2024 Volume 33 Issue 8
Article Contents

Zhiyao Hu, Qixian Li, Xuanchen Zhang, He-Bin Zhang, Long-Gang Huang, Yong-Chun Liu. Nonlinear time-reversal interferometry with arbitrary quadratic collective-spin interaction[J]. Chinese Physics B, 2024, 33(8): 080601. doi: 10.1088/1674-1056/ad4ff7
Citation: Zhiyao Hu, Qixian Li, Xuanchen Zhang, He-Bin Zhang, Long-Gang Huang, Yong-Chun Liu. Nonlinear time-reversal interferometry with arbitrary quadratic collective-spin interaction[J]. Chinese Physics B, 2024, 33(8): 080601. doi: 10.1088/1674-1056/ad4ff7

Nonlinear time-reversal interferometry with arbitrary quadratic collective-spin interaction

  • Received Date: 21/04/2024
    Accepted Date: 20/05/2024
    Available Online: 01/07/2024
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(8)

Article Metrics

Article views(717) PDF downloads(5) Cited by(0)

Access History

Nonlinear time-reversal interferometry with arbitrary quadratic collective-spin interaction

Abstract: Atomic nonlinear interferometry has wide applications in quantum metrology and quantum information science. Here we propose a nonlinear time-reversal interferometry scheme with high robustness and metrological gain based on the spin squeezing generated by arbitrary quadratic collective-spin interaction, which could be described by the Lipkin–Meshkov–Glick (LMG) model. We optimize the squeezing process, encoding process, and anti-squeezing process, finding that the two particular cases of the LMG model, one-axis twisting and two-axis twisting outperform in robustness and precision, respectively. Moreover, we propose a Floquet driving method to realize equivalent time reverse in the atomic system, which leads to high performance in precision, robustness, and operability. Our study sets a benchmark for achieving high precision and high robustness in atomic nonlinear interferometry.

1.   Introduction
  • Improving precision and system robustness is always the main task in quantum metrology and quantum sensing. Projection noise, originating from the quantum fluctuations in the measured populations, is a fundamental limitation for the improvement of precision.[1–5] To break through the limit, many-body entanglement is usually needed,[6–16] while spin squeezing[1,17] is a widely used method. Squeezed spin states (SSSs) possess good properties of reduced spin fluctuations in certain directions, thus having a variety of applications in high-precision measurements and quantum information science.[18–33]

    Despite squeezed spin states having great properties in reducing projection noise for quantum metrology, in real experiments, the precision we could achieve is still greatly restricted by other various types of noises, including the detection noise. Under certain ranges of detection noise, the metrological gain using squeezed spin states will decrease excessively.[34] To address this problem, an echo-like scheme similar to nonlinear interferometry is proposed.[35–44] In this scheme, apart from the squeezing process and encoding process, a quasi-time-reverse evolution called interaction-based readout (IBR) is added before detection, which will significantly improve the robustness of the system to the detection noise.[34,45–53] The added reversal control will improve the robustness of the system to the detection noise. However, the previous studies mostly focus on IBR generated by one-axis twisting (OAT)[17,34] and two-axis twisting (TAT),[17,48] which are very special cases of collective-spin interaction. Besides, introducing the IBR scheme will lead to a tradeoff between precision and robustness, which is very important but seldom investigated. Moreover, most of the previous studies assume that there exists a time-reversal control, which is not always easy to implement in experiments.

    In this article, we investigate the performance of the nonlinear interferometry with arbitrary quadratic collective-spin interaction and propose the optimal scheme for high precision and high robustness in quantum metrology with the atomic system. To improve the precision and robustness of the system, we find the optimal squeezing process, phase encoding process, and anti-squeezing process with different anisotropic parameters for quadratic interaction. Through both analytical and numerical analyses, we also demonstrate that the highest precision and highest robustness could be achieved simultaneously for the TAT interaction. Moreover, we propose a Floquet driving method to achieve the equivalent time-reversal control, which is implementable for arbitrary quadratic collective-spin interaction. We also show that the method is robust to imperfect pulse from the Floquet driving according to numerical simulations.

    The paper is organized as follows. In Section 2, we first explain how the nonlinear interferometry works, introducing our system model and optimizing the interferometer processes with quadratic collective-spin interaction, which can be described by the Lipkin–Meshkov–Glick (LMG) model. In Section 3, we propose that we could generate equivalent time reversal by making use of Floquet driving. In Section 4, we show that the method is robust to different kinds of noises, followed by the summary.

2.   System model and nonlinear interferometer scheme
  • As shown in Fig. 1, the whole process of the nonlinear interferometer scheme could be viewed as three parts. First, a coherent spin state (CSS) is prepared and it evolves under U1 for t1, in which the entanglement is generated and the state turns to a squeezed or over-squeezed state. Following that is the encoding part, where the estimated value ϕ is encoded through Uϕ,θ. After that comes the readout part, during which the squeezed spin state evolves under a time-reversed dynamics U2 for t2, followed by a measurement. Now we consider how to optimize this scheme to achieve high precision and high robustness.

  • Consider a system of mutually interacting spin-1/2 particles described by the following Hamiltonian:

    where σαj is the Pauli operator of the j-th spin and α,β ∈ {x,y,z}. The parameter χαβ characterizes the strength of the interaction in different directions. To ensure the Hermicity of the Hamiltonian, we have χαβ = χβα. Here we have the assumption that the interactions between individual spins are the same, and this assumption holds when there are all-to-all interactions, which is valid under some systems such as nuclear system,[54] cavity QED,[44] and ion trap.[55]

    Now we introduce the collective spin operators Sα=(ℏ/2)∑jσαj with commutation relations [Sα,Sβ]=iℏεαβγSγ, where α, β, γ denote the components in any three orthogonal directions, and εαβγ is the Levi–Civita symbol, and we let ℏ = 1. H preserves the magnitude of the total spin S2=∑αSα2, namely, [H,S2] = 0. By applying a linear transformation to the collective spin operators and redefining the coordinate axes we can prove that this Hamiltonian is equivalent to

    which could be described as the LMG model.[54,56] In the expression of HLMG(χ, γ), χ is the strength of the interaction which describes the rate at which the system evolves, and γ is the anisotropic parameter reflecting the symmetry of the system. OAT and TAT could be viewed as special cases when γ = 0 and γ = 0.5.[57]

  • The first step of the nonlinear interferometer scheme is to generate entanglement. Consider a system with N interacting spin-1/2 particles, we choose a CSS as the initial state, which can be described by

    where ϑ is the angle between the z-axis and the collective-spin vector (polar angle), while φ is the angle between the x-axis and the vertical plane containing the collective-spin vector (azimuth angle). In the following, we set ϑ = π/2 and φ = π/2, namely |y^〉, as the initial state.

    Firstly, the system undergoes a unitary evolution U1, namely dynamics with the LMG Hamiltonian HLMG(χ, γ) for t1 in our strategy, evolving to a squeezed or over-squeezed state. For the system with different anisotropic parameters γ, the corresponding t1 to get the state with the highest precision varies as well. We need to optimize the evolution time t1 in order to get the highest precision for estimating the value of the unknown parameter, which is bounded by quantum Cramer–Rao bound (QCRB)[58]

    where k is the number of independent repetitions (in the following we set k = 1), and F[ρλ] is the quantum Fisher information (QFI)[59–62] defined as

    where ρλ^=∑κqκ|κ〉〈κ| is the λ-dependent density matrix and both the eigenvalues qκ ≥ 0 and the associated eigenvectors |κ〉 depend on λ, and λ is the known parameter to be estimated. For the system and parameterization we are interested in, as a pure state, the QFI simply becomes

    where Sn = sinϑ cosφSx + sinϑ sinφSy + cosφSz. To set a higher bound for the precision of parameter estimation, t1 should be set as the time realizing the highest QFI. For a certain γ, change the evolution time t1′ and find the corresponding maximum QFI: F[t1′,γ]. The maximum of F[t1′,γ] then is the largest QFI with this γ, denoted as F[γ]max, and the corresponding t1′ is the t1. Now we change the γ, and track how F[γ]max and t1 change when γ varies from 0 to 0.5.

    The results are shown in Fig. 2. We find that F[γ]max and t1 monotonically depend on γ for 0 ≤ γ ≤ 0.5. When γ = 0.5, the system attains its maximum F[γ]max, and the corresponding t1 is also the shortest, corresponding to the TAT interaction. What is worth mentioning is that the range of t1 we choose is around the best squeezing time, while the OAT interaction HOAT=χSx2 could generate GHZ states at t = π/(2χ), the QFI of which could reach the Heisenberg limit.

  • The entanglement of the system is generated through U1, followed by the application of the perturbation spin rotation Uϕ,θ = e−iϕSθ, which also plays an important role in the nonlinear interferometry. Contrary to the traditional belief that the encoding part only has an effect on the precision so that the axis chosen to encode the parameter could simply choose the direction with minimum fluctuation, the encoding part has a vital influence on the robustness of the system to the detection noise if an anti-squeezing operator U2 is applied before performing the measurement.

    In our model, the squeezing happens in the x–z plane, so Sθ should be chosen in this plane to fully exploit the metrology-enhanced property. Sθ could be expressed as Sθ = Sx sin θ + Sz cos θ, where θ is the angle between the z-axis and Sθ. The sensitivity for estimating ϕ and the robustness to detection noise of the system vary as θ takes different values. In the actual experiments, different measurements have different emphases. In some conditions, the precision might be the priority, while under certain circumstances the robustness against the measurement noise deserves extra attention. To quantify the robustness of the system to the detection noise, here we introduce the magnification factor G as

    The numerator (denominator) represents the magnitude of the mean-spin measurement with (without) the time-reversal control, Sm is the spin component that we apply collective mean-spin measurement, m is a continuous parameter corresponding to an angle in the x–z plane, and G represents the phase magnification benefited from the time-reversed dynamics (here we assume that a perfect time-reversed dynamics is followed after the encoding process). A higher G indicates the influence of detection noise will have less influence on the signal, indicating the robustness to detection noise.[34,48]

    We denote the corresponding θ as θr and θp with the directions that achieve the highest precision for estimation and robustness to detection noise, respectively. As γ changes, θr and θp would change as well. The analytical solution of the dependence of θr and θp with respect to γ is too complex to conduct, but we can give a brief and reasonably accurate estimate of θr. We use the mean-field approximation to describe the time evolution after the parameter is encoded. The detail is shown in Appendix A. The calculation results in a brief expression of θr:

    which has been proved to have enough accuracy when γ is not too small. The comparison of the numerical solution and mean-field results is shown in Fig. 3(a). We can see that our approximation is accurate enough, especially when γ approaches 0.5. When γ = 0.5, θr becomes π/4, which is consistent with previous research.[48]

    To get the estimated value of θp, we perform the numerical analysis by calculating the fluctuation for each θ after U1 and finding the minimum one, taking the corresponding θ as θp for this γ. The result is shown in Fig. 3(b). Compared to θr, we show that for a certain γ, the corresponding θr and θp have an approximate relationship

    We find that γ = 0.5 is a special situation when θr and θp take the same value π/4, which indicates that for the TAT interaction, we can achieve the best robustness and precision simultaneously by setting Sθ=(2/2)Sx+(2/2)Sz.

  • The time-reversed nonlinear dynamics is the key to our strategy. Applying HLMG(−χ, γ) for t2, we could perform the anti-squeezing process to get the final state to measure. An appropriate t2 is important in this step. To determine the optimal t2 to achieve higher precision, we define the metrological gain as ΔG to investigate the effect of U2 for t2:[51]

    where (Δϕ)SQL=1/N is the standard quantum limit (SQL)[63] and Δϕ is given by the error propagation formula

    where ΔSmϕ is the standard deviation of Sm and ∂ϕ = d/dϕ. ∂ϕ〈Smϕ〉 can be replaced by its first order approximation with respect to ϕ:

    where we define S(U) = U†SU for brevity.

    Previously we proposed two directions of Sθ, and the dependence of metrological gain on t2 and γ with different directions of Sθ is shown in Fig. 5, respectively. We show that with γ ranging from 0 to 0.5, to reach the maximum metrological gain, the time t2 could always be chosen near t2 = t1, which is conducive for us to determine the reverse time t2 for convenience. Besides, we find that TAT outperforms OAT in precision with the reversed dynamics when the unknown parameter is encoded through the axis of Sθr and Sθp, which is consistent with the previous study.[48]

3.   Generation of equivalent time reverse
  • Realizing the interaction-based readout scheme requires implementable time-reversal control, which is usually not a trivial task in experiments. Inspired by the previous study, we propose that we could realize equivalent time reversal by making use of multiple π/2 pulses, which can be realized using the coupling term ΩαSα (α = x, y,z).[56,64,65] By making use of a multi-pulse sequence along the α-axis (α = x,y,z), we can rotate the spin along the α-axis and affect the dynamic of squeezing. A π/2 pulse corresponds to ∫−∞+∞Ωα(t)dt=π/2, which leads to the result that Rα,−π/2eitχSβ2Rα,π/2=eitχSκ2, where Rα,θ = e−iθSα and κ is the axis that perpendicular to the α-axis and β-axis. The multi-pulse sequence is periodic, and the frequency is determined by γ and the axis we choose.

    To generate equivalent rime reversal U2 = eiHLMGt, one way is to generate H = −HLMG + kS2 since [H, S2] = 0, the total spin will not have an effect on the squeezing properties while k is any constant. The whole time reversal process could be viewed as the repetitions of Fig. 6(a), and each period is made up of the following: a −π/2 pulse along the α-axis, a free evolution for δt1, a π/2 along the α-axis, a −π/2 along the β-axis, a free evolution for δt2, and a π/2 along the β-axis. The period is tc = δt1+δt2, neglecting the time needed for applying the four π/2 pulses. Figure 6(b) shows that the cyclic ±π/2 could be viewed as rotations on the Bloch sphere. By adjusting the relationship between δt1 and δt2, we can transform the LMG interaction into equivalent time-reversed dynamics.

    One potential strategy to generate time reversal is to apply multiple pulses along the y-axis and z-axis. For HLMG=χ(Sx2+γSy2), with the z-axis to be the α-axis and y-axis to be the β-axis, the time evolution operation for a single period could be expressed as

    Using the Baker–Campbell–Hausdorff formula, we find Uyz≈e−iχ[Sx2(γδt2)+Sy2(γδt1+δt2)+Sz2(δt1)] for small δt1 and δt2.[66] In order to achieve equivalent time reverse, the relationship between γ and δt2/δt1 should satisfy

    Accordingly, we obtain the effective Hamiltonian

    With the anisotropic parameter γ ranging from 0 to 0.5, χeff=−(γ2−γ+1)−γ2−2γ+2<0, which means that we can generate equivalent time reverse by applying pulse sequences along the y-axis and z-axis.

4.   Noise analysis
  • Now we analyze the performance of the time-reversal interferometry scheme under various kinds of noises, and here we focus on analyzing the impact of detection noise as well as the imperfect pulse. The former could be weakened by applying the time-reversal control, while the latter is introduced by the multiple pulse sequences involved in our scheme.

  • The effect of detection noise can be described by the measuring operator M^,

    where |m〉 is the m-th eigen state of Sz and

    determined by parameter σ. This setup means if the system is in the |m〉 state, execute the measurement and there exists a probability of Γm,m′ to mistake |m〉 for |m′〉. Γm,m′ has a form of Gaussian convolution, which makes sense experimentally and mathematically. Now we define noise strength N=e−1/(2σ)2, M^ then could be expressed as

    The M^ could be regarded as an identity operator I^ attached by higher order term of N, which would cause a correction term in the angular sensitivity Δϕ. Now we define the robustness coefficient R as

    which reflects the response of measurement results to the detection noise. A larger R means less response, namely, better robustness. When N is not too large (for example, N=0.1, which means a probability of 0.2 to make an error just in the detection procedure), we can get the value of R semi-analytically. For simplicity, we take t2 = t1. The detail is shown in Appendix B.

    The conditions when Sθ is towards θr and θp are calculated respectively. The result is shown in Fig. 7(a). We can find that in most of the cases, the robustness is better when Sθ is towards θr. More precisely, when Sθ is towards θp, the additional fluctuation caused by detection noise is about 10 times as large as that of θr, especially when γ is small. When γ gets larger, this gap is narrowing. When γ = 0.5, the two curves coincide, for θr = θp then.

    In order to measure the noise suppression effect of our scheme, we also calculate the relative robustness coefficient Rrel = R–R0, where R0 is the robustness coefficient when U2 is identity (no time-reversed dynamic is applied). The result is shown in Fig. 7(b). It can be concluded that for both θr and θp, the IBR method can reduce the effect of noise to 10−2–10−3, which reflects the superiority of our scheme in noise reduction. Conversely, the Rrel of θp performs better than that of θr, and when γ = 0.5 it is the same as the two conditions.

  • Apart from the detection noise, noise from imperfect pulses should also be taken into consideration. Firstly, our scheme assumes that the duration of each period is small enough to ignore the high-order term, which means the frequency for the pulse period is very high. Besides, the pulse area, pulse separation, and pulse phase are not always perfect. To verify the robustness of our scheme, we perform numerical simulations by adding Gaussian stochastic noises, i.e., assuming the fluctuating pulse areas, pulse separations and pulse phase are subject to Gaussian distribution with a standard deviation of different ranges of the average value.

    To quantify the influence of imperfect pulse, as shown in Fig. 8, by adding the noise into the pulse parameters and repeating the simulations 100 times, we show that under the circumstances of the noise for pulse frequency higher than 500χ, noise in the pulse area less than 0.5%, noise in pulse separation less than 5%, and noise in pulse phase less than 0.1%, our method can almost achieve the optimal performance under perfect time reverse, indicating the robustness of the pulse scheme, and set the bound for experimentalists to realize the dynamics. It is worth mentioning that dynamical decoupling sequences could also help cancel the noise.[67–69]

    As for the spin decoherence, our method will not introduce new resources of decoherence but expend the experimental time, and we consider the situation that the coherence time is long enough to ignore the impact of spin decoherence,[50] so we ignore the impact of extending the evolution time for spin decoherence for simplification.

5.   Conclusion
  • In summary, we investigate the performance of nonlinear time-reversal interferometry with quadratic collective-spin interaction. We design nonlinear interferometry based on that and optimize each part of this scheme with high precision and high robustness with different anisotropic parameters. We find no tradeoffs in reaching the highest precision and highest robustness with the TAT interaction, which is an exceptional case in quadratic collective-spin interaction, and we prove it through numerical and analytic analyses. Besides, we find that OAT outperforms in robustness while TAT outperforms in precision. To achieve the time reversal in this interferometry, we propose a Floquet-driving method to generate the equivalent time reversal. With HLMG=Sx2+γSy2, we find that by making use of multi-pulse sequences along y-axis and z-axis, we could achieve effective time reverse. We also show that our scheme is robust to different kinds of noises, including detection noise and imperfect pulses. Our work will significantly deepen the insight of making use of atomic systems to achieve high precision and high robustness in nonlinear interferometry and push the frontier of quantum metrology.

Appendix A: Mean-field approximation in perturbation encoding
  • For the LMG model, we have

    which means S2=Sx2+Sy2+Sz2 is constant during the evolution. Thus HLMG(χ, γ) is equivalent to

    When the unknown parameter is encoded through Sθ, suppose Sα satisfies Sα = 〈Sα〉 + Sα1, where α = x,z and Sα1 is the first order small quantity compared to Sα. Applying this approximation to Eq. (A2), we can get the mean-field approximation of HLMG(χ, γ) as

    For ϕ far smaller than S, we can use S to represent 〈Sy〉. Consider (〈Sx〉, 〈Sz〉) as a point (x,z) on the phase plane. Applying Heisenberg equation to Sx and Sz, we have

    here we use the commutation relation of the collective spin operators

    where α ,β ,γ denote the components in any three orthogonal directions, and εαβγ is the Levi–Civita symbol. Combining Eqs. (A4) and (A5), we get the evolution equation of x and z

    Solve the equations, we have

    where A, B, C, and D are undetermined coefficients. When Sϕ,θ has just added, we have

    apply which to Eq. (A8) we can determine the expression for x:

    Similarly, we can get the expression for z:

    To ensure 〈Sϕ〉 reaches its maximum, we can equivalently ascertain a θ that makes the point (x, z) on the phase plane as far away from the origin as possible during the evolution. The distance from (x, z) to the origin is

    where we omit the terms with e−2χSγ(1−γ)t. Equation (A13) has maximum when

    reaches its maximum, which leads to our result

Appendix B: Semi-analytical solution of robustness coefficient R
  • When N is not too large, the main source of error is mistaking |m〉 for |m±1〉. The other term of M^ would vanish because of the higher order of N. Thus M^ can be approximated by

    where M^0 is simply the identity operator and M^1=∑m∈ℕ,m±1≥0N|m±1〉〈m|. In matrix form, M^0 can be denoted as an identity matrix I, and M^1 can be denoted as a matrix M1 with non-zero element M1(k,k+1) = M1(k,k–1) = 1. Taking M into consideration, the final state turns into MU2e−iSθϕU1|y^〉. Applying Eq. (B1) into Eq. (12), we can get the expression of 〈Smϕ〉 under the perturbation of M^1:

    where we define {M1, Sm} = M1Sm + SmM1. Expanding to the first order of N, then we can get an approximation of ∂ϕ〈Smϕ〉:

    Applying Eq. (B3) to Eq. (11) and expanding to the first order of N, we can get the corrected Δϕ:

    and go further to figure out R.

Figure (8)  Reference (69)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return