SLIDE 1 Analytic and nearly optimal self-testing bounds for the Clauser-Horne-Shimony-Holt and Mermin inequalities
- Phys. Rev. Lett. 117, 070402 (2016)
[arXiv:1604.08176]
Jed Kaniewski
QMATH, Department of Mathematical Sciences University of Copenhagen, Universitetsparken 5, 2100 Copenhagen Ø, Denmark
17 June 2016 CEQIP ’16
SLIDE 2
QMATH http://qmath.ku.dk/
SLIDE 3
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 4
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 5
What is self-testing?
Bell scenario x a y b Pr[a, b|x, y]
SLIDE 6 What is self-testing?
Bell scenario x a y b Pr[a, b|x, y] Def.: Pr[a, b|x, y] is local if Pr[a, b|x, y] =
p(λ) p(a|x, λ) p(b|y, λ). Otherwise = ⇒ nonlocal or it violates (some) Bell inequality
SLIDE 7 What is self-testing?
Obs.: Separable states give local statistics (for all measurements) ρAB =
pλαλ ⊗ βλ, Pr[a, b|x, y] = tr
a ⊗ Qy b)ρAB
pλ · tr(P x
a αλ)
· tr(Qy
bβλ)
.
SLIDE 8
What is self-testing?
ρAB is separable = ⇒ statistics are local Pr[a, b|x, y] is nonlocal = ⇒ ρAB is entangled
SLIDE 9
What is self-testing?
ρAB is separable = ⇒ statistics are local Pr[a, b|x, y] is nonlocal = ⇒ ρAB is entangled smart! anything more specific?
SLIDE 10
What is self-testing?
ρAB is separable = ⇒ statistics are local Pr[a, b|x, y] is nonlocal = ⇒ ρAB is entangled smart! anything more specific? sure! let me google it for you...
SLIDE 11 What is self-testing?
Given Pr[a, b|x, y] = tr
a ⊗ Qy b)ρAB
- deduce properties of ρAB, {P x
a }, {Qy b}
SLIDE 12 What is self-testing?
Given Pr[a, b|x, y] = tr
a ⊗ Qy b)ρAB
- deduce properties of ρAB, {P x
a }, {Qy b}
We don’t assume that ρAB is pure and it’s important! (ask me if you want to know more)
SLIDE 13 What is self-testing?
Given Pr[a, b|x, y] = tr
a ⊗ Qy b)ρAB
- deduce properties of ρAB, {P x
a }, {Qy b}
We don’t assume that ρAB is pure and it’s important! (ask me if you want to know more)
- ften only promised some Bell violation
- abxy
cxy
ab Pr[a, b|x, y] = β
SLIDE 14 What is self-testing?
Example: the CHSH inequality [Popescu, Rohrlich ’92] βCHSH :=
(−1)a+b+xy Pr[a, b|x, y] for a, b, x, y ∈ {0, 1} βCHSH = 2 √ 2 (max) = ⇒ ρAB ≃ ΦAB for |ΦAB =
1 √ 2(|00 + |11).
SLIDE 15 What is self-testing?
Example: the CHSH inequality [Popescu, Rohrlich ’92] βCHSH :=
(−1)a+b+xy Pr[a, b|x, y] for a, b, x, y ∈ {0, 1} βCHSH = 2 √ 2 (max) = ⇒ ρAB ≃ ΦAB for |ΦAB =
1 √ 2(|00 + |11).
ρAB = ΦAB Inherent limitations
SLIDE 16 What is self-testing?
Example: the CHSH inequality [Popescu, Rohrlich ’92] βCHSH :=
(−1)a+b+xy Pr[a, b|x, y] for a, b, x, y ∈ {0, 1} βCHSH = 2 √ 2 (max) = ⇒ ρAB ≃ ΦAB for |ΦAB =
1 √ 2(|00 + |11).
ρAB = ΦAB Inherent limitations
- cannot see auxiliary systems (measurements act trivially)
⊗ τA′B′
SLIDE 17 What is self-testing?
Example: the CHSH inequality [Popescu, Rohrlich ’92] βCHSH :=
(−1)a+b+xy Pr[a, b|x, y] for a, b, x, y ∈ {0, 1} βCHSH = 2 √ 2 (max) = ⇒ ρAB ≃ ΦAB for |ΦAB =
1 √ 2(|00 + |11).
ρAB = ΦAB Inherent limitations
- cannot see auxiliary systems (measurements act trivially)
⊗ τA′B′
- cannot see local unitaries
U( )U † for U = UAA′ ⊗ UBB′
SLIDE 18 What is self-testing?
Example: the CHSH inequality [Popescu, Rohrlich ’92] βCHSH :=
(−1)a+b+xy Pr[a, b|x, y] for a, b, x, y ∈ {0, 1} βCHSH = 2 √ 2 (max) = ⇒ ρAB ≃ ΦAB for |ΦAB =
1 √ 2(|00 + |11).
ρAB = ΦAB Inherent limitations
- cannot see auxiliary systems (measurements act trivially)
⊗ τA′B′
- cannot see local unitaries
U( )U † for U = UAA′ ⊗ UBB′ Necessary... but also sufficient!
SLIDE 19 What is self-testing?
ab Pr[a, b|x, y] = β
SLIDE 20 What is self-testing?
ab Pr[a, b|x, y] = β
? ρAB state certification
SLIDE 21 What is self-testing?
ab Pr[a, b|x, y] = β
? ρAB state certification ? P x
a , Qy b(ρAB)
measurement certification
SLIDE 22 What is self-testing?
ab Pr[a, b|x, y] = β
? ρAB state certification ? P x
a , Qy b(ρAB)
measurement certification
SLIDE 23 What is self-testing?
ab Pr[a, b|x, y] = β
? ρAB state certification ? P x
a , Qy b(ρAB)
measurement certification Which states can be certified?
Ψ?
SLIDE 24 What is self-testing?
ab Pr[a, b|x, y] = β
? ρAB state certification ? P x
a , Qy b(ρAB)
measurement certification Which states can be certified?
Ψ?
SLIDE 25
What is self-testing?
What is experimentally-relevant? The CHSH inequality: βC = 2 and βQ = 2 √ 2
SLIDE 26
What is self-testing?
What is experimentally-relevant? The CHSH inequality: βC = 2 and βQ = 2 √ 2 Non-trivial bounds for... [Bardyn et al. ’09]: β ≥ 1 + √ 2 ≈ 2.41 [McKague et al. ’12]: β ≥ βQ − 2 · 10−5
SLIDE 27
What is self-testing?
What is experimentally-relevant? The CHSH inequality: βC = 2 and βQ = 2 √ 2 Non-trivial bounds for... [Bardyn et al. ’09]: β ≥ 1 + √ 2 ≈ 2.41 [McKague et al. ’12]: β ≥ βQ − 2 · 10−5
SLIDE 28
What is self-testing?
What is experimentally-relevant? The CHSH inequality: βC = 2 and βQ = 2 √ 2 Non-trivial bounds for... [Bardyn et al. ’09]: β ≥ 1 + √ 2 ≈ 2.41 [McKague et al. ’12]: β ≥ βQ − 2 · 10−5 The loophole-free Bell experiment from Delft β = 2.4 ± 0.2
SLIDE 29
What is self-testing?
What is experimentally-relevant? The CHSH inequality: βC = 2 and βQ = 2 √ 2 Non-trivial bounds for... [Bardyn et al. ’09]: β ≥ 1 + √ 2 ≈ 2.41 [McKague et al. ’12]: β ≥ βQ − 2 · 10−5 The loophole-free Bell experiment from Delft β = 2.4 ± 0.2 4 orders of magnitude off!
SLIDE 30
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 31
Previous results
Self-testable the singlet [McKague et al. ’12] graph states [McKague ’14] high-dimensional maximally entangled state [Slofstra ’11, Yang, Navascués ’13, McKague ’16, Salavrakos et al. ’16 + . . . ] non-maximally entangled states of 2 qubits [Bamps, Pironio ’15] Only for almost perfect statistics (ε ≈ 10−4).
SLIDE 32 Previous results
Self-testable the singlet [McKague et al. ’12] graph states [McKague ’14] high-dimensional maximally entangled state [Slofstra ’11, Yang, Navascués ’13, McKague ’16, Salavrakos et al. ’16 + . . . ] non-maximally entangled states of 2 qubits [Bamps, Pironio ’15] Only for almost perfect statistics (ε ≈ 10−4). Experimentally-relevant robustness a single analytic result for the singlet-CHSH case [Bardyn et
swap trick: a numerical method, versatile but computationally expensive (so far up to 4 qubits or 2 qutrits) [Yang et al. ’14, Bancal et al. ’15]
[see arXiv:1604.08176 for references]
SLIDE 33
New findings
New approach for analytic self-testing bounds improvement for the CHSH and Mermin3 Mermin3 is actually tight (!) self-testing problem operator inequalities
SLIDE 34
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 35 Self-testing from operator inequalities
Extractability of ΨA′B′ from ρAB Ξ(ρAB → ΨA′B′) := maxΛA,ΛB F
local extraction channels fidelity
SLIDE 36 Self-testing from operator inequalities
Extractability of ΨA′B′ from ρAB Ξ(ρAB → ΨA′B′) := maxΛA,ΛB F
local extraction channels fidelity Obs1: Ξ(ρAB → ΨA′B′) = 1 ⇐ ⇒ ρAB = V (ΨA′B′ ⊗ σA′′B′′)V † for V = VA′A′′→A ⊗ VB′B′′→B
SLIDE 37 Self-testing from operator inequalities
Extractability of ΨA′B′ from ρAB Ξ(ρAB → ΨA′B′) := maxΛA,ΛB F
local extraction channels fidelity Obs1: Ξ(ρAB → ΨA′B′) = 1 ⇐ ⇒ ρAB = V (ΨA′B′ ⊗ σA′′B′′)V † for V = VA′A′′→A ⊗ VB′B′′→B Obs2: Ξ(ρAB → ΨA′B′) ∈ [λ2
max, 1]
largest Schmidt coefficient
SLIDE 38
Self-testing from operator inequalities
Idea: measurement operators extraction channels! Analytical bound of [Bardyn et al.] in 2 steps [1] Solve the problem for 2 qbits (local measurements determine a local unitary correction) [2] Use Jordan’s lemma to argue that it holds in arbitrary dimension
SLIDE 39 Self-testing from operator inequalities
Idea: measurement operators extraction channels! Analytical bound of [Bardyn et al.] in 2 steps [1] Solve the problem for 2 qbits (local measurements determine a local unitary correction) [2] Use Jordan’s lemma to argue that it holds in arbitrary dimension
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 β 0.25 0.50 0.75 1.00 [C. E. Bardyn, T. C. H. Liew, S. Massar,
- M. McKague, and V. Scarani.
- Phys. Rev. A, 80(6), 2009. arXiv:0907.2170]
SLIDE 40
Self-testing from operator inequalities
Refined approach: assume ΛA := ΛA({P x
a })
and ΛB := ΛB({Qy
b}).
SLIDE 41 Self-testing from operator inequalities
Refined approach: assume ΛA := ΛA({P x
a })
and ΛB := ΛB({Qy
b}).
for ΨA′B′ pure F
= (ΛA ⊗ ΛB)(ρAB), ΨA′B′ = ρAB, (Λ†
A ⊗ Λ† B)(ΨA′B′) = tr(KρAB)
for K = (Λ†
A ⊗ Λ† B)(ΨA′B′)
important: K depends only on ΨA′B′, {P x
a }, {Qy b}, not on ρAB!
SLIDE 42 Self-testing from operator inequalities
Refined approach: assume ΛA := ΛA({P x
a })
and ΛB := ΛB({Qy
b}).
for ΨA′B′ pure F
= (ΛA ⊗ ΛB)(ρAB), ΨA′B′ = ρAB, (Λ†
A ⊗ Λ† B)(ΨA′B′) = tr(KρAB)
for K = (Λ†
A ⊗ Λ† B)(ΨA′B′)
important: K depends only on ΨA′B′, {P x
a }, {Qy b}, not on ρAB!
... just like the Bell operator W =
cxy
abP x a ⊗ Qy b.
SLIDE 43
Self-testing from operator inequalities
Forget the input state ρAB! Want s, µ ∈ R such that K ≥ sW + µ1 holds for all possible measurement operators
SLIDE 44 Self-testing from operator inequalities
Forget the input state ρAB! Want s, µ ∈ R such that K ≥ sW + µ1 holds for all possible measurement operators Challenging!... but if works then tr(KρAB) ≥ s tr(WρAB) + µ tr(ρAB) equivalent to F
≥ sβ + µ precisely a (linear) self-testing statement!
SLIDE 45
Self-testing from operator inequalities
Main technical challenge: find channels and s, µ such that K ≥ sW + µ1 holds for all possible measurement operators
SLIDE 46
Self-testing from operator inequalities
Main technical challenge: find channels and s, µ such that K ≥ sW + µ1 holds for all possible measurement operators Jordan’s lemma: any two binary, projective measurements can be simultaneously block-diagonalised into 2 × 2 blocks (at most) each block parametrised by an angle a ∈ [0, π/2] (up to unitary) this becomes tractable: 1-parameter per party
SLIDE 47
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 48 CHSH self-testing: proof in 4 steps
1 Extraction channels: angle-dependent dephasing
full partial none partial full
SLIDE 49 CHSH self-testing: proof in 4 steps
1 Extraction channels: angle-dependent dephasing
full partial none partial full
2 Find suitable s, µ (numerics): s = (4 + 5
√ 2)/16 and µ = −(1 + 2 √ 2)/4
SLIDE 50 CHSH self-testing: proof in 4 steps
1 Extraction channels: angle-dependent dephasing
full partial none partial full
2 Find suitable s, µ (numerics): s = (4 + 5
√ 2)/16 and µ = −(1 + 2 √ 2)/4
3 Prove
K(a, b) ≥ sW(a, b) + µ1 for all a, b ∈ [0, π/2] (2-parameter family of 4 × 4 matrices)
SLIDE 51
CHSH self-testing: proof in 4 steps
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 β 0.25 0.50 0.75 1.00 Bardyn et al. [BLM+09] Bancal et al. [BNS+15] current work
SLIDE 52
CHSH self-testing: proof in 4 steps
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 β 0.25 0.50 0.75 1.00 Bardyn et al. [BLM+09] Bancal et al. [BNS+15] current work
non-trivial for β > (16 + 14 √ 2)/17 ≈ 2.11 ?
SLIDE 53 Mermin3 self-testing: proof in 4 steps
1 Same extraction channels 2 Find suitable s, µ (numerics): s = (2 +
√ 2)/8 and µ = −1/ √ 2
3 Prove
K(a, b, c) ≥ sW(a, b, c) + µ1 for all a, b, c ∈ [0, π/2] (3-parameter family of 8 × 8 matrices)
SLIDE 54
Mermin3 self-testing: proof in 4 steps
2.8 3.0 3.2 3.4 3.6 3.8 4.0 β 0.25 0.50 0.75 1.00 P´ al et al. [PVN14] current work
SLIDE 55
Mermin3 self-testing: proof in 4 steps
2.8 3.0 3.2 3.4 3.6 3.8 4.0 β 0.25 0.50 0.75 1.00 P´ al et al. [PVN14] current work
tight :)
SLIDE 56
Outline
What is self-testing? Previous results and new findings Self-testing from operator inequalities Two examples: the CHSH and Mermin3 inequalities Summary and future work
SLIDE 57
Summary and future work
Summary self-testing from operator inequalities improvements for the CHSH and Mermin3 inequalities first provably tight self-testing statement
SLIDE 58
Summary and future work
Summary self-testing from operator inequalities improvements for the CHSH and Mermin3 inequalities first provably tight self-testing statement Future work Merminn
?
= ⇒ GHZn state (preliminary numerics) tilted CHSH
?
= ⇒ non-maximally entangled 2-qubit states [project in progress with Tim Coopmans and Christian Schaffner] beyond Jordan’s lemma? apply this approach to steering?
SLIDE 59
So you can really certify quantum states without trusting the devices at all? Yes, Pooh, quantum mechanics is very strange and nobody really understands it but let’s talk about it some other day...