pr olog u e ÔÔÔb eam u s u p sco tt y!Õ Õ · ou t lin e qubits and qu an t u m information ßo...
TRANSCRIPT
PROLOGUE
‘‘BEAM US UP SCOTTY!’’‘‘BEAM US UP SCOTTY!’’‘‘BEAM US UP SCOTTY!’’
•••
‘‘HOW DO I DO THAT?’’‘‘HOW DO I DO THAT?’’
•••
‘‘HERE’S THE CODE’’‘‘HERE’S THE CODE’’‘‘HERE’S THE CODE’’
1
The physicist’s description
Alice has an ‘unknown’ qubit |φ〉 and wants to send itto Bob. They have the ability to communicate classicalbits, and they share an entangled pair in the EPR-state, that is 1√
2(|00〉 + |11〉), which Alice produced
by first applying a Hadamard-gate 1√2
„1 11 −1
«to
the first qubit of a qubit pair in the ground state |00〉,and by then applying a CNOT-gate, that is
|00〉 #→ |00〉 |01〉 #→ |01〉 |10〉 #→ |11〉 |11〉 #→ |10〉,
then she sends the first qubit of the pair to Bob.
To teleport her qubit, Alice first performs a bipar-tite measurement on the unknown qubit and her halfof the entangled pair in the Bell-base, that isn
|0x〉+ (−1)z|1(1− x)〉˛̨̨
x, z ∈ {0, 1}o
,
where we denote the four possible outcomes of themeasurement by xz. Then she sends the 2-bit out-come xz to Bob using the classical channel.
Then, if x = 1, Bob performs the unitary oper-
ation σx =
„0 11 0
«on its half of the shared en-
tangled pair, and he also performs a unitary operation
σz =
„1 00 −1
«on it if z = 1. Now Bob’s half of
the initially entangled pair is in state |φ〉.
2
Ours
ρQ ; 1⊗ !1" ;α ; 〈#βi$〉i=4i=1 ⊗ 1 ; DIST ; ⊕i=4
i=1 β−1i
3
I.e.
Q
produce EPR-pair
Q⊗ (Q∗⊗Q)
(1⊗ !1") ◦ ρ
!
spatial relocation
(Q⊗Q∗)⊗Q
α
!
Bell-base measurement
(4 · I)⊗Q
〈#βi$〉i=4i=1 ⊗ 1
!
classical communication
4 · Q
DIST
!
unitary correction
4 · Q
⊕i=4i=1β−1
i
!
4
The physicist’s proof
In the case that the measurement outcome of the Bell-base measurement is xz, for
Pxz := 〈0x + (−1)z1(1− x)|−〉|0x + (−1)z1(1− x)〉we have to apply Pxz ⊗ id to the input state
|φ〉 ⊗ 1√2(|00〉+ |11〉) .
Setting |φ〉 := φ0|0〉+ φ1|1〉 we rewrite the input as
1√2(φ0|000〉+ φ0|011〉+ φ1|100〉+ φ1|111〉)
=1√2(φ0
Xx=0,1
|0xx〉+ φ1X
x=0,1
|1(1− x)(1− x)〉)
and application of Pxz ⊗ id then yields
1√2|0x + (−1)z1(1− x)〉 ⊗ (φ0|x〉+ (−1)zφ1|1− x〉).
There are four cases concerning the unitary correctionsUxz which have to be applied. For x = z = 0 the thirdqubit is φ0|0〉 + φ1|1〉 = |φ〉. If x = 0 and z = 1 it is
φ0|0〉 − φ1|1〉 which after applying σz =
„1 00 −1
«becomes |φ〉. If x = 1 it is φ0|1〉 + (−1)zφ1|0〉 which
after applying σx =
„0 11 0
«brings us back to the
previous two cases. Hence the result follows.
5
Ours
=
=β−1
i β−1i
βi
βi
1
6
Quantum formalism := vonNeumann [1932]
“I would like to make a confession which mayseem immoral: I do not believe absolutely inHilbert space no more.”
von Neumann [1935]
???von Neumann QM
! high-level languagelow-level language
Defects of von Neumann’s quantum formal-ism for Quantum Informatics:
• Types do not reflect kinds,
• Classical data-flow isn’t addressed,
• Quantum data-flow i.e. the featurewhich enables teleportation, is hidden.
7
OUTLINE
Qubits and quantum information flow.
B. Coecke. “The logic of entanglement”. OUCL-
PRG-RR-03-12 http://web.comlab.ox.ac.uk/oucl/
publications/tr/rr-03-12.html
From von Neumann quantum mechan-ics to high-level quantum mechanics.
S. Abramsky and B. Coecke. “A categorical seman-
tics of quantum protocols”. 19th IEEE Conference on
Logic in Computer Science (LiCS’04) www.arXiv.org/
quant-ph/0402130
On transdisciplinary opportunities andmethodological contributions of com-puter science to the other sciences.
8
A bit admits
• two values 0, 1,
• it is freely readable,
• it admits arbitrary transformations.
A qubit admits
• a sphere of values spanned by |0〉, |1〉,
|1〉
|0〉
|ψ〉|+〉
|−〉
• a measurement of it
– has two outcomes |−〉, |+〉,– changes the value |ψ〉,
• it admits unitary transformations i.e. op-posites and angles are preserved.
9
|ψ〉
|+〉
|−〉
θ−θ+
P+ : |ψ〉 &→ |+〉 P− : |ψ〉 &→ |−〉have chance prob(θ&) for & ∈ {+,−}. On thesphere Q we obtain partial constant maps
P& : Q → Q :: |ψ〉 &→ | & 〉.We know the value after a measurement, butnot what it was before it.
• Bad: measurements destroy data.
• Good: measurements act on data.
Quantum computing and quantum protocols:acrobatics between “the good” and “the bad”.
10
Q.M. in a nutshell
Systems are described by vectors (up to ascalar) of an inner-product space over C.
Compoundness is described by −⊗−.
Operations is described by unitaries.
Measurements “are described” by self-adjointoperators H : H→ H i.e.
H = a1 · P1 + . . . + a1 · Pn .
The measurement process constitutes:
• One Pi in (P1, . . . ,Pn) happens.
• The observer receives the token ai.
• The probability of this is |Pi(ψ)|2.
11
Quantum teleportation [Bennett et al. 1993]
|Ψ〉
{Px}x
Ux
|φ〉
|φ〉
Continuous data transmission through 2 bits?
So what causes the magic?
A pair of qubits isn’t described by a pair
|ψ1,ψ2〉 ∈ Q1 ×Q2
but by a function
|f : Q1→ Q2〉 ∈ Q1 ⊗Q2
12
Indeed, the tensor product H1⊗H2 of Hilbertspaces H1,H2 is the Hilbert space spanned by
|11〉 · · · |1n〉...
. . ....
|k1〉 · · · |kn〉hence,
∑ij
mij |ij〉 +←→ m11 · · · m1n
.... . .
...mk1 · · · mkn
+←→ | i 〉 &→
∑j
mij | j 〉
e.g. for k = n,
|11〉+ . . . + |nn〉 +←→ | i 〉 &→ | i 〉
Pairs |ψ1,ψ2〉 are a special case in H1⊗ H2
— reducing to |ij〉 in a well-chosen base.
13
The identity
|id : Q → Q 〉 ∈ Q⊗Q
is the Einstein-Podolsky-Rosen state.
A measurement of Q⊗Q has four outcomes
|f1〉, |f2〉, |f3〉, |f4〉cf.
|00〉, |01〉, |10〉, |11〉and corresponding
Pf : Q⊗Q → Q⊗Q :: |g〉 &→ | f 〉e.g.
Pid : Q⊗Q → Q⊗Q :: |g〉 &→ |id〉produces the EPR state.
Do Q⊗Q-functions compose (in some way)?
They do!
14
|φin〉
|φout〉?
f1
f2
f3
f4
φout =(f2 ; f1 ; f †3 ; f †
2 ; f4 ; f3)(φin)
15
f1
f2
f3
f4
f2 ; f1 ; f †3 ; f †
2 ; f4 ; f3
16
Permitted
Forbidden
17
id
id
id ; id = id
⇒ Teleportation
18
1 ≤ i ≤ 4
id
βi
β−1i
βi ; id; β−1i = id
19
f
βi
γi
βi ; f ; γi = f
⇒ Logic-gateteleportation
20
id
id id
id ; id; id = id
⇒ Entanglementswapping
21
f2
f1
φin
φout = (f2 ◦ f1)(φin)
((Pf1⊗ 1) ◦ (1⊗Pf2)
)(φin ⊗ . . .)
= . . .⊗(f2◦f1)(φin)
Hilbert space ! setlinear function ! relationtensor product ! product
PR := R×R
={((x1, y1), (x2, y2))
∣∣ x1Ry1, x2Ry2
}(sin, · , · )((PR1⊗1)◦(1⊗PR2)
)( · , · , sout)
⇒ sin(R2◦R1)sout
22
High-level quantum mechanics
A category C has objects A,B,C, . . . andfor each two objects A,B a set of morphismsC(A,B). It also has idA ∈ C(A,A) andcomposition g ◦ f when types match.
Why categories? Since they allow to deal ex-plicitly with processes e.g.
Logic ProgrammingPropositions I/O-Types
Proofs Programs
Variables (e.g. qubits Q) ∼ objects
Processes (e.g. measurements) ∼ morphisms
Compoundness (e.g. Q⊗Q) ∼ morphism sets
23
We also want an associative connective ⊗which encodes compound systems as objects.
A symmetric monoidal category is a cat-egory in which we can “plug things together”,things being both objects and morphisms:
A⊗B f1 ⊗ f2 : A1 ⊗A2 → B1 ⊗B2
There exist natural operations e.g. swapping,
σA,B : A⊗B → B ⊗A,
which satisfy obvious behavioral rules e.g.
A1 ⊗A2f1 ⊗ f2" B1 ⊗B2
A2 ⊗A1
σA1,A2
! f2 ⊗ f1" B2 ⊗B1
σB1,B2
!
24
There is some locality for components
A1 ⊗A2f1 ⊗ id" B1 ⊗A2
A1 ⊗B2
id⊗ f2
!
f1 ⊗ id" B1 ⊗B2
id⊗ f2
!
But we don’t assume a pair-like structure, i.e.
whole /= ∑components
This turns out to comprise the absence of
A∆−→ A⊗A A⊗B
π−→ A
25
No-cloning & No-deleting
Wooters-Zurek 1982; Pati-Braunstein 2000
Crucial for secure quantum cryptography.
Is this connected to resource sensitive lan-guages and recent models for concurrency andinteraction in distributed computation?
Yes
Categories will enable us to capture the quan-tum information flow (= geometry of the line)syntactically, and compare it to the resource-sensitive and concurrent CS semantics.
No-cloning & no-deleting refute existence of
A∆−→ A⊗A A⊗B
π−→ A
A 0 A ∧A A ∧B 0 A
What more do we need for quantum theory?
26
Dissection of Pf
id
id
f
f
Pf = f ⊗ id ; #id$ ; !id" ; id⊗ f
(cf. unraveling a program into commands)
27
The algebra of entanglement
1. a monoidal involution on objects,
A &→ A∗
2. for each involutive pair a morphism
!id" : I → A⊗A∗
3. a monoidal duality on morphisms,
f : A → B &→ f† : B → A
such that
id
id
=
28
=
=
=
29
Scalars:
• Unit 1 as idI.
• Integers as #idA$ ◦ !idA"
• Reals as s ◦ s†
• Probabilities [0, 1] as |〈φ | ψ〉|.contained in C(I, I) for I the ⊗-unit, and eachscalar induces “natural” scalar multiplication.
Unitarity:U† = U−1
Inner-product:
〈φ | ψ〉 := φ† ◦ ψ
for φ : I →A and ψ : I → B. We have:
〈U ◦ ψ | U ◦ φ〉 = 〈ψ | φ〉〈f† ◦ ψ | φ〉 = 〈ψ | f ◦ φ〉
30
Exampleobjects: sets
morphisms: relations
tensor: cartesian product
Additional data:
X∗ := X R† := Rc
η := {(∗, (x, x)) | x ∈ X} ⊆ {∗}× (X ×X)
Matrices in any involutive abelian semiringprovide a model generalizing relations, thatis, matrices in the Boolean semiring.
Other examples are inner-product spaces, freecompact closed categories generated by self-dual categories, n-cobordisms, . . .
31
What do we miss to have all ingredi-ents of von Neumann’s formalism?
An operation ⊕ on C which now stands for
“plugging histories together”
where history = picture = branch = world,and which has pair-like nature (= biproduct).
Disjunctive/probabilistic content:
Q⊕Q
Producing a “sphere” from two “germs”:
Q 4 I⊕ I
32
A qubit measurement becomes
〈P+,P−〉 : Q → Q⊕Q
whereP& = π†
& ◦ π& : Q → Q
with〈π†
+,π†−〉 : Q → I⊕ I
unitary.
Classical information flow is distributivity,
(I⊕ I)⊗QDIST" (I⊗Q)⊕ (I⊗Q) 4 Q⊕Q
(Q1 ⊕Q2)⊗QDIST" (Q1 ⊗Q)⊕ (Q2 ⊗Q)
33
Theorem.
Q ============== Q
produce EPR-pair
Q⊗ (Q∗⊗Q)
(id⊗ !id") ◦ ρ
!
spatial relocation
(Q⊗Q∗)⊗Q
+!
Bell-base measurement
(4 · I)⊗Q
〈#βi$〉i=4i=1 ⊗ id
!
classical communication
4 · Q
DIST
!
unitary correction
4 · Q
〈 id 〉i=4i=1
!=========== 4 · Q
⊕i=4i=1β−1
i
!
i.e.
ρQ ; id⊗!id" ; α ; 〈#βi$〉i=4i=1⊗id ; DIST ;⊕i=4
i=1β−1i = id
34
=
=β−1
i β−1i
βi
βi
1
35
Beyond von Neumann’s 1932 Q.M.
We increased resolution. We now see thequantum information flow. We extendedscope. We added classical data and informa-tion flow. Types reflect kinds, and thereare many more of them.
The formalism is high-level. We have acompact purely formal language and intuitivegeometrical proofs, but with a categorical log-ical and syntactic counterpart, namely linearlogic with dual self-dual connectives.
We emphasized essence. We abstractover non-crucial things and allowed new de-grees of axiomatic freedom, enabling rela-tional reasoning: sets, relations and thecartesian product satisfy the axiomatics.
A bonus. The probability rule can be de-rived in our formalism while in von Neumann’sformalism it is a postulate.
36
Direct applications
Intuitive analysis and design of computa-tional schemes and protocols, and general op-erational reasoning about quantum systems.
A step stone towards automated analysis,design and reasoning about quantum systems
E.g. measurement based quantum com-puting is both from soft- and hardware per-spective very promising, and relies completelyon the structure of entanglement.
In particular, qualifying and quantifying mul-tipartite entanglement and obtaining abetter understanding of its properties is thecurrent holy grail of quantum information.
37
A transdisciplinary perspective
What did we do so far? We definitely did
• high-level quantum informatics,
but, also
• physics in informatic perspective.
Why did it take 60 years and 6 persons for
• teleportation to be invented?
Why did it took 70 years to have
• a complete quantum formalism and logic
• which recognizes the compositional na-ture of entanglement,
and as such
• trivializes teleportation?
38
Since many of the tools were either
• not available (yet), or,
• not being considered (yet).
Now they do exist and were developed by
• computer scientists.
So why didn’t physicist invent them?
• Many physicists couldn’t care less.
• They looked in the wrong direction.
• They used inappropriate methodologies.
• They followed old-fashion paradigms.
39
Computer science has something to offer toother sciences — other than “the computer”!
• Logical and structural reasoning.
• The operational methodology.
• Fresh paradigms.
E.g.
• Interaction and concurrency.
• Open systems.
• Qualitative reasoning about information.
• Continuous vs. discrete.
• Hybrid systems.
40
EPILOGUE
id
βi
β−1i
41