Sie sind auf Seite 1von 24

The Quantum Measurement Process; natural

emergence of classicality in measurement


processes

Mathijs de Bruin
University of Amsterdam

June 25, 2008


Abstract

This report discusses the role of measurement in quantum mechanics and its
interpretative aspects. It will review a realistic model of the measurement of a
spin- 12 particle, displaying spontaneous disappearance of so-called ‘Schrödinger
cat’-terms from the density matrix of the measured system — thus suggesting
the unnecessity of an explicit ‘collapse’-postulate in the quantum formalism.
Contents
1 Introduction 2

2 Measurement Processes 3

3 State description in Quantum Mechanics 4

4 Schrödinger cat terms 5

5 Realistic Example of a Measurement: a spin- 21 particle 6


5.1 Classical description . . . . . . . . . . . . . . . . . . . . . . . . . 6
5.2 Quantum mechanical description . . . . . . . . . . . . . . . . . . 11
5.3 Disappearance of Schrödinger cat terms . . . . . . . . . . . . . . 12

6 Conclusion 15

A The Density Matrix 16


A.1 Dirac notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A.2 Spinor notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A.3 Observables and probabilities . . . . . . . . . . . . . . . . . . . . 17
A.4 Spin- 12 particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
A.5 Measurable parameters and corresponding state vectors . . . . . 18
A.6 Von Neumann equation . . . . . . . . . . . . . . . . . . . . . . . 19
A.7 Time evolution of spin- 12 particle in constant B . . . . . . . . . . 20
A.8 Multi-partite systems . . . . . . . . . . . . . . . . . . . . . . . . . 21

References 22

1
1 Introduction
In this report we will be investigating some aspects of measurement processes
in quantum mechanics and relate them to macroscopic measurement processes.
Doing so we will review the important role of state representation in quantum
measurement processes.
Stressing the fact that quantum mechanics can only be related to empirical
reality in a statistical manner, we will argue for the Statistical Interpretation of
quantum mechanics (as upheld by Albert Einstein, amongst others) as opposed
to the generally more accepted Copenhagen Interpretation (as popularized by
Niels Bohr). [7, 9, 10, 13]
To emphasize this point we will discuss a realistic example of a spin- 12 mea-
surement process as modeled by Armen Allahverdyan, Roger Balian and Theo
Nieuwenhuizen. [1–4] Of particular interest here is the disappearance of so-called
‘Schrödinger cat’-terms, the off-diagonal elements of the density matrix, causing
the measured quantum system to naturally evolve into a system with a classical
probability distribution — without requiring any measurement postulates.

2
2 Measurement Processes
Before giving a more extensive discussion of the specificities of quantum mea-
surements it seems sensible to point out some aspects of measurement processes
in general. What is it that we do upon performing a measurement, what could
make a measurement distinct form ordinary types of physical interaction?
Let us begin by defining a distinction between the system to be measured
(S) and the measurement apparatus (A). Before we will be performing any
measurement we expect (S) to be in some unknown state described by ρS (0)1 .
Likewise, the initial state of (A) will be described by ρA (0).
For a known state ρS (t) of the system we would like to define q, the quantity
to be measured. Similarly, for (A) we define the quantity Q, the pointer variable,
which can be obtained from a known state ρA (t) of the measurement apparatus.
Now at a time t = θ we would require the state ρA (θ) to be in such a way
correlated to ρS (θ) that upon determining Q we have acquired some information
about q.
Furthermore, we would ideally require the post-measurement correlation be-
tween (S) and (A) to be such that, were we to repeat the measurement with
(S) in the same initial state, the measurement would yield the same value of Q,
regardless of the initial state of (A).
So whereas initially no correlation existed between (S) and (A), some inter-
action between them will allow there to emerge a correlation between Q and q.
In an ideal measurement this would mean that observing a value of the pointer
value Q, for example a measuring rod pointing to a certain value on a scale,
would also provide us with the exact value of q at time t = θ.
Lastly, we would expect the value of Q not to change after the measurement,
so we can afterwards perform a registration its value at t = θ. For example, this
could mean noting down the number of measured events during some (past) time
span from the display of a counter. Here, we immediately see the necessity of
the stability of the macro state of ρA corresponding to the value of the pointer
variable; after the measurement this value should not change anymore as to
allow for the possibility of registration of the result of the measurement.
1 Where ρ (t) denotes an abstract state description of (S), initially with t = 0. Later we will
S
specify ρS (t) to be a density matrix – at this point however this would mean an unnecessary
loss of generality.

3
3 State description in Quantum Mechanics
Whereas classical measurements can be described in a deterministic manner,
in quantum mechanics this is generally not so. The description of a quantum
mechanical measurement yields us only a probability distribution P (q) for a
measurable quantity q.
In order to account for this in the generalized measurement process described
in the previous section, we will need to propagate the statistical nature of q in
(S) to Q in (A). If P (q = x) is defined as the probability of finding x as the
eigenvalue of q, we would ideally expect there to be a unique value of X so that
P (q = x) = P (Q = X). This, however seems to be in conflict with the condition
of registration we posed in the same section; its value should allow for the regis-
tration of Q and it must therefore not adhere to a statistical distribution. This
leaves us to the paradoxical conclusion that, while the description of a measur-
able quantity prior to measurement has the form of a statistical distribution,
the result of a single measurement clearly does not. This transition is what is
often called the ‘collapse’ of a quantum system.
We could try to solve this paradox by assuming a quantum description merely
to adhere to an ensemble of systems in similarly prepared states, instead of
taking it to also apply to a single system which, after measurement, would end up
in single definite state.2 The former corresponds to the statistical interpretation
of quantum mechanics brought forward in [7, 13], whereas the latter is often
associated with the Copenhagen interpretation as implicitly assumed in [11].
In the statistical interpretation of quantum measurements the so-called ‘col-
lapse’ is simply seen as the selection of a subensemble of states corresponding to
the measured value, not an instant ‘magical’ transition from a statistical distri-
bution into a single state as proponents of the Copenhagen interpretation would
like us to believe. Furthermore, by attesting to the statistical interpretation we
can realistically model the process of ‘collapse’ in an actual measurement and
this is exactly what we will be trying to demonstrate in the current report,
based upon the work of Armen Allahverdyan, Roger Balian and Theo Nieuwen-
huizen. [1–4]
Formally, the difference between the two interpretations is reflected by the
kind of state descriptions they consider as being more fundamental. Whereas
the Copenhagen interpretation deems the state P of any system to be exhaus-
tively described by a single state vector |Ψi = k |ϕk i, with ϕk a spectrum
of orthogonal
P states, the statistical interpretation makes use of a density ma-
trix3 ρ = k pk |ϕk i hϕk | where ϕk are not necessarily orthogonal states. Thus
allowing for a complete description of statistically measurable states.4
Corresponding to a single density matrix there can be multiple, mutually
exclusive mixtures of state vectors, while we know a density matrix to contain
all the information that we can measure. This exposes a problem referred to
as the preferred base paradox, where the arbitrary choice of a base seems to
influence the outcome of a measurement.
2 Note that the former assumes all quantum measurements to be repeated and thus statis-

tical in nature which is, given modern day practice, a fair assumption.
3 For a brief introduction into density matrices, please refer to appendix A.
4 As opposed to state vectors, which assume individually measurable states.

4
4 Schrödinger cat terms
While at first sight it might seem fair to view quantum physics as a mere gen-
eralization of classical statistical mechanics, it is easily demonstrated that this
view is incorrect. If we tend to view an individual measurement as the selection
of a subensemble out of an ensemble of similarly prepared systems we run into
problems when our system is prepared in a state which corresponds to none of
the measurable situations.
Consider for example a system (S) consisting of a single spin- 12 particle,
prepared in a pure state along the x-axis. This allows us to exhaustively describe
its state 5
|χi = x↑ |↑(x) i + x↓ |↓(x) i where (1)
x↑ , x↓ ∈ C and |x↑ |2 + |x↓ |2 = 1
but upon a single measurement along the z-axis the only measurable values
are the eigenvalues corresponding to eigenfunctions along that particular axis.
Somehow our system seems to have been changed by our measurement into
|χ0 i = |↑(z) i or |χ0 i = |↓(z) i (2)
depending on what value we actually measured; respectively spin up or down
along the z-axis. Here, we would find the corresponding probabilities to be
P (↑(z) ) = |x↑ − x↓ |2 and P (↓(z) ) = |x↑ + x↓ |2 .
In the Statistical Interpretation this process corresponds to a change in our
density matrix ρ from (see Appendix A.5).
|x↑ − x↓ |2 (x↑ − x↓ ) (x↑ + x↓ )∗
 
1
ρS = (3)
2 (x↑ − x↓ )∗ (x↑ + x↓ ) |x↑ + x↓ |2
to
|x↑ − x↓ |2
 
1 0
ρ0S = . (4)
2 0 |x↑ + x↓ |2
This corresponds to a classical probability distribution with the same proba-
bilities of finding either one of the measurable states. Thus the result of an
individual measurement along the z-axis will not be fundamentally different
from a measurement performed on an analogous classical system.
In conclusion; the state description ρS is changed by our measurement. In
the vector description initially given we see that we can only give account of
a single step in the measurement process; it seems as though our system has
discontinuously evolved from a superposition state into the eigenstate of an
eigenvalue along a different axis. In the density matrix description, however,
we see that the measurement process for an individual measurement actually
consists of two steps. First there is the disappearance of off-diagonal components
from the density matrix, causing the system state to ‘collapse’ to the measured
(z)-basis. Secondly there is the selection of a subensemble pertaining to the
measured state which is, albeit discontinuous, conceptually the same as any
measurement described with statistics.
5 Where |↑(i) i and |↓(i) i are the respective up and down eigenfunctions along an axis (i),

as formulated in Appendix A.4.

5
5 Realistic Example of a Measurement: a spin- 21
particle
In [1–6,12] Armen Allahverdyan, Roger Balian and Theo Nieuwenhuizen propose
an exactly solvable model for measurement of the spin state of a spin- 12 particle
which satisfies the mentioned requirements:

• (A) and (S) should be completely uncorrelated before the measurement.


• After the measurement there should, ideally, be a full correlation between
(A) and (S) for the measured quantity.
• The state of (A) after the measurement should be stable as to allow for
macroscopic registration of the measured value.

In this case, the measurement of a spin- 21 particle will be performed using an


apparatus (A) consisting of a magnet (M) modeled as an Ising-system coupled
to a phonon bath (B) for cooling of (M). The system (S) consists simply of a
spin- 12 particle, coupled only to the spins of (M). A schematic display of the
measurement setup is shown in Figure 1.

Figure 1: Schematic display of the measurement setup. Here (S) represents the measured
system and (A) the measurement apparatus consisting of magnet (M) and phonon bath (B).
The circles are spin- 12 particles and the waves represent harmonic oscillators in the ohmic
bath.

5.1 Classical description


First we will give an explanation of the modeled measurement process in the
classical limit, so we can describe the system using ordinary statistical physics.
Once the general measurement process has become clear, a quantum mechan-
ical model of the same process will be brought forward in which the so-called
Schrödinger cat terms naturally disappear from the density matrix describing
the system (S).
As we have seen before, the physics of a measurement process require (A) to
be initially in a metastable state. After the measurement, we expect to find (A)
in a stable state corresponding to a measurable state of (S). Now when we cool
down a large enough Ising system we can get it to display exactly this behavior;

6
there are exactly two measurable states of (S), up and down, and when the
temperature T lies below the critical temperature Tc for (M), we find that there
exactly two stable states with hmi = m+ ≈ 1 or hmi = m− ≈ −1 and one
metastable state with hmi = 0. This will cause hmi to naturally evolve from 0
into either m+ or m− .
F N -1

-0.18

-0.20

-0.22

-0.24
<m>
-1.0 -0.5 -0.26 0.5 1.0

-0.28

-0.30

Figure 2: Free energy of (A) as a function of hmi with g = 0 for temperatures T = 0.42J
(lower), T = Tc = 0.362949J (middle) and T = 0.3J (top), based on formula (13).

For this process to work (M) should consist of a large number of particles
N  1 having mean-field interactions in quartets in order to have a first order
phase transition. This yields us an exactly solvable phase transition of the
system under the influence of the weak magnetic field caused by the spin- 12
particle in (S), resulting in a correlation between (A) and (S). Because the
eventual state of (S) is stable we will also find the third condition, registrability,
to be satisfied.
We can illustrate this phase transition by calculating the free energy F =
H −T S for (A).6 In this description we will leave the Ohmic bath out as its only
function is to cool the magnet down, enabling it to reach a thermodynamically
stable post-measurement state. We can now proceed and calculate the magnet’s
Hamiltonian H = HM + HSM , where HM is the magnet’s self-Hamiltonian and
HSM the Hamiltonian due to the coupling of (M) to (S) during the measurement.
Since the system-apparatus coupling is initially disabled (so HSM = 0), it
makes sense to start out with HM
N
J X (i) (j) (k) (l)
HM = − σz σz σz σz
4N 3
ijkl=1
1
= − N Jm4 (5)
4
N
1 X (n)
m = σ . (6)
N n=1 z
(n)
Here, σz ∈ {−1, 1} denotes the spin of the individual particles and m repre-
sents the (fluctuating) magnetization. J is the coupling constant for the particles
6 Taking for T the temperature and S as the magnet’s entropy.

7
in the magnet.
Continuing, we will need to find the magnet’s entropy. This is exactly the
same as the entropy for any two-state system with N particles which we can
calculate using Stirling’s approximation
N
N± = (1 ± m) (7)
2
N!
Ω = (8)
N+ !N− !
S = log Ω
= log N ! − log N+ ! − log N− !
(N  1) ≈ N log N − N+ log N+ − N− log N− (9)
 
1+m 2 1−m 2
= N log + log . (10)
2 1+m 2 1−m

Where in the classical case the ensemble average m ≡ hmi. This should be
(close to) 0 initially, in order for an unbiased measurement to be performed.
This can more easily be understood by looking at the graph in Figure 2.
Thus before the system-apparatus coupling, taking into account that in the
classical case m = hmi, the free energy of (M) is

F = H − TS
 
1 4 1+m 2 1−m 2
= − N Jm − N T ln + ln . (11)
4 2 1+m 2 1−m
Now that we have an expression for the free energy of the magnet we can intro-
duce the interaction between (S) and (A) in order to describe the measurement.
Here sz ∈ {+1, −1} corresponds to the spin of the measured particle and g is a
chosen coupling constant
N
X
HSM = −gsz σz(n)
n=1
= −gsz N m (12)

so the Hamiltonian becomes H = HM + HSM ,7 which simply adds up to the


free energy per particle
 
1 4 1+m 2 1−m 2
F/N = − Jm − T ln + ln
4 2 1+m 2 1−m
−gsz m (13)

causing a sole dependency on m and sz . Where the sz dependency causes the


free energy to slope down on the side where the sign of m equals that of sz , as
can be seen in Figure 3. This makes it possible to overcome the peak we have
seen in Figure 2 around |m| = 0.7, enabling the magnet to reach a stable state
with average magnetization of m+ or m− , corresponding to the spin state of
the measured particle.
7 Note that we leave out the self-Hamiltonian H for (S) since the measured particle does
S
not interact with itself.

8
While before the measurement the coupling constant g is zero, during the
measurement it is essential that it has the right nonzero value for (A) to reach
a stable state correlated to the spin state of (S). After either one of these stable
positions has been reached the coupling g between (S) and (A) can be turned
off while (A) will remain in the same macroscopic state.

F N -1 F N -1
-0.10 -0.10

-0.15 -0.15

-0.20 -0.20

<m> <m>
-1.0 -0.5 0.5 1.0 -1.0 -0.5 0.5 1.0

-0.30 -0.30

-0.35 -0.35

-0.40 -0.40

(a) (b)

Figure 3: Free energy as a function of m for T = Tc with sz = +1 for different coupling


constants. (a): g = 0.00 J, g = 0.06 J and g = 0.12 J corresponding to zero, medium and
highest negative slopes of the dashed lines. (b): critical conditions for the phase transition
from m = 0 to m = m+ with g = gc = 0.09035 J.

As we would expect from looking at the free energy, the evolution of the
system can be macroscopically described with an equation involving only m(t).
This is indeed the case, the expected function has been presented in [2]
 
m
ṁ(t) = γh 1 − (14)
tanh βh
1
h = gsz + Jm3 ; β = (15)
kT
with γ  ~−1 a small constant characterizing the coupling to the bath. The
(numerical) solution of this equation complies with our earlier requirements; it
yields a stable value of m ≈ ±1 corresponding to the spin states sz = ±1 of the
measured particle.
This process of time evolution has been illustrated in Figure 4, for different
values of the constants J and γ. As can be seen from the graphs the main
effect of these constants lies in the time it takes for the magnet to enter a
macroscopically stable state (hence; the total time it takes for a measurement
to take place). In Figure 5 we have shown the time it takes for a measurement
to complete as a function of these constants. Notice here that for small values
of J and γ the measurements take increasingly longer, quickly approaching well
over 100s for J < 0.05 or γ ≈ 0.

9
mHtL mHtL
1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

tHsL tHsL
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0

(a) Plot with J = 1 and from left to (b) Plot with γ = 10 and from left to
right γ = 50, 40, . . . , 10. right J = 0.7, 0.8, . . . , 1.1.

Figure 4: Several graphs of the time evolution of the magnetization for sz = +1, with
T = Tc ≈ 0.362949J and g = 0.12J. Due symmetry the situation for sz = −1 would be
completely identical, apart from the sign of m.

J
0.5

0.4

0.3

0.2

0.1
10
40 20
60
70 30
50

Γ
0 20 40 60 80 100

Figure 5: Equipartite lines corresponding to the time it takes for the magnet to evolve from
initial state m = 0 to m = 0.9 as a function of γ and J. The numbers next to the lines are the
times in seconds. Just like before we used T = Tc ≈ 0.362949J and g = 0.12J. Please take
note that the small ripples in the lines are most likely due to rounding errors in our numerical
approximation.

10
5.2 Quantum mechanical description
A quantum mechanical description of the measurement process can be made
analogous to the description in the macroscopic limit. In this case the totality of
system and apparatus is described by a density matrix D(t), which can initially
be written as
D(0) = r(0) ⊗ RA (0) (16)
RA (0) = RM (0) ⊗ RB (0) (17)
with r(0) the initial spin state of the particle in (S) and RA (0) that of (A),
consisting of the magnet state RM (0) and the bath state RB (0).
Analogous to the classical description we require that, initially, hmi = 0 so
that
1  (1) (N )

RM (0) = σ̂ 0 ⊗ ... ⊗ σ̂ 0 (18)
2N 

1 0
σ̂0 = (19)
0 1
which corresponds to an equal superposition of the up and down eigenstates.8
For the aspects of the measurement process being discussed here RB (t), the
description of the bath, can and will be left out of picture.
(n)
The Hamiltonians for (S) and (M) can by found by substitution of σz by
(n)
σ̂z , the Pauli spin operator on the z-basis for a particular particle (n) in the
magnet.9 For HS we can replace sz by ŝz , the same spin operator working on
the basis of the first particle. This results in a Hamiltonian operator on the
basis of the whole system
H = HM + HSM + HB
1
= − N J m̂4 − gŝz N m̂ + HB (20)
4
with
N
1 X (n)
m̂ = σ̂ . (21)
N n=1 z

We can now calculate how the off-diagonal components of the system’s den-
sity matrix r(t) will quickly reduce to zero under the mere influence of the
coupling to the apparatus spins.
This can be done by assuming that for small t the spins in RA (t) are still
separable, so
 
D↑↑ (t) D↑↓ (t)
D(t) =
D↓↑ (t) D↓↓ (t)
 
r↑↑ (0)RA (t) r↑↓ (0)RA (t)
= (22)
r↓↑ (0)RA (t) r↓↓ (0)RA (t)
RA (t) = d(1) (t) ⊗ . . . ⊗ d(N ) (t). (23)
8 See Appendix A.5.
9 See Appendix A.8.

11
where d(m) represents the still separable spin state of the mth particle.
We can do this by writing out the von Neumann equation i~ dρ
dt = [Ĥ, ρ],
10

using only HSM as our Hamiltonian

i~Ḋ(t) = [H, D(t)] = HD(t) − D(t)H


= −gN (ŝz m̂D(t) − D(t)ŝz m̂)
= −gN (m̂ŝz D(t) − D(t)ŝz m̂) (24)
 
D↑↑ (t) D↑↓ (t)
ŝz D(t) = (25)
−D↓↑ (t) −D↓↓ (t)
 
D↑↑ (t) −D↑↓ (t)
D(t)ŝz = . (26)
D↓↑ (t) −D↓↓ (t)

5.3 Disappearance of Schrödinger cat terms


Since we are only interested in partially solving this equation, we can take just
the off-diagonal component of the full density matrix D(t), and write

i~Ḋ↑↓ (t) = −gN (m̂D↑↓ (t) + D↑↓ (t)m̂)


= −gN r↑↓ (0)(m̂RA (t) + RA (t)m̂)
N 
X 
= −g r↑↓ (0) σz(n) RA (t) + RA (t)σz(n) . (27)
n=1

Now we can calculate and solve the differential equation for just two particles
(N = 2), which can then be generalized to any large number particles (N  2).
To do this we will first need to write out the upper-right component of the
density matrix

D↑↓ (t) = r↑↓ (0) [d(1) (t) ⊗ d(2) (t)] (28)


Ḋ↑↓ (t) = r↑↓ (0) [d˙(1) (t) ⊗ d(2) (t) + d(1) (t) ⊗ d˙(2) (t)]. (29)

Substituting this into Equation (27) brings us


h
i~Ḋ↑↓ (t) = −g r↑↓ (0) σ̂z d(1) (t) ⊗ d(2) (t) + d(1) (t)σ̂z ⊗ d(2) (t)
i
+ d(1) (t) ⊗ σ̂z d(2) (t) + d(1) (t) ⊗ d(2) (t)σ̂z
= i~ r↑↓ (0) [d˙(1) (t) ⊗ d(2) (t) + d(1) (t) ⊗ d˙(2) (t)] (30)

allowing us to calculate the differential equation for the density matrix of an


individual particle d(t) in the magnet
 
d↑↑ (t) d↑↓ (t)
d(t) = (31)
d↓↑ (t) d↓↓ (t)
−g[σ̂z d + dσ̂z ] = i~ d˙ (32)
˙ ˙
   
d (t) 0 d (t) d↑↓ (t)
−2g ↑↑ = i~ ˙↑↑ . (33)
0 d↓↓ (t) d↓↑ (t) d˙↓↓ (t)
10 See Appendix A.6.

12
This leads us to a few easily solvable equations

d˙↑↓ (t) = d˙↓↑ (t) = 0 (34)


i~ d˙↑↑ (t) = −2g d↑↑ (t). (35)

Solving these with Equation (18) as border condition, so


1
d(0) = σ̂0 (36)
2
we find
1 2gt/~
d↑↑ (t) = e (37)
2
1 −2gt/~
d↓↓ (t) = e (38)
2
where Equation (38) can be deduced in exactly the same way from D↓↑ .
Finally, for small t the behavior of the off-diagonal components of the mea-
sured particle can be separated by a partial trace over the apparatus basis.
Because this same method works for any N , we find that for large N the time
evolution approaches a Gaussian decay, as shown in Figure 6.
r­¯ HtL r­¯ HtL
1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

2 gt 2 gt
0 1 2 3 4 Ñ 0 1 2 3 4 Ñ

Figure 6: Short-term evolution of the off-diagonal ‘Schrödinger cat’ terms of the density
matrix of (S) (dashed line) and the exponential approximating it for large N (solid line).
Left: N = 2, right: N = 20.

We can calculate this decay as follows

r↑↓ (t) = trM D(t)


= r↑↓ (0)[tr d(n) (t)]N
= r↑↓ (0)[cos(2gt/~) + i sin(2gt/~) + cos(−2gt/~) + i sin(−2gt/~)
= r↑↓ (0)[cos(2gt/~)]N
2 2
g /~2
≈ r↑↓ (0)e−2N t (39)

where trM denotes the partial trace over M as defined in Appendix A.8.

13
2 2
g /~2
That cosN 2gt/~ ≈ e−2N t for large N can be seen by expansion
" ∞
#N
X (−1)n
cosN (2gt/~) = (2gt/~)2n
n=0
(2n)!
≈ [1 − 2g 2 t2 /~2 ]N
≈ 1 − 2N g 2 t2 /~2 (40)

2 2
g /~2
X 1
e−2N t = (−2N t2 g 2 /~2 )n
n=0
n!
≈ 1 − 2N g 2 t2 /~2 (41)

However, the ‘decay’ according to cosN shows regular recurrence. The


Schrödinger cat terms do not seem to disappear at all! We could, though,
make the realistic assumption that there is statistical spread in the values of
the coupling constant g for each particle. This way a spread in phase amongst
the cosines would be introduced, quickly leading to the disappearance of the
off-diagonal terms, as can be seen in Figure 7.
r­¯ HtL
1.0

0.8

0.6

0.4

0.2

2 gt
0 2 4 6 8 10 Ñ

Figure 7: Short-term evolution of off-diagonal terms for the density matrix of (S) with
different coupling constants g for the different particles. The couplings where selected from
a normal distribution with mean µ = 1. The dashed peaks with constant amplitude is for
N = 10 with spread σ = 0, the dashed one with decaying peaks has a deviation σ = 0.1 with
N = 10. The solid line also has σ = 0.1 but for N = 20.

The behavior thus demonstrated is exactly what we had expected; due to


the scale difference between the measurement apparatus (A) and the measured
system (S), the quantum nature of the system disappears upon measurement.
Without any ‘metaphysical’ interference, in the form of ‘magical’ collapse or
otherwise. It seems we have witnessed classicality emerge at macroscopic scales.
Also, the decay of the off-diagonal terms takes place on a very short time
scale — there is no need to give account for more than the possibility of reg-
istration in our measurement process, the final state of the apparatus will be
both stable and classical.

14
6 Conclusion
It seems as though, by adhering to the Statistical Interpretation, we can realisti-
cally model a quantum measurement process. It does not require any postulates
in the formalism and there is no discontinuity in the time evolution of the sys-
tem. However, finding realistic models for quantum measurements like the one
discussed in this report is complicated.
As of now the lack of realism in the description of many quantum measure-
ments does not seem to limit their empirical value. Due to this, the choice of
interpretation seems neutral with regards to the physics concerned. Neverthe-
less this choice has a huge impact on the conceptual framework underlying the
formalism of quantum physics.

15
A The Density Matrix
In the density matrix notation there is the clear distinction between a pure
state, which can be fully represented by a unique state vector in Dirac notation
and a mixed state for which no univocal state vector representation exists. It
follows that a mixed state can at best be described by a density matrix. A more
extensive description of the density matrix formalism can be found in [8].

A.1 Dirac notation


2
P P
Taking a normalized state vector |Ψi = k ak |ψk i ; k |ak | = 1 ,representing
a pure state, we can write its density matrix ρ as follows

ρkk0 = |ψk i hψk0 | (42)


! !
X X
ρ = |Ψi hΨ| = ak |ψk i a∗k0 hψk0 | (43)
k k0
X X
= ak a∗k0 |ψk i hψk0 | = ak a∗k0 ρkk0 (44)
k,k0 k,k0
X X
= |ak |2 ρkk + ak a∗k0 ρkk0 . (45)
k k6=k0

A.2 Spinor notation


In spinor notation, where
   
δ0k a0
|ψk i =  ...  ; |Ψi =  ..  (46)
  
. 
δnk ak
(47)

which amounts to
|a0 |2 a0 a∗1 a0 a∗k
 
  ...
a0 ∗
 ..  a1 a0
  |a1 |2 ... a1 a∗k 
ρ = . a∗0 ∗
. . . ak =  . (48)

.. .. .. 
 .. . . . 
ak
ak a∗0 ak a∗1 ... |ak |2
 
δ0k 0 ... 0
 0 δ1k ... 0 
ρkk =  . (49)
 
.. .. ..
 ..

. . . 
0 0 0 δnk
X
2
T r(ρ) = |ak | = 1. (50)
k

16
A.3 Observables and probabilities
For an observable Ô we find
X
ρÔ = ak a∗k0 ρkk0 Ô (51)
k,k0
X X
= |ak |2 ρkk Ô + ak a∗k0 ρkk0 Ô (52)
k k6=k0
X X
= |ak |2 |ψk i hψk | Ô + ak a∗k0 |ψk i hψk0 | Ô. (53)
k k6=k0

And since we know that in Dirac notation


! !
X X
hÔi = hΨ|Ô|Ψi = a∗k hψk | Ô ak0 |ψk0 i (54)
k k0
X
= ak a∗k0 hψk |Ô|ψk0 i (55)
k,k0
X X
= |ak |2 hψk |Ô|ψk i + hψk |Ô|ψk0 i . (56)
k k6=k0

Now, if {|ψk i} are orthonormal eigenfunctions of Ô with eigenvalues λk


hψk |ψk0 i = δkk0 (57)
hψk |Ô|ψk i = λk hψk |ψk i = λk (58)
thus
X
T r(ρÔ) = |ak |2 λk (59)
k
X
= |ak |2 hψk |Ô|ψk i (60)
k

= hÔi . (61)
Likewise
P (O = λk ) = hψk |Ψ|ψk i = |ak |2 (62)
and
X
P (O ∈ {λk }) = |ak |2 = T r{λk } (ρ) (63)
{k}

with {λk } a set of eigenvalues and T r{λk } the sum of the corresponding diagonal
components.

A.4 Spin- 21 particles

The normalized spin-operator σ̂ for a spin- 12 particle can be written as σ̂ =


(σx , σy , σx ) with σi the respective Pauli-matrices
     
0 1 0 −i 1 0
σx = ; σy = ; σz = (64)
1 0 i 0 0 −1

17
all having eigenvalues {+1, −1} and respective eigenfunctions
   
(x) 1 1 (x) 1 −1
|χ+ i = √ ; |χ− i = √ (65)
2 1 2 1
   
(y) 1 −i (y) 1 i
|χ+ i = √ ; |χ− i = √ (66)
2 1 2 1
   
(z) 1 (z) 0
|χ+ i = ; |χ− i = (67)
0 1
Accordingly any spin state can be written as a superposition of the two eigen-
functions
(i) (i) (i) (i)
|Ψ(i) i = a1 |χ+ i + a2 |χ− i (68)
and its corresponding density matrix would be
!
(i) (i) (i)∗
(i) (i) |a1 |2 a1 a2
ρi = |Ψ i hΨ | = (i)∗ (i) (i) . (69)
a1 a2 |a2 |2
Now suppose we would have a beam of particles consisting of a mixture of pure
states |Ψi along two axes, with respective weights {Wi } then we get
ρ = Wa ρa + Wb ρb (70)
!
(a) (b) (a) (a)∗ (b) (b)∗
Wa |a1 |2 + Wb |a1 |2 Wa a1 a2 + Wb a1 a2
= (a)∗ (a) (b)∗ (b) (a) (b) (71)
Wa a1 a2 + Wb a1 a2 Wa |a2 |2 + Wb |a2 |2
In this case we can calculate the expectation value of the spin for the whole
beam in exactly the same way as we did before
hσ̂i = (hσx i , hσy i , hσz i) = (T r(σx ρ), T r(σy ρ, T r(σz ρ)) (72)
(a) (a)∗ (a)∗ (a) (b) (b)∗ (b)∗ (b)
= (Wa [a1 a2 + a1 a2 ] + Wb [a1 a2 + a1 a2 ],
(a) (a)∗ (a)∗ (a) (b) (b)∗ (b)∗ (b)
iWa [a1 a2 − a1 a2 ] + iWb [a1 a2 − a1 a2 ],
(a) (a) (b) (b)
Wa [|a1 |2 − |a2 |2 ] + Wb [|a1 |2 − |a2 |2 ]. (73)
If we want to calculate the probability of finding a state |ϕi we proceed as we
did before, we ‘sandwich’ the density matrix
P (|ϕi) = hϕ|ρ|ϕi (74)
= hϕ|Wa ρa + Wb ρb |ϕi (75)
= Wa hϕ|ρa |ϕi + Wb hϕ|ρb |ϕi (76)
= Wa hϕ|Ψ(a) i hΨ(a) |ϕi + Wb hϕ|Ψ(b) i hΨ(b) |ϕi (77)
(a) 2 (b) 2
= Wa | hϕ|Ψ i | + Wb | hϕ|Ψ i| (78)

A.5 Measurable parameters and corresponding state vec-


tors
Because we know T r(ρ) = 1, we can express the density matrix in the expecta-
tion values
 
1 1 + hσz i hσx i − i hσy i
ρ = (79)
2 hσx i + i hσy i 1 − hσz i

18
from which it follows that there are only three parameters required to completely
specify the spin state of a spin- 12 ensemble. Furthermore, it becomes clear that
there is generally no unique way to decompose this density matrix in matrices
corresponding to pure states; we cannot simply derive state vectors describing
the system in a univocal way.
For example, consider a density matrix composing a mixture of spin states
on the same axis 11
1 (z) (z) 1 (z) (z)
ρ = |χ i hχ+ | + |χ− i hχ− | (80)
2 +   2 
1 1 0 0 0
= + (81)
2 0 0 0 1
 
1 1 0
= (82)
2 0 1

and the following mixture of states on different axes


1 (z) (z) 1 (z) (z) 1 (x) (x) 1 (x) (x)
ρ = |χ+ i hχ+ | + |χ− i hχ− | + |χ+ i hχ+ | + |χ− i hχ− | (83)
6    6   3   3
1 1 0 0 0 1 1 1 1 −1
= + + + (84)
6 0 0 0 1 6 1 1 −1 1
 
1 1 0
= . (85)
2 0 1

Since both mixtures have the same density matrix, there is no combination of
spin measurements possible that distinguishes between them — and thus there
is no unique combination of state vectors able to describe this system without
making unwarranted assumptions about the state.

A.6 Von Neumann equation


We can also find the time evolution of a density matrix ρ when we know the
corresponding Hamiltonian, using the Von Neumann-equation

i~ = [Ĥ, ρ] (86)
dt
= Ĥρ − ρĤ (87)

which can be seen as a generalization of the Schrödinger equation i~ d|Ψi


dt =
H |Ψi.
11 This example originates from [8], pp. 15.

19
When we take H(t) = H(0) we find
X
|Ψi = ak |ψk i ; H |ψk i = λk (88)
k
X
H |Ψi = ak λk |ψk i = (hΨ| H)† ≡ |Ψ2 i (89)
k
ρ = |Ψi hΨ| (90)
[H, ρ] = Hρ − ρH (91)
= |Ψ2 i hΨ| − |Ψi hΨ2 | (92)
λ0 |a0 |2 λ0 a0 a∗1 . . . λ0 a0 a∗k
 
 λ1 a1 a∗0 λ1 |a1 |2 . . . λ1 a1 a∗k 
= ..  −
 
 .. .. . .
 . . . . 
λk ak a∗0 λk ak a∗1 . . . λk |ak |2
λ0 |a0 |2 λ1 a0 a∗1 . . . λk a0 a∗k
 
λ0 a1 a∗0 λ1 |a1 |2 . . . λk a1 a∗k 
(93)
 
 .. .. .. .. 
 . . . . 
∗ ∗
λ0 ak a0 λ1 ak a1 . . . λk |ak |2
(λ0 − λ1 )a0 a∗1 . . . (λ0 − λk )a0 a∗k
 
0
 (λ1 − λ0 )a1 a∗0 0 ... (λ1 − λk )a1 a∗k 
=  . (94)
 
 .. .. .. ..
 . . . . 
(λk − λ0 )ak a∗0 (λk − λ1 )ak a∗1 ... 0

A.7 Time evolution of spin- 21 particle in constant B

For the spin- 12 particle from Equation (79) in a constant magnetic field B =
(0, 0, B0 ) the Hamiltonian is
H = µḂ (95)
= B0 σz (96)
and so, taking u(t) = hσu i, we find
  
B0 1 0 1 + z(t) x(t) − iy(t)
Hρ − ρH =
2 0 −1 x(t) + iy(t) 1 − z(t)
  
B0 1 + z(t) x(t) − iy(t) 1 0
− (97)
2 x(t) + iy(t) 1 − z(t) 0 −1
 
0 x(t) − iy(t)
= B0 (98)
−x(t) − iy(t) 0
0
x (t) − iy 0 (t)
0
 
i~ z (t)
= . (99)
2 x0 (t) + iy 0 (t) −z 0 (t)
This can be solved to
x(t) − iy(t) = Ae−2iB0 t/~ (100)
= (x(t) − iy(t))∗ (101)
z(t) = B (102)
A, B ∈ R (103)

20
which is exactly what we would have found using the time dependent Schrödinger
equation for a constant H.

A.8 Multi-partite systems


The description of multi-partite systems in the density matrix formalism is
much alike the state vector formalism. For two separable systems (A) and (B)
described by density matrices ρA and ρB we would write the combined density
matrix as

ρ = ρA ⊗ ρB (104)

which would, for two spin- 21 systems, imply the Kronecker product between the
respective density matrices
   
a1 b1 a2 b2
ρ = ⊗ (105)
c1 d1 c2 d 2
 
a1 a2 a1 b2 a2 b1 b1 b2
 a1 c2 a1 d2 b1 c2 b1 d2 
=   a2 c1 b2 c1 a2 d1 b2 d1  .
 (106)
c1 c2 c1 d2 c2 d1 d1 d2

Holding on to the same condition of separability we can also retrieve the original
state of either one of the particles by ‘tracing out’ other particles

trB ρ = ρA (107)
trA ρ = ρB (108)

which generalizes well to systems with more particles.


Similarly, the Pauli spin operator for the (z)-axis working on the second
particle might be represented as

σ̂z(B) = σ̂0 ⊗ σ̂z (109)


   
1 0 1 0
= ⊗ (110)
0 1 0 −1
 
1 0 0 0
 0 −1 0 0 
= 
 0 0
 (111)
1 0 
0 0 0 −1

which would generalize to


(1) (n−1) (n+1) (N )
σ̂z(n) = σ̂0 ⊗ . . . ⊗ σ̂0 ⊗ σz(n) ⊗ σ0 ⊗ . . . ⊗ σ̂0 (112)

for an N -particle system.

21
References
[1] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. Curie-weiss
model of the quantum measurement process. Europhysics Letters (EPL),
61(4):452–458, 2003.

[2] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. The quan-


tum measurement process: an exactly solvable model. arXiv.org:cond-
mat/0309188, 2003.
[3] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. Determin-
ing a quantum state by means of a single apparatus. Phys. Rev. Lett.,
92(12):120402, Mar 2004.
[4] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. The quan-
tum measurement process in an exactly solvable model. arXiv:cond-
mat/0408316v1, 2004.

[5] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. Dynamics of a


quantum measurement. Physica E: Low-dimensional Systems and Nanos-
tructures, 29(1-2):261–271, 2005.
[6] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen. The quan-
tum measurement process: Lessons from an exactly solvable model.
arXiv:quant-ph/0702135v2, 2007.
[7] L. E. Ballentine. The statistical interpretation of quantum mechanics. Rev.
Mod. Phys., 42(4):358–381, Oct 1970.
[8] K. Blum. Density Matrix Theory and Applications (Physics of atoms and
molecules). Plenum Pub Corp, 1981.

[9] N. Bohr. Can quantum-mechanical description of physical reality be con-


sidered complete? Phys. Rev., 48(8):696–702, Oct 1935.
[10] A. Einstein, B. Podolsky, and N. Rosen. Can quantum-mechanical descrip-
tion of physical reality be considered complete? Phys. Rev., 47(10):777–
780, May 1935.
[11] D. J. Griffiths. Introduction to Quantum Mechanics (2nd Edition). Ben-
jamin Cummings, 2004.
[12] A. Khrennikov, editor. The quantum measurement process in an exactly
solvable model, volume 750. AIP, 2005.

[13] K. R. Popper. Quantum mechanics without the observer. In M. Bunge,


editor, Quantum Theory and Reality. Springer, 1967.

22

Das könnte Ihnen auch gefallen