effect of oxygen vacancies on spin-dependent tunneling in Fe/MgO/Fe magnetic tunnel junctions, Applied Physics Letters 90, 072502 (2007).DOI
  24. [24]G. X. Miao, Y. J. Park, J. S. Moodera, M. Seibt, G. Eilers and M. Münzenberg, Disturbance of Tunneling Coherence by Oxygen Vacancy in Epitaxial $\mathrm{Fe}/\mathrm{MgO}/\mathrm{Fe}$ Magnetic Tunnel Junctions, Physical Review Letters 100, 246803 (2008).DOI
  25. [25]Y. Ke, K. Xia and H. Guo, Oxygen-Vacancy-Induced Diffusive Scattering in $\mathrm{Fe}/\mathrm{MgO}/\mathrm{Fe}$ Magnetic Tunnel Junctions, Physical Review Letters 105, 236801 (2010).DOI
  26. [26]E. Y. Tsymbal and D. G. Pettifor, Local impurity-assisted conductance in magnetic tunnel junctions, Physical Review B 64, 212401 (2001).DOI
  27. [27]T. König, G. H. Simon, H. P. Rust and M. Heyde, Work Function Measurements of Thin Oxide Films on Metals—MgO on Ag(001), The Journal of Physical Chemistry C 113, 11301 (2009).
  28. [28]J. M. Teixeira, J. Ventura, F. Carpinteiro, J. P. Araujo, J. B. Sousa, P. Wisniowski and P. P. Freitas, The effect of pinhole formation/growth on the tunnel magnetoresistance of MgO-based magnetic tunnel junctions, Journal of Applied Physics 106, 073707 (2009).DOI
  29. [29]T. König, G. H. Simon, H. P. Rust, G. Pacchioni, M. Heyde and H. J. Freund, Measuring the Charge State of Point Defects on MgO/Ag(001), Journal of the American Chemical Society 131, 17544 (2009).DOI
  30. [30]J. Mathon and A. Umerski, Theory of tunneling magnetoresistance of an epitaxial Fe/MgO/Fe(001) junction, Physical Review B 63, 220403 (2001).DOI
  31. [31]W. H. Butler, X. G. Zhang, T. C. Schulthess and J. M. MacLaren, Spin-dependent tunneling conductance of $\mathrm{Fe}|\mathrm{MgO}|\mathrm{Fe}$ sandwiches, Physical Review B 63, 054416 (2001).DOI
  32. [32]W. H. Butler, Tunneling magnetoresistance from a symmetry filtering effect, Science and Technology of Advanced Materials 9, 014106 (2008).pubmed.ncbi.nlm.nih.gov/27877932
  33. [33]J. P. Velev, M. Y. Zhuravlev, K. D. Belashchenko, S. S. Jaswal, E. Y. Tsymbal, T. Katayama and S. Yuasa, Defect-Mediated Properties of Magnetic Tunnel Junctions, IEEE Transactions on Magnetics 43, 2770 (2007).DOI
  34. [34]J. W. Gadzuk, Resonance Tunneling Through Impurity States in Metal-Insulator-Metal Junctions, Journal of Applied Physics 41, 286 (1970).DOI
  35. [35]J. M. Teixeira, J. Ventura, J. P. Araujo, J. B. Sousa, P. Wisniowski, S. Cardoso and P. P. Freitas, Resonant Tunneling through Electronic Trapping States in Thin MgO Magnetic Junctions, Physical Review Letters 106 (2011).pubmed.ncbi.nlm.nih.gov/21668184
  36. [36]A. S. Landsmann, M. Weger, J. Maurer, R. Boge, A. Ludwig, S. Heuser, C. Cirelli, L. Gallmann and U. Keller, Ultrafast resolution of tunneling delay time, Optica 1, 343 (2014).
  37. [37]J. Mathon and A. Umerski, Theory of resonant tunneling in an epitaxial Fe/Au/MgO/Au/Fe(001) junction, Physical Review B 71 (2005).
  38. [38]S. Yuasa, T. Nagahama and Y. Suzuki, Spin-Polarized Resonant Tunneling in Magnetic Tunnel Junctions, Science 297, 234 (2002).
  39. [39]Z.-Y. Lu, X. G. Zhang and S. T. Pantelides, Spin-Dependent Resonant Tunneling through Quantum-Well States in Magnetic Metallic Thin Films, Physical Review Letters 94 (2005).pubmed.ncbi.nlm.nih.gov/16090286
  40. [40]M. M. Glazov, P. S. Alekseev, M. A. Odnoblyudov, V. M. Chistyakov, S. A. Tarasenko and I. N. Yassievich, Spin-dependent resonant tunneling in symmetrical double-barrier structures, Physical Review B 71, 155313 (2005).DOI
  41. [41]T. S. Seifert, N. M. Tran, O. Gueckstock, S. M. Rouzegar, L. Nadвornik, S. Jaswal, G. Jakob, V. V. Temnov, M. Münzenberg, M. Wolf et al., Terahertz spectroscopy for all-optical spintronic characterization of the spin-Hall-effect metals Pt, W and Cu80Ir20, Journal of Physics D: Applied Physics 51, 364003 (2018).
  42. [42]Y. Xu, D. Ephron and M. R. Beasley, Directed inelastic hopping of electrons through metal-insulator-metal tunnel junctions, Physical Review B 52, 2843 (1995).pubmed.ncbi.nlm.nih.gov/9981356
  43. [43]A. Melnikov, L. Brandt, N. Liebling, M. Ribow, I. Mertig and G. Woltersdorf, Ultrafast spin transport and control of spin current pulse shape in metallic multilayers, Physical Review B 106 (2022).DOI
  44. [44]K. Lee, D.-K. Lee, D. Yang, R. Mishra, D.-J. Kim, S. Liu, Q. Xiong, S. K. Kim, K.-J. Lee and H. Yang, Superluminal-like magnon propagation in antiferromagnetic NiO at nanoscale distances, Nature Nanotechnology 16, 1337 (2021).DOI
  45. [45]Y. Sasaki, G. Li, T. Moriyama, T. Ono, R. V. Mikhaylovskiy, A. V. Kimel and S. Mizukami, Laser stimulated THz emission from Pt/CoO/FeCoB, Applied Physics Letters 117 (2020).DOI
  46. [46]O. Gueckstock, R. L. Seeger, T. S. Seifert, S. Auffret, S. Gambarelli, J. N. Kirchhof, K. I. Bolotin, V. Baltz, T. Kampfrath and L. Nádvorník, Impact of gigahertz and terahertz transport regimes on spin propagation and conversion in the antiferromagnet IrMn, Applied Physics Letters 120 (2022).DOI

Paper details

vlnc:W_e3vnurf5
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