Intrinsic Spin Accumulation Mechanisms

• Intrinsic spin accumulation is the buildup of nonequilibrium spin polarization arising from a system’s inherent spin–orbit coupling, Berry curvature, and geometric phase characteristics.
• It manifests as spatially nonuniform accumulations—such as edge spin density, chemical potential splitting, or thermal imbalances—that can be quantitatively probed in electronic, thermal, and photonic platforms.
• Research indicates that intrinsic spin accumulation influences device performance by altering transport properties and magnetoelectronic behavior, opening pathways for advanced spintronic applications.

Searching arXiv for the cited papers and closely related work on intrinsic spin accumulation. {"query":"Intrinsic spin accumulation arXiv (Ling et al., 2014, Dejene et al., 2013, Liu et al., 2018, Hartmann et al., 2020, Sadjina et al., 2011)", "max_results": 10} Intrinsic spin accumulation is the nonequilibrium build-up of spin polarization, spin-resolved electrochemical imbalance, or spin-resolved thermal imbalance generated by a system’s own internal spin structure rather than by impurity-driven asymmetries or spin-selective external interfaces. Depending on the platform, the accumulated quantity may be an edge spin density, an electrochemical-potential splitting $\mu_s=\mu_\uparrow-\mu_\downarrow$, a spin heat accumulation $T_s=T_\uparrow-T_\downarrow$, or an optical separation of opposite helicities. In electronic transport, the term usually denotes accumulation driven by band-structure spin–orbit coupling, two-channel spin transport, or interfacial spin bias; in photonics it denotes accumulation produced by the internal geometric phase structure of a beam; in thermal transport it denotes spin imbalance produced by spin-dependent heat currents alone (Dejene et al., 2013, Ling et al., 2014, Shitade, 29 Sep 2025).

1. Conceptual scope and basic observables

The central observable is not unique across subfields, but the common structure is a spin-resolved nonequilibrium variable that becomes spatially nonuniform and often peaks at boundaries. In metallic spin transport, the canonical quantities are the electrical spin accumulation $\mu_s=\mu_\uparrow-\mu_\downarrow$ and the spin heat accumulation $T_s=T_\uparrow-T_\downarrow$. In spin–orbit systems with boundaries, the observable is typically an edge spin density $s^a(x)$. In thermal and magnetic formulations that avoid ambiguous spin-current operators, the response is written directly as a spin-density response to gradients such as $\partial_i E_j$ or $\partial_i(-\partial_j T)$ (Dejene et al., 2013, Shitade, 2022, Shitade, 29 Sep 2025).

Platform	Driving agent	Accumulated quantity
Metallic spin valves	Charge current or heat current	$\mu_s$, $T_s$
Spin–orbit electronic channels	Charge current or electric-field gradient	Edge spin density $s^a$
Thermal spin transport	Temperature gradient	$T_s=T_\uparrow-T_\downarrow$0 or $T_s=T_\uparrow-T_\downarrow$1
Photonic spin Hall systems	PB/vortex phase with broken symmetry	Opposite edge accumulation of $T_s=T_\uparrow-T_\downarrow$2 and $T_s=T_\uparrow-T_\downarrow$3

A persistent source of confusion is the word “intrinsic.” In spin Hall physics it often means Berry-curvature or band-geometry origin, as opposed to skew scattering or side-jump. In two-channel metallic transport it can instead mean that the accumulation is generated by the internal spin-up/spin-down transport structure of the device. In photonics it refers to spin splitting that is independent of light–matter interaction parameters and originates in the beam’s internal polarization geometry. In magnetic systems with SOC or noncollinear order, the term has additionally become tied to gauge-invariant formulations based on spin density rather than on a nonunique spin current (Ling et al., 2014, Dejene et al., 2013, Shitade, 29 Sep 2025).

2. Electronic edge accumulation from intrinsic spin–orbit structure

In diffusive Rashba systems, intrinsic spin accumulation can arise through “intrinsic spin swapping,” in which spin–orbit-induced precession couples spin components in the diffusion equations so that a primary injected spin current generates a transverse secondary spin current whose polarization is effectively interchanged with the direction of flow. For injected $T_s=T_\uparrow-T_\downarrow$4, the dynamics generates a transverse $T_s=T_\uparrow-T_\downarrow$5 and therefore a $T_s=T_\uparrow-T_\downarrow$6-polarized edge accumulation; for injected $T_s=T_\uparrow-T_\downarrow$7, it generates $T_s=T_\uparrow-T_\downarrow$8 and an out-of-plane edge accumulation. In the wide-strip regime the swapped transverse current can be of the order of the injected primary current, and the maximum magnitude reported for $T_s=T_\uparrow-T_\downarrow$9 is approximately $\mu_s=\mu_\uparrow-\mu_\downarrow$0 (Sadjina et al., 2011).

A distinct intrinsic mechanism appears in high-mobility symmetric quantum wells with two subbands. There the relevant coupling is an inter-subband spin–orbit interaction between even- and odd-parity subbands, present even though the usual Rashba term is absent in an inversion-symmetric structure. The edge spin density takes the form

$\mu_s=\mu_\uparrow-\mu_\downarrow$1

with $\mu_s=\mu_\uparrow-\mu_\downarrow$2, and the inter-subband gap $\mu_s=\mu_\uparrow-\mu_\downarrow$3 strongly controls the magnitude through $\mu_s=\mu_\uparrow-\mu_\downarrow$4. The paper estimates that changing $\mu_s=\mu_\uparrow-\mu_\downarrow$5 from zero up to $\mu_s=\mu_\uparrow-\mu_\downarrow$6 K changes the magnitude of the effect by three orders of magnitude, a result introduced in the context of the experiment by Hernandez et al. on symmetric bilayer GaAs structures (Khaetskii et al., 2016).

Current-induced edge accumulation was also directly imaged at room temperature in Bi$\mu_s=\mu_\uparrow-\mu_\downarrow$7Se$\mu_s=\mu_\uparrow-\mu_\downarrow$8, BiSbTeSe$\mu_s=\mu_\uparrow-\mu_\downarrow$9, and Pt. The observed signal is an out-of-plane spin polarization of opposite sign at opposite channel edges, reversing sign with current reversal and scaling linearly with current. For Bi$T_s=T_\uparrow-T_\downarrow$0Se$T_s=T_\uparrow-T_\downarrow$1 and BiSbTeSe$T_s=T_\uparrow-T_\downarrow$2, the extracted spin Hall angles were $T_s=T_\uparrow-T_\downarrow$3 and $T_s=T_\uparrow-T_\downarrow$4, respectively, and Hanle analysis yielded $T_s=T_\uparrow-T_\downarrow$5 ps in Bi$T_s=T_\uparrow-T_\downarrow$6Se$T_s=T_\uparrow-T_\downarrow$7 and $T_s=T_\uparrow-T_\downarrow$8 ps in BiSbTeSe$T_s=T_\uparrow-T_\downarrow$9. The spatial symmetry rules out a dominant uniform Edelstein origin for the measured signal, although the optical imaging alone does not separate intrinsic and extrinsic SHE contributions quantitatively (Liu et al., 2018).

A further refinement comes from nonequilibrium Green-function calculations for disordered Pt films. In the bulklike region, the calculations recover the intrinsic spin Hall conductivity of Pt, but near the surfaces they show that the effective transverse decay length for intrinsically generated spin accumulation differs strongly from the conventional longitudinal spin-diffusion length. At the Fermi energy, the reported transverse decay lengths are approximately $s^a(x)$0, $s^a(x)$1, and $s^a(x)$2 nm for disorder strengths $s^a(x)$3, $s^a(x)$4, and $s^a(x)$5 mRy, while the corresponding longitudinal values are approximately $s^a(x)$6, $s^a(x)$7, and $s^a(x)$8 nm. This places the transverse length scale close to the mean free path and implies a breakdown of standard drift–diffusion modeling for intrinsic SHE-driven transverse transport near surfaces (Belashchenko et al., 2023).

3. Two-channel metallic accumulations and device-level manifestations

In metallic spin valves, intrinsic spin accumulation appears most directly in the two-channel model. Dejene et al. treated both the familiar electrical accumulation,

$s^a(x)$9

and the thermal analogue,

$\partial_i E_j$0

Their diffusion equations,

$\partial_i E_j$1

separate spin-flip relaxation from interspin energy relaxation. In Py/Cu/Py nanopillar spin valves, fitting the spin heat valve signal with a 3D finite-element model gave $\partial_i E_j$2 mK at room temperature and $\partial_i E_j$3 mK at $\partial_i E_j$4 K, with $\partial_i E_j$5 nm at room temperature and $\partial_i E_j$6 nm at $\partial_i E_j$7 K. These values were reported as up to about $\partial_i E_j$8 of the total temperature bias across the pillar (Dejene et al., 2013).

Within the same Valet–Fert framework, spin accumulation can be re-expressed as a capacitance-like storage of spin-channel imbalance. In an FM/NM junction, the spin-dependent chemical-potential splitting $\partial_i E_j$9 under quasi-neutrality permits the two spin channels to be treated as the two plates of a capacitor. The resulting spin-accumulation capacitance for each layer is

$\partial_i(-\partial_j T)$0

and the stored energy is the splitting energy,

$\partial_i(-\partial_j T)$1

rather than electrostatic field energy. This formulation reinterprets the low-frequency imaginary part of magnetoimpedance as arising from spin accumulation rather than from ordinary geometric capacitance (Zhu et al., 2017).

Device-internal spin accumulation can also modify Coulomb-blockade electrostatics. In a ferromagnetic single-electron transistor with Co/Al$\partial_i(-\partial_j T)$2O$\partial_i(-\partial_j T)$3/Al/Py/Al/Al$\partial_i(-\partial_j T)$4O$\partial_i(-\partial_j T)$5/Co junctions measured at about $\partial_i(-\partial_j T)$6 mK, non-equilibrium spin accumulation in the Al layer near an antiparallel-aligned junction produces a tiny interfacial charge dipole and an additional series capacitance $\partial_i(-\partial_j T)$7. The extra charging scale is

$\partial_i(-\partial_j T)$8

and the resulting tunnel magnetocapacitance reaches approximately $\partial_i(-\partial_j T)$9. The modeling associated this with a spin diffusion length of approximately $\mu_s$0 nm and a dipole crossing position $\mu_s$1 nm in Al (Lee et al., 2015).

A different experimental route quantified spin accumulation in Ta/Co/Pt trilayers by its contribution to harmonic Hall signals. In Ta($\mu_s$2)/Co(2 nm)/Pt(5 nm), the extracted spin-accumulation coefficient $\mu_s$3 was approximately $\mu_s$4 emu/cc per $\mu_s$5 A/m$\mu_s$6 for $\mu_s$7 nm and approximately $\mu_s$8 emu/cc per $\mu_s$9 A/m$T_s$0 for $T_s$1 nm. At $T_s$2 A/m$T_s$3, the inferred spin accumulation in Co exceeded ten percent of the local magnetization for multiple Ta thicknesses, reaching about $T_s$4 for $T_s$5 nm. The thickness trend of $T_s$6 followed the previously reported spin Hall angle trend of the Ta/Pt system (Luo et al., 2017).

4. Magnetic media, interfacial spin bias, and gauge-invariant formulations

At a normal-metal|magnetic-insulator interface, a spin accumulation $T_s$7 can reshape the equilibrium magnetization texture of the insulator. Within Landau–Lifshitz–Gilbert phenomenology combined with magnetoelectronic circuit theory, the static solution for a semi-infinite easy-axis magnetic insulator is a virtual, or image, domain wall whose center lies outside the magnet. For spin accumulation perpendicular to the easy axis, the theory predicts a threshold

$T_s$8

above which the interfacial magnetization acquires a component parallel to $T_s$9. In the Pt|YIG parameter set used in the paper, the corresponding critical electric field was estimated as $s^a$0 V/m, and the threshold produces kinks in spin Hall magnetoresistance and onset behavior in nonlocal magnon transport (Hartmann et al., 2020).

A broader longitudinal phenomenology for magnets introduces two variables: the equilibrium magnetization $s^a$1 and a nonequilibrium spin accumulation $s^a$2. In this framework, $s^a$3 does not diffuse but may decay, while $s^a$4 both diffuses and decays. Thermodynamic consistency requires a new exchange constant

$s^a$5

and the coupled dynamics lead to two decay rates at fixed wavevector instead of one. In the low-$s^a$6 limit, the slow mode becomes diffusive with an effective diffusion constant

$s^a$7

so diffusion is slowed because transport proceeds through the sequence $s^a$8, diffusion of $s^a$9, and $T_s=T_\uparrow-T_\downarrow$00 (Saslow et al., 2021).

In magnetic systems with SOC or noncollinear order, the conventional spin current is not uniquely defined, which has motivated gauge-invariant formulations based directly on the spin density response to electric-field gradients. The intrinsic spin accumulation coefficient for the magnetic spin Hall effect is written as

$T_s=T_\uparrow-T_\downarrow$01

where the intrinsic interband contribution can remain finite in insulators. Applied to altermagnets with magnetic point group $T_s=T_\uparrow-T_\downarrow$02, the theory predicts nonzero intrinsic spin accumulation in metallic RuO$T_s=T_\uparrow-T_\downarrow$03 and insulating MnF$T_s=T_\uparrow-T_\downarrow$04. In insulating MnF$T_s=T_\uparrow-T_\downarrow$05, the reported values include

$T_s=T_\uparrow-T_\downarrow$06

without SOC, and with SOC

$T_s=T_\uparrow-T_\downarrow$07

This directly contrasts with the magnetic spin Hall conductivity, which vanishes in insulators because it is an extrinsic intraband Fermi-surface response (Shitade, 29 Sep 2025).

5. Thermal and photonic analogues

Thermal spin accumulation is generated when the two spin channels respond differently to a temperature gradient. In the two-channel phenomenology of Wegrowe, Drouhin, and Lacour, open-circuit conditions in a conductor imply

$T_s=T_\uparrow-T_\downarrow$08

so that

$T_s=T_\uparrow-T_\downarrow$09

The resulting diffusion equation,

$T_s=T_\uparrow-T_\downarrow$10

shows that heat currents alone generate spin accumulation in conductors, while an analogous equation with length scale $T_s=T_\uparrow-T_\downarrow$11 describes insulators in which the heat carriers are spin-dependent excitations rather than charge carriers (Wegrowe et al., 2012).

A more microscopic Green-function treatment was later developed for the spin Nernst effect. There the response is formulated directly as a spin-density response to the gradient of the temperature gradient,

$T_s=T_\uparrow-T_\downarrow$12

which avoids ambiguities in spin-current definitions and magnetization corrections. Under time-reversal symmetry and in the absence of inelastic scattering, the generalized Mott relation becomes

$T_s=T_\uparrow-T_\downarrow$13

Within this framework, thermal spin accumulation vanishes for the three-dimensional Luttinger model with $T_s=T_\uparrow-T_\downarrow$14-function nonmagnetic disorder in the first Born approximation, but is nonzero for the two-dimensional Rashba model below the Rashba band crossing (Shitade, 2022).

In photonics, intrinsic spin accumulation has an optical analogue in the intrinsic photonic spin Hall effect. A cylindrical vector beam can be decomposed as

$T_s=T_\uparrow-T_\downarrow$15

so the two circular components carry opposite vortex phases $T_s=T_\uparrow-T_\downarrow$16, equivalent to a spin-dependent Pancharatnam–Berry phase. In a perfectly rotationally symmetric beam these phases superpose without net splitting, but when rotational symmetry is broken by a fan-shaped aperture the noncontinuous azimuthal phase generates a transverse momentum shift,

$T_s=T_\uparrow-T_\downarrow$17

and therefore edge accumulation of opposite helicities. The reported splitting is tunable with topological charge $T_s=T_\uparrow-T_\downarrow$18, reverses sign with $T_s=T_\uparrow-T_\downarrow$19, increases with aperture angle $T_s=T_\uparrow-T_\downarrow$20, and reaches millimetre scale for $T_s=T_\uparrow-T_\downarrow$21 with $T_s=T_\uparrow-T_\downarrow$22, far larger than the wavelength $T_s=T_\uparrow-T_\downarrow$23 nm (Ling et al., 2014).

6. Terminological distinctions, limitations, and open questions

A recurring misconception is that intrinsic spin accumulation is synonymous with one specific microscopic mechanism. The literature instead uses the term for several internally generated nonequilibrium spin states: Berry-curvature-driven edge accumulation in spin–orbit materials, two-channel electrochemical or thermal imbalance in metallic spin valves, thermally induced spin imbalance in conductors and insulators, and geometric-phase-driven helicity separation in photonics. A plausible implication is that the unifying criterion is not the precise Hamiltonian term but the absence of reliance on impurity asymmetry or explicitly spin-selective boundary scattering (Dejene et al., 2013, Ling et al., 2014, Shitade, 29 Sep 2025).

The topic is also shaped by several formal and experimental limitations. In magnetic systems with SOC or noncollinear order, spin current is not unique, which motivates ISA formalisms based on directly observable spin density. In metallic Pt, NEGF results show that near-surface transverse transport generated by the intrinsic SHE is not captured by standard drift–diffusion, even when modified boundary conditions are added. In spin caloritronics, the measured $T_s=T_\uparrow-T_\downarrow$24 should be interpreted as an effective temperature difference parametrizing slightly non-thermal spin distributions, rather than as fully thermalized spin reservoirs. In optical imaging of current-induced accumulation in topological insulators, the spatial symmetry establishes SHE-driven edge accumulation but does not quantitatively separate intrinsic and extrinsic contributions (Shitade, 29 Sep 2025, Belashchenko et al., 2023, Dejene et al., 2013, Liu et al., 2018).

Open problems follow directly from these limits. The magnetic ISA formalism leaves open whether the antisymmetric part of the response tensor obeys a Streda-type relation analogous to the symmetric octupolar relation. Thermal formulations point toward extensions to magnetic spin Nernst physics and to boundary-sensitive probes that directly image spin density rather than infer spin current. Photonic realizations suggest optimization of sector geometry, propagation distance, and ring radius for tailored spin accumulation profiles. In metallic and spin–orbit systems, a central unresolved issue is when transverse intrinsic accumulation can be reduced to a single effective diffusion length and when a microscopic nonequilibrium Green-function treatment is indispensable (Shitade, 29 Sep 2025, Shitade, 2022, Ling et al., 2014, Belashchenko et al., 2023).