pre- and post-selection paradoxes, measurement-disturbance and contextuality in quantum mechanics
DESCRIPTION
Talk from circa 2004 on pre- and post-selection paradoxes in quantum theory and their relationship to contextuality. Based on the results of Phys. Rev. Lett. 95, 200405 (2005) (http://arxiv.org/abs/quant-ph/0412178) and Proceedings of QS 2004, Int. J. Theor. Phys. 44:1977-1987 (2005) (http://arxiv.org/abs/quant-ph/0412179). This talk was recorded and can be viewed online at http://pirsa.org/04110008/TRANSCRIPT
April 13, 2023 M. S. Leifer - Perimeter Institute
1
Pre- and Post-Selection Paradoxes, Measurement-Disturbance and
Contextuality in Quantum Mechanics
M. S. Leifer
(joint work with R. W. Spekkens)
10th November 2004
Coming soon to an arXiv/quant-ph near you!
April 13, 2023 M. S. Leifer - Perimeter Institute
2
Background and Motivation
Pre- and Post-Selection was introduced by Aharonov, Bergmann and Lebowitz (ABL) in 1964.
Two-State-Vector Formalism (TSVF) (see Aharonov & Vaidman, 2001)
Foundational motivation - Time symmetry in the quantum measurement process.
Pre-selection
MeasurementPost-
selection
tpre t tpost
time
pre postMeasurement
April 13, 2023 M. S. Leifer - Perimeter Institute
3
Background and Motivation
Suggests possibilities for interpretation of QM.
Reality of both states - Aharonov (2001):
“Even at present, before the ‘future’ measurements, the backward evolving quantum state [...] exists! It exists in the same way as the quantum state evolving from the past exists. An element of arbitrariness: ‘Why this particular outcome and not some other?’ might discourage, but the alternative [...] – the collapse of the quantum wave – is clearly worse than that.”
Counterfactual interpretation - Mohrhoff (2000):
“ABL probabilities are based on a complete set of facts, and are therefore objective, only if none of the measurements to the possible results of which they are assigned is actually performed (that is, only if between the ‘preparation’ or preselection and the ‘retroparation’ or postselection no measurement is performed).”
pre post
pre postMeasurement
April 13, 2023 M. S. Leifer - Perimeter Institute
4
Background and Motivation
The counterfactual interpretation has several difficulties
We will see that “Logical PPS paradoxes” would imply that reality is contextual.
Further, a similar interpretation of “classical” probability would also imply that reality is contextual.
Reason - It does not take the possibility that measurements might disturb the system into account.
There is no compelling reason to treat the ABL probabilities counterfactually.
See Kastner Phil. Sci. 70, 145 (2003) for refs.
Main results:
Logical PPS paradoxes can be explained in noncontextual hidden variable theories, provided measurements disturb the values of the hidden variables.
However, for every logical PPS paradox there is a related proof of contextuality, which can be constructed from the same set of measurements.
April 13, 2023 M. S. Leifer - Perimeter Institute
5
Why should we care about PPS systems? Despite the controversy over interpretation, the TSVF
agrees with QM for actual measurements.
TSVF has yielded results of interest to quantum information theorists: The Mean King’s Problem (Vaidman et. al., 1987) Cryptography protocols (Bub, 2000, Botero & Reznik,
1999)
The formalism has been extended to more general types of pre- and post-selection (Aharovov & Reznik, 1995). e.g. correlated pre- and post-selections of the type used in
standard quantum cryptography (BB84, B92, etc.)
There is an analog of entanglement, which might be regarded as a resource for qinfo.
It seems likely that many PPS effects could be exploited in quantum information protocols. To find out if this is the case, we need to know whether
these effects can be simulated “classically” and, if so, how efficiently? This means that we should re-examine the foundational questions with this in mind.
April 13, 2023 M. S. Leifer - Perimeter Institute
6
Outline
1) Operational Notation
2) Pre- and Post-Selection in Quantum Theory1) Quantum Measurement Formalism
2) The ABL Rule
3) The “Three-Box Paradox”
3) Hidden Variable Theories1) The “Partitioned-Box Paradox”
2) General Formalism of HVTs
3) Pre- and Post-Selection in HVTs
4) Noncontextuality
5) Noncontextuality, PPS and Disturbance
4) A Surprising Theorem1) Example: “Three-box” and Clifton
2) Example: “Failure of the Product Rule”
5) Conclusions and Open Questions
April 13, 2023 M. S. Leifer - Perimeter Institute
7
1) Operational Notation
M - A measurement.
XM2{1, 2, ..., n} - The outcome of M.
pre/post - Refers to pre/post-selection.
Apre/Apost - The occurrence of successful pre/post-selection.
Goal: Compute p(XM = j | Apre, Apost, M)
Pre-selection
MPost-
selection
Apre Apost
XM
April 13, 2023 M. S. Leifer - Perimeter Institute
8
2) Pre- and Post-Selection in Quantum Theory2.1) Quantum Measurement Formalism
States: Density operators, , on ad-dimensional H. S.
Sharp quantum measurements
Statistical aspect of M given by a PVM {PM,j}
PM,j2 = PM,j j PM,j = I
Born rule: p(XM = j) = Tr(PM,j)
Transformation aspect given by CP-maps {EM,j}
EM,j#(I) = PM,j where Tr(E#(A)B) = Tr(AE(B))
On obtaining XM = j:
! EM,j() / Tr(EM,j())
Lüders Rule (projection postulate)
EM,j() = PM,j PM,j
April 13, 2023 M. S. Leifer - Perimeter Institute
9
2.2) The ABL Rule
Assumptions Density matrix is I/d prior to pre-selection. Pre-selection measurement obeys Lüders Rule.
Notation
Apre (Apost) corresponds to a projector pre (post).
The measurement M corresponds to a PVM {PM,j}
and CP-maps {EM,j}.
kpreM,post
preM,post
preMMprepost
preMMprepost
postpreM
ΠΠTr
ΠΠTr
M,AM,,AA
M,AM,,AA
M,A,A
k
j
k
kXpkXp
jXpjXp
jXp
E
E
April 13, 2023 M. S. Leifer - Perimeter Institute
10
2.2) The ABL Rule
Special cases:
Intermediate Lüders rule
Rank-1 pre- and post-selection projectors
pre = | pre i h pre | post = | post i h post |
kkk
jj
PP
PPjXp
M,preM,post
M,preM,postpostpreM Tr
TrM,A,A
kk
j
P
PjXp 2
preM,post
2
preM,post
postpreM
ψψ
ψψM,A,A
April 13, 2023 M. S. Leifer - Perimeter Institute
11
2.3) The “Three-Box Paradox”
| 1 i | 2 i | 3 i
Pre-selection: |pre i = | 1 i + | 2 i + | 3 i
Post-selection: |post i = | 1 i + | 2 i - | 3 i
Intermediate measurements:
M: {PM,1 = | 1 ih 1 |, PM,2 = | 2 i h 2 | + | 3 ih 3 | }
N: {PN,1 = | 2 ih 2 |, PN,2 = | 1 ih 1 | + | 3 ih 3 | }
p(XM = 1 | Apre, Apost, M) = p(XN = 1 | Apre, Apost, N) = 1
Reason: h post | PM,2 | pre i = h post | PN,2 | pre i = 0
Counterfactual interpretation: David Blaine is in both boxes at the same time!
April 13, 2023 M. S. Leifer - Perimeter Institute
12
3) Hidden Variable Theories
3.1) The “Partitioned-Box Paradox”
Left verification
Similarly for Right Front and Back verification.
F
B
L R
or
?
? ?
?
April 13, 2023 M. S. Leifer - Perimeter Institute
13
3.1) The “Partitioned Box Paradox”
Pre-selection: Successful Front Verification.
Post-selection: Successful Back Verification
Possible Intermediate measurements:
Left Verification or Right Verification
LV case
Counterfactual interpretation: The ball is both on the left and the right.
Correct interpretation: Different measurements disturb the ball in different ways.
??
??
?
?
April 13, 2023 M. S. Leifer - Perimeter Institute
14
3.2) General Formalism of HVTs
Set of ontic states 2
States correspond to probability distributions ().
Assumption: Outcome determinism for sharp measurements.
Measurement M corresponds to:
Statistical aspect: indicator functions {M,j}
M,j() = 0 or 1 j M,j() = 1
p (XM = j) = s M,j() () d
Transformation aspect: Stochastic transformations {DM,j}
DM,j(,) is prob. of transition from to
Define transition matrices M,j(,) = DM,j(,) M,j()
sM,j(,) d = M,j()
April 13, 2023 M. S. Leifer - Perimeter Institute
15
3.2) General Formalism of HVTs
Update rule
3.3) Pre- and Post-Selection in HVTs
Pre-selection prepares pre.
Post-selection corresponds to post.
Intermediate measurement M: {M,j}, {M,j}
By Bayes' Theorem:
dλdωωμωλ,
dωωμωλ,λμ
M,
M,
M
j
jjX
kk
jjXp
dωdλωμωλ,Γλχ
dωdλωμωλ,ΓλχM,A,A
preM,post
preM,post
postpreM
April 13, 2023 M. S. Leifer - Perimeter Institute
16
3.4) Noncontextuality
If there is an outcome j of M and an outcome k of N that have the same probability for all preparations then
M,j = N,k
In QM equivalence both outcomes associated to the same projector P.
M,j = N,k = P
projectors associated to unique properties.
for any distribution can assign probabilities to projectors via p(P) = sP()()d
Algebraic constraints
0,1
1
10
null
PpIp
PpPIp
Pp PQpQpPpPQQPp
QpPQpPpPQp
QP
,
0,For
April 13, 2023 M. S. Leifer - Perimeter Institute
17
3.5) Noncontextuality, PPS and Disturbance We can also try to apply these constraints to ABL
probabilities.
Definition: We have a Logical PPS paradox whenever the ABL rule assigns probability 1 or 0 to the outcomes of measurements, such that associating the same probabilities to the projectors corresponding to those outcomes would violate the algebraic constraints.
The Three-Box paradox is an example of this.
You could explain Logical PPS paradoxes via contextuality. However, this is not required as our “partitioned box” example shows.
Reproducing ABL predictions does place nontrivial constraints on the transitions M,j.
They must be nontrivial M,j(,). (,) M,j().
Outcomes associated with the same projector, but different CP-maps, must be associated with different transitions.
April 13, 2023 M. S. Leifer - Perimeter Institute
18
4) A Surprising Theorem
Whenever a set of projectors have ABL probabilities that give rise to a logical PPS paradox, and the pre- and post-selection projectors are not orthogonal, one can use fine grainings of the projectors, together with the pre- and post-selection projectors to show that a MNHVT is impossible.
4.1) Three box paradox and Clifton’s proof
1
1
1
1
1
1
0
0
1
1
1
0
1
1
0
0
1
0
1
0
1
1
0
1
pre post
April 13, 2023 M. S. Leifer - Perimeter Institute
19
4.2) Failure of the Product Rule
For observables B = j j Pj in an MNHVT we can
define values by
v(B) = k iff p(Pk) = 1
Algebraic conditions imply that for [A,B] = 0
v(B) + v(C) = v(B + C) Sum Rule
v(BC) = v(B)v(C) Product Rule
Similarly we can define ABL values
vABL(B) = k iff p(XB = j | Apre, Apost, B) = 1
These can violate Sum and Product rules even in noncontextual models, but whenever they do we can find a related proof of contextuality.
April 13, 2023 M. S. Leifer - Perimeter Institute
20
4.2) Failure of the Product Rule
Pre-select: | pre i = 2-1/2 ( | 0 i | 0 i + | 1 i | 1 i )
Post-select: | post i = | 0 i | + i
Intermediate observables:
X I vABL(X I) = 1
I Z vABL(I Z) = 1
X Z vABL(X Z) = -1
1
01
0
10
10
01
01
10
10
01ZXI
102 21
April 13, 2023 M. S. Leifer - Perimeter Institute
21
4.2) Failure of the Product Rule
pre post
0
0
11
10
1
102 21
102 21
Projectors onto Eigenvalues
Observable +1 -1
X I P+ P-
I Z Q+ Q-
X Z R+ = P+ Q+ + P- Q- R+ = P+ Q- + P- Q+
Q-
P-
R+
April 13, 2023 M. S. Leifer - Perimeter Institute
22
5) Conclusions and Open Questions
We have shown that:
The existence of Logical PPS paradoxes in a theory does not imply contextuality.
Nonetheless, each Logical PPS paradox is related to a proof of contextuality.
Possible reason for discrepancy:
HVTs have to satisfy additional constraints in order to reproduce quantum predictions.
Conjecture: Existence of Logical PPS paradoxes + some suitable analog of the uncertainty principle implies contextuality.
Justification: Toy theory of Spekkens has such an analog, is noncontextual and seems to be devoid of logical PPS paradoxes.
Applications of logical PPS paradoxes in quantum information theory?
Use of our results in axiomatics of quantum theory?