super-duper-activation of quantum zero-error capacities
DESCRIPTION
Super-Duper-Activation of Quantum Zero-Error Capacities. Toby Cubitt, Jenxin Chen, Aram Harrow arXiv:0906.2547 and Toby Cubitt, Graeme Smith arXiv:0912.2737. (qu)bits. (qu)bits. …. …. n. Channel Capacities. (qu)bits. (qu)bits. …. …. n. Zero-Error Channel Capacities. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/1.jpg)
Toby Cubitt, Jenxin Chen, Aram Harrow
arXiv:0906.2547and
Toby Cubitt, Graeme Smith
arXiv:0912.2737
Super-Duper-Activation of Quantum Zero-Error Capacities
![Page 2: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/2.jpg)
Channel Capacities
(qu)bits (qu)bits…
n
…
![Page 3: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/3.jpg)
Zero-Error Channel Capacities
(qu)bits (qu)bits…
n
…
![Page 4: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/4.jpg)
Regularization required Additivity violation Superactivation
Cla
ssical
chan
nels
Qu
antu
m
chan
nels
Cla
ssica
lQ
ua
ntum
?
Status of Zero-Error Capacity Theory
(trivial)
[Alon ’97]
[Shannon ’56]
[Duan, ’09]
?
![Page 5: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/5.jpg)
Superactivation of the Zero-Error Classical Capacity of Quantum Channels
• TheoremFor any satisfying , there exist channels such that:
• Each channel maps and has Kraus operators.
• Each channel individually has no zero-error classical capacity at all.
• The joint channel has positive zero-error classical capacity.
![Page 6: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/6.jpg)
Status of Zero-Error Capacity TheoryRegularization required Additivity violation Superactivation
Cla
ssical
chan
nels
Qu
antu
m
chan
nels
Cla
ssica
lQ
ua
ntum
(trivial)
[Alon ’97]
[Shannon ’56]
![Page 7: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/7.jpg)
Superduperactivation of Zero-Error Capacities of Quantum Channels
• TheoremFor any satisfying , there exist channels such that:
• Each channel maps and has Kraus operators.
• Each channel individually has no classical zero-error capacity
• The joint channel even has positive quantum zero-error capacity
(hence, also no quantum zero-error).
(hence, every other capacity is also positive).
![Page 8: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/8.jpg)
1. Reduce super(duper)activation to question about existence of certain subspaces.
2. Show that such subspaces exist.
![Page 9: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/9.jpg)
![Page 10: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/10.jpg)
Reducuctio ad Subspace
• Want two channels such that:
A channel has positive zero-error capacity iff there exist two different inputs that are mapped to outputs states that are perfectly distinguishable.
![Page 11: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/11.jpg)
Reducuctio ad Subspace
• Want two channels such that:
• The zero-error capacity is non-zero iff
• Conversely, the zero-error capacity is zero iff
![Page 12: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/12.jpg)
Reducuctio ad Subspace
• Want two channels such that:
• The zero-error capacity is non-zero iff
• Conversely, the zero-error capacity is zero iff
![Page 13: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/13.jpg)
Reducuctio ad Subspace
• Want two channels such that:
Translate these into statements about the supports of the Choi-Jamiołkowski matrices of the composite maps
![Page 14: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/14.jpg)
Reducuctio ad Subspace
• Want a bipartite subspace such that:
(This is similar to the argument for p = 0 minimum output entropies in[Cubitt, Harrow, Leung, Montanaro, Winter, CMP 284, 281 (2008)])
![Page 15: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/15.jpg)
(Almost) any old subspace will do!
• Def.
The set of subspaces that have a tensor power whose orthogonal complement
does contains a product state.
=
The subspaces we don’t want.
![Page 16: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/16.jpg)
(Almost) any old subspace will do!
• Proof intuition:Think of (unnormalised) bipartite subspaces as (projective) matrix spaces:
Product states$ rank-1 matrices$ all order-2 minors vanish$ set of simultaneous polynomials$ Segre variety
• Lemma: Ed is an algebraic set≡ defined by simultaneous polynomial equations≡ Zariski closed in the Grassmanian)
![Page 17: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/17.jpg)
(Almost) any old subspace will do!
• An algebraic set set is either measure zero (in the usual Haar measure) or it is the entire space.
(Intuitively, it’s defined by some algebraic constraints, so either the constraints are trivial, or they restrict the set to some lower-dimensional manifold.)
GrdGrd
Ed
Ed
![Page 18: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/18.jpg)
(Almost) any old subspace will do!
• To show Ed is measure zero, just have to exhibit one subspace that’s not contained in Ed.
Grd
Ed
Grd
• Use a subspace whose orthogonal complement is spanned by an unextendible product basis (UPB), which exist for a large range of dimensions (! mild dimension constraints)
• Lemma: Tensor products of UPBs are again UPBs.
UPB
Ed
![Page 19: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/19.jpg)
(Almost) any old subspace will do!
• The set of “bad” subspaces is zero-measure (in the
usual Haar measure on the Grassmanian), so the set of
“good” ones is full measure ! pick a one at random!
Grd
Ed
![Page 20: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/20.jpg)
Reducuctio ad Subspace
• Want a bipartite subspace such that:
pick one atrandom!
Subspaces obeyingsymmetry constraints
also form an algebraic set
![Page 21: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/21.jpg)
(Almost) any subspace will do!
• Want a bipartite subspace such that:
full measure
non-zeromeasure
pick one atrandom!
![Page 22: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/22.jpg)
![Page 23: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/23.jpg)
Superduperactivation of Zero-Error Capacities of Quantum Channels
• Translates into one additional constraint on the
subspace :
full measure
non-zeromeasure
pick one atrandom!
![Page 24: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/24.jpg)
Conclusions
• Unwind everything(!) to get channels that give super(duper)activation of the asymptotic zero-error capacity.
• Quantum channels not only behave in this extremely weird way for quantum information [Smith, Yard ’07],but also for classical information too.
• A corollary of this result is that the regularized Rényi-0 entropy is non-additive. If we could prove the same thing for the regularized Rényi-1 (von Neumann) entropy, would prove non-additivity of the classical Shannon capacity of quantum channels.
![Page 25: Super-Duper-Activation of Quantum Zero-Error Capacities](https://reader035.vdocuments.us/reader035/viewer/2022062222/568157d3550346895dc559a8/html5/thumbnails/25.jpg)
Status of Shannon Capacity TheoryRegularization required Additivity violation Superactivation
Cla
ssical
chan
nels
Qu
antu
m cha
nne
ls
Cla
ssical
Qu
antu
m
* probably
[Shannon ’48]
[Shannon ’48]
(trivial)
*
[Hasting ’08]
? (trivial)
*
[DiVincenzo, Shor, Smolin ’98]
[Smith, Yard ’08]