Squashed Entanglement Nicholas LeCompte based on Squashed - - PowerPoint PPT Presentation

squashed entanglement
SMART_READER_LITE
LIVE PREVIEW

Squashed Entanglement Nicholas LeCompte based on Squashed - - PowerPoint PPT Presentation

Entanglement measures Squashed Entanglement Nicholas LeCompte based on Squashed Entanglement - An Additive Entanglement Measure (M. Christandl, A. Winter, quant-ph/0308088 ), and A paradigm for entanglement theory based on quantum


slide-1
SLIDE 1

Entanglement measures

Squashed Entanglement

Nicholas LeCompte based on ”Squashed Entanglement” - An Additive Entanglement Measure (M. Christandl, A. Winter, quant-ph/0308088), and A paradigm for entanglement theory based on quantum communication (J. Oppenheim arXiv:0801.0458)

Nicholas LeCompte Squashed entanglement

slide-2
SLIDE 2

Entanglement measures

(Classical) intrinsic information

◮ The intrinsic information between two random variables X, Y ,

given a third r.v. Z, is given by I(X; Y ↓ Z) := inf

¯ Z (I(X; Y |¯

Z)) with ¯ Z governed by a conditional probability distribution P¯

Z|Z. ◮ Idea: two parties have access to X and Y , and want to

generate a secret key even if an adversary has access to some additional knowledge (Z).

Nicholas LeCompte Squashed entanglement

slide-3
SLIDE 3

Entanglement measures

(Classical) intrinsic information

◮ The intrinsic information between two random variables X, Y ,

given a third r.v. Z, is given by I(X; Y ↓ Z) := inf

¯ Z (I(X; Y |¯

Z)) with ¯ Z governed by a conditional probability distribution P¯

Z|Z. ◮ Idea: two parties have access to X and Y , and want to

generate a secret key even if an adversary has access to some additional knowledge (Z).

◮ Mauer, Wolf 1999: The intrinsic information upper-bounds

the rate at which the key can be extracted.

Nicholas LeCompte Squashed entanglement

slide-4
SLIDE 4

Entanglement measures

A quantum analogue (Christandl, Winter, quant-ph/0308088)

◮ Replace classical mutual conditional infos with quantum, P¯ Z|Z

with quantum channel: Esq(ρAB) = inf

Λ:B(HC )→B(HE )

1 2I(A; B|E) : ρABE = (id ⊗ Λ)|ΨΨ|ABC

  • ◮ Here, |ΨABC is a purification of ρAB, and ρABE is simply a

tripartite extension; ρAB = TrE(ρABE).

Nicholas LeCompte Squashed entanglement

slide-5
SLIDE 5

Entanglement measures

A quantum analogue (Christandl, Winter, quant-ph/0308088)

◮ Replace classical mutual conditional infos with quantum, P¯ Z|Z

with quantum channel: Esq(ρAB) = inf

Λ:B(HC )→B(HE )

1 2I(A; B|E) : ρABE = (id ⊗ Λ)|ΨΨ|ABC

  • ◮ Here, |ΨABC is a purification of ρAB, and ρABE is simply a

tripartite extension; ρAB = TrE(ρABE).

◮ Esq(ρAB) is the squashed entanglement.

Nicholas LeCompte Squashed entanglement

slide-6
SLIDE 6

Entanglement measures

Squashed entanglement

◮ Equivalently,

Esq(ρAB) = inf

ρABE

1 2I(A; B|E) : TrE(ρABE) = ρAB

  • ◮ (equivalence follows by the equivalence of purifications up to a

unitary map)

Nicholas LeCompte Squashed entanglement

slide-7
SLIDE 7

Entanglement measures

Squashed entanglement

◮ Equivalently,

Esq(ρAB) = inf

ρABE

1 2I(A; B|E) : TrE(ρABE) = ρAB

  • ◮ (equivalence follows by the equivalence of purifications up to a

unitary map)

◮ Sanity check: If ρAB = |ψψ|AB, all tripartite extensions are

given by ρAB ⊗ ρE, so Esq(ρAB) = S(ρA) = E(|ψ).

Nicholas LeCompte Squashed entanglement

slide-8
SLIDE 8

Entanglement measures

Some basic properties

◮ Esq has basic properties we want in an entanglement measure,

e.g.:

Nicholas LeCompte Squashed entanglement

slide-9
SLIDE 9

Entanglement measures

Some basic properties

◮ Esq has basic properties we want in an entanglement measure,

e.g.:

◮ does not increase under LOCC ◮ is convex ◮ is continuous (by Alicki-Fannes theorem, quant-ph/0312081

[or Lecture 10])

Nicholas LeCompte Squashed entanglement

slide-10
SLIDE 10

Entanglement measures

Some basic properties

◮ Esq has basic properties we want in an entanglement measure,

e.g.:

◮ does not increase under LOCC ◮ is convex ◮ is continuous (by Alicki-Fannes theorem, quant-ph/0312081

[or Lecture 10])

◮ It’s also superadditive, and additive on tensor products.

Nicholas LeCompte Squashed entanglement

slide-11
SLIDE 11

Entanglement measures

Relations to other entanglement measures

◮ Recall that the intrinsic information upper-bounds the rate at

which the key can be extracted given an adversary has access to some information.

Nicholas LeCompte Squashed entanglement

slide-12
SLIDE 12

Entanglement measures

Relations to other entanglement measures

◮ Recall that the intrinsic information upper-bounds the rate at

which the key can be extracted given an adversary has access to some information.

◮ Prop. ED(ρAB) ≤ Esq(ρAB), where ED is the distillable

entanglement.

◮ (Recall ED(ρ) is the limiting ratio n/m of n Bell pairs created

from m copies of ρ with LOCC.)

Nicholas LeCompte Squashed entanglement

slide-13
SLIDE 13

Entanglement measures

Relations to other entanglement measures

◮ Recall that the intrinsic information upper-bounds the rate at

which the key can be extracted given an adversary has access to some information.

◮ Prop. ED(ρAB) ≤ Esq(ρAB), where ED is the distillable

entanglement.

◮ (Recall ED(ρ) is the limiting ratio n/m of n Bell pairs created

from m copies of ρ with LOCC.)

◮ The squashed entanglement is an analogous upper-bound

(viewing maximally entangled states as secret quantum correlations).

Nicholas LeCompte Squashed entanglement

slide-14
SLIDE 14

Entanglement measures

Proof sketch: ED(ρAB) ≤ Esq(ρAB)

◮ Start with an arbitrary distillation protocol, taking

(ρAB)⊗n →LOCC ΨAB such that ||ΨAB − |kk|AB||1 ≤ ε, where |k is a rank-k maximally entangled state.

Nicholas LeCompte Squashed entanglement

slide-15
SLIDE 15

Entanglement measures

Proof sketch: ED(ρAB) ≤ Esq(ρAB)

◮ Start with an arbitrary distillation protocol, taking

(ρAB)⊗n →LOCC ΨAB such that ||ΨAB − |kk|AB||1 ≤ ε, where |k is a rank-k maximally entangled state.

◮ By superadditivity and monotonicity under LOCC,

nEsq(ρAB) ≥ Esq(ΨAB).

Nicholas LeCompte Squashed entanglement

slide-16
SLIDE 16

Entanglement measures

Proof sketch: ED(ρAB) ≤ Esq(ρAB)

◮ Start with an arbitrary distillation protocol, taking

(ρAB)⊗n →LOCC ΨAB such that ||ΨAB − |kk|AB||1 ≤ ε, where |k is a rank-k maximally entangled state.

◮ By superadditivity and monotonicity under LOCC,

nEsq(ρAB) ≥ Esq(ΨAB).

◮ Note Esq(|kk|) = log k (since |k is maximally entangled),

and using Fannes’ inequality, one can show that for any extension ΨABE, 1

2I(A; B|E) ≥ log k − f (ǫ) log(k) for some

function f with limǫ→0 f (ǫ) = 0.

Nicholas LeCompte Squashed entanglement

slide-17
SLIDE 17

Entanglement measures

Proof sketch: ED(ρAB) ≤ Esq(ρAB)

◮ Start with an arbitrary distillation protocol, taking

(ρAB)⊗n →LOCC ΨAB such that ||ΨAB − |kk|AB||1 ≤ ε, where |k is a rank-k maximally entangled state.

◮ By superadditivity and monotonicity under LOCC,

nEsq(ρAB) ≥ Esq(ΨAB).

◮ Note Esq(|kk|) = log k (since |k is maximally entangled),

and using Fannes’ inequality, one can show that for any extension ΨABE, 1

2I(A; B|E) ≥ log k − f (ǫ) log(k) for some

function f with limǫ→0 f (ǫ) = 0.

◮ Hence Esq(ρAB) ≥ 1 n(1 − f (ǫ)) log k.

Nicholas LeCompte Squashed entanglement

slide-18
SLIDE 18

Entanglement measures

Relations to other entanglement measures

◮ Recall that the entanglement of formation is given by

EF(ρAB) = inf

ρ=

i pi|ψiψi|

  • i

piE(|ψi) = inf 1 2

  • i

piI(|ψiA; |ψiB)

Nicholas LeCompte Squashed entanglement

slide-19
SLIDE 19

Entanglement measures

Relations to other entanglement measures

◮ Recall that the entanglement of formation is given by

EF(ρAB) = inf

ρ=

i pi|ψiψi|

  • i

piE(|ψi) = inf 1 2

  • i

piI(|ψiA; |ψiB)

◮ Consider the following extension

ρABE =

  • i

pi|ψiψi|AB ⊗ |ii|E

Nicholas LeCompte Squashed entanglement

slide-20
SLIDE 20

Entanglement measures

Relations to other entanglement measures

◮ Recall that the entanglement of formation is given by

EF(ρAB) = inf

ρ=

i pi|ψiψi|

  • i

piE(|ψi) = inf 1 2

  • i

piI(|ψiA; |ψiB)

◮ Consider the following extension

ρABE =

  • i

pi|ψiψi|AB ⊗ |ii|E

◮ Then

1 2

  • i

piI(|ψiA; |ψiB) = 1 2I(A; B|E).

Nicholas LeCompte Squashed entanglement

slide-21
SLIDE 21

Entanglement measures

Relations to other entanglement measures

◮ Hence the entanglement of formation can be thought of as

the infinum of I(A; B|E) over a certain type of tripartite extensions.

Nicholas LeCompte Squashed entanglement

slide-22
SLIDE 22

Entanglement measures

Relations to other entanglement measures

◮ Hence the entanglement of formation can be thought of as

the infinum of I(A; B|E) over a certain type of tripartite extensions.

◮ It follows that Esq(ρAB) ≤ EF(ρAB).

Nicholas LeCompte Squashed entanglement

slide-23
SLIDE 23

Entanglement measures

Relations to other entanglement measures

◮ Hence the entanglement of formation can be thought of as

the infinum of I(A; B|E) over a certain type of tripartite extensions.

◮ It follows that Esq(ρAB) ≤ EF(ρAB). ◮ Invoking additivity, the entanglement cost

EC(ρAB) = lim

n→∞

1 nEF((ρAB)⊗n) is also bounded below by Esq(ρAB).

Nicholas LeCompte Squashed entanglement

slide-24
SLIDE 24

Entanglement measures

Another interpretation of squashed entanglement (Oppenheim, arXiv:0801.0458)

◮ Suppose Alice and Bob share a state ρAB, and Alice wants to

send her part to Eve. How many qubits does she need?

◮ We allow Alice and Eve to have as much side information

(e.g., ancillas) as they like.

Nicholas LeCompte Squashed entanglement

slide-25
SLIDE 25

Entanglement measures

Another interpretation of squashed entanglement (Oppenheim, arXiv:0801.0458)

◮ Suppose Alice and Bob share a state ρAB, and Alice wants to

send her part to Eve. How many qubits does she need?

◮ We allow Alice and Eve to have as much side information

(e.g., ancillas) as they like.

◮ However, the amount of useful information is limited, since

Alice’s share of the state may be entangled with Bob’s.

Nicholas LeCompte Squashed entanglement

slide-26
SLIDE 26

Entanglement measures

Another interpretation of squashed entanglement (Oppenheim, arXiv:0801.0458)

◮ Suppose Alice and Bob share a state ρAB, and Alice wants to

send her part to Eve. How many qubits does she need?

◮ We allow Alice and Eve to have as much side information

(e.g., ancillas) as they like.

◮ However, the amount of useful information is limited, since

Alice’s share of the state may be entangled with Bob’s.

◮ Let QA′E(ρAB) be the rate of communication required to send

ρA to a receiver who holds ρC, with ρA′ held by the sender.

Nicholas LeCompte Squashed entanglement

slide-27
SLIDE 27

Entanglement measures

State redistribution and merging (Oppenheim, arXiv:0805.1065, and Lecture 16)

◮ Using state merging, the minimum amount of quantum

communication required is QA′C = 1 2I(A : B|C) E (entanglement) = 1 2I(A : A′) − 1 2I(A : C)

Nicholas LeCompte Squashed entanglement

slide-28
SLIDE 28

Entanglement measures

State redistribution and merging (Oppenheim, arXiv:0805.1065, and Lecture 16)

◮ Using state merging, the minimum amount of quantum

communication required is QA′C = 1 2I(A : B|C) E (entanglement) = 1 2I(A : A′) − 1 2I(A : C)

◮ To clear up some confusion: in terms of the merging mother

protocol, Bob is the reference here.

Nicholas LeCompte Squashed entanglement

slide-29
SLIDE 29

Entanglement measures

Back to squashed entanglement

◮ Allowing arbitrary side information, the entanglement E is

certainly attainable.

Nicholas LeCompte Squashed entanglement

slide-30
SLIDE 30

Entanglement measures

Back to squashed entanglement

◮ Allowing arbitrary side information, the entanglement E is

certainly attainable.

◮ So, let ρS be the side information shared between Alice and

Eve (so that ρABS is pure), with Alice’s share being ρA′ and Eve’s ρC.

Nicholas LeCompte Squashed entanglement

slide-31
SLIDE 31

Entanglement measures

Back to squashed entanglement

◮ Allowing arbitrary side information, the entanglement E is

certainly attainable.

◮ So, let ρS be the side information shared between Alice and

Eve (so that ρABS is pure), with Alice’s share being ρA′ and Eve’s ρC.

◮ Then

inf

ρA′C

QA′C = inf

ρA′C

1 2I(A : B|C) = inf

Λ:B(S)→B(C)

1 2I(A : B|Λ(S)) = Esq(ρAB)

Nicholas LeCompte Squashed entanglement

slide-32
SLIDE 32

Entanglement measures

Back to squashed entanglement

◮ Allowing arbitrary side information, the entanglement E is

certainly attainable.

◮ So, let ρS be the side information shared between Alice and

Eve (so that ρABS is pure), with Alice’s share being ρA′ and Eve’s ρC.

◮ Then

inf

ρA′C

QA′C = inf

ρA′C

1 2I(A : B|C) = inf

Λ:B(S)→B(C)

1 2I(A : B|Λ(S)) = Esq(ρAB)

◮ So the squashed entanglement is the fastest rate Alice can

send a state to Eve given arbitrary side information.

Nicholas LeCompte Squashed entanglement

slide-33
SLIDE 33

Entanglement measures

Sanity check

◮ If ρAB = |ψψ|, Esq(ρ(AB)) = E(|ψ). ◮ Likewise, any pure state extension ρABS must have S

completely uncorrelated with AB.

Nicholas LeCompte Squashed entanglement

slide-34
SLIDE 34

Entanglement measures

Sanity check

◮ If ρAB = |ψψ|, Esq(ρ(AB)) = E(|ψ). ◮ Likewise, any pure state extension ρABS must have S

completely uncorrelated with AB.

◮ If Alice and Bob share a separable state ρAB, written as a

convex combination of separable pure states, Esq(ρAB) = 0:

Nicholas LeCompte Squashed entanglement

slide-35
SLIDE 35

Entanglement measures

Sanity check

◮ If ρAB = |ψψ|, Esq(ρ(AB)) = E(|ψ). ◮ Likewise, any pure state extension ρABS must have S

completely uncorrelated with AB.

◮ If Alice and Bob share a separable state ρAB, written as a

convex combination of separable pure states, Esq(ρAB) = 0:

◮ The extension

ρABE =

  • i

pi|ψiψi|A ⊗ |φiφi|B ⊗ |ii|E has zero quantum conditional mutual information.

Nicholas LeCompte Squashed entanglement

slide-36
SLIDE 36

Entanglement measures

Sanity check

◮ If ρAB = |ψψ|, Esq(ρ(AB)) = E(|ψ). ◮ Likewise, any pure state extension ρABS must have S

completely uncorrelated with AB.

◮ If Alice and Bob share a separable state ρAB, written as a

convex combination of separable pure states, Esq(ρAB) = 0:

◮ The extension

ρABE =

  • i

pi|ψiψi|A ⊗ |φiφi|B ⊗ |ii|E has zero quantum conditional mutual information.

◮ Of course, the rate also vanishes - teleportation does the trick.

Nicholas LeCompte Squashed entanglement

slide-37
SLIDE 37

Entanglement measures

Why “squashed” entanglement?

Nicholas LeCompte Squashed entanglement

slide-38
SLIDE 38

Entanglement measures

Why “squashed” entanglement?

◮ Christandl and Winter:

“Our name for this functional comes from the idea that the right choice of a conditioning system reduces the quantum mutual information between A and B, thus “squashing out” the nonquantum correlations.”

Nicholas LeCompte Squashed entanglement

slide-39
SLIDE 39

Entanglement measures

Why “squashed” entanglement?

◮ Christandl and Winter:

“Our name for this functional comes from the idea that the right choice of a conditioning system reduces the quantum mutual information between A and B, thus “squashing out” the nonquantum correlations.”

◮ Oppenheim defines the “puffed entanglement” as the

supremum of I(A : B|E) over all extensions E, with the

  • perational interpretation of the rate required given the side

information is as “useless” as possible.

Nicholas LeCompte Squashed entanglement

slide-40
SLIDE 40

Entanglement measures

Thank you!

Questions?

Nicholas LeCompte Squashed entanglement