
2
linear dynamics underscores the utility of this scheme. The
analysis presented in this letter is generic, pertinent to a wide
class of systems with a parametrically driven component. This
is because the DDI is ubiquitous. The investigated paradigm
would apply, for example, to nonlinear Kerr boson systems
driven far from equilibrium [44].
To set the stage, we shortly recapitulate the problem of two
spatially separated macroscopic ferrite samples of YIG cou-
pled to the microcavity field. Owing to an effective cavity-
mediated coupling between the two spheres, an external driv-
ing field applied to the first YIG sample would elicit a spin
current from the second one. The system Hamiltonian, in the
reference frame of the driving field, assumes the form [17, 45]
H0/~= ∆ca†a+
2
X
j=1
∆jm†
jmj+
2
X
j=1
gj[a†mj+am†]
+iΩ(m†
1−m1),(1)
where ais the annihilation operator representing the cavity,
∆c=ωc−ωdis the cavity detuning, m1,m2are the two Kittel
modes representing magnonic excitations in the two samples,
and ∆1=ω1−ωd,∆2=ω2−ωdthe respective detunings.
For each j, the parameter gj=(√5/2)γepNjBvac denotes the
coherent magnon-photon interaction strength, with γebeing
the gyromagnetic ratio, Bvac =rµ0~ωc
2Vc
the magnetic field
of vacuum, and Njthe total number of spins in the sample.
Plus, Ω = γe
2r5µ0ρ1d1Dp
3cis the Rabi frequency of the ap-
plied drive, where ρ1and d1are the respective spin density
and diameter of the first sample, while Dpis incident power
of the applied drive. To understand this transfer of spin-wave
excitations at the level of transitions among energy levels, we
note that in the absence of the extrinsic magnon drive, the
Hamiltonian is excitation-preserving, with only an oscillatory
energy transfer between the magnon and the cavity modes.
Now let us label the eigenstates of the noninteracting sys-
tem as |na,n1,n2i, where na,n1and n2indicate populations
of the cavity and the two magnon modes respectively. If a
weak coherent drive (at the single-photon level) on m1excites
the system into the state |0,1,0i, energy would be exchanged
back and forth among the energy levels |0,1,0i,|1,0,0i, and
|0,0,1i, provided the dissipation is negligible. The transfer
happens via the pathway |0,1,0i→|1,0,0i→|0,0,1i, and
this simple scheme can be easily extended to the case of co-
herent drives.
Amplified spin current: We now demonstrate the impact
of a parametric drive applied to the cavity on the associated
transfer efficiency in the system considered above. Precisely,
we would be leveraging the potential of parametrically en-
hanced spin-photon interactions to amplify the spin currents
from magnetic samples loaded into a cavity resonator. The
new schematic is portrayed in Fig. 1, where the cavity field
is now parametrically driven. The preceding Hamiltonian has
to be supplemented by an additional contribution of the form
FIG. 1: Schematic of two ferrimagnetic samples of YIG, co-
herently coupled to a single-mode cavity, which is driven ex-
ternally by a two-photon parametric drive. A uniform bias
magnetic field B0applied to either of the YIG spheres gener-
ates the corresponding Kittel mode and the YIG1 is driven
externally by a coherent drive of low photon occupancy.
Hp/~=(G/2)(a2+a†2), so that the Hamiltonian for the para-
metric system [22, 23, 25] would be given by H=H0+Hp.
To quell the time-dependence of the Hamiltonian, the fre-
quency of the applied magnon drive ωdhas been set equal
to ωp/2. Our objective is to investigate the steady-state spin-
current response from the second YIG. Under the semiclas-
sical approximation, the dynamical equations for the mode
amplitudes at the level of mean fields can be obtained from
the master equation of the system. These equations can be
condensed in the form of a 6 ×6-matrix-differential equation
˙
X=−iHeffX+ ΩFin,(2)
where X=(a m1m2a†m†
1m†
2)T,Heff= H(0)
effJ
−J−H∗(0)
eff!
is a 6 ×6 coupling matrix, and Fin =(0 1 0 0 1 0)T.
The expectation-value notations h.ihave been dropped for
brevity. The constituent block elements of Heffare given by
H(0)
eff=
∆c−iκg1g2
g1∆1−iγ10
g20∆2−iγ2
and J=
G0 0
0 0 0
0 0 0
, where
2κ, 2γ1,2γ2are the respective relaxation rates of the cavity
mode and the two magnon modes. Thus H(0)
effdenotes the cou-
pling matrix in the absence of the parametric drive, i.e., with
Gset equal to 0. Eq. (4) would permit a steady-state so-
lution insofar as the eigenmodes of Heffhave decaying char-
acter. Subject to this assumption, the steady-state spin-current
response from the second magnetic sample could be expressed
as M=|m2|2, wherein m2=Pj=2,5−iH−1
eff2j
Ω.
To keep the analysis straightforward, we henceforth work
with the assumption that ∆1= ∆2= ∆c= ∆,κ=γ1=γ2=γ,
and g1=g2=g. As was just stated, the stability of the steady
state hinges on the imaginary parts of the eigenvalues, which
we label as λi’s. This is formally equivalent to the Routh-
Hurwitz criterion for the stability of nonlinear dynamics. The
stable and the unstable regimes for this parametric model are
numerically plotted in Fig. 2(a), for the ratio g/κ =2, with
the demarcating partition between the two indentified by the