30.04.2013 Views

ENTANGLEMENT OF GAUSSIAN STATES Gerardo Adesso

ENTANGLEMENT OF GAUSSIAN STATES Gerardo Adesso

ENTANGLEMENT OF GAUSSIAN STATES Gerardo Adesso

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

13.2. Entanglement distribution in Gaussian valence bond states 225<br />

13.2.1. Short-range correlations<br />

Let us consider a building block γ with s = smin ≡ (x + 1)/2. It is straightforward<br />

to evaluate, as a function of x, the GVBS in Eq. (13.3) for an arbitrary number<br />

of modes (we omit the CM here, as no particular insight can be drawn from the<br />

explicit expressions of the covariances). By repeatedly applying the PPT criterion,<br />

one can analytically check that each reduced two-mode block γout i,j is separable for<br />

|i−j| > 1, which means that the output GVBS Γ out exhibits bipartite entanglement<br />

only between nearest neighbor modes, for any value of x > 1 (for x = 1 we obtain<br />

a product state).<br />

While this certainly entails that Γ out is genuinely multiparty entangled, due to<br />

the translational invariance, it is interesting to observe that, without feeding entanglement<br />

in the input port γss of the original building block, the range of quantum<br />

correlations in the output GVBS is minimum. The pairwise entanglement between<br />

nearest neighbors will naturally decrease with increasing number of modes, being<br />

frustrated by the overall symmetry and by the intrinsic limitations on entanglement<br />

sharing (the so-called monogamy constraints [GA10], see Chapter 6). We can<br />

study the asymptotic scaling of this entanglement in the limit x → ∞. One finds<br />

that the corresponding partially-transposed symplectic eigenvalue ˜νi,i+1 is equal to<br />

(N − 2)/N for even N, and [(N − 2)/N] 1/2 for odd N: neighboring sites are thus<br />

considerably more entangled if the ring size is even-numbered. Such frustration<br />

effect on entanglement in odd-sized rings, already devised in a similar context in<br />

Ref. [272], is quite puzzling. An explanation may follow from counting arguments<br />

applied to the number of parameters (which are related to the degree of pairwise<br />

entanglement) characterizing a block-diagonal pure state on harmonic lattices (see<br />

Sec. 11.2.1), as we will now show.<br />

13.2.1.1. Valence bond representability and entanglement frustration. Let us make a<br />

brief digression. It is conjectured that all pure N-mode Gaussian states can be<br />

described as GVBS [202]. Here we provide a lower bound on the number M of<br />

ancillary bonds required to accomplish this task, as a function of N. We restrict<br />

to ground states of harmonic chains with spring-like interactions, i.e. to the blockdiagonal<br />

Gaussian states of Sec. 11.2.1, which have been proven to rely on N(N −<br />

1)/2 parameters [GA14], and which are GVBS with a CM of the form Eq. (13.4).<br />

With a simple counting argument using Eq. (11.1), the total number of parameters<br />

of the initial chain Γ of building blocks should be at least equal to that of the<br />

target state, i.e.<br />

N(2M + 1)(2M)/2 ≥ N(N − 1)/2 ,<br />

which means M ≥ IntPart[( √ 4N − 3 − 1)/4]. This implies, for instance, that to<br />

describe general pure states with at least N > 7 modes, a single EPR bond per<br />

site is no more enough (even though the simplest case of M = 1 yields interesting<br />

families of N-mode GVBS for any N, as shown in the following).<br />

The minimum M scales as N 1/2 , diverging in the field limit N → ∞. As infinitely<br />

many bonds would be necessary (and maybe not even sufficient) to describe<br />

general infinite harmonic chains, the valence bond formalism is probably not helpful<br />

to prove or disprove statements related to the entropic area scaling law [187] for

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!