Encoded Universality and Decoherence-Free Subspaces Control Seminar – Quantum Information Science and Technology UC Berkeley, Feb. 9, 2004 Julia Kempe CS Division and Dept. of Chemistry, University of California, Berkeley LRI, Universite de Paris-Sud, Paris, France
Towards nanotechnology Size of the components Number of components Speed Gordon Moore 1965 prevent or use quantum effects ? Theoretical limitations reached in 2020 !!! Apparition of quantum phenomena
Information is physical! Use the laws of quantum mechanics for the basic components of an information processing machine! Quantum computing Quantum cryptography Quantum information …
Main applications zCryptography yProtocol of unconditionally secure secret key distribution [Bennett, Brassard 84] Implementation : ~ 100 km zQuantum information yTeleportation [B, B, Crépeau, Jozsa, Peres, Wooters 93] Implementation [Bouwmeester, Pan, Mattle, Eibl, Weinfurter, Zeilinger 97] zAlgorithms yFactoring, discrete logarithm,... [Shor 94] yDatabase search [Grover 96] Num. of qubits ? 1995 : 2, 1998 : 3, 2002 : 8 [Chuang (IBM)] - 10 [Los Alamos]
Overview zBasic notions of quantum computing zStandard solutions: yUniversal gate set yQuantum Error Correcting Codes (QECC) zEncoded Universality zDecoherence-free Subsystems
The qubit Classical bit: b {0,1} Probabilistic bit: probability distribution d R + {0,1} such that || d || 1 =1. d=(p,1-p) with p [0,1] Quantum bit: | C {0,1} such that || | || 2 =1. | = |0 + |1 with | | 2 + | | 2 = 1 (Dirac notation)
Qubit evolution zMeasure: reads and modifies Measure | | 2| | 2 | | 2| | 2 |0 + |1 |0|0 |1|1 Superposition Probability distribution Unitary transformation: U C 2 2 such that UU † =Id | U | ’ = U | unitary reversible: U| U†U† |
Example Superposition: Measure: Measure 1/3 2/3 |0|0 |1|1 |
Example Superposition: Measure: Unitary transformations: yNOT: |0 |1 yHadamard: Measure 1/3 2/3 |0|0 |1|1 | U | ’ = U | H
Quantum computer: n qubits zn qubits tensor product | C {0,1} n such that || | || 2 =1. | = x {0,1} n x |x with x | x | 2 = 1 zMeasure zPartial Measure Measure |x | 2|x | 2 x {0,1} n x |x |x|x Measure Second bit = 0 (| | 2 + | | 2 ) |00 + |01 + |10 + |11
Quantum computer: n qubits n qubits tensor product | C {0,1} n such that || | || 2 =1. | = x {0,1} n x |x with x | x | 2 = 1 zMeasure zPartial Measure zUnitary transformation | U| with U U(2 n ) ex: XOR= Measure |x | 2|x | 2 x {0,1} n x |x |x|x Measure Second bit = 0 (| | 2 + | | 2 ) |00 + |01 + |10 + |11 |00 |01 |10 |11 + |i |j |i |XOR(i,j)
Quantum computing a function Let f: {0,1} n {0,1} m x f(x) Reversible: R f :{0,1} n+m {0,1} n+m (x,y) (x,y f(x)) Quantum: U f U(2 n+m ): C n+m C n+m |x |y |x |y f(x)
Simplest Quantum Algorithm: Deutsch’s Problem Input: function f:{0,1} {0,1} (in black box) Question: f constant (f(0)=f(1)) or balanced (f(0) f(1)) ? Quantum black box (reversible): Algorithm: one query only!!! f |x|x |y|y |x|x |y f(x) f H H H |0|0 |1|1 Measure |0 -constant |1 -balanced
Simplest Quantum Algorithm: Deutsch’s Problem Input: function f:{0,1} {0,1} (in black box) Question: f constant (f(0)=f(1)) or balanced (f(0) f(1)) ? Quantum black box (reversible): Algorithm: one query only!!! f |x|x |y|y |x|x |y f(x) f H H H |0|0 |1|1 Measure |0 -constant |1 -balanced =0 if f balanced =0 if f constant
Where are we in practice? Several physical architectures have been proposed: zNMR, solid state, ion traps, superconducting qubits, optical cavities, photons zQubit = nuclear spin of atoms in a molecule (NMR), nuclear spin of (doped) atoms in a silicon donor (solid state), vibrational degrees of freedom of ions (ion trap), flux-degree of freedom (superconductors), polarisation (photons)
Five requirements for the implementation of quantum computation* 1.A scalable physical system with well characterized qubits 2.The ability to initialise the qubits to a simple state, such as |00…0 3.A “universal” set of quantum gates 4.Long relevant decoherence times, much longer than the gate operation time 5.A qubit-specific measurement capability * D. DiVincenzo, 97
Overview zBasic notions of quantum computing zStandard solutions: yUniversal gate set yQuantum Error Correcting Codes (QECC) zEncoded Universality zDecoherence-free Subsystems
Five requirements for the implementation of quantum computation* 1.A scalable physical system with well characterized qubits 2.The ability to initialise the qubits to a simple state, such as |00…0 3.A “universal” set of quantum gates 4.Long relevant decoherence times, much longer than the gate operation time 5.A qubit-specific measurement capability * D. DiVincenzo, 97
Universal computation Classical circuit model: Quantum circuit model: evaluates boolean functions can be constructed from universal local gates (ex.: NAND, COPY) 010…1010…1 bits 0000 unitary transformations U qubits |0 |1 |0 Measure U
Quantum circuits + + =U Barenco et al. ’95: Single-qubit gates and CNOT generate every unitary transformation!
Five requirements for the implementation of quantum computation* 1.A scalable physical system with well characterized qubits 2.The ability to initialise the qubits to a simple state, such as |00…0 3.A “universal” set of quantum gates 4.Long relevant decoherence times, much longer than the gate operation time 5.A qubit-specific measurement capability * D. DiVincenzo, 97
Main obstacle – decoherence! zQuantum information is very fragile zAny interaction with the environment disturbs the stored information zSolutions: quantum error correcting codes, decoherence-free subspaces, fault- tolerant computation …
Active noise protection: quantum error correction Quantum information is very fragile -how maintain quantum coherence? Quantum Error-Correcting Codes (QECCS)! Easiest classical code: repetition code: 0 111 Can correct one bitflip error (majority) 100,010,001 ,101,110 111
Active noise protection: quantum error correction Quantum information is very fragile -how maintain quantum coherence? Quantum Error-Correcting Codes (QECCS)! Easiest classical code: repetition code: 0 111 Can correct one bitflip error (majority) 100,010,001 ,101,110 111 Quantum: Quantum: many possible errors (bit flip, phase error, measurements…) “Quantum” repetition code? Impossible – no cloning principle.
Active noise protection: quantum error correction No cloning principle: No cloning principle: It is impossible to copy a quantum state.Proof: if then and By linearity:
Active noise protection: quantum error correction 3-qubit QECC: protects against phase errors: 3-qubit QECC: protects against bitflip errors: 9-qubit QECC: protection against all one qubit errors (Shor-code):
Overview zBasic notions of quantum computing zStandard solutions: yUniversal gate set yQuantum Error Correcting Codes (QECC) zEncoded Universality zDecoherence-free Subsystems
Quantum circuits + + =U Barenco et al. ’95: Single-qubit gates and CNOT generate every unitary transformation!
Hamiltonians=Interactions + + =U Unitaries are generated by Hamiltonians:
Hamiltonians=Interactions + + =U Single-qubit gates and CNOT generate every unitary transformation! Unitaries are generated by Hamiltonians: Hamiltonians describe the interactions (of qubits) in physical systems. Quantum engineers tweak the Hamiltonians to produce single qubit gates and CNOT.
Universality: The problem “Easy” and “hard” interactions (system-dependent) “Easy”: intrinsic interactions “natural” to the system, easy to tune, rapid “Hard”: slower, require higher device complexity, high decoherence System“Easy”“Hard” Photon- qubits Single qubit operations (linear optics) Two qubit operations (non-linearity, non-deterministic qc) Solid state Two qubit operations (ex. J-gates) Single qubit operations (ex. local focused B-fields) “Coherent- state” qubits Two qubit operations (beam-splitter) Single qubit operations (non-linearity, non-det.) Can we avoid “hard” interactions?
Almost every interaction is universal! Deutsch et al.(’95), Lloyd (‘95) : Almost any interaction on two qubits is universal. In the generic sense. Does not include the most frequent interactions. Nature is not generic! H ij H ji qubit iqubit jqubit iqubit j
Change of paradigm Traditionally: manipulate the physical system* to produce + + H 1,H 2,... * Independent of system’s natural talents (fast, robust interactions) often difficult, certain gates can only be implemented with noise; high decoherence...
Change of paradigm Traditionally: manipulate the physical system to produce + + Universal encoded computation Universal encoded computation: interactions given by the physical system find a way to make them universal HHH Encoding? H 1,H 2,...
Classical « Analogy » Two coins can only flip the two coins together « encode » « 0 »- « 1 » flip Encoded « coin »
Language of Hamiltonians Which interactions are universal? Given = H 1, H 2,…, H n can one generate any unitary transformation (exactly or approximatively)? U(t) = exp(iHt)
Language of Hamiltonians Which interactions are universal? Given = H 1, H 2,…, H n can one generate any unitary transformation (exactly or approximatively)? Possible compositions: scalar multiple 1) scalar multiple 2) linear combination 3) Lie bracket H has to generate Lie algebra su(N) of the unitary group SU(N)! U(t) = exp(iHt)
Lie Algebra of H Lie(H) closed under: 1) scalar multiplication 2) linear combination 3) Lie-bracket H has to generate Lie algebra su(N) of the unitary group SU(N)!
EX: Heisenberg interaction zomnipresent in solid state physics (« Easy ») zis not universal: had to be supplemented with single qubit gates On three qubits: E 12 E 13 E 23 (Pauli matrices)
The algebra L 3 (E) of E (3 qubits) the algebra L 3 (E) splits into irreducible components as: L 3 (E) L 3 (E) S 1 I 4 S 2 I 2 su(2) S 2 Encoded qubit ? su(2) 2 2 Lie algebra of E :
The algebra L 3 (E) of E (3 qubits) the algebra L 3 (E) splits into irreducible components as: L 3 (E) L 3 (E) S 1 I 4 S 2 I 2 su(2) S 2 Encoded qubit ? su(2) 2 2 Simulation of all operations of one qubit (su(2)) with L 3 (E) on the encoded qubit ! Lie algebra of E :
The algebra L n (E) of E (n qubits) the algebra L n (E) splits into irreducible components as: L n (E) ...
The algebra L n (E) of E (n qubits) the algebra L n (E) splits into irreducible components as: L n (E) ... Commutant L’ of L n (E) : L’ is generated by (« spin » algebra su(2)) As a Lie algebra L’ splits into irreducible representations of su(2).
Useful theorem... Universal computation “for free”? Let L be a †-closed algebra closed under multiplication and linear combination. Then the underlying space H is isomorphic to and L and its commutant L’ split as: where M(C d ) (M(C n )) is the algebra of all matrices on C d (C n ).
Useful theorem Let L be a †-closed algebra closed under multiplication and linear combination. Then the underlying space H is isomorphic to and L and its commutant L’ split as: where M(C d ) (M(C n )) is the algebra of all matrices on C d (C n ). The multiplicative algebra is not at our disposition! However the Lie algebra splits into irreducible components in the same basis: NO!
Problem of “Encoded Universality” Given an ensemble of generators H with Lie algebra Lie(H) which splits as can one find a component s.t. contains su(n j )? Encode the quantum information into the corresponding sub-space. dimension: n j... Yes
EX: The algebra L n (E) of E (n qubits) the algebra L n (E) splits into irreducible components as: L n (E) ... Results: Results: E is universal with encoding* introduce tensor structure, ex. blocks with 3 qubits** *J.K., D.Bacon, D.A.Lidar, K.B.Whaley, Phys. Rev. A 63: (2001) **D.DiVincenzo, D.Bacon, J.K., K.B.Whaley, NATURE 408 (2000) 19 operations for CNOT, 4 operations for 1-qubit
Ex: Heisenberg interaction On three qubits: E 12 E 13 E 23 (Pauli matrices) su(2) 2 2 Proof idea: su(2) su(2)=Lie( x, y, z ) [ x, y ]= i z [ y, z ]= i x [ z, x ]= i y [ i, j ]= i ijk k
Overview zBasic notions of quantum computing zStandard solutions: yUniversal gate set yQuantum Error Correcting Codes (QECC) zEncoded Universality zDecoherence-free Subsystems
Decoherence-free subsystems (DFS) Decoherence-free subsystems (DFS) (“Dual” approach to encoded universality) zinteraction described by system-environment Hamiltonian: System (quantum computer) Environment interaction causes noise Decoherence system environmt.
DFS z Start initially decoupled from environment z Coupling terms in H I perturb the unitary evolution and evolve system into a mixed state system environmt.
DFS z Start initially decoupled from environment z Coupling terms in H I perturb the unitary evolution and evolve system into a mixed state z Solutions: Active (error correction) - QECC: deal well with independent errors on qubits Passive (error avoidance) – DFS: find a subspace of the system space over which evolution stays unitary, unperturbed, correlated noise noise-free subspace, exploits symmetries in the noise system environmt.
Classical « Analogy » (again!) Two coins Adversary (noise) can only flip the right coin encode a « noise-free » coin « 0 »- « 1 » flip Adversary (noise) encoded 2 states of coin Manipulate encoded coin with global flips!
More formal the S (and the algebra generated by them) split as: If we only keep the states within one component then we can encode into and the coupling H I will not affect the encoded states!... S S encode into here system environmt.
eXAMPLE: Collective Decoherence splits into irreducible components as:... (« spin » algebra su(2))
eXAMPLE: Collective Decoherence splits into irreducible components as:... (« spin » algebra su(2)) DF- Subspaces 1-dim Start decoupled: unaffectd by H I Effective encoding: (=0)
eXAMPLE: Collective Decoherence splits into irreducible components as:... (« spin » algebra su(2)) DF- Subsystem* 2j+1-dim Each C 2j+1 unaffectd by H I *Knill, Laflamme, Viola, PRL’01
3 qubits 3 qubits splits as: S S 4 I 1 S 2 I 2 2 H I doesn’t mix the two spaces! Commutant of the S is Lie(E)!
Problem of DFS Given an interaction Hamiltonian where the S split as find a component such that n j is large enough to encode quantum states ( “=“ DFS).
Problem of DFS Given an interaction Hamiltonian where the S split as find a component such that n j is large enough to encode quantum states ( “=“ DFS). Moreover, in order to universally compute over the DFS, we need the commutant ! Computing with protected states DF-state has to stay within DF-space during the entire operation gate Hamiltonian has to commute with all the S
Computing on DFS Computing with protected states DF-state has to stay within DF-space during the entire operation gate Hamiltonian has to commute with all the S ... S S
Computing on DFS Computing with protected states DF-state has to stay within DF-space during the entire operation gate Hamiltonian has to commute with all the S ... S S DF-Subspace:
Same as “Encoded Universality”! Given an ensemble of generators H with Lie algebra L(H) which splits as can one find a component s.t. contains su(n j )? Encode the quantum information into the corresponding sub-space. dimension: n j... Yes
Summary-Outlook DFS and Encoded Universality related concepts – representation theory of Lie algebras DFS: so far solved only for specific noise models – collective decoherence Encoded Universality: encodings so far for variants of the exchange interaction OPEN: other interactions/noise models hybrid DFS-QECC (so far concatentation)
References Encoded Universality: J. Kempe, D. Bacon, D.P. DiVincenzo, K.B. Whaley, “ Encoded universality from a single physical interaction”, in «Quantum Information and Computation»; Special Issue, Vol. 1, 2001, quant-ph/ D. Bacon, J. Kempe, D.P. DiVincenzo, D.A. Lidar, K.B. Whaley, “ Encoded Universality in Physical Implementations of a Quantum Computer”, Proceedings of IQC ’01, Australia, quant-ph/ D.P. DiVincenzo, D. Bacon, J. Kempe, K.B. Whaley, “ Universal Quantum Computation with the Exchange Interaction”, NATURE 408, 339 (2000), quant-ph/ J. Vala and K.B. Whaley, “Encoded Universality with Generalized Anisotropic Exchange Interactions”, Earlier related work on DFSs: D. Bacon, D. Lidar, K.B. Whaley, Phys. Rev. Lett. (2000) J. Kempe, D. Bacon, D. Lidar, K.B. Whaley, Phys. Rev. A 63: (2001) D. Bacon, J. Kempe, D. Lidar, K.B. Whaley, Phys. Rev. Lett. (2000)