Papers
Topics
Authors
Recent
Search
2000 character limit reached

In-Circuit Raman Sideband Cooling

Updated 4 July 2026
  • In-Circuit Raman sideband cooling is a technique that uses modulated coupling between a low-frequency target resonator and a high-frequency auxiliary resonator to transfer and remove excitations.
  • It employs an effective beam-splitter Hamiltonian that preferentially enables anti-Stokes processes in the resolved-sideband regime, significantly suppressing unwanted heating channels.
  • This method mirrors trapped-ion cooling, demonstrating how coherent sideband transfer not only cools to near-ground-state but also facilitates precise state control in quantum systems.

Searching arXiv for the cited papers and closely related work to ground the article. In-circuit Raman sideband cooling denotes a sideband-cooling protocol implemented within a superconducting circuit by frequency conversion between two linearly coupled harmonic modes: a low-frequency target resonator and a high-frequency auxiliary resonator whose coupling is modulated at their difference frequency. In this formulation, the effective interaction is a beam-splitter, or “Raman,” coupling that removes excitations from the target by converting them into excitations of the auxiliary, which is itself damped to a cold bath (Jacobs et al., 2010). A closely related use of the term “Raman sideband cooling” appears in trapped-ion systems, where optical Raman transitions between internal states are combined with motional sidebands; for a single 138^{138}Ba+^+ ion, this was implemented using the two S1/2S_{1/2} Zeeman sublevels and near-resonant 493 nm light, reaching nˉ≈0.17\bar{n}\approx 0.17 and nˉ=0.15(6)\bar{n}=0.15(6) depending on the diagnostic method (Seck et al., 2016). Taken together, these implementations show that “Raman sideband cooling” names an operational structure—coherent sideband transfer followed by dissipation—even when the microscopic hardware differs.

1. Conceptual definition and physical setting

In the superconducting-circuit realization, the system consists of two harmonic modes: a “mechanical” resonator of frequency ω≡ωm\omega \equiv \omega_m with annihilation operator aa, and a “superconducting” auxiliary LC or stripline resonator of frequency Ω≡ωc\Omega \equiv \omega_c with annihilation operator bb. Their coupling is modulated at the difference frequency Δ≡Ω−ω\Delta \equiv \Omega - \omega, with

H(t)=ℏ ω a†a+ℏ Ω b†b+ℏ g0cos⁡[(Ω−ω)t] (a+a†)(b+b†).H(t) = \hbar\,\omega\,a^\dagger a + \hbar\,\Omega\,b^\dagger b + \hbar\,g_0\cos\bigl[(\Omega-\omega)t\bigr]\,(a + a^\dagger)(b + b^\dagger).

Under the rotating-wave approximation, this becomes the effective beam-splitter interaction

Heff=ℏ g0 (a b†+a† b),H_{\rm eff} = \hbar\,g_0\,(a\,b^\dagger + a^\dagger\,b),

which transfers quanta from the target resonator to the auxiliary (Jacobs et al., 2010).

The trapped-ion realization provides a useful optical analogue. There, the two ground-state levels are ∣↑⟩≡∣S1/2,mj=+1/2⟩|\uparrow\rangle \equiv |S_{1/2},m_j=+1/2\rangle and ∣↓⟩≡∣S1/2,mj=−1/2⟩|\downarrow\rangle \equiv |S_{1/2},m_j=-1/2\rangle, separated by a Zeeman splitting ωZ=2π×10.97(2) MHz\omega_Z = 2\pi\times 10.97(2)\,\text{MHz} at B=3.919(7) GB=3.919(7)\,\text{G}. The Raman coupling to the axial mode of frequency ωz=2π×1.14 MHz\omega_z=2\pi\times 1.14\,\text{MHz} is described by

HI=ℏ Ωeff σx ei Δk⋅z+h.c.,H_I=\hbar\,\Omega_{\rm eff}\,\sigma_x\,e^{i\,\Delta k\cdot z}+h.c.,

with Ωeff=2π×89(2) kHz\Omega_{\rm eff}=2\pi\times 89(2)\,\text{kHz} and measured Lamb–Dicke parameter η=0.118(2)\eta=0.118(2) (Seck et al., 2016).

A plausible implication is that the term “Raman” is being used at two different microscopic levels. In the ion case it refers directly to a two-photon optical Raman process between internal states. In the circuit case it labels the effective resonant exchange interaction generated by modulated coupling. The shared content is not the optical mechanism but the coherent sideband-mediated transfer of energy into a dissipative auxiliary channel.

2. Hamiltonian structure and the sideband-resolved regime

For the superconducting circuit, moving to the interaction picture with respect to H0=ℏωa†a+ℏΩb†bH_0=\hbar\omega a^\dagger a+\hbar\Omega b^\dagger b yields

HI(t)=ℏ g0[a b† e−i(Ω−ω)t+a† b e+i(Ω−ω)t+a b e−i(Ω+ω)t+a† b† e+i(Ω+ω)t].H_I(t)=\hbar\,g_0\Bigl[a\,b^\dagger\,e^{-i(\Omega-\omega)t} + a^\dagger\,b\,e^{+i(\Omega-\omega)t} + a\,b\,e^{-i(\Omega+\omega)t} + a^\dagger\,b^\dagger\,e^{+i(\Omega+\omega)t}\Bigr].

Because the coupling is modulated at Δ=Ω−ω\Delta=\Omega-\omega, the exchange terms become time-independent while the counter-rotating terms oscillate at approximately 2ω2\omega. Under the usual rotating-wave approximation, valid for g0,κ,γ≪ωg_0,\kappa,\gamma \ll \omega, the latter are dropped, leaving the resonant exchange Hamiltonian HeffH_{\rm eff} (Jacobs et al., 2010).

The cooling and heating processes are then naturally identified as anti-Stokes and Stokes scattering channels. In the resolved-sideband regime κ≪ω\kappa \ll \omega and on resonance, the anti-Stokes rate is

Γ−≡AAS=g02 Sbb(−ω)≃g02κ,\Gamma_- \equiv A_{\rm AS} = g_0^2\,S_{bb}(-\omega) \simeq \frac{g_0^2}{\kappa},

while the Stokes rate is

Γ+≡AS=g02 Sbb(+ω)≃g02κ4ω2.\Gamma_+ \equiv A_{\rm S} = g_0^2\,S_{bb}(+\omega) \simeq \frac{g_0^2\kappa}{4\omega^2}.

The net cooling rate is therefore

Γnet=Γ−−Γ+≃g02κ(1−κ216ω2),\Gamma_{\rm net}=\Gamma_- - \Gamma_+ \simeq \frac{g_0^2}{\kappa}\Bigl(1-\frac{\kappa^2}{16\omega^2}\Bigr),

and the steady-state phonon number in the presence of mechanical heating γnT\gamma n_T is

nˉf=γnT+Γ+γ+Γ−−Γ+.\bar n_f=\frac{\gamma n_T+\Gamma_+}{\gamma+\Gamma_- - \Gamma_+}.

These expressions make explicit that the circuit protocol is a sideband-cooling scheme only when the auxiliary linewidth is sufficiently narrow compared with the target frequency (Jacobs et al., 2010).

The trapped-ion description has the same structural ingredients, but in a different language. Expanding Δk⋅z≈η(a+a†)\Delta k\cdot z \approx \eta(a+a^\dagger) in the Lamb–Dicke regime produces the usual red and blue sideband couplings, with

ΩredSB=η Ωeffn,ΩblueSB=η Ωeffn+1.\Omega_{\rm redSB}=\eta\,\Omega_{\rm eff}\sqrt{n}, \qquad \Omega_{\rm blueSB}=\eta\,\Omega_{\rm eff}\sqrt{n+1}.

This suggests that the unifying mathematical object across both platforms is the suppression of unwanted heating channels relative to a resonant exchange term that preferentially removes a single quantum from the mode of interest (Seck et al., 2016).

3. Cooling cycle, rates, and approach to the ground state

In the circuit implementation, the auxiliary mode bb is damped to a cold bath at rate κ\kappa, while the target mode aa is damped to a hot bath at rate γ\gamma with thermal occupancy nTn_T. The cooling mechanism is therefore indirect: excitations are first coherently transferred from aa to bb through HeffH_{\rm eff} and are then irreversibly lost through the damping of bb (Jacobs et al., 2010).

Using the numerical example given for typical circuit-QED parameters,

  • ω/2π=20 MHz\omega/2\pi = 20\,\text{MHz},
  • Qm=105⇒γ≈2π⋅200 HzQ_m=10^5 \Rightarrow \gamma \approx 2\pi\cdot 200\,\text{Hz},
  • Ω/2π=5 GHz\Omega/2\pi = 5\,\text{GHz},
  • κ=106 s−1\kappa = 10^6\,\text{s}^{-1},
  • g0=2×106 s−1g_0 = 2\times 10^6\,\text{s}^{-1},

one obtains

  • Γ−≃4×106 s−1\Gamma_- \simeq 4\times 10^6\,\text{s}^{-1},
  • Γ+≲102 s−1\Gamma_+ \lesssim 10^2\,\text{s}^{-1},
  • Γnet≃4×106 s−1\Gamma_{\rm net} \simeq 4\times 10^6\,\text{s}^{-1}.

For a mechanical bath at T=20 mKT=20\,\text{mK}, corresponding to nT≈20n_T\approx 20, the final occupancy is estimated as

nˉf≃γnTΓ−≃2×10−3,\bar n_f \simeq \frac{\gamma n_T}{\Gamma_-} \simeq 2\times 10^{-3},

so that ground-state occupation, in the sense ⟨n⟩≪1\langle n\rangle \ll 1, is achieved (Jacobs et al., 2010).

The trapped-ion implementation gives an experimentally resolved time-domain counterpart. After 20 ms20\,\text{ms} of Doppler cooling with 493 nm light detuned by approximately −Γ/2-\Gamma/2 and 650 nm repump light, the system is switched to a near-resonant Raman configuration for a total cooling time Tcool=1T_{\rm cool}=1–10 ms10\,\text{ms}. During this interval, the σ+\sigma^+ pump beam optically pumps ∣↓⟩→∣↑⟩|\downarrow\rangle \to |\uparrow\rangle while the π\pi probe drives ∣↑,n⟩→∣↓,n−1⟩|\uparrow,n\rangle \to |\downarrow,n-1\rangle on the red sideband. The repetition is continuous rather than pulse-by-pulse. The steady-state mean phonon number is estimated as nˉ≈0.17\bar n \approx 0.17 after approximately 10 ms10\,\text{ms} of cooling, and the temporal approach is described by

nˉ(t)=nˉ∞+[nˉ(0)−nˉ∞]e−(Γcool−Γheat)t,\bar n(t)=\bar n_\infty + [\bar n(0)-\bar n_\infty]e^{-(\Gamma_{\rm cool}-\Gamma_{\rm heat})t},

with Γcool−Γheat≈100 s−1\Gamma_{\rm cool}-\Gamma_{\rm heat}\approx 100\,\text{s}^{-1} inferred from equilibration in ≲10 ms\lesssim 10\,\text{ms} (Seck et al., 2016).

A common misconception is that sideband cooling is only a steady-state dissipative process. The circuit analysis explicitly states that the same effective resonant coupling also generates near-perfect state-swapping, while the ion experiment shows that continuous repumping can be used instead of discrete Raman pulse sequences. The methods differ in timing structure, but both exploit the same asymmetry between cooling and heating channels.

4. State swapping and coherent control

Under the effective Hamiltonian

Heff=ℏ g0(a b†+a†b),H_{\rm eff}=\hbar\,g_0(a\,b^\dagger + a^\dagger b),

and neglecting damping over a short interval, the Heisenberg evolution is

a(τ)=cos⁡(g0τ) a(0)−isin⁡(g0τ) b(0),b(τ)=cos⁡(g0τ) b(0)−isin⁡(g0τ) a(0).a(\tau)=\cos(g_0\tau)\,a(0)-i\sin(g_0\tau)\,b(0), \qquad b(\tau)=\cos(g_0\tau)\,b(0)-i\sin(g_0\tau)\,a(0).

At τ=π/(2g0)\tau=\pi/(2g_0), one obtains

a(τ)=−i b(0),b(τ)=−i a(0),a(\tau)=-i\,b(0), \qquad b(\tau)=-i\,a(0),

which is an exact SWAP up to a phase. A single Raman-swap pulse of this duration therefore prepares the target in the ground state if the auxiliary is initially in ∣0⟩|0\rangle (Jacobs et al., 2010).

The same work further treats the auxiliary as a coherent quantum feedback controller. If the auxiliary is driven by a classical coherent tone of amplitude β\beta at resonance, the linearized Heisenberg–Langevin equations are

a˙=−(γ/2) a−i g0 b+γ ain,\dot a = -(\gamma/2)\,a - i\,g_0\,b + \sqrt{\gamma}\,a_{\rm in},

b˙=−(κ/2) b−i g0 a+κ bin+β.\dot b = -(\kappa/2)\,b - i\,g_0\,a + \sqrt{\kappa}\,b_{\rm in} + \beta.

For γ≪κ\gamma\ll\kappa and g0<κ/4g_0<\kappa/4, the steady-state coherent amplitudes show that by choosing the phase and amplitude of β\beta, one can place ⟨a⟩\langle a\rangle anywhere in phase space, while the residual variance remains the cooled thermal occupancy nˉf\bar n_f (Jacobs et al., 2010).

This dual role of the auxiliary—as cold bath and controller—distinguishes the in-circuit formulation from the standard presentation of laser cooling. A plausible implication is that “Raman sideband cooling” in circuits is not merely a refrigeration primitive but also a control primitive, since the same exchange Hamiltonian that removes energy also transduces coherent drive information without measurement.

5. Relation to optical Raman sideband cooling in trapped ions

The 138^{138}Ba+^+ experiment provides a concrete reference point for what “Raman sideband cooling” means in the more conventional AMO setting. The S1/2S_{1/2} ground state is Zeeman split into mj=+1/2m_j=+1/2 and mj=−1/2m_j=-1/2 by an applied field, and the relevant optical transitions are S1/2→P1/2S_{1/2}\to P_{1/2} at 493 nm, D3/2→P1/2D_{3/2}\to P_{1/2} at 650 nm, and shelving/deshelving transitions at 455 nm and 614 nm. Near-resonant Raman light is produced by splitting the 493 nm beam into two arms called “Raman pump” and “Raman probe,” with both beams red-detuned by approximately 80 MHz from S1/2→P1/2S_{1/2}\to P_{1/2}. The beam geometry is chosen so that Δk⃗\Delta\vec{k} lies along the trap zz-axis, maximizing coupling to the axial mode (Seck et al., 2016).

The measured carrier Rabi frequencies are

Ωpump=2π×14.9(3) MHz,Ωprobe=2π×1.07(2) MHz,\Omega_{\rm pump}=2\pi\times 14.9(3)\,\text{MHz}, \qquad \Omega_{\rm probe}=2\pi\times 1.07(2)\,\text{MHz},

which produce

Ωeff=ΩpumpΩprobe2Δ1=2π×89(2) kHz.\Omega_{\rm eff}=\frac{\Omega_{\rm pump}\Omega_{\rm probe}}{2\Delta_1}=2\pi\times 89(2)\,\text{kHz}.

For motional-state detection, a second 493 nm ECDL is tuned Δfar≈−59 GHz\Delta_{\rm far}\approx -59\,\text{GHz} from S1/2→P1/2S_{1/2}\to P_{1/2}, yielding a measured carrier Rabi frequency Ωeff,car=2π×50.92(3) kHz\Omega_{\rm eff,car}=2\pi\times 50.92(3)\,\text{kHz}, blue-sideband Rabi frequency Ωeff,bsb=2π×6.02(8) kHz\Omega_{\rm eff,bsb}=2\pi\times 6.02(8)\,\text{kHz}, and decoherence rate γϕ≈2π×7.7(3) kHz\gamma_\phi\approx 2\pi\times 7.7(3)\,\text{kHz}, corresponding to a carrier coherence time of approximately 130 μs130\,\mu\text{s} (Seck et al., 2016).

The detection sequence is also explicitly specified: optical pumping to ∣↑⟩|\uparrow\rangle in 10 μs10\,\mu\text{s}, application of a far-off-resonant Raman π\pi-pulse on the red or blue sideband, protection of ∣↓⟩|\downarrow\rangle population by optical pumping into the D3/2D_{3/2} manifold, shelving of remaining S1/2S_{1/2} population in D5/2D_{5/2} with 455 nm light, and fluorescence discrimination with Doppler light. Fits to red- and blue-sideband shelving probabilities yield nˉ=0.15(6)\bar n=0.15(6), in excellent agreement with the near-resonant estimate of $0.17(3)$ (Seck et al., 2016).

The comparison with the circuit scheme can be summarized briefly:

Aspect Superconducting circuit 138^{138}Ba+^+ ion
Target mode Mechanical resonator Axial motional mode
Auxiliary degree of freedom LC or stripline resonator Zeeman sublevels plus optical pumping
Effective interaction Heff=ℏg0(ab†+a†b)H_{\rm eff}=\hbar g_0(a b^\dagger + a^\dagger b) Raman red-sideband coupling
Dissipative channel Auxiliary damping at rate κ\kappa Spontaneous scattering / optical pumping
Example final occupancy nˉf≃2×10−3\bar n_f \simeq 2\times 10^{-3} nˉ≈0.17\bar n\approx 0.17, nˉ=0.15(6)\bar n=0.15(6)

This suggests that the circuit and ion protocols are best viewed as hardware-specific realizations of the same sideband-cooling template: a resonant exchange step that lowers motional excitation, followed by a reset mechanism for the auxiliary subsystem.

6. Integration, resource requirements, and limitations

For the trapped-ion implementation, near-resonant Raman sideband cooling requires only the existing 493 nm Doppler laser and two additional AOM channels, rather than a new laser wavelength. The far-off-resonant detection path requires one more 493 nm ECDL and two extra AOM channels, with all RF drives generated from a single multi-channel DDS. The description further notes that on-chip photonic beam splitters and integrated AOMs could, in principle, further miniaturize the arrangement; magnetic-field stabilization could lock the Zeeman splitting to less than 1 kHz drift; and multi-ion or multi-mode operation would require beam steering or multiple AOM channels, while preserving the basic building blocks of the setup (Seck et al., 2016).

For the superconducting-circuit implementation, the decisive operating conditions are encoded directly in the model: the rotating-wave approximation requires g0,κ,γ≪ωg_0,\kappa,\gamma \ll \omega, and efficient cooling requires the resolved-sideband regime κ≪ω\kappa \ll \omega. The heating channel scales as Γ+≃g02κ/(4ω2)\Gamma_+ \simeq g_0^2\kappa/(4\omega^2), so the suppression of Stokes processes is inseparable from spectral resolution. At the same time, the coherent control analysis is given for γ≪κ\gamma\ll\kappa and g0<κ/4g_0<\kappa/4 (Jacobs et al., 2010).

An objective source of confusion in the literature is the phrase “in-circuit Raman sideband cooling.” The circuit paper describes the relevant interaction as a beam-splitter, or “Raman,” interaction generated by modulating a linear coupling, whereas the ion paper uses near-resonant and far-off-resonant optical Raman transitions between Zeeman sublevels. The commonality is therefore operational rather than microscopic. A plausible implication is that the phrase is most precise when it emphasizes the effective sideband exchange mechanism inside the circuit, not an optical Raman process in the usual atomic-physics sense.

Within the numerical and experimental examples provided, the two implementations also illustrate different performance regimes. The circuit analysis estimates nˉf≃2×10−3\bar n_f \simeq 2\times 10^{-3} under representative parameters, while the 138^{138}Ba+^+ protocol attains a steady-state occupancy nˉ≈0.17\bar n\approx 0.17 in ≲10 ms\lesssim 10\,\text{ms} and motional analysis contrast greater than 40%, which is described as sufficient for quantum logic spectroscopy with 138^{138}Ba+^+ without a narrow 1.76 μ\mum laser or hyperfine structure (Jacobs et al., 2010, Seck et al., 2016).

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to In-Circuit Raman Sideband Cooling.