This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Near Heisenberg limited parameter estimation precision by a dipolar Bose gas reservoir engineering

Qing-Shou Tan College of Physics and Electronic Engineering, Hainan Normal University, Haikou 571158, China    Ji-Bing Yuan Department of Physics and Electronic Information Science, Hengyang Normal University, Hengyang 421002, China    Guang-Ri Jin grjin@bjtu.edu.cn Department of Physics, Beijing Jiaotong University, Beijing 100044, China    Le-Man Kuang lmkuang@hunnu.edu.cn Key Laboratory of Low-Dimensional Quantum Structures and Quantum Control of Ministry of Education, and Department of Physics, Hunan Normal University, Changsha 410081, China
(August 10, 2025)
Abstract

We propose a scheme to obtain the Heisenberg limited parameter estimation precision by immersing atoms in a thermally equilibrated quasi-one-dimensional dipolar Bose-Einstein condensate reservoir. We show that the collisions between the dipolar atoms and the immersed atoms can result in a controllable nonlinear interaction through tuning the relative strength and the sign of the dipolar and contact interaction. We find that the repulsive dipolar interaction reservoir is preferential for the spin squeezing and the appearance of an entangled non-Gaussian state. As an useful resource for quantum metrology, we also show that the non-Gaussian state results in the phase estimation precision in the Heisenberg scaling, outperforming that of the spin-squeezed state.

pacs:

I introduction

One of main goals of quantum metrology is to achieve parameter (or phase) estimation precision beyond the shot-noise limit Caves1981 ; Yurke1985 ; Holland1993 ; Giovannetti2011 ; Dorner2009 ; Sanders1995 ; Ma2011 ; Ma2011B ; Humphreys2013 ; Dowling2008 ; Lvovsky2009 . Atomic spin squeezed states (SSSs) play an important role in quantum phase estimation and have been widely studied  Kitagawa1993 ; Wineland1994 ; Momer ; Sorensen2001 ; Poulsen2001 ; Jin2009 ; Jin2010 ; Guehne2009 ; Wang2003 ; Liu2011 in the past few decades ever since the pioneer work of Kitagawa and Ueda Kitagawa1993 , who showed that the SSSs can be dynamically generated from the so-called one-axis twisting interaction (OAT) among spin-1/21/2 particles bou ; kas . As useful quantum resource, the SSSs have been proposed to achieve such a sub-shot-noise limited phase sensitivity Wineland1994 ; Momer ; Sorensen2001 ; Poulsen2001 ; Jin2009 ; Jin2010 ; Guehne2009 ; Wang2003 ; Liu2011 . Recently, Stroble et al. strobel experimentally demonstrated that the OAT can also generate entangled non-Gaussian states (ENGSs), which can outperform the spin-squeezed state. The OAT interaction have been proposed and demonstrated using ion traps Momer , Rydeberg atoms macri , nitrogen-vacancy centers Bennett and atomic Bose-Einstein condensates (BECs).

BECs due to their unique coherence properties and the controllable nonlinearity bou ; kas , have attracted much attention for quantum metrology. Experimental realizations of the OAT model has been proposed and demonstrated through Feshbach resonances gros or spatially separating the components of BECs rie . Besides, the atomic BECs also often as the reservoirs suitable for engineering is considered widely Palzer ; Will ; Cirone ; Spethmann ; Scelle ; Zipkes ; Schmid ; Balewski ; Recati ; Ng ; bargill ; Bru . For instance, one can drastically enhance the OAT interaction by placing a two-state condensate in a completely different special BEC reservoir bargill .

So far, the studies of bosonic atoms for metrology are mainly focused on ss-wave contact interaction. However, for ultracold atoms there also exists long-rang magnetic dipole-dipole interaction (MDDI) yi2000 ; Goral2002 ; yi ; yi2007 ; lu2 . In experiments, dipolar BECs have been realized for atoms with large magnetic dipole moments Lu ; Aikawa . Furthermore, both the sign and the strength of the effective dipolar interaction can be tuned via a fast rotating orienting field Giovanazzi ; Griesmaier2006 ; Lahaye . Very recently, Yuan et al. have used the quasi-2D dipolar BEC as reservoir engineering to study the non-Markovian dynamics of an impurity atom yuan . Therefore, the effects of the MDDI should be considered in the realizations of the OAT model based on dipolar BEC.

In this paper, we realize the OAT model induced by the reservoir dephasing noise, which has been widely viewed as one of the main obstacles for quantum metrology. We consider the dynamics of two-mode BEC consisting of NN atoms coupled to a one-dimensional (1D) dipolar Bose gas reservoir. It show that, the collisions interaction between the dipolar BEC reservoir and the immersed atoms can be described by a spin-boson model. Through calculating the SS ξR\xi_{R}, and the quantum Fisher information (QFI) FQF_{Q}, we find that the dephasing noise can produce SSSs and ENGSs. And the degree of the SS and entanglement both depends on the relative strength and sign of the dipolar and contact interaction. In other words, the repulsive dipolar interaction reservoir can induce better SS and ENGSs. Compared with spin squeezed states, ENGSs can last for a very long time under the dephasing noise. It can monotonically increase in the regimes without SS (ξR>1\xi_{R}>1), next successively undergoes metastable entangled states and entanglement suddenly increase, corresponding to FQN2/2F_{Q}\simeq N^{2}/2 and FQN2F_{Q}\sim N^{2}, respectively. According to Cramér-Rao theorem, Δθ1/FQ\Delta\theta\geq 1/\sqrt{F_{Q}}, we know that FQ>NF_{Q}>N means that the states are entangled and useful for sub-shot-noise-limited phase-estimation precision; and FQ=N2F_{Q}=N^{2} is the maximal entangled states, corresponding to the Heisenberg limit. This confirms that the phase estimation sensitivity can approach to Heisenberg limit, when using ENGSs for metrology.

The paper is organized as follows. In Sec. II, we give the model of immersed atoms interacting with the thermally equilibrated quasi-1D dipolar BEC reservoir. In Sec. III, we study the dynamics evolution of the atoms due to the dephasing noise. The SS and entanglement dynamical behaviors are discussed in Sec. IV and Sec. V. Finally, we draw our conclusion in Sec. VI.

Refer to caption
Figure 1: Schematic diagrams of (a) NN two-level atoms (red) immersed in a quasi-1D dipolar gas (green) and (b) the tuning of the dipole-dipole interaction via a fast rotating orienting field. (d)-(e) show the Wigner function of the initial coherent spin state, the spin squeezed state, and entangled non-Gaussian state (similar to the spin cat state) for the two-level atom system. The negative values of Wigner function correspond to the quantum states.

II formulation

We consider a system of NN two states Rb87{}^{87}\mathrm{Rb} atoms with up ||F=2,mF=1|\uparrow\rangle\equiv|F=2,m_{F}=-1\rangle and down states ||F=1,mF=1|\downarrow\rangle\equiv|F=1,m_{F}=1\rangle immersed in a quasi-1D dipolar gas reservoir. And the system atoms are confined in a harmonic trap that is independent of the internal states [see Fig.1(a)]. In general, the interaction between the system atoms and the reservoir is described by the Hamiltonian

H=HA+HB+HAB,{H}={H}_{A}+{H}_{B}+{H}_{AB}, (1)

where HAH_{A} is the two-state atomic Hamiltonian, HBH_{B} is the dipolar gas reservoir Hamiltonian, and HABH_{AB} describes their interaction.

II.1 Two level atom system

The spin Hamiltonian provides the most intuitive description in the internal case of two-mode trapped atomic BEC

HA=λJz+χJz2.{H}_{A}=\lambda{J}_{z}+\chi{J}_{z}^{2}. (2)

Here, we define the pseudo-spin operator 𝐉(Jx,Jy,Jz){\mathbf{J}}\equiv({J}_{x},{J}_{y},{J}_{z}) based on space orbitals as 𝐉=(a,a)σ(a,a)T/2{\mathbf{J}}=(a^{\dagger}_{\uparrow},a^{\dagger}_{\downarrow}){\mathbf{\sigma}}(a_{\uparrow},a_{\downarrow})^{T}/2, where σ\mathbf{\sigma} is the Pauli matrices, and a^(a^)\hat{a}_{\uparrow}(\hat{a}_{\downarrow}) denotes the annihilation operators of the atom states. The energy difference of the two states λ\lambda and nonlinearity χ\chi depend on the mean filed wave-function of the two modes. We assume that the two modes have the same spatial orbital

ΦA=(πA2)3/4e(x2+y2+z2)/(2A2),\Phi_{A}=(\pi\ell_{A}^{2})^{-3/4}e^{-(x^{2}+y^{2}+z^{2})/(2\ell_{A}^{2})}, (3)

where A=/(mAωA)\ell_{A}=\sqrt{\hbar/(m_{A}\omega_{A})} with ωA\omega_{A} being the trap frequency, and mAm_{A} being the mass of atom. Therefore, χ=(g11+g222g12)/[2(2π)3/2A3]\chi=(g_{11}+g_{22}-2g_{12})/[2(2\pi)^{3/2}\ell_{A}^{3}] with coupling constants gij=4π2aij/mAg_{ij}=4\pi\hbar^{2}a_{ij}/m_{A} and aija_{ij} being ss-wave scattering length. For Rb87{}^{87}\mathrm{Rb} and the chosen hyperfine states, aija_{ij} are almost equal: a11=100.44a0,a22=95.47a0a_{11}=100.44a_{0},a_{22}=95.47a_{0} and a12=97.7a0a_{12}=97.7a_{0}, with a0a_{0} being the Bohr radius. Then, the nonlinearity χ\chi is close to zero. We point out that χ\chi is tunable via Feshbach resonance, but the price of these methods is significantly increased atom losses bargill . Below, we choose χ=0,\chi=0, and apply a dipolar gas reservoir to induce a stronger nonlinearity interaction.

II.2 Bogoliubov modes of quasi-1D dipolar gas reservoir

In second-quantized form, the many-body Hamiltonian of the 1D dipolar BEC is

HB\displaystyle{H}_{B} =\displaystyle= 𝑑xΨ^B(x)h^Ψ^B(x)\displaystyle\int dx\hat{\Psi}_{B}^{\dagger}(x)\hat{h}\hat{\Psi}_{B}(x) (4)
+12𝑑x𝑑xΨ^B(x)Ψ^B(x)V(xx)Ψ^B(x)Ψ^B(x),\displaystyle+\frac{1}{2}\int dxdx^{\prime}\hat{\Psi}_{B}^{\dagger}(x)\hat{\Psi}_{B}^{\dagger}(x^{\prime})V(x-x^{\prime})\hat{\Psi}_{B}(x^{\prime})\hat{\Psi}_{B}(x),

where Ψ^B(x)\hat{\Psi}_{B}(x) is the field operator and h^=222mBx2\hat{h}=-\frac{\hbar^{2}\partial^{2}}{2m_{B}\partial x^{2}} is the single-particle Hamiltonian with mBm_{B} being the mass of the atom. Here, we have assumed that the dipolar BEC to be confined in a cylindrically symmetric trap with a transverse trapping frequency ω\omega_{\perp} and negligible longitudinal confinement ωx\omega_{x} along the xx direction, i.e., ω/ωx1\omega_{\perp}/\omega_{x}\gg 1. In three dimensions, the two-body interaction is

V3D(𝐫)=gBδ(𝐫)+3cd4π13(μ^𝐦𝐫^)2r3,V^{3D}(\mathbf{r})=g_{B}\delta(\mathbf{r})+\frac{3c_{d}}{4\pi}\frac{1-3(\mathbf{\hat{\mu}_{m}\cdot\hat{r}})^{2}}{r^{3}}, (5)

where the contact interaction strength is gB=4π2aB/mBg_{B}=4\pi\hbar^{2}a_{B}/m_{B} with aBa_{B} being the ss-wave scattering length; the dipolar interaction strength is cd=4π2add/mBc_{d}=4\pi\hbar^{2}a_{dd}/m_{B}, where add=μ0μm2mB/(12π2)a_{dd}=\mu_{0}\mu_{m}^{2}m_{B}/(12\pi\hbar^{2}) is a length scale characterizing the MDDI with μ0\mu_{0} the vacuum permeability, μm\mu_{m} the magnetic dipole moment; here 𝐫^=𝐫/r\hat{\mathbf{r}}=\mathbf{r}/r is a unit vector.

To obtain the effective 1D interaction potential, V(xx)V(x-x^{\prime}), in Hamiltonian (4). We assume that the transverse wave function of all the reservoir atoms is

Ψ(y,z)=(πB2)1/2e(y2+z2)/(2B2),\Psi_{\perp}(y,z)=(\pi\ell_{B}^{2})^{-1/2}e^{-(y^{2}+z^{2})/(2\ell_{B}^{2})}, (6)

with B/(mBω)\ell_{B}\equiv\sqrt{\hbar/(m_{B}\omega_{\perp})} being the width of the Gaussian function. By integrating out the yy and zz variables, we can obtain the Fourier transform of the 1D interaction potential (as shown in Appendix A)

V~1D(k)=gB2πB2[1ϵddν~(k)]\displaystyle\tilde{V}_{1D}(k)=\frac{g_{B}}{2\pi\ell_{B}^{2}}\left[1-\epsilon_{dd}\tilde{\nu}(k)\right] (7)

with ϵddcd/gB=add/aB\epsilon_{dd}\equiv c_{d}/g_{B}=a_{dd}/a_{B}, where

ν~(k)=132k2B2exp[k2B22]Γ(0,k2B22),\displaystyle\tilde{\nu}(k)=1-\frac{3}{2}k^{2}\ell_{B}^{2}\exp\left[\frac{k^{2}\ell_{B}^{2}}{2}\right]\Gamma\left(0,\frac{k^{2}\ell_{B}^{2}}{2}\right), (8)

with Γ(0,x)\Gamma(0,x) being the incomplete Gamma function.

To proceed, in the degenerate regime, the bosonic field can be decomposed as

Ψ^B(𝐫)=Ψ(y,z)[n0+1Lk(ukb^keikxvkb^keikx)],\hat{\Psi}_{B}(\mathbf{r})=\Psi_{\perp}(y,z)\left[\sqrt{n_{0}}+\frac{1}{\sqrt{L}}\sum_{k}\left(u_{k}\hat{b}_{k}e^{ikx}-v_{k}\hat{b}^{\dagger}_{k}e^{-ikx}\right)\right], (9)

with n0n_{0} being the condensate linear density, LL the length of the reservoir, b^k\hat{b}_{k} (b^k\hat{b}^{\dagger}_{k}) the annihilation (creation) operators of the Bogoliubov modes with momentum kk. And its Bogoliubov modes are

uk\displaystyle u_{k} =\displaystyle= 1/2(εk/Ek+Ek/εk),\displaystyle 1/2\left(\sqrt{\varepsilon_{k}/E_{k}}+\sqrt{E_{k}/\varepsilon_{k}}\right),
vk\displaystyle v_{k} =\displaystyle= 1/2(εk/EkEk/εk),\displaystyle 1/2\left(\sqrt{\varepsilon_{k}/E_{k}}-\sqrt{E_{k}/\varepsilon_{k}}\right), (10)

with Ek=2k2/(2mB)E_{k}=\hbar^{2}k^{2}/(2m_{B}) being the free-particle energy. Where the excitation energy is Cirone

εk\displaystyle\varepsilon_{k} =\displaystyle= Ek2+2n0Ekν~(k)\displaystyle\sqrt{E_{k}^{2}+2n_{0}E_{k}\tilde{\nu}(k)} (11)
=\displaystyle= 12ω(kB)4+η(kB)2[1ϵddν~D(k)],\displaystyle\frac{1}{2}{\hbar\omega_{\perp}}\sqrt{(k\ell_{B})^{4}+\eta(k\ell_{B})^{2}\left[1-\epsilon_{dd}\tilde{\nu}_{D}(k)\right]},

with dimensionless parameters η=8n0aB\eta=8n_{0}a_{B}. Hence, the Hamiltonian for the collective excitations is

HB=k0εkbkbk.{H}_{B}^{\prime}=\sum_{k\neq 0}\varepsilon_{k}b_{k}^{{\dagger}}b_{k}. (12)

The sum over Bogoliubov modes exclude the zero mode and will act as the reservoir under our model.

II.3 Interaction Hamiltonian

We assume that the reservoir atoms are coupled with the up state ||\uparrow\rangle of the system atoms via a Raman transition Cirone ; yuan

HAB\displaystyle{H}_{AB} =\displaystyle= gABa^a^𝑑𝐫|ΦA(𝐫)|2Ψ^B(𝐫)Ψ^B(𝐫),\displaystyle g_{AB}\hat{a}_{\uparrow}^{{\dagger}}\hat{a}_{\uparrow}\int d\mathbf{r}|{\Phi}_{A}(\mathbf{r})|^{2}\hat{\Psi}_{B}^{{\dagger}}(\mathbf{r})\hat{\Psi}_{B}(\mathbf{r}), (13)

where gAB=2π2aAB/mABg_{AB}=2\pi\hbar^{2}a_{AB}/m_{AB} with the atoms and reservoir scattering length aABa_{AB} and reduced mass mAB=mAmB/(mA+mB).m_{AB}=m_{A}m_{B}/(m_{A}+m_{B}). By substituting Eqs. (3) and (9) into the above interaction Hamiltonian and omitting the square terms about b^k\hat{b}_{k} and b^k\hat{b}^{\dagger}_{k}, we have

HAB\displaystyle{H}_{AB} \displaystyle\simeq δa^a^+a^a^kgk(b^k+b^k),\displaystyle\delta_{\uparrow}\hat{a}_{\uparrow}^{{\dagger}}\hat{a}_{\uparrow}+\hat{a}_{\uparrow}^{{\dagger}}\hat{a}_{\uparrow}\sum_{k}g_{k}\left(\hat{b}_{k}+\hat{b}_{k}^{{\dagger}}\right), (14)

where

δ\displaystyle\delta_{\uparrow} =\displaystyle= gABn0𝑑y𝑑z|ΨB(y,z)|2|ΦA(y,z)|2𝑑x|ΦA(x)|2\displaystyle{g_{AB}n_{0}}\int dydz\left|\Psi_{B}(y,z)\right|^{2}\left|\Phi_{A}(y,z)\right|^{2}\int dx\left|\Phi_{A}(x)\right|^{2} (15)
=\displaystyle= 22aABn0mAB(A2+B2),\displaystyle\frac{2\hbar^{2}a_{AB}n_{0}}{m_{AB}(\ell_{A}^{2}+\ell_{B}^{2})},

and

gk=22aABmAB(A2+B2)n0EkLϵkexp(k2A24).\displaystyle g_{k}=\frac{2\hbar^{2}a_{AB}}{m_{AB}(\ell_{A}^{2}+\ell_{B}^{2})}\sqrt{\frac{n_{0}E_{k}}{L\epsilon_{k}}}{\exp\left(-\frac{k^{2}\ell_{A}^{2}}{4}\right)}. (16)

III System dynamical evolution

In the interaction picture with respect to HB{H}_{B}^{\prime}, the total Hamiltonian is

HI(t)=(λ+δ)Jz+Nkgk(bkeiωkt+bkeiωkt)iΓlossN,H_{I}(t)=(\lambda+\delta_{\uparrow})J_{z}+N_{\uparrow}\sum_{k}g_{k}(b_{k}^{{\dagger}}e^{i\omega_{k}t}+b_{k}e^{-i\omega_{k}t})-i\Gamma_{\mathrm{loss}}N_{\uparrow}, (17)

which is a non-Hermitian dephasing spin-boson model with N=(Jz+N/2)N_{\uparrow}=\left(J_{z}+N/2\right) being the up state number operator. Here the non-Hermitian term Γloss\Gamma_{\mathrm{loss}} is phenomenally introduced to describe the one-boby particle loss rate, owing to inelastic collisions between the system atoms and the noncondensed thermal atoms. It results in the particles be kicked out from the system. Such a kind of loss is a typical dissipation effect and has been widely studied strobel ; pawlowski ; hao ; spehner ; huang .

By using of Magnus expansion bargill , the time evolution operator can be read as U(t)=eitHeff,U(t)=e^{-itH_{\mathrm{eff}}}, where the effective Hamiltonian is (see Appendix B for details)

Heff=λJz+Δ(t)Jz2+iJzk(αkbkαbk)iΓlossN,H_{\mathrm{eff}}=\lambda^{\prime}J_{z}+\Delta(t)J_{z}^{2}+iJ_{z}\sum_{k}(\alpha_{k}b_{k}^{\dagger}-\alpha^{\ast}b_{k})-i\Gamma_{\mathrm{loss}}N_{\uparrow}, (18)

with λλ+δNΔ(t)\lambda^{\prime}\equiv\lambda+\delta_{\uparrow}-N\Delta(t) and αkgk(1eiωkt)/(tωk)\alpha_{k}\equiv g_{k}(1-e^{i\omega_{k}t})/(t{\omega_{k}}). From the above equation, we can find that the collisional interaction between the atoms and reservoir induce a nonlinear term Jz2\propto J_{z}^{2}, corresponding to the OAT Hamiltonian, and the noise induced nonlinear strength is Breuer

Δ(t)=1t0𝑑ωJ(ω)ωtsin(ωt)ω2.\displaystyle\Delta(t)=\frac{1}{t}\int_{0}^{\infty}d\omega J(\omega)\frac{\omega t-\sin(\omega t)}{\omega^{2}}. (19)

Here, the reservoir spectral density J(ω)J(\omega) defined as J(ω)=k0|gk|2δ(ωεk/)J(\omega)=\sum_{k\neq 0}\left|g_{k}\right|^{2}\delta(\omega-\varepsilon_{k}/\hbar). In the continuum limit, L1k(2π)1𝑑kL^{-1}\sum_{k}\rightarrow(2\pi)^{-1}\int dk, we have

J(ω)\displaystyle J(\omega) =\displaystyle= Θω3B30𝑑kk2ek2A2/2ε(k)δ(ωε(k))\displaystyle{\Theta}\hbar\omega_{\perp}^{3}\ell_{B}^{3}\int_{0}^{\infty}dk\frac{k^{2}e^{-k^{2}\ell_{A}^{2}/2}}{\varepsilon(k)}\delta\left(\omega-\frac{\varepsilon(k)}{\hbar}\right) (20)
=\displaystyle= Θω3B3if(ki(ω))ω|dε(k)dk|k=ki(ω)1,\displaystyle{\Theta}\hbar\omega_{\perp}^{3}\ell_{B}^{3}\sum_{i}\frac{f(k_{i}(\omega))}{\omega}\left|\frac{d\varepsilon(k)}{dk}\right|_{k=k_{i}(\omega)}^{-1},

where f(k)k2ek2A2/2f(k)\equiv k^{2}e^{-k^{2}\ell_{A}^{2}/2} with ki(ω)k_{i}(\omega) being the roots of the equation ε(k)=ω,\varepsilon(k)=\hbar\omega, and the dimensionless parameter is

Θ=n0B3aAB2(mA+mB)2πmA2(A2+B2)2.{\Theta}=\frac{n_{0}\ell_{B}^{3}a_{AB}^{2}(m_{A}+m_{B})^{2}}{\pi m_{A}^{2}\left(\ell_{A}^{2}+\ell_{B}^{2}\right)^{2}}. (21)

Assuming that the initial state of the total system is given by

ρT(0)=|Φ(0)AΦ(0)|ρB,\rho_{T}(0)=\left|\Phi(0)\right\rangle_{A}\left\langle\Phi(0)\right|\otimes\rho_{B}, (22)

where |Φ(0)A12N/2(|+|)N=mcm(0)|j,m\left|\Phi(0)\right\rangle_{A}\equiv\frac{1}{2^{N/2}}\left(\left|\uparrow\right\rangle+\left|\downarrow\right\rangle\right)^{\otimes N}=\sum_{m}c_{m}(0)\left|j,m\right\rangle is CSS, with the probability amplitudes cm=2j(C2jj+m)1/2c_{m}=2^{-j}\left(C_{2j}^{j+m}\right)^{1/2} and total spin j=N/2j=N/2 for a system consisting of NN condensated atoms. And the density matrix of reservoir read as

ρB=Πk[1exp(βωk)]exp(βωkbkbk),\rho_{B}=\Pi_{k}[1-\exp(-\beta\omega_{k})]\exp(-\beta\omega_{k}b_{k}^{{\dagger}}b_{k}), (23)

with β\beta the inverse temperature. With the help of Eq. (18), the time-evolution reduced matrix elements of the atom system at any later time tt is found by tracing over the reservoir degrees of freedom

ρjm,jnA(t)\displaystyle\rho_{jm,jn}^{A}(t) =\displaystyle= eitλ(mn)eitΔ(t)(m2n2)\displaystyle e^{-it\lambda^{\prime}(m-n)}e^{it\Delta(t)(m^{2}-n^{2})} (24)
×eΓlosst(m+n+N)et(mn)2γ(t)ρjm,jnA(0),\displaystyle\times e^{-\Gamma_{\mathrm{loss}}t(m+n+N)}e^{-t(m-n)^{2}\gamma(t)}\rho_{jm,jn}^{A}(0),

with decoherence function

γ(t)\displaystyle\gamma(t) =\displaystyle= 1t0𝑑ωJ(ω)coth(ω2kBT)1cos(ωt)ω2,\displaystyle\frac{1}{t}\int_{0}^{\infty}d\omega J(\omega)\coth\left(\frac{\hbar\omega}{2k_{B}T}\right)\frac{1-\cos(\omega t)}{\omega^{2}}, (25)

which is a function of temperature TT.

In Eq. (24), Δ(t)\Delta(t) and γ(t)\gamma(t), respectively, correspond to the unitary and non-unitary evolution due to the effects of the reservoir. Equations (19) and (25) show that both Δ(t)\Delta(t) and γ(t)\gamma(t) depend on the reservoir spectral density J(ω)J(\omega), which can be controlled by tuning the MDDI of the dipolar Bose gas reservoir  Giovanazzi ; Griesmaier2006 ; Lahaye . This is the main difference from Feshbach resonances method whose control is on the system atoms directly and will strongly enhance three-body loss near resonances regime beside the one-body loss. Therefore, in our model the one-body particle loss due to the inelastic collisions of the noncondensed atoms maybe the main loss mechanism that need considering. When temperature is low enough the values of one-body loss rate Γloss\Gamma_{\mathrm{loss}} will be very small, since there is only a small number of thermal atoms with sufficient energy to knock atoms out of condensate Dalfovo . For simplicity, hereafter we only consider the case of T0T\rightarrow 0, and choose Γloss\Gamma_{\mathrm{loss}} as a free parameter.

IV Spin squeezing and entangled non-Gaussian spin states

IV.1 Spin squeezing parameter

Now, a state is regarded as squeezed if the variance of one spin component normal to the mean spin vector 𝑱=Tr[𝑱ρA(t)]\langle{\boldsymbol{J}}\rangle={\mathrm{Tr}}\left[\boldsymbol{J}\rho^{A}(t)\right] is lower than the Heisenberg limited value. The SS parameter defined by Wineland is Wineland1994

ξR2=N(ΔJn^)min2|𝐉|2,\xi_{R}^{2}=\frac{N(\Delta J_{\hat{n}_{\perp}})_{\mathrm{min}}^{2}}{|\langle{\mathbf{J}}\rangle|^{2}},

where (ΔJn^)min(\Delta J_{\hat{n}_{\perp}})_{\mathrm{min}} represents the minimal variance of the spin component perpendicular to the mean spin direction r^0𝐉/|𝐉|\hat{r}_{0}\equiv\langle{\mathbf{J}}\rangle/|\langle{\mathbf{J}}\rangle|. Where the mean spin is |𝐉|=Jx2+Jx2+Jz2|\langle{\mathbf{J}}\rangle|=\sqrt{\left\langle J_{x}\right\rangle^{2}+\left\langle J_{x}\right\rangle^{2}+\left\langle J_{z}\right\rangle^{2}}. A state is spin squeezed if ξR2<1\xi_{R}^{2}<1. In addition, the smaller ξR2\xi_{R}^{2} is, the stronger the squeezing is. If Γloss=0\Gamma_{\mathrm{loss}}=0, the squeezing parameter can be evaluated explicitly

ξR2=4+(N1)(A~A~2+B~2)4e2tγ(t)[cos(tΔ(t))]2N2,\xi_{R}^{2}=\frac{4+(N-1)\left(\tilde{A}-\sqrt{\tilde{A}^{2}+\tilde{B}^{2}}\right)}{4e^{-2t\gamma(t)}\left[\cos(t\Delta(t))\right]^{2N-2}}, (26)

which does not depend on λ\lambda^{\prime}. And the optimally squeezed direction is ϕopt=[π+tan1(B~/A~)]\phi_{\rm opt}=[\pi+\tan^{-1}(\tilde{B}/\tilde{A})], with

A~\displaystyle\tilde{A} =\displaystyle= 1cosN2[2(tΔ(t))]exp[4tγ(t)],\displaystyle 1-\cos^{N-2}[2(t\Delta(t))]\exp[-4t\gamma(t)],
B~\displaystyle\tilde{B} =\displaystyle= 4sin[tΔ(t)]cosN2[tΔ(t)]exp[tγ(t)].\displaystyle-4\sin[t\Delta(t)]\cos^{N-2}[t\Delta(t)]\exp[-t\gamma(t)]. (27)

When considering the one-body losses, Γloss0,\Gamma_{\mathrm{loss}}\neq 0, the form of SS has given in Appendix A.

Equation (26) indicates that the dephasing noise plays two different roles: On one hand, it can generate the SS by creating the nonlinear interaction Δ(t)\Delta(t); on the other hand, it degrades the degree of SS via the decoherence function γ(t)\gamma(t).

Refer to caption
Figure 2: Time dependence of Δ(t)\Delta(t) and γ(t)\gamma(t) for different ϵdd\epsilon_{dd}. Here we choose Θ=1.5×102\Theta=1.5\times 10^{-2} and η=5\eta=5. Note that for repulsive MDDI the values of noise-induced nonlinear interaction Δ(t)\Delta(t) approach to their steady values Δ()\Delta(\infty), while decoherence function γ(t)\gamma(t) is decreasing with time.
Refer to caption
Figure 3: Spin squeezing dynamics of two-mode BEC consisting of N=100N=100 atoms coupled to a 1D dipolar Bose gas reservoir. (a) Squeezing parameter for different values of ϵdd\epsilon_{dd}: from top to bottom ϵdd=1,0,1\epsilon_{dd}=1,0,-1. Here Γloss=0\Gamma_{\mathrm{loss}}=0. (b) Squeezing parameter for different values of loss parameters with ϵdd=1\epsilon_{dd}=-1, from top to bottom Γloss=0.01Δ(),0.002Δ()\Gamma_{\mathrm{loss}}=0.01\Delta(\infty),0.002\Delta(\infty), and 0.

IV.2 QFI and entanglement of non-Gaussian spin states

To investigate the entanglement created by the dipolar BEC reservoir, we can also introduce the QFI. In general, the states that are entangled and useful for sub-shot-noise-limited parameter-estimation precision is identified by the QFI criterion FQ>NF_{Q}>N. The QFI FQF_{Q} with respect to θ\theta, acquired by an SU(2) rotation, can be described as Ma2011B

FQ[ρ(θ,t),J^n]=n𝑪nT,F_{Q}[\rho(\theta,t),\hat{J}_{\vec{n}}]=\vec{n}\boldsymbol{C}\vec{n}^{T}, (28)

where ρ(θ,t)=exp(iθJn)ρ(t)exp(iθJ(n))\rho(\theta,t)=\exp(-i\theta J_{\vec{n}})\rho(t)\exp(i\theta J_{(}\vec{n})) with n\vec{n} being the optimal rotation direction, and the matrix element for the symmetric matrix 𝑪\boldsymbol{C} is

Ckl=ij(pipj)2pi+pj[i|Jk|jj|Jl|i+i|Jl|jj|Jk|i,C_{kl}=\sum_{i\neq j}\frac{(p_{i}-p_{j})^{2}}{p_{i}+p_{j}}[\left\langle i\right|J_{k}\left|j\right\rangle\left\langle j\right|J_{l}\left|i\right\rangle+\left\langle i\right|J_{l}\left|j\right\rangle\left\langle j\right|J_{k}\left|i\right\rangle, (29)

with pi(|i)p_{i}(|i\rangle) being the eigenvalues (eigenvectors) of ρ(θ,t)\rho(\theta,t).

For simplicity, we first consider the case of small NN, e.g., N=2N=2. When setting λ=0\lambda^{\prime}=0 and neglecting the particle loss, e.g., Γloss=0\Gamma_{\mathrm{loss}}=0, the QFI can be calculated analytically

FQ[ρ(θ,t),J^n]=max[Cxx,C],F_{Q}[\rho(\theta,t),\hat{J}_{n}]=\max[C_{xx},C_{\perp}], (30)

with

Cxx=4sinh2[2γ(t)t]+16e2γ(t)t1+3e4γ(t)t[116cos2[Δ(t)t]e6γ(t)t[1e4γ(t)t]2+16]C_{xx}=\frac{4\sinh^{2}[2\gamma(t)t]+16e^{2\gamma(t)t}}{1+3e^{4\gamma(t)t}}\left[1-\frac{16\cos^{2}[\Delta(t)t]}{e^{-6\gamma(t)t}[1-e^{4\gamma(t)t}]^{2}+16}\right] (31)

in xx-axis direction, and

C=Cyy+Czz+(Cyy+Czz)2+4Cyz22C_{\perp}=\frac{C_{yy}+C_{zz}+\sqrt{(C_{yy}+C_{zz})^{2}+4C_{yz}^{2}}}{2} (32)

in yzyz plane, which can also be obtained by using of Eq. (29) (see Appendix B).

The maximal QFI can be found in xx axis direction

FQmax=4sinh2[2γ(t)topt]+16e2γ(t)topt1+3e4γ(t)toptF_{Q}^{\mathrm{max}}=\frac{4\sinh^{2}[2\gamma(t)t_{\mathrm{opt}}]+16e^{2\gamma(t)t_{\mathrm{opt}}}}{1+3e^{4\gamma(t)t_{\mathrm{opt}}}} (33)

when choosing optimal interrogation time topt=π/[2Δ(t)]t_{\mathrm{opt}}=\pi/[2\Delta(t)]. Equation (33) reveals that the values of γ(t)\gamma(t) is directly related to the QFI. One finds FQmaxN2F_{Q}^{\mathrm{max}}\rightarrow N^{2} (the Heisenberg limit) if γ(t)0\gamma(t)\rightarrow 0. Fortunately, Fig. 2(b) indicates that the nearly neglectable γ(t)\gamma(t) can be obtained when the dipolar Bose gas reservoir with repulsive MDDI. Therefore, the main limitation of Heisenberg scaling is one-body loss mechanism in our scheme. Next, we consider large NN and the case of Γloss0\Gamma_{\mathrm{loss}}\neq 0 numerically.

V Results and discussion

As a concrete example, we consider a BEC reservoir of Dy162{}^{162}\mathrm{Dy} atoms, for which we have μm=9.9μB\mu_{m}=9.9\mu_{B}  and aB=112a0a_{B}=112a_{0} with μB\mu_{B} the Bohr magneton. This means that add131a0a_{dd}\simeq 131a_{0}\ and dipolar interaction is mainly attractive since ϵdd>1,\epsilon_{dd}>1, then the attraction is stronger than the short-range repulsion. However, not only the contact interaction strength can be tunable via Feshbach resonance, but also both the sign and the strength of the effective dipolar interaction can be tuned by the use of a fast rotating orienting field. Later, we will consider the values of ϵdd[1,1]\epsilon_{dd}\in[-1,1], which is repulsive (attractive) MDDI for ϵdd<0\epsilon_{dd}<0 (ϵdd>0\epsilon_{dd}>0).

Refer to caption
Figure 4: (a) Time dependence of QFI amplification rate with repulsive interaction ϵdd=1\epsilon_{dd}=-1. We show three cases with N=10,30N=10,30 and 5050. (b) The maximal QFI amplification rate FQmax/NF^{\mathrm{max}}_{Q}/N as a function of atom number NN with ϵdd=1\epsilon_{dd}=-1. (c) QFI amplification rate and (d) the corresponding optimal evolution time ωtopt\omega_{\perp}t_{\mathrm{opt}} with respect to ϵdd\epsilon_{dd} when N=100N=100. Here Γloss=0\Gamma_{\mathrm{loss}}=0.

Numerically, it is convenient to introduce the dimensionless units: ω\hbar\omega_{\perp} for energy, ω1\omega_{\perp}^{-1} for time, and B=[/(mω)]1/2\ell_{B}=[\hbar/(m\omega_{\perp})]^{1/2} for length. To obtain the values of dimensionless parameter η\eta and Θ\Theta, we assume a quasi-1D trap with ωx=2π×20Hz\omega_{x}=2\pi\times 20\mathrm{Hz} and ω=2π×103Hz\omega_{\perp}=2\pi\times 10^{3}\mathrm{Hz}; and the corresponding harmonic oscillator width is B=A2.5×107m\ell_{B}=\ell_{A}\simeq 2.5\times 10^{-7}\mathrm{m}. We assume that the linear density of the quasi-1D condensate is n0=108m1n_{0}=10^{8}\mathrm{m^{-1}}, and the ss-wave scattering length between Rb and Dy atoms is aAB5nma_{AB}\sim\mathrm{5nm} yuan . Then, we shall take η=5\eta=5 and Θ=1.5×102\Theta=1.5\times 10^{-2} in the results presented below.

Since both the SS and QFI depend on γ(t)\gamma(t) and Δ(t),\Delta(t), let us first investigate the time dependence of dephasing factor. From Fig. 2, we can clearly see that the squeezing rate Δ(t)\Delta(t) is nearly constant, Δ()\Delta(\infty). Whereas γ(t)\gamma(t) is decreasing with time, what is more we can get very small γ(t)\gamma(t) (e.g.10310410^{-3}-10^{-4}) when ϵdd<0\epsilon_{dd}<0. Comparing Δ(t)\Delta(t) and γ(t)\gamma(t), we can also find that the values of Δ(t)\Delta(t) (squeezing rate) is larger than γ(t)\gamma(t) (dispersive rate) for ϵdd<0\epsilon_{dd}<0, which means that we can obtain strong squeezing and large QFI by the reservoir’s engineering with repulsive MDDI.

In Fig. 3, we plot the SS ξR2\xi^{2}_{R} dynamics of two-mode BEC consisting NN atoms coupled to a 1D dipolar Bose gas reservoir for various ϵdd\epsilon_{dd} values. As is shown in Fig. 3(a), the optimal squeezing can be achieved within short time scale, and after a transient time, it is lost (ξR2>1\xi_{R}^{2}>1) and then ENGSs are produced. However, for repulsive-dipolar-interaction reservoir, we can obtain stronger SS, due to their smaller dispersive rate γ(t)\gamma(t). In Fig. 3(b), we further plot the time evolution of ξR2\xi^{2}_{R} for various atom loss rates. It indicates that the squeezing degree degraded with the increasing of the atom loss rate.

Figure 4 illustrates the QFI amplification rates with respect to the initial state (CSS) FQ/N{F}_{Q}/N. In Fig. 4(a), we present the time dependence of QFI amplification rates with repulsive interaction. We see that differ from the case of SS, the amplified QFI can last for a very long time, which means that the ENGSs can be produced and achieve the maximal even in the regimes without squeezing. The optimal QFI first monotonically increasing and then reach a metastable N/2\sim N/2 in the yzyz plane. Subsequently, the QFI suddenly increasing in the xx-axis direction at the optimal interrogation time toptt_{\mathrm{opt}}. As shown in Fig. 4(b), the maximal amplification rate FQmax/NF^{\mathrm{max}}_{Q}/N is proportional to the atom number NN, and the scale factor is N\sim N, which is the Heisenberg scale. What is more, compared with the attractive MDDI reservoir the repulsive interaction can induce larger QFI, this result is presented in Fig. 4(c).

Comparison of Figs. 4(a) and 5(a), we can see that for not too large atom loss rates Γloss\Gamma_{\mathrm{loss}} the optimal evolution time is topt=π/[2Δ(t)]t_{\mathrm{opt}}=\pi/[2\Delta(t)], which is the same as the case of N=2N=2.

Refer to caption
Figure 5: (a) QFI vs time ωt\omega_{\perp}t for various Γloss\Gamma_{\mathrm{loss}} with N=2N=2. The solid line (black) corresponds to the analytic solution given in Eqs. (30-33). (b) The maximal QFI as a function of atom number NN for various Γloss\Gamma_{\mathrm{loss}}. The shaded area indicates the regime between shot-noise limit and Heisenberg limit. (c) The rates of the Heisenberg limit FQmax/N2F^{\mathrm{max}}_{Q}/N^{2} with respect to atom number NN for different values of loss parameters Γloss\Gamma_{\mathrm{loss}}. Other parameters are ϵdd=1\epsilon_{dd}=-1, from top to bottom Γloss=0,0.001Δ(),0.002Δ(),\Gamma_{\mathrm{loss}}=0,0.001\Delta(\infty),0.002\Delta(\infty), and 0.005Δ()0.005\Delta(\infty).

Figures 5(b) and (c) illustrate the QFI with respect to atom number NN for different values of loss parameters. As shown in Fig. 5(b), under the values of Γloss\Gamma_{\mathrm{loss}} we considered, we can obtain near-Heisenberg scaling. Figures 5(c) plots the rates of the Heisenberg limit FQmax/N2F^{\mathrm{max}}_{Q}/N^{2} with respect to atom number NN for different values of loss parameters Γloss\Gamma_{\mathrm{loss}}. It indicates that for not too large Γloss\Gamma_{\mathrm{loss}}, we can obtain the Heisenberg scaling only with a prefactor. We can also see, when increasing the loss rate the values of FQmax/N2F^{\mathrm{max}}_{Q}/N^{2} degrades with the increasing of number of atoms NN, since the collective dissipate rare NΓlossN\Gamma_{\mathrm{loss}} depends on NN. Fortunately, for the low temperature limit we considered the values of Γloss\Gamma_{\mathrm{loss}} is not too large, hence we can still obtain the robust sub-shot-noise-limited phase sensitivity.

Refer to caption
Figure 6: (a) The fidelity between our optimal ENGSs and spin cat state with respect to ϵdd\epsilon_{dd} for different values of atom number NN. Here we choose Γloss=0\Gamma_{\rm loss}=0. (b) The fidelity as a function of atom number NN for various Γloss\Gamma_{\rm loss} with ϵdd=1\epsilon_{dd}=-1.

To demonstrate the near Heisenberg-limited sensitivity with the ENGSs realized by our model, we calculate the fidelity between the optimal ENGSs and spin cat states Momer , which are the maximal entangled states and have the Heisenberg-limited sensitivity for metrology. It is given by

|Ψcat=12(|π2,0eiπ2(N+1)|π2,π),|\Psi\rangle_{\mathrm{cat}}=\frac{1}{\sqrt{2}}\left(\left|\frac{\pi}{2},0\right\rangle-e^{-i\frac{\pi}{2}(N+1)}\left|\frac{\pi}{2},\pi\right\rangle\right), (34)

where |θ0,ϕ0eiθ0(Jxsinϕ0Jycosϕ0)|j,j|\theta_{0},\phi_{0}\rangle\equiv e^{i\theta_{0}(J_{x}\sin\phi_{0}-J_{y}\cos\phi_{0})|j,j\rangle} is the CSS. With the definition of fidelity, we have

ϱ=trϱcat1/2ρA(topt)ϱcat1/2,\mathcal{F}_{\varrho}=\mathrm{tr}\sqrt{\varrho_{\mathrm{cat}}^{1/2}\rho^{A}(t_{\rm opt})\varrho_{\mathrm{cat}}^{1/2}}, (35)

where ϱcat=|ΨcatΨ|\varrho_{\mathrm{cat}}=|\Psi\rangle_{\mathrm{cat}}\langle\Psi|.

Form Fig. 6, we can find that the fidelity depends on both the ϵdd\epsilon_{dd} and particle number NN. The maximal values occurs at ϵdd=1\epsilon_{dd}=-1. Because of the dissipative rate γ(t)\gamma(t), the fidelity decrease with the increase of NN. As shown in Fig. 6(a), when Γloss=0\Gamma_{\rm loss}=0 we get ϱ0.94,0.85\mathcal{F}_{\varrho}\simeq 0.94,0.85, and 0.800.80 for N=10,30N=10,30, and 5050. And the effect of Γloss\Gamma_{\rm loss} on fidelity also is shown in Fig. 6(b).

VI conclusion

In summary, we have realized the SSSs and ENGSs by immersing atoms in a thermally equilibrated quasi-1D dipolar BEC reservoir. It has demonstrated that the repulsive-dipolar-interaction reservoir can induce better SS and entanglement. We have shown that owing to the dephasing noise, even in the regimes without SS the ENGSs can successively undergo highly metastable entangled states and entanglement suddenly increase. To explain the highly sensitivity for metrology, we calculated the fidelity between the optimal ENGSs and spin cat states, and found that the optimal ENGS is similar to the spin cat state. It has confirmed that by the use of ENGSs for metrology, the phase estimation sensitivity can surpass that by SS and even approach to Heisenberg limit for neglectable atom loss rates. The effect of the atom loss rate as a free parameter has also been considered.

Finally, we give two remarks on the above obtained results: First, these results we have obtained in this paper are based on the negligible spatial evolution of the immersed condensate wave functions. Usually this assumption is enough to capture the basic processes and physics, and detailed consideration of the impact of spatial dynamics can be investigated by adopting the multi-configurational time-dependent Hartree for bosons method stre2006 . Second, the scheme we proposed in this work can also suit the case that the system atoms are weak or no interaction two-level impurity atoms which are not condensate.

Acknowledgements.
This work was supported by the NSFC under Grant No. 11547159. G.R.J. acknowledges support from the Major Research Plan of the NSFC (Grant No. 91636108).

Appendix A Derivation of Eq. (7)

Here, we present a detailed derivation of the Fourier transform of the effective 1D interaction potential. It can be obtained by integrating out the yy and zz variables as

V~1D(k)\displaystyle\tilde{V}_{1D}(k) =\displaystyle= 12π𝑑y𝑑z|Ψ(y,z)|2Fyz1[Fyz[|Ψ(y,z)|2]V~(k)]\displaystyle\frac{1}{2\pi}\int dydz\left|\Psi_{\perp}(y,z)\right|^{2}F_{yz}^{-1}\left[F_{yz}\left[\left|\Psi_{\perp}(y,z)\right|^{2}\right]\tilde{V}(k)\right] (36)
=\displaystyle= 12π𝑑y𝑑z|Ψ(y,z)|2Fyz1[12πe(ky2lB2+kz2lB2)/4[gBcd(13(μ^m𝐞^k)2)]].\displaystyle\frac{1}{2\pi}\int dydz\left|\Psi_{\perp}(y,z)\right|^{2}F_{yz}^{-1}\left[\frac{1}{2\pi}e^{-(k_{y}^{2}l_{B}^{2}+k_{z}^{2}l_{B}^{2})/4}\left[g_{B}-c_{d}\left(1-3(\mathbf{\hat{\mu}}_{m}\mathbf{\cdot\hat{e}}_{k})^{2}\right)\right]\right].

When assuming the dipole moments lie on the xzxz plane forming an angle φ\varphi to xx axis, i.e.,

μ^m=(cosφ,0,sinφ).\mathbf{\hat{\mu}}_{m}=(\cos\varphi,0,\sin\varphi). (37)

Then, we have

V~1D(k)\displaystyle\tilde{V}_{1D}(k) =\displaystyle= 1(2π)2𝑑ϕ𝑑kke(k2lB2)/2[gBcd(13(kcosφ+kcosϕsinφ)2k2+k2)]\displaystyle\frac{1}{\left(2\pi\right)^{2}}\int d\phi\int dk_{\perp}k_{\perp}e^{-(k_{\perp}^{2}l_{B}^{2})/2}\left[g_{B}-c_{d}\left(1-3\frac{(k\cos\varphi+k_{\perp}\cos\phi\sin\varphi)^{2}}{k^{2}+k_{\perp}^{2}}\right)\right] (38)
=\displaystyle= gB2πlB2cd2πlB2(132sin2φ)[132k2lB2exp(k2lB22)Γ(0,k2lB22)]\displaystyle\frac{g_{B}}{2\pi l_{B}^{2}}-\frac{c_{d}}{2\pi l_{B}^{2}}\left(1-\frac{3}{2}\sin^{2}\varphi\right)\left[1-\frac{3}{2}k^{2}l_{B}^{2}\exp\left(\frac{k^{2}l_{B}^{2}}{2}\right)\Gamma\left(0,\frac{k^{2}l_{B}^{2}}{2}\right)\right]
=\displaystyle= gB2πlB2c~d2πlB2[132k2lB2exp(k2lB22)Γ(0,k2lB22)]\displaystyle\frac{g_{B}}{2\pi l_{B}^{2}}-\frac{\tilde{c}_{d}}{2\pi l_{B}^{2}}\left[1-\frac{3}{2}k^{2}l_{B}^{2}\exp\left(\frac{k^{2}l_{B}^{2}}{2}\right)\Gamma\left(0,\frac{k^{2}l_{B}^{2}}{2}\right)\right]
=\displaystyle= gB2πlB2{1ϵ~dd[132k2lB2exp(k2lB22)Γ(0,k2lB22)]},\displaystyle\frac{g_{B}}{2\pi l_{B}^{2}}\left\{1-\tilde{\epsilon}_{dd}\left[1-\frac{3}{2}k^{2}l_{B}^{2}\exp\left(\frac{k^{2}l_{B}^{2}}{2}\right)\Gamma\left(0,\frac{k^{2}l_{B}^{2}}{2}\right)\right]\right\},

where ϵ~dd=c~d/gB\tilde{\epsilon}_{dd}=\tilde{c}_{d}/g_{B} with k=ky2+kz2k_{\perp}=\sqrt{k_{y}^{2}+k_{z}^{2}} and c~d=cd(132sin2φ)\tilde{c}_{d}=c_{d}\left(1-\frac{3}{2}\sin^{2}\varphi\right). Clearly, the effective 1D dipolar interaction vanishes at the magical angle αm=54.74\alpha_{m}=54.74^{\circ}, and it is attractive (repulsive) for α<αm(α>αm)\alpha<\alpha_{m}(\alpha>\alpha_{m}). In the main text, we have dropped the tilde on c~d\tilde{c}_{d} and ϵ~dd\tilde{\epsilon}_{dd}, and will only consider the values of ϵdd[1,1]\epsilon_{dd}\in[-1,1].

Appendix B Time evolution operator U(t)U(t)

The time evolution operator can be obtained by using Magnus expansion

U(t)T+exp[i0tHI(t)𝑑t]=exp[n=1(i)nn!Fn(t)].U(t)\equiv\mathrm{T}_{+}\exp\left[-i\int_{0}^{t}H_{I}(t^{\prime})dt^{\prime}\right]=\exp\left[\sum_{n=1}^{\infty}\frac{(-i)^{n}}{n!}F_{n}(t)\right]. (39)

Note that only the below first two terms of the expansion are non-zero

F1(t)\displaystyle F_{1}(t) =\displaystyle= 0tHI(t)𝑑t=λtJz+(Jz+N2)tk0t(gkbkeiωkt+gkbkeiωkt)\displaystyle\int_{0}^{t}H_{I}(t^{\prime})dt^{\prime}=\lambda tJ_{z}+\left(J_{z}+\frac{N}{2}\right)t\sum_{k}\int_{0}^{t}\left(g_{k}b_{k}^{{\dagger}}e^{i\omega_{k}t^{\prime}}+g_{k}^{\ast}b_{k}e^{-i\omega_{k}t^{\prime}}\right) (40)
=\displaystyle= λtJz+(Jz+N2)k(αkbkαkbk)iΓloss,\displaystyle\lambda tJ_{z}+\left(J_{z}+\frac{N}{2}\right)\sum_{k}(\alpha_{k}b_{k}^{{\dagger}}-\alpha_{k}^{\ast}b_{k})-i\Gamma_{\mathrm{loss}},
F2(t)\displaystyle F_{2}(t) =\displaystyle= 0t𝑑s0s𝑑s[HI(s),HI(s)]\displaystyle\int_{0}^{t}ds\int_{0}^{s}ds^{\prime}[H_{I}(s),H_{I}(s^{\prime})] (41)
=\displaystyle= 2iN2k|gk|20t𝑑s0s𝑑ssinωk(ss)=2iN2tΔ(t),\displaystyle-2iN_{\uparrow}^{2}\sum_{k}\left|g_{k}\right|^{2}\int_{0}^{t}ds\int_{0}^{s}ds^{\prime}\sin\omega_{k}(s-s^{\prime})=-2iN_{\uparrow}^{2}t\Delta(t),

with the amplitudes αk=igk0teiωks𝑑s/t=gk(1eiωks)/ωkt\alpha_{k}=-ig_{k}\int_{0}^{t}e^{i\omega_{k}s}ds/t=g_{k}(1-e^{i\omega_{k}s})/\omega_{k}t, since [HI(s),HI(s)]=2iN2k|gk|2sinωk(ss),[H_{I}(s),H_{I}(s^{\prime})]=-2iN_{\uparrow}^{2}\sum_{k}\left|g_{k}\right|^{2}\sin\omega_{k}(s-s^{\prime}), which commutes with the high order terms. It is worth to point out that the commutator of the interaction Hamiltonian at two different times is an operator but not a CC number as considering in the single bit case, which can induce the nonlinear interaction; the noise-induced nonlinear interaction strengthen Δ(t)\Delta(t) can recast as

Δ(t)\displaystyle\Delta(t) =\displaystyle= 1tk|gk|20t𝑑s0s𝑑ssinωk(ss)=1t0𝑑ωJ(ω)0t𝑑s0s𝑑ssinω(ss)\displaystyle\frac{1}{t}\sum_{k}\left|g_{k}\right|^{2}\int_{0}^{t}ds\int_{0}^{s}ds^{\prime}\sin\omega_{k}(s-s^{\prime})=\frac{1}{t}\int_{0}^{\infty}d\omega J(\omega)\int_{0}^{t}ds\int_{0}^{s}ds^{\prime}\sin\omega(s-s^{\prime}) (42)
=\displaystyle= 1t0𝑑ωJ(ω)ωtsin(ωt)ω2,\displaystyle\frac{1}{t}\int_{0}^{\infty}d\omega J(\omega)\frac{\omega t-\sin(\omega t)}{\omega^{2}},

where we have used the relation k|gk|20𝑑ωJ(ω)\sum_{k}\left|g_{k}\right|^{2}\rightarrow\int_{0}^{\infty}d\omega J(\omega). Then,

U(t)\displaystyle U(t) =\displaystyle= exp[iF1(t)12F2(t)]=exp(iλtJz)exp[(Jz+N2)k(αkbkαkbk)]exp[itN2Δ(t)]etΓloss\displaystyle\exp\left[-iF_{1}(t)-\frac{1}{2}F_{2}(t)\right]=\exp(-i\lambda tJ_{z})\exp\left[\left(J_{z}+\frac{N}{2}\right)\sum_{k}(\alpha_{k}b_{k}^{{\dagger}}-\alpha_{k}^{\ast}b_{k})\right]\exp[itN_{\uparrow}^{2}\Delta(t)]e^{-t\Gamma_{\mathrm{loss}}} (43)
=\displaystyle= exp[itλJz]exp[itΔ(t)Jz2]exp(tΓloss)exp[iϕ0(t)]exp[Jzk(αkbkαkbk)],\displaystyle\exp\left[-it\lambda^{\prime}J_{z}\right]\exp\left[it\Delta(t)J_{z}^{2}\right]\exp(-t\Gamma_{\mathrm{loss}})\exp[i\phi_{0}(t)]\exp\left[J_{z}\sum_{k}(\alpha_{k}b_{k}^{{\dagger}}-\alpha_{k}^{\ast}b_{k})\right],

where λ=λNΔ(t)\lambda^{\prime}=\lambda-N\Delta(t) and ϕ0(t)\phi_{0}(t) is the global phase and will be dropped.

Appendix C Spin squeezing with Γloss0\Gamma_{\mathrm{loss}}\neq 0

In this Appendix, we present a detailed of SS with Γloss0\Gamma_{\mathrm{loss}}\neq 0. To this end, we assume, without loss of generality, that n^0=(sinϑcosϕ,sinϑsinϕ,cosϑ)\hat{n}_{0}=(\sin\vartheta\cos\phi,\sin\vartheta\sin\phi,\cos\vartheta), where ϑ=tan1(Jx2+Jy2/Jz)\vartheta=\tan^{-1}\left(\sqrt{\langle J_{x}\rangle^{2}+\langle J_{y}\rangle^{2}}/\langle J_{z}\rangle\right) and ϕ=tan1(Jy/Jx)\phi=\tan^{-1}\left(\langle J_{y}\rangle/\langle J_{x}\rangle\right) are polar and azimuthal angles, respectively. We then define two mutually perpendicular unit vectors n^1=(sinϕ,cosϕ,0)\hat{n}_{1}=(-\sin\phi,\cos\phi,0) and n^2=(cosϑcosϕ,cosϑsinϕ,sinϑ)\hat{n}_{2}=(\cos\vartheta\cos\phi,\cos\vartheta\sin\phi,-\sin\vartheta). Clearly, both n^1\hat{n}_{1} and n^2\hat{n}_{2} are perpendicular to n^0\hat{n}_{0} such that (n^1,n^2,n^0)(\hat{n}_{1},\hat{n}_{2},\hat{n}_{0}) form a right-hand frame. Now, the minimal fluctuation of a spin component perpendicular to the mean spin is

(ΔJn^)min2=12[CA2+B2],(\Delta J_{\hat{n}_{\perp}})_{\mathrm{min}}^{2}=\frac{1}{2}\left[C-\sqrt{A^{2}+B^{2}}\right], (44)

and the mean spin is

|𝐉|=Jx2+Jx2+Jz2=|J+|2+Jz2,|\langle{\mathbf{J}}\rangle|=\sqrt{\left\langle J_{x}\right\rangle^{2}+\left\langle J_{x}\right\rangle^{2}+\left\langle J_{z}\right\rangle^{2}}=\sqrt{\left|\langle J_{+}\right\rangle|^{2}+\left\langle J_{z}\right\rangle^{2}}, (45)

where

A\displaystyle A =\displaystyle= sin2ϑ2[j(j+1)3Jz2](1+cos2ϑ)2Re[J+2e2iϕ]+sinϑcosθRe[J+(2Jz+1)eiϕ],\displaystyle\frac{\sin^{2}\vartheta}{2}\left[j(j+1)-3\langle J_{z}^{2}\rangle\right]\frac{(1+\cos^{2}\vartheta)}{2}{\mathrm{Re}}[\langle J_{+}^{2}\rangle e^{-2i\phi}]+\sin\vartheta\cos\theta{\mathrm{Re}}[\langle J_{+}(2J_{z}+1)\rangle e^{-i\phi}],
B\displaystyle B =\displaystyle= cosϑIm[J+2e2iϕ]+sinϑIm[J+(2Jz+1)eiϕ],\displaystyle-\cos\vartheta{\mathrm{Im}}[\langle J_{+}^{2}\rangle e^{-2i\phi}]+\sin\vartheta{\mathrm{Im}}[\left\langle J_{+}(2J_{z}+1)\right\rangle e^{-i\phi}],
C\displaystyle C =\displaystyle= j(j+1)Jz2Re[J+2e2iϕ]sin2ϑ2[j(j+1)3Jz2]+(1+cos2ϑ)2Re[J+2e2iϕ]sin(2ϑ)2Re[J+(2Jz+1)eiϕ],\displaystyle j(j+1)-\langle J_{z}^{2}\rangle-{\mathrm{Re}}[\langle J_{+}^{2}\rangle e^{-2i\phi}]-\frac{\sin^{2}\vartheta}{2}[j(j+1)-3\langle J_{z}^{2}\rangle]+\frac{(1+\cos^{2}\vartheta)}{2}{\mathrm{Re}}[\langle J_{+}^{2}\rangle e^{-2i\phi}]-\frac{\sin(2\vartheta)}{2}{\mathrm{Re}}[\langle J_{+}(2J_{z}+1)\rangle e^{-i\phi}],

with

J+\displaystyle\left\langle J_{+}\right\rangle =\displaystyle= jeitλeNΓlosstetγ(t){cos[tΔ(t)]cosh(Γlosst)+isin[tΔ(t)]sinh(Γlosst)}2j1\displaystyle je^{it\lambda^{\prime}}e^{-N\Gamma_{\mathrm{loss}}t}e^{-t\gamma(t)}\left\{\cos[t\Delta(t)]\cosh(\Gamma_{\mathrm{loss}}t)+i\sin[t\Delta(t)]\sinh(\Gamma_{\mathrm{loss}}t)\right\}^{2j-1}
J+2\displaystyle\left\langle J_{+}^{2}\right\rangle =\displaystyle= j(j12)eNΓlosste4tγ(t)e2itλ{cos[2(tΔ(t))]cosh(Γt)+isin[2(tΔ(t))]sinh(Γlosst)}2j2,\displaystyle j\left(j-\frac{1}{2}\right)e^{-N\Gamma_{\mathrm{loss}}t}e^{-4t\gamma(t)}e^{2it\lambda^{\prime}}\{\cos[2(t\Delta(t))]\cosh(\Gamma t)+i\sin[2(t\Delta(t))]\sinh(\Gamma_{\mathrm{loss}}t)\}^{2j-2},
Jz\displaystyle\left\langle J_{z}\right\rangle =\displaystyle= j2jeNΓlosstsinh(2Γlosst)[1+cosh(2Γlosst)]j1\displaystyle\frac{-j}{2^{j}}e^{-N\Gamma_{\mathrm{loss}}t}\sinh(2\Gamma_{\mathrm{loss}}t)[1+\cosh(2\Gamma_{\mathrm{loss}}t)]^{j-1}
Jz2\displaystyle\left\langle J_{z}^{2}\right\rangle =\displaystyle= jeNΓlosst2j(1+e2Γlosst)2[1+cosh(2Γlosst)]j[2e2Γlosst+j(1e2Γlosst)2],\displaystyle\frac{je^{-N\Gamma_{\mathrm{loss}}t}}{2^{j}(1+e^{2\Gamma_{\mathrm{loss}}t})^{2}}[1+\cosh(2\Gamma_{\mathrm{loss}}t)]^{j}[2e^{2\Gamma_{\mathrm{loss}}t}+j(1-e^{2\Gamma_{\mathrm{loss}}t})^{2}],
J+(2Jz+1)\displaystyle\left\langle J_{+}(2J_{z}+1)\right\rangle =\displaystyle= 2j(j12)eitλeNΓlosstetγ(t){cos[tΔ(t)]cosh(Γlosst)+isin[tΔ(t)]sinh(Γlosst)}2j2\displaystyle 2j\left(j-\frac{1}{2}\right)e^{it\lambda^{\prime}}e^{-N\Gamma_{\mathrm{loss}}t}e^{-t\gamma(t)}\left\{\cos[t\Delta(t)]\cosh(\Gamma_{\mathrm{loss}}t)+i\sin[t\Delta(t)]\sinh(\Gamma_{\mathrm{loss}}t)\right\}^{2j-2}
×{cos[tΔ(t)]sinh(Γlosst)+isin[tΔ(t)]cosh(Γlosst)}.\displaystyle\times\left\{-\cos[t\Delta(t)]\sinh(\Gamma_{\mathrm{loss}}t)+i\sin[t\Delta(t)]\cosh(\Gamma_{\mathrm{loss}}t)\right\}.

Appendix D Matrix elements of CC_{\perp} for N=2N=2

The matrix elements for the symmetric matrix 𝑪\boldsymbol{C}  in yzyz plane are

Cyy\displaystyle C_{yy} =\displaystyle= 4[|β+|2(pp)2p+p+|β|2(pp+)2p+p+],\displaystyle 4\left[\frac{\left|\beta_{+}\right|^{2}(p-p_{-})^{2}}{p+p_{-}}+\frac{\left|\beta_{-}\right|^{2}(p-p_{+})^{2}}{p+p_{+}}\right],
Czz\displaystyle C_{zz} =\displaystyle= 4[|α+|2(pp)2p+p+|α|2(pp+)2p+p+],\displaystyle 4\left[\frac{\left|\alpha_{+}\right|^{2}(p-p_{-})^{2}}{p+p_{-}}+\frac{\left|\alpha_{-}\right|^{2}(p-p_{+})^{2}}{p+p_{+}}\right],
Cyz\displaystyle C_{yz} =\displaystyle= 42[(pp)2α+Imβ+p+p+(pp+)2αImβp+p+],\displaystyle 4\sqrt{2}\left[\frac{(p-p_{-})^{2}\alpha_{+}\mathrm{Im}\beta_{+}}{p+p_{-}}+\frac{(p-p_{+})^{2}\alpha_{-}\mathrm{Im}\beta_{-}}{p+p_{+}}\right], (47)

with

p\displaystyle p =\displaystyle= 14(1e4tγ(t)),p±=18e4tγ(t)[1+3e4tγ(t)±Ξ],\displaystyle\frac{1}{4}(1-e^{-4t\gamma(t)}),\hskip 14.22636ptp_{\pm}=\frac{1}{8}e^{-4t\gamma(t)}\left[1+3e^{4t\gamma(t)}\pm\Xi\right],
Ξ\displaystyle\Xi =\displaystyle= (1e4tγ(t))2+16e6tγ(t),\displaystyle\sqrt{(1-e^{4t\gamma(t)})^{2}+16e^{6t\gamma(t)}}, (48)
α±\displaystyle\alpha_{\pm} =\displaystyle= 2216+e6tγ(t)[1e4tγ(t)±Ξ]2,β±=eitΔ(t)e3tγ(t)[1e4tγ(t)±Ξ]16+e6tγ(t)[1e4tγ(t)±Ξ]2.\displaystyle\frac{2\sqrt{2}}{\sqrt{16+e^{-6t\gamma(t)}\left[1-e^{4t\gamma(t)}\pm\Xi\right]^{2}}},\hskip 14.22636pt\beta_{\pm}=\frac{-e^{it\Delta(t)}e^{-3t\gamma(t)}\left[1-e^{4t\gamma(t)}\pm\Xi\right]}{\sqrt{16+e^{-6t\gamma(t)}\left[1-e^{-4t\gamma(t)}\pm\Xi\right]^{2}}}. (49)

References

  • (1) C. M. Caves, Phys. Rev. D 23, 1693 (1981).
  • (2) B. Yurke, S. L. McCall, and J. R. Klauder, Phys. Rev. A 33, 4033 (1986).
  • (3) M. J. Holland and K. Burnett, Phys. Rev. Lett. 71, 1355 (1993).
  • (4) U. Dorner, R. Demkowicz-Dobrzanski, B. J. Smith, J. S. Lundeen, W. Wasilewski, K. Banaszek, and I. A. Walmsley, Phys. Rev. Lett. 102, 040403 (2009).
  • (5) V. Giovannetti, S. Lloyd, and L. Maccone, Nat. Photon. 5, 222 (2011).
  • (6) B. C. Sanders and G. J. Milburn, Phys. Rev. Lett. 75, 2944 (1995).
  • (7) J. Ma, X. Wang, C. P. Sun, and F. Nori, Phys. Rep. 509, 89 (2011).
  • (8) J. Ma, Y. X. Huang, X. Wang, and C. P. Sun, Phys. Rev. A 84, 022302 (2011).
  • (9) P. C. Humphreys, M. Barbieri, A. Datta, and I. A. Walmsley, Phys. Rev. Lett. 111, 070403 (2013).
  • (10) J. P. Dowling, Contemp. Phys. 49, 125 (2008).
  • (11) A. I. Lvovsky, B. C. Sanders, and W. Tittel, Nat. Photon. 3, 706 (2009)
  • (12) M. Kitagawa and M. Ueda, Phys. Rev. A 47, 5138 (1993).
  • (13) D. J. Wineland, J. J. Bollinger, W. M. Itano, D. J. Heinzen, Phys. Rev. A 50, 67 (1994).
  • (14) K. Mømer and A. Søensen, Phys. Rev. Lett. 82, 1835 (1999).
  • (15) A. Sørensen, L-M. Duan, J. I. Cirac and P. Zoller, Nature 409, 4 (2001).
  • (16) U. V. Poulsen and K. Mølmer, Phys. Rev. A 64, 013616 (2001).
  • (17) G. R. Jin, Y. C. Liu and W. M. Liu, New J. Phys. 11, 073049 (2009).
  • (18) G. R. Jin, Y. An, T. Yan, and Z. S. Lu, Phys. Rev. A 82, 063622 (2010).
  • (19) O. Guehne and G. Tóth, Phys. Rep. 474, 1 (2009).
  • (20) X. Wang and B. C. Sanders, Phys. Rev. A 68, 012101 (2003).
  • (21) Y. C. Liu, Z. F. Xu, G. R. Jin, and L. You, Phys. Rev. Lett. 107, 013601 (2011).
  • (22) H. Strobel, W. Muessel, D. Linnemann, T. Zibold, D. B. Hume, L. Pezzé, A. Smerzi and M. K. Oberthaler, science 345, 424 (2014).
  • (23) T. Macrì, A. Smerzi, and L. Pezzé, Phys. Rev. A 94, 010102 ( 2016).
  • (24) S. D. Bennett, N.Y. Yao, J. Otterbach, P. Zoller, P. Rabl, and M. D. Lukin, Phys. Rev. Lett. 110, 156402 (2013).
  • (25) P. Bouyer and M. A. Kasevich, Phys. Rev. A 56, R1083 (1997).
  • (26) M. A. Kasevich, Science, 298, 1363 (2002).
  • (27) C. Gross, T. Zibold, E. Nicklas, J. Estève and M. K. Oberthaler, Nature (London) 464, 1165 (2010).
  • (28) M. F. Riedel, P. Böhi, Y. Li, T. W. Hansch, A. Sinatra and P. Treutlein, Nature (London) 464, 1170 (2010).
  • (29) S. Palzer, C. Zipkes, C. Sias, and M. Köhl, Phys. Rev. Lett. 103, 150601 (2009).
  • (30) M. A. Cirone, G. De. Chiara, G. M. Palma, and A. Recati, New J. Phys. 11, 103055 (2009).
  • (31) S. Will, T. Best, S. Braun, U. Schneider, and I. Bloch, Phys. Rev. Lett. 106, 115305 (2011).
  • (32) N. Spethmann, F. Kindermann, S. John, C. Weber, D. Meschede, and A. Widera, Phys. Rev. Lett. 109, 235301 (2012).
  • (33) R. Scelle, T. Rentrop, A. Trautmann, T. Schuster, and M. K. Oberthaler, Phys. Rev. Lett. 111, 070401 (2013).
  • (34) C. Zipkes, S. Palzer, C. Sias, and M. Köhl, Nature (London) 464, 388 (2010).
  • (35) S. Schmid, A. Harter, and J. H. Denschlag, Phys. Rev. Lett. 105, 133202 (2010).
  • (36) J. B. Balewski, A. T. Krupp, A. Gaj, D. Peter, H. P. Büchler, R. Löw, S. Hofferberth, and T. Pfau, Nature (London) 502, 664 (2013).
  • (37) A. Recati, P. O. Fedichev, W. Zwerger, J. von Delft, and P. Zoller, Phys. Rev. Lett. 94, 040404 (2005).
  • (38) H. T. Ng and S. Bose, Phys. Rev. A 78, 023610 (2008).
  • (39) N. Bar-Gill, D. D. Bhaktavatsala Rao, and G. Kurizki, Phys. Rev. Lett. 107, 010404 (2011).
  • (40) M. Bruderer and D. Jaksch, New J. Phys. 8, 87 (2006).
  • (41) S. Yi and L. You, Phys. Rev. A 61, 041604(R) (2000); 63, 053607 (2001).
  • (42) K. Goral, L. Santos, and M. Lewenstein, Phys. Rev. Lett. 88, 170406 (2002)
  • (43) S. Yi, L. You, and H. Pu, Phys. Rev. Lett. 93, 040403 (2004).
  • (44) S. Yi, T. Li, and C. P. Sun, Phys. Rev. Lett. 98, 260405 (2007).
  • (45) H.-Y. Lu and S. Yi, Sci China-Phys Mech Astron 55, 1535 (2012).
  • (46) M. Lu, N. Q. Burdick, S. H. Youn, and B. L. Lev, Phys. Rev. Lett. 107, 190401 (2011).
  • (47) K. Aikawa, A. Frisch, M. Mark, S. Baier, A. Rietzler, R. Grimm, and F. Ferlaino, Phys. Rev. Lett. 108, 210401 (2012).
  • (48) S. Giovanazzi, A. Görlitz, and T. Pfau, Phys. Rev. Lett. 89, 130401 (2002).
  • (49) T. Lahaye, C. Menotti, L. Santos, M. Lewenstein, and T. Pfau, Rep. Prog. Phys. 72, 126401 (2009).
  • (50) A. Griesmaier, J. Stuhler, T. Koch, M. Fattori, T. Pfau, and S. Giovanazzi, Phys. Rev. Lett. 97, 250402 (2006).
  • (51) J. B. Yuan, H.-J Xing, L.-M. Kuang, and S. Yi, Phys Rev A 95, 033610 (2017).
  • (52) K. Pawlowski, K. Rzazewski, Phys Rev A 81, 013620 (2010).
  • (53) Y. Hao and Q. Gu, Phys Rev A 83, 043620 (2011).
  • (54) D. Spehner, K. Pawlowski, G. Ferrini, and A. Minguzzi, Eur. Phys. J. B 87, 156 (2014).
  • (55) J. Huang, X. Qin, H. Zhong, Y. Ke and C. Lee, Sci Rep. 5, 17894 (2015).
  • (56) F. Dalfovo, S. Giorgini, L. P. Pitaevskii, and S. Stringari, Rev. Mod. Phys. 71, 463 (1999).
  • (57) H. P. Breuer and F. Petruccione, The Theory of Open Quantum Systems (Oxford University Press, Oxford, 2007).
  • (58) A. I. Streltsov, O. E. Alon, L. S. Cederbaum, Phys Rev A 73, 063626 (2006); O. E. Alon, A. I. Streltsov, L. S. Cederbaum, Phys Rev A 77, 033613 (2008).