Tsirelson's problem and linear system games
Transcript of Tsirelson's problem and linear system games
![Page 1: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/1.jpg)
Tsirelson’s problem and linear system games
William Slofstra
IQC, University of Waterloo
January 20th, 2017
includes joint work with Richard Cleve and Li Liu
Tsirelson’s problem and linear system games William Slofstra
![Page 2: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/2.jpg)
A speculative question
Conventional wisdom: Finite time / volume / energy / etc. =)can always describe nature by finite-dimensional Hilbert spaces
But... many models in quantum mechanics and quantum field
theory require infinite-dimensional Hilbert spaces (e.g. CCR)
Could nature be “intrinsically” infinite-dimensional?
Answer: Probably not
But if it was... could we recognize that fact in an experiment?
(For instance, in a Bell-type experiment?)
Tsirelson’s problem and linear system games William Slofstra
![Page 3: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/3.jpg)
A speculative question
Conventional wisdom: Finite time / volume / energy / etc. =)can always describe nature by finite-dimensional Hilbert spaces
But... many models in quantum mechanics and quantum field
theory require infinite-dimensional Hilbert spaces (e.g. CCR)
Could nature be “intrinsically” infinite-dimensional?
Answer: Probably not
But if it was... could we recognize that fact in an experiment?
(For instance, in a Bell-type experiment?)
Tsirelson’s problem and linear system games William Slofstra
![Page 4: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/4.jpg)
A speculative question
Conventional wisdom: Finite time / volume / energy / etc. =)can always describe nature by finite-dimensional Hilbert spaces
But... many models in quantum mechanics and quantum field
theory require infinite-dimensional Hilbert spaces (e.g. CCR)
Could nature be “intrinsically” infinite-dimensional?
Answer: Probably not
But if it was... could we recognize that fact in an experiment?
(For instance, in a Bell-type experiment?)
Tsirelson’s problem and linear system games William Slofstra
![Page 5: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/5.jpg)
A speculative question
Conventional wisdom: Finite time / volume / energy / etc. =)can always describe nature by finite-dimensional Hilbert spaces
But... many models in quantum mechanics and quantum field
theory require infinite-dimensional Hilbert spaces (e.g. CCR)
Could nature be “intrinsically” infinite-dimensional?
Answer: Probably not
But if it was... could we recognize that fact in an experiment?
(For instance, in a Bell-type experiment?)
Tsirelson’s problem and linear system games William Slofstra
![Page 6: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/6.jpg)
Non-local games (aka Bell-type experiments)
Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
Win/lose based on outputs a, band inputs x , y
Alice and Bob must cooperate
to win
Winning conditions known in
advance
Complication: players cannot
communicate while the game is
in progress
Tsirelson’s problem and linear system games William Slofstra
![Page 7: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/7.jpg)
Non-local games ct’d
Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
Suppose game is played many
times, with inputs drawn from
some public distribution ⇡
To outside observer, Alice and
Bob’s strategy is described by:
P(a, b|x , y) = the probability of
output (a, b) on input (x , y)
Correlation matrix: collection of
numbers {P(a, b|x , y)}
Tsirelson’s problem and linear system games William Slofstra
![Page 8: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/8.jpg)
What can P(a, b|x , y) be?Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
P(a, b|x , y) = the probability of output (a, b) oninput (x , y)
n questions, m answers: {P(a, b|x , y)} ⇢ Rm
2
n
2
Classically
P(a, b|x , y) = pxa
· qyb
Probability that Alice outputs a on input x
Same for Bob
Tsirelson’s problem and linear system games William Slofstra
![Page 9: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/9.jpg)
What can P(a, b|x , y) be?Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
P(a, b|x , y) = the probability of output (a, b) oninput (x , y)
n questions, m answers: {P(a, b|x , y)} ⇢ Rm
2
n
2
Classically
P(a, b|x , y) = Pi
�i
· pxia
· qyib
Probability that Alice outputs a on input x
Same for Bob
Shared randomness
Tsirelson’s problem and linear system games William Slofstra
![Page 10: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/10.jpg)
What can P(a, b|x , y) be?Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
P(a, b|x , y) = the probability of output (a, b) oninput (x , y)
n questions, m answers: {P(a, b|x , y)} ⇢ Rm
2
n
2
Quantum
P(a, b|x , y) = h | Mx
a
⌦ Ny
b
| iAlice’s measurement on input x
Bob’s measurement on input y
shared state on H1
⌦ H2
Tsirelson’s problem and linear system games William Slofstra
![Page 11: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/11.jpg)
What can P(a, b|x , y) be?Referee
Alice Bob
Referee
Win
Lose
x
y
a
b
P(a, b|x , y) = the probability of output (a, b) oninput (x , y)
n questions, m answers: {P(a, b|x , y)} ⇢ Rm
2
n
2
Quantum
P(a, b|x , y) = h | Mx
a
⌦ Ny
b
| i
tensor product
Why? axiom of quantum mechanics for composite systems
Tsirelson’s problem and linear system games William Slofstra
![Page 12: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/12.jpg)
Bell inequalities
Cq
Cc
(all diagrams are schematic)
Cc
(m, n) = set of classical correlation matrices
Cq
(m, n) = set of quantum correlation matrices
Both are convex subsets of Rm
2
n
2
.
Tsirelson’s problem and linear system games William Slofstra
![Page 13: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/13.jpg)
Bell inequalities ct’d
Cq
Cc
!(G ,P) = !c
(G )
!(G ,P) = !q
(G )
!(G ,P) = probability of winning game G with correlation P
!c
(G ) = maximum winning probability for P 2 Cc
(m, n)
!q
(G ) = same thing but with Cq
(m, n)
Tsirelson’s problem and linear system games William Slofstra
![Page 14: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/14.jpg)
Bell inequalities ct’d
Cq
Cc
!(G ,P) = !c
(G )
!(G ,P) = !q
(G )
If !c
(G ) < !q
(G ), then
(1) Cc
( Cq
, and
(2) we can (theoretically) show this in an experiment
Bell’s theorem + many experiments: this happens!
Tsirelson’s problem and linear system games William Slofstra
![Page 15: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/15.jpg)
Bell inequalities ct’d
Cq
Cc
!(G ,P) = !c
(G )
!(G ,P) = !q
(G )
If !c
(G ) < !q
(G ), then
(1) Cc
( Cq
, and
(2) we can (theoretically) show this in an experiment
Bell’s theorem + many experiments: this happens!
Tsirelson’s problem and linear system games William Slofstra
![Page 16: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/16.jpg)
Finite versus infinite-dimensional
Quantum correlations:
P(a, b|x , y) = h |Mx
a
⌦ Ny
b
| iwhere | i 2 H
1
⌦ H2
Correlation set Cq
:
H1
, H2
must be finite-dimensional
(but, no bound on dimension)
Correlation set Cqs
:
H1
, H2
allowed to be infinite-dimensional
(the ‘s’ stands for ‘spatial tensor product’)
Tsirelson’s problem and linear system games William Slofstra
![Page 17: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/17.jpg)
Finite versus infinite-dimensional
Quantum correlations:
P(a, b|x , y) = h |Mx
a
⌦ Ny
b
| iwhere | i 2 H
1
⌦ H2
Correlation set Cq
:
H1
, H2
must be finite-dimensional
(but, no bound on dimension)
Correlation set Cqs
:
H1
, H2
allowed to be infinite-dimensional
(the ‘s’ stands for ‘spatial tensor product’)
Tsirelson’s problem and linear system games William Slofstra
![Page 18: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/18.jpg)
Finite versus infinite-dimensional
Quantum correlations:
P(a, b|x , y) = h |Mx
a
⌦ Ny
b
| iwhere | i 2 H
1
⌦ H2
Correlation set Cq
:
H1
, H2
must be finite-dimensional
(but, no bound on dimension)
Correlation set Cqs
:
H1
, H2
allowed to be infinite-dimensional
(the ‘s’ stands for ‘spatial tensor product’)
Tsirelson’s problem and linear system games William Slofstra
![Page 19: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/19.jpg)
Finite versus infinite-dimensional ct’d
Can we separate Cq
from Cqs
with a Bell inequality?
Cqs
Cq
NO!This is the wrong picture
Tsirelson’s problem and linear system games William Slofstra
![Page 20: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/20.jpg)
Finite versus infinite-dimensional ct’d
Can we separate Cq
from Cqs
with a Bell inequality?
Cqs
Cq
NO!This is the wrong picture
Tsirelson’s problem and linear system games William Slofstra
![Page 21: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/21.jpg)
How is this picture wrong?
Cqs
Cq
Cq
and Cqs
are not known to be closed.
Even worse: Cqs
= Cq
New correlation set Cqa
:= Cq
contains limits of finite-dimensional correlations
indistinguishable from Cq
and Cqs
in experiment
Tsirelson’s problem and linear system games William Slofstra
![Page 22: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/22.jpg)
How is this picture wrong?
Cqs
Cq
Cq
and Cqs
are not known to be closed.
Even worse: Cqs
= Cq
New correlation set Cqa
:= Cq
contains limits of finite-dimensional correlations
indistinguishable from Cq
and Cqs
in experiment
Tsirelson’s problem and linear system games William Slofstra
![Page 23: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/23.jpg)
How is this picture wrong?
Cqs
Cq
Cq
and Cqs
are not known to be closed.
Even worse: Cqs
= Cq
New correlation set Cqa
:= Cq
contains limits of finite-dimensional correlations
indistinguishable from Cq
and Cqs
in experiment
Tsirelson’s problem and linear system games William Slofstra
![Page 24: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/24.jpg)
The real picture
Could look like:
Cq
in Cq
in Cqs
but not Cq
in Cqa
but not Cqs
We know Cq
✓ Cqs
✓ Cqa
. . . but nothing else!
Fortunately, this is not the end of the story
We’ve assumed that H = HA
⌦HB
... maybe this is too restrictive
Tsirelson’s problem and linear system games William Slofstra
![Page 25: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/25.jpg)
The real picture
Could look like:
Cq
in Cq
in Cqs
but not Cq
in Cqa
but not Cqs
We know Cq
✓ Cqs
✓ Cqa
. . . but nothing else!
Fortunately, this is not the end of the story
We’ve assumed that H = HA
⌦HB
... maybe this is too restrictive
Tsirelson’s problem and linear system games William Slofstra
![Page 26: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/26.jpg)
Commuting-operator model
Another model of composite systems
Correlation set Cqc
:
P(a, b|x , y) = h |Mx
a
· Ny
b
| i
where
(1) | i belongs to a joint Hilbert space H(possibly infinite-dimensional)
(2) Measurements commute: Mx
a
Ny
b
= Ny
b
Mx
a
for all x , y , a, b
‘qc’ stands for ‘quantum-commuting’
Tsirelson’s problem and linear system games William Slofstra
![Page 27: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/27.jpg)
What do we know about Cqc
Correlation set Cqc
: P(a, b|x , y) = h |Mx
a
· Ny
b
| i
Cqc
is closed!
Get a hierarchy Cq
✓ Cqs
✓ Cqa
✓ Cqc
of convex sets
If H is finite-dimensional, then {P(a, b|x , y)} 2 Cq
Can find H1
, H2
such that H = H1
⌦ H2
,
Mx
a
⇠=
eMx
a
⌦ I and Ny
b
⇠=
I ⌦ eNy
b
for all x , y , a, b
This argument doesn’t work if H is infinite-dimensional
Tsirelson’s problem and linear system games William Slofstra
![Page 28: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/28.jpg)
What do we know about Cqc
Correlation set Cqc
: P(a, b|x , y) = h |Mx
a
· Ny
b
| i
Cqc
is closed!
Get a hierarchy Cq
✓ Cqs
✓ Cqa
✓ Cqc
of convex sets
If H is finite-dimensional, then {P(a, b|x , y)} 2 Cq
Can find H1
, H2
such that H = H1
⌦ H2
,
Mx
a
⇠=
eMx
a
⌦ I and Ny
b
⇠=
I ⌦ eNy
b
for all x , y , a, b
This argument doesn’t work if H is infinite-dimensional
Tsirelson’s problem and linear system games William Slofstra
![Page 29: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/29.jpg)
Tsirelson’s problem(s)
Cq
✓ Cqs
✓ Cqa
✓ Cqc
strong
weak
tensor product
commuting operator
Tsirelson problems: is Ct
, t 2 {q, qs, qa} equal to Cqc
These are fundamental questions
1 Comparing two axiom systems:
Strong Tsirelson: is Cq
= Cqc
?
2 Is !q
(G ) < !qc
(G ) for any game?
Equivalent to weak Tsirelson: is Cqa
= Cqc
?
Tsirelson’s problem and linear system games William Slofstra
![Page 30: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/30.jpg)
Tsirelson’s problem(s)
Cq
✓ Cqs
✓ Cqa
✓ Cqc
strong
weak
tensor product
commuting operator
Tsirelson problems: is Ct
, t 2 {q, qs, qa} equal to Cqc
These are fundamental questions
1 Comparing two axiom systems:
Strong Tsirelson: is Cq
= Cqc
?
2 Is !q
(G ) < !qc
(G ) for any game?
Equivalent to weak Tsirelson: is Cqa
= Cqc
?
Tsirelson’s problem and linear system games William Slofstra
![Page 31: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/31.jpg)
Tsirelson’s problem(s)
Cq
✓ Cqs
✓ Cqa
✓ Cqc
strong
weak
tensor product
commuting operator
Tsirelson problems: is Ct
, t 2 {q, qs, qa} equal to Cqc
These are fundamental questions
1 Comparing two axiom systems:
Strong Tsirelson: is Cq
= Cqc
?
2 Is !q
(G ) < !qc
(G ) for any game?
Equivalent to weak Tsirelson: is Cqa
= Cqc
?
Tsirelson’s problem and linear system games William Slofstra
![Page 32: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/32.jpg)
What do we know?
Cq
✓ Cqs
✓ Cqa
✓ Cqc
strong
weak
Theorem (Ozawa, JNPPSW, Fr)
Cqa
= Cqc
if and only if Connes’ embedding problem is true
Theorem (S)
Cqs
6= Cqc
Tsirelson’s problem and linear system games William Slofstra
![Page 33: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/33.jpg)
What do we know?
Cq
✓ Cqs
✓ Cqa
✓ Cqc
strong
weak
Theorem (Ozawa, JNPPSW, Fr)
Cqa
= Cqc
if and only if Connes’ embedding problem is true
Theorem (S)
Cqs
6= Cqc
Tsirelson’s problem and linear system games William Slofstra
![Page 34: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/34.jpg)
Other fundamental questions
1 Resource question:
A non-local game G is a computational task
Bell’s theorem: can do better with entanglement
Can G be played optimally with finite Hilbert space
dimension?
Yes () Cq
= Cqa
(in other words, is Cq
closed?)
Variants of games: finite dimensions do not su�ce
[LTW13],[MV14],[RV15]
2 Can we compute !q
(G ) or !qc
(G )?
(what is the power of MIP⇤?)
Tsirelson’s problem and linear system games William Slofstra
![Page 35: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/35.jpg)
Other fundamental questions
1 Resource question:
A non-local game G is a computational task
Bell’s theorem: can do better with entanglement
Can G be played optimally with finite Hilbert space
dimension?
Yes () Cq
= Cqa
(in other words, is Cq
closed?)
Variants of games: finite dimensions do not su�ce
[LTW13],[MV14],[RV15]
2 Can we compute !q
(G ) or !qc
(G )?
(what is the power of MIP⇤?)
Tsirelson’s problem and linear system games William Slofstra
![Page 36: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/36.jpg)
What do we know?
Question: can we compute !q
(G ) or !qc
(G )?
Brute force search through strategies on HA
= HB
= Cn
,
converges to !q
(from below)
Navascues, Pironio, Acın: Given a non-local game, there is a
hierarchy of SDPs which converge in value to !qc
(from above)
In both cases, no way to tell how close we are to the correct answer
Theorem (S)
It is undecidable to tell if !qc < 1
General cases of other questions completely open!
Tsirelson’s problem and linear system games William Slofstra
![Page 37: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/37.jpg)
What do we know?
Question: can we compute !q
(G ) or !qc
(G )?
Brute force search through strategies on HA
= HB
= Cn
,
converges to !q
(from below)
Navascues, Pironio, Acın: Given a non-local game, there is a
hierarchy of SDPs which converge in value to !qc
(from above)
In both cases, no way to tell how close we are to the correct answer
Theorem (S)
It is undecidable to tell if !qc < 1
General cases of other questions completely open!
Tsirelson’s problem and linear system games William Slofstra
![Page 38: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/38.jpg)
What do we know?
Question: can we compute !q
(G ) or !qc
(G )?
Brute force search through strategies on HA
= HB
= Cn
,
converges to !q
(from below)
Navascues, Pironio, Acın: Given a non-local game, there is a
hierarchy of SDPs which converge in value to !qc
(from above)
In both cases, no way to tell how close we are to the correct answer
Theorem (S)
It is undecidable to tell if !qc < 1
General cases of other questions completely open!
Tsirelson’s problem and linear system games William Slofstra
![Page 39: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/39.jpg)
What do we know?
Question: can we compute !q
(G ) or !qc
(G )?
Brute force search through strategies on HA
= HB
= Cn
,
converges to !q
(from below)
Navascues, Pironio, Acın: Given a non-local game, there is a
hierarchy of SDPs which converge in value to !qc
(from above)
In both cases, no way to tell how close we are to the correct answer
Theorem (S)
It is undecidable to tell if !qc < 1
General cases of other questions completely open!
Tsirelson’s problem and linear system games William Slofstra
![Page 40: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/40.jpg)
What do we know?
Question: can we compute !q
(G ) or !qc
(G )?
Brute force search through strategies on HA
= HB
= Cn
,
converges to !q
(from below)
Navascues, Pironio, Acın: Given a non-local game, there is a
hierarchy of SDPs which converge in value to !qc
(from above)
In both cases, no way to tell how close we are to the correct answer
Theorem (S)
It is undecidable to tell if !qc < 1
General cases of other questions completely open!
Tsirelson’s problem and linear system games William Slofstra
![Page 41: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/41.jpg)
Undecidability
Theorem (S)
It is undecidable to tell if !qc < 1
NPA hierarchy: there is no computable function
L : Games ! N
such that !qc
(G ) = L(G )th level of NPA hierarchy
We still don’t know: can we compute !qc
(G ) to within some given
error?
(Ji ’16: this problem is MIP⇤-complete)
If weak Tsirelson is true, then !qc is computable in this stronger
sense
Tsirelson’s problem and linear system games William Slofstra
![Page 42: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/42.jpg)
Undecidability
Theorem (S)
It is undecidable to tell if !qc < 1
NPA hierarchy: there is no computable function
L : Games ! N
such that !qc
(G ) = L(G )th level of NPA hierarchy
We still don’t know: can we compute !qc
(G ) to within some given
error?
(Ji ’16: this problem is MIP⇤-complete)
If weak Tsirelson is true, then !qc is computable in this stronger
sense
Tsirelson’s problem and linear system games William Slofstra
![Page 43: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/43.jpg)
Undecidability
Theorem (S)
It is undecidable to tell if !qc < 1
NPA hierarchy: there is no computable function
L : Games ! N
such that !qc
(G ) = L(G )th level of NPA hierarchy
We still don’t know: can we compute !qc
(G ) to within some given
error?
(Ji ’16: this problem is MIP⇤-complete)
If weak Tsirelson is true, then !qc is computable in this stronger
sense
Tsirelson’s problem and linear system games William Slofstra
![Page 44: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/44.jpg)
Undecidability comes from exact error?
Comparison point: Can decide if optimal value of finite SDP is < 1
(very ine�cient algorithm)
More generally: first-order logic for field of real numbers is
decidable
Contrast: first-order logic for integers and rationals is undecidable
Consequence of undecidability of !qc < 1 due to Tobias Fritz:
quantum logic (first order theory for projections on
Hilbert spaces) is undecidable
Tsirelson’s problem and linear system games William Slofstra
![Page 45: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/45.jpg)
Undecidability comes from exact error?
Comparison point: Can decide if optimal value of finite SDP is < 1
(very ine�cient algorithm)
More generally: first-order logic for field of real numbers is
decidable
Contrast: first-order logic for integers and rationals is undecidable
Consequence of undecidability of !qc < 1 due to Tobias Fritz:
quantum logic (first order theory for projections on
Hilbert spaces) is undecidable
Tsirelson’s problem and linear system games William Slofstra
![Page 46: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/46.jpg)
Undecidability comes from exact error?
Comparison point: Can decide if optimal value of finite SDP is < 1
(very ine�cient algorithm)
More generally: first-order logic for field of real numbers is
decidable
Contrast: first-order logic for integers and rationals is undecidable
Consequence of undecidability of !qc < 1 due to Tobias Fritz:
quantum logic (first order theory for projections on
Hilbert spaces) is undecidable
Tsirelson’s problem and linear system games William Slofstra
![Page 47: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/47.jpg)
Quantum logic is undecidable
Theorem (Tobias Fritz)
The following problem is undecidable:Given n � 1 and a collection of subsets C of {1, . . . , n}, determineif there are self-adjoint projections P
1
, . . . ,Pn
such that
X
i2SPi
= I , Pi
Pj
= Pj
Pi
= 0 if i 6= j 2 S
for all S 2 C.Proof: follows from undecidability of !qc < 1
Builds on Acın-Fritz-Leverrier-Sainz ’15.
Tsirelson’s problem and linear system games William Slofstra
![Page 48: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/48.jpg)
Two theorems
Theorem (S)
Cqs
6= Cqc
Theorem (S)
It is undecidable to tell if !qc
< 1
Theorems look very di↵erent...
But: proof follows from a single theorem in group theory
Connection with group theory comes from linear system games
Tsirelson’s problem and linear system games William Slofstra
![Page 49: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/49.jpg)
Two theorems
Theorem (S)
Cqs
6= Cqc
Theorem (S)
It is undecidable to tell if !qc
< 1
Theorems look very di↵erent...
But: proof follows from a single theorem in group theory
Connection with group theory comes from linear system games
Tsirelson’s problem and linear system games William Slofstra
![Page 50: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/50.jpg)
Linear system games
Start with m ⇥ n linear system Ax = b over Z2
Inputs:
•Alice receives 1 i m (an equation)
•Bob receives 1 j n (a variable)
Outputs:
•Alice outputs an assignment a
k
for all variables xk
with
Aik
6= 0
•Bob outputs an assignment b
j
for xj
They win if:
• Aij
= 0 (assignment irrelevant) or
• Aij
6= 0 and aj
= bj
(assignment consistent)
Such games go back to Mermin-Peres magic square, more recently
studied by Cleve-Mittal, Ji, Arkhipov
Tsirelson’s problem and linear system games William Slofstra
![Page 51: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/51.jpg)
Linear system games
Start with m ⇥ n linear system Ax = b over Z2
Inputs:
•Alice receives 1 i m (an equation)
•Bob receives 1 j n (a variable)
Outputs:
•Alice outputs an assignment a
k
for all variables xk
with
Aik
6= 0
•Bob outputs an assignment b
j
for xj
They win if:
• Aij
= 0 (assignment irrelevant) or
• Aij
6= 0 and aj
= bj
(assignment consistent)
Such games go back to Mermin-Peres magic square, more recently
studied by Cleve-Mittal, Ji, Arkhipov
Tsirelson’s problem and linear system games William Slofstra
![Page 52: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/52.jpg)
Linear system games
Start with m ⇥ n linear system Ax = b over Z2
Inputs:
•Alice receives 1 i m (an equation)
•Bob receives 1 j n (a variable)
Outputs:
•Alice outputs an assignment a
k
for all variables xk
with
Aik
6= 0
•Bob outputs an assignment b
j
for xj
They win if:
• Aij
= 0 (assignment irrelevant) or
• Aij
6= 0 and aj
= bj
(assignment consistent)
Such games go back to Mermin-Peres magic square, more recently
studied by Cleve-Mittal, Ji, Arkhipov
Tsirelson’s problem and linear system games William Slofstra
![Page 53: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/53.jpg)
Linear system games
Start with m ⇥ n linear system Ax = b over Z2
Inputs:
•Alice receives 1 i m (an equation)
•Bob receives 1 j n (a variable)
Outputs:
•Alice outputs an assignment a
k
for all variables xk
with
Aik
6= 0
•Bob outputs an assignment b
j
for xj
They win if:
• Aij
= 0 (assignment irrelevant) or
• Aij
6= 0 and aj
= bj
(assignment consistent)
Such games go back to Mermin-Peres magic square, more recently
studied by Cleve-Mittal, Ji, Arkhipov
Tsirelson’s problem and linear system games William Slofstra
![Page 54: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/54.jpg)
Linear system games
Start with m ⇥ n linear system Ax = b over Z2
Inputs:
•Alice receives 1 i m (an equation)
•Bob receives 1 j n (a variable)
Outputs:
•Alice outputs an assignment a
k
for all variables xk
with
Aik
6= 0
•Bob outputs an assignment b
j
for xj
They win if:
• Aij
= 0 (assignment irrelevant) or
• Aij
6= 0 and aj
= bj
(assignment consistent)
Such games go back to Mermin-Peres magic square, more recently
studied by Cleve-Mittal, Ji, Arkhipov
Tsirelson’s problem and linear system games William Slofstra
![Page 55: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/55.jpg)
Quantum solutions of Ax = b
Observables Xj
such that
1 X 2
j
= I for all j
2
Qn
j=1
XA
ij
j
= (�I )bi for all i
3 If Aij
,Aik
6= 0, then Xj
Xk
= Xk
Xj
(We’ve written linear equations multiplicatively)
Theorem (Cleve-Mittal,Cleve-Liu-S)
Let G be the game for linear system Ax = b. Then:
• G has a perfect strategy in Cqs
if and only if Ax = b has afinite-dimensional quantum solution
• G has a perfect strategy in Cqc
if and only if Ax = b has aquantum solution
Tsirelson’s problem and linear system games William Slofstra
![Page 56: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/56.jpg)
Quantum solutions of Ax = b
Observables Xj
such that
1 X 2
j
= I for all j
2
Qn
j=1
XA
ij
j
= (�I )bi for all i
3 If Aij
,Aik
6= 0, then Xj
Xk
= Xk
Xj
(We’ve written linear equations multiplicatively)
Theorem (Cleve-Mittal,Cleve-Liu-S)
Let G be the game for linear system Ax = b. Then:
• G has a perfect strategy in Cqs
if and only if Ax = b has afinite-dimensional quantum solution
• G has a perfect strategy in Cqc
if and only if Ax = b has aquantum solution
Tsirelson’s problem and linear system games William Slofstra
![Page 57: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/57.jpg)
Group theory ct’d
The solution group � of Ax = b is the group generated by
X1
, . . . ,Xn
, J such that
1 X 2
j
= [Xj
, J] = J2 = e for all j
2
Qn
j=1
XA
ij
j
= Jbi for all i
3 If Aij
,Aik
6= 0, then [Xj
,Xk
] = e
where [a, b] = aba�1b�1
, e = group identity
Theorem (Cleve-Mittal,Cleve-Liu-S)
Let G be the game for linear system Ax = b. Then:
• G has a perfect strategy in Cqs
if and only if � has afinite-dimensional representation with J 6= I
• G has a perfect strategy in Cqc
if and only if J 6= e in �
Tsirelson’s problem and linear system games William Slofstra
![Page 58: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/58.jpg)
Groups and local compatibility
Suppose we can write down any group relations we want...
But: generators in the relation will be forced to commute!
Call this condition local compatibility
Local compatibility is (a priori) a very strong constraint
For instance, S3
is generated by a, b subject to the relations
a2 = b2 = e, (ab)3 = e
If ab = ba, then (ab)3 = a3b3 = ab
So relations imply a = b, and S3
becomes Z2
Tsirelson’s problem and linear system games William Slofstra
![Page 59: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/59.jpg)
Groups and local compatibility
Suppose we can write down any group relations we want...
But: generators in the relation will be forced to commute!
Call this condition local compatibility
Local compatibility is (a priori) a very strong constraint
For instance, S3
is generated by a, b subject to the relations
a2 = b2 = e, (ab)3 = e
If ab = ba, then (ab)3 = a3b3 = ab
So relations imply a = b, and S3
becomes Z2
Tsirelson’s problem and linear system games William Slofstra
![Page 60: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/60.jpg)
Groups and local compatibility
Suppose we can write down any group relations we want...
But: generators in the relation will be forced to commute!
Call this condition local compatibility
Local compatibility is (a priori) a very strong constraint
For instance, S3
is generated by a, b subject to the relations
a2 = b2 = e, (ab)3 = e
If ab = ba, then (ab)3 = a3b3 = ab
So relations imply a = b, and S3
becomes Z2
Tsirelson’s problem and linear system games William Slofstra
![Page 61: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/61.jpg)
Group embedding theorem
Solution groups satisfy local compatibility
Nonetheless:
Solution groups are as complicated as general groups
Theorem (S)
Let G be any finitely-presented group, and suppose we are given J0
in the center of G such that J20
= e.
Then there is an injective homomorphism � : G ,! �, where � isthe solution group of a linear system Ax = b, with �(J
0
) = J.
Tsirelson’s problem and linear system games William Slofstra
![Page 62: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/62.jpg)
How do we prove the embedding theorem?
Linear system Ax = b over Z2
equivalent to labelled hypergraph:
Edges are variables
Vertices are equations
v is adjacent to e if and only if Ave
6= 0
v is labelled by bi
2 Z2
Given finitely-presented group G , we get � from a linear system
But what linear system?
Can answer this pictorially by writing down a hypergraph?
Tsirelson’s problem and linear system games William Slofstra
![Page 63: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/63.jpg)
How do we prove the embedding theorem?
Linear system Ax = b over Z2
equivalent to labelled hypergraph:
Edges are variables
Vertices are equations
v is adjacent to e if and only if Ave
6= 0
v is labelled by bi
2 Z2
Given finitely-presented group G , we get � from a linear system
But what linear system?
Can answer this pictorially by writing down a hypergraph?
Tsirelson’s problem and linear system games William Slofstra
![Page 64: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/64.jpg)
The hypergraph by example
zy uvx
hx , y , z , u, v : xyxz = xuvu = e = x2 = y2 = · · · = v2i
Tsirelson’s problem and linear system games William Slofstra
![Page 65: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/65.jpg)
Further directions
1 Further refinements to address Cq
vs Cqa
2 Is !q
(G ) < 1 decidable?
3 Embedding theorem: for any f.p. group G , get a non-local
game such that Alice and Bob are forced to use G to play
perfectly
(Caveat: but might need to use infinite-dimensional
commuting-operator strategy to achieve this)
Applications to self-testing / device independent protocols?
Tsirelson’s problem and linear system games William Slofstra
![Page 66: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/66.jpg)
Further directions
1 Further refinements to address Cq
vs Cqa
2 Is !q
(G ) < 1 decidable?
3 Embedding theorem: for any f.p. group G , get a non-local
game such that Alice and Bob are forced to use G to play
perfectly
(Caveat: but might need to use infinite-dimensional
commuting-operator strategy to achieve this)
Applications to self-testing / device independent protocols?
Tsirelson’s problem and linear system games William Slofstra
![Page 67: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/67.jpg)
The end
hx , y , z , u, v : xyxz = xuvu = e = x2 = y2 = · · · = v2i
Thank-you!
Tsirelson’s problem and linear system games William Slofstra
![Page 68: Tsirelson's problem and linear system games](https://reader034.fdocuments.net/reader034/viewer/2022052111/62886d3f270c94574e0d7d09/html5/thumbnails/68.jpg)
Extra slide: Higman’s group
G = ha, b, c , d : aba�1
= b2, bcb�1
= c2, cdc�1
= d2, dad�1
= a2
Only finite-dimensional representation is the trivial representation
On the other hand, a, b, c , d are all non-trivial in G
Tsirelson’s problem and linear system games William Slofstra