Present extraction
To extract the in-plane sheet present IC flowing contained in the pattern from the measured terahertz sign STHz, we first measure our set-up response perform HSE by having a reference electro-optic emitter (50 μm GaP on a 500 μm glass substrate) on the similar place because the pattern, which yields a reference terahertz sign (S_{mathrm{THz}}^{mathrm{ref}}). By calculating the emitted terahertz electrical area from that reference emitter Eref, HSE is set by fixing the convolution ({{{S}}}_{{rm{THz}}}^{{rm{ref}}}left({{t}}proper)=left({{{H}}}_{{{SE}}} * {{{E}}}^{{rm{ref}}}proper)left({{t}}proper)) (equation (2)) for HSE (ref. 43). Additional measured inputs for this calculation are the excitation spot dimension with a full-width at half-maximum of twenty-two μm, the excitation pulse vitality of 1.9 nJ and a Fourier-transform-limited pump pulse with a spectrum centred at 800 nm and a full-width at half-maximum of 110 nm. We carry out the deconvolution instantly within the time area by recasting it as a matrix equation16.
Subsequent, the electrical area E instantly behind the pattern is obtained from the recorded terahertz sign STHz with the assistance of the derived perform HSE by fixing the analogous equation STHz(t) = (HSE * E)(t) for E. Lastly, the sheet cost present (Supplementary Desk 1) as proven in Fig. 3 is derived from a generalized Ohm’s legislation28, which within the frequency area at frequency ω/2π reads
$$Eleft(omega proper)={eZ}left(omega proper){I}_{{{C}}}left(omega proper).$$
(3)
Right here, −e is the electron cost and the pattern impedance Z(ω) is given by Z0/[1 + nsub + Z0dσ(ω)] with the free-space impedance Z0, the substrate refractive index nsub ≈ 2 (ref. 56) and the metal-stack thickness d = dFM + dPM. The measured imply pattern conductivity σ (Supplementary Desk 1) is roughly frequency-independent as a result of giant Drude scattering charge (Supplementary Fig. 2). To allow a comparability of terahertz currents from totally different samples, we normalize IC by the absorbed fluence within the FM. The information proven in Fig. 3 have been obtained in a dry-air environment.
Pattern preparation
The FM|PM samples (FM = Ni and Py, PM = Pt, Ti, Cu and W) are fabricated on glass substrates (thickness 500 μm) or thermally oxidized Si substrates (625 μm) by radio-frequency magnetron sputtering below an Ar environment of 6N purity. The pattern construction and thickness are described in Supplementary Desk 1. For the sputtering, the bottom stress within the chamber is decrease than 5 × 10−7 Pa. To keep away from oxidation, SiO2 (thickness 4 nm) is sputtered on the floor of the movies. All sputtering processes are carried out at room temperature. The W movies are predominantly within the β-phase for dW < 10 nm, with an α-phase content material that grows with dW and dominates for dW > 10 nm (ref. 9).
Estimate of digital temperatures
We calculate the digital temperature enhance ΔTe0 upon pump-pulse absorption by
$$Delta {T}_{{rm{e}}0}=sqrt{{T}_{0}^{2}+frac{2{F}_{l}}{gamma d}}-{T}_{0}.$$
(4)
Right here, T0 = 300 Okay is the ambient temperature, Fl is the absorbed fluence within the respective layer l = FM or PM (Supplementary Desk 1), d is the layer thickness and γTe is the precise digital warmth capability at digital temperature Te with γ = 300 J m−3 Okay−2 for W, 320 J m−3 Okay−2 for Ni, 330 J m−3 Okay−2 for Ti and 90 J m−3 Okay−2 for Pt (ref. 57).
To acquire the absorbed fluences in every layer, we observe that the pump electrical area is nearly fixed all through the pattern (Supplementary Fig. 9). Due to this fact, the native pump absorption scales solely with the imaginary half Imε of the dielectric perform ε at a wavelength of 800 nm, which equals 22.07 for Ni, 9.31 for Pt, 19.41 for Ti and 19.71 for W (ref. 58). Consequently, the absorbed fluence is set by
$${F}_{l}={F}_{{rm{tot}}}frac{{d}_{l}{rm{Im}}{varepsilon }_{l}}{{d}_{{rm{FM}}}{rm{Im}}{varepsilon }_{{rm{FM}}}+{d}_{{rm{PM}}}{rm{Im}}{varepsilon }_{{rm{PM}}}}$$
(5)
with the whole absorbed fluence Ftot that’s obtained from the absorbed pump energy (Supplementary Desk 1) and the beam dimension on the pattern (as described beforehand).
Experimental error estimation
The error bars for the utmost place tmax of the transient cost present IC(t) (Fig. 4b) are estimated as ±20% of the read-off delay worth (round markers in Fig. 4a), however a minimum of 5 fs. The uncertainty within the relative amplitude of IC(tmax) (Fig. 4c) and the relative space of IC(t) (Fig. 4e) is, respectively, estimated as ±20% and ±10%, each reflecting the standard signal-to-noise ratio of the extracted present traces (Fig. 4a). The error bars for the full-width of IC(t) at half-maximum (Fig. 4d) are obtained from the uncertainty of the delay (Fig. 4b) with subsequent multiplication by (sqrt{2}), which accounts for the error propagation of a distinction of two portions.
Mannequin of L transport
To mannequin the ballistic present within the PM, we assume {that a} δ(t)-like transient L accumulation within the FM generates an digital wave packet, which has orbital angular momentum ΔLokay0 alongside the course of the FM magnetization M and imply wave vector okay within the PM proper behind the FM/PM interface, that’s, at z = 0+ (Fig. 5a).
Within the case of purely ballistic transport, this wave packet propagates into the PM bulk in keeping with ΔLokay(z, t) = ΔLokay0 δ(z − vokayzt), the place vokayz is the z part of the wave-packet group velocity. Be aware that we prohibit ourselves to okay values with non-negative vokayz values. The entire pump-induced L present density flowing into the depth of the PM is for z > 0 given by the sum
$${r}_{z}left(tright)=mathop{sum }limits_{{bf{okay}},{v}_{{bf{okay}}z}ge 0}Delta {L}_{{bf{okay}}0}{v}_{{bf{okay}}z}updeltaleft(z-{v}_{{bf{okay}}z}tright).$$
(6)
Assuming that ΔLokay0 arises from states not too removed from the Fermi vitality, the summation of equation (6) is roughly proportional to an integration over the Fermi floor elements with vokayz ≥ 0. One obtains
$${r}_{z}(t)={mathrm{{e}}}^{-t/tau }{int }_{0}^{infty }{rm{d}}{v}_{z}w({v}_{z}){v}_{z}updelta (z-{v}_{z}t)$$
(7)
the place z > 0, and
$$w({v}_{z})=sum _{{bf{okay}},{v}_{{bf{okay}}z}ge 0}Delta {L}_{{bf{okay}}0}updelta ({v}_{{bf{okay}}z}-{v}_{z})$$
(8)
is the L weight of the z-axis group velocity vz. In equation (7), we phenomenologically account for the comfort of the ballistic present with time fixed τ by introducing the issue e−t/τ. Performing the combination of equation (7) yields
$${r}_{z}left(tright)=frac{{{rm{e}}}^{-t/tau }}{t}frac{z}{t}wleft(frac{z}{t}proper).$$
(9)
To find out a believable form of w(v), we observe that ΔLokay0 is non-zero inside a number of 0.1 eV across the Fermi vitality EF owing to the width of the photoexcited and quickly enjoyable electron distribution33. Assuming a spherical Fermi floor and isotropic ΔLokay0, we now have ΔLokay0 ∝ δ(Eokay − EF) and vokayz = vFcosθ, the place Eokay is the band construction, vF is the Fermi velocity and θ is the angle between okay and the z axis. Due to this fact, after turning equation (8) into an integral, the integrand (Delta {L}_{{bf{okay}}0}updeltaleft({v}_{{rm{F}}}{{cos }}theta -{v}_{z}proper)) ({{rm{d}}^{3}{bf{okay}}}) turns into proportional to (updeltaleft({v}_{{rm{F}}}{{cos }}theta -{v}_{z}proper)) ({rm{d}}{{cos }}theta) in spherical coordinates, resulting in
$$wleft({v}_{z}proper)propto {int }_{0}^{1}{rm{d}}{{cos }}theta updeltaleft({v}_{{rm{F}}}{{cos }}theta -{v}_{z}proper)propto varTheta left({v}_{{rm{F}}}-{v}_{z}proper),$$
(10)
the place Θ is the Heaviside step perform. In different phrases, all velocities ({v}_{z}={v}_{{rm{F}}}{{cos }}theta) from 0 to vF have equal weight.
Within the case of purely diffusive transport, we use the L diffusion equation for μL (ref. 23). With a localized accumulation μL(z, t) ∝ δ(z) at time t ≈ 0, the buildup disperses in keeping with the well-known resolution
$${mu }_{L}left(z,tright)propto frac{1}{sqrt{Dt}}{{exp }}left(-frac{{z}^{2}}{4Dt}proper)$$
(11)
for z > 0 and t > 0. Right here, D is the diffusion coefficient that equals ({v}_{L}^{2}tau /3) within the case of a spherical okay-space floor carrying the L wave packets, vL is their group velocity and τ is their velocity rest time. To find out the present density, we apply Fick’s legislation23, jL = −D∂μL/∂z to equation (11) and acquire
$${r}_{z}left(tright)propto varTheta left(tright)frac{z}{t}frac{1}{sqrt{Dt}}{{exp }}left(-frac{{z}^{2}}{4Dt}proper).$$
(12)
Ab initio estimate of the L velocity
The orbital velocity is estimated for the majority W in a body-centred cubic construction. The ab initio self-consistent calculation of the digital states is carried out inside density purposeful principle through the use of the FLEUR code2, which implements the full-potential linearly augmented plane-wave (FLAPW) technique59. The alternate–correlation impact is included within the scheme of the generalized gradient approximation through the use of the Perdew–Burke–Ernzerhof purposeful60. The lattice parameter of the cubic unit cell is about to five.96a0, the place a0 is the Bohr radius. For the muffin-tin potential, we set RMT = 2.5a0 for the radius and lmax = 12 for the utmost of the harmonic growth. Additional, we set the plane-wave cut-offs for the interstitial area to (4.0{a}_{0}^{-1}), (10.1{a}_{0}^{-1}) and (12.2{a}_{0}^{-1}) for the idea set, the alternate–correlation purposeful and the cost density, respectively. For the okay-points, a 16 × 16 × 16 Monkhorst–Pack mesh is outlined.
From the converged digital construction, we receive the maximally localized Wannier capabilities through the use of the WANNIER90 code61. We use 18 Wannier states with (s,{p}_{x},{p}_{y},{p}_{z},{d}_{{z}^{2}},{d}_{{x}^{2}-{y}^{2}},{d}_{xy},{d}_{yz},{d}_{zx}) symmetries for spin up and down because the preliminary guess. The utmost of the interior (frozen) vitality window is about 5 eV above the Fermi vitality for the disentanglement, and the outer vitality window is outlined by the minimal and most energies of the 36 valence states obtained from the FLAPW calculation. The Hamiltonian, place, orbital-angular-momentum operators and spin-angular-momentum operators are reworked from the FLAPW foundation into the maximally localized Wannier perform foundation. The ensuing digital band construction and texture of the orbital-angular-momentum operator L are displayed in Supplementary Fig. 9.
From this sensible tight-binding mannequin, the orbital-momentum-weighted velocity averaged over the Fermi floor (FS) is calculated by
$${langle {v}_{alpha }{L}_{beta }rangle }_{{rm{FS}}}=frac{{sum }_{n{bf{okay}}}{,f}_{n{bf{okay}}}^{,{prime} }langle n{bf{okay}}|({v}_{alpha }{L}_{beta }+{L}_{beta }{v}_{alpha })/2|n{bf{okay}}rangle }{{sum }_{n{bf{okay}}}{,f}_{n{bf{okay}}}^{,{prime} }},$$
(13)
the place vα and Lβ are the α part of the rate and the β part of the orbital-angular-momentum operators, respectively. The |nokay〉 is the eigenstate of the Hamiltonian with vitality Enokay and band index n, and ({f}_{n{bf{okay}}}^{,{prime} }) is the vitality by-product of the Fermi–Dirac distribution perform. To polarize Lβ, we add a small orbital Zeeman coupling alongside the β course to the naked Hamiltonian. We verify that the results of equation (13) modifications by lower than 1% when the orbital Zeeman splitting is elevated from 10 meV to 30 meV. The okay-space integrals in equation (13) are carried out on a 256 × 256 × 256 mesh.
The orbital velocity is estimated by
$${leftlangle {v}_{alpha }^{{L}_{beta }}rightrangle }_{{rm{FS}}}=frac{{leftlangle {v}_{alpha }{L}_{beta }rightrangle }_{{rm{FS}}}}{sqrt{{leftlangle {L}_{beta }^{2}rightrangle }_{{rm{FS}}}}},$$
(14)
the place ({langle {L}_{beta }^{2}rangle }_{{rm{FS}}}) is obtained by equation (13), however with vα changed by Lβ. The result’s proven in Supplementary Fig. 10.
Particulars of ab initio LCC calculations
A skinny W stack of 19 body-centred cubic (110) atomic layers is calculated by the self-consistent ab initio technique utilizing the identical FLAPW parameters as for the majority calculation given beforehand, aside from the mesh of okay-points for which we use a 24 × 24 Monkhorst–Pack mesh. For Wannierization, we receive 342 maximally localized Wannier capabilities, ranging from Wannier states with (s,{p}_{x},{p}_{y},{p}_{z},{d}_{{z}^{2}},{d}_{{x}^{2}-{y}^{2}},{d}_{xy},{d}_{yz},{d}_{zx}) symmetries for spin up and down because the preliminary guess. We outline the utmost of the interior window at 2 eV above the Fermi vitality.
From the Hamiltonian of the W skinny movie, the okay-space orbital-angular-momentum texture on the Fermi floor is obtained by
$${langle {{bf{L}}}_{{rm{high}}}rangle }_{{rm{FS}}}({bf{okay}})=-4{okay}_{{rm{B}}}Tsum _{n}{f}_{n{bf{okay}}}^{,{prime} }{langle {{bf{L}}}_{{rm{high}}}rangle }_{n{bf{okay}}},$$
(15)
the place ({langle {{bf{L}}}_{{rm{high}}}rangle }_{n{bf{okay}}}=langle n{bf{okay}}|{{bf{L}}}_{{rm{high}}}|n{bf{okay}}rangle) is the expectation worth of the orbital angular momentum for the 2 atoms on the highest floor, T = 300 Okay is temperature and okayB is the Boltzmann fixed.
To calculate the cost present on account of LCC, we contemplate the orbital-dependent chemical potential
$$frac{{varepsilon }_{beta gamma }}{2}({r}_{beta }{L}_{gamma }+{L}_{gamma }{r}_{beta })$$
(16)
as a perturbation. Right here, Lγ is the γ part of the orbital-angular-momentum operator, and rβ is the β part of the place operator, which is well-defined alongside the z axis (Fig. 1), and ({varepsilon }_{beta gamma }) may be interpreted as an orbital-dependent electrical area. The charge-current density alongside the α course is given by the Kubo components
$$langle {,j}_{alpha }rangle =-frac{e}{V}sum _{{bf{okay}}nn^{prime} }({,f}_{n{boldsymbol{okay}}}-{f}_{n^{prime} {bf{okay}}}){rm{Re}}frac{langle n{bf{okay}}|{v}_{alpha }|n^{prime} {bf{okay}}rangle langle n^{prime} {bf{okay}}|V|n^{prime} {bf{okay}}rangle }{{E}_{n{bf{okay}}}-{E}_{n^{prime} {bf{okay}}}+{rm{i}}varGamma }$$
(17)
the place V is the amount of the system and Γ = 25 meV is a phenomenological broadening parameter. The LCC response is characterised by the tensor
$${sigma }_{{{{LC}}}{rm{C}},alpha beta }^{{L}_{gamma }}=-frac{e}{2V}mathop{sum }limits_{{bf{okay}}n{n}^{{prime} }}left({,f}_{n{bf{okay}}}-{f}_{{n}^{{prime} }{bf{okay}}}proper){rm{Re}}frac{leftlangle n{bf{okay}},|,{v}_{alpha },|,{n}^{{prime} }{bf{okay}}rightrangle leftlangle {n}^{{prime} }{bf{okay}},|,left({r}_{beta }{L}_{gamma }+{L}_{gamma }{r}_{beta }proper),|,{n}^{{prime} }{bf{okay}}rightrangle }{{E}_{n{bf{okay}}}-{E}_{{n}^{{prime} }{bf{okay}}}+{rm{i}}varGamma },$$
(18)
which relates (langle {j}_{a}rangle) and εβγ by (langle {j}_{alpha }rangle ={sigma }_{{LCmathrm{C}},alpha beta }^{{L}_{gamma }} epsilon_{beta gamma}). The okay-space integral is carried out on a 400 × 400 mesh. The z-resolved LCC response is proven in Supplementary Fig. 11.
Reporting abstract
Additional data on analysis design is accessible within the Nature Portfolio Reporting Abstract linked to this text.