Overcoming the SQL in gravitational wave detectors using spin systems with negative effective mass

F.Ya.Khalili∗ Faculty of , M.V. Lomonosov Moscow State University, 119991 Moscow, Russia and Russian Center, Skolkovo 143025, Russia

E.S.Polzik† Niels Bohr Institute, University of Copenhagen, 2100, Denmark

Quantum back action (QBA) of a measurement limits the precision of observation of the of a free mass. This profound effect dabbed the "Heisenberg microscope" in the early days of quantum , leads to the standard (SQL) [1] stemming from the balance between the measurement sensitivity and the QBA. Here we consider the measurement of motion of a free mass performed in a quantum reference frame with an effective negative mass which is not limited by QBA. As a result, the disturbance on the motion of a free mass can be measured beyond SQL. QBA-limited detection of motion for a free mass is extremely challenging, but there are devices where this effect is expected to play an essential role, namely, gravitational wave detectors (GWD) such as LIGO and VIRGO. Recent reports on observation of gravitational waves [2–6] have opened new horizons in cosmology and astrophysics. Here we present a general idea and a detailed numerical analysis for QBA-evading measurement of the gravitational wave effect on the GWD mirrors which can be considered free masses under relevant conditions. The measurement is performed by two entangled beams of light probing the GWD and an auxiliary atomic spin ensemble, respectively. The latter plays a role of a free negative mass. We show that under realistic conditions the sensitivity of the GWD can be significantly increased over the entire frequency band of interest.

a. Introduction Position of a mass as a function of time, is the origin of the QBA noise. QBA dominates for longer xˆ(t) = xˆ(0) + xÛˆ(0)t = xˆ(0) + pˆ(0)t/m, cannot be determined measurement times, as its intensity is proportional to the free precisely because xˆ(0) and pˆ(0) are non-commuting operators mass mechanical susceptibility of suspended mirrors of GW and if one of them is measured, the other one is disturbed detectors, χ = −1/Ω2. For the Advanced LIGO, the radiation by QBA. However, if the position is measured relatively to pressure (QBA) noise will be a major limitation [21]. a reference frame associated with another quantum system described by xˆ0(0) and pˆ0(0) and negative effective mass −m, Suppression of both the shot noise and the QBA, and thus then xˆ(t)−xˆ0(t) = xˆ(0)−xˆ0(0)+(xÛˆ(0)−xÛˆ0(0))t = xˆ(0)−xˆ0(0)+ overcoming the SQL, requires more advanced methods than or- (pˆ(0) + pˆ0(0))t/m. In this case, the relative position is defined dinary frequency-independent squeezing. Methods proposed by commuting operators xˆ(0) − xˆ0(0) and pˆ(0) + pˆ0(0) and towards this goal to-date [22–25] are challenging as they in- hence can be QBA-free. Narrowband QBA evading methods volve large scale installations and/or modifications of the GWD for high frequency oscillators have been proposed and utilized core optics. experimentally [7–12]. In [13, 14] the negative mass approach has been analyzed from an information-theoretical perspective. Here we show that a suitably designed atomic spin ensemble In this Letter we propose an idea and present a detailed provides a reference frame in which a broadband quantum analysis for application of the measurement of a free mass in noise reduction for motion of free masses, such as the GWD a negative mass reference frame to gravitational wave detec- mirrors, is possible. tors (GWDs). Due to the unique technical parameters of the GWDs, their massive mirrors can be considered free masses in b. The scheme. The schematic of the proposed experi- the relevant range of time and frequency. At the same time sen- mental realization for detection of the free mass motion in sitivity of state-of-the-art interferometric GWDs, such as the negative mass reference frame is presented in Figure1. Advanced LIGO [15], Advanced VIRGO [16], and GEO600 Two quantum measurements are performed in parallel, the arXiv:1710.10405v1 [quant-ph] 28 Oct 2017 [17] is to a major extent limited by quantum fluctuations of the measurement of the position of the end mirrors of the GW probing light. interferometer with the optical field aˆI , and the measurement For short measurement times (a higher frequency range) on the auxiliary atomic spin ensemble with the field aˆS. The the sensitivity of measurement of motion is dominated by the two fields are centered at wavelengths λI and λS, respectively, phase shot noise of light. This noise can be suppressed [18] where λI is determined by the probing laser of the GWD by using a phase-squeezed state of light, as demonstrated at (presently 1064nm) and λS - by an atomic resonant transition. the GEO600 GW detector [19, 20]. However, suppression As we shall show below if aˆI and aˆS are in an entangled state, of the phase fluctuations leads to proportional increase of the both the shot noise and the radiation pressure (QBA) noise amplitude fluctuations, or the radiation pressure noise which contributions to the joint measurement on the two systems can be suppressed.

In the absence of optical losses, with the interferometer ∗ [email protected] tuned on resonance, the phase quadrature of the light mode ex- † [email protected] ˆ s iting the interferometer, b1, measured by a homodyne detector 2

FIG. 1. Setup for a GWD beyond the SQL with the negative mass FIG. 2. Macroscopic spin oscillator. a). A collective spin of an spin system. The GWD and the atomic system are probed with en- ensemble of atoms in magnetic field shown as a vector on a Bloch tangled light modes aS, aI (dashed lines). The modes are generated sphere. b) Normalized spin components orthogonal to the mean spin through sum frequency generation (SFG) of the GWI laser and an direction are equivalent to canonical operators. c). For the geometry auxiliary laser at the atomic frequency λS, and the subsequent para- shown in a) the first of the spin oscillator has the energy metric downconversion (PDC). Combined signals from detectors DS which is below the state with no excitations corresponding to the and DI allow for back action free measurement. PBS — negative mass oscillator. d). Mechanical oscillator spectrum. beam splitter; FR — Faraday rotator.

where ΩS is the Larmor frequency. The first term is an irrele- DI is [25, 26]: vant constant energy offset due to the mean spin polarization. √ κ + iΩ 2κ Θχ 2κ Θ F + F The second term is equivalent to the Hamiltonian of a mechan- bˆ s = I aˆs + I aˆc + I χ s√ T , (1) ical oscillator Hˆ with a negative mass and spring constant. I − Ω I 2 I − Ω M κI i (κI − iΩ) κI i ~m Each quantum of excitation in the negative mass spin oscil- where aˆs, aˆc are the phase (sine) and amplitude (cosine) lator physically corresponds to a deexcitation of the inverted I I Ω quadratures of the incident light, Fs is the signal force, for spin population from its highest by ~ S. Prepa- example from GW, and FT is a sum of the thermal force, seis- ration of the collective spin in the energetically lowest Zeeman mic noise and other technical noise sources (notations used state realizes instead a positive mass and spring constant spin throughout this paper are listed in TableI). The first term oscillator. describes the shot noise and the second one — the QBA noise. Interaction of light with a spin in magnetic field placed If the incident light is in a coherent or in a squeezed state, inside a resonator with the finess F can be cast in the form s c similar to that for the mechanical oscillator (see Supplementary then the quadratures aˆI , aˆI are uncorrelated and their spectral densities are equal to e−2r /2 and e2r /2. It is easy to show Information): [18, 25] that in this case the spectral density of the sum of the √ ˆ s ˆs θ χ ˆc θ χ fˆ , shot noise and QBA quantum noise (normalized to to signal bS = aS + 2 SaS + 2 S S (4) F 2 force Fs) cannot be smaller than the force SQL SSQL = ~mΩ . s c where aˆS, aˆS are the phase (sine) and amplitude (cosine) Typically, it is recast as the equivalent position SQL: quadratures of the input light mode in polarization orthog- F onal to the linearly polarized driving optical field, θ = Ω Γ , S ~ S S x SQL Γ = γ d is the spin oscillator read out rate, d = 2F σNa SSQL = = . (2) S S 0 0 π A (mΩ2)2 mΩ2 is the cavity enhanced resonant optical depth of the spin en- γ2 Φ Let us now introduce the second quantum system consisting σ opt semble, γS = A ∆2 is the spin bandwidth dominated by the of a multi-atom spin ensemble. If the spins are optically polar- opt optically induced decoherence, N is the atoms number, σ - the ized along a certain direction x (Figure2) the collective spin a atomic optical crossection, A — the spin ensemble crossec- has a large average projection J = |hJˆ i|/~  1 [8, 27]. Its x x tion, γ — optical transition bandwidth, ∆ - optical field normalized y, z quantum components form canonical variables opt opt √ √ detuning from atomic resonance, Φ — photon flux, XˆS = Jˆz/ ~Jx, PˆS = −Jˆy/ ~Jx, satisfying the commutation relation [XˆS, PˆS] = i. In terms of those variables, the Hamilto- 2 2 −1 χS = −[(γS − iΩ) + ΩS] (5) nian for the ensemble placed in magnetic field oriented along x becomes is the effective susceptibility of the spin oscillator, and fˆS is ~Ω a normalized thermal force acting on the spin [12, 27]. An Hˆ = ~Ω J − S (Xˆ 2 + Pˆ2), (3) S S x 2 S S interesting and useful feature of a spin oscillator is that it is 3

Notation Quantity Value, Adv.LIGO Value, 10-m log 15 r Squeezing factor ↔ 15 db 2 L Interferometer arms length 4000 m 10 m m Mirrors mass 40 kg 0.1 kg κI Interferometer half-bandwidth 2π × 500 Hz 2π × 2000 Hz Ic Optical power circulating in each of the arms 840 kW 1 kW 8ω I Θ = o c normalized optical power (2π × 100)3 s−3 (2π × 575)3 s−3 mcL ΩS Atomic system eigen frequency 2π × 3 Hz 2π × 30 Hz γS Atomic system damping rate 2π × 3 Hz 2π × 30 Hz

TABLE I. The main notations used throughout this paper

possible to provide the effective temperature of the noise fˆS close to zero even if the collective spin is formed by a gas of atoms at room temperature [27]. Under such conditions the spectral density of this noise force corresponds to zero point fluctuations: −1 SS = | Im χS | = 2|Ω|γS . (6) ˆ s The sine quadratires bI,S of the output beams are measured by the homodyne detectors DI,S, respectively. Let us first consider an ideal case within the frequency band Ω  κI when the detectors’ outputs are added to measure Θ ˆ s ˆ s s s 2 χ c c bI + bS = aˆI + aˆS + aˆI + 2θ χSaˆS κI r 2Θ Fs + FT √ + χ √ + 2θ χS fˆS tanh 2r . (7) κI ~m s s If the probe fields are perfectly entangled such that aˆI = −aˆS c c and aˆI = aˆS, the response of the spin system is characterized by the effective negative mass and matches the response of the interferometer Θχ = −θ χS , (8) κI then both the phase shot noise and the QBA are cancelled in the combined measurement record, resulting in the measurement sensitivity not limited by SQL. FIG. 3. Spectral densities of quantum noise. Thin solid line - Entangled probe fields can be prepared by nonlinear opti- SQL; dashed orange curve - SQL-limited GWD noise; solid green cal transformations (sum frequency generation and parametric curve - hybrid GWD/spin system. Top: Advaced LIGO, bottom: down conversion) as shown in Figure1[28, 29]. A small frac- 10-m prototype (see TableI). In all cases, 2.5% of input losses and tion of the GW detector laser and a laser locked to an atomic 2.5% of output losses are assumed for both GWD and atomic spin transition generate a pump beam through the sum frequency channels. In addition, 0.01% of intracavity roundtrip losses for the GWD and 0.3% for the spin system are assumed. Inset: dashed line generation in χ(2) medium. This beam is then used to pump — susceptibility function χ for a free mass; solid line — absolute a parametric downconversion process (PDC) in which two- value of susceptibility of the spin system with ΩS = γS = 2π × 3 Hz. mode squeezed vacuum modes aˆI,S at wavelengths λ1 and λ2 are generated [29, 30] satisfying c c c aˆI,S = ˆzI,S cosh r + ˆzS,I sinh r , (9a) For a finite degree of entanglement, the outputs of the de- s s s tectors DS and DI should be added with the optimal weigth aˆI,S = ˆzI,S cosh r − ˆzS,I sinh r . (9b) function, to obtain the readout with the suppressed quantum c,s c,s Here ˆz1 and ˆz2 correspond to two independent vacuum noise. It is straightforward to show using Eqs. (1,4), that in fields. the lossless case, within the bandwidth of the interferometer, 4

|Ω|  κI , (where the QBA is significant) the weigth factor is power broadening to γS/2π = 3 Hz the linewidth will be dom- equal to tanh 2r. In this case the combined output is equal to inated by the readout and we will achieve the required value of ΓS/γS. s s ˆz cosh r + ˆzs sinh r In the 10-m prototype GWD case, the best sensitivity fre- bˆ s + bˆ s tanh 2r = I I S cosh 2r quency band is shifted to upper frequencies by about one order c c Ω 2Θχ ˆzI cosh r − ˆzs sinh r of magnitude, which relaxes requrements for S, γS propor- + tionally, see TableI. In this case, the value of F ≈ 35 is κI cosh 2r r sufficient. 2Θ Fs + FT √ + χ √ + 2θ χS fˆS tanh 2r , (10) In Figure3, the resulting quantum noise spectral densities κI ~m calculated in SI are shown for realistic optical losses listed in the figure caption. In the inset, we plot the susceptibility which corresponds to the noise supprression factor for both the functions for a free mass and for the spin system matching shot noise and the QBA noise equal to cosh 2r. Importantly, the free mass over the entire frequency band of interest. It the λ and λ do not have to be phase locked to each I S follows from these results that the proposed scheme allows other. Rather only the phases between the local oscillators to achieve the sensitivity gain about 6 db across the entire and the respective laser beams should be stabilized, so that the sensitivity band of interest, which corresponds to almost an correct quadratures are detected. order of magnutude gain in the “visible” part of Universe and c. Numerical estimates. The quantum noise spectral a proportional increase of the event rate. density of the considered scheme is calculated in the SI, tak- ing into account the optical losses in the interferometer and d. Conclusion. We present a way to suppress quantum the spin system for two sets of parameters, one of which ap- noise in gravitational wave interferometers by adding a spin proximately corresponds to the design goals of the Advanced system into the detection path. The proposed method allows LIGO [21] and the other one — to the Hannover 10-m proto- for broadband detection sensitivity beyond the Standard Quan- type interferometer [31, 32], see TableI. We show that the spin tum Limit across the entire frequency bandwidth relevant for system allows for a broadband sensitivity beyond the SQL for gravitational wave observation. In comparison to the earlier both interferometers. proposals for beyond the SQL GWD which use either an exter- The critical parameters of the spin system are deduced from nal filtering cavity [23] or utilize the GW interferometer as an condition (8) using the parameters of the corresponding GWDs effective filtering cavity [24] our approach has an advantage and of the atomic spin system reported in [12]. Consider, for of being completely compatible with existing GWDs and thus example, the Advanced LIGO interferometer. Its projected not requiring complex alterations in the GWD’s core optics. circulating power (TableI) corresponds to the normalized QBA-evading measurement paves the road towards generation power Θ/(2π)3 ≈ (100 Hz)3 with the interferometer band- of an entangled state of the multi-kilogram GWD mirrors and atomic spins which would be of fundamental interest due to width κI /2π ≈ 500 Hz [21]. Tuning the Larmor frequency the sheer size of the objects involved. of spins to ΩS/2π = 3 Hz, we arrive at the requirement for ΓS/2π ≈ 600 Hz. Note that from Eq. (4) we can infer that on resonance the ratio of the noise contribution (last term) to the shot noise is equal to half the ratio of the QBA (second term) to the ground state noise and is given by ACKNOWLEDGMENTS ΓS/γS = d0. In [33] we achieved this ratio and hence the d0 ≈ 2 for a single pass interaction in the atomic cells with We acknowledge motivating discussions with Nergis Maval- the length of 4 cm. Increasing the length of the cell to 10 cm vala and Antonios Kontos. We acknowledge also helpful com- and placing the spin ensemble in an optical resonator with ments from Yiqui Ma. E.S.P acknowledges funding by the the finesse of F ≈ 70 will provide d0 ≈ 200. The 25 mm ERC grant INTERFACE, by ARO grant W911NF and the room temperature Caesium cells with advanced wall coating John Templeton Foundation. F.K. acknowledges funding by [27, 34] have the intrinsic linewidth < 1 Hz . With optical the RFBR grants 16-52-10069 and 16-52-12031.

[1] V.B.Braginsky, F.Ya.Khalili, Quantum Measurement, Cam- [8] K. Hammerer, M. Aspelmeyer, E. S. Polzik, and P. Zoller, Phys. bridge University Press, 1992. Rev. Lett. 102, 020501 (2009). [2] B.Abbott et al, Phys. Rev. Lett. 116, 061102 (2016). [9] K. Zhang, P. Meystre, and W. Zhang, Phys. Rev. A 88, 043632 [3] B.Abbott et al, Phys. Rev. Lett. 116, 241103 (2016). (2013). [4] B.Abbott et al, Phys. Rev. Lett. 118, 221101 (2017). [10] M. J. Woolley and A. A. Clerk, Phys. Rev. A 87, 063846 (2013). [5] B.Abbott et al, Phys. Rev. Lett. 119, 141101 (2017). [11] C. F. Ockeloen-Korppi et al., Phys. Rev. Lett. 117, 140401 [6] B.Abbott et al, Phys. Rev. Lett. 119, 161101 (2017). (2016). [7] B. Julsgaard, A. Kozhekin, and E. S. Polzik, Nature 413, 400 [12] C. B. Møller et al., Nature 547, 191 (2017). (2001). [13] M. Tsang and C. M. Caves, Phys. Rev. Lett. 105, 123601 (2010). [14] M. Tsang and C. M. Caves, Phys. Rev. X 2, 031016 (2012). 5

[15] http://www.ligo.caltech.edu. SUPPLEMENTARY INFORMATION [16] http://www.virgo-gw.eu. [17] http://www.geo600.org. 1. Input/output relation [18] C. M. Caves, Phys. Rev. D 23, 1693 (1981). [19] J.Abadie et al, Nature Physics 7, 962 (2011). [20] H. Grote et al., Phys. Rev. Lett. 110, 181101 (2013). With optical losses taken into account, Eqs. (1, 4) of the [21] J.Aasi et al, Classical and 32, 074001 (2015). main text can be shown [s1, s2] to take the following form: [22] Y.Chen, Physical Review D 67, 122004 (2003).    [23] H.J.Kimble, Yu.Levin, A.B.Matsko, K.S.Thorne and s √ s 2κIcΘχ c s Θχ s bˆ = ηI3 RI cˆ + cˆ + TI nˆ + nˆ S.P.Vyatchanin, Physical Review D 65, 022002 (2001). I I `2 I I2 ` I2 [24] Y. Ma et al., Nature Physics 13, 776 (2017). r  2κIcΘ χFs p [25] S.L.Danilishin, F.Ya.Khalili, Living Reviews in Relativity 15 + + 1 − ηI3 nˆ I3 , (11a) (2012). ~m ` [26] A.Buonanno, Y.Chen, Physical Review D 67, 062002 (2003). s √  s c s c bˆ = ηS3 RScˆ + 2ηS2θ χScˆ + TS(nˆ + θ χSnˆ ) [27] K. Hammerer, A. S. Sørensen, and E. S. Polzik, Rev. Mod. Phys. S S S S2 S2 p  p 82, 1041 (2010). + 2ηS2θ χS fS + 1 − ηS3 nˆS3 . (11b) [28] D. F. Walls and G. J.Milburn, , Springer-Verlag, Berlin, 2008. Here [29] C. Schori, J. L. Sørensen, and E. S. Polzik, Phys. Rev. A 66, √ cˆc,s η c,s p − η c,s , 033802 (2002). I = I1 aˆI + 1 I1 nˆ I1 (12a) [30] N. P. Georgiades, E. S. Polzik, K. Edamatsu, H. J. Kimble, and c,s √ c,s p c,s cˆ = ηS1 aˆ + 1 − ηS1 nˆ (12b) A. S. Parkins, Phys. Rev. Lett. 75, 3426 (1995). S S S1 [31] http://www.aei.mpg.de/18478/04_10m_Prototype. are the effective incident light quadratures for the interferom- [32] Tobias Westphal et al, arXiv::1111.7252 (2011). c,s c,s c,s c,s eter and the spin system cavity, nˆ I1 and nˆ I3 , nˆS1 and nˆ I3 are [33] W. Wasilewski et al., Phys. Rev. Lett. 104, 133601 (2010). the vacuum noise operators associated with the input and the [34] M. V. Balabas et al., Opt. Express 18, 5825 (2010). c,s c,s output losses, nˆS2 , nˆ I2 are the vacuum noise operators due to the internal optical losses in the interferometer/spin system cavity,

` = κI − iΩ, (13a) √ 2κ 2 κIc κIl R = Ic − 1, T = , (13b) I ` I ` √ TS − AS 2 TS AS RS = , TS = . (13c) TS + AS TS + AS Other notations are listed in TableII. It follows from (11a) that the signal force estimate without detection on the spin system is given by s ~ − Ω ˜ m κI i ˆ s ˆ Fs I = bI = Fs + FI , (14) 2κIcΘ χ where s " ~m ` 2κ Θχ  Θχ  Fˆ = R cˆs + Ic cˆc + T nˆ s + nˆ c I I I 2 I I I2 I2 2κIcΘ χ ` ` s # 1 − ηI3 + nˆ I3 (15) ηI3 is the quantum noise in the interferometer channel alone.

2. Spectral densities

Spectral densities of the input noise components [see Eq. (9) of the main text] are defined by the degree of two-mode squeez- ing cosh 2r S[aˆc] = S[aˆs] = S[aˆc ] = S[aˆs ] = . (16a) I I S S 2 6

Notation Quantity Value

ηI1 quantum efficiency of the input path of the interferometer 97.5% ηI3 quantum efficiency of the output path of the interferometer 97.5% AI Light losses per bounce in the interferometer 0.01% cA κ = (1 − η )κ = I part of the interferometer half-bandwidth due to the optical losses Il I2 I 4L κIc = ηI2κI part of the interferometer half-bandwidth due to the coupling ηS1 quantum efficiency of the input path of the spin system system 97.5% ηS3 quantum efficiency of the output path of the atomic system 97.5% AS Light losses per bounce in the spin system cavity 0.3% TS Transmissivity of the coupling mirror of the spin system cavity TS ηS2 = TS + AS

TABLE II. Notation related to the optical losses

The only non-zero cross-correlation spectral densities are The force estimate for the hybrid interferometer-spin system is equal to c c sinh 2r s s sinh 2r S[aˆI aˆS] = , S[aˆI aˆS] = − . (16b) ˜ ˜ ˆ s 2 2 Fs = Fs I + αbS , (21) c,s Therefore, spectral densities of cˆI,S and their (non-zero) cross- where correlation spectral densities are equal to, respectively SI Spin α = − . (22) c s ρI + 1 SSpin S[cˆI ] = S[cˆI ] = , (17a) 2 The resulting effective position noise spectral density is equal c s ρS + 1 to S[cˆS] = S[cˆS] = , (17b) 2 2 ! 2  2  ρ ρ |SI Spin| ~m|`| |σI Spin| c c IS s s IS S = | χ|2 S − = σ − . (23) S[cˆI cˆS] = , S[cˆI cˆS] = − , (17c) I I 2 2 SSpin 4κIcΘ σSpin where 2 ρI = 2ηI1 sinh r , (18a) 3. Balancing of the back actions 2 ρS = 2ηS1 sinh r , (18b) √ Consider the leading in er terms in Eqs. (20). In this ap- ρ η η r . IS = I1 S1 sinh 2 (18c) proximation, The force spectral densities for the interferometer and the spin 2r ηI1e system (15, 11b) and their cross-correlation spectral density ρI ≈ , (24a) 2 are: 2r ηS1e 2 ρ ≈ , (24b) ~m|`| S 2 SI = σI , (19a) Θ| |2 √ 2r 4κIc χ ηI1ηS1 e η ρ ≈ , (24c) S = S3 σ , (19b) IS 2 Spin 2 Spin s 1 ~mηS3 ` η e2r  4η2 κ2Θ2| χ|2  SI Spin = σI Spin . (19c) σ ≈ I1 |R |2 + I2 I , (25a) 2 2κIcΘ χ I 2 I |`|4 2r   Here ηS1e 2 2 2 2 σSpin ≈ RS + 4ηS2θ | χS | , (25b) 4η κ2Θ2| χ|2 2 2 1 I2 I √ σI = |RI | ρI + + (ηI2 ρI + 1), (20a) 2r ∗ 4 ηI1ηS1 e  4ηI2ηS2κI Θθ χ χ  ηI3 |`| σ ≈ −R R + S . (25c) I Spin I S 2 2 1 2 2 2 ` σSpin = RS ρS + + 4ηS2θ | χS | (ηS2 ρS + 1) ηS3 and +4ηS2θ| Im χS | , (20b) 2 RSηI2κI Θχ ∗ + R η θ χ  4κ η Θθ χ χ  2 2r 2 I S2 S Ic S2 S |σI Spin| ηI1e ` σI Spin = −RI RS + ρIS . (20c) σ − ≈ . (26) `2 I σ 2 R2 2 2| |2 Spin S + 4ηS2θ χS 7

In order to compensate the noise, the numerator of this equation where θSP is the single-pass value of the coupling factor θ, has to be equal to zero. The simplified frequency-independent Eq. (27) can be presented as follows: form of this condition is the following: 2 2 TS − 2bTS − AS = 0, (29) where ηS2 ηI2 Θ θ = . (27) 2η − 1 2κ θ 2ηS2 − 1 2ηI2 − 1 κI b = I2 I SP . (30) ηI2 Θ

Therefore, for given θSP and AS, the back actions balancing is Taking into account that provided by q 2 2 TS = b + b + A . (31) 4 2F S θ = θSP = θSP , (28) TS + AS π

[s1] S.L.Danilishin, F.Ya.Khalili, Living Reviews in Relativity 15 [s2] K. Hammerer, A. S. Sørensen, and E. S. Polzik, Rev. Mod. Phys. (2012). 82, 1041 (2010).