POVM

In functional analysis and quantum measurement theory, a POVM (Positive Operator Valued Measure) is a measure whose values are non-negative self-adjoint operators on a Hilbert space. It is the most general formulation of a measurement in the theory of quantum physics. The need for the POVM formalism arises from the fact that projective measurements on a larger system will act on a subsystem in ways that cannot be described by projective measurement on the subsystem alone. They are used in the field of quantum information. Quantum t-designs have been recently introduced to POVMs and SIC-POVMs as a means of providing a simple and elegant formulation of the field in a general setting, since a SIC-POVM is a type of spherical t-design.[1]

In rough analogy, a POVM is to a projective measurement what a density matrix is to a pure state. Density matrices can describe part of a larger system that is in a pure state (see purification of quantum state); analogously, POVMs on a physical system can describe the effect of a projective measurement performed on a larger system.

Contents

Definition

In the simplest case, a POVM is a set of Hermitian positive semidefinite operators \{F_i\} on a Hilbert space  \mathcal{H} that sum to unity,

\sum_{i=1}^n F_i = \operatorname{I}_H.

This formula is similar to the decomposition of a Hilbert space by a set of orthogonal projectors:

\sum_{i=1}^N E_i = \operatorname{I}_H, \quad E_i E_j = \delta_{i j} E_i.

An important difference is that the elements of a POVM are not necessarily orthogonal, with the consequence that the number of elements in the POVM, n, can be larger than the dimension, N, of the Hilbert space they act in.

In general, POVMs can be defined in situations where outcomes can occur in a non-discrete space. The relevant fact is that measurements determine probability measures on the outcome space:

Definition. Let (X, M) be measurable space; that is M is a σ-algebra of subsets of X. A POVM is a function F defined on M whose values are bounded non-negative self-adjoint operators on a Hilbert space H such that F(X) = IH and for every ξ \in H,

 E \mapsto \langle F(E) \xi \mid \xi \rangle

is a non-negative countably additive measure on the σ-algebra M.

This definition should be contrasted with that for the projection-valued measure, which is very similar, except that, in the projection-valued measure, the F are required to be projection operators.

POVMs and measurement

As in the theory of projective measurement, the probability the outcome associated with measurement of operator F_i occurs is

 P(i)={\rm tr}(\rho F_i),\;

where \rho is the density matrix of the measured system.

Such a measurement can be carried out by doing a projective measurement in a larger Hilbert space. Let us extend the Hilbert space H_A to H_A \oplus H^\perp_A and perform the measurement defined by the projection operators \{\hat{\pi}_i\}. The probability of the outcome associated with \hat{\pi}_i is

 P(i)={\rm tr}(\rho \hat{\pi}_i)={\rm tr}(\rho \hat{\pi}_A \hat{\pi}_i \hat{\pi}_A),\;

where \hat{\pi}_A is the orthogonal projection taking H_A \oplus H^\perp_A to H_A. In the original Hilbert space H_A, this is a POVM with operators given by F_i=\hat{\pi}_A \hat{\pi}_i \hat{\pi}_A. Neumark's dilation theorem guarantees that any POVM can be implemented in this manner.

In practice, POVMs are usually performed by coupling the original system to an ancilla. For an ancilla prepared in a pure state |0\rangle_B, this is a special case of the above; the Hilbert space is extended by the states |\phi\rangle_A \otimes |\psi\rangle_B where \langle\psi|0\rangle_B=0.

Post-measurement state

Consider the case where the ancilla is initially a pure state |0\rangle_B. We entangle the ancilla with the system, taking

 |\psi\rangle_A |0\rangle_B \rightarrow \sum_i M_i |\psi\rangle_A |i\rangle_B,

and perform a projective measurement on the ancilla in the \{|i\rangle_B\} basis. The operators of the resulting POVM are given by

 F_i = M_i ^\dagger M_i .

Since the M_i are not required to be positive, there are an infinite number of solutions to this equation. This means that there are infinite different experimental apparatuses that give the same probabilities for the outcomes. Since the post-measurement state of the system

\rho' = {M_i \rho M_i^\dagger \over {\rm tr}(M_i \rho M_i^\dagger)}

depends on the M_i, in general it cannot be inferred from the POVM alone.

Another difference from the projective measurements is that a POVM is not repeatable. If \rho' is subjected to the same measurement, the new state is

\rho'' = {M_i \rho' M_i^\dagger \over {\rm tr}(M_i \rho' M_i^\dagger)} = {M_i M_i \rho M_i^\dagger M_i^\dagger \over {\rm tr}(M_i M_i \rho M_i^\dagger M_i^\dagger)}

which is equal to \rho' iff M_i^2 = M_i, that is, if the POVM reduces to a projective measurement.

This gives rises to many interesting effects, amongst them the quantum anti-Zeno effect.

Neumark's dilation theorem

Note: An alternate spelling of this is "Naimark's Theorem"

Neumark's dilation theorem is the classification result for POVM's. It states that a POVM can be "lifted" by an operator map of the form V*(·)V to a projection-valued measure. In the physical context, this means that measuring a POVM consisting of a set of n > N rank-one operators acting on a N-dimensional Hilbert space can always be achieved by performing a projective measurement on a Hilbert space of dimension n.

Quantum properties of measurements

A recent work[2] shows that the properties of a measurement are not revealed by the POVM element corresponding to the measurement, but by its pre-measurement state. This one is the main tool of the retrodictive approach of quantum physics in which we make predictions about state preparations leading to a measurement result. We show,[2][3] that this state simply corresponds to the normalized POVM element:


\hat{\rho}_{retr}^{[n]}=\frac{\hat{\Pi}_{n}}{\mathrm{Tr}\lbrace\hat{\Pi}_{n}\rbrace}.

We can make predictions about preparations leading to the result 'n' by using an expression similar to Born's rule:


\mathrm{Pr}\left(m\vert n\right)=\mathrm{Tr}\lbrace\hat{\rho}_{retr}^{[n]}\hat{\Theta}_{m}\rbrace,

in which \hat{\Theta}_{m} is a hermitian and positive operator corresponding to a proposition about the state of the measured system just after its preparation in some a state \hat{\rho}_{m}.[2] Such an approach allows us to determine in which kind of states the system was prepared for leading to the result 'n'.

Thus, the non-classicality of a measurement corresponds to the non-classicality of its pre-measurement state, for which such a notion can be measured by different signatures of non-classicality. The projective character of a measurement can be measured by its projectivity \pi_{n} which is the purity of its pre-measurement state:


\pi_{n}=\mathrm{Tr}\left[\left(\hat{\rho}_{retr}^{[n]}\right)^{2}\right].

The measurement is projective when its pre-measurement state is a pure quantum state \vert\psi_{n}\rangle (\pi_{n}=1). Thus, the corresponding POVM element is given by:


\hat{\Pi}_{n}=\eta_{n}\vert\psi_{n}\rangle\langle\psi_{n}\vert,

where \eta_{n}=\mathrm{Tr}\lbrace\Pi_{n}\rbrace is in fact the detection efficiency of the state \vert\psi_{n}\rangle, since Born's rule leads to \mathrm{Pr}\left(n\vert\psi_{n}\right)=\eta_{n}. Therefore, the measurement can be projective but non-ideal, which is an important distinction with the usual definition of projective measurements.

An example: Unambiguous quantum state discrimination

The task of unambiguous quantum state discrimination (UQSD) is to discern conclusively which state, of given set of pure states, a quantum system (which we call the input) is in. The impossibility of perfectly discriminating between a set of non-orthogonal states is the basis for quantum information protocols such as quantum cryptography, quantum coin-flipping, and quantum money. This example will show that a POVM has a higher success probability for performing UQSD than any possible projective measurement.

First let us consider a trivial case. Take a set that consists of two orthogonal states  |\psi\rang  and   |\psi^T\rang . A projective measurement of the form,

\hat{A}= a|\psi^T\rang\lang\psi^T| %2B b|\psi\rang\lang\psi|,

will result in eigenvalue a only when the system is in   |\psi^T\rang and eigenvalue b only when the system is in   |\psi\rang . In addition, the measurement always discriminates between the two states (i.e. with 100% probability). This latter ability is unnecessary for UQSD and, in fact, is impossible for anything but orthogonal states. Now consider a set that consists of two states  |\psi\rang  and   |\phi\rang in two-dimensional Hilbert space that are not orthogonal. i.e.,

 |\lang\phi|\psi\rang| = cos(\theta),

for  \theta > 0. These could be states of a system such as the spin of spin-1/2 particle (e.g. an electron), or the polarization of a photon. Assuming that the system has an equal likelihood of being in each of these two states, the best strategy for UQSD using only projective measurement is to perform each of the following measurements,

\hat{\pi}_{\psi^T}= |\psi^T\rang\lang\psi^T|,
\hat{\pi}_{\phi^T}= |\phi^T\rang\lang\phi^T|,

50% of the time. If \hat{\pi}_{\phi^T} is measured and results in an eigenvalue of 1, than it is certain that the state must have been in  |\psi\rang  . However, an eigenvalue of zero is now an inconclusive result since this can come about from the system could being in either of the two states in the set. Similarly, a result of 1 for \hat{\pi}_{\psi^T} indicates conclusively that the system is in   |\phi\rang and 0 is inconclusive. The probability that this strategy returns a conclusive result is,

 P_{proj}=\frac{1-|\lang\phi|\psi\rang|^2}{2}.

In contrast, a strategy based on POVMs has a greater probability of success given by,

 P_{POVM}=1-|\lang\phi|\psi\rang|.

This is the minimum allowed by the rules of quantum indeterminacy and the uncertainty principle. This strategy is based on a POVM consisting of,

\hat{F}_{\psi}=\frac{1-|\phi\rang\lang\phi|}{1%2B|\lang\phi|\psi\rang|}
\hat{F}_{\phi}=\frac{1-|\psi\rang\lang\psi|}{1%2B|\lang\phi|\psi\rang|}
\hat{F}_{inconcl.}= 1-\hat{F}_{\psi}-\hat{F}_{\phi},

where the result associated with  \hat{F}_{i} indicates the system is in state i with certainty.

These POVMs can be created by extending the two-dimensional Hilbert space. This can be visualized as follows: The two states fall in the x-y plane with an angle of θ between them and the space is extended in the z-direction. (The total space is the direct sum of spaces defined by the z-direction and the x-y plane.) The measurement first unitarily rotates the states towards the z-axis so that  |\psi\rang  has no component along the y-direction and  |\phi\rang  has no component along the x-direction. At this point, the three elements of the POVM correspond to projective measurements along x-direction, y-direction and z-direction, respectively.

For a specific example, take a stream of photons, each of which are polarized along either the horizontal direction or at 45 degrees. On average there are equal numbers of horizontal and 45 degree photons. The projective strategy corresponds to passing the photons through a polarizer in either the vertical direction or -45 degree direction. If the photon passes through the vertical polarizer it must have been at 45 degrees and vice versa. The success probability is (1-1/2)/2=25\%. The POVM strategy for this example is more complicated and requires another optical mode (known as an ancilla). It has a success probability of 1-1/\sqrt{2}=29.3\%.

See also

References

  1. ^ http://arxiv.org/abs/quant-ph/0310075
  2. ^ a b c Taoufik Amri, Quantum behavior of measurement apparatus, arXiv:1001.3032 (2010).
  3. ^ S. M. Barnett et al. arXiv:0106139 (2001).