数理解析研究所講究録

第 859 巻 1994 年 127-142

127

Phase operator problem and infinitesimal analysis

MASANAO OZAWA
Department of Mathematics, College of General Education,
Nagoya University, Nagoya 464, Japan

Abstract
To find an operator representation of the phase variable of a single-mode
electromagnetic field, the Schrodinger representation is extended to a larger
Hilbert space including states with infinite excitation by nonstandard analy-
sis, so that the self-adjoint phase operator is acting on the extended Hilbert
space. It is shown that this phase operator is a Naimark extension of the
optimal probability operator-valued measure for the phase parameter already
known in the quantum estimation theory, and is naturally considered as the
controversial limit of the phase operators on finite dimensional spaces recently
proposed by Pegg and Barnett. Eventually, two of recent promising attempts
to obtain the correct statistics of the phase variable in quantum mechanics
is synthesized in the present framework based on new mathematical ideas
concerning the number system including infinite and infinitesimal numbers.

1. Introduction

The existence and properties of a self-adjoint operator on a Hilbert space, corre-

sponding to the phase of the electromagnetic field, has provoked many discussions
for some time, since Dirac [7] first discussed the problem. According to the unique-
ness theorem of the irreducible representations of the canonical commutation rela-
tions due to von Neumann, the commutation relation between the number operator
and the phase operator, which Dirac presupposed by employing the correspondence
between commutators and classical Poisson brackets, cannot be realized. Further,
Susskind and Glogower [31] clearly demonstrated that the polar decomposition of
the annihilation operator into the unitary operator of the exponential of the phase
and the square root of the number operator presupposed by Dirac is also impossi-
ble. However, Pegg and Barnett [28] recently made an interesting proposal for the
problem. They constructed a phase operator on a finite, but supposedly very large,
128

dimensional space and showed that it obeys a certain new commutation relation
with the number operator. According to their proposal, the correct statistical pre-
diction is given after the limit process making the dimension infinity, although they
failed to find the operator after the limit process.
On the other hand, another approach to the problem has been established in
quantum estimation theory $[15,18]$ . This theory discusses a quite general type of
optimization problems of quantum measurements. The statistics of measurements
are represented in this theory by the probability operator-valued measures (POM)
on Hilbert spaces which extend the conventional description by the self-adjoint op-
erators. In this approach, the optimum POM of the estimation problem of the phase
parameter was found by Helstrom [14] and mathematically rigorous development of
this approach is given by Holevo [18].
A promising aspect of these two approaches is that the statistics of the phase
variable obtained by Pegg and Barnett coincides naturally with the one represented
by the optimum POM of the phase parameter. This shows, however, that con-
trary to their claim the limit of their exponential phase operator is nothing but
the well-known Susskind-Glogower exponential phase operator [31], as long as the
limit is taken in the original Hilbert space. Since the limit process within the orig-
inal Hilbert space does not preserve the function calculus and destroys the desired
properties of the phase operator such as unitarity of the exponentials, this limit
has never the features which Pegg and Barnett [28] described intuitively. Accord-
ing to the Naimark theorem, every POM can be extended to a projection-valued
measure on a larger Hilbert space, which gives rise to a self-adjoint operator, by the
spectral theory, representing an observable in the standard formulation of quantum
mechanics. This suggests that there is the phase operator, somewhere beyond the
Schr\"odinger representation, which is the intuitive limit of the Pegg-Barnett phase
operators as well as being an extension of the optimal POM. Thus in order to real-
ize the Pegg and Barnett phase operator on the dimension at infinity, we need an
alternative mathematical construction other than limits in the Hilbert space.
For this purpose nonstandard analysis will be used in this paper. The nonstan-
dard analysis was invented by Robinson [29] and has yielded rigorous and fruitful
mathematics of infinite and infinitesimal numbers as well as has realized the Leibniz
infinitesimal calculus. We shall construct a natural extension of the Schr\"odinger
representation and show that the desired phase operator exists on this extended
Hilbert space. The Hilbert space of this extension of the Schr\"odinger representation
is the direct sum of the original Schr\"odinger representation and the space of the
 129

states with infinite excitation which are naturally considered as the classical lim-
its of the ordinary quantum states. Thus our construction supports the following
heuristic reasoning on why the phase operator does not exist in the Schr\"odinger
representation: The unitary operator of one quantity , from a pair of canon- $e^{i\hat{A}}$
$A$

ically conjugate quantities and $B$ , changes an eigenvector of the operator

$A$
to $\hat{B}$

another eigenvector in such a way as the eigenvalue changes in some magnitude, say
. Similarly,
$\delta B$
changes an eigenvector of by
$e^{i\hat{B}}$

. Then, it occurs either that $\hat{A}$

$\delta A$

both and are finite non-infinitesimal numbers or that one of them is infinite
$\delta A$ $\delta B$

and the other is infinitesimal. The first case is the case of so-called the Schr\"odinger
pairs such as the position-momentum pair. The second case arises when one of the
pair is a quantized quantity. In this case, the operator of the quantity conjugate to
the quantized quantity changes the eigenvalue in magnitude of an infinite number
and cannot be represented by an operator on the state space of the Schr\"odinger
representation which contains only eigenstates with a finite eigenvalue. This is true,
even if the state space is understood as the Schwarz space extending the Hilbert
space, since every self-adjoint operator defined on this space has a complete system
of eigenvectors with finite eigenvalues.
For bibliography on the phase operator problem, we shall refer to the refer-
ences of [27,2,4]. For the quantum estimation theory, [35,13,14,15,17,18], and for
recent development of quantum measurement theory, [6,5,25,26]. Applications of
nonstandard analysis to physics is not new and has been developed in such pa-
pers as [3,8,9,12,20,32,21,22,33,34,24,11,10], and in such a monograph as [1]. For
mathematical foundations, we shall refer to the following monographs [29,30,19,1].

2. The Susskind-Glogower Phase operators

The single-mode electromagnetic field is a well-known physical system which has

been modeled by the quantum mechanical harmonic oscillator with unit mass. Let
be the Hilbert space of the Schr\"odinger representation of the quantum mechanical
$\mathcal{H}$

harmonic oscillator. Denote by the algebra of all bounded operators on {. Let

$\mathcal{L}(\mathcal{H})$ $j$

and be the position and momentum operators on . The annihilation operator

$\hat{q}$ $\hat{p}$
$\mathcal{H}$

is defined by
$\hat{a}$

$\hat{a}=\frac{1}{\sqrt{2h\omega}}(\omega\hat{q}+i\hat{p})$
, (2.1)
130

where is the angular frequency, and its adjoint

$\omega$
$\hat{a}^{\dagger}$

is the creation operator. Then

the number operator is defined by $\hat{N}$

$\hat{N}=\hat{a}^{\dagger_{\hat{O}}}$

. (2.2)

The number operator has the complete orthonormal basis $\{|n\rangle|n=1,2, \ldots\}$
$\hat{N}$

of for which
$\mathcal{H}$

} $=n|n\rangle$ . The Hamiltonian of the system is given by

$\hat{N}|n$ $\hat{H}$ $\hat{H}=$

.
$h \omega(\hat{N}+\frac{1}{2})$

In his original description of the quantized electromagnetic field, Dirac [7] postu-
lated the existence of a self-adjoint phase operator such that the unitary exponen- $\hat{\phi}$

tial operator of would appear in the polar decomposition of the annihilation

$e^{i\overline{\phi}}$
$\hat{\phi}$

operator
. (2.3) $\hat{a}=e^{i\overline{\phi}}\hat{N}^{-1/2}$

The difficulty with this approach were clearly pointed out by Susskind and Glogower
[31] by showing that the polar decomposition of cannot be realized by any unitary $\hat{a}$

operator. Instead, they introduced the isometries representing the exponentials of

the phase variable
$\hat{e}^{i\phi}$

$=$ $(\hat{N}+1)^{-1/2}\hat{a}$
, (2.4)
$\hat{e}^{-i\phi}$
$=$ $\hat{a}^{\dagger}(\hat{N}+1)^{-1/2}$
, (2.5)

and the self-adjoint operators representing the sine and cosine of the phase variable

$\overline{\cos}\phi$
$=$ $\frac{1}{2}(\hat{e}^{i\phi}+\hat{e}^{-i\phi})$
, (2.6)
$s\overline{in}\phi$
$=$ $\frac{1}{2i}(\hat{e}^{i\phi}-\hat{e}^{-i\phi})$
. (2.7)

Note that the places of the carets in these Susskind-Glogower operators suggest
that these operators are not derived by the function calculus of a certain self-adjoint
operator corresponding to . These operators are considered to behave well in the $\phi$

limit of large amplitudes but they fail to define well-behaved operators for periodic
functions of the phase in the quantum regime. Thus we cannot derive the correct
statistics of the phase variable from these operators. However, it turns out that that
these operators give the correct mean values of the corresponding quantities , $e^{\pm i\phi}$

and
$\sin\phi$ . A systematic method for obtaining the correct statistics needs a
$\cos\phi$

new mathematical concept extending the self-adjoint operators, which is described

in the next section.
 131

3. Quantum estimation problem of the phase parameter

A positive operator-valued measure on the Borel field of the real line $P(d\theta)$ $\mathcal{B}(R)$

$R$ with values in is called a probability operator-measure (POM) if $P(R)=1$ .

$\mathcal{L}(H)$

Now, we shall start with the following presupposition: Corresponding to any mea-
surable physical quantity $X$ , there is a $POMP_{X}(d\theta)$ such that the probability dis-
tribution of $X$ in state is given by the probability measure {
$\psi\in \mathcal{H}$
, $\psi|P_{X}(d\theta)|\psi\rangle$

which predicts the statistics of outcomes of ideal measurements of $X$ in . In the $\psi$

conventional framework of quantum mechanics, every observable which has the cor-
responding self-adjoint operator in has the POM as its spectral measure. However, $\mathcal{H}$

there may be some physical quantities which are considered to have the POM but
no self-adjoint operators in . For measurability of POM’s, it is known that for any
$\mathcal{H}$

POM on , there is another Hilbert space , a unit vector

$P(d\theta)$ $\mathcal{H}$

, a unitary $\mathcal{K}$
$\xi\in \mathcal{K}$

operator on $U$
, and a self-adjoint operator in with spectral measure
$\mathcal{H}\otimes \mathcal{K}$ $\mathcal{K}$
$E(d\theta)$

satisfying
{ ,
$\psi|P(d\theta)|\psi\rangle$
(3.1)
$=\{\psi\otimes\xi|U^{\dagger}(1\otimes E(d\theta))U|\psi\otimes\xi\}$

for all . This suggests that for any POM

$\psi\in \mathcal{H}[25]$ there is an experi- $P(d\theta)$

ment, with the outcome to be predicted by statistically, which consists of $P(d\theta)$

the following process; 1) preparation of the apparatus, described by , in state , $\mathcal{K}$

$\xi$

2) interaction, described by , between the object, described by , and the appa- $U$ $\mathcal{H}$

ratus, 3) measurement of the observable in the apparatus corresponding to . $E(d\theta)$

We shall call any experiment consisting of the above process and satisfying (3.1) as
a measurement of POM . Thus, our presupposition is a conservative exten-
$P(d\theta)$

sion of the standard formulation of quantum mechanics, in the sense that, if every
observable can be measured, so can any quantity corresponding to a POM.
The determination of the statistics of the phase variable is thus reduced to deter-
mination of POM corresponding to the phase variable. This problem is solved in the
quantum estimation theory as follows. Since the phase is canonically conjugate to
the action variable in classical mechanics, the number operator is the infinitesimal
generator of the phase shift operators . Thus the POM corresponding to $e^{i\theta\hat{N}}$
$P(d\theta)$

the phase variable should satisfy the relations

$e^{-i\theta\hat{N}}P(B)e^{i\theta\hat{N}}=P(B_{-\theta})$
,
(3.2)
$B,$ $B_{-\theta}\subset[0,2\pi$ ), $B_{-\theta}=B-\theta$ $(mod 2\pi)$ .

Any POM satisfying (3.2) is called a covariant $POM$. It is well known that there is
no self-adj oint operators such that their spectral measures satisfy the above relations
132

but there are many solutions among general POM’s. In order to select the optimum
one, consider the following estimation problem of the phase parameter . Let us $\theta$

given a reference state , which is supposed to be modulated by a phase $\psi\in \mathcal{H}$

shifter with unknown shift parameter $\theta(0\leq\theta<2\pi)$ so that the outgoing state is
. The problem is to find an experiment in state
$\psi_{\theta}=e^{i\theta\hat{N}}\psi$
which gives the best $\psi_{\theta}$

estimate of the parameter , and is equivalently to find a measurement in the state $\theta$

the outcome of which is the best estimate of the parameter . The relevance
$\psi_{\theta}$
$\overline{\theta}$
$\theta$

of this problem to our problem is as follows. Suppose that the reference state $\psi$

were the phase eigenstate $\psi=|\phi=0$ }. Then the outgoing state would also be
the phase eigenstate }, for which the the best estimator would give the $\psi_{\theta}=|\phi=\theta$

estimate with probability 1. Thus in this case the best estimator would be
$\overline{\theta}=\theta$

the measurement of the phase variable. Thus for a POM to represent the $P(d\theta)$

phase variable, it is necessary that it is the optimum estimator of this estimation

problem for reference states which approximate a phase eigenstate well. Although
we do not know what and where are phase eigenstates, we can reach the essentially
unique solution. For any POM , the joint probability distribution $P(d\theta)$

$p(d \theta, d\overline{\theta})=\{\psi_{\theta}|P_{\phi}(d\overline{\theta})|\psi_{\theta}\}\frac{d\theta}{2\pi}$

(3.3)

gives naturally the joint probability distribution of the true parameter and the $\theta$

estimate . Given an appropriate error function

$\overline{\theta}$

, which gives the penalty $W(\theta-\overline{\theta})$

for the case , the optimum estimator should minimize the average error
$\theta\neq\overline{\theta}$

$\int_{0}^{2\pi}W(\theta-\overline{\theta})p(d\theta, d\overline{\theta})$
. (3.4)

This type of optimization problem has been studied in quantum estimation theory
$[15,18]$ . The common optimum solution, which is also a covariant POM, for a large

class of error functions such as $W(x)=4 \sin 2\frac{x}{2}$ or $W(x)=-\delta(x)$ , where is the $\delta(x)$

periodic -function, is the following POM

$\delta$

: $P(d\theta)$

{ $n|P(d\theta)|n’\rangle$ $=e^{i(\alpha_{n}-\alpha_{n’})}e^{i(n-n’)\theta} \frac{d\theta}{2\pi}$

, (3.5)

where $(n=0,1, \ldots)$ is the number basis and

$|n\rangle$
. Since this problem $\alpha_{n}=\arg\{n|\psi\rangle$

is not of the estimation of the absolute phase, the optimum solutions depend on
the phase factors of the reference state . However, this dependence can be $\alpha_{n}$
$\psi$

interpreted to reflect the arbitrariness of our choice of the phase eigenstate , $|\phi=0\rangle$

and each choice from the optimum POM’s determines a unique $|\phi=0$ } among $P(d\theta)$
 133

physically equivalent alternatives. For simplicity, we choose the solution for $\alpha_{n}=1$

$(n=0,1, \ldots)$ and denote it by , i.e., $P_{\phi}(d\theta)$

{ $n|P_{\phi}(d\theta)|n’\rangle$ $=e^{i(n-n’)\theta} \frac{d\theta}{2\pi}$

. (3.6)

We will call as the phase $POM$.

$P_{\phi}(d\theta)$

From this result, we can calculate the mean value of any bounded Borel function
$f(\phi)$ of the phase variable . Indeed, let $\phi$

$f \overline{(\phi})=\int_{0}^{2\pi}f(\theta)P_{\phi}(d\theta)$
. (3.7)

Then the mean value of the quantity in a state is given by . An $f(\phi)$ $\psi$
$\{\psi|f\overline{(\phi})|\psi\}$

interesting result from this is that the Susskind-Glogower phase operators coincide
with the operators defined in this way [18, p. 141], i.e.,

$\hat{e}^{\pm i\phi}$

$=$ $\int_{0}^{2\pi}e^{\pm i\theta}P_{\phi(d\theta)}=e^{\overline{\pm i}\phi}$

, (3.8)

$\overline{\cos}\phi$
$=$ $\int_{0}^{2\pi}\cos\theta P_{\phi}(d\theta)=c\overline{os}\phi$
, (3.9)

$s\overline{in}\phi=$ $\int_{0}^{2\pi}\sin\theta P_{\phi}(d\theta)=s\overline{in}\phi$

. (3.10)

Thus their operators give the correct mean values of $f(\phi)=e^{\pm i\phi},$ and $\sin\phi$ $\cos\phi$

respectively, but none of their powers.

The phase POM gives the correct statistics for all functions of phase
$P_{\phi}(d\theta)$

variable but it gives little information about algebraic relations between other ob-
servables.

4. The approach due to Pegg and Barnett

Now we shall turn to the recent proposal due to Pegg and Barnett [28]. They start
with the s-dimensional subspace of $H$ spanned by with $n=0,1,$ $s-1$ .
$\Psi_{s}$ $|n\rangle$
$\ldots,$

For $\theta_{m}=m\triangle\theta(m=0,1, \ldots, s-1)$ , where , their approximate phase $\triangle\theta=2\pi/s$

state is
$| \theta_{m}\rangle=s^{-1/2}\sum_{n=0}^{s-1}e^{in\theta_{m}}|n\}$
, (4.1)

and their approximate phase operator $\hat{\phi}_{s}$

on $\Psi_{s}$
is

$\hat{\phi}_{s}=\sum_{m=0}^{s-1}\theta_{m}|\theta_{m}\}\{\theta_{m}|$
. (4.2)
134

Their intrinsic proposal is that the mean value of the quantity $f(\phi)$ in state is the $\psi$

limit of as $sarrow\infty$ . Then for state

$\langle\psi|f(\hat{\phi}_{s})|\psi\rangle$
} $(k<\infty)$ , we have $\psi=\Sigma_{n=0}^{k}c_{n}|n$

$\lim_{sarrow\infty}\langle\psi|f(\hat{\phi}_{s})|\psi\rangle$ $=$ $\sum_{n,n=0}^{k}c_{n}c_{n}^{*},\lim_{sarrow\infty}\sum_{m=0}^{s-1}f(\theta_{m})e^{i(n-n’)\theta_{m_{\frac{\triangle\theta}{2\pi}}}}$

$\int_{0}^{2\pi}f(\theta)e^{i(n-n’)\theta}\frac{d\theta}{2\pi}$
$=$ $\sum_{n,n=0}^{k}c_{n}c_{n}^{*},$

$=$ $\int_{0}^{2\pi}f(\theta)\{\psi|P_{\phi}(d\theta)|\psi\}$

$=\{\psi|f\overline{(\phi})|\psi\rangle$
,

for any continuous function on $f(\theta)$ $[0,2\pi]$ . Thus, their expectations are the same
as those given by the phase POM $P_{\phi}(d\theta)$
, and we have

$\lim_{sarrow\infty}f(\hat{\phi}_{s})=\int_{0}^{2\pi}f(\theta)P_{\phi}(d\theta)=f\overline{(\phi})$
, (4.3)

in the weak operator topology. In particular, the limit of their exponential phase
operators are the Susskind-Glogower exponential operators, i.e.,
$\lim_{sarrow\infty}e^{\pm i\hat{\phi}_{3}}=e^{\overline{\pm i}\phi}$
. (4.4)

We have, therefore, shown that the statistics of the phase variable obtained by
Pegg and Barnett coincides with the statistics obtained by the phase POM, and that
the limit of the Pegg-Barnett unitary phase operators on finite dimensional spaces
is nothing but the Susskind-Glogower exponential operators, as long as the limit is
taken in the original Hilbert space.

5. Hyperfinite extension of the Schr\"odinger representation

For the basic framework of nonstandard analysis we shall refer to Hurd-Loeb [19]
and Stroyan-Luxemburg [30] as standard textbooks. In what follows we assume
that our nonstandard universe is an -saturated bounded elementary extension of $\aleph_{1}$

a superstructure which contains . To avoid confusions, we shall use symbol \dag er for $\mathcal{H}$

the adjoint operation of operators and the involution on -algebra. $a^{*}$

Let $(E,p)$ be an internal normed linear space with norm over the hypercomplex $p$

number field *C. We define the principal galaxy of $(E,p)$ and the principal $E_{G}$

monad of $(E,p)$ as follows:

$E_{M}$

$E_{G}$ $=$
{ $x\in E|p(x)$ is finite}, (5.1)
$E_{\Lambda I}$ $=$
{ $x\in E|p(x)$ is infinitesimal}. (5.2)
 135

Then both and are linear spaces over the complex number field C. Let
$E_{G}$ $E_{M}$

$\hat{E}=E_{G}/E_{M}$ and $\hat{p}([x])=\circ(p(x))$ for $x\in E_{G}$ , where $[x]=x+E_{M}$ , and o stands for

the standard part map on the finite hypercomplex numbers. Then becomes $(\hat{E},\hat{p})$

a normed linear space over , called the nonstandard hull of a normed linear space $C$

$(E,p)$ . Under the assumption of -saturation, the hull completeness theorem holds $\aleph_{1}$

and concludes that the nonstandard hull is a Banach space. $(\hat{E},\hat{p})$

Let be the nonstandard extension of and

$\mathcal{H}$

its nonstandard hull. Then $\mathcal{H}$

$\overline{\mathcal{H}}$ $\overline{*\mathcal{H}}$

is a Hilbert space with inner product { . Let be a nonstandard $[\xi]|[\eta]\rangle$ $=\circ(^{*}\{\xi|\eta\})$ $\nu$

natural number and the internal -dimensional subspace of spanned by the $\mathcal{D}$
$\nu$
$\mathcal{H}$

hyperfinite set $\{|n\}|n=0,1,$ }. Then the nonstandard hull cont.ains $\ldots,$

$\iota/-1$
$\hat{\mathcal{D}}$

$\mathcal{H}$

as a closed subspace by the canonical isometric embedding which maps to $V_{H}$ $\xi\in \mathcal{H}$

. Since the family of all

$[^{*}\xi]\in\hat{\mathcal{D}}$
$[|n\}]$ with $0\leq n\leq\nu-1$ is an orthonormal family
in , the external dimension of is at least the cardinality of the continuum. The
$\hat{D}$ $\hat{\mathcal{D}}$

nonstandard hulls and are, thus, Hilbert spaces which satisfy the relations $\overline{*\mathcal{H}}$ $\hat{\mathcal{D}}$

$7- t\subset\hat{\mathcal{D}}\subset\overline{*\mathcal{H}}$

. (5.3)

Let be the algebra of internal linear operators on . Then $A$ is a hyperfinite

$\mathcal{A}$ $\mathcal{D}$

dimensional internal *-algebra over *C. For $x\in A$ , let be the internal uniform $\Vert x\Vert$

norm of . Let be the principal galaxy of

$x$ and the principal monad.
$\mathcal{A}_{G}$ $(\mathcal{A}, \Vert\cdot\Vert)$ $\mathcal{A}_{M}$

Obviously, is $a*$ -algebra and is a self-adjoint two sided ideal of

$\mathcal{A}_{G}$
. Thus, $A_{M}$ $\mathcal{A}_{G}$

from the hull completeness theorem, the nonstandard hull of becomes $\hat{\mathcal{A}}$

$(\mathcal{A}, \Vert\cdot\Vert)$

a Banach -algebra with norm I $*$

. Then the norm . I satisfies the $[x]\Vert=\circ\Vert x\Vert$ $\Vert$

-condition, i.e., 1
$C^{*}$
for all , and hence is a $[x]^{*}[x]\Vert=\Vert[x]\Vert^{2}$ $[x]\in\hat{\mathcal{A}}$ $(\hat{\mathcal{A}}, \Vert\cdot\Vert)$ $C^{*}-$

algebra. The internal algebra is internally *-isomorphic with the internal matrix $\mathcal{A}$

algebra $M(\nu, *c)$ of all matrices over $*c$ by a matrix representation of linear $\nu\cross\nu$

operators. In [16], it is shown that the -algebra has a closed maximal ideal $C^{*}$
$\hat{4}$
$\mathcal{I}$

defined by
$\mathcal{I}=$
{ $[x] \in\hat{\mathcal{A}}|(\frac{1}{\nu}Tr[x^{*}x])^{1/2}$
is infinitesimal}, (5.4)

where Tr stands for the trace of the corresponding matrix, such that is a type $\hat{\mathcal{A}}/\mathcal{I}$

factor which is not separably representable nor approximately finite.

$II_{1}$

Any internal operator leaves and invariant and gives rise to a $x\in \mathcal{A}_{G}$ $\mathcal{D}_{G}$ $\mathcal{D}_{M}$

bounded operator on such that $\lambda(x)[\xi]=[x\xi]$ for all . Then the$\lambda(x)$

$\hat{\mathcal{D}}$

$\xi\in \mathcal{D}_{G}$

correspondence $[x]\mapsto\lambda(x)$ for defines a faithful -representation of the $x\in \mathcal{A}_{G}$

$*$

-algebra on the Hilbert space .

$C^{*}$
$\hat{\mathcal{A}}$ $\hat{\mathcal{D}}$

Let be the internal projection from

$P_{D}$
onto . Any bounded operator on $*\mathcal{H}$ $\mathcal{D}$
$T$

has the nonstandard extension

$\mathcal{H}$

which is an internal bounded linear operator $*\tau$

136

on . Denote the restriction of

$*H$ to by . Then it is easy to see that $P_{D^{*}}T$ $\mathcal{D}$
$*\tau_{D}$

. We denote the operator A

$*\tau_{D}\in \mathcal{A}_{G}$
on by , which is called the standard $(^{*}T_{D})$
$\hat{\mathcal{D}}$

$T_{D}$

hyperfinite extension of to . Then $T_{D}=T$ on and . Properties $T$

$\hat{\mathcal{D}}$

$\mathcal{H}$
$\Vert T_{D}\Vert=\Vert T\Vert$

of such extensions from to are studied by Moore [23] extensively. Now the $\mathcal{H}$
$\hat{\mathcal{D}}$

following statement is easily established; cf. [23, Lemma 1.3],

Theorem 5.1. The mapping : $T\mapsto T_{D}$ is a completely positive isometric $\mathcal{E}_{D}$

injection from to and the mapping : is a completely

$\mathcal{L}(\mathcal{H})$
$\hat{\mathcal{A}}$

$\mathcal{E}_{H}$ $\lambda(T)\mapsto V_{H}^{\dagger}\lambda(T)V_{H}$

positive surjection of onto . The composition is the identity map on $\hat{\mathcal{A}}$

$\mathcal{L}(\mathcal{H})$ $\mathcal{E}_{H}\mathcal{E}_{D}$

and the composition

$\mathcal{L}(\mathcal{H})_{f}$
is the norm one projection from onto . $\mathcal{E}_{D}\mathcal{E}_{H}$
$\hat{\mathcal{A}}$

$\mathcal{E}_{D}(\mathcal{L}(\mathcal{H}))$

6. The phase operator in the hyperfinite extension

In this section, we will show that there is a bounded self-adjoint operator $\hat{\phi}$

in $\hat{A}$

satisfying the following two conditions:

(P1) The spectral measure $E_{\phi}$
of $\hat{\phi}$

satisfies the relation

$P_{\phi}(d\theta)=V_{H^{\uparrow}}E_{\phi}(d\theta)V_{H}$
, (6.1)

where $P_{\phi}$
is the phase POM.
(P2) The Susskind-Glogower phase operators are given by the relations:
$\hat{e}^{i\phi}$

$=$
$V_{H^{\uparrow}}e^{i\hat{\phi}}V_{H}$
, (6.2)
$\hat{e}^{-i\phi}$
$=$
$V_{H}^{\dagger}e^{-i\hat{\phi}}V_{H}$
, (6.3)
$\overline{\cos}\phi$ $=$ $V_{H}^{\dagger}\cos\hat{\phi}V_{H}$
, (6.4)
$s\overline{in}\phi$
$=$ $V_{H^{1}}\sin\hat{\phi}V_{H}$
. (6.5)

Let $\triangle\theta=2\pi/\nu$
, and for each
$\theta_{m}=m\triangle\theta$
$m(m=0,1, \ldots, \nu-1)$ . The internal
phase eigenstate in
$|\theta_{m}\rangle$
$\mathcal{D}$

is defined by

$| \theta_{m}\rangle=\nu^{-1/2}\sum_{n=0}^{t/-1}e^{in\theta_{m}}|n\rangle$
. (6.6)

Then we have
$\{\theta_{m}|\theta_{m’}\}=\delta_{m,m’}$ . (6.7)
The internal phase operator $\hat{\phi}_{I}$

on $\mathcal{D}$

is defined by

$\hat{\phi}_{I}=\sum_{m=0}^{\iota/-1}\theta_{m}$

I $\theta_{m}\rangle$
\langle $\theta_{m}|$
. (6.8)
 137

Then the internal phase operator has the spectrum $\{2m\pi/\nu|m=0,1, \ldots, \nu-1\}$ $\hat{\phi}_{I}$

and hence is in . Thus we have the self-adjoint operator on , denoted by

$\mathcal{A}_{G}$
$\lambda(\hat{\phi}_{I})$
$\hat{\mathcal{D}}$

. Denote by
$\hat{\phi}$

the spectrum of and the point spectrum (eigenvalues)

$\Lambda(\hat{\phi})$ $\hat{\phi}$
$\Pi_{0}(\hat{\emptyset})$

of . $\hat{\phi}$

Theorem 6.1. We have $\Lambda(\hat{\phi})=\Pi_{0}(\hat{\emptyset})=[0,2\pi]$

. For each $\theta(\leq\theta\leq 2\pi)$ , the
vector with is an eigenvector
$[|\theta_{m}\}]$ $\theta_{m}\approx\theta$
of for $\hat{\phi}$

the eigenvalue . $\theta$

Proof. Let $\theta\in[0,2\pi]$ and $\theta_{m}\approx\theta$

. Then obviously,
$\hat{\phi}[|\theta_{m}\}]=[\hat{\phi}_{I}|\theta_{m}\}]=[\theta_{m}|\theta_{m}\}]=\circ\theta_{m}[|\theta_{m}\rangle]=\theta[|\theta_{m}\}]$
,
and hence is an eigenvector of corresponding to eigenvalue . Thus we have
$[|\theta_{m}\rangle]$
$\hat{\phi}$ $\theta$

$[0,2\pi]\subset\Pi_{0}(\hat{\phi})$
. Since , we have for any unit vector
$0\leq\theta_{m}\leq 2\pi$ $0\leq\{\psi|\hat{\phi}_{I}|\psi\}\leq 2\pi$

$\psi\in \mathcal{D}$
, and hence
$0\leq\langle[\psi]|\hat{\phi}|[\psi]\rangle=\circ\langle\psi|\hat{\phi}_{I}|\psi\}\leq 2\pi$
.
It follows that $0\leq\hat{\phi}\leq 2\pi 1$
, so that $\Lambda(\hat{\phi})\subset[0,2\pi]$
. Therefore, $\Lambda(\hat{\phi})=[0,2\pi]=$

.
$\Pi_{0}(\hat{\emptyset})$ $\square$

For each $\theta(0\leq\theta\leq 2\pi)$ and $n\in N$ , define $F(\theta, n)$ to be the internal projection

$F( \theta, n)=\sum_{\theta_{m}\leq\theta+n^{-1}}|\theta_{m}\rangle\{\theta_{m}|$

.

Then $F(\theta, n)=0$ if $\theta+n^{-1}<0$ , and for each the sequence is a monotone $\theta$
$\{\hat{F}(\theta, n)\}$

decreasing sequence of projections on . Define to be the strong limit of $\hat{\mathcal{D}}$

$E_{\phi}(\theta)$

. Then $\{E_{\phi}(\theta)|\theta\in[0,2\pi]\}$ is the spectral resolution for

$\hat{F}(\theta, n)$
, Theorem $\hat{\phi}[23$

4.1].

Theorem 6.2. The bounded self-adjoint operator $\hat{\phi}$

in $\hat{\mathcal{A}}$

with its spectral reso-

lution satisfies conditions (P1) and (P2).
$E_{\phi}(\theta)$

Proof. Let $n,$ $n’\in N$ . We have

$\{n|E_{\phi}(\theta)|n’\}$ $=$ $\lim_{karrow\infty}\{n|\hat{F}(\theta, k)|n’\rangle$

$=$ $\lim_{karrow\infty}\circ\{n|F(\theta, k)|n’\}$

$=$ $\lim_{karrow\infty}\circ\sum_{\theta_{m}\leq\theta+k^{-1}}e^{i(n-n’)\theta_{m_{\frac{\triangle\theta}{2\pi}}}}$

$\lim_{karrow\infty}\int_{0}^{\theta+k^{-1}}e^{i(n-n’)\overline{\theta}_{\frac{d\overline{\theta}}{2\pi}}}$
$=$

$=$
$\int_{0}^{\theta}e^{i(n-n’)\overline{\theta}}\frac{d\overline{\theta}}{2\pi}$

.
138

From (3.6), this concludes condition (P1). To prove (P2), by the obvious relations,
it suffices to show (6.3). We have

$\{n|e^{-i\hat{\phi}}|n’\}$
$=$ $\int_{0}^{2\pi}e^{-i\theta}\langle n|dE_{\phi}(\theta)|n’$
}
$\int_{0}^{2\pi}e^{i(n-n’-1)\theta}\frac{d\theta}{2\pi}$
$=$

$=$ $\delta_{n,n’+1}$ ,

whence (6.3) is obtained from the relation

$\{n|(\hat{N}+1)^{-\frac{1}{2}}\hat{a}|n’\rangle$
$=$ $\langle n|n’+1\rangle$
.

$\square$

7. Macroscopic states in the hyperfinite extension

We have extended the Schr\"odinger representation on to its hyperfinite extension $\mathcal{H}$

. The following theorem shows that the augmented states in

$\hat{\mathcal{D}}$

can be in- $\hat{\mathcal{D}}\ominus \mathcal{H}$

terpreted naturally as states representing the macroscopic limits of the quantum

mechanical states in . $\mathcal{H}$

Theorem 7.1. Let . Suppose that \langle $n|T|n’$ } $(n, n’\in N)$ is a Cauchy
$T\in \mathcal{L}(\mathcal{H})$

sequence in and ‘. Then for any $k’\in\{0,1, \ldots, \nu-1\}\backslash N$ the standard
$n$ $n$ $k,$

hyperfinite extension \in \^A of satisfies the relation $T_{D}$ $T$

{ $k|T_{D}|k’\rangle$ $=$ $\lim_{n,n’arrow\infty}\{n|T|n’\rangle$ . (7.1)

Proof. Let and $L= \lim_{n,n’arrow\infty}a_{n,n’}$ . Let $*a_{m,m’}$

$a_{n,n’}=\langle n|T|n’\rangle$ for $n,$ $n’\in N$ ,

$(m, m’\in*N)$ be the nonstandard extension of the sequence $a_{n,n’}(n, n’\in N)$ .
Then, since is a Cauchy sequence, $L\approx a_{m,m’}$ for all $m,$ $m’\in*N\backslash N$ . Let
$a_{n,n’}$

$k,$$k’\in\{0,1, \ldots, \nu-1\}\backslash$ ’N. By transfer principle, , and hence $*a_{k,k’}=\langle k|^{*}T|k’\rangle$

$L\approx\{k|^{*}T|k’$ ). It follows that $L=\circ\{k|^{*}T|k’\rangle$ . $=\{k|T_{D}|k’\rangle$ $\square$

Remark. Suppose that the nonstandard universe is constructed by a bounded

ultrapower of a superstructure based on $R$ , using the index set $I=N$ and a free
ultrafilter . Then any hyperfinite number $k\in*N\backslash N$ is represented by a sequence
$\mathcal{U}$

$s(i)(i\in N)$ of natural numbers in such a way that two sequences $s(i)$ and $s’(i)$
represents the same hyperfinite number if and only if $\{i\in N|s(i)=s’(i)\}\in \mathcal{U}$ . $k$

Let $m$ be a hyperfinite number corresponding to a sequence $s(i)$ . Let $a(n)$ be a

 139

bounded sequence of complex numbers. Then the standard part of $*a_{m}$ coincides
with the ultralimit of the subsequence $a(s(i))$ of $a(n)$ , i.e.,

$\circ(^{*}a_{m})=\lim_{iarrow \mathcal{U}}a(s(i))$ . (7.2)

Thus, for any $T\in \mathcal{L}(\mathcal{H})$

and $k,$ $k’\in*N\backslash N$ , we have,

{ $k|T_{D}|k’\rangle$
$= \lim_{i,jarrow \mathcal{U}}\{s(i)|T|s’(j)\}$ , (7.3)

provided $kk’$ are represented by sequences $s(i),$ $s(j)$ .

By the above theorem, it is obvious that the space contains microscopic states, $\hat{\mathcal{D}}$

macroscopic states and superpositions of those states. In the conventional physics,

microscopic properties and macroscopic properties are discussed separately in quan-
tum mechanics and in classical mechanics (or quantum mechanics with $harrow\infty$ ).
However, these approaches cannot mention the quantum mechanical coherence be-
tween microscopic states and macroscopic states. Our new representation has just
realized one of such a coherent description of quantum and classical mechanics –

only in such a representation the phase operator behaves as a self-adjoint operator.

A potential application of this representation other than phase sensitive effects is
the measurement problem, where the unitary time evolution in an amplifier evolves
from a microscopic state to a state with infinite excitation which can be observed
macroscopically just like Schr\"odinger’s cat. An application of our framework to this
area will be published elsewhere.

Acknowledgments
The author thanks Roy J. Glauber, Horace P. Yuen and Izumi Ojima for helpful
discussions.

References

[1] S. Albeverio et. al. Nonstandard methods in stochastic analysis and mathemat-
ical physics. Academic Press, Orland, 1986.
[2] S. M. Barnett and D. T. Pegg. Phase in quantum optics. J. Phys. A: Math.
Gen., 19, 3849-3862, (1986).
[3] Ph. Blanchard and J. Tarski. Renormalizable interactions in two dimensional
sharp-time fields. Acta Physica Austriaca, 49, (1978).
140

[4] P. Carruthers and M. M. Nieto. Phase and angle variables in quantum me-
chanics. Rev. Mod. Phys., 40, 411-440, (1968).
[5] E. B. Davies. Quantum Theory of Open Systems. Academic Press, London,
1976.
[6] E. B. Davies and J. T. Lewis. An operational approach to quantum probability.
Commun. Math. Phys., 17, 239-260, 1970.
[7] P. A. M. Dirac. The quantum theory of the emission and absorption of radia-
tion. Proc. R. Soc. London, Ser., A, 114,243-265, (1927).
[8] M. O. Farrukh. Application of nonstandard analysis to quantum mechanics. J.
Math. Phys., 16, 177-200, (1975).
[9] J. E. Fenstad. Nonstandard methods in stochastic analysis and mathematical
physics. Jber. d. Dt. Math. Verein., 82, (1980).
[10] R. Fittler. More nonstandard quantum electrodynamics. Helv. Phys. Acta, 60,
881-902, (1987).
[11] R. Fittler. Some nonstandard quantum electrodynamics. Helv. Phys. Acta, 57,
579-609, (1984).
[12] C. E. Francis. Applications of nonstandard analysis to relativistic quantum
mechanics I. J. Phys. A: Math. Gen., 14, 2539-2551, (1981).
[13] C. W. Helstrom. Cram\’er-rao inequality for operator-valued measures in quan-
tum mechanics. Int. J. Theor. Phys., 8, 361-376, (1973).
[14] C. W. Helstrom. Estimation of a displacement parameter of a quantum system.
Int. J. Theor. Phys., 11(6), 357-378, (1974).
[15] C. W. Helstrom. Quantum Detection and Estimation Theory. Academic Press,
New York, 1976.
[16] T. Hinokuma and M. Ozawa. Conversion from nonstandard matrix algebras to
standard factors of type . Submitted to Ill. J. Math.
$II_{1}$

[17] A. S. Holevo. Estimation of shift parameters of a quantum state. Rep. Math.

Phys., 13, 287-307, (1978).
[18] A. S. Holevo. Probabilistic and Statistical Aspects of Quantum Theory. North-
Holland, Amsterdam, 1982.
[19] A. E. Hurd and P. A. Loeb. An Introduction to Nonstandard Real Analysis.
Academic Press, Orland, 1985.
 141

[20] R. Kambe. On quantum field theory I. Prog. of Theor. Phys., 52, 688-706,
(1974).
[21] P. J. Kelemen. Applications of nonstandard analysis to quantum mechanics and
quantum field theory. In A. E. Hurd and P. Loeb, edit-ors, Victoria Symposium
on Nonstandard Analysis 1972, Lecture Notes in Math. 369, pages 116-121,
Springer-Verlag, Berlin, 1974.
[22] P. J. Kelemen and A. Robinson. The nonstandard $\lambda$

: $\phi_{2}^{4}(x)$
: model. I, II. J.
Math. Phys., 13, 1870-1874, 1875-1878, (1972).
[23] L. C. Moore, Jr. Hyperfinite extensions of bounded operators on a separable
Hilbert space. Trans. Amer. Math. Soc., 218, 285-295, (1976).
[24] T. Nakamura. A nonstandard representation of Feynman’s path integrals. J.
Math. Phys., 32, 457-463, (1991).
[25] M. Ozawa. Quantum measuring processes of continuous observables. J. Math.
Phys., 25, 79-87, (1984).
[26] M. Ozawa. Realization of measurement and the standard quantum limit. In P.
Tombesi and E. R. Pike, editors, Squeezed and Nonclassical Light, pages 263-
286, Plenum, New York, 1989.
[27] D. T. Pegg and S. M. Barnett. Phase properties of the quantized single-mode
electromagnetic field. Phys. Rev. A, 39,1665-1675, (1989).
[28] D. T. Pegg and S. M. Barnett. Unitary phase operator in quantum mechanics.
Europhys. Lett., 6, 483-487, (1988).
[29] A. Robinson. Non-Standard Analysis. North-Holland, Amsterdam, 1962.
[30] K. D. Stroyan and W. A. J. Luxemburg. Introduction to the Theory of In-
finitesimals. Academic Press, New York, 1976.
[31] L. Susskind and J. Glogower. Quantum mechanical phase and time operator.
Physics, 1 (1), 49-61, (1964).
[32] G. Takeuti. Dirac space. Proc. Japan Acad., 38, 414-418, (1962).
[33] J. Tarski. Short introduction to nonstandard analysis and its physical applica-
tions. In, editor, Many degrees of freedom in field theory, Plenum, New York,
1978.
[34] J. K. Thurber and J. Katz. Applications of fractional powers of delta functions.
In A. Hurd and P. Loeb, editors, Victoria Symposium on Nonstandard Analysis
142

1972. Lecture Notes in Math. Vol. 369, pages 272-302, Springer-Verlag, Berlin,
1974.
[35] H. P. Yuen and M. Lax. Multiple parameter quantum estimation and measure-
ment of nonselfadjoint operators. Trans. IEEE, IT-19, 740-750, (1973).