text
stringlengths 0
44.4k
|
---|
For later reference let us note that the matrix Jr, with matrix elements |
(Jr)st=Jrst (28) |
is the adjoint representative of Π rin the SIC-projector basis: |
adΠrΠs= [Πr,Πs] =d2/summationdisplay |
t=1JrstΠt (29) |
It can be seen that all the interesting features of the tensor Grs(respectively, |
the tensors Trst,JrstandRrst) are contained in the order-2 angle tensor θrs(re- |
spectively, the order-3 angle tensor θrst). It is also easy to see that, for any unitary |
U, the transformation |
Πr→UΠrU†(30) |
leaves the angle tensors invariant. This suggests that we shift our focus from indi- |
vidual SIC-POVMs to families of unitarily equivalent SIC-POVMs—SIC- families, |
as we will call them for short. |
We begin our investigation in Section 2by giving necessary and sufficient con- |
ditions for an arbitrary tensor θrs(respectively θrst) to be the rank-2 (respectively |
rank-3) angle tensor corresponding to a SIC-family. We also show t hat either angle |
tensor uniquely determines the corresponding SIC-family. Finally we describe a6 |
method for reconstructing the SIC-family, starting from a knowle dge of either of |
the two angle tensors. |
In Sections 3,4and5we prove the central result of this paper: namely, that |
the existence of a SIC-POVM in dimension dis equivalent to the existence of a |
certain very special set of matrices in the adjoint representation of gl(d,C). In |
Section3we show that, for any SIC-POVM, the adjoint matrices Jrhave the |
spectral decomposition |
Jr=Qr−QT |
r (31) |
whereQris a rankd−1 projector which has the remarkable property of being |
orthogonal to its own transpose: |
QrQT |
r= 0 (32) |
We refer to this feature of the adjoint matrices as the Q-QTproperty. In Section 3 |
we also show that from a knowledge of the Jmatrices it is possible to reconstruct |
the corresponding SIC-family. In Section 4we characterize the general class of |
projectors which have the property of being orthogonal to their own transpose. |
Then, in Section 5, we prove a converse of the result established in Section 3. The |
Q-QTproperty is not completely equivalent to the property of being a SIC set. |
However, it turns out that it is, in a certain sense, very nearly equiv alent. To be |
more specific: let Lrbe any set of d2Hermitian operators which constitute a basis |
for gl(d,C) and letCrbe the adjoint representative of Lrin this basis. Then the |
necessary and sufficient condition for the Crto have the spectral decomposition |
Cr=Qr−QT |
r (33) |
whereQris a rankd−1 projector such that QrQT |
r= 0 is that there exists a |
SIC set Π rsuch thatLr=ǫr(Πr+αI) for some fixed number α∈Rand signs |
ǫr=±1. In particular, the existence of an Hermitian basis for gl( d,C) having the |
Q-QTproperty is both necesary and sufficient for the existence of a SIC -POVM in |
dimensiond. |
In Section 6we digress briefly, and consider sl( d,C) (the Lie algebra consisting |
of all trace-zero d×dcomplex matrices). As we have explained, this paper is |
motivated by the hope that a Lie algebraic perspective will cast light o n the SIC- |
existence problem, rather than by an interest in Lie algebras as suc h. We focus on |
gl(d,C) because that is the casewherethe connection with SIC-POVMsse ems most |
straightforward. However a SIC-POVM also gives rise to an interes ting geometrical |
structure in sl( d,C), as we show in Section 6. |
In Section 7we derive a number of additional identities satisfied by the Jand |
Qmatrices. |
The complex projectors Qr,QT |
rand the real projector Qr+QT |
rdefine three |
families of subspaces. It turns out that there are some interestin g geometrical |
relationships between these subspaces, which we study in Section 8. |
Finally, in Section 9we show that, with the appropriate choice of gauge, the |
Gram matrix corresponding to a Weyl-Heisenberg covariant SIC-fa mily has a fea- |
ture analogous to the Q-QTproperty, which we call the P-PTproperty. It is an |
open question whether this result generalizes to other SIC-families , not covariant |
with respect to the Weyl-Heisenberg group.7 |
2.The Angle Tensors |
The purpose of this section is to establish the necessary and sufficie nt conditions |
for an arbitrary tensor θrs(respectively θrst) to be the order-2 (respectively order- |
3) angle tensor for a SIC-family. We will also show that either one of t he angle |
tensors is enough to uniquely determine the SIC-family. Moreover, we will describe |
explicit procedures for reconstructing the family, starting from a knowledge of one |
of the angle tensors. |
We begin by considering the general class of POVMs (not just SIC-P OVMs) |
which consist of d2rank-1 elements. A POVM of this type is thus defined by a set |
ofd2vectors|ξ1/an}bracketri}ht,...,|ξd2/an}bracketri}htwith the property |
d2/summationdisplay |
r=1|ξr/an}bracketri}ht/an}bracketle{tξr|=I (34) |
Note that/summationtextd2 |
r=1/vextenddouble/vextenddouble|ξr/an}bracketri}ht/vextenddouble/vextenddouble2=d, so the vectors |ξr/an}bracketri}htcannot all be normalized. In the |
particular case of a SIC-POVM the vectors all have the same norm/vextenddouble/vextenddouble|ξr/an}bracketri}ht/vextenddouble/vextenddouble=1√ |
d. |
However in the general case they may have different norms. |
Given a set of such vectors consider the Gram matrix |
Prs=/an}bracketle{tξr|ξs/an}bracketri}ht (35) |
Clearly the Gram matrix cannot determine the POVM uniquely since if Uis any |
unitary operator then the vectors U|ξr/an}bracketri}htwill define another POVM having the same |
Gram matrix. However, the theorem we now prove shows that this is the only free- |
dom. In other words, the Gram matrix fixes the POVM up to unitary e quivalence. |
The theorem also provides us with a criterion for deciding whether an arbitrary |
d2×d2matrixPis the Gram matrix corresponding to a POVM of the specified |
type. As a corollary this will give us a criterion for deciding whether an arbitrary |
tensorθrsis specifically the order-2 angle tensor for a SIC-family. |
Theorem 1. LetPbe anyd2×d2Hermitian matrix. Then the following conditions |
are equivalent: |
(1)Pis a rankdprojector. |
(2)Psatisfies the trace identities |
Tr(P) = Tr(P2) = Tr(P3) = Tr(P4) =d (36) |
(3)Pis the Gram matrix for a set of d2vectors|ξr/an}bracketri}ht(not all normalized) such |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.