Anti-Stokes Brillouin-Mandelstam Scattering

• Anti-Stokes Brillouin-Mandelstam scattering is defined as the inelastic process where photons absorb phonons, resulting in a blue-shifted optical signal.
• It operates across spontaneous, stimulated, and cascaded regimes in platforms such as optical fibers, silicon waveguides, and plasmas, featuring unique phase matching and modal asymmetry.
• The phenomenon enables applications like phonon annihilation for Brillouin cooling, nonreciprocal transmission, microwave-to-optical transduction, and coherent photon-phonon exchange.

Searching arXiv for recent and foundational papers on anti-Stokes Brillouin-Mandelstam scattering and related Brillouin platforms. Anti-Stokes Brillouin-Mandelstam scattering is the inelastic light-sound process in which a photon gains energy by absorbing a phonon and is shifted to higher frequency. Within the Brillouin-Mandelstam framework, it appears in spontaneous, stimulated, forward, backward, intra-modal, and inter-modal forms, and it can be mediated by thermally populated acoustic modes, coherent acoustic phonon pulses, guided acoustic waves, ion-acoustic waves, or electromechanically driven phonons. Across these settings, the anti-Stokes channel is not merely the blue-shifted counterpart of Stokes scattering; the supplied literature shows that it can be decoupled from Stokes scattering by phase matching, made to interfere coherently with Stokes pathways, used to annihilate phonons for cooling, and exploited for nonreciprocity, microwave-to-optical transduction, quantum state exchange, and analog computation (Lai et al., 2020, Kittlaus et al., 2016, Liu et al., 2019, Chen et al., 2016, Zoubi, 2022).

1. Scattering channel, conservation laws, and modal asymmetry

In the standard distinction, Stokes Brillouin scattering is the process in which the photon loses energy by emitting a phonon, whereas anti-Stokes Brillouin scattering is the process in which the photon gains energy by absorbing a phonon (Lai et al., 2020). In a forward-scattering linear waveguide, the anti-Stokes interaction is written as a three-wave mixing process among a pump optical field $a_1(z,t)$, an anti-Stokes optical field $a_2(z,t)$, and an acoustic field $b(z,t)$, with conservation laws

$$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$

The coupled equations

$$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$

make explicit that the anti-Stokes field and the phonon mode form the central dissipative subsystem in the cooling problem (Chen et al., 2016).

A recurring result in the literature is that the symmetry between Stokes and anti-Stokes channels is highly platform-dependent rather than universal. In stimulated inter-modal Brillouin scattering in a multimode silicon waveguide, the Stokes and anti-Stokes phase-matching conditions are

$$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$

and the paper states explicitly that $q_s \neq q_{as}$. Because the two processes require different phonons, the dynamics are decoupled and single-sideband operation becomes possible (Kittlaus et al., 2016). In the silicon nanowire treatment of inter-modal forward Brillouin scattering, the same decoupling is expressed by the relation $q_s \neq q_{as}$ together with $\Omega_s=\Omega_{as}=\Omega$, so the symmetry breaking comes from wavevector selection rather than phonon frequency splitting (Zoubi, 2022).

This body of work suggests that anti-Stokes Brillouin-Mandelstam scattering is best understood not as a fixed mirror image of Stokes scattering, but as one branch of a photon-phonon interaction network whose accessibility is controlled by phase matching, modal dispersion, and the available acoustic state.

2. Reflection-induced anti-Stokes/Stokes interference and complex Fano resonance

A particularly detailed treatment arises in ultrafast acoustics, where anti-Stokes Brillouin-Mandelstam scattering is one branch of coherent light scattering from a coherent acoustic phonon pulse. In that setting the acoustic field is a phase-correlated strain packet $\eta(z-vt)$, not a random thermal phonon bath, and reflection of the coherent acoustic phonon pulse at an interface abruptly reverses the phonon momentum. The paper identifies that reversal as the mechanism by which the optical response switches from a single branch to a simultaneous mixture of anti-Stokes and Stokes scattering, allowing the two pathways to interfere in one detection event (Lai et al., 2020).

For a monochromatic probe of frequency $a_2(z,t)$0, the relative reflectance is written as

$a_2(z,t)$1

The two terms in brackets correspond, respectively, to the anti-Stokes and Stokes branches. In frequency space the scattering cross-section becomes

$a_2(z,t)$2

with

$a_2(z,t)$3

The formulation makes the anti-Stokes/Stokes structure explicit: the resonances are centered at opposite detunings, and their overlap depends on $a_2(z,t)$4, $a_2(z,t)$5, and the acoustic reflection coefficient $a_2(z,t)$6 (Lai et al., 2020).

The same paper rewrites the superposition as a complex Fano resonance,

$a_2(z,t)$7

Here the real part of $a_2(z,t)$8 is associated with resonance overlap, while the imaginary part encodes losses through reflection. A Fano profile appears only when the Stokes and anti-Stokes resonances overlap sufficiently in frequency, which requires substantial optical absorption $a_2(z,t)$9 and a finite reflection coefficient $b(z,t)$0. In the weak-absorption limit the overlap disappears and the asymmetric profile collapses toward independent Lorentzians (Lai et al., 2020).

The experimental verification used picosecond ultrasonic pump-probe measurements on a $b(z,t)$1 nm tungsten film on Si with a tunable Ti:sapphire oscillator generating $b(z,t)$2 fs pulses in a one-color geometry. Acoustic echoes appeared at about $b(z,t)$3 ps and $b(z,t)$4 ps, and the Fourier transforms of those echoes exhibited Fano-type asymmetric dips. The model reproduced the measured spectra using $b(z,t)$5 and literature optical and photoelastic constants. The same formalism was then extended to dissipation, modeled by reducing $b(z,t)$6, and decoherence, modeled through a random phase at reflection with standard deviation $b(z,t)$7, with distinct trajectories of the complex Fano parameter in the complex plane (Lai et al., 2020).

3. Stimulated, cascaded, and envelope-based anti-Stokes regimes

In high-intensity laser-plasma interaction, anti-Stokes Brillouin-Mandelstam scattering emerges as part of a stimulated Brillouin scattering cascade rather than as a primary instability. The cascade begins with

$b(z,t)$8

followed by

$b(z,t)$9

and, once the second-stage ion-acoustic wave is sufficiently strong, the pump can generate a higher-frequency anti-Stokes branch through

$$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$0

The paper identifies this branch as stimulated anti-Stokes Brillouin scattering, or SABS, and reports that it competes directly with further SBS rescattering in determining the total SBS reflectivity (Feng et al., 2017).

The plasma analysis uses the matching relations

$$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$1

together with the approximations $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$2 and $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$3 for backward scattering. For the carbon plasma parameters considered, the authors obtain $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$4, so the anti-Stokes ladder is approximately

$$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$5

The simulations show anti-Stokes peaks at about $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$6, $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$7, $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$8, whereas the Stokes cascade appears near $$\omega_1 + \Omega = \omega_2, \qquad k_1 + q = k_2.$$9, $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$0, $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$1. In the reported regime, cascade and SABS are negligible below about $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$2 and become important around $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$3 (Feng et al., 2017).

A complementary theoretical development concerns reduced-envelope descriptions of cascaded Brillouin dynamics. In a backward SBS comb model, the multitude of pump, Stokes, and anti-Stokes optical fields are represented by one forward optical envelope $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$4, one backward optical envelope $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$5, and one acoustic envelope $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$6, all referenced to a single carrier frequency $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$7. In that framework, anti-Stokes orders are not inserted as separate fundamental fields; they arise through four-wave mixing once there are at least two co-propagating pump and Stokes waves, and the frequency ladder is $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$8 with positive $$a_{2t}+v a_{2z}=-\gamma a_2-ig b, \qquad b_t+v_b b_z=-\Gamma b-ig^* a_2+\xi$$9 corresponding to anti-Stokes orders (Dong et al., 2015).

An analogous compact description exists for cascaded forward Brillouin scattering. The optical field is expanded as

$$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$0

with $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$1 denoting Stokes orders and $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$2 denoting anti-Stokes orders. The coupled sideband equations

$$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$3

show that anti-Stokes couplings are intrinsic to the same ladder as Stokes couplings. For negligible optical dispersion, the exact solution gives Bessel-function sidebands $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$4, so the cascade is a pure phase modulation containing symmetric Stokes and anti-Stokes orders. With weak optical dispersion, that symmetry is broken and amplitude modulation becomes observable (Wolff et al., 2016).

4. Integrated photonics, single-sideband operation, and electromechanical anti-Stokes generation

Integrated photonics has provided several settings in which the anti-Stokes branch can be selected rather than merely observed. In silicon, stimulated inter-modal Brillouin scattering couples a symmetric optical mode $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$5, an anti-symmetric optical mode $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$6, and a guided acoustic mode $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$7. Because the Stokes and anti-Stokes phonons are distinct, symmetry is broken between the two processes, permitting unidirectional single-sideband coupling between only two optical fields. The reported device demonstrated $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$8 dB of optical gain, over $$q_s = k_1(\omega_p)-k_2(\omega_p-\Omega_s), \qquad q_{as} = k_1(\omega_p+\Omega_{as})-k_2(\omega_p),$$9 dB of net amplification, and $q_s \neq q_{as}$0 single-sideband energy transfer between two optical modes in a pure silicon waveguide. The fitted Brillouin gain coefficient was $q_s \neq q_{as}$1, the simulated value was $q_s \neq q_{as}$2, and the resonance occurred at $q_s \neq q_{as}$3 with $q_s \neq q_{as}$4 (Kittlaus et al., 2016).

A different route uses electromechanical excitation of the phonon field. In a suspended AlN optomechanical waveguide, acoustic phonons of $q_s \neq q_{as}$5 GHz are excited by an interdigital transducer and scatter a counter-propagating optical carrier into a single anti-Stokes sideband. The backward Brillouin phase matching is stated as the Bragg condition

$q_s \neq q_{as}$6

For the stated parameters the simulated acoustic frequency is $q_s \neq q_{as}$7 GHz, the main acoustic resonance in the experiment is $q_s \neq q_{as}$8 GHz, and the phase-matching curve in the wavelength-frequency map has slope $q_s \neq q_{as}$9. In the key measurement only one sideband peak was observed, shifted by exactly $q_s \neq q_{as}$0 GHz above the carrier, and no Stokes sideband was observed. The device was therefore described as an acousto-optic single-sideband modulator with tunable transmission bands of about $q_s \neq q_{as}$1–$q_s \neq q_{as}$2 MHz and tuning spans of about $q_s \neq q_{as}$3 MHz in one scheme and $q_s \neq q_{as}$4 MHz in another (Liu et al., 2019).

In dual nanoweb waveguides, exact phase matching can be engineered for a cascade of Stokes and anti-Stokes orders mediated by flexural phonons. The Brillouin-like intermodal channel alternates between slow and fast optical supermodes, with even-order sidebands in the slow mode and odd-order sidebands in the fast mode. The same work predicts continuous-wave generation of frequency combs without any kind of cavity and periodic reversal of the energy flow between mechanical and optical modes (Noskov et al., 2017).

These platforms collectively show that anti-Stokes Brillouin-Mandelstam scattering can be promoted to a design variable: sideband asymmetry can be imposed by modal dispersion, by unidirectional guided phonons, or by engineered flexural resonances.

5. Phonon annihilation, Brillouin cooling, and coherent photon-phonon exchange

Because anti-Stokes scattering annihilates phonons, it is the cooling branch of Brillouin optomechanics. In the resonator-based Brillouin cooling analysis, the required phase matching is

$q_s \neq q_{as}$5

and the three-mode interaction has the form

$q_s \neq q_{as}$6

The central rate-equation result is

$q_s \neq q_{as}$7

so increasing the dissipation of the optical anti-Stokes resonance improves the cooling ratio by allowing anti-Stokes photons to leave the cavity efficiently. For a $q_s \neq q_{as}$8 silica sphere with optical $q_s \neq q_{as}$9, a mechanical mode near $\Omega_s=\Omega_{as}=\Omega$0 MHz, and a telecom pump at $\Omega_s=\Omega_{as}=\Omega$1, the theory predicts cooling ratios of about $\Omega_s=\Omega_{as}=\Omega$2, and with deliberately reduced anti-Stokes $\Omega_s=\Omega_{as}=\Omega$3 the paper states that cooling ratios above $\Omega_s=\Omega_{as}=\Omega$4 may be achievable (Tomes et al., 2011).

In a linear waveguide, the relevant control parameters are spatial rather than purely temporal. The paper defines

$\Omega_s=\Omega_{as}=\Omega$5

and finds the large-coupling limit

$\Omega_s=\Omega_{as}=\Omega$6

Measurable cooling requires the phonon spatial loss rate to be of the same order as the spatial optical loss rate. The paper states that appreciable cooling may occur if the Brillouin gain reaches the order of $\Omega_s=\Omega_{as}=\Omega$7, with modest pump power of a few mW. It also distinguishes low-group-velocity phonons, for which no appreciable cooling occurs even with $\Omega_s=\Omega_{as}=\Omega$8, from high-group-velocity phonons, for which the phonon spectral density at $\Omega_s=\Omega_{as}=\Omega$9 can reach about $\eta(z-vt)$0 at long distance and, in a high-$\eta(z-vt)$1 example, the cooling ratio saturates at about $\eta(z-vt)$2 (Chen et al., 2016).

The anti-Stokes channel also supports coherent photon-phonon exchange rather than only dissipation. In stimulated inter-modal forward Brillouin scattering in silicon nanowires, the linearized anti-Stokes Hamiltonian is

$\eta(z-vt)$3

which is of beam-splitter type rather than pair-creation type. Diagonalization yields polaritonic collective modes with resonance splitting

$\eta(z-vt)$4

The numerical regime quoted in the paper is $\eta(z-vt)$5, $\eta(z-vt)$6, $\eta(z-vt)$7, $\eta(z-vt)$8, $\eta(z-vt)$9, and $a_2(z,t)$00, so that $a_2(z,t)$01 and the anti-Stokes polaritons lie in the strong-coupling regime. The same paper notes that cooling to about $a_2(z,t)$02 mK is needed to suppress thermal occupation of $a_2(z,t)$03 GHz phonons and approach the quantum regime (Zoubi, 2022).

6. Detection, nonreciprocity, and emerging applications

Anti-Stokes Brillouin-Mandelstam scattering is observable even without external acoustic excitation when thermally populated phonons provide the acoustic field. In a few-mode step-index fiber, heterodyne detection resolved both Stokes and anti-Stokes components simultaneously for forward and backward scattering and for both intra-modal and inter-modal processes. The phase-matching conditions were written as

$a_2(z,t)$04

with forward scattering additionally satisfying

$a_2(z,t)$05

The reported platform was a $a_2(z,t)$06-km SMF-28 Ultra fiber at $a_2(z,t)$07 nm supporting the $a_2(z,t)$08 and $a_2(z,t)$09 modes. The strongest backward gain was $a_2(z,t)$10 for the $a_2(z,t)$11 mode, while the strongest forward gain observed was $a_2(z,t)$12. In forward scattering the Stokes and anti-Stokes components were nearly identical in strength in the weak-gain regime, whereas in backward scattering the anti-Stokes signal was much weaker than Stokes (Kikuchi et al., 10 Jan 2026).

A different anti-Stokes analogue appears in a space-time modulated dielectric medium. There the traveling modulation $a_2(z,t)$13 couples an incident electromagnetic wave to Bloch-Floquet harmonics, and for backward propagation the dominant strong coupling is to the $a_2(z,t)$14 harmonic,

$a_2(z,t)$15

The paper interprets the backward directional bandgap as an Anti-Stokes scattering center and gives the condition for non-overlap of forward and backward gaps as

$a_2(z,t)$16

Although the forward and backward bandgaps generally occur at different frequencies, their widths and insertion losses are equal, and the backward case produces a reflected blue-shifted wave that bounces back to the source (ELnaggar et al., 2017).

The anti-Stokes channel has also been assigned explicit device functionality. In an optical coprocessor based on spontaneous Brillouin scattering, ring resonators support anti-Stokes optical modes pumped via spontaneous Brillouin scattering on thermal phonons. After tracing out the phonon reservoirs, the stationary waveguide anti-Stokes occupancy is

$a_2(z,t)$17

so the anti-Stokes output is a weighted sum of thermal phonon occupancies. In the strong phonon-pumping and weak-waveguide-loss regime, the simplified weights are

$a_2(z,t)$18

which defines a matrix-vector multiplication. The same paper estimates parallelism over frequency channels with

$a_2(z,t)$19

and states $a_2(z,t)$20 for high-$a_2(z,t)$21 resonators (Vovchenko et al., 17 Dec 2025).

A common misconception is that anti-Stokes Brillouin-Mandelstam scattering is always a weak, derivative, or parasitic branch. The literature assembled here shows instead that it can be the dominant diagnostic signature of interface reflection, a direct competitor in SBS saturation, the selected sideband in integrated modulation, the cooling channel in traveling-wave and resonator optomechanics, the coherent exchange channel in photon-phonon polariton formation, the reflected branch of nonreciprocal space-time scattering, and the observable thermal-noise counterpart of Stokes scattering in few-mode fibers (Lai et al., 2020, Feng et al., 2017, Liu et al., 2019, Chen et al., 2016, Zoubi, 2022, Kikuchi et al., 10 Jan 2026).