Anda di halaman 1dari 4

Maximal entanglement achievable by controlled dynamics

Alessio Serafini1 and Stefano Mancini2


1
Department of Physics & Astronomy, University College London, Gower Street, London WC1E 6BT, United Kingdom
2
Dipartimento di Fisica, Università di Camerino, I-62032 Camerino, Italy
(Dated: October 13, 2009)
We consider the feedback control of quantum systems comprised of any number of bosonic degrees
of freedom. We derive a general upper bound for the logarithmic negativity achievable, at steady
state, with continuous Gaussian measurements on the environment and linear driving on the system.
Our results apply to rotating wave system-bath couplings and to any quadratic system’s Hamiltonian.
Furthermore, we apply this upper bound to parametric processes, show it to be tight, and compare it
to feedback strategies limited to local measurements.
arXiv:0910.2205v1 [quant-ph] 12 Oct 2009

PACS numbers: 03.67.Bg, 02.30.Yy, 42.50.Dv

The field of quantum control is central in the current are affine in phase-space: Ĥ = (1/2)x̂⊤Hx̂ − x̂⊤ ΩBu(t),
rise of quantum technologies [1, 2]. In particular, the where the “Hamiltonian matrix” H is real and symmetric
control of the coherent resources of quantum states is of and B is real. The second term of Ĥ is a ‘linear driving’
crucial practical interest. Most valuable, and delicate, proportional to a time-dependent input u(t): this term
among such resources is certainly quantum entangle- will describe the control exerted over the system.
ment, whose control is a primary requisite for quantum The system is considered to be open and such that
information and communication [3, 4]. each degree of freedom has its own channel to interact
In this paper, we focus on bosonic quantum systems with the environment. Though thermal noise can also
subject to quadratic Hamiltonians and losses, and de- be treated along the lines we will present here, in this
rive bounds on the maximal entanglement achievable, study we specialise for simplicity to pure losses, which
between specific bipartitions, through continuous feed- are the main source of decoherence in quantum optical
back based on general Gaussian measurements and lin- settings. We will thus assume a beam splitter-like (“ro-
ear driving [5]. The class of dynamics and feedback tating wave”) interaction a j b†j + a†j b j between each mode
strategies covered in our study is of great practical rel- and the associated mode of the bath, with ladder op-
evance in quantum optics, and is applicable to more erator b j . Under the conditions set out above, the first
general continuous variable systems. Because entan- moments of the canonical operators evolve according to
glement is not a linear figure of merit in the quantum dhx̂i/dt = Ahx̂i + Bu(t), while the second moments obey
state’s parameters, one cannot address the questions we
are about to tackle with standard optimisation tools, like dσ/dt = Aσ + σA⊤ + 1. (2)
semi-definite programming [6], but rather requires the Here, the “drift matrix” A = (ΩH−1)/2, and 1 is the iden-
more detailed, specific analysis we shall present. tity matrix with dimension clear from the context. We
Notation – We consider systems of N degrees of free- will only address stable systems, for which (A + AT ) < 0.
doms described by pairs of canonical operators: ⊤ defin- Note that, for Gaussian states, these equations describe
ing
h a vector
i of operators x̂ = q̂1 , p̂ 1 , ..., q̂N , p̂ N , one has the complete dynamics of the system.
x̂ j , x̂k = iΩ jk , where Ω is the (2N) × (2N) symplectic As customary in the context of feedback control, we
   
form: Ω jk = δ j+1,k 1 − (−1) j /2 − δ j,k+1 1 + (−1) j /2, in will now assume that the degrees of freedom of the envi-
√ ronment can be continuously monitored on time-scales
terms of Kroneker deltas δ j,k . Also, a j = (x̂ j + ip̂ j )/ 2.
which are short with respect to the system’s response
For a system with such a phase-space structure we
time [8]. Because of the rotating wave interaction be-
can define “Gaussian states” as the states with a Gaus-
tween system and environment, instantaneous Gaussian
sian Wigner function. These states are completely deter-
measurements on the environment degrees of freedom
mined by the vector of means hx̂i, and by the covariance
results into measuring the operators â 1 + â† Υ , where
 
 ⊤
matrix (CM) σ, with entries σ jk = h∆x̂ j ∆x̂k i + h∆x̂ j ∆x̂k i ,
the vector â = (a1 , . . . , aN )⊤ contains all the annihila-
where ∆ô = (ô − hôi) for operator ô. The – always nec-
tion operators of the system, and the complex matrix
essary – Robertson-Schrödinger uncertainty relation is
Υ parametrises the Gaussian measurement. In turn, Υ
also sufficient for Gaussian states to be physical [7]:
defines the so called “unravelling matrix” U, given by
σ + iΩ ≥ 0 . (1) 1 1 + Re [Υ] Im [Υ]
!
U≡ . (3)
2 Im [Υ] 1 − Re [Υ]
We will consider Hamiltonians Ĥ that are at most of the
second-order in x̂, so that their resulting free evolutions The only conditions on Υ are that U be symmetric and
2

positive semi-definite. The outcome of the measure- Lemma 1 (Bound on smallest symplectic eigenvalue)
ments on the environment √ is recorded as a “current” y = The smallest partially transposed symplectic eigenvalue ν̃− of
Chx̂i+ dwdt , where C = 2U 1/2
C̄ and C̄ jk = (δ2 j−1,k +δ2(j−N),k ) a generic CM σ is bounded from below as follows
for j, k ∈ [1, . . . , 2N]. Finally, dw is a vector of real Wiener
increments satisfying dwdw⊤ = 1dt. ν̃2− ≥ λ↑1 λ↑2 , (5)
The conditional evolution of the moments under such
continuous measurements can also be derived by stan- λ↑1 and λ↑2 being the two smallest eigenvalues of σ.
dard techniques (Itô calculus). It amounts to a diffusive
equation with a stochastic component for the first mo- Next, the uncertainty principle entails:
ments hx̂i, and to a deterministic Riccati equation for the
second moments [6]. In our reasonings to follow, we will Lemma 2 (Uncertainty relation for CMs’ eigenvalues)
not make use of the details of such equations directly. We Let { λ↑j } and { λ↓j } be, respectively, the 2N increasingly-
will be interested in stable systems, and will determine ordered and decreasingly-ordered eigenvalues of an N-mode
the maximal entanglement achievable at steady state. CM σ. Then one has:
Hence, all we need to remark is that the set of CMs {σ∞ }
that are stabilising solutions [9] of the Riccati equation for λ↑j λ↓j ≥ 1 for 1 ≤ j ≤ N. (6)
the second moments can be characterised as follows [6]:
As an immediate corollary of Lemma 2, one obtains
Aσ ∞ + σ ∞ AT + 1 ≥ 0 . (4)
1
λ↑1 λ↑2 ≥ . (7)
Together with Inequality (1), this relationship completely λ↓1 λ↓2
determines the set of stabilising solutions of our condi-
tional dynamics. Lemma 3 (Bound on eigenvalues of steady state CMs)
The final ingredient to be added is the possible de- Let σ ∞ be a conditional CM at steady state obtained under
pendence of the linear drive u(t) on the history of the continuous Gaussian measurements, pure losses and a
measurement record y(s) for s < t, which affects both Hamiltonian matrix H. The product of the two largest
first and second moments of the unconditional, ‘average’, eigenvalues λ↓1 and λ↓2 of σ ∞ is bounded as follows:
evolution (whereas second moments of the conditional
states are unaffected by the linear driving), and closes 1
λ↓1 λ↓2 ≤ , (8)
the control loop. We will denote the unconditional state α↑1 α↑2
by ̺. Note that, for our class of dynamics, ̺ is a statisti-
cal mixture of states with the same conditional CM σ ∞ , where {α↑j } are the (strictly positive) eigenvalues of (−A − A⊤)
obeying Inequality (4), and varying first moments. For
Gaussian states, this implies that ̺ can be obtained from in increasing order, and A = 12 (ΩH − 1).
a Gaussian state ̺0 with CM σ ∞ and vanishing first mo-
The chain of Inequalities (5), (7) and (8) leads to
ments by local operations and classical communication
alone: ̺ = L(̺0 ), where L is some LOCC map.
ν̃2− ≥ α↑1 α↑2 , (9)
The typical aim of control over some time interval is
to minimise the expected value of a cost function [1, 9]. which constrains the maximal logarithmic negativity
Our cost function will be the entanglement of Gaussian achievable for states conditioned by Gaussian measure-
multi-mode steady states for bipartitions of 1 versus ments. Further, and more generally, one has:
(N − 1) modes and ‘bi-symmetric’ bipartitions (i.e., in-
variant under the permutation of local modes). Such an Proposition 1 (Maximal unconditional entanglement)
entanglement can be quantified by the logarithmic neg- The logarithmic negativity EN (̺) of any 1 versus (N − 1)
ativity EN , which is in turn determined, for a Gaussian modes or bi-symmetric bipartition of an unconditional steady
state with CM σ, as − log2 ν̃− , where ν̃2− is the smallest state achievable by continuous Gaussian feedback and linear
eigenvalue of (σ Ω̃σ Ω̃T ), being Ω̃ the (skew-symmetric) driving is bounded by:
partial transposition of Ω [10, 11]. Clearly, ν̃− is not a
quadratic cost function (i.e., it is not linear in σ). This 1
 
EN (̺) ≤ max 0, − log2 (α↑1 α↑2 ) . (10)
is why, albeit dealing with linear systems with Gaussian 2
noise, we cannot resort to optimisation methods mutu-
ated from classical LQG control problems [6]. Applications – Clearly, the eigenvalues {α↑j } can be analyt-
General results – Hereafter, we present our main findings ically or numerically determined for general quadratic
as three lemmas leading to a final proposition. Proofs of Hamiltonians, describing a wide variety of systems of
these statements may be found in appendix. practical interest (bosonic atoms, trapped ions, nano-
Our investigation starts off from a corollary of Ref. [12]: mechanical resonators and Josephson junctions, just to
3

mention a few). Here, we will focus on yet another inter- assumed for the unravelling matrix U. This applies to
esting case where our results apply and can be handled situations where the environments of the two local sub-
analytically: the case of parametric interactions, which systems can be combined before being measured (like,
are the state of the art technology to generate continuous e.g., for a parametric crystal in a cavity). We intend now
variable entanglement in quantum optics. to provide a lower bound on the entanglement achiev-
The parametric interaction between modes j and k is able under local control, where the environmental de-
described by the Hamiltonian χ(x̂ j p̂k + p̂ j x̂k ) [13]. We will grees of freedom pertaining to the separate subsystems
assume equal interaction strengths χ between any pair of cannot be combined, and compare it to the upper bound
modes, consider a (m+n)-mode bipartition, and describe we obtained above. To this end, we will adopt direct
analytically the scaling of the control of the entanglement (Markovian) feedback [5] and set u(t) = Fy(t). The un-
with the number of modes m and n (we have m + n = N). conditional evolution of the system is then described by
In point of fact, we will see that our bounds in this
case are tight, and yield the actual optimal entanglement dσ/dt = A′ σ + σA′T + D′ , (13)
achievable by Gaussian filtering. Due to the symmetry
with drift and diffusion matrices modified as A′ =
of the system under the exchange of any two modes, the
Ā + BFC and D′ = 1 − CT FT BT − BFC + 2BFFT BT . We
entanglement between the m- and the n-modes subsys-
also choose a specific form of U and BF. Since in the free
tems can be reduced to two-mode entanglement [14]: a
dynamics, governed by the drift matrix of Eq. (11), the
local symplectic transformation exists that turns the ma-
quadratures p1 and p2 are less noisy than q1 and q2 , it
trix A into an equivalent two-mode drift matrix Ā, plus
is advantageous to measure locally p1 and p2 and drive
a direct sum of decoupled single-mode matrices that are
with the respective currents the quadratures q2 and q1 .
irrelevant for the entanglement. The matrix Ā reads:
However, due to the possible asymmetry of the two sub-


 (m − 1)χ 0 mnχ
 systems for m , n, we have to consider different driving
√0 
0 −(m − 1)χ 0 − mnχ  1 amplitudes µ1 and µ2 for their quadratures. All this
 
Ā =  √

 − . √
mnχ 0 (n − 1)χ 0  2 corresponds
√ to setting U 33 = U 44 = 1, 2(BF)24 = µ2 ,
 √
0 − mnχ 0 −(n − 1)χ 2(BF)43 = µ1 , and all other entries of U and BF vanish-
 
(11) ing. We can then find the steady state solution of Eq. (13)
1 as a function of the two feedback amplitudes µ1 and µ2 ,
For the system to be stable one must require: χ < 2(N−1) .
As A is symmetric and invertible, the ‘free’ steady and evaluate its logarithmic negativity. It turns out that
state CM σ f can be promptly determined from Eq. (2): the maximum logarithmic negativity at steady state is
σ f = −A−1 /2. Its smallest partially transposed sym- attained for µ2 = µ1 n/m. Hence, we are left with the en-
plectic eigenvalue, which determines the asymptotic tanglement depending on one parameter, over which we
entanglement in the absence of control, is given by minimise numerically in the stable region, determined
1/[(1 + 2χ)(1 + 2(N − 1)χ)]. Instead, the bound of In- by (A′ + A′T ) < 0. As a case of study, we have considered
equality (9) for any steady state CM σ ∞ with Gaussian a system of 6 modes and summarised the results in Fig. 1.
feedback control reads Local control is very close to optimal global control in the
case of a balanced bipartition. However, the more un-
ν̃2− ≥ (1 − 2χ) [1 − 2(N − 1)χ] . (12) balanced the bipartition, the more degraded the control,
although numerics indicate that infinite entanglement
This lower bound is attained by the CM
can always be retrieved close to instability.
RT diag(α4 , α3 , α1 , α2 )R, where R is the orthogonal
Conclusion – We have derived bounds on the entangle-
transformation that diagonalises Ā and {α j } are the
ment achievable, at steady state and for various biparti-
eigenvalues of −Ā in increasing order. This solution
tions, in multimode linear bosonic systems subjected to
also saturates the Inequalities (4) and (1). Both the free
continuous feedback control. Our investigation not only
asymptotic entanglement and the optimal one under
applies to quantum information processing and state en-
Gaussian filtering have thus been obtained analyti-
gineering, as shown, but also yields a technique for op-
cally. Quite remarkably, in the symmetric situation
timisation problems lying beyond the LQG scenario.
considered, none of these quantities depends on the
Appendix – Proofs of mathematical statements. Henceforth,
chosen bipartition, but only on the total number of
|vi will stand for a unit vector in the phase space Γ and
modes involved: the same amount of entanglement can
hv| will be its dual under the Euclidean scalar product.
be obtained regardless of the bipartition considered.
Also, note that the uncertainty relation (1) is equivalent
Note also that the ‘optimal unravelling’, that is the
to the two following conditions [12, 15]:
matrix U granting maximal entanglement, may also be
straightforwardly derived from the optimal state [6]. σ 1/2 ΩT σΩσ 1/2 ≥ 1 , and σ > 0 . (14)
Local control – Such an optimal entanglement is in gen-
eral achieved by filtering the system through global mea- Proof of Lemma 1. The squared symplectic eigenvalue
surements on the environment, as no restrictions were ν̃2− is the smallest eigenvalue of the matrix σ 1/2 Ω̃T σ Ω̃σ 1/2 :
4

infhw|vi=0 hw|K|wihv|K|vi = λ↑1 λ↑2 , where λ↑j is the j-th small-


1.0 est eigenvalue of the positive matrix K.

ν2−
0.8

0.6

Proof of Proposition 1. As seen previously, ̺ = L(̺0 ),


0.4
where L is a LOCC operation and ̺0 a Gaussian state with
0.2 a CM which is a stabilisinghsolution of (2). Thusi EN (̺) =
↑ ↑
EN (L(̺0 )) ≤ EN (̺0 ) ≤ max 0, − log2 (α1 α2 )/2 , where (9),
0.0
the formula EN = − log(ν̃− ) (holding for 1 − (N − 1) and
0.00 0.02 0.04 0.06 0.08
χ 0.10
bi-symmetric bipartitions), and the monotonicity of EN
FIG. 1: Squared symplectic eigenvalue ν̃2− at steady state for under LOCC [17] have been invoked.
a system of 6 modes (ν̃− → 0 implies infinite entanglement).
Green (lighter) curves depict ν̃− in the absence of control (from
top to bottom: 1:5, 2:4, and 3:3 modes bipartition); bleu (darker)
curves refer to numerically optimised local feedback (from top
to bottom: 1:5, 2:4, and 3:3 modes bipartition); the red curve is We acknowledge financial support from the EU through the
the analytical lower bound (12) achievable by global control. FET-Open Project HIP (FP7-ICT-221899).

ν̃2− = min|vi hv|σ1/2 Ω̃T σ Ω̃σ 1/2 |vi. For √


each |vi, one can
define the unit vector |wi = Ω̃σ 1/2 |vi/ hv|σ|vi, such that
hv|σ1/2 |wi = 0 (due to the antisymmetry of Ω̃) and [1] H. M. Wiseman and G. J. Milburn, Quantum measurement
and control (Cambridge University Press, in press).
[2] W. P. Smith et al., Phys. Rev. Lett. 89, 133601, (2002); M. A.
ν̃2− = minhv|σ|vihw|σ|wi ≥ min hv|σ|vihw|σ|wi = λ↑1 λ↑2 .
|vi |vi,|wi Armen et al., ibid. 89, 133602 (2002); S. Chaudhury et al.,
ibid. 99, 163002 (2007); R. L. Cook, P. J. Martin and J. M.
The last equality is easily verified once hv|σ1/2 |wi = 0 and Geremia, Nature 446, 774 (2007); B. L. Higgins et. al., ibid.
σ > 0 are enforced, and completes the proof. 450, 393 (2007); B. L. Higgins et al., arXiv:0909.1572.
[3] S. Mancini and J. Wang, Eur. Phys. J. D 32, 257 (2005); S.
Proof of Lemma 2. Once √ again, for any |vi ∈ Γ one can
Mancini, Phys. Rev. A 73, 010304(R) (2006).
define |wi = Ωσ 1/2 |vi/ hv|σ|vi, so that the Robertson [4] S. Mancini and H. M. Wiseman, Phys. Rev. A 75, 012330
Schrödinger Inequality (14) can be recast as (2007).
[5] H. M. Wiseman and G. J. Milburn, Phys. Rev. Lett. 70, 548
(1993); Phys. Rev. A 49, 1350 (1994).
hv|σ|vihw|σ|wi ≥ 1 ∀ |vi ∈ Γ.
[6] H. M. Wiseman and A. C. Doherty, Phys. Rev. Lett. 94,
070405 (2005).
We will now denote by |v j i the eigenvectors corre- [7] A. S. Holevo, IEEE Trans. Inf. Theor. IT21 533 (1975); R.
sponding to the increasingly ordered eigenvalues of σ: Simon, N. Mukunda, and B. Dutta, Phys. Rev. A 49, 1567
σ|v j i = λ↑j |v j i. Let us consider a vector |vi belonging to (1994).
the subspace, which we shall denote Γk , spanned by the [8] H. M. Wiseman and L. Diósi, Chem. Phys. 268, 91 (2001).
[9] K. Zhou, J. C. Doyle, and K. Glover, Robust and Optimal
k smallest eigenvectors of σ {|v ji}, for j ≤ k. Clearly one
Control (Prentice-Hall, New Jersey, 1996).
has hv|σ|vi ≤ λ↑k . The inequality above then leads to [10] Partial transposition on a (m + n)-mode bipartition is de-
scribed in phase space by T = 1 ⊕ σ⊕n z , where the identity
λ↑k hw|σ|wi ≥ hv|σ|vihw|σ|wi ≥ 1 ∀ |vi ∈ Γk , act on the first m modes while the direct sum of reflections
σz acts on the other n modes. One then has Ω̃ = TΩT.
which must be satisfied by all the vectors |wi belonging [11] For such bipartitions, only one partially transposed sym-
plectic eigenvalue contributes to the logarithmic negativ-
to the k-dimensional linear subspace ΩΓk (defined as the
ity. See A. Serafini, Phys. Rev. Lett. 96, 110402 (2006).
subspace spanned by the k orthogonal vectors Ω|vk i): [12] M. M. Wolf, J. Eisert, and M. B. Plenio, Phys. Rev. Lett. 90,
047904 (2003).
λ↑k hw|σ|wi ≥ 1 ∀|wi ∈ ΩΓk . [13] M. D. Reid, Phys. Rev. A 4440, 913 (1989).
[14] A. Serafini, G. Adesso, and F. Illuminati, Phys. Rev. A 71,
By Poincaré Inequality [16], a vector |wi must exist in 032349 (2005).
ΩΓk for which hw|σ|wi ≤ λ↓k , such that λ↑k λ↓k ≥ 1. [15] S. Pirandola et al., Phys. Rev. A 79, 052327 (2009).
[16] ∀ k-dimensional subspace Σk and hermitian M, ∃ |vi ∈ Σk
Proof of Lemma 3. This statement is a consequence such that hv|vi = 1 and hv|M|vi ≤ λ↓k (M). See, e.g., R. Bathia,
of Inequality (4) applied to the two eigenvectors cor- Matrix Analysis (Springer, New York, 1996), page 58.
responding to λ↓1 and λ↓2 , and of the relationship: [17] M. B. Plenio, Phys. Rev. Lett. 95, 090503 (2005).

Anda mungkin juga menyukai