
2
FIG. 1. Schematic view of 1D model of a FMp/NM/FMaspin
valve where a spin-unpolarized current pulse I(t)—carrying
charge Q=RI(t)dt =Neecomprised of one (Ne= 1) or
more (Ne>1) electrons—is injected into the polarizing FMp
layer. After traversing it to become spin-polarized, it im-
pinges onto the analyzing FMalayer where it transfers part
of its spin angular momentum onto the localized spins via
quantum STT [55,56]. Unlike in conventional Slonczewski-
Berger STT studies [52]—where localized spins within FMp
and FMalayers are modeled by classical vectors [51]—we re-
tain a fully quantum description based on spin operators.
Their ground state expectation values ⟨ˆ
Si⟩(t≤0) are de-
picted by red arrows and are arranged into a collinear but
antiparallel geometry in which conventional STT is identi-
cally zero [52]. This process dynamically generates a mixed
entangled quantum state of the FMp∪FMasubsystem. To
account for its dissipation and phase decoherence, we cou-
ple each spin to its own bosonic bath [57,58], where all such
baths are kept in equilibrium at temperature Tbb. Delayed X-
ray pulses, with incoming momentum and energy (ℏki,ℏωi),
are assumed to be shined during and after the current pulse
duration. X-rays, with momentum and energy (ℏkf,ℏωf),
scattered off the FMp∪FMasubsystem facilitate a witness-
ing scheme of nonequilibrium entanglement we propose based
on extraction [45] of time-dependent quantum Fisher infor-
mation from trRIXS response function [66–68].
sure effects inside microcavities [25–27], have relied upon
the use of photons as mediators of quantum correlations
[28,29]. Recently, a greater interest has been placed on
trying to reproduce these effects in a solid state setup. In
effect, a small ∼10 [30] or moderate ∼103[31] number of
spins in solids have already been entangled at distances of
a few lattice constants, while predictions [32,33] of long-
range entanglement among spin-1/2 probes in strongly
correlated systems have also been recently realized in ex-
periments with superconducting flux circuits [34]. More-
over, several recent theoretical works [35–38] prescribe a
way to entangle much larger spin ensembles (N≳1016)
residing within two distant spheres carved out of a fer-
rimagnetic insulator, using the cavity photon modes as
entanglement mediators. It is important to note that
quantum entanglement of a macroscopic number of de-
grees of freedom is ubiquitous in the ground or low-lying
excited states of strongly electron-correlated materials,
such as superconductors [11], quantum spin liquids [39],
antiferromagnets [12,40–43], and Hubbard model mate-
rials [44,45]. However, in practice, it is extremely chal-
lenging to isolate different subsystems of such systems
and then probe their mutual entanglement.
At first sight, it seems that none of the plethora
of nonequilibrium spin-dependent phenomena, involving
itinerant electrons and localized spins in typical spin-
tronic devices, like spin valves (SVs) and magnetic tunnel
junctions (MTJs), would be useful for investigating large-
scale entanglement of well-separated quantum units. An-
ticipating the deception of this expectation, we still recall
that SVs and MTJs are composed of two macroscopic
FM layers hosting a very large number of localized spins,
usually derived from d-orbitals of Fe, Ni, or Co. These
FM layers are separated by a few nanometers thick NM
spacer (such as Cu) in SVs (as illustrated schematically
by our 1D model in Fig. 1) or by an insulating barrier
(such as MgO) in the case of MTJs. At standard room-
temperature, the value of these localized spins within
FMs are typically S > 1, which falls outside [47] of the
“ultra-quantum” limit, where quantum corrections to the
S2(1 + 1/S) eigenvalue of the ˆ
S2
ioperator are signifi-
cant [48]. This fact has been used to intuitively (but not
rigorously [49]) justify the modeling of spin dynamics [50]
and injected electronic currents [51] in the presence of
magnetic fields, by means of the Landau-Lifshitz-Gilbert
(LLG) equation [50,51], which treats localized spins as
classical vectors of fixed length. The extended LLG equa-
tion [51] includes a conventional (Slonczewski-Berger)
spin-transfer torque (STT) [52] term describing spin an-
gular momentum exchange at a semiclassical level. Such
a term may be phenomenological, as in classical micro-
magnetics codes [51], or it can be computed microscopi-
cally from some steady-state [53] or time-dependent [54]
single-particle quantum transport theory.
Defying conventional wisdom, recent experiments at
ultralow temperatures (T∼1 K) [55] have observed
current-driven magnetization dynamics in SVs that
started from collinear magnetizations in the two FM
layers. In this situation, the conventional STT is iden-
tically zero and the system’s dynamics cannot be un-
derstood within the LLG paradigm. This has motivated
the development of a quantum STT theory, where both
the flowing electronic spins and localized spins must be
treated quantum-mechanically [56,59–62]. Even though
a ferromagnet in equilibrium remains in a separable (un-
entangled) quantum state under various externally im-
posed conditions [63], a single FM layer can be driven by
a spin-polarized current to experience a quantum STT
which induces a dynamical build-up of long-range entan-
glement [56,64]. The simplest signature of such entan-
glement is a shrinking in the expectation value of the
localized spin magnitude, i.e.,|⟨ˆ
Si⟩(t)|<Sℏ(ibeing the
site of the crystalline lattice). In some circumstances,
these expectation values can even be reduced to zero [56],
which further explains the failure of the classical LLG
equation [51,52] to describe the STT-driven magnetiza-
tion dynamics on these systems [49,65]. The quantum
nature of the problem then calls for a full-fledged many-
body approach that captures the intrinsic quantum na-
ture of localized spins and thus goes beyond the common