08.04.2013 Views

Quantum Information Theory with Gaussian Systems

Quantum Information Theory with Gaussian Systems

Quantum Information Theory with Gaussian Systems

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

3 Optimal cloners for coherent states<br />

If one of the states is pure, as in our case, this expression reduces to f(ρ1, ρ2) =<br />

tr[ρ1 ρ2]. We employ this functional to quantify the quality of the clones <strong>with</strong> respect<br />

to the input state.<br />

For example, we could require that the overlap between the joint output T∗(ρ) of<br />

the cloner and a tensor product of n perfect copies of the input state ρ becomes as<br />

large as possible. This is accomplished by maximizing the joint fidelity<br />

fjoint(T, ρ) = tr T∗(ρ)ρ ⊗n . (3.1)<br />

Since this criterion compares the complete output state, including correlations between<br />

subsystems, <strong>with</strong> a tensor product state, which is noncorrelated, it might be<br />

too strong. Instead, one could measure the quality of individual clones by comparing<br />

a single output subsystem, e.g. the i-th, to the input state <strong>with</strong> an appropriate<br />

fidelity expression:<br />

fi(T, ρ) = tr T∗(ρ)( ⊗ · · · ⊗ ⊗ ρ (i) ⊗ · · · ⊗ ) , (3.2)<br />

where the upper index (i) indicates the position in the tensor product. However, a<br />

single suchone-clone-onlyfidelity could be trivially put to one by a cloner which<br />

does essentially nothing, but merely returns the input state in the i-th subsystem of<br />

the output and yields a suitable fixed stated for the other subsystems. So, optimizing<br />

the fidelities for all i in sequence would result in different cloners for each fidelity. To<br />

avoid this, we optimize over a weighted sum of such fidelities, n i=1 λi fi(T, ρ) <strong>with</strong><br />

positive weights λi. The relative weights determine which clones should resemble the<br />

input state more closely and thus allow to describe nonsymmetric cloners.<br />

For a similar reason it is not useful to optimize the cloner for each input state<br />

separately, because that would yield a source which perfectly produces the respective<br />

quantum state. Instead, we can either consider the average or the worst-case<br />

quality <strong>with</strong> respect to an ensemble of input states. However, both approaches face<br />

conceptual difficulties. In the first case, the process of averaging over the pure <strong>Gaussian</strong><br />

states is not well defined, because this amounts to averaging over the group<br />

Sp(2n,Ê) of symplectic transformations, which is noncompact. In the latter case,<br />

for every given cloner squeezed states exist which for sufficiently large squeezing<br />

bring the fidelity arbitrarily close to zero (see end of Section 3.4.1). While this can<br />

in principle be compensated for a fixed and known squeezing by a modified cloner<br />

(desqueeze, clone unsqueezed state and resqueeze output), it is not possible to circumvent<br />

this behavior for arbitrary, unknown squeezing. We address the problem by<br />

optimizing the cloner only for coherent states, which constitute a subset of all pure<br />

<strong>Gaussian</strong> states.<br />

As the figure of merit, we choose the worst-case fidelities fjoint(T) and fi(T),<br />

defined as the infima of (3.1) and (3.2) over the set coh = |ξ〉〈ξ| <br />

ξ ∈ Ξin of all<br />

coherent states,<br />

34<br />

fjoint(T) = inf<br />

ρ∈coh fjoint(T, ρ) and fi(T) = inf<br />

ρ∈coh fi(T, ρ).

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

Saved successfully!

Ooh no, something went wrong!