quantum information theory: present status and future directions

44
Quantum Information Theory: Present Status and Future Directions Julia Kempe Julia Kempe CNRS & LRI, Univ. de Paris-Sud, Orsay, France Newton Institute, Cambridge, August 24 th , 2004 Complexity of Local Hamilton Complexity of Local Hamilton

Upload: hope-meyers

Post on 31-Dec-2015

20 views

Category:

Documents


0 download

DESCRIPTION

Newton Institute, Cambridge, August 24 th , 2004. Quantum Information Theory: Present Status and Future Directions. The Complexity of Local Hamiltonians. Julia Kempe CNRS & LRI, Univ. de Paris-Sud, Orsay, France. Also implies:. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Quantum Information Theory:  Present Status and Future Directions

Quantum Information Theory: Present Status and Future Directions

Julia KempeJulia KempeCNRS & LRI, Univ. de Paris-Sud, Orsay, France

Newton Institute, Cambridge, August 24th, 2004

The Complexity of Local HamiltoniansThe Complexity of Local Hamiltonians

Page 2: Quantum Information Theory:  Present Status and Future Directions

Joint work with Joint work with Oded Oded RegevRegev and and Alexei KitaevAlexei Kitaev

Result: 2-local Hamiltonian is QMA complete

J. K., Alexei Kitaev and Oded Regev, quant-ph/0406180

2-local adiabatic computation is equivalent to standard quantum computation

Also implies:

Page 3: Quantum Information Theory:  Present Status and Future Directions

OutlineOutline

A Bit of History• QMA • Local Hamiltonians

Previous Constructions

The 3-qubit Gadget

Implications• Adiabatic computation• Other applications of the technique

Page 4: Quantum Information Theory:  Present Status and Future Directions

A Bit of (ancient) A Bit of (ancient) HistoryHistory

Complexity Theory:

• classify “easy” and “hard”

Page 5: Quantum Information Theory:  Present Status and Future Directions

A Bit of (ancient) A Bit of (ancient) HistoryHistory

NP – Nondeterministic Polynomial Time:

Def. L NP if there is a poly-time verifier V and a polynomial p s.t.

p( x )

p( x )

y {0,1} V(x,y)=1

y {0,1} V(x,y)=0

x L

x L

V“yes” instance: x L

witness: y1 (accept)

V“no” instance: x L

for all “witnesses” y0 (reject)

Page 6: Quantum Information Theory:  Present Status and Future Directions

Example: SATExample: SAT 1 1 2 3 3 4 5,..., ...nx x x x x x x x Formula:

SAT iff there is a satisfying assignment for x1,…,xn

(i.e. all clauses true simultaneously).

0 1 1

10 0 0

1

0 - false1 - true0 = 1, 1 = 0

V= (y)“yes” instance: SAT

witness: y=011000…1 (true, accept)

“no” instance: SAT

for all “witnesses” y=010110…

0 (false, reject)V= (y)

Page 7: Quantum Information Theory:  Present Status and Future Directions

NP completeNP completeA language is NP complete if it is in NP and as hard as any other problem in NP.

Cook-Levin Theorem: SAT is NP-complete

L SAT

y=011000…y1 1V

x0

SAT

y=010110…0

L

y

x NP NP-complete

Page 8: Quantum Information Theory:  Present Status and Future Directions

NP completeNP complete

Cook-Levin Theorem: 3SAT is NP-complete

1 2 3 3 4 5 ...x x x x x x

3SAT: 3 variables per clause

3 variables

2SAT is in P (there is a poly time algorithm).

MAX2SAT is NP-complete

MAX2SAT:Input: Formula with 2 variables per clause, number mOutput: 1 (accept) if there is an assignment that violates m clauses

0 (reject) all assignments violate >m clauses

Page 9: Quantum Information Theory:  Present Status and Future Directions

QMAQMA

V“yes” instance: x Lyes 1 (accept)

V“no” instance: x Lno

witness: |

for all “witnesses” | 0 (reject)

prob 1-

0 (reject) prob

1 (accept) prob

prob 1-

QMA – Quantum Merlin Artur = BQNP = “Quantum NP”

Def. L QMA if there is a poly-time quantum verifier V and

a polynomial p s.t.

p( x )

p( x )

prob V x, =1 1

prob V x, =1

x L

x L

2

2

C

C

Page 10: Quantum Information Theory:  Present Status and Future Directions

More recent (quantum) HistoryMore recent (quantum) HistoryQMA – Quantum Merlin Artur = BQNP

Def. L QMA if there is a poly-time quantum verifier V and

a polynomial p s.th.

p( x )

p( x )

prob V x, =1 1

prob V x, =1

x L

x L

2

2

C

C

•First studied in [Knill’96] and [Kitaev’99] – called it BQNP• “QMA” coined by [Watrous’00] – also: group-nonmembership QMA

Kitaev’s quantum Cook-Levin Theorem (’99): Local Hamiltonian is QMA-complete.

Page 11: Quantum Information Theory:  Present Status and Future Directions

Local HamiltoniansLocal Hamiltonians

Def. k-local Hamiltonian problem:

Input: k-local Hamiltonian , , Hi acts on k qubits, a<b constantsPromise:

• smallest eigenvalue of H either a or b (b-a const.)Output:

• 1 if H has eigenvalue a• 0 if all eigenvalues of H b

( )

1

poly n

ii

H H

iH ( )poly n

Page 12: Quantum Information Theory:  Present Status and Future Directions

Local HamiltoniansLocal Hamiltonians

1 2 3 3 4 5 ...x x x x x x

Intuition:

Formula:

Penalties for: x1x2x3 = 010 x3x4x5 = 100 …

Satisfying assignment is groundstate of

ii

H HEnergy-penalty 1 for each unsatisfied constraint.

x1x2 … xn| H |x1x2 … xn = #unsatisfied constraints

Hamiltonians: 1,2,3010 010

3,4,5100 100,

H1 H2 local Hamiltonians

Page 13: Quantum Information Theory:  Present Status and Future Directions

NP and QMANP and QMANP-completeness: QMA-completeness?

x1

x2

…y1

y2

…00…

1x

y

0

Verifier V:

input

witness

ancilla

Page 14: Quantum Information Theory:  Present Status and Future Directions

NP and QMANP and QMANP-completeness: QMA-completeness?

y1

y2

00…

1

y

0

Verifier Vx :

3-clauses check:

• propagation

• output

z01

z02

z03

z04

z0N z1N z2NzTN

t = 0 1 2 3 4 … T

|

|0 |0…

C

H

|1

C H

|0|1 … |T

?

Verifier Ux :

ancilla qubits

witness

ancilla

• input

ancilla

No local way to check!

Page 15: Quantum Information Theory:  Present Status and Future Directions

NP and QMANP and QMANP-completeness: QMA-completeness?

y1

y2

00…

1

y

0

Verifier Vx :

3-clauses check:

• propagation

• output

|

|0 |0…

C

H

|1

C H

?

Verifier Ux :

ancilla qubits

witness

ancilla

• input

||0=|0 |1 … |T

+ + ++

|0 |1 |2 |T

| |0|0+|1|1+…+ |T|T

witness = sum over history

Page 16: Quantum Information Theory:  Present Status and Future Directions

NP-completeness: QMA-completeness:

• 3SAT is NP-complete

• 2SAT is in P

• log|x|-local Hamiltonian is QMA-compl. [Kitaev’99]• 5-local Hamiltonian is QMA-compl. [Kitaev’99]

• 3-local Hamiltonian is QMA-compl. [KempeRegev’02]

• but: MAX2SAT is NP-complete • 2-local Hamiltonian is NP-hard

2-local Hamiltonian????

• 1-local Hamiltonian is in P

More recent (quantum) HistoryMore recent (quantum) History

Is 2-local Hamiltonian QMA-complete??

Page 17: Quantum Information Theory:  Present Status and Future Directions

OutlineOutline

A Bit of History• QMA • Local Hamiltonians

Previous Constructions

The 3-qubit Gadget

Implications• Adiabatic computation• Other applications of the technique

Page 18: Quantum Information Theory:  Present Status and Future Directions

Kitaev’s log-local ConstructionKitaev’s log-local Construction

Local Hamiltonians check: H= Jin Hin + Jprop Hprop + Hout

| |1

Verifier Ux :

witness = sum over historym

N-m

TT=poly(N)

• input

• propagation

• output

1

1 1 0 0N

in ii m

H

1

1-1 -1 -1 -1

2

T

prop t tt

H I t t I t t U t t U t t

11 0 0outH T T T

Computation qubits

Time register {|0, |1,…, |T}

Page 19: Quantum Information Theory:  Present Status and Future Directions

Kitaev’s log-local ConstructionKitaev’s log-local ConstructionH= Jin Hin + Jprop Hprop + HoutVerifier: Ux=UTUT-1…U1

To show: If Ux accepts with prob. 1- on input |,0, then H has eigenvalue . If Ux accepts with prob. on all |,0, then all eigenvalues of H ½-.

Page 20: Quantum Information Theory:  Present Status and Future Directions

Completeness Completeness H= Jin Hin + Jprop Hprop + HoutVerifier: Ux=UTUT-1…U1

To show: If Ux accepts with prob. 1- on input |,0, then H has eigenvalue . If Ux accepts with prob. on all |,0, then all eigenvalues of H ½-.

1

1 1 0 0N

in ii m

H

†,

1 1

1-1 -1 -1 -1

2

T T

prop t t prop tt t

H I t t I t t U t t U t t H

11 0 0outH T T T

|Hin| =0

†, 1 -1 1 1 -1 1... ... -1 ... ... -1 0prop t t t t t t tH U U t U U t U U t U U U U t

|Hprop| =0

|Hout| 0 11 1

10 1 1 ...

1T TT T

T

Page 21: Quantum Information Theory:  Present Status and Future Directions

5-local Hamiltonians5-local HamiltoniansLog-local terms: , -1 , -1t t t t t t

Idea (Kitaev): unary |t | 11…100…0 t T-t

|tt| |1010|t,t+1

|tt-1| |110100|t-1,t,t+1

Penalise illegal time states: ,01 01clock i j

i j

H I

clock - space of legal time-states is preserved (invariant)

Page 22: Quantum Information Theory:  Present Status and Future Directions

3-local Hamiltonians3-local Hamiltonians5-local terms: |tt-1| |110100|t-1,t,t+1 ,

01 01clock i ji j

H I

Idea [KR’02]: |110100|t-1,t,t+1 |10|t

clock clock KitaevH J H H

(|10|t)|clock = |tt-1|

Give a high energy penalty to illegal time statesto effectively prevent transitions outside clock :

clock

Page 23: Quantum Information Theory:  Present Status and Future Directions

OutlineOutline

A Bit of History• QMA • Local Hamiltonians

Previous Constructions

The 3-qubit Gadget

Implications• Adiabatic computation• Other applications of the technique

Page 24: Quantum Information Theory:  Present Status and Future Directions

Idea: use perturbation theory to obtain effective 3-local Hamiltonians from 2-local ones by restricting

to subspaces

H’ = H + V

Spectrum: H…

0 groundspace S

Energy gap: ||H||>>||V||

What is the effective Hamiltonian in the lower part of the spectrum?

Three-qubit gadgetThree-qubit gadget

Page 25: Quantum Information Theory:  Present Status and Future Directions

Perturbation TheoryPerturbation Theory

H’ = H + V

Spectrum: H

0 groundspace S

Energy gap: ||H||>>||V||

S

Case 1: Energy gap >>> ||V|| V V

VV V

S

S

V-- - restriction of V to S

V++ - restriction of V to S

What is the effective Hamiltonian in the lower part of the spectrum?

Projection Lemma: Heff = V-- (same spectrum) =O(||V||2/)

Page 26: Quantum Information Theory:  Present Status and Future Directions

Perturbation TheoryPerturbation Theory

H’ = H + V

Spectrum: H

0 groundspace S

Energy gap: ||H||>>||V||

Theorem:

2 3 1

1 1 1...

n

eff n

VH V V V V V V V V V V O

S

What is the effective Hamiltonian in the lower part of the spectrum?

Case 2: Fine tune the energy gap > ||V|| V V

VV V

S

S

V-- - restriction of V to S

V++ - restriction of V to S

Page 27: Quantum Information Theory:  Present Status and Future Directions

Perturbation TheoryPerturbation Theory

H

0 groundspace S

Energy gap:

Theorem:

2 3 1

1 1 1...

n

eff n

VH V V V V V V V V V V O

S

First orderSecond order

Third order

The lower spectrum of H’ is close to the spectrum of Heff (under certain conditions).

H’ = H + V

Page 28: Quantum Information Theory:  Present Status and Future Directions

Perturbation TheoryPerturbation Theory

H

0 groundspace S

Energy gap:

Theorem: 21

...eff

VH V V V V O

S

First order: ||V||2 <<

The lower eigenvalues (<||V||) of H’ are close to the eigenvalues of Heff (under certain conditions).

Projection Lemma

H’ = H + V

Page 29: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

H=P1P2P3

3-local

1

32

1

32

B

A

C

ZZ

ZZ

ZZ

P1XA

P2XB P3XC

Terms in H’ are 2-local

Heff=P1P2P3

3-local

Page 30: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadgetH’ = H + V

Energy gap:

S={|000, |111}

S={|001,|010,|100, |110,|101,|011}

0

=-3B

A

C

ZZ

ZZ

ZZ

3

34 A B B C A CH Z Z Z Z Z Z I

Page 31: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

C

H’ = H + V

Energy gap:

S={|000, |111}

S={|001,|010,|100, |110,|101,|011}

0

=-3

2 P2XB3P3XC

1P1XA

Theorem: 4 32

1 1effH V V V V V V O V

Second order: S S

SV-+ V+-

Third order: S S

S V-+ V+-

V++S

First order: S SV--

V VV

V V

Page 32: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

C

Energy gap:

S={|000, |111}

S={|001,|010,|100, |110,|101,|011}

0

=-3

2 P2XB3P3XC

1P1XA

Theorem:

Second order: S S

SV-+ V+-Ex.: P1XA P1XA

|000

|100

|000

1 000 100 ...V P 2

1 000 000 ...V V P

4 32

1 1effH V V V V V V O V

Page 33: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

C

H’ = H + V

Energy gap:

S={|000, |111}

S={|001,|010,|100, |110,|101,|011}

0

=-3

2 P2XB3P3XC

1P1XA

Theorem:

Third order: S S

S V-+ V+-

V++S

Ex.: P1XA P3XC

|000

|100 |110

|111

P2XB

1 2 3 000 111 ...V V V PP P

4 32

1 1effH V V V V V V O V

Page 34: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

CH’ = H + V

2 P2XB3P3XC

1P1XA

1 2 2 21 2 3P P P 2

1 2 3A B CV PX P X P X

0V

1 12

2

000 100 111 011

000 010 ...

P PV

P

21 2010 110 100 110 ...V P P

Theorem:

4 32

1 1 effH V V V V V V O V

3

34 A B B C A CH Z Z Z Z Z Z I

4

2 2 21 2 33

0 SP P P I

6

1 2 36

3 000 111 111 000PP P O

1 2 2 21 2 3 SP P P I

Page 35: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

CH’ = H + V

2 P2XB3P3XC

1P1XA

1 2 2 21 2 3P P P 2

1 2 3A B CV PX P X P X

V VV

V V

0V

1 12

2

000 100 111 011

000 010 ...

P PV

P

21 2010 110 100 110 ...V P P

Theorem:

3

34 A B B C A CH Z Z Z Z Z Z I

1 2 2 21 2 3 SP P P I

4 32

1 1 effH V V V V V V O V

6

1 2 36

3 000 111 111 000PP P O

Page 36: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

CH’ = H + V

2 P2XB3P3XC

1P1XA

1 2 2 21 2 3P P P 2

1 2 3A B CV PX P X P X

V VV

V V

0V

1 12

2

000 100 111 011

000 010 ...

P PV

P

21 2010 110 100 110 ...V P P

Theorem:

4 32

1 1 effH V V V V V V O V

3

34 A B B C A CH Z Z Z Z Z Z I

1 2 33 SPP P X O

1 2 2 21 2 3 SP P P I

effH

Page 37: Quantum Information Theory:  Present Status and Future Directions

Three-qubit gadgetThree-qubit gadget

B

A

CH’ = H + V

2 P2XB3P3XC

1P1XA

1 2 2 21 2 3P P P 2

1 2 3A B CV PX P X P X

Theorem:

4 32

1 1 effH V V V V V V O V

3

34 A B B C A CH Z Z Z Z Z Z I

1 2 33 SPP P X O effH

=-3

0

H

0

-1V

Heff

const.

Page 38: Quantum Information Theory:  Present Status and Future Directions

2-local Hamiltonian is QMA-complete2-local Hamiltonian is QMA-complete

• start with the QMA-complete 3-local Hamiltonian

• replace each 3-local term by 3-qubit gadget

Page 39: Quantum Information Theory:  Present Status and Future Directions

OutlineOutline

A Bit of History• QMA • Local Hamiltonians

Previous Constructions

The 3-qubit Gadget

Implications• Adiabatic computation• Other applications of the technique

Page 40: Quantum Information Theory:  Present Status and Future Directions

Implications for Adiabatic ComputationImplications for Adiabatic ComputationAdiabatic computation [Farhi et al.’00]:

• “track” the groundstate of a slowly varying Hamiltonian

Standard quantum circuit:

|0…0 |T

T gates

*D. Aharonov, W.  van Dam, J. Kempe, Z. Landau, S. Lloyd, O. Regev: "Adiabatic Quantum Computation is Equivalent to Standard Quantum Computation", lanl-report quant-ph/0405098

Adiabatic simulation*:

Hinitial

•groundstate |0…0 |0

Hfinal

•groundstate

H(t) = (1-t/T’)Hinitial +t/T’ Hfinal

T’=poly(T):

If gap 0(H(t))-1(H(t)) between groundstate and first excited state is 1/poly(T)

Page 41: Quantum Information Theory:  Present Status and Future Directions

Implications for Adiabatic ComputationImplications for Adiabatic Computation

2-local adiabatic computation is equivalent to standard quantum computation

Our result also implies:

*D. Aharonov, W.  van Dam, J. Kempe, Z. Landau, S. Lloyd, O. Regev: "Adiabatic Quantum Computation is Equivalent to Standard Quantum Computation", lanl-report quant-ph/0405098

|0…0 |0 adiabat

H(t) = (1-t/T’) Hin + t/T’ HpropLog-local*:

Replace with 2-local: H(t) = (1-t/T’)(Hin+Jclock Hclock) + t/T’(Hprop

gadget+Jclock Hclock)

Page 42: Quantum Information Theory:  Present Status and Future Directions

Other applications of the Other applications of the gadgetgadget

(work in progress)(work in progress)“Interaction at a distance”:

H=P1P2Heff=P1P2

-1P1XA -1P2XA

-2ZA

“Proxy Interaction”: (with A. Landahl)

H=Z1X2

only XX,YY,ZZ availableHeff=Z1X2

-2YAYB-1Z1ZA -1X2XB

Useful for Hamiltonian-based quantum architectures

Page 43: Quantum Information Theory:  Present Status and Future Directions

ReferencesReferences

Quantum Complexity :J. Kempe, A. Kitaev, O. Regev: “The Complexity of the local

Hamiltonian Problem”, quant-ph/0406180, to appear in Proc. FSTTCS’04

J. Kempe and O. Regev: "3-Local Hamiltonian is QMA-complete", Quantum Information and Computation, Vol. 3 (3), p.258-64 (2003), lanl-report quant-ph/0302079

Adiabatic Computation :D. Aharonov, W.  van Dam, J. Kempe, Z. Landau, S. Lloyd, O.

Regev: "Adiabatic Quantum Computationis Equivalent to Standard Quantum Computation", lanl-report quant-ph/0405098, to appear in FOCS’04

*Photo: Oded Regev: Ladybug reading “3-local Hamiltonian” paper

Page 44: Quantum Information Theory:  Present Status and Future Directions

MAMA

V“yes” instance: x Lyes

witness: y1 (accept)

V“no” instance: x Lno

for all “witnesses” y

0 (reject)

MA – Merlin-Artur:Def. L MA if there is a poly-time verifier V and a polynomial p s.t.

p( x )

p( x )

y {0,1} prob V(x,y)=1 1

y {0,1} prob V(x,y)=1

yes

no

x L

x L

0 (reject)

prob 1-prob

1 (accept) prob

prob 1-