Phenomenological Distinction Between Unitary and Antiunitary Symmetry Operators
Phenomenological Distinction Between Unitary and Antiunitary Symmetry Operators
Operators
Eugene P. Wigner
Downloaded 30 Dec 2012 to 128.148.252.35. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
JOURNAL OF MATHEMATICAL PHYSICS VOLUME 1, NUMBER 5 SEPTEMBER-QCTOBER, 1960
It is well known that one always can find as many orthogonal states (i.e., states between which the
transition probability is zero) as the·Hilbert space has dimensions which are invariant under a given unitary
transformation. The corresponding vectors are characteristic vectors of the unitary operator. In contrast,
most antiunitary operators leave not more than one state invariant. However, if there are two orthogonal
invariant states, a consideration of the states for which the transition probability is j into both invariant
states surely provides a distinction. In the antiunitary case, one of these states is also invariant, another
one is transformed into an orthogonal state, the rest are in between. In the unitary case, the transition
probability between original state and transformed state is the same for all states for which the transition
probability is i into two orthogonal states. This provides a "directly observable" distinction between
unitary and antiunitary transformations.
Downloaded 30 Dec 2012 to 128.148.252.35. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
UNITARY AND ANTIUNITARY OPERATORS 415
2. It follows that the structure of the set of invariant
Let us consider, first, the states which are left states is, in general, very different for unitary and
unchanged by the transformation. For these, antiunitary operators. An antiunitary operator may
have no invariant state, or it may have only one.
Tq;=wq;, (5) However, no matter how many invariant states it has,
they form a continuous manifold each member of
and we shall call them invariant states. If T is unitary, which can be changed continuously into any other.
q; is one of its characteristic vectors. It follows that If the antiunitary operator has n orthogonal invariant
there is at least one set of mutually orthogonal states
vectors the manifold of invariant states is an n-1
which contains as many members as any set of mutually dimensional continuous manifold. Its state vector can
orthogonal states contains. In particular, unless there be described by n real coefficients the sum of the
are at least two orthogonal invariant states, the squares of which is 1. There is no arbitrar~ phase
symmetry operator cannot be unitary and must be, factor in this case because (8) already determllles the
therefore, antiunitary. phase factor.
If the spectrum of the unitary operator is simple,
the invariant states are isolated, but if it has a char- 3.
acteristic value of multiplicity 1, the corresponding The number and topological properties of the
invariant states form a continuous manifold with 21- 2 invariant states actually permit a phenomenological
parameters. The characteristic functions have 1 com-
distinction between unitary and antiunitary trans-
plex or 21 real parameters but the normalization formations. The following distinction is, however, more
condition subjects these to one real equation and o~e direct.
real parameter, characterizing the phase factor, IS Consider two orthogonal invariant states, q;l and q;2.
physically meaningless. . If there are no such, the transformation is. surely
Let us consider now an antiunitary transformatIOn. antiunitary. Next, consider the states for whIch the
We decompose the state vector into the invariant and transition probability is t into both q;l and q;2. The
characteristic vectors of the operator T state vectors of these states are
q;= L: ak'Vlk+ L: b.,k'V.,k. (6)
(14)
k .,k
If T q;=w' q; (i.e., q; is an invariant state), we can Consider finally the transition probability of these
consider q;'=w'lq; and have states into the states Tif;a,
q; will satisfy (8) if It is independent of a, Le., the same for all the states
(10) if;a with transition probabilities t into q;1 and q;2.
and Let us assume next that Tis antiunitary. In this case
(11)
(18)
Since (11) must hold for all wand all k, w can be replaced
therein by w* to give and the transition probability becomes
(w*)ib.,k*=b.,.k or b.,.k* = wib.,k. (12) P(a)=tl (q;I+e ia q;2, q;1+e- ia q(2) 1 2=tll+e-2ia I 2
Insertion of the latter expression into (11) gives =Hl+cos2a). (19)
(13)
It varies, for the states in question, between 0 and 1.
. ..J'1 It follows that if the symmetry operator The preceding argument can yet be greatly
Slllce Wr- • d 'b d generalized. In the unitary case, i.f th~ transition
is antiunitary, the invariant states. can be ~scn. e
by invariant vectors, i.e., by real llllear conblllatIOns probabilities into n orthogonal Invanant states
2
(/!l, (/!2, " ' , (/!n are prescribed to be r12, r2 , ••• , rn2 so
of the 'Vlk.
Downloaded 30 Dec 2012 to 128.148.252.35. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
416 EUGENE P. WIGNER
probability from any of these Y;al" 'an into the cor- assume every value between o and 1 for suitably
responding TY;al" 'an is the same, namely, chosen a.
The striking difference in the relation of the original
) (Y;al" 'an,TY;al" 'an) )2
and transformed states to each other shows particularly
= I (L Tke iak ipk,L Tke iak ())k ipk) [2 clearly how definite the relations in question are in
= IL Tk2())k[2. (22) either case.
Downloaded 30 Dec 2012 to 128.148.252.35. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions