Multipartite entanglement detection via projective tensor norms
Based on joint work with Maria Anastasia Jivulescu and Ion Nechita, available at arXiv:2010.06365
Cécilia Lancien
Institut de Mathématiques de Toulouse & CNRS
Mathematical Quantum Physics & Algebra seminar, Innsbruck – March 25 2021
Motivations
A stateρonCd1⊗ · · · ⊗Cdmis calledseparableif it is a convex combination of product states:
ρ=
r
∑
k=1
λkρ1k⊗ · · · ⊗ρmk,with (
λk>0,16k6r,∑rk=1λk=1 ρikstate onCdi,16k6r,16i6m . Otherwise it is calledentangled.
−→If a quantum system is in a separable state, there is no intrinsically quantum correlation between its subsystems. It thus does not provide any advantage over a classical system in information processing tasks.
Known:The problem of deciding whether a given multipartite quantum state is entangled or separable (and even just approximate versions of it) is computationally hard(Gharibian).
−→Solution in practice: Look for necessary conditions to separability, which are easier to check than separability itself, akaentanglement criteria.
Our goal:Design and study a class of entanglement criteria, based on the idea of applyinglocal contractionsto an input multipartite state and computing a suitabletensor normof the output.
Motivations
A stateρonCd1⊗ · · · ⊗Cdmis calledseparableif it is a convex combination of product states:
ρ=
r
∑
k=1
λkρ1k⊗ · · · ⊗ρmk,with (
λk>0,16k6r,∑rk=1λk=1 ρikstate onCdi,16k6r,16i6m . Otherwise it is calledentangled.
−→If a quantum system is in a separable state, there is no intrinsically quantum correlation between its subsystems. It thus does not provide any advantage over a classical system in information processing tasks.
Known:The problem of deciding whether a given multipartite quantum state is entangled or separable (and even just approximate versions of it) is computationally hard(Gharibian).
−→Solution in practice: Look for necessary conditions to separability, which are easier to check than separability itself, akaentanglement criteria.
Our goal:Design and study a class of entanglement criteria, based on the idea of applyinglocal contractionsto an input multipartite state and computing a suitabletensor normof the output.
Motivations
A stateρonCd1⊗ · · · ⊗Cdmis calledseparableif it is a convex combination of product states:
ρ=
r
∑
k=1
λkρ1k⊗ · · · ⊗ρmk,with (
λk>0,16k6r,∑rk=1λk=1 ρikstate onCdi,16k6r,16i6m . Otherwise it is calledentangled.
−→If a quantum system is in a separable state, there is no intrinsically quantum correlation between its subsystems. It thus does not provide any advantage over a classical system in information processing tasks.
Known:The problem of deciding whether a given multipartite quantum state is entangled or separable (and even just approximate versions of it) is computationally hard(Gharibian).
−→Solution in practice: Look for necessary conditions to separability, which are easier to check than separability itself, akaentanglement criteria.
Our goal:Design and study a class of entanglement criteria, based on the idea of applyinglocal contractionsto an input multipartite state and computing a suitabletensor normof the output.
Plan
1 Tensor norms and entanglement
2 Detecting entanglement with testers: definitions and first examples
3 Entanglement testers in the bipartite setting
4 Entanglement testers in the multipartite setting
Tensor norms in Banach spaces
LetA1, . . . ,Ambe Banach spaces. Givenx∈A1⊗ · · · ⊗Am, itsprojective tensor normis kxkA1⊗π···⊗πAm:= inf
( r
∑
k=1
|αk|:aik∈Ai,kaikk61,x=
r
∑
k=1
αka1k⊗ · · · ⊗amk,r∈N )
,
and itsinjective tensor normis
kxkA1⊗ε···⊗εAm:= sup
b1⊗ · · · ⊗bm|x
:bi∈A∗i,kbik61 .
These norms are dual to one another: for allx∈A1⊗ · · · ⊗Am, kxkA1⊗π···⊗πAm= sup
hy|xi:kykA∗
1⊗ε···⊗εA∗m61 , kxkA1⊗ε···⊗εAm= sup
hy|xi:kykA∗
1⊗π···⊗πA∗m61 .
The projective and injective norms are examples oftensor norms: for alla1∈A1, . . . ,am∈Am, ka1⊗ · · · ⊗amkA1⊗π···⊗πAm=ka1⊗ · · · ⊗amkA1⊗ε···⊗εAm=ka1k · · · kamk. And they are extremal among such norms: for any other tensor normk · konA1⊗ · · · ⊗Am,
k · kA1⊗ε···⊗εAm6k · k6k · kA1⊗π···⊗πAm.
Tensor norms in Banach spaces
LetA1, . . . ,Ambe Banach spaces. Givenx∈A1⊗ · · · ⊗Am, itsprojective tensor normis kxkA1⊗π···⊗πAm:= inf
( r
∑
k=1
|αk|:aik∈Ai,kaikk61,x=
r
∑
k=1
αka1k⊗ · · · ⊗amk,r∈N )
,
and itsinjective tensor normis
kxkA1⊗ε···⊗εAm:= sup
b1⊗ · · · ⊗bm|x
:bi∈A∗i,kbik61 . These norms are dual to one another: for allx∈A1⊗ · · · ⊗Am,
kxkA1⊗π···⊗πAm= sup
hy|xi:kykA∗
1⊗ε···⊗εA∗m61 , kxkA1⊗ε···⊗εAm= sup
hy|xi:kykA∗
1⊗π···⊗πA∗m61 .
The projective and injective norms are examples oftensor norms: for alla1∈A1, . . . ,am∈Am, ka1⊗ · · · ⊗amkA1⊗π···⊗πAm=ka1⊗ · · · ⊗amkA1⊗ε···⊗εAm=ka1k · · · kamk. And they are extremal among such norms: for any other tensor normk · konA1⊗ · · · ⊗Am,
k · kA1⊗ε···⊗εAm6k · k6k · kA1⊗π···⊗πAm.
Tensor norms in Banach spaces
LetA1, . . . ,Ambe Banach spaces. Givenx∈A1⊗ · · · ⊗Am, itsprojective tensor normis kxkA1⊗π···⊗πAm:= inf
( r
∑
k=1
|αk|:aik∈Ai,kaikk61,x=
r
∑
k=1
αka1k⊗ · · · ⊗amk,r∈N )
,
and itsinjective tensor normis
kxkA1⊗ε···⊗εAm:= sup
b1⊗ · · · ⊗bm|x
:bi∈A∗i,kbik61 . These norms are dual to one another: for allx∈A1⊗ · · · ⊗Am,
kxkA1⊗π···⊗πAm= sup
hy|xi:kykA∗
1⊗ε···⊗εA∗m61 , kxkA1⊗ε···⊗εAm= sup
hy|xi:kykA∗
1⊗π···⊗πA∗m61 .
The projective and injective norms are examples oftensor norms: for alla1∈A1, . . . ,am∈Am, ka1⊗ · · · ⊗amkA1⊗π···⊗πAm=ka1⊗ · · · ⊗amkA1⊗ε···⊗εAm=ka1k · · · kamk.
And they are extremal among such norms: for any other tensor normk · konA1⊗ · · · ⊗Am, k · kA1⊗ε···⊗εAm6k · k6k · kA1⊗π···⊗πAm.
Characterizing entanglement through tensor norms
•Pure state entanglement:
Banach spaces
Cdi,k · k`di 2
, 16i6m.
Notation:∀x∈Cd,kxk`d
2:= ∑dk=1|xk|21/2 vector 2-norm A pure stateψ∈Cd1⊗ · · · ⊗Cdmis s.t.kψk`d1···dm
2
=1.
And it is separable iffkψk`d1
2⊗ε···⊗ε`dm2 =kψk`d1
2⊗π···⊗π`dm2 =1, where by definition:
kψk`d1
2⊗ε···⊗ε`dm2 := supn
ϕ1⊗ · · · ⊗ϕm|ψ
:ϕi∈Cdi,kϕik`di 2
=1 o
, kψk`d1
2⊗π···⊗π`dm2 := inf r
∑
k=1
|αk|:χik∈Cdi,kχikk`di 2
=1,ψ= ∑r
k=1
αkχ1k⊗ · · · ⊗χmk
.
−→Geometric measure of entanglement(Shimony, Wei/Goldbart, Zhu/Chen/Hayashi): G(ψ) :=−log supn
|
ϕ1⊗ · · · ⊗ϕm|ψ
|2:ϕi∈Cdi,kϕik`di 2
=1 o
=−2logkψk`d1
2⊗ε···⊗ε`dm2 .
•Mixed state entanglement: Banach spaces
M
di(C),k · kSdi1
, 16i6m.
Notation:∀X∈
M
d(C),kXkSd1:= Tr|X| matrix 1-norm A mixed stateρ∈
M
d1(C)⊗ · · · ⊗M
dm(C)is s.t.ρ>0 andkρkSd11···dm
=1. And it is separable iffkρkSd1
1⊗π···⊗πSdm1 =1(Rudolph, Pérez-García), where by definition: kρkS1d1⊗π···⊗πSdm1
:= inf r
∑
k=1
|αk|:τik∈
M
di(C),kτikkSdi1
=1,ρ= ∑r
k=1
αkτ1k⊗ · · · ⊗τmk
.
Characterizing entanglement through tensor norms
•Pure state entanglement:
Banach spaces
Cdi,k · k`di 2
, 16i6m.
Notation:∀x∈Cd,kxk`d
2:= ∑dk=1|xk|21/2 vector 2-norm A pure stateψ∈Cd1⊗ · · · ⊗Cdmis s.t.kψk`d1···dm
2
=1.
And it is separable iffkψk`d1
2⊗ε···⊗ε`dm2 =kψk`d1
2⊗π···⊗π`dm2 =1, where by definition:
kψk`d1
2⊗ε···⊗ε`dm2 := supn
ϕ1⊗ · · · ⊗ϕm|ψ
:ϕi∈Cdi,kϕik`di 2
=1 o
, kψk`d1
2⊗π···⊗π`dm2 := inf r
∑
k=1
|αk|:χik∈Cdi,kχikk`di 2
=1,ψ= ∑r
k=1
αkχ1k⊗ · · · ⊗χmk
.
−→Geometric measure of entanglement(Shimony, Wei/Goldbart, Zhu/Chen/Hayashi): G(ψ) :=−log supn
|
ϕ1⊗ · · · ⊗ϕm|ψ
|2:ϕi∈Cdi,kϕik`di 2
=1 o
=−2logkψk`d1
2⊗ε···⊗ε`dm2 .
•Mixed state entanglement: Banach spaces
M
di(C),k · kSdi1
, 16i6m.
Notation:∀X∈
M
d(C),kXkSd1:= Tr|X| matrix 1-norm A mixed stateρ∈
M
d1(C)⊗ · · · ⊗M
dm(C)is s.t.ρ>0 andkρkSd11···dm
=1. And it is separable iffkρkSd1
1⊗π···⊗πSdm1 =1(Rudolph, Pérez-García), where by definition: kρkS1d1⊗π···⊗πSdm1
:= inf r
∑
k=1
|αk|:τik∈
M
di(C),kτikkSdi1
=1,ρ= ∑r
k=1
αkτ1k⊗ · · · ⊗τmk
.
Characterizing entanglement through tensor norms
•Pure state entanglement:
Banach spaces
Cdi,k · k`di 2
, 16i6m.
Notation:∀x∈Cd,kxk`d
2:= ∑dk=1|xk|21/2 vector 2-norm A pure stateψ∈Cd1⊗ · · · ⊗Cdmis s.t.kψk`d1···dm
2
=1.
And it is separable iffkψk`d1
2⊗ε···⊗ε`dm2 =kψk`d1
2⊗π···⊗π`dm2 =1, where by definition:
kψk`d1
2⊗ε···⊗ε`dm2 := supn
ϕ1⊗ · · · ⊗ϕm|ψ
:ϕi∈Cdi,kϕik`di 2
=1 o
, kψk`d1
2⊗π···⊗π`dm2 := inf r
∑
k=1
|αk|:χik∈Cdi,kχikk`di 2
=1,ψ= ∑r
k=1
αkχ1k⊗ · · · ⊗χmk
.
−→Geometric measure of entanglement(Shimony, Wei/Goldbart, Zhu/Chen/Hayashi): G(ψ) :=−log supn
|
ϕ1⊗ · · · ⊗ϕm|ψ
|2:ϕi∈Cdi,kϕik`di 2
=1 o
=−2logkψk`d1
2⊗ε···⊗ε`dm2 .
•Mixed state entanglement:
Banach spaces
M
di(C),k · kSdi1
, 16i6m.
Notation:∀X∈
M
d(C),kXkSd1:= Tr|X| matrix 1-norm A mixed stateρ∈
M
d1(C)⊗ · · · ⊗M
dm(C)is s.t.ρ>0 andkρkSd11···dm
=1.
And it is separable iffkρkSd1
1⊗π···⊗πSdm1 =1(Rudolph, Pérez-García), where by definition:
kρkS1d1⊗π···⊗πSdm1
:= inf r
∑
k=1
|αk|:τik∈
M
di(C),kτikkSdi1
=1,ρ= ∑r
k=1
αkτ1k⊗ · · · ⊗τmk
.
Plan
1 Tensor norms and entanglement
2 Detecting entanglement with testers: definitions and first examples
3 Entanglement testers in the bipartite setting
4 Entanglement testers in the multipartite setting
Entanglement testers
LetE1, . . . ,En∈
M
d(C)and let{|1i, . . . ,|ni}be an o.n.b. ofCn. Define:E
:X∈M
d(C)7→n
∑
k=1
Tr(Ek∗X)|ki ∈Cn.
If
E
is s.tkE
kSd1→`n2:= maxn
k
E
(X)k`n2 :kXkSd 1=1
o
=1, we call it an(entanglement) tester.
Observation:k
E
kSd1→`n2= max n
(Tr(TE∗X⊗X∗))1/2 :kXkSd
1 =1 o
=1.
TE:= ∑n
k=1
Ek⊗Ek∗,test operatorassociated to
E
A tester
E
:M
d(C)→Cn: EE X
Its associated test operatorTE:Cd⊗Cd→Cd⊗Cd:
E E∗
TE =
4 3
2 1
Detecting entanglement with testers
Theorem (Multipartite entanglement criterion based on testers)
Let
E
i:M
di(C)→Cni, 16i6m, be testers. For anyX∈M
d1(C)⊗ · · · ⊗M
dm(C), we have kE
1⊗ · · · ⊗E
m(X)k`n12⊗π···⊗π`nm2 6kXk
Sd11⊗π···⊗πS1dm
.
Hence, for any stateρonCd1⊗ · · · ⊗Cdm, k
E
1⊗ · · · ⊗E
m(ρ)k`n12⊗π···⊗π`nm2 >1 =⇒ kρk
S1d1⊗π···⊗πSdm1
>1 ⇐⇒ ρentangled.
Proof idea:k
E
1⊗ · · · ⊗E
mkSd11⊗π···⊗πS1dm→`n21⊗π···⊗π`nm2 =k
E
1kSd11→`n21· · · k
E
mkSdm1 →`nm2 =1.
factorization property The action of testers
E
i:M
di(C)→Cni, 16i6m,on a stateρ∈
M
d1(C)⊗ · · · ⊗M
dm(C):ρ d1 d1 d2 d2
dm dm E1
E2
Em n1
n2
nm
Practical interest of tester-based entanglement criteria
Entanglement criterion based on reducing the study of mixed state entanglement (checking if an (Sd1)⊗πmnorm is>1) to that of pure state entanglement (checking if an(`n2)⊗πmnorm is>1).
Observation:`n2⊗π`n2≡S1n. Identification|xi=
n
∑
k,l=1
xkl|kli ∈Cn⊗Cn↔X=
n
∑
k,l=1
xkl|kihl| ∈
M
n(C), s.t.kxk`n2⊗π`n2=kXkSn1 (Schmidt decomposition↔singular value decomposition).•Bipartite case: Checking if testers
E
,F
:M
d(C)→Cndetect the entanglement of a stateρon Cd⊗Cdconsists in computing the`n2⊗π`n2norm ofE
⊗F
(ρ)∈Cn⊗Cn, i.e. itsS1nnorm if seen as an element ofM
n(C).−→This is much easier than computing theS1d⊗πS1dnorm ofρ.
•Multipartite case: The computation of an(Sd1)⊗πmnorm, i.e. of an(`d2⊗π`d2)⊗πm≡(`d2)⊗π2m norm, is reduced to the computation of an(`n2)⊗πmnorm. associativity of⊗π
−→Reduction by a factor 2 of the number of factors (at the cost of potentially increasing the dimension of each of them fromdton>d...)
Practical interest of tester-based entanglement criteria
Entanglement criterion based on reducing the study of mixed state entanglement (checking if an (Sd1)⊗πmnorm is>1) to that of pure state entanglement (checking if an(`n2)⊗πmnorm is>1).
Observation:`n2⊗π`n2≡S1n. Identification|xi=
n
∑
k,l=1
xkl|kli ∈Cn⊗Cn↔X=
n
∑
k,l=1
xkl|kihl| ∈
M
n(C), s.t.kxk`n2⊗π`n2=kXkSn1 (Schmidt decomposition↔singular value decomposition).•Bipartite case: Checking if testers
E
,F
:M
d(C)→Cndetect the entanglement of a stateρon Cd⊗Cdconsists in computing the`n2⊗π`n2norm ofE
⊗F
(ρ)∈Cn⊗Cn, i.e. itsS1nnorm if seen as an element ofM
n(C).−→This is much easier than computing theS1d⊗πS1dnorm ofρ.
•Multipartite case: The computation of an(Sd1)⊗πmnorm, i.e. of an(`d2⊗π`d2)⊗πm≡(`d2)⊗π2m norm, is reduced to the computation of an(`n2)⊗πmnorm. associativity of⊗π
−→Reduction by a factor 2 of the number of factors (at the cost of potentially increasing the dimension of each of them fromdton>d...)
Practical interest of tester-based entanglement criteria
Entanglement criterion based on reducing the study of mixed state entanglement (checking if an (Sd1)⊗πmnorm is>1) to that of pure state entanglement (checking if an(`n2)⊗πmnorm is>1).
Observation:`n2⊗π`n2≡S1n. Identification|xi=
n
∑
k,l=1
xkl|kli ∈Cn⊗Cn↔X=
n
∑
k,l=1
xkl|kihl| ∈
M
n(C), s.t.kxk`n2⊗π`n2=kXkSn1 (Schmidt decomposition↔singular value decomposition).•Bipartite case: Checking if testers
E
,F
:M
d(C)→Cndetect the entanglement of a stateρon Cd⊗Cdconsists in computing the`n2⊗π`n2norm ofE
⊗F
(ρ)∈Cn⊗Cn, i.e. itsS1nnorm if seen as an element ofM
n(C).−→This is much easier than computing theS1d⊗πS1dnorm ofρ.
•Multipartite case: The computation of an(Sd1)⊗πmnorm, i.e. of an(`d2⊗π`d2)⊗πm≡(`d2)⊗π2m norm, is reduced to the computation of an(`n2)⊗πmnorm. associativity of⊗π
−→Reduction by a factor 2 of the number of factors (at the cost of potentially increasing the dimension of each of them fromdton>d...)
Important examples of testers
•Maps defined from matrix bases:
Let{G1, . . . ,Gd2}be an o.n.b. of
M
d(C)and define:G
:X∈M
d(C)7→d2
∑
k=1
Tr(G∗kX)|ki ∈Cd2. Clearlyk
G
(X)k`d22
=kXkSd
26kXkSd
1. So
G
is indeed a tester.Example:
R
:X∈M
d(C)7→d
∑
k,k0=1
Tr(R∗kk0X)|kk0i ∈Cd2, whereRkk0:=|kihk0|, 16k,k06d.
Associated test operator:TR=
d
∑
k,k0=1
Rkk0⊗Rkk∗0=
d
∑
k,k0=1
|kihk0| ⊗ |k0ihk|=F.
•Maps defined from vector 2-designs:
Let{|x1i, . . . ,|xd2i}be a symmetric spherical 2-design ofCd, i.e.d12
d2
∑
k=1
|xkihxk|⊗2=d(Id++F1).
SetSk:= qd+1
2d |xkihxk|, 16k6d2, and define:
S
:X∈M
d(C)7→d2
∑
k=1
Tr(Sk∗X)|ki ∈Cd2. We havek
S
(X)k`d22
= 12(|TrX|2
| {z }
6kXk2
Sd1
+ Tr|X|2
| {z }
=kXk2
Sd2
6kXk2
Sd1
)1/2
6kXkSd
1. So
S
is indeed a tester.Associated test operator:TS=
d2
∑
k=1
Sk⊗Sk∗=d+1 2d
d2
∑
k=1
|xkihxk|⊗2=I+F 2 .
Important examples of testers
•Maps defined from matrix bases:
Let{G1, . . . ,Gd2}be an o.n.b. of
M
d(C)and define:G
:X∈M
d(C)7→d2
∑
k=1
Tr(G∗kX)|ki ∈Cd2. Clearlyk
G
(X)k`d22
=kXkSd
26kXkSd
1. So
G
is indeed a tester.Example:
R
:X∈M
d(C)7→d
∑
k,k0=1
Tr(R∗kk0X)|kk0i ∈Cd2, whereRkk0:=|kihk0|, 16k,k06d.
Associated test operator:TR=
d
∑
k,k0=1
Rkk0⊗Rkk∗0=
d
∑
k,k0=1
|kihk0| ⊗ |k0ihk|=F.
•Maps defined from vector 2-designs:
Let{|x1i, . . . ,|xd2i}be a symmetric spherical 2-design ofCd, i.e.d12
d2
∑
k=1
|xkihxk|⊗2=d(Id++F1).
SetSk:=
qd+1
2d |xkihxk|, 16k6d2, and define:
S
:X∈M
d(C)7→d2
∑
k=1
Tr(Sk∗X)|ki ∈Cd2. We havek
S
(X)k`d22
= 12(|TrX|2
| {z }
6kXk2
Sd1
+ Tr|X|2
| {z }
=kXk2
Sd2
6kXk2
Sd1
)1/2
6kXkSd
1. So
S
is indeed a tester.Associated test operator:TS=
d2
∑
k=1
Sk⊗Sk∗=d+1 2d
d2
∑
k=1
|xkihxk|⊗2=I+F 2 .
Realignment and SIC POVM entanglement criteria in the tester framework
•Let
R
:X∈M
d(C)7→d
∑
k,k0=1
hk0|X|ki|kk0i ∈Cd2be the matrix unit tester.
For any stateρonCd⊗Cd, we have:
R
⊗2(ρ) = ∑dk,k0,l,l0=1
hk0l0|ρ|kli|kk0i|ll0i. Seen as a matrix, this isρR, the realignment ofρ.
ρR
ρR = ρ
−→Computing the`d22⊗π`d22norm of
R
⊗2(ρ)is computing theS1d2norm ofρR. The tester criterion based onR
is therealignment criterion(Chen/Wu, Rudolph).•Let
S
:X∈M
d(C)7→ rd+12d
d2
∑
k=1
hxk|X|xki|ki ∈Cd2be the 2-design tester.
For any stateρonCd⊗Cd, we have:
S
⊗2(ρ) =d2d+1 d2
∑
k,l=1
hxkxl|ρ|xkxli|ki|li. Seen as a matrix, this isρS, the SIC POVM transformation ofρ.
−→Computing the`d22⊗π`d22norm of
S
⊗2(ρ)is computing theSd12norm ofρS. The tester criterion based onS
is theSIC POVM criterion(Shang/Asadian/Zhu/Gühne).Interest:The realignment and SIC POVM criteria are originally defined only for bipartite states. Viewing them inside the tester framework allows for an immediate multipartite generalization.
Realignment and SIC POVM entanglement criteria in the tester framework
•Let
R
:X∈M
d(C)7→d
∑
k,k0=1
hk0|X|ki|kk0i ∈Cd2be the matrix unit tester.
For any stateρonCd⊗Cd, we have:
R
⊗2(ρ) = ∑dk,k0,l,l0=1
hk0l0|ρ|kli|kk0i|ll0i. Seen as a matrix, this isρR, the realignment ofρ.
ρR
ρR = ρ
−→Computing the`d22⊗π`d22norm of
R
⊗2(ρ)is computing theS1d2norm ofρR. The tester criterion based onR
is therealignment criterion(Chen/Wu, Rudolph).•Let
S
:X∈M
d(C)7→rd+1 2d
d2
∑
k=1
hxk|X|xki|ki ∈Cd2be the 2-design tester.
For any stateρonCd⊗Cd, we have:
S
⊗2(ρ) =d2d+1 d2
∑
k,l=1
hxkxl|ρ|xkxli|ki|li. Seen as a matrix, this isρS, the SIC POVM transformation ofρ.
−→Computing the`d22⊗π`d22norm of
S
⊗2(ρ)is computing theSd12norm ofρS. The tester criterion based onS
is theSIC POVM criterion(Shang/Asadian/Zhu/Gühne).Interest:The realignment and SIC POVM criteria are originally defined only for bipartite states. Viewing them inside the tester framework allows for an immediate multipartite generalization.
Realignment and SIC POVM entanglement criteria in the tester framework
•Let
R
:X∈M
d(C)7→d
∑
k,k0=1
hk0|X|ki|kk0i ∈Cd2be the matrix unit tester.
For any stateρonCd⊗Cd, we have:
R
⊗2(ρ) = ∑dk,k0,l,l0=1
hk0l0|ρ|kli|kk0i|ll0i. Seen as a matrix, this isρR, the realignment ofρ.
ρR
ρR = ρ
−→Computing the`d22⊗π`d22norm of
R
⊗2(ρ)is computing theS1d2norm ofρR. The tester criterion based onR
is therealignment criterion(Chen/Wu, Rudolph).•Let
S
:X∈M
d(C)7→rd+1 2d
d2
∑
k=1
hxk|X|xki|ki ∈Cd2be the 2-design tester.
For any stateρonCd⊗Cd, we have:
S
⊗2(ρ) =d2d+1 d2
∑
k,l=1
hxkxl|ρ|xkxli|ki|li. Seen as a matrix, this isρS, the SIC POVM transformation ofρ.
−→Computing the`d22⊗π`d22norm of
S
⊗2(ρ)is computing theSd12norm ofρS. The tester criterion based onS
is theSIC POVM criterion(Shang/Asadian/Zhu/Gühne).Interest:The realignment and SIC POVM criteria are originally defined only for bipartite states.
Viewing them inside the tester framework allows for an immediate multipartite generalization.
Plan
1 Tensor norms and entanglement
2 Detecting entanglement with testers: definitions and first examples
3 Entanglement testers in the bipartite setting
4 Entanglement testers in the multipartite setting
Detecting entanglement of bipartite pure states with symmetric testers
A tester
E
:M
d(C)→Cnis calledsymmetricif there existα,β∈Rs.t.TE=αF+βI.Fact:The set of admissible pairs(α,β)is{(α,1−α) :06α61} ∪ {(1,β) :−1/d6β60}. Examples:
R
andS
are of this form, with(α,β) = (1,0)and(α,β) = (1/2,1/2), respectively.[Note: A symmetric tester necessarily hasn>d2. So
R
andS
are ‘minimal’ symmetric testers.]Theorem (Performance of symmetric testers on bipartite pure states)
Let
E
:M
d(C)→Cnbe a symmetric tester, with admissible parameters(α,β). For any pure stateϕ∈Cd⊗Cd, with Schmidt decomposition|ϕi=∑rk=1√
λk|ekfki, we have k
E
⊗2(|ϕihϕ|)k`n2⊗π`n2=α+β+2α∑
16k<l6r
p λkλl.
In particular: k
R
⊗2(|ϕihϕ|)k`d22⊗π`d22=1+2 ∑
16k<l6r
√
λkλlandk
S
⊗2(|ϕihϕ|)k`d22⊗π`d22=1+ ∑
16k<l6r
√ λkλl. Sok
R
⊗2(|ϕihϕ|)k`d22⊗π`d22>k
S
⊗2(|ϕihϕ|)k`d22 ⊗π`d22>1, with strict inequalities wheneverr>1.
−→
R
andS
detect all entangled pure states, andR
always performs ‘better’ at it thanS
.In fact:k
R
⊗2(|ϕihϕ|)k`d22⊗π`d22=k|ϕihϕ|kSd
1⊗πSd1. So