From Complexity to Universality in Quantum Computation Kae Nemoto National Institute of Informatics, Tokyo How Complex is Quantum Computation? Can a quantum system be probabilistically simulated by a classical universal computer? … the answer is certainly, No! — Richard P. Feynman (1982) N particles of d-level ϕ1 ⊗Λ ⊗ ϕ N U ⎯ ⎯→ d ∑Λ j1 =1 d 1 N c ϕ ⊗ Λ ⊗ ϕ ∑ j1Κ jN j N =1 The number of coefficient is in the order of d N Hilbert space is a big place. — Carlton Caves Does the complexity of quantum computation come from Hilbert space? What makes quantum computation distinct from classical computation? 2 Complexity • How can we examine complexity of a quantum system? The initial state Hilbert space Hilbert space Starting from the initial state at time t, we observe where the state goes. Can we say that the right case is more complicated than the left one? 3 Separable and entangled • Consider only pure states, – If the time evolution does not generate any entanglement, then the state description is rather simple. ϕ ⊗Λ ⊗ ϕ 1 SU ( d ) ⊗Λ ⊗ SU ( d ) ⎯⎯ ⎯ ⎯ ⎯⎯→ N d d N c ϕ ⊗ Λ ⊗ c ϕ ∑ j1 ∑ jN 1 j1 =1 j N =1 The number of coefficient is in the order of dN. – If it is possible to generate entanglement, the state description must stay in the complicated form: ϕ ⊗Λ ⊗ ϕ 1 N SU ( d N ) ⎯⎯ ⎯ ⎯→ d ∑Λ j1 =1 d N 1 c ϕ ⊗ Λ ⊗ ϕ ∑ j1Κ jN j N =1 The number of coefficient is in the order of d N. 4 Restricted operations might allow us a simple description • • • • Restricting the allowed operations to local unitary operators, we can have a simple description of states. The previous example suggests that a system without entanglement can be represented simply. Is there any other examples which allows us to have a simple description of states? Is it possible to include entangled states in such a simple description? The schematic image of a simple state description case A case which apparently requires a complex description The initial state Hilbert space Hilbert space Any way to retrieve the simple description? If we can do this systematically, then we can evaluate how complex our system is.5 Preparation for the stabilizer formalism State representation Operator representation Qubit case: Pauli Group Basis states 0, Pauli matrices: 1 Find a biggest subgroup which does not change the state (including the global phase). 0 : Z0 = 0 (Z 1 ⎛0 X = ⎜⎜ ⎝1 ⎛0 Y = ⎜⎜ ⎝i 1⎞ ⎟ 0 ⎟⎠ −i⎞ ⎟ 0 ⎟⎠ ⎛1 0 ⎞ ⎟⎟ Z = ⎜⎜ 0 1 − ⎝ ⎠ Unity matrix: ⎛1 0⎞ ⎟⎟ I = ⎜⎜ 0 1 ⎝ ⎠ ⊗ {± 1, i} = −1 ) Ex. For a given state |0> the subgroup ( I, Z ) does not change the state. 0 ↔ (I , Z ) ↔ Z The subgroup can labeled by the set of generators. 6 Stabilizer Formalism • • • • The stabilizer formalism can be a powerful tool to describe quantum states using group theory. The group of principal interest is the Pauli group. Start with 0 in the computational basis. – A subgroup S = {I , Z } stabilizes the initial state 0 . – We specify the subgroup by the generator Z . – We can describe a state via its stabilizer. Generalized it to multi-qubit systems: A state: • 0 ⊗Λ ⊗ 0 The stabilizer generators: Z1 , Z 2 , Κ , Z n Now, we use the stabilizer formalism to analyze dynamics of systems. 7 Unitary gates and the stabilizer formalism • Unitary gates can be considered as a map from one stabilizer to another. – U is a unitary gate and the initial state is 0 . Schrödinger picture State description Stabilizer description 0 Z U U U0 UZU + Heisenberg picture – We may compute how the unitary gate affects the generators of the stabilizer. – The gate U transforms the state <Z > to a linear sum of Pauli group elements. 8 Special classes of unitary gates • In general, Z U UZU + UZU + = α X + β Y + γ Z , α , β , γ ∈ C The gate operation transforms the generator Z to an operator outside of Pauli group. • • • In general, the stabilizer description is typically as complex as the state representation. The final state typically requires 2n amplitudes to be specified. How can we seek simplicity in quantum computational circuits? – If we can restrict gate operations to ones which preserve the Pauli group, then the stabilizer description remains simple. Z U UZU + U : Hadamard, then UZU + = X 9 Complexity of a gate set • • Using the stabilizer formalism, is there any interesting sets of restricted gates? Find a set of gates which preserve the Pauli group. The initial state can go only some part of Hilbert space. Hilbert space 1 … 0 … n Cf. Local unitary operations cannot transform a separable state to a entangled state. → the initial state has to stay In some part of the Hilbert space. 10 Restrict Dynamics • Seek the class of operations such that the complexity of the stabilizer description remains linear in n. Gate Operations which preserve the Pauli group. H X Z H= → → S Z X X Z 1 ⎛1 1 ⎞ ⎜⎜ ⎟⎟ 2 ⎝ 1 − 1⎠ → → Y Z 1 S = Z = ⎛⎜ ⎝0 Pauli operators: → → 0⎞ i ⎟⎠ ⎛1 ⎜0 CNOT = ⎜ ⎜0 ⎝0 Y X X Z CNOT X1 → X1 X 2 X2 → X2 Z1 → Z1 Z 2 → Z1 Z 2 X −Z X Z → → 0 1 0 0 0 0 0 1 0⎞ 0⎟ ⎟ 1⎟ 0⎠ Z −X −Z X Z → → −X Z 11 One Qubit S,Z 0 H X 0 +1 0 +i1 Y Y S S 0 −i 1 X,Z,H Y,Z S S X X,Y 0 −1 H 1 S,Z 12 Example: the swap circuit The swap circuit + 1 )⊗ 0 Initial State 0 ⊗( 0 + 1 ) 00 + 11 CNOT (0 CNOT CNOT ⎧U CNOT ( X 1 ⊗ I 2 )U CNOT + CNOT operation: ⎨ + ( ) ⊗ U I Z U CNOT 2 ⎩ CNOT 1 X1 X1 X 2 Z2 Z1 Z 2 0 ⊗( 0 + 1 ) = X1 ⊗ X 2 = Z1 ⊗ Z 2 X2 Z1 ⎧ X 1 ⊗ X 2 ( 00 + 11 ) ⎨ ⎩ Z1 ⊗ Z 2 ( 00 + 11 ) Final State X2 Z1 = = 11 + 00 00 + 11 13 Gottesman-Knill (GK) Theorem • GK analyzed: the complexity of quantum computational circuits using stabilizers. GK theorem: A quantum circuit which starts with a state in the computational basis, and consists of only H Hadamard ⎫ S Phase ⎪⎪ CNOT Controlled-NOT ⎬⎪ X,Y,Z Pauli ⎪⎭ gates, and projective measurements in the computational basis may be efficiently simulated on a classical computer. • GK theorem specifies a class of quantum computations which can be efficiently simulated on a classical computer. 14 Qubit to Qunat Qubit • Most widely and deeply investigated computational model. 0,1 Group contraction to Qunats Easy to translate to each other Qudit • • 0 , 1 , Λ , d −1 Qunat (Continuous Variables) q and p Squeezed states Infinite squeezing d-level systems Properties of qudit systems are quite similar to qubit systems in many ways. Squeeze in one direction. Squeezing Hamiltonian: ( Hˆ = χ a + 2 + a 2 ) Gaussian states 15 Qunat quantum computation • Generalize the Pauli group for qunat quantum computation and seek the gate operations which preserve the generalized Pauli group. – The generalized Pauli group: { } The generators qˆ , pˆ , Iˆ , [qˆ, pˆ ] = iηδ ij Iˆ generate the single Pauli operators X (q) = e i − qpˆ η , Z ( p) = e i pqˆ η • X(q) is a position-translation operator (by the amount of q). • Z(p) is a momentum boost operator (kicking the moment by the amount of p). The action of single Pauli operators on the computational basis of position eigenstates i X (q ) s = s + q , Z ( p ) s = exp( sp ) s η 16 The action of the quibit operators The qubit operators can be generalized to ones for continuous variables based on their action: Hadamard gate H : X Z → → Z X Phase gate S : X Z → → Y Z → → → → X1 ⊗ X 2 I1 ⊗ X 2 Z1 ⊗ I 2 Z1 ⊗ Z 2 CNOT gate : X1 ⊗ I2 I1 ⊗ X 2 Z1 ⊗ I 2 I1 ⊗ Z 2 17 To preserve the generalized Pauil group – The special class of unitary operators to preserve the generalized Pauli group: Fourier transform gate, Phase gate, and SUM gate. Action Fourier Transform F : ⎞ ⎛i π 2 F = exp⎜ qˆ + pˆ 2 ⎟ ⎠ ⎝η 4 ( ) Phase gate P : ⎛ i ⎞ P(η ) = exp⎜ η qˆ 2 ⎟ ⎝ 2η ⎠ SUM gate: ) ⎞ ⎛ i SUM = exp⎜ − qˆ1 ⊗ p2 ⎟ ⎝ η ⎠ X (q ) Z ( p) X (q ) Z ( p) → → X 1 (q ) ⊗ I 2 Z1 ( p ) ⊗ I 2 I1 ⊗ X 2 (q ) I1 ⊗ Z 2 ( p ) → → e → → → → Z (q ) X ( p) −1 i ηq 2 2η X (q) Z (η q ) Z ( p) X 1 (q ) ⊗ X 2 (q ) Z1 ( p ) ⊗ I 2 I1 ⊗ X 2 (q ) −1 Z1 ( p ) ⊗ Z 2 ( p ) 18 Generalized Gottesman-Knill Theorem • The Generalized GK theorem for qunat quantum computation. – The generalized Pauli group: a set of single-oscillator Pauli operators. – The special class of unitary operators to preserve the generalized Pauli group: • Fourier transform gate → Hadamard gate • Phase gate → Phase gate S • SUM gate → CNOT gate – Possible measurement: measurements in position- or momentumeigenstate basis with finite losses and classical feed-forward. • Physics viewpoint: the simplicity is maintained by quadratic Hamiltonians which preserve the Gaussian features of the states. 19 Entanglement is not enough Shall we think such a computation circuit is simple, or rather complicated? • • Simple: A computation circuit which satisfy the GK requirements is simple enough to efficiently simulate on a classical computer. Complex: Such a circuit can generate maximally entangled states! The states which can be generated by the gate set of the GK theorem does not cover the entire Hilbert space. Hilbert space Hilbert space The initial state • Requirements for true quantum computation is much more complicated than just entanglement. • We need other measures to distinguish truly quantum computation from ones which can be simulated on a classical computer. 20 Beyond GK: Universal sets of gates It is known for qubit computation that The gate set of GK theorem Gates in GK Theorem + π/8-gate Universal gates Universal sets of gates efficiently classically simulated Qudit SUM gate, etc. + arbitrary single-dit transformation Qubit CNOT, Hadamard, etc. + Single-bit (π/8) rotation Qunat Quadratic Hamiltonian • • + third order or beyond The set of these gates in the GK theorem are not necessarily easy to realize in a physical system. Such a classification is nothing to do with physical difficulties in realization. 21 Gate sets and measurement strategies • • • • We have only considered projective measurements in the computational basis as our measurement strategies. However if we allow us to use other type of measurement, how may we change the possible quantum computation? For instance, Lloyd and Braunstein have shown that ideal photon counting is enough to complement quadratic Hamiltonian gates to be universal. Or we may move to measurement based quantum computation, where we do not need a universal set of gates to perform universal quantum computation. A gate set Universal set of gates + measurements 22 Non-standard quantum computation Divicenzo’s Criteria Criterion 1: An identifiable set of qubits, scalable in number for large scale quantum computing. Criterion 2: Accurate preparation of initial states. Criterion 3: The computation occurs through control of H and needs conditional dynamics. (Universal set of gates) Criterion 4: The error probability during an individual gate must be small. Criterion 5: Projective quantum measurements must be possible on chosen qubits. Operator-based QC Hamiltonian-based QC • Adiabatic passage quantum computation Gate-based QC • Quadratic Hamiltonian + Ideal photon counting Measurement-based QC • Cluster-state quantum computation Other approaches • Quantum Cellular automata • Global addressing 23 Universal quantum computation Qubit computation Universal gate sets: • CNOT + Single-rotation • CNOT + Hadamard + Phase + π/8-gate • etc. This non-universal set can simulate qubit quantum computation • • Non-universal gates: Controlled-rotation etc. It generates no states with imaginary components This set of gates is universal as rebit quantum computation but not universal in qubit computation. + measurements • • CNOT + Hadamard + Phase etc. These sets cannot simulate qubit quantum computation. 24 References • • • Stabilizer formalism: – M. A. Nielsen and I. L. Chuang, “Quantum Computation and Quantum Information”, Cambridge University Press, Cambridge (2000). The GK thorems: – D. Gottesman, “The Hesenberg Representation of Quantum Computers”, Group22: Proceedings of the XXII international Colloquium on Group Theoretical Methods in Physics, eds. S. P. Corney et al, (Cambridge, MA, International Press, 1999), p. 32, quant-ph/9807006, D. Gottesman, quantph/9802007. – S. D. Bartlett, B. C. Sanders, S. L. Braunstien, K. Nemoto, Phys. Rev. Lett., 88, 097904 (2002) . Universal quantum computation: – A. Barenco, et. al., Phys. Rev. A, 52, 3457 (1995). – S. Lloyd, S. L. Braunstein, Phys. Rev. Lett., 82 1784 (1999). – Y. Shi, quant-ph/0205115. – E. Farhi, et al., Science 292, 472 (2001). – M. A. Nielsen, “Optical quantum computation using cluster states”, quantph/0402005. 25