Hartle’s Frequency Operator Revisited

Patrick Das Gupta

Department of Physics and Astrophysics,
University of Delhi, Delhi - 110 007
e-mail: patrick@srb.org.in

1 Introduction
According to quantum theory, the physical state of a quantum system is represented
by a normalized state-vector lying in a Hilbert space, and in the absence of any mea-
surement performed on the system, as time passes, the state evolves unitarily in a
deterministic manner, ordained by the Hamiltonian of the given system. If a measure-
ment of an observable of the system is made at some instant, its outcome in general
cannot be predicted with certainty except that it is always one of the eigenvalues of
the self-adjoint operator corresponding to the measured observable. However, mak-
ing use of the Born rule, the probability of finding a particular eigenvalue to be the
outcome of the measurement can definitely be forecast. These aspects of quantum
theory have been verified in myriads of experiments.
There is a subtle difference between the roles played by observables in classical
and quantum theories, other than the limitation imposed by uncertainty principle in
the case of a pair of non-commuting observables in the latter. In classical mechanics,
the operational definition of an observable itself suggests the manner in which one
should proceed to measure it. To take an example, if one were to measure the velocity
of a particle at time t, one could just measure the positions at times t and t + ∆t, for
an infinitesmally tiny ∆t, then take the difference and divide it by ∆t, since that is
how velocity is defined in the first place.
While in quantum theory, an observable has an abstract representation in the
form of a hermitian operator which by itself does not suggest a way to measure it.
In order to measure the self-adjoint operator, one frequently needs to invoke classical
concepts and design a suitable interaction between the quantum observable (or,a
related operator like its canonical conjugate etc.) and a pointer variable. To cite
an example, the all too familiar spin angular momentum operator Ŝ = 21 h̄⃗σ for an
electron (σi , i=1,2,3, being the Pauli matrices) does not give us any clue as to how to
measure it. For the measurement of electron spin in a Stern-Gerlach experiment like

1
set up, one relies on the classical notion that a spinning charge particle is associated
with a magnetic moment ⃗µ so that in the presence of an inhomogeneous magnetic
⃗ r), the interaction energy is ⃗µ.B
field B(⃗ ⃗ which leads to a coupling between the spin
degree of freedom and the position (‘pointer variable’) of the electron.
Although from a practical point of view quantum mechanics is a very successful
theory, issues related to measurements continue to perplex one since the inception of
the theory. For instance, any measurement after all is a physical interaction between
the system and the apparatus involving a definite interaction Hamiltonian, so that
the combined system ought to evolve with time according to the Schrodinger equation
in a deterministic way. But then, why are outcomes of any observation probabilistic
in nature? Stated differently, from where does the Born rule come from? In the
framework of Many-Worlds Interpretation (MWI) of quantum mechanics (QM), there
have been several attempts to obtain the Born rule without invoking any probabilistic
”collapse of the state-vector to an eigenstate” during a measurement [1,2]. In another
context, Hartle had described an elegant scheme of treating an ensemble of infinite
number of identical systems as an individual quantum system with a well-defined
frequency observable [3]. In the following section, we discuss the status of Born rule
in Hartle’s approach.

2 Hartle’s frequency operator

We begin with a quantum system described by a Hilbert space H and an observable
K̂ defined on it. For simplicity, we assume that K̂ has a discrete spectrum so that,

K̂ = ki |i⟩, i = 1, 2, 3, ..... (1a)

with,
⟨i|j⟩ = δij (1b)

and,
∑
∞
|i⟩ ⟨i| = 1 , (1c)
i=1

where ki are the eigenvalues that are observed when K̂ is measured.

A physical state of the system is described by |ψ⟩, which is an element of H, and

2
can be linearly expanded in terms of the orthonormal basis vectors {|i⟩, i = 1, 2 . . . },

∑
∞
|ψ⟩ = ci |i⟩ , (2a)
i=1

with,
⟨ψ|ψ⟩ = 1 , (2b)

and,
ci = ⟨i|ψ⟩ . (2c)

Since, predictions of QM concerning outcomes of measurements have been tested by

making use of large number of identically prepared systems, we need to formulate the
problem accordingly. An ensemble of N identical systems in QM is represented by the
Hilbert space formed out of the tensor product of individual spaces H ×H ×· · ·×H ≡
H N . If we consider now N identical systems, each being specified by the state-vector
|ψ⟩, then as an individual quantum system, the ensemble is described by an element
of H N given by the direct product of |ψ⟩s,

|(ψ)N ⟩ ≡ |ψ⟩1 |ψ⟩2 . . . |ψ⟩N (3a)

In the limit N → ∞, the state-vector for the ensemble is given by,

|(ψ)∞ ⟩ ≡ |ψ⟩1 |ψ⟩2 . . . |ψ⟩N |ψ⟩N +1 . . . (3b)

For the lth system of the ensemble, we use the notation in which |il ⟩ is an eigenstate
of K̂ with eigenvalue kil , so that {|il ⟩, il = 1, 2, . . . } is an orthonormal basis cor-
responding to the lth system (see eqs.(1a)-(1c)). The Hilbert space H N is therefore
spanned by the orthonormal vectors {|i1 ⟩|i2 ⟩ . . . |iN ⟩, i1 , i2 , · · · = 1, 2, . . . }.
These direct product of eigenstates can be used to construct a frequency operator
FNj for the eigenvalue kj of K̂ as follows,
∑
F̂Nj ≡ fj |i1 ⟩|i2 ⟩ . . . |iN ⟩⟨iN |⟨iN −1 | . . . ⟨i2 |⟨i1 | , (4a)
i1 ,i2 ,...,iN

where,
1 ∑
N
fj ≡ δji (4b)
N l=1 l
is clearly the frequency of il being equal to j in {i1 , i2 , . . . , iN }.

3
 It is easy to see that |i′1 ⟩|i′2 ⟩ . . . |i′N ⟩ is an eigenstate of FNj corresponding to the
eigenvalue being the frequency of i′l equal to j for l = 1, 2, . . . , N in {i′1 , i′2 , . . . , i′N },
since from eqs.(4a-b) and the orthonormality of {|il ⟩, il = 1, 2, . . . } we get,

∑ 1 ∑
N
F̂Nj |i′1 ⟩|i′2 ⟩ . . . |i′N ⟩ = δji |i1 ⟩|i2 ⟩ . . . |iN ⟩δi1 i′1 δi2 i′2 . . . δiN i′N (5a)
i1 ,i2 ,...,iN
N l=1 l

( )
1 ∑
N
′ ′ ′
= δji′ |i ⟩|i ⟩ . . . |iN ⟩ , (5b)
N l=1 l 1 2

thus, vindicating that FNj indeed is a frequency operator.

For later purposes, it is useful to express the frequency operator as,
( )
1 ∑
j
F̂N = |i1 ⟩|i2 ⟩ . . . |iN ⟩ δji1 + δji2 + · · · + δjiN ⟨iN |⟨iN −1 | . . . ⟨i2 |⟨i1 |
N i ,i ,...,i
1 2 N

{ ∑
1
= |j⟩11 ⟨j| |i2 ⟩⟨i2 ||i3 ⟩⟨i3 | . . . |iN ⟩⟨iN |+
N i2 ,i3 ,...,iN
∑
+|j⟩22 ⟨j| |i1 ⟩⟨i1 ||i3 ⟩⟨i3 | . . . |iN ⟩⟨iN | + · · · +
i1 ,i3 ,...,iN

∑ }
+|j⟩N N ⟨j| |i1 ⟩⟨i1 ||i2 ⟩⟨i2 | . . . |iN −1 ⟩⟨iN −1 |
i1 ,i2 ,...,iN −1
{ }
1
= |j⟩11 ⟨j| + |j⟩22 ⟨j| + · · · + |j⟩N N ⟨j| (6)
N
The last step follows from eq.(1c). The operation of F̂Nj on |(ψ)∞ ⟩ is defined by,

F̂Nj |(ψ)∞ ⟩ ≡ (F̂Nj |(ψ)N ⟩)|ψ⟩N +1 |ψ⟩N +2 . . . (7a)

Using eq.(6) and eq.(2c), we obtain,

{
cj
F̂N |(ψ) ⟩ =
j N
|j⟩1 |ψ⟩2 |ψ⟩3 . . . |ψ⟩N + |ψ⟩1 |j⟩2 |ψ⟩3 . . . |ψ⟩N +
N
}
+ · · · + |ψ⟩1 |ψ⟩2 . . . |ψ⟩N −1 |j⟩N , (7b)

so that, from eq.(7a), we have,

{
∞ cj
F̂N |(ψ) ⟩ =
j
|j⟩1 |ψ⟩2 |ψ⟩3 . . . |ψ⟩N + |ψ⟩1 |j⟩2 |ψ⟩3 . . . |ψ⟩N +
N

4
 }
+ · · · + |ψ⟩1 |ψ⟩2 . . . |ψ⟩N −1 |j⟩N |ψ⟩N +1 |ψ⟩N +2 . . . (7c)

Following Hartle [3], we may ask how close is the state-vector F̂Nj |(ψ)∞ ⟩ to |cj |2 |(ψ)∞ ⟩
in the limit N → ∞? Now,
( )
||F̂Nj |(ψ)∞ ⟩−|cj |2 |(ψ)∞ ⟩||2 ≡ F̂Nj |(ψ)∞ ⟩−|cj |2 |(ψ)∞ ⟩, F̂Nj |(ψ)∞ ⟩−|cj |2 |(ψ)∞ ⟩ (8a)
( ) ( )
= F̂Nj |(ψ)∞ ⟩, F̂Nj |(ψ)∞ ⟩ − 2|cj | 2
|(ψ) ∞
⟩, F̂Nj |(ψ)∞ ⟩ + |cj |4 (8b)

The last term in eq.(8b) follows from the fact that ⟨(ψ)∞ |(ψ)∞ ⟩ = 1 because of eqs.
(2b) and (3b). Again, eqs.(2b), (2c), (3b) and (7c) leads to,
( ) { }
∞ ∞ cj
|(ψ) ⟩, F̂N |(ψ) ⟩ =
j
1 ⟨ψ|j⟩1 + 2 ⟨ψ|j⟩2 + · · · + N ⟨ψ|j⟩N = |cj |2 (8c)
N
From eqs.(7c) and (2b), we get,
( ) {
∞ ∞ |cj |2
F̂N |(ψ) ⟩, F̂N |(ψ) ⟩ = 2 1 ⟨j|2 ⟨ψ|3 ⟨ψ| . . . N ⟨ψ|+
j j
N
}
+1 ⟨ψ|2 ⟨j|3 ⟨ψ| . . . N ⟨ψ| + · · · + 1 ⟨ψ|2 ⟨ψ|3 ⟨ψ| . . . N −1 ⟨ψ|N ⟨j| ×
{ }
× |j⟩1 |ψ⟩2 |ψ⟩3 . . . |ψ⟩N + |ψ⟩1 |j⟩2 |ψ⟩3 . . . |ψ⟩N + · · · + |ψ⟩1 |ψ⟩2 . . . |ψ⟩N −1 |j⟩N (9a)

|cj |2
{N + N (N − 1)|cj |2 }
= (9b)
N2
Hence, employing eqs.(8c) and (9b) in eq.(8b), entails,
|cj |2
||F̂Nj |(ψ)∞ ⟩ − |cj |2 |(ψ)∞ ⟩||2 = {1 − |cj |2 } (9c)
N
From eq.(9c), it is evident that as N → ∞, we have ||F̂Nj |(ψ)∞ ⟩ − |cj |2 |(ψ)∞ ⟩||2 → 0.
This is a remarkable result in the sense that no matter what |ψ⟩ is, for every
eigenvalue kj of the observable K̂, the distance between the state F̂Nj |(ψ)∞ ⟩ and the
Born probability times |(ψ)∞ ⟩ can be made arbitrarily small by considering sufficiently
large ensemble. But this by no means implies that as N → ∞, the state-vector
F̂Nj |(ψ)∞ ⟩ → |cj |2 |(ψ)∞ ⟩. In fact, Squires’ paper demonstrates that the vanishing of
the left hand side of eq.(9c) does not entail that F̂Nj |(ψ)∞ ⟩ = |cj |2 |(ψ)∞ ⟩ as N → ∞
[4]. As much is hinted by the expression in the right hand side of eq.(7b).
The hope of obtaining the Born probabilities |cj |2 , j = 1, 2, . . . as eigenvalues of
the frequency operator remains unfulfilled in this approach. Hence, the enigma of the
Born rule continues to be wrapped in a riddle!

5
 References
[1] Everett, H., 1957, ”Relative State Formulation of Quantum Mechanics”, Re-
views of Modern Physics, 29: 454-462
[2] DeWitt, B. S., and N. Graham (eds.), 1973, The Many-Worlds Interpretation
of Quantum Mechanics, Princeton: Princeton University Press.
[3] Hartle, J. B., 1968, ”Quantum Mechanics of Individual Systems”, Am. Jour.
Phys., 36: 704-712
[4] Squires, E. J., 1990, Phys.Lett.A, 145, 67