s
Transkrypt
s
• solid-state implementations of quantum computing • quantum tomography • quantum error correction • quantum algorithms • quantum teleportation • quantum cryptography • quantum entanglement • quantum logic gates Keywords summer semester 2008 http://zon8.physd.amu.edu.pl/∼miran e-mail: [email protected] Adam Miranowicz (with an emphasis on optical implementations) Methods of Quantum Information Processing 2 1 8. Quantum Computing and Communications by Sandor Imre, Ferenc Balazs 7. Quantum Computing by Josef Gruska 6. Quantum Information Processing edited by Gerd Leuchs, Thomas Beth 5. Explorations in Quantum Computing by Colin P. Williams, Scott H. Clearwater 4. Quantum Computing by Mika Hirvensalo 3. Introduction to Quantum Computation and Information edited by Hoi-Kwong Lo, Tim Spiller, Sandu Popescu 2. Approaching Quantum Computing by Dan C. Marinescu, Gabriela M. Marinescu 1. Quantum Computing by Joachim Stolze, Dieter Suter Other Textbooks on Quantum Computing 4. Quantum mechanical description of linear optics J. Skaar, J.C.G. Escartin, H. Landro, Am. J. Phys. 72, 1385-1391 (2005). 4 3. Quantum optical systems for the implementation of quantum information processing, T.C. Ralph, free downloads at http://arxiv.org/quant-ph/0609038. 2. Linear optics quantum computation: an overview, C.R. Myers, R. Laflamme, free downloads at http://arxiv.org/quant-ph/0512104. 1. Linear optical quantum computing, P. Kok, W.J. Munro, K. Nemoto, T.C. Ralph, J. P. Dowling, G.J. Milburn, free downloads at http://arxiv.org/quant-ph/0512071. Review articles on quantum-optical computing: by Michael A. Nielsen and Isaac L. Chuang (Cambridge, 2000) 1. Quantum Computation and Quantum Information Basic textbook: 3 9. An Introduction to Quantum Computing Algorithms by Arthur O. Pittenger 10. Quantum Computing by M. Nakahara, Tetsuo Ohmi 11. Principles of Quantum Computation and Information - Vol.1 by Giuliano Benenti, Giulio Casati, Giuliano Strini 5 12. Principles of Quantum Computation And Information: Basic Tools And Special Topics by Giuliano Benenti 13. Quantum Optics for Quantum Information Processing edited by Paolo Mataloni 14. Lectures on Quantum Information edited by Dagmar Bruß, Gerd Leuchs 15. A Short Introduction to Quantum Information and Quantum Computation by Michel Le Bellac 6 16. Quantum Computation and Quantum Communication by Mladen Pavicic 17. Scalable Quantum Computers: Paving the Way to Realization edited by Samuel L. Braunstein, Hoi-Kwong Lo 18. Fundamentals of Quantum Information edited by Dieter Heiss 19. Experimental Aspects of Quantum Computing edited by Henry O. Everitt 20. Quantum Computing: Where Do We Want to Go Tomorrow edited by Samuel L. Braunstein 21. Quantum Information by Gernot Alber et al. 22. The Physics of Quantum Information edited by Dirk Bouwmeester, Artur K. Ekert, Anton Zeilinger 23. Temple of Quantum Computing by Riley T. Perry (free downloads at www.toqc.com) 24. Quantum Computation edited by Samuel J. Lomonaco, Jr. (free downloads at www.cs.umbc.edu/∼lomonaco) 25. Lecture Notes for Physics: Quantum Information and Computation by John Preskill (free downloads at www.theory.caltech.edu/people/preskill/ph229). |gi - ground state |ei - excited state – physical notation 0 , 1 |ei = 1 0 |gi = 0 1 σ̂z ≡ 7 8 σ̂z = σ̂ †σ̂−σ̂σ̂ † = |eihe|−|gihg| 1 0 . 0 −1 a two-level atom/system – information notation |ei = 0 −i , i 0 1 , 0 σ̂y ≡ |gi = 0 1 , 1 0 • Pauli matrices σ̂x ≡ • rasing (σ̂ †) and lowering (σ̂) energy operators, σ̂x = atomic-transition operators σ̂ − iσ̂y + iσ̂ x y = |eihg|, σ̂ = = |gihe|, 2 2 σ̂ † = qubit = quantum bit • the smallest unit of quantum information • physically realized by a 2-level quantum system whose two basic states are conventionally labelled |0i and |1i • By contrast to classical bits, a qubit can be in an arbitrary superposition of ’0’ and ’1’: |ψi = α|0i + β|1i with normalization condition |α|2 + |β|2 = 1. • matrix representation of qubit states: 1 0 |0i = , |1i = 0 1 1 0 α +β = 0 1 β |ψi = α|0i + β|1i = α cn|ni Ŷ (α|0i + β|1i) = 0 −i i 0 k = 0, 1 01 10 examples 0 1 X̂|1i = X̂ = = |0i 1 0 01 α β X̂(α|0i + β|1i) = = = β|0i + α|1i 10 β α X̂|ki = |1 ⊕ ki, X̂ ≡ σ̂x = Pauli X̂ gate = quantum NOT gate = bit flip they are for quantum computers what classical logic gates are in conventional computers k = 0, 1 example 0 −i α −β =i = −iβ|0i + iα|1i i 0 β α 1 1 1 −1 Ĥ|−i = |1i examples 1 1 1 1 1 1 Ĥ|0i = √2 = √12 (|0i + |1i) ≡ |+i = √2 1 −1 1 0 1 1 1 0 1 1 √ √ Ĥ|1i = 2 = √12 (|0i − |1i) ≡ |−i = 2 1 −1 −1 1 1 1 1 2 1 1 1 1 √ =2 Ĥ|+i = √2 = = |0i 1 −1 2 1 0 0 Ĥ = √1 2 Hadamard gate Ŝ = 1 0 0 i basic quantum circuits operating on a small number of qubits Ŷ |ki = i(−1)k |1 ⊕ ki, Ŷ ≡ σ̂X = phase gate 10 example 1 0 α α = = α|0i − β|1i 0 −1 β −β Pauli Ŷ gate = phase flip + bit flip Ẑ(α|0i + β|1i) = k = 0, 1 1 0 0 −1 Ẑ|ki = (−1)k |ki, Ẑ ≡ σ̂z = Pauli Ẑ gate = phase flip quantum (logic) gates optical qudits are spanned in d-dimensional Fock space ... 5D qudit = (qu)quintit 4D qudit = (qu)quartit = ququart 3D qudit = qutrit 2D qudit = qubit special cases n=0 n=0 with normalization condition d−1 X |cn|2 = 1 |ψid = d−1 X d-dimensional quantum states, generalized qubits qudits = qunits 9 12 11 x |ψ y 2 |1i |0 |1 |0i + eiφ sin 2 |0 + |1 θ θ |0 Bloch representation/sphere z θ φ |ψi = cos 2 qubit rotations • rotations on Bloch sphere R̂n(2θ) ≡ exp (−iθn · σ̂ ) = σ̂I cos θ − in · σ̂ sin θ = σ̂I cos θ − i(nxσ̂x + ny σ̂y + nz σ̂z ) sin θ, • Pauli matrices (and identity matrix) 0 1 σ̂x ≡ = |0ih1| + |1ih0|, 1 0 0 −i σ̂y ≡ = −i|0ih1| + i|1ih0|, i 0 1 0 σ̂z ≡ = |0ih0| − |1ih1|, 0 −1 1 0 = |0ih0| + |1ih1| 0 1 σ̂I ≡ Iˆ = σ̂ = (σ̂x, σ̂y , σ̂z ) 13 + i |1 2 14 • rotations about k = x, y, z axes – special cases of R̂n(θ) R̂k (θ) = exp −i θ2 σ̂k = σ̂I cos 2θ − iσ̂k sin 2θ or explicitly cos 2θ −i sin 2θ X̂(θ) ≡ R̂x(θ) = −i sin 2θ cos 2θ cos 2θ − sin θ2 Ŷ (θ) ≡ R̂y (θ) = sin θ2 cos 2θ e−iθ/2 0 0 eiθ/2 Ẑ(θ) ≡ R̂z (θ) = • Do we need all XY Z-rotations? No! e.g. we can omit Z-rotation as Ẑ(θ) = X̂( π2 )Ŷ (θ)X̂(− π2 ) = Ŷ ( π2 )X̂(−θ)Ŷ (− π2 ). • any single-qubit gate can be written as a decomposition ZY (or, equivalently, XY or XZ) Û = eiαR̂n(θ) = eiαẐ(θ1)Ŷ (θ2)Ẑ(θ3) θ θ |0i + eiφ sin |1i 2 2 Bloch vector (Pauli vector) for a qubit ρ11 ρ12 ρ21 ρ22 = 12 (σ̂I + rBloch · σ̂) hσ̂i Tr[ρ̂σ̂] (Tr[ρ̂σ̂x], Tr[ρ̂σ̂y ], Tr[ρ̂σ̂z ]) (2Reρ21, 2Imρ21, ρ11 − ρ22) = 21 (σ̂I + rxσ̂x + ry σ̂y + rz σ̂z ) (so-called Pauli basis) rBloch ≡ (rx, ry , rz ) = (sin θ cos φ, sin θ sin φ, cos θ) |ψi = cos • in any pure state we have ρ̂ = • in any mixed state we get rBloch = = = = 15 16 then let ⊗ a11 a12 , a21 a22 a11 a12 a21 a22 b11 b12 b21 b22 B= b11 b12 b21 b22 b11 b12 b11 b12 a , a 12 11 b21 b22 b21 b22 = b b b b a21 11 12 , a22 11 12 b21 b22 b21 b22 a11b11 a11b12 a12b11 a12b12 a11b21 a11b22 a12b21 a12b22 = a21b11 a11b12 a22b11 a12b12 a21b21 a11b22 a22b21 a12b22 A⊗B = A= Kronecker tensor product |rBloch| = 1 – for pure state, |rBloch| < 1 – for mixed state • Properties: = ρ12 + ρ21 = ρ∗21 + ρ21 = 2Re ρ21 = 2Re ρ12 0 1 · 1 0 + 2rx + 0 + 0) ρ11 ρ12 ρ21 ρ22 ρ12 ρ11 = Tr ρ22 ρ21 Tr[ρ̂σ̂x] = Tr moreover = rx = 1 2 (0 = 12 (Tr[σ̂x] + rxTr[σ̂I ] + ry Tr[−iσ̂z ] + rz Tr[iσ̂y ]) = 21 (Tr[σ̂I σ̂x] + rxTr[σ̂xσ̂x] + ry Tr[σ̂y σ̂x] + rz Tr[σ̂z σ̂x]) = 21 Tr[σ̂I σ̂x + rxσ̂xσ̂x + ry σ̂y σ̂x + rz σ̂z σ̂x] Tr[ρ̂σ̂x] = Tr[ 21 (σ̂I + rxσ̂x + ry σ̂y + rz σ̂z )σ̂x] • Exemplary proof: 18 17 =α α β = γ δ 1 0 0 0 0 1 0 0 = α 0 + β 0 + γ 1 + δ 0 0 0 0 1 1 1 1 0 0 1 0 0 ⊗ +β ⊗ +γ ⊗ +δ ⊗ 0 0 0 1 1 0 1 1 |ψi = α|00i + β|01i + γ|10i + δ|11i matrix representation 1 0 1 1 |00i = ⊗ = 0 , 0 0 0 0 0 0 0 0 1 |01i = 0 , |10i = 1 , |11i = 0 , 1 0 0 matrix representation |ψi = |ψ ′iA ⊗ |ψ ′′iB ≡ |ψ ′iA|ψ ′′iB ≡ |ψ ′ψ ′′iAB ≡ |ψ ′ψ ′′i notation of two-mode states |ψi = α|00i + β|01i + γ|10i + δ|11i two-qubit pure states 20 19 Universal set of gates for quantum computing • rotations of single qubits • any nontrivial two-qubit gate (e.g., CNOT, NS, CZ) CZ |0, 0i |0, 1i |1, 0i −|1, 1i CNOT |0, 0i |0, 1i |1, 1i |1, 0i CZ and CNOT gates CZ = controlled phase gate (CPhase) control bit target bit |0i |0i |1i |0i |1i |0i |1i |1i = controlled sign gate (CSign) or equivalently CZ :|q1, q2i → (−1)q1q2 |q1, q2i CNOT : |q1, q2i → |q1, q1 ⊕ q2i where qk = 0, 1 and q1 ⊕ q2 = mod (q1 + q2, 2). Introduction to quantum-optical cryptography basic cryptographic terms Vernam protocol quantum key distribution BB84 protocol security and no-cloning theorem a note quantum cryptography is, probably, the most important application of quantum optics nowadays 22 21 24 23 basic cryptographic terms (I) Plaintext a string of numbers (letters) of our digital alphabet Encryption a computation which is usually quick and easy to perform Decryption quick and easy computation is only when some additional information is available otherwise it is very long and time consuming computation Key a set of instructions to encrypt and decrypt a message, e.g., randomly chosen series of numbers known to Alice and Bob only 25 01 06 11 16 21 26 31 36 A D H Ł Ó Ś X _ Ą E I M P T Y - 03 08 13 18 23 28 33 38 B Ę J N Q U Z ? 04 09 14 19 24 29 34 39 C F K Ń R V ś , cleartext: A D A M plaintext: 01060117 02 07 12 17 22 27 32 37 05 10 15 20 25 30 35 40 exemplary digital alphabet Ć G L O S W Ź . cleartext plaintext ciphertext plaintext cleartext 27 Decryption using key k2 (private key) Encryption using key k1 (public key) 28 Decryption using the same key k key distribution Encryption using key k asymmetric algorithms = public-key algorithms cleartext cryptology = cryptography + cryptoanalysis 26 plaintext ciphertext plaintext cleartext symmetric algorithms cryptosystem = cryptographic algorithm + all keys + all cryptograms + all cleartexts cipher 1. encryption scheme = cryptographic algorithm = math. function for encryption and decryption using a key 2. encrypted message = ciphertext = cryptogram basic cryptographic terms (II) eavesdropping on quantum system 31 first paper on quantum cryptography 29 1983 (1970 !) monitoring disturbs quantum information no-cloning theorem quantum teleportation seems to be also impossible ??? superluminal communication by using entangled states is impossible quantum cryptography is secure This is one of the most fundamental theorems of quantum mechanics and quantum information [Wootters, śurek and Dieks (1982)] It is impossible to make a copy of an unknown quantum state. 32 Eve cannot clone the information as she does not know the state of the „carrier” of information by Stephen Wiesner first description of quantum coding How to print banknotes, which cannot be counterfeited How to combine two messages such that by reading one of them, the other is automatically destroyed 30 eavesdropping on classical system two steps: 1. Eve makes a copy (clone) of the information carrier 2. and reads information from the copy passive monitoring of classical information is possible photons with slanted polarization photons polarized vertically photons polarized horizontally crystal enables discrimination of polarizations perpendicular to each other for example Iceland spar (calcite) or BBO a birefringent crystal 34 1. It is possible to disinguish two polarizations at a = 0o and 90o 2. It is possible to disinguish two polarizations at a = 45o and 135o 3. It is possible to change quickly polarization (e.g. by Pockels cell) 4. BUT it is impossible to measure simultaneously four polarizations at a=0o, 90o, 45o, 135o Heisenberg uncertainty principle passive eavesdropping is impossible 33 BB84 cipher Bob PROBLEM: How can Alice and Bob establish their common key? 2 stages: 1. quantum key distribution 2. classical encryption e.g.Vernam protocol Alice 1. quantum private for key distribution 2. classical public (e.g. internet) for cipher transmission key 36 35 scheme of Bennett & Brassard (1984) = BB84 protocol two channels: Vernam cipher/protocol (1918) = Che Guevara cipher = one-time pad algorithm addition modulo N (e.g. 40) key 1. a series of randomly chosen number 2. physically safe 3. not shorter than the length of message Vernam cipher is unbreakable if the above conditions are satisfied. example of Vernam protocol A M _ M I R A N O W I C Z _ Z 37 38 O N 16 19 24 03 13 24 07 25 10 23 20 19 22 38 14 16 12 16 11 Key is a series of random numbers: D cleartext: A plaintext: 01 06 01 17 36 17 12 24 01 18 20 30 12 04 33 36 33 20 18 -------------------------------------------------------------------------------------------- Sum: 17 25 25 20 49 41 19 49 11 41 40 49 34 42 47 52 45 36 29 Sum mod (40): 17 25 25 20 09 01 19 09 11 01 40 09 34 02 07 12 05 36 29 cipher BB84 (1) convention and / (a = 45o ) | (a = 90o ) and \ (a = 135o ) - (a = 0o ) 07 08 09 10 11 12 13 14 15 1.Alice sends photons 01 02 03 04 05 06 39 => bit 1 => bit 0 + + x + x x x + x + + x x + x base 40 | - \ | / / \ | / - | \ \ - \ polarization 1 0 1 1 0 0 1 1 0 0 1 1 1 0 1 bit BB84 (2) 2. Bob randomly chooses the measurement base polarization of Alice’s photons 07 08 09 10 11 12 13 14 15 | - \ | / / \ | / - | \ \ - \ Bob’s base 01 02 03 04 05 06 + x + + x x + + x + x x + + x | \ - | / / | | / - / \ - - \ polarization of photons after measurement 42 07 08 09 10 11 12 13 14 15 1 . . 1 0 0 . 1 0 0 . 1 . 0 1 Bob’s series 1 . . 1 0 0 . 1 0 0 . 1 . 0 1 Alice’s series 01 02 03 04 05 06 4. Alice and Bob keep only those results obtained for the same bases BB84 (4) test OK OK BB84 (6) OK Bob’s series * . . 1 * 0 . 1 0 * . 1 . * 1 1 0 1 0 1 1 thus the secret key known only to Alice and Bob is Alice’s series 07 08 09 10 11 12 13 14 15 44 * . . 1 * 0 . 1 0 * . 1 . * 1 01 02 03 04 05 06 6. They reject the bits for the tested photons OK 1 . . 1 0 0 . 1 0 0 . 1 . 0 1 Bob’s series + x + + x x + + x + x x + + x Bob’s bases y n n y y y n y y y n y n y y 1 . . 1 0 0 . 1 0 0 . 1 . 0 1 Alice’s series + + x + x x x + x + + x x + x Alice’s bases 07 08 09 10 11 12 13 14 15 01 02 03 04 05 06 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 3. Alice and Bob publicly compare bases BB84 (5) 43 5. Alice and Bob publicly check results for some photons say 1, 5, 10, 14th. BB84 (3) 41 Test of agreement 1. Alice and Bob compare arbitrary subset of their data. (Obviously, the tested subset is then not used for the key.) 2. Methods: (a) testing bit after bit (b) testing parity e.g 20 times ⇒ (1/2)20 ∼ 0.000001 3. If the subset reveals eavesdropping, all the data of Alice and Bob are rejected and the BB84 protocol is repeated. 4. privacy amplification 45 46 via e.g. Bennett-Brassard-Robert, Ekert et al. or Horodecki et al. schemes Eve’s strategy of eavesdropping (I) Eve measures a photon sent by Alice (e.g. by using another calcite crystal) QUESTION What is the probability that a single photon was measured by Eve, but Alice and Bob have not realized it? ANSWER P = 43 the probability is surprisingly high! Base Polarization Probability 47 = 1/8 strategy of eavesdropping Alice + Eve + 1/2 Bob + | 1/2 = 1/8 Bob + Alice + | Eve x / 1/2*1/2 Bob + | 1/2 48 = 1/2 Alice + | Eve x \ 1/2*1/2 security of BB84 for 1 photon P1=3/4 so for n photons Pn=(3/4)n e.g. P2=(3/4)2 ~ 0.56 P10=(3/4)10 ~ 0.06 P20=(3/4)20 ~ 0.003 P100=(3/4)100 ~ 10-13 and for 1000 photons P1000=(3/4)1000 ~10-125 49 1997 Grover algorithm for searching databases: How to search the keys more effectively? 1994 Shor algorithm for finding discrete logarithms: How to break ElGamal cryptosystem? 1994 Shor algorithm for number factorization: How to break cryptosystems of RSA, Rabin, Williams, Blum-Goldwasser,…? 1985 Deutsch (Deutsch-Jozsa / DJ) algorithm: How to see both sides of a coin simultaneously? quantum algorithms and cryptography 50 2004 Englert et al. protocol claimed to be the most efficient nowadays (Singapore protocol) 1992 Bennett protocol using any two nonorthogonal states (B92) 1992 Bennett-Brassard-Mermin protocol a la Ekert protocol but without Bell’s inequality 1991 Ekert protocol based Bell’s inequality (E91) 1984 Bennett-Brassard protocol (BB84) famous protocols of quantum key distribution A bipartite mixed state is P separable if ρAB = i piρiA ⊗ ρiB P entangled if ρAB 6= i piρiA ⊗ ρiB P where i pi = 1 and pi ≥ 0 for all i. separable if |ψAB i = |ψAi ⊗ |ψB i entangled if |ψAB i = 6 |ψAi ⊗ |ψB i A bipartite pure state is definition have to be described with reference to each other. the quantum states of two or more objects (possibly spatially separated) it is a quantum phenomenon in which (Schrödinger 1935) quantum entanglement = inseparability = Verschränkung = entwinement quantum cryptography is a branch of physics - mainly of quantum optics classical cryptography is a branch of mathematics „Information is inevitably tied to a physical representation and therefore to restrictions and possibilities related to the laws of physics” [R. Landauer] information is physical 52 51 quantum entanglement & Einstein-Podolsky-Rosen paradox → → S 1+ S 2 = 0 2 S 1y + S 2 y = 0 S → 53 How to ‘violate’ uncertainty relation ? Can we measure exactly both compontents of spin? 2 !2 S 1z 4 1 var S 1 x var S 1 y ≥ → S S 1 x + S 2 x = 0, 54 one of the triplet states singlet state So let’s measure S 1 x i S 2 y to determine S 2 x i S 1 y Bell states (EPR states) maximally entangled two-qubit states ΦA = Ψ ( − ) ΦB = Ψ ( + ) ΦC = Φ ( − ) 1 = ( 0 x 1 y − 1 x 0 y) 2 1 = ( 0 x 1 y + 1 x 0 y) 2 1 ( 0 x 0 y − 1 x 1 y) = 2 1 ( 0 x 0 y + 1 x 1 y) 2 ΦD = Φ ( + ) = 55 GHZ (Greenberger-Horne-Zeilinger) states maximally? entangled states of 3 qubits ' ΨGHZ ' '' ΨGHZ √1 2 56 1 1 '' ( ( 001 ± 110 ) 000 ± 111 ) = ΨGHZ = 2 2 1 iv 1 ( 100 ± 011 ) Ψ ( 010 ± 101 ) = GHZ = 2 2 W states 1 ( 100 + 010 + 001 ) ΦW = 3 1 ( 011 + 101 + 110 ) 3 ΨW = √1 (|00i 2 0 1 , |Ψ(+)i = √12 (|01i + |10i) = √12 1 0 1 0 − |11i) = |Φ(+)i = 0 1 Bell states in matrix representation √1 2 0 1 , |Ψ(−)i = √12 (|01i − |10i) = √12 −1 0 1 0 , − |11i) = |Φ(−)i = 0 −1 √1 (|00i 2 a GHZ state in matrix representation √1 2 √1 ([10000000]T 2 + [00000001]T ) = √1 [10000001]T 2 |Ψ′i = √12 (|000i + |111i) 1 1 1 0 0 0 ⊗ ⊗ + ⊗ ⊗ 0 0 0 1 1 1 = = quantum limit: λ/(2N ) for N -photon absorption classical limit: λ/2 • interferometric lithography beyond diffraction limit [Boto et al.’00] • better clock synchronization [Jozsa et al.’00, Chuang’00] • to improve frequency standards [Huelga et al. ’97] • quantum noise reduction in spectroscopy [Wineland et al.’92] to increase precision of quantum measurements: • remote state preparation [Bennett et al.’01] • telecloning [Murao et al. 99] in quantum state engineering: • to simplify communication complexity [Cleve & Buhrman’97] in quantum communication: Applications of quantum entanglement (II) • quantum watermarking [Imoto et al.’05] • quantum authentication [Ljunggren et al.’00] • secret sharing [Żukowski et al.’98] • privacy amplification [Bennett et al.’96, Deutsch et al.’96] • quantum key distribution [Ekert’91, Bennett et al.’92] in quantum cryptography: • quantum error correction [Shor’95, Steane’96] • superfast [Shor’94] and fast [Grover’97] algorithms • entanglement swapping [Żukowski et al.’93] • quantum teleportation [Bennett et al.’93] • dense coding [Bennett and Wiesner’92] in quantum information theory: Applications of quantum entanglement (I) 58 57 |01i−|10i √ 2 60 where = aa′|00i + ab′|01i + ba′|10i + bb′|11i = (a|0i + b|1i) ⊗ (a′|0i + b′|1i) which admits no solution. 1 1 aa′ = 0, ab′ = √ , ba′ = − √ , bb′ = 0 2 2 So we get set of equations a, a′, b, b′ ∈ C |a|2 + |b|2 = |a′|2 + |b′|2 = 1 |01i−|10i √ 2 proof that it cannot be written as a product state |ψi = example of inseparable state - singlet state 1 1 |ψi = [|0iA(|0iB + |1iB ) + |1iA(|0iB + |1iB )] = (|0iA + |1iA)(|0iB + |1iB ) 2 2 |0iA + |1iA |0iB + |1iB √ √ · ≡ |+iA|+iB = 2 2 1 |ψi = (|00i + |01i + |10i + |11i) 2 proof: example 2 |+iB 1 |0iB + |1iB 1 √ = |1iA|+iB |ψi = √ (|1iA|0iB +|1iA|1iB ) = √ |1iA(|0iB +|1iB ) = |1iA 2 2 | {z2 } proof: 1 |ψi = √ (|10i + |11i) 2 example 1 examples of separable states 59 example of inseparable (Bell-like) state proof: =|−iB 1 |ψi = (|00i + |01i + |10i − |11i) 2 =|+iB |0i − |1i |0+i + |1−i |0i + |1i 1 B B B B √ √ √ +|1iA = |ψi = √ |0iA 2 2 {z2 } | {z2 } | =|−i this state can be obtained from a Bell state by applying Hadamard gate to 2nd qubit |00i + |11i √ |ψi = HB 2 =|+i 1 = √ (|0iA(H|0i)B + |1iA(H|1i)B ) | {z } | {z } 2 1 = √ (|0iA|+iB + |1iA|−iB ) 2 generation of all Bell states from one of them |00i + |11i √ 2 e.g. from |Φ(+)i = flip qubit B = apply NOT gate to qubit B: |00i + |11i |01i + |10i √ √ = |Ψ(+)i = XB 2 2 phase flip qubit B: |00i + |11i |00i − |11i √ √ = = |Φ(−)i 2 2 flip + phase flip qubit B: |00i + |11i |01i − |10i √ √ =i = i|Ψ(−)i 2 2 ZB YB 61 62 How to generate Bell states from |00i Bell state |Φ(+)i : = #1 |0i + |1i |00i + |10i √ = UCNOT √ |0i = U CNOT 2 2 |0, 0 ⊕ 0i + |1, 0 ⊕ 1i √ = 2 |00i + |11i √ = |Φ(+)i 2 UCNOTHA|00i = UCNOT| + 0i Bell state |Φ(−)i : #2 |0, 0 ⊕ 0i − |1, 0 ⊕ 1i |00i − |11i √ √ = = |Φ(−)i 2 2 UCNOTHAXA|00i = UCNOTHA|10i = UCNOT| − 0i = |01i + |10i √ = |Ψ(+)i 2 |0, 1 ⊕ 0i + |1, 1 ⊕ 1i √ 2 How to generate Bell states from |00i Bell state |Ψ(+)i : = = UCNOT| + 0i = UCNOTHAXB |00i = UCNOTHA|01i Bell state |Ψ(−)i : UCNOTHAXAXB |00i = UCNOTHA|11i |0, 1 ⊕ 0i − |1, 1 ⊕ 1i √ 2 |01i − |10i √ = |Ψ(−)i 2 = UCNOT| − 1i = = 63 64 l qubit B H CNOT Bellkl |0li + (−1)k |1, 1 ⊕ li √ = |Bellkl i 2 k qubit B H CNOT Bellkl' CNOT’ in | out -------00 | 00 01 | 11 10 | 10 11 | 01 SWAP in | out -------00 | 00 01 | 10 10 | 01 11 | 11 66 65 CNOT CNOT’ SWAP ÛSWAP|ψini = c0|00i + c2|01i + c1|10i + c3|11i ≡ c0|0i + c2|1i + c1|2i + c3|3i ′ |ψini = c0|00i + c3|01i + c2|10i + c1|11i ≡ c0|0i + c3|1i + c2|2i + c1|3i ÛCNOT ÛCNOT|ψini = c0|00i + c1|01i + c3|10i + c2|11i ≡ c0|0i + c1|1i + c3|2i + c2|3i |ψini = c0|00i + c1|01i + c2|10i + c3|11i ≡ c0|0i + c1|1i + c2|2i + c3|3i CNOT in | out -------00 | 00 01 | 01 10 | 11 11 | 10 CNOT and SWAP gates l qubit A ′ UCNOT HB |lki = |Bell’kl i obviously, we can obtain Bell states by: k qubit A UCNOTHA|kli = general formula (k, l = 0, 1) Generation of Bell states 0 0 1 0 0 0 0 1 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 c0 c0 c1 c2 0 = 0 c 2 c 1 1 c3 c3 0 c0 c0 c1 c3 1 = 0 c 2 c 2 0 c3 c1 0 c0 c0 c1 c1 0 = 1 c 2 c 3 0 c3 c2 xy = ( 1 ↑ x ↓ y− ↓ x ↑ 2 y ) xy = ( y y ) ) 68 xy = 1 2 (0 x 1 y + e iχ 1 x 0 y ) projection of separable state onto entangled one Φ output light of beam splitter 50:50 with a single input photon Ψ 1 H x V y + e iχ V x H 2 1 Φ xy = H x H y + e iχ V x V 2 ( light generated by parametric down conventer (PDC II) Ψ decay of a particle with spin 0 into 2 particles with spin ½ Generation of Bell states c0 1 c1 0 ÛSWAP c2 = 0 0 c3 c0 1 c ′ 1=0 ÛCNOT c2 0 c3 0 c0 1 c1 0 ÛCNOT c2 = 0 c3 0 CNOT and SWAP gates in matrix representation 67 Entanglement from a mystery to a physical resource 69 • An entangled wave-function does not describe the physical reality in a complete way. [Einstein,Podolsky,Rosen] • For an entangled state is the best possible knowledge of the whole does not include the best possible knowledge of its parts. [E. Schrödinger] 70 • Entanglement is a “fundamental resource of Nature, of comparable importance to energy, information, entropy, or any other fundamental resource.” [M. Nielsen, I. Chuang] Entanglement is... • a correlation that is stronger than any classical correlation. [J. Bell] • a correlation that contradicts the theory of elements of reality. [D. Mermin] • a trick that quantum magicians use to produce phenomena that cannot be imitated by classical magicians. [A. Peres] • a resource that enables quantum teleportation. [C. Bennett] • a global structure of the wavefunction that allows for faster algorithms. [P. Shor] • a tool for secure communication. [A. Ekert] • the need for first applications of positive maps in physics. [Horodecki family] [collected by Dagmar Bruß, quant-ph/0110078] → S 1 → S 1+ S → S = 0 2 → 2 − | ↓i | ↑i | ↑↓i − | ↓↑i 1 2 √ √ ≡ 2 2 1 2 71 b 72 filter 2 Measurement of entangled spins and Bell inequalities Stern-Gerlach filter a filter 1 N++(α, β) – number of outcomes when spin +1 was measured in filter 1 at angle α and in filter 2 at angle β N – total number of outcomes | ↑i1| ↓i2 probability P++(α, β) = N++(α, β)/N • QM predictions |ψ (−)i = for electrons in singlet state then α−β 1 P++(α, β) = P−−(α, β) = sin2 2 2 1 2 α−β cos 2 2 P+−(α, β) = P−+(α, β) = spins of opposite values 1 4 P+−(α, α) = P−+(α, α) = ⇒ measurement of 2 independent spins +α P++(α, α) = P−−(α, α) = 0, ⇒ special cases: π 2 P++(α, β) = P+−(α, β) = P−+(α, β) = P−−(α, β) = ⇒ 1. α = β 2. β = ⇒ 74 73 • “the deepest discovery in history of science” [Stapp] as enables to decide experimentally a dispute between Einstein and Bohr. • violation of a Bell inequality implies that the world is nonrealistic or/and nonlocal – measurement by filter in position 1 does not influence the result of measurement by filter in a distant place 2 3. locality – physical objects have properties independent of whether we measure them or not 2. realism – this is very natural assumption – from large number of measurements probability can be determined 1. reasoning by induction Bell assumptions P[a_, b_] := 1/2*Sin[(a - b)/2]^2 LHS = P[0, Pi/4] + P[Pi/4, Pi/2] RHS = P[0, Pi/2] Mathematica: LHS = sin2(π/8) = 0.146 · · · ≥ \ RHS = 0.25 occurs, e.g., for α = 0, β = π/4, γ = π/2 • violation of Bell inequality P++(α, β) + P++(β, γ) ≥ P++(α, γ) N++(α, β) + N++(β, γ) ≥ N++(α, γ) • Bell inequality for 3 series of measurements PM C C PM IˆB − C 75 hIˆA+IˆB+i + hIˆA−IˆB−i + hIˆA+IˆB−i + hIˆA−IˆB+i h(IˆA+ − IˆA−)(IˆB+ − IˆB−)i h(IˆA+ + IˆA−)(IˆB+ + IˆB−)i hIˆA+IˆB+i + hIˆA−IˆB−i − hIˆA+IˆB−i − hIˆA−IˆB+i 76 S = E(φ′A, φ′B ) − E(φ′A, φ′′B ) + E(φ′′A, φ′B ) + E(φ′′A, φ′′B ) should be satisfied for realistic local theories. where |S| ≤ 2 • Bell inequality according to Clauser-Horne-Shimony-Holt at any angles φA, φB . = E(φA, φB ) = • What is measured in the experiment? PM – photon multipliers PA, PB – polarizing beam splitters = optical analogues of Stern-Gerlach filters Iˆk ∼ n̂k – intensity of kth beam S – source of photons C – correlation systems Key: C IˆA− Aspect et al. tests of Bell-CHSH inequalities (1984) PA ˆ PB ˆ I B+ IA IˆB IˆA+ PM PM S |HiA|V iB − |V iA|HiB √ |Ψ(−)i = 2 • polarization singlet state • QM predicts that 77 E(φA, φB ) = P++(φA, φB ) + P−−(φA, φB ) − P+−(φA, φB ) − P−+(φA, φB ) where Pjk (φA, φB ) in the former slides • maximal violation of Bell inequality √ max |S| = 2 2 II. loopholes in BI assumptions 1. independence assumption: 79 There are physical processes independent of the Bell experiment that can be used as an effective source of randomness. III. interpretational loopholes Accepting that Bell’s theorem is true then either locality or realism might be false. 1. universe is non-local but real e.g. Bohmian hidden variable theory is explicitly non-local and contextual, but still fairly natural looking. QM is just a set of recipes for calculating probabilities of measurement results. the standard Copenhagen interpretation: 2. operationalism of quantum mechanism (QM) which can be obtained for the following angles of polarizers π π π ′ ′ ′′ ′′ φA = 0, φB = , φA = , φB =3 4 2 4 as π √ π 3π 3π S = −3 cos2 + 3 sin2 = −2 2 ≈ −2.8 + cos2 − sin2 8 8 8 8 loopholes in Aspect’s et al. experiment (1982) 1. a fair-sample assumption • BI test more ideal than ever by Zeilinger et al. (PRL 1998) on “Violation of Bell’s inequality under strict Einstein locality conditions”. The necessary space-like separation of the observations was not achieved by sufficient physical distance between the measurement stations. (ii) no true space-like separation Aspect et al. switched the directions of polarization analysers after the photons left the source, but they used periodic sinusoidal switching, which is predictable. Analyzers were not randomly rotated during the flight of the particles. 2. causality (i) no true randomization All experiments so far detect only a small subset of all pairs created thus it is necessary to assume that the pairs registered are a fair sample of all pairs emitted. 80 It says nothing about an underlying physical reality, which may not even exist, and therefore nothing about its locality. 78 • Note: the same setup of Aspect was used for testing violation of Schwarz inequality possible loopholes of Bell’s inequalities (BI) • in the present experimental BI tests • in the proof or assumptions of BI • in the BI interpretation I. experimental loopholes there has been no loophole-free experimental test of BI. 1. detection efficiency and fair-sample assumption: None experimental test does not detect 100% the particle pairs emitted. Thus it is not clear that the pairs registered are a fair sample of all pairs emitted. 2. causality: The choice of measurement settings of Alice and Bob should be truly random and placed at sufficient physical distance. i pi ⊗ ρiB = 1 and pi ≥ 0 for all i. i piρiA = 1 and pi ≥ 0 for all i. pi|ψAi ihψAi | ⊗ |ψBi ihψBi | i pi P i X i pi = 1 and pi ≥ 0 for all i. i p ||Â|| = Tr|Â| = Tr( †Â) trace norm If ρ̂AB acts in finite-dimensional Hilbert spaces then the approximation part is redundant. systems of finite dimensions where P A state ρ̂AB is separable iff it can be written or approximated (in, e.g., trace norm) by X ρ̂AB = piρiA ⊗ ρiB General definition of bipartite entanglement These definitions are equivalent, as ρiA (and ρiB ) can be expanded in terms eigenvectors X ρjA = qj |ψAj ihψAj | P where j qj = 1 and qj ≥ 0 for all i. where ρAB = Def. 2. A state ρAB is separable iff where P ρAB = X Two definitions of entanglement Def. 1. A state ρAB is separable iff 82 81 ihΨ (+) |= i=1 5 X pi|aiihai| ⊗ |biihbi| 1−p 1̂ ⊗ 1̂ + p|Ψ(±)ihΨ(±)| 4 |01i ± |10i √ 2 B(ρ̂W ) > 0 for all √1 2 <p≤1 and violates Bell inequality E(ρ̂W ) > 0 for all 13 < p ≤ 1 the Werner state is entangled 1 0 1̂ ⊗ 1̂ ≡ 1̂2 ⊗ 1̂2 = 1̂4 = |00ih00| + |01ih01| + |10ih10| + |11ih11| = 0 0 and maximally mixed state |Ψ(±)i = is mixture of maximally entangled state (±) ρ̂W = Werner state thus it is not a convex combination of product states and the state is entangled. note that p4 = p5 < 0 |0i + |1i √ 2 |0i + ei2π/3|1i √ |a2i = |b2i = 2 |0i + e−i2π/3|1i √ |a3i = |b3i = 2 |a4i = |0i, |b4i = |1i, |a5i = |1i, |b5i = |0i |a1i = |b1i = 2 1 p1 = p2 = p3 = , p4 = p5 = − 3 2 |Ψ (+) e.g. the ‘triplet’ state Can we decompose Bell states into separable states? 0 1 0 0 0 0 1 0 0 0 0 1 84 83 = i=1 7 X pi|aiihai| ⊗ |biihbi| decomposition of Werner state (+) ρ̂W 2 1 − 3p 1−p p1 = p2 = p3 = , p4 = p5 = , p6 = p7 = 3 4 4 = = = = |0i, |1i, |0i, |1i, |0i + |1i |a1i = |b1i = √ 2 |0i + ei2π/3|1i √ |a2i = |b2i = 2 |0i + e−i2π/3|1i √ |a3i = |b3i = 2 |b4i = |1i |b5i = |0i |b6i = |0i |b7i = |1i |a4i |a5i |a6i |a7i clearly, the state is separable only for p ≤ 13 when all pi ≥ 0. 1−p 1̂ ⊗ 1̂ + p|ψihψ| = 4 Werner and Werner-like states ρ̂W (p) |01i−|10i √ 2 original Werner state |ψi → |Ψ(−)i = |00i+|11i √ 2 isotropic state (Werner-like state) |ψi → |Φ(+)i = |ψi → |Ψ(+)i = |00i−|11i √ 2 |01i+|10i √ 2 other Werner-like states |ψi → |Φ(−)i = 1−p d2 (+) 1̂⊗d + p|Φd ihΦd | (+) two qudit (d-dimensional) isotropic state ρ̂W (p) = √1 d where − d21−1 ≤ p ≤ 1 Pd−1 i=0 |iii (+) |Φd i = 85 86 properties of Werner states (−) 87 88 1. They are so-called maximally entangled mixed states (MEMS) • entanglement E(ρW ) cannot be increased by any unitary operations, • entanglement E(ρW ) is maximal for a given linear entropy (and vice versa) [Ishizaka, Hiroshima’00, Munro et al.’01] (−) 2. Original Werner state exhibits U ⊗ U invariance: Û ⊗ Û ρ̂W = ρ̂W √ 3. All entangled Werner states (even for p ∈ (1/3, 1/ 2i) can be used for quantum-information processing including teleportation [Popescu’94, Lee,Kim’00] nonlocality quantum states are called nonlocal if there is no local unhidden variable model of their behavior. Thus a measurement of the whole can reveal more information about the system’s state than any sequence of classically coordinated measurements of the parts. 1. nonlocality without entanglement [Bennett et al. 1999] There are orthogonal sets of product states of two qutrits that cannot be reliably distinguished by a pair of separated observers ignorant of which of the states has been presented to them, even if the observers are allowed to communicate by LOCC. Note: LOCC = LQCC local quantum operations and classical communication 1 √ |uiatom|aliveicat + |diatom|deadicat 2 where |di – decayed state, |ui – undecayed state. • state of the isolated atom after a time equal to its half-time is 1 √ |ui + |di 2 – automatic device to release the poison when the atom decays – radioactive atom – a bottle with poison gas – a cat image a chamber containing: • gedanken experiment of Schrödinger cat – can we talk about of superposition, entanglement or smearing of classical objects? 92 • Note: the above states are not mixtures of coherent states 1 ρ̂mix = |αihα| + | − αih−α| 6= |ψihψ| 2 1 |ψYSi = √ |αi ± i| − αi 2 – Yurke-Stoler states for ϕ = ±π/2: ∞ X 1 α2n+1 p |ψ−i = N− |αi − | − αi = p |2n + 1i ≡ |1iL sinh(α2) n=0 (2n + 1)! – odd coherent state for ϕ = π: ∞ X α2n 1 p |2ni ≡ |0iL |ψ+i = N+ |αi + | − αi = p cosh(α2) n=0 (2n)! – even coherent state for ϕ = 0: • questions Schrödinger cats: N = [2 + 2 cos ϕ exp(−2α2)]−1/2 where N is a normalization constant assuming α ∈ R: |ψi = N [|αi + eiϕ| − αi] • e.g. superposition of two coherent states special cases of single-mode – how to interpret superposition of states and entanglement? 91 Note: the definition can be applied to a single mode, thus not necessarily has to be related to entanglement it is a superposition of two macroscopically distinct states (modern) definition of Schrödinger cat (state) the „quantum chamber” has to be opened to check the state of the „cat”. to understand it, one has to include the observer or measurement process: • Copenhagen interpretation so the state is • but the atom is entangled with the cat via the apparatus Schrödinger cat paradox 90 Note: two-qubit pure states violate Bell inequality iff they are entangled. thus they can be described in terms realistic local theories are entangled and satisfy Bell inequality e.g. Werner states and isotropic states for √ p ∈ (1/3, 1/ 2i two-qubit mixed states can be entangled without violating Bell inequality [Werner 1989] 2’. entanglement without violating Bell inequality (although there are some serious doubts about it) are often identified nonlocality and violation Bell-inequality 2. entanglement without nonlocality 89 Problems 93 two-mode Schrödinger cats Hamiltonians (for simplicity assume g = |g|) 1. hamiltonian for a degenerate process Ĥdeg = h̄g[â2 + (â2)†] 2. hamiltonian for a nondegenerate process of type I Ĥ1 = h̄gâs† âi† + h.c. Note: signal (s) and idler (i) have the same polarization ωp pump r ks r kp χ ( 2) nonlinear crystal r ki momentum conservation idler 95 ω s signal ωi ωs ωi energy conservation ωp spontaneous parametric down-conversion (SPDC) 96 a nonlinear optical phenomenon in which a nonlinear crystal splits incoming photons into pairs of photons of lower energy. spontaneous parametric down-conversion (SPDC) – How to generate Bell polarization states? – How to generate single photons? 94 continuous variable (CV) version of two-qubit Bell states 1 |Φ(+)i = √ (|00iL + |11iL) = N (|α, βi + | − α, −βi) 2 1 |Φ(−)i = √ (|00iL − |11iL) = N (|α, −βi + | − α, βi) 2 1 |Ψ(+)i = √ (|01iL + |10iL) = N (|α, βi − | − α, −βi) 2 1 |Ψ(−)i = √ (|01iL + |10iL) = N (|α, −βi − | − α, βi) 2 assumption α = β exemplary proof of a CV Bell state definition of logical qubits (α = β): |0iL = N+ |βi + | − βi |1iL = N− |βi − | − βi so 2N+N− |α, βi + | − α, −βi +|α, βi − |α, −βi − | − α, βi + | − α, −βi N N− h + √ |α, βi + |α, −βi + | − α, βi + | − α, −βi 2 √ i i 1 N N− h + √ (|00iL+|11iL) = √ (|αi+|−αi)(|βi+|−βi)+(|αi−|−αi)(|βi−|−βi) 2 2 = = 97 signal pump idler o-ray e-ray 98 pump "red" "green" "blue" type II parametric down converter orange red deep red type I parametric down converter pump type II parametric down converter pump type I parametric down converter |ψ(0)i = |0iV s|0iHs|0iV i|0iHi iτ |ψ(t)i = (1 − 12 τ 2)|0is|0ii − √ |1is|1ii 2 |ψ(0)i = |0is|0ii |1iV s|0iHs ≡ |V is, |0iV s|1iHs ≡ |His, 1 |Ψ(+)i = √ (|V is|Hii + |His|V ii) 2 the second term is a polarization Bell state iτ |ψ(t)i = (1 − 12 τ 2)|0is|0ii − √ (|V is|Hii + |His|V ii) 2 thus we have and analogously for mode i |0iV s|0iHs ≡ |0is, • notation: |ψ(t)i = (1 − 12 τ 2)|0iV s|0iHs|0iV i|0iHi iτ − √ (|1iV s|0iHs|0iV i|1iHi + |0iV s|1iHs|1iV i|0iHi) 2 – type II process: where τ = gt – type I process: " 2# t 1 t 1 |ψ(0)i Ĥt |ψ(0)i ≈ 1 + Ĥ + Ĥ |ψ(t)i = exp ih̄ ih̄ 2 ih̄ short-time evolution Note: signal and idler modes have perpendicular polarizations â†V s – creation operator for signal mode of vertical polarization etc. V (H) – linear vertical (horizontal) polarization where Ĥ2 = h̄g(â†V sâ†Hi + â†Hsâ†V i) + h.c. 3. hamiltonian for a nondegenerate process of type II 100 99 101 How to generate other polarization Bell states? – place half-wave plate (HWP) on the path of one beams: 1 1 |Ψ(+)i = √ (|V is|Hii + |His|V ii) → √ (|His|Hii + |V is|V ii) = |Φ(+)i 2 2 – use phase shifter or simply rotate crystal to change α: 1 |Ψ(+)i → |ψ (α)i = √ (|V is|Hii + eiα |His|V ii) 2 • this is how all four Bell states can be obtained including: 1 |Ψ(−)i = √ (|V is|Hii − |His|V ii) 2 1 |Φ(−)i = √ (|His|Hii − |V is|V ii) 2 the experimental method was developed by Kwiat, Zeilinger et al. (1995) using crystals of BBO type II. 102 standard matrix representations of qubit states 1 1 1 −1 so far we have been applying the orthonormal basis 1 0 |0i = , |1i = 0 1 √1 2 satisfying the identity resolution 1 0 1 0 0 0 1 0 [10] + [01] = + = 0 1 0 0 0 1 0 1 Iˆ = |0ih0| + |1ih1| = 0 1 1 0 then the NOT gate is X̂ = |0ih1| + |1ih0| = and the Hadamard gate is Ĥ = |0ih+| + |1ih−| = |+ih0| + |−ih1| = can we define other representations? 1 2 √1 2 1 1 1 1 1 −1 1 0 [11] + 21 [1, −1] = 12 + 21 = 1 −1 1 1 −1 1 0 1 identity resolution then 1 =0 h0|1i = 21 [1, 1] −1 |0i = let’s use the orthonormal basis 1 1 |1i = √12 1 −1 another matrix representation of qubit states Iˆ = NOT gate 1 2 0 1 1 1 X̂ = |0ih1| + |1ih0| = 12 [1, −1] + 21 [1, 1] 1 −1 1 −1 1 1 1 0 + 21 = 1 −1 −1 −1 0 −1 = Hadamard gate |+i = √1 (|0i 2 √1 (|0i 2 − |1i) = note that 1 + |1i) = 0 |−i = so phase-flip gate Ĥ = |0ih+| + |1ih−| 1 1 1 1 [10] + √ [01] = √ 2 1 2 −1 1 0 1 1 1 1 1 1 0 +√ =√ = √ 2 1 0 2 0 −1 2 1 −1 1 2 1 1 Ẑ = |0ih0| − |1ih1| = 1 [1, 1] − 1 [1, −1] 2 2 1 −1 1 1 1 −1 0 1 − 21 = 1 1 −1 1 1 0 = 103 104 or equivalently ′ |0i cos θ sin θ |0i = − sin θ cos θ |1i |1i′ orthonormal basis cos θ − sin θ , |1i = sin θ cos θ CNOT a 00 + b 11 |1i = c|xi + d|yi = ac|xxi + ad|xyi + bc|yxi + bd|yyi − ac|xxi − ad|yxi − bc|xyi − bd|yyi √ 2 ad(|xyi − |yxi) − bc(|xyi − |yxi) √ = 2 |xyi − |yxi |xyi − |yxi a b |xyi − |yxi √ √ √ = det = eiφ = (ad − bc) c d 2 2 2 |01i − |10i (a|xi + b|yi) ⊗ (c|xi + d|yi) − (c|xi + d|yi) ⊗ (a|xi + b|yi) √ √ = 2 2 proof |0i = a|xi + b|yi, unitary transformation proof |00, 900i − |900, 00i |θ, θ + 900i − |θ + 900, θi √ √ = 2 2 it does not exist a general quantum cloning device as the device can clone only orthogonal states. thus either |ψ1i = |ψ2i or |ψ1i is orthogonal to |ψ2i equation x = x2 has only trivial solutions: x = 0 and x = 1 so Û |ψ2i ⊗ |si = |ψ2i ⊗ |ψ2i so † Û |ψ2i ⊗ |si = hψ1|ψ2i LHS = hs| ⊗ hψ1|Û RHS = hψ1| ⊗ hψ1| |ψ2i ⊗ |ψ2i = hψ1|ψ2i2 Û |ψ1i ⊗ |si = |ψ1i ⊗ |ψ1i let us clone two states It is impossible to copy (clone, reproduce) an unknown quantum pure state. 108 Measurements of the polarization singlet state at different angles of the crystal now measure one of the qubits... = a2|00i + ab(|01i + |10i) + b2|11i Û a|0i + b|1i ⊗ |si → a|0i + b|1i ⊗ a|0i + b|1i requirement for a cloning device It is a cloning device iff ab = 0. now measure one of the qubits... 0 a 0 +b 1 Quantum no-cloning theorem: 106 CNOT as a cloning device? ÛCNOT a|0i + b|1i ⊗ |0i = a|0, 0 ⊕ 0i + b|1, 0 ⊕ 1i = a|00i + b|11i 107 singlet state and rotations of analyzer Iˆ = cos θ − sin θ [cos θ, sin θ] + [− sin θ, cos θ] sin θ cos θ cos2 θ sin θ cos θ sin2 θ − sin θ cos θ 1 0 = + = sin θ cos θ sin2 θ − sin θ cos θ cos2 θ 0 1 − sin(2θ) cos(2θ) NOT gate: X̂ = cos(2θ) sin(2θ) − sin(2θ) + cos(2θ), sin(2θ) + cos(2θ) 1 Hadamard gate: Ĥ = √2 sin(2θ) + cos(2θ), sin(2θ) − cos(2θ) identity resolution in terms of standard qubit representations |0i = [1; 0] and |1i = [0; 1] |0i = rotated matrix representation of qubit states 105 Further questions 1. Can we clone unknown mixed quantum states? Impossible! 2. What about cloning by non-unitary operations? Impossible! 3. What about optimal approximate cloning machines? For example, a cloning machine of Bužek and Hillery (1996) 4. Can we clone known quantum states? Yes. 5. Can we clone known quantum states by local operations? In general impossible! 109 Quantum no-deleting theorem: Ûdel|ψi|ψi|Ai = |ψi|Si|Aψ i we try to find a unitary operation Ûdel such that proof It is not (time) reverse of quantum no-cloning theorem. note It is impossible to delete an unknown quantum state against a copy. where |ψi = a|0i + b|1i - arbitrary qubit state |Ai – initial state of ancilla |Aψ i – final state of ancilla (dependent on |ψi) |Si – some standard final state of the qubit Let us analyze polarization states |ψi = a|Hi + b|V i We expect Ûdel|Hi|Hi|Ai = |Hi|Si|AH i Ûdel|V i|V i|Ai = |V i|Si|AV i 110 |1> Quantum cloning and stimulated emission ω 01 Ûdel Thus, we have LHS = = = = RHS = a|Hi|Si|Aψ i + b|V i|Si|Aψ i Ûdel|ψi|ψi|Ai Ûdel(a|Hi + b|V i) ⊗ (a|Hi + b|V i)|Ai Ûdel[a2|Hi|Hi + ab(|Hi|V i + |V i|Hi) + b2|V i|V i]|Ai √ a2|Hi|Si|AH i + 2ab|Bi + b2|V i|Si|AV i where |Bi is some combined input-ancilla state. |Hi|V i + |V i|Hi √ |Ai = |Bi 2 |0> For each perfect clone there is also one randomly polarized, spontaneously emitted, photon. Another question Suppose that we have several copies of a photon in an unknown state. Is it possible to delete the information content of one or more of these photons by a physical process? having only solution for |Aψ i = a|AH i + b|AV i ∀ψ hAψ |Aψ i = 1 ⇒ hAH |AV i = 0 111 112 |Bi = |Hi|Si|AV i + |V i|Si|AH i √ 2 114 [it might be you...] 4. quantum no-deleting for mixed states ??? Given a general mixed state for a quantum system, there are no physical means for broadcasting that state onto two separate quantum systems, even when the states need only be reproduced marginally on the separate systems. specifically It is impossible to copy an unknown quantum mixed state. [Barnum, Caves, Fuchs, Jozsa, Schumacher (1996)] 3. quantum no-broadcasting It is impossible to delete an unknown quantum pure state. [Pati, Braunstein (2000)] 2. quantum no-deleting It is impossible to copy an unknown quantum pure state. [Wootters, Żurek, Dieks (1982)] 1. quantum no-cloning no-go theorems |Aψ i = a|AH i + b|AV i. “deleting” is just swapping onto 2D subspace of ancilla as 113 √ |Hi|Si|AV i + |V i|Si|AH i √ a2|Hi|Si|AH i + 2ab + b2|V i|Si|AV i 2 a2|Hi|Si|AH i + ab|Hi|Si|AV i + ab|V i|Si|AH i + b2|V i|Si|AV i a|Hi|Si a|AH i + b|AV i + b|V i|Si a|AH i + b|AV i a|Hi + b|V i |Si|Aψ i = RHS Thus it is seen that = = = LHS = so and EPR EPR B PM PM 115 ( 1 H 2 V A −V B A H B )= ( 1 + 2 A − B − + EPR EPR EPR EPR PM PM PM PM 116 A − B ) BIT 0 BIT 0 BIT 1 BIT 1 a patent for superluminal communication Ψ = PM = photon multiplier EPR = source of the polarization singlet state (e.g. BBO) A superluminal communication? Alice Bob (super)dense coding [Bennett and Wiesner 1992] + |1iA|1iB √ 2 + |1iA|1iB √ 2 117 Ẑ|Φ(+)i = |Φ(−)i |Φ(+)i ⇒ 118 X̂|Φ(+)i = |Ψ(+)i iŶ |Φ(+)i = |Ψ(−)i ⇒ ⇒ ⇒ How to send 2 bits of information by transmitting 1 qubit? |0iA|0iB 1. Alice and Bob share 2 qubits in an EPR state e.g. |Φ(+)i = 2. if Alice wants to send a message: 00 - then she does nothing 01 - then flips phase of her qubit (Pauli Z gate) 10 - then flips her qubit (Pauli X gate) 11 - then flips & phase-flips her qubit (Pauli iY gate) 3. Alice sends her qubit to Bob 4. Bob measures both qubits using Bell state analyzer superdense coding - a detailed analysis |0iA|0iB 1. Alice takes qubit A and Bob takes qubit B from an EPR pair: |Φ(+)i = 2. Alice encodes her message by applying proper gate to her qubit A: |00i → IˆA|Φ(+)i = |Φ(+)i |10i + |01i √ = |Ψ(+)i 2 |10i + |01i −|10i + |01i √ √ = = |Ψ(−)i 2 2 |00i − |11i √ = = |Φ(−)i 2 |01i → X̂ A|Φ(+)i = |10i → Ẑ A|Φ(+)i |11i → ẐA X̂A|Φ(+)i = ẐA 2. Alice sends qubit A to Bob. So, he has now two entangled qubits A and B. • H • 89:; ?>=< fe fe |βmni |mi |ni 119 120 |00i + |11i |0, 0 ⊕ 0i + |1, 1 ⊕ 1i |00i + |10i |0i + |1i √ √ √ = = = √ |0i = |+, 0i 2 2 2 2 3. Bob applies CNOT to remove entanglement between qubits A and B: ÛCNOT |10i + |01i |11i + |01i |1i + |0i √ √ = = √ |1i = |+, 1i ÛCNOT 2 2 2 |00i − |11i |00i − |10i |0i − |1i √ √ = = √ |0i = |−, 0i ÛCNOT 2 2 2 |01i − |10i |01i − |11i |0i − |1i √ √ = = √ |1i = |−, 1i 2 2 2 ÛCNOT 4. Bob applies Hadamard gate to qubit A: ĤA|+, 0i = |00i ĤA|+, 1i = |01i ĤA|−, 0i = |10i ĤA|−, 1i = |11i 4. Bob measures qubits A and B in the standard basis. H q-circuit for Bell-state generation |ni |mi |βmni 89:; ?>=< for m, n = 0, 1 |β10i = |Φ(−)i, where |β01i = |Ψ(+)i, or compactly |0, ni + (−1)m|1, 1 − ni √ 2 |β11i = |Ψ(−)i q-circuit for Bell-state analysis |β00i = |Φ(+)i, |βmn i = 121 in 1 BSM ABCD Alice 2 EPR source 3 ABCD Bob U out Bennett, Brassard, Crepeau, Jozsa, Peres, Wootters (1993) principles of teleportation 122 any quantum circuit can be realized using only teleportation and singlequbit operations. • Universal quantum computation via teleportation: • Teleportation can provide the quantum channel for communication between quantum computers. • Super-luminal communication via teleportation is impossible. (why?) • Teleportation does not violate the no-cloning theorem. (why?) • Teleportation is not a transfer of an object itself, neither its energy etc. • Teleportation is a transfer of information about a quantum system without measuring it! Remarks: and transmission of some classical information. over large distances via entangled particles a method to transfer (information about) unknown quantum states What is quantum teleportation? = (a 0 1 + b 1 1)⊗ 2 0 21 3− 1 2 0 3 12 12 ΦC ΦD where 12 12 3 x → (−1) x +1 x ⊕ 1 phase flip + bit flip bit flip x → − x ⊕ 1 phase flip x → (−1) x x OK 124 Φ A = Ψ ( − ) , ΦB = Ψ ( + ) , ΦC = Φ ( − ) , Φ D = Φ ( + ) ⇒ − iσy (a 1 3 − b 0 ) ) 3 σz (a 0 3 − b 1 3 ) a 0 3+b1 3 ⇒ − σx (− a 1 3 − b 0 ⇒ ⇒ measurement in Bell basis ΦB ΦA 23 1 1 ΦA 12 ⊗ (a 0 3 + b 1 3 ) − ΦB 12 ⊗ (a 0 3 − b 1 3 ) 2 2 1 1 + ΦC 12 ⊗ (a 1 3 + b 0 3) + ΦD 12 ⊗ (a 1 3 − b 0 3 ) 2 2 =− in 1 ⊗ ΦA Problem: How to teleport state of qubit 1 to qubit 3 ? Assumption: qubits 2 & 3 are in the singlet state explanation of teleportation 123 12hΨ Quantum teleportation - a detailed analysis Let us analyze teleportation of a polarization state as in Zeilinger’s experiment |φi1 = a|0i1 + b|1i1 = a|Hi1 + b|V i1 via the singlet state 1 |Ψ(−)i23 = √ (|HV i23 − |V Hi23) 2 main idea Let’s expand the total initial state |φi1|Ψ(−)i23 in the Bell basis 1 |Ψ(±)i = √ (|HV i ± |V Hi) 2 1 |Φ(±)i = √ (|HHi ± |V V i) 2 note that 125 126 |Ψ(+)ihΨ(+)| + |Ψ(−)ihΨ(−)| + |Φ(+)ihΦ(+)| + |Φ(−)ihΦ(−)| = I4 ≡ id(4) which leads to |φi1|Ψ(−)i23 = (|Ψ(−)ihΨ(−)| + |Ψ(+)ihΨ(+)| + |Φ(−)ihΦ(−)| + |Φ(+)ihΦ(+)|)12|φi1|Ψ(−)i23 = |Ψ(−)i12(12hΨ(−)|φi1|Ψ(−)i23) + |Ψ(+)i12(12hΨ(+)|φi1|Ψ(−)i23) + |Φ(−)i12(12hΦ(−)|φi1|Ψ(−)i23) + |Φ(+)i12(12hΦ(+)|φi1|Ψ(−)i23) |φi1|Ψ(−)i23 Term 1: (−) 1 1 = √ ( hHV | − hV H|)(a|Hi + b|V i ) √ (|HV i − |V Hi ) 12 12 1 1 23 23 2 2 1 = a|HHV i − a|HV Hi + b|V HV i − b|V V Hi 12hHV | − 12hV H| 2 1 − a 12hHV |HV Hi − b 12hV H|V HV i = 2 1 = − a|Hi3 + b|V i3 2 1 = − |φi3 2 12hΨ |φi1|Ψ(−)i23 Term 2: (+) since Term 3: (−) |φi1|Ψ(−)i23 12hΦ since Term 4: (+) |φi1|Ψ(−)i23 12hΦ 1 0 0 −1 a b = a −b = a|Hi − b|V i 127 1 = a|HHV i − a|HV Hi + b|V HV i − b|V V Hi 12hHV | +12 hV H| 2 1 = − a 12hHV |HV Hi + b 12hV H|V HV i 2 1 1 = − (a|Hi3 − b|V i3) = − Ẑ|φi3 2 2 Ẑ|φi3 = 0 1 1 0 a b = b a = b|Hi + a|V i 128 1 = a|HHV i − a|HV Hi + b|V HV i − b|V V Hi 12hHH| −12 hV V | 2 1 a hHH|HHV i + b 12hV V |V V Hi = 12 2 1 1 (a|V i3 + b|Hi3) = X̂|φi3 2 2 = X̂|φi3 = a b = 0 −1 1 0 a b = −b a = −b|Hi + a|V i 1 a|HHV i − a|HV Hi + b|V HV i − b|V V Hi = 12hHH| +12 hV V | 2 1 a hHH|HHV i − b 12hV V |V V Hi = 12 2 1 1 (a|V i3 − b|Hi3) = (−iŶ )|φi3 2 2 = 0 −i i 0 since (−iŶ = X̂ Ẑ) −iŶ |φi3 = −i 1 (−) |Ψ i12|φi3 + |Ψ(+)i12Ẑ|φi3 − X̂|Φ(−)i12|φi3 + |Φ(+)i12(iŶ )|φi3 2 1 −|Ψ(−)i12|φi3 − |Ψ(+)i12Ẑ|φi3 + X̂|Φ(−)i12|φi3 + |Φ(+)i12(−iŶ )|φi3 2 thus we have |φi1|Ψ(−)i23 = = − Now, Alice performs a Bell-state measurement (BSM) - a projective measurement in the Bell-state basis (on particles 1 and 2). n 2 3 |Ψ(−)i23 Xn • Zm • |φi3 PL P D1 H fe fe n m 4 BBO 1 BS 3 2 Alice D2 D4 130 D3 EOM PBS Bob D 2 in (no photon) Alice = a 0 + b 1 BS1 0 out D1 (1 photon) = BS2 a 0 Bob teleportation of optical qubits using quantum scissors PBS – polarizing beam splitter b 1 1 + 132 PC – polarization controller to correct extra rotation of polarization in fibres Xn BS – beam splitter EOM – electro-optic modulator to perform Bob’s unitary operation |φi3 131 BBO – β-barium borate used to generate entangled photon pair (wavelength 788 nm) by spontaneous parametric down-conversion (PDC) PL – pulsed laser (wavelength 394 nm, rate 76 MHz) RF - unit – classical channel F – optical fibre (length 800m) – quantum channel qubit – polarization states |Hi and |V i of photon Key: experiment of Zeilinger at al. (2004) teleportation across the River Danube • Zm • Zeilinger's experiment i23 89:; ?>=< • D0 trigger |Ψ (−) |φi1 or more explicitly where m, n ∈ {0, 1} and BSM stands for the Bell state measurement BSM m 1 |φi1 q-circuit for quantum teleportation 129 FD VA 0 PDC D0 out BS1 in realization of quantum scissors BS delay - ‘trombone’ system D0 - acts as a trigger D0, D1, D2 – photon-counting detectors BS, BS1, BS2 – beam splitters CCL – coincidence counter and logic VA – variable attenuator with narrow-band filter PDC – parametric down conversion crystal FD – frequency doubler PL – pulsed laser experimental quantum scissors for teleportation and generation of qubit states Key: PL ay D el BS2 D1 133 D2 134 CCL 1 1 1 Bell state generator Z π 2 hide detector hide detector -1 hide -1 2 π experiment of Blatt et al. (2004) UX π 2 , i0 + 1 135 detector 2 = 100 ms , T delay = 10 ms 136 Z X UX-1 deterministic teleportation of atomic states hide Blatt et al. experiment (2004) 1 ≡ S1/ 2 ( m J = − 1 / 2 ) 3 ions of 40Ca+ states 0 ≡ D5 / 2 (m J = −1 / 2) 2 0 + 1 H ≡ D5 / 2 (m J = −5 / 2) Ux 1 → 1 , 0 , .lifetime (66.7%<) 75% (<87%) T teleport = 2 ms , T Bell.state fidelity: |e>n laser mode La D2 (L) wn Ca BS Alicja Dn wn (C) Cb 138 |g>n cavity mode Lb Ab |r>n 3-level atom Aa D1 Bob teleportation via cavity decay 137 4 139 4 1 ABCD 2 EPR sources BSM 3 ABCD entanglement swapping U 140 5. Thus, although Alice and Claire never interacted with each other, their particles A and C are now entangled. 4. Claire, as in the standard teleportation protocol, applies proper unitary operation to C. So, the state of B1 is teleported to C. 3. Bob performs BSM on his particles B1 and B2 and informs Claire about his measurement results. 2. Initially particles A and B1 are entangled, and so B2 and C. 1. Alice has particle A, Bob two particles B1, B2, and Claire particle C. Simple description • quantum repeaters can be based on entanglement swapping • it is quantum teleportation of unknown entangled quantum state Remarks: a method to establish perfect entanglement between remote parties What is the entanglement swapping? 3 141 What can be interesting about linear optics? Can we construct a quantum computer composed only of: teleportation - theory and experiments T 1935 quantum entanglement - Schrödinger, and Einstein, Podolsky, Rosen • beam splitters • phase shifters • single-photon sources • photodetectors [3] Briegel and Raussendorf [PRL 2001]: “A one-way quantum computer” [4] Zeilinger et al. [Nature 2005]: “Experimental one-way quantum computing” How to encode optical qubits • single-rail qubits logical values of qubits are encoded by number of photons |0iL = |0i – vacuum |1iL = |1i – single-photon state • two-rail qubits encoding in spatial modes |0iL = |1i ⊗ |0i = |10i – photon in the first mode |1iL = |0i ⊗ |1i = |01i – photon in the second mode or vice versa • polarization qubits encoding in polarization modes of qubits |0iL = |Hi = |1h, 0v i – photon with horizontal polarization 143 144 “[Experimental] demonstration of an all-optical quantum controlled-NOT gate” [2] O’Brien et al. [Nature 2003]: “A scheme for efficient quantum computation with linear optics” [1] Knill, Laflamme and Milburn [Nature 2001]: Yes! T 1982 quantum no-cloning - Wootters, Żurek and Dieks T 1993 quantum teleportation - Bennett,Brassard,Crepeau,Jozsa,Peres,Wootters T 1993 entanglement swapping - Żukowski, Zeilinger, Horne, Ekert E 1997 limited (conditional) optical teleportation - Zeilinger et al. E 1998 unconditional optical teleportation - Furusawa, Kimble, Polzik et al. E 1998 optical entanglement swapping - Zeilinger et al. T 1999 universal quantum computation via teleportation - Gottesman and Chuang E 2004 unconditional teleportation of atomic states - (i) Barrett, Wineland at al. and (ii) Riebe, Blatt at al. E 2006 unconditional teleportation between light and matter - Polzik et al. E 2006 two-qubit teleportation - Zhang, Pan et al. Key: T - theory, E - experiment 142 Introduction to linear-optical quantum computing encoding optical qubits linear-optical elements scattering matrices conditional measurements optical single-qubit gates optical two-qubit gates a motto efficient quantum computation with linear optics is possible! |1iL = |V i = |0h, 1v i – photon with vertical polarization or vice versa B′′ = B = eiθ0 t ir ir t t exp(iθt) r exp(iθr ) −r exp(−iθr ) t exp(−iθt) b̂1 â2 1 2 n1 n2 ψ out light transmitted from any side of BS does not change its phase light reflected from the ‘black’ surface does not change its phase the phase of light reflected from the ‘white’ surface is π-shifted • convention |ψini - input state, e.g. |ψini = |n1i1|n2i2 ≡ |n1, n2i |ψouti - output state, b̂1,2 – annihilation operators for output modes â1,2 – annihilation operators for input modes b̂2 â1 scheme of (asymmetrical) BS θt,r,0 – phase shifts 1 = T + R = t2 + r 2 R ≡ ρ = r – reflectance = reflectivity = reflection coefficient 2 T ≡ τ = t2 – transmittance = transmittivity= transmission coefficient r – [probability] amplitude of reflection t – [probability] amplitude of transmission where t r , −r t the most general form of BS matrix B′ = typical BS matrix Beam splitter (BS) = partially-transparent mirror 146 145 b̂1 b̂2 ψ in S b̂N b̂2 b̂1 â1 â2 BS j=1 N X Sij âj |ψini = |n1, . . . , nN i ≡ |ni for N -input Fock states b̂ ≡ [b̂1, b̂2, . . . , b̂N ]T â ≡ [â1, â2, . . . , âN ]T • in vector notation b̂j – annihilation operator for the jth output mode âj – annihilation operator for the jth input mode Û – unitary operator describing evolution of N input modes Sij – elements of unitary scattering matrix S where âi → b̂i = Û †âiÛ = |ψouti = Û |ψini transformations of states and operators in MP S – scattering matrix â N â1 â2 linear optical multiport (MP) â2 â1 equivalent schemes of BS 148 b̂1 ψ b̂2 out 147 the MP so âi† = j X j Sjib̂j† Û † Sjib̂j† ↠= Û b̂†Û † = ST b̂† b̂ = Û †âÛ = S â transformations can be written compactly: or explicitly N i=1 N X SjiÛ b̂j† Û † Sjiâj† – is the total number of photons in system. n2 nN – multiple sum j1, j2, ..., jM n1 149 150 • How to express output operators in terms of input operators without knowing explicitly Û ? j X j X Û âi†Û † = Û = = general transformations in MP i ni Y 1 √ (Û âi†Û †)ni |0i = ni ! i=1 ni N N N M Y X X Y 1 1 √ Sjiâj† |0i = √ Sj x ↠|0i ni! j=1 n1! · · · nN ! j=1 l=1 l l jl = (↠)n2 (↠)nN (↠)n1 = Û √1 Û †Û √2 Û †Û · · · Û †Û √N Û †Û |0i n1 ! n2 ! nN ! i=1 Y (â†)ni (↠)n1 (↠)n2 (↠)nN √i |0i = Û √1 √2 · · · √N |0i ni ! n1 ! n2 ! nN ! input state |ψini = |n1, . . . , nN i ≡ |ni output state P |ψouti = Û |ψini = Û where M = j {xl } ≡ (1, ..., 1}, 2, ..., 2}, ..., |N, {z ..., N}) for l = 1, ..., M | {z | {z P → t r −r t =|00i Sj1âj† |00i |ψouti = t|10i − r|01i S = B′ = |ψini = |10i j √ 2rt(|20i − |02i) + (t2 − r2)|11i 151 152 Exemplary transformations of states by beam-splitter assuming that • example 1: proof: |ψouti = Û |ψini = Û |10i = Û â1† |00i =1 = Û â1† |{z} Û †Û |00i P = Û â1† Û † Û |00i | {z } | {z } = = S11â1† |00i + S21â2† |00i = S11|10i + S21|01i = t|10i − r|01i so |10i → t|10i − r|01i √ in special case for 50:50 BS (i.e., t = r = 1/ 2) |01i → r|10i + t|01i 1 |10i → √ (|10i − |01i) ≡ |Ψ(−)i (the singlet state) 2 • example 2: |11i → for 50:50 BS 1 |01i → √ (|10i + |01i) ≡ |Ψ(+)i (the ′triplet′ state) 2 • example 3: for 50:50 BS 1 |11i → √ (|20i − |02i) 2 note: state |11i is not generated – it is so-called photon coalescence or photon ‘bunching’ (but not in Mandel’s sense) √ 2rt|11i + r2|02i √ 1 |20i → |20i − 2|11i + |02i 2 |20i → t2|20i − √ 1 √ 3|30i − |21i − |12i + 3|03i |21i → √ 2 2 153 â2 â1 â2 â1 for 50:50 BS r 3 1 1 |0, 4i − |2, 2i + 2 2 2 r 3 |4, 0i 2 t − r r t t r S= − r t convention 1. â2 â1 b̂2 b̂1 â2 â1 b̂1 b̂2 r t t − r − r t t r convention 2. scattering matrices for BS 1 |22i → 2 b̂2 b̂1 b̂2 b̂1 154 i √ h |22i → (r4 −4r2t2 +t4)|2, 2i+ 6rt rt|0, 4i+(r2 −t2)(|1, 3i−|3, 1i)+rt|4, 0i • example 6: for 50:50 BS • example 5: √ √ |21i → 3rt2|30i + t(t2 − 2r2)|21i + (r3 − 2rt2)|12i + 3r2t|03i for 50:50 BS • example 4: e iθ Sθ = 0 0 r 0 S= t 0 â4 â3 0 t 1 0 0 −r 0 0 BS θ BS1 BS2 S = S 2 Sθ S1 = S 2 S1Sθ t2 r2 0 b̂2 b̂1 b̂3 0 0 − r2 1 0 0 1 0, S1 = 0 r1 t1 , S 2 = t 2 0 0 t1 − r1 0 1 â3 â2 â1 b̂4 b̂3 0 0 0 1 exemplary scattering matrix (II) 1 0 0 S = 0 r t 0 t − r b̂3 b̂2 â2 b̂2 â2 â3 b̂1 â1 b̂1 â1 BS exemplary scattering matrices (I) 0 0 1 156 155 159 Example of von Neumann projective measurement 157 von Neumann projective measurement a projection synthesis via conditional measurements general two-qubit pure state prob1(0) ≡ prob(|0i1) = |c0|2 + |c1|2 c |0i + c1|1i p0 ≡ |φ0i |c0|2 + |c1|2 √ ... is the renormalization. and post measurement state is where 2) probability of measuring 1st qubit in |1i: prob(|1i1) = |c2|2 + |c3|2 if |ψi = c0|00i + c1|01i + c2|10i + c3|11i the post measurement state is c |0i + c3|1i p2 ≡ |φ1i |c2|2 + |c3|2 160 let us assume that the kth subsystem (e.g. mode, qubit) of a bipartite quantum system is represented by complete orthonormal set of states |mi then: |ψi = c0|00i + c1|01i + c2|10i + c3|11i normalization projector = projection operator = projection valued (PV) measure = orthogonal measurement operator is (k ′ ) [P̂m(k), P̂m′ ] = 0 1) probability of measuring 1st qubit in |0i: 1 = |c0|2 + |c1|2 + |c2|2 + |c3|2 158 P̂m(k) = |mikk hm| for the kth subsystem corresponding to the measurement outcome m probability of the measurement outcome m ˆ prob1(m) = hψ|P̂m(1) ⊗ I|ψi prob2(m) = hψ|Iˆ ⊗ P̂m(2)|ψi (2) state after projection/measurement (1) ˆ P̂m ⊗ I|ψi Iˆ ⊗ P̂m |ψi |φ(1)i = q , |φ(2)i = q (1) (2) ˆ hψ|P̂m ⊗ I)|ψi hψ|Iˆ ⊗ P̂m |ψi √ ... is the renormalization. where requirements for the projectors then = δmm′ , non-negative, Hermitian, orthogonal and summing up to identity: X ˆ P̂m(k) = (P̂m(k))†, = I, P̂m(k) m ′ k hm|m ik X 3) probability of measuring 2nd qubit in |0i: prob(|0i2) = |c0|2 + |c2|2 prob(m) = 1 if |ψi = c0|00i + c1|01i + c2|10i + c3|11i m measurement outcomes corresponding to non-orthogonal states do not commute and thus are not simultaneously observable - analogously 4) probability of measuring 2nd qubit in |1i: c |0i + c3|1i p1 |c1|2 + |c3|2 the post measurement state is Note: number of projectors = dimension of Hilbert space • Can we apply a more general type of measurement? Yes! positive operator valued measures (POVM, POM) describe measurement outcomes associated with non-orthogonal states ... and we will discuss it later! 161 # ⇒ b̂†1 b̂†2 " ST = =S T # S11 S21 S12 S22 â†1 â†2 ψ φa no photons no photons ψ Answer: Proper conditional measurements should be performed |φai ∼ c1|0i + c2|1i and |φbi ∼ c5|0i + c7|1i? Problem: How to get single-mode states φb no photons 1 photon |ψi = c1|000i+c2|001i+c3|010i+c4|011i+c5|100i+c6|101i+c7|110i+c8|111i 0 1 ψ S = S3 S 2 S1 0 1 0 0, S 3 = 0 r3 0 t3 1 0 t3 − r3 t2 r2 0 − r2 1 0 0 S1 = 0 r1 t1 , S 2 = t 2 0 0 t1 − r1 BS3 b̂2 „0” „1” ψ' b̂3 BS1 BS2 b̂1 â3 â2 â1 |ψi = c0|0i + c1|1i + c2|2i −→ |ψ ′i = c0|0i + c1|1i − c2|2i example: Let’s analyze a three-mode state 163 164 |100i → S11â†1|000i + S21â†2|000i + S31â†3|000i = S11|100i + S21|010i + S31|001i e.g. † â†1 b̂†1 â†1 â S11 S21 S31 † † † †1 â2 → b̂2 = ST â2 = S12 S22 S32 â2 S13 S23 S33 â†3 b̂†3 â†3 â†3 • Analogously for a three-mode system (six-port system) holds: probabilistic quantum state engineering via conditional measurements 162 S11 S12 S21 S22 → " |01i → S12â†1|00i + S22â†2|00i = S12|10i + S22|01i # |10i → S11â†1|00i + S21â†2|00i = S11|10i + S21|01i S= â†1 â†2 nonlinear sign gate (NS) p p prob(|0i1)|0i ⊗ |φ0i + prob(|1i1)|1i ⊗ |φ1i e.g. where " • As we know, for a two-mode system (four-port system) holds: projection synthesis = p c0|0i + c1|1i p 2 c2|0i + c3|1i = |c0|2 + |c1|2 |0i ⊗ p + |c2| + |c3|2 |1i ⊗ p | {z } {z } |c0|2 + |c1|2 | |c |2 + |c |2 | | 2 {z 3 } {z } = |0i ⊗ (c0|0i + c1|1i) + |1i ⊗ (c2|0i + c3|1i) |ψi = c0|00i + c1|01i + c2|10i + c3|11i another description of projection synthesis • scattering matrix Mathematica: S1 := {{1, 0, 0}, {0, r1, t1}, {0, t1, -r1}} S2 := {{-r2, t2, 0}, {t2, r2, 0}, {0, 0, 1}} S3 := {{1, 0, 0}, {0, r3, t3}, {0, t3, -r3}} S = S3.S2.S1 thus, after multiplication, one gets S11 S12 S13 −r2, t2 r 1 , t1 t2 S = S21 S22 S23 = t2r3, t1t3 + r1r2r3, −t3r1 + t1r2r3 ↔ Û S31 S32 S33 t2 t3 , t3 r 1 r 2 − t1 r 3 , t1 t3 r 2 + r 1 r 3 • probability amplitudes hn10|Û |n10i Problem: find such reflection amplitudes r1, r2 and r3 for BS that p 1 − ri2 h010|Û |010i = h110|Û |110i = −h210|Û |210i ≡ c Note: transmission amplitudes for BS are calculated from ti = 165 166 Û |010i = (S12â1† + S22â2† + S32â3† )|000i = S12|100i + S22|010i + S32|001i • |n10i → c|n10i for n = 0 so h010|Û |010i = S12h010|100i+S22h010|010i+S32h010|001i = S22 = t1t3+r1r2r3 = c k,l=1 = (S11S22 + S21S12)|110i + · · · = (S11S22â1† â2† + S21S12â2† â1† + · · · )|000i 3 X S S √k1 l2 âk† âl†|000i Û |110i = 1!1! • |n10i → c|n10i for n = 1 so h110|Û |110i = (S11S22 + S21S12)h110|110i + · · · = S11S22 + S21S12 = −r2c + (t2r3)(t2r1) = −r2c + r1r3t22 3 X S k1Sl1Sm2 † † † √ âk âl âm|000i 2!1! k,l,m=1 2|210i + 2S11S21S12) â12†â2† |000i + · · · | √ {z } 167 2† † † 2† † † † √1 (S 2 S22â â +S21S11S12â â +S11S21S12â â â +· · · )|000i 1 2 2 1 1 2 1 2 11 √1 (S 2 S22 2 11 2 = (S11 S22 + 2S11S21S12)|210i + · · · = = Û |210i = • |n10i → −c|n10i for n = 2 so 2 h210|Û |210i = S11 S22 + 2S11S21S12 = r22c − 2r2(t2r3)(t2r1) = r22c − 2r1r2r3t22 finally, we get set of equations 168 c = t1t3 + r1r2r3 = −r2c + (t2r3)(t2r1) = −(r22c − 2r1r2r3t22) √ for which the solutions are r1 = r3 = √ 1 √ , r2 = 2 − 1 4−2 2 simplified implementation of NS gate 1 ψ BS2 ψ' „1” „0” BS2 ψ' 0 BS1 „0” ψ 1 „1” 0 BS1 (less number of optical elements but also lower probability of success!) (a) (b) NS BS 50:50 π π BS2 BS1 BS3 „1” „1” t1 = t2 = t3 = cos θ where θ = 54.74 and t4 = cos φ where φ = 17.63 transmission amplitudes for BSs: target qubit 1 1 control qubit another implementation of CZ gate BS 50:50 NS implementation of CZ gate BS4 170 169 n2 a^ 2 a^ 4 B2 P2 P6 B3 P3 P4 ^ b1 B4 P5 φ b2 ^ 1 0 0 t3 B3 = 0 −r3 0 0 0 0 0 1 0 0 0 t5 r 5 0 −r5 t5 Mk = diag[exp(iζδ1k ), exp(iζδ2k ), exp(iζδ3k ), exp(iζδ4k )] mirror Pk = diag[exp(iξk ), 1, 1, 1] dla k = 1, 2, 6, Pk = diag[1, exp(iξk ), 1, 1] dla k = 3, 4, P5 = diag[1, 1, exp(iξ5), 1]. phase shifters r2 0 t2 0 0 0 , 0 1 1 0 B5 = 0 0 0 1 0 0 beam splitters S = P6B5P5B4P4B3P3B2P2B1P1 0 r3 t3 0 D3 b^ 4 D2 b3 B5 ^ unitary transformations ψ a^3 B1 P1 0 t2 0 0 , B2 = −r2 0 1 0 1 0 0 0 0 t4 0 r 4 B4 = 0 0 1 0 , 0 −r4 0 t4 0 0 1 0 n3 r1 t1 0 0 1 t1 −r1 B1 = 0 0 n1 a^ multiport Mach-Zehnder interferometer 0 0 , 0 1 172 D4 171 output state projection synthesis cn(d)(T, ξ) = hnN2N3N4|Û |n1n2n3ni |φouti1 = N 2hN2| 3hN3| 4hN4|Û |n1i1|n2i2|n3i3|ψini4 = N where amplitudes depend on: projection synthesis transmittances T ≡ [t12, t22, t32, t42, t52] phase shifts ξ ≡ [ξ1, ξ2, ξ3, ξ4, ξ5] input Fock states |n1i, |n2i and |n3i3 measurements results N2, N3, N4 a^1 B1 a^ 2 B2 a^3 > a^ 4 B3 P6 ^ b1 n=0 d−1 X B5 b^ 3 > 173 174 cn(d)γn|ni b^ 4 1 photon 1 photon 1 photon |φ ^ b 2 B4 for exemplary input states and measurement outcomes |1> |1> |1> |ψ a { α|10iab + β|01iab { 175 176 α|Hi + β|V i interchanging polarization qubits and two-rail qubits α|Hi + β|V i b PBS – polarizing beam splitter (e.g. calcite crystal) transmits |Hi and reflects |V i (or vice versa) HWP(π/4) – half-wave plate (polarization rotator) changes photon polarization |V i ↔ |Hi. How to rotate polarization qubits? wave plate = waveplate = retarder a birefringent crystal (with a properly chosen thickness) changes polarization of a light by shifting its phase between two perpendicular polarization components. cos(2β) sin(2β) sin(2β) − cos(2β) half-wave plate (HWP) Ŝλ/2(β) = exp (i 3 π) cos(2β) − i sin(2β) √4 sin(2β) − cos(2β) − i 2 quarter-wave plate (QWP) Ŝλ/4(β) = cos(2β) sin(2β) sin(2β) − cos(2β) 0 1 1 0 ≡ σ̂X 1 0 0 −1 ≡ σ̂Z rotation of polarization in a waveplate 1 1 1 Ŝλ/2(π/8) = √ ≡ Ĥ 2 1 −1 Hadamard gate Ŝλ/2(0) = Pauli Z gate = phase flip Ŝλ/2(π/4) = NOT gate = Pauli X gate = bit flip Ŝλ/2(β) = special values β for HWP source: Wikipedia 178 177 V HWP 2 45-PBS c ψ in b d ψ out ψ out PBS ψ in 45-PBS HH + VV control a PBS control implementation of polarization CNOT gate β ≈ 61.50 for HWP2 HWP 1 β ≈ 150.50 for HWP1 ψ in PBS „0” implementation of polarization NS gate target target „V” PBS 180 179 ψ out â1 â2 â1 â2 b̂2 â1 â2 â1 â2 b̂1 b̂2 b̂1 ir t t ir − r t t r 50% ? 181 182 convention 2. symmetric vs asymmetric BSs convention 1. t ir S= ir t t r S= − r t BS ? beam-splitter and interference BS classical coin tossing 50% 0 1 BS but here classical intuition fails 0 1 (all BSs are 50-50) b̂1 b̂2 b̂1 b̂2 1 0 1 0 1 100% 0% 100% 0% 100% 0% 183 0 184 symmetric Mach-Zehnder interferometers so setup 3: so setup 2: as this is the |01i ≡ |1iL |0iL → |1iL. = iX̂ 1 1 1 1 −1 1 B1 = √ , B2 = √ 2 1 −1 2 1 1 0 −1 S = B2B1 = = −iŶ 1 0 |0iL → |0iL. i 1 0 1 =i 1 i 1 0 |0iL → i|1iL ∼ = |1iL 1 i 1 2 1 i 1 −1 1 √ B1 = B2 = 2 1 1 1 0 S = B2B1 = = Iˆ 0 1 S = B2B1 = 1 i 1 B1 = B2 = √ 2 1 i |0iL → S11|0iL + S21|1iL. √ N OT gate! setup 1: one gets |10i ≡ |0iL, |10i → S11|10i + S21|01i. Equivalently, in terms of two-rail qubit notation transforms We have already shown that 2x2 scattering matrix S11 S12 S= S21 S22 a single photon in Mach-Zehnder interferometer 186 185 0 1 0 1 0 1 θ θ θ θ θ θ θ sin 2 2 θ cos 2 2 θ cos 2 2 θ sin 2 2 θ 188 sin 2 2 θ cos 2 2 Mach-Zehnder interferometers with phase shifter 187 Pθ = 1 0 0 eiθ S = B2Pθ B1 + S21|1iL = f−|0iL + if+|1iL 189 1 where f± = (eiθ ± 1) 2 Mach-Zehnder interferometer with phase shifter total scattering matrix → |ψi = S11|0iL θ 2 – prob. of click in upper detector θ prob(0L) = |Lh0|ψi|2 = |f+|2 = cos2 2 θ 2 prob(1L) = |Lh1|ψi|2 = |f−|2 = sin2 1 −1 1 1 0 1 1 f−, −f+ = 0 eiθ 1 −1 f+, −f− 2 1 1 |0iL → f−|0iL + f+|1iL 190 – prob. of click in lower detector f+ 1 −1 1 1 0 −1 1 f+ f− = 0 eiθ 1 1 2 1 1 f− θ 2 i|0i + |1i i|0i + eiθ |1iL L L L √ √ → → Pθ → → B2 → f−|0iL+if+|1iL 2 2 |0iL 1 i 1 1 0 i 1 f− if+ S= = , 1 i if+ −f− 0 eiθ 2 1 i setup 1: so → B1 or explicitly |0iL thus prob(0L) = |Lh0|ψi|2 = |f−|2 = sin2 S= |0iL → |ψi = S11|0iL + S21|1iL = f+|0iL + f−|1iL S= prob(1L) = |Lh1|ψi|2 = |f+|2 = cos2 setup 2: so and setup 3: so and θ prob(0 ) = sin2 L 2 θ 2 prob(1L) = cos2 1 1 QIP with simple linear-optical systems 1. quantum key distribution (QKD) based on Mach-Zehnder interferometer using Bennett’s protocol B92 2. quantum-state engineering and teleportation of qubit and qutrit states using quantum scissors device Bob θB θ B = 00 ,900 Dlow Dup B92 protocol (Bennett 1992) Alice θA θ A = 00 ,900 ,1800 ,2700 191 192 Dup Dlow 0o 90o 0o 90o 0o 90o 0o 90o 1 2 1 2 1 − − 194 (n = 0, 1), 6. In case of full agreement of the tested photons, the secret key is formed by the remaining bits. Otherwise the protocol is repeated. Those photons are rejected from the key. 5. Alice and Bob publicly check their results for some photons. Alice and Bob keep only those deterministic results. which implies the deterministic detection of photons. θA − θB = n 180o 4. Alice publicly informs Bob, when their phase shifts fulfill the condition 3. Bob through a public channel informs Alice about his phase shifts without, of course, telling his measurement outcomes. 2. Bob, just before measuring each Alice’s photon, randomly applies phase shift 90o or 0o. 1. Alice sends photons each time randomly choosing one of 4 phase shifts (rotations): 0o, 90o 180o, 270o. 1 1 2 1 2 1 2 1 2 0 0 0 1 1 0 0 1 2 1 2 0 − − bit 1 1 θB Pup Plow QKD using B92 protocol 0o 0o 90o 90o 180o 180o 270o 270o θA probabilities of click in upper and lower detectors θ + θ θ + θ A B A B Pup = sin2 , Plow = cos2 2 2 quantum key distribution (QKD) using B92 protocol 193 D D 2 2 = a 0 + b 1 BS1 0 out D1 (1 photon) = BS2 a 0 Bob in (1 photon) Alice = a 0 + b 1 BS1 D1 (0 photon) 0 out = BS2 a 0 Bob phase flip in teleportation in (no photon) Alice teleportation of optical qubits using quantum scissors b 1 1 − b 1 196 1 + 195 + out D1(1 photon) + b2 2 BS1 (79:21) b1 1 ≅ b1 1 Bob + 197 1 b2 2 198 BS2 (79:21) b0 0 1 + teleportation of optical qutrits b0 0 Alice ≅ D (1 photon) 2 in quantum scissors device Fundamental Concepts • Entanglement • Non-locality • Conditional measurement Possible Applications • Qubit generation • Teleportation of superposition states • Quantum state engineering • Conversion of a classical state into a non-classical one • Cryptography quantum scissors Input: coherent state 2 0 +α1 1+ α B2 B2 ψ D2 1 b̂3 2 b̂ b̂ â1 φ 199 D2 D1 D1 200 n 2 ∞ α − α /2 n α =e ∑ n! n=0 2 2 α − α /2 2 + .... =e 0 +α 1 + 2 φ Φ = â 1 â 2 â3 B1 B1 â 2 n2 3 â Output: qubit state n3 ψ n2 n3 B1 â3 â2 â1 n2 B1 â2 θ θ B2 B2 r2 S = B2 Pθ B1 = t 2 0 1 0 1 0 0 B1 = 0 − r1 t1 , Pθ = 0 eiθ 0 0 0 t1 r1 θ ψ b̂3 b̂2 b̂1 â1 0 r2 0, B2 = t 2 0 1 B2 e r1r2 t1 iθ − eiθ r1t 2 0 0 1 202 eiθ t1t 2 − e iθ r2t1 r1 B1 : 1, 0, 0 , 0, r1, t1 , 0, t1, r1 B2 : r2, t2, 0 , t2, r2, 0 , 0, 0, 1 P: 1, 0, 0 , 0, Exp i theta , 0 , 0, 0, 1 S B2.P.B1 MatrixForm 0 − r2 t2 φ N 2 clicks N1 clicks N1 clicks 201 N 2 clicks scattering matrix for quantum scissors n3 n2 ψ B1 Mathematica: n3 â3 φ 203 h010|Û |010i = S22 = eiθ r1r2 = (S21S32 + S31S22)|011i + · · · iθ |φ10 01i ∼ e t1t2 γ0 |0i + r1 r2 γ1 |1i, 01 |φ01i ∼ −eiθ t1r2γ0|0i + r1t2γ1|1i, iθ |φ10 10i ∼ −e r1t2 γ0 |0i + t1 r4 γ1 |1i Analogously, we can show Other examples where x ∼ y means that x and y are equal up a renormalization constant. iθ |φ01 10i ∼ e r1 r2 γ0 |0i + t1 t2γ1 |1i h011|Û |110i = S21S32 + S31S22 = t2t1 + 0 · S22 = t1t2 thus we have so 204 3 X Sk1Sl2 † † √ âk âl |000i = (S21S32â†2â†3 + S31S22â†3â†2 + · · · )|000i Û |110i = 1!1! k,l=1 • |n10i → c|01ni for n = 1 so U |010i = (S12â†1 + S22â†2 + S32â†3)|000i = S12|100i + S22|010i + S32|001i after projective measurements of N1 and N2 photons in the respective modes. Example for |φ01 10i: • |n10i → c|01ni for n = 0 1 N2 |φi ≡ |φN n2n3 i =1 hN1 | 2 hN2 |Φi single-mode output state: |Φi three-mode output state (before projection): |ψ, n2, n3i where |ψi ∼ γ0|0i + γ1|1i input state: output states of the quantum scissors D 2 PPB quantum scissors a 0 + b 1 + c 2 + ... BS1 D1(1 photon) 0 BS2 205 206 1 out ∝ a 0 + b 1 Bob - truncation and teleportation of qubit states Alice = (no photon) in quantum state truncation to qubit states Note that the same solutions are for |ψi ∼ γ0|0i + γ1|1i + γ2|2i + · · · Why? Since, in this case, the single-mode output state |φi is always a superposition of |0i and |1i for any |ψi. conditions for perfect qubit-state truncation and teleportation 01 10 t1 = r2 and θ = 0 for |φ10 i and |φ01 i 01 10 t1 = t2 and θ = π for |φ01 i and |φ10 i optimized solutions correspond to the highest probability of successful truncation and teleportation. = t2 1 =√ 2 In the all 4 cases, the optimized solutions are for the 50:50 BSs: t1 = + b 1 + + 1 ! b 1 Bob + d 3 + ... 207 1 c 2 208 BS2 (79:21) ∝ a 0 c 2 BS1 (79:21) out D1 (1 photon) a 0 r1 t1 1 − 3(r1t1)2 + truncation and teleportation of qutrit states Alice D (1 photon) 2 in |φi ∼ γ0|0i + γ1|1i + γ2|2i quantum scissors for qutrit states general output state assuming n1 = n2 = N1 = N2 = 1 is 11 |φ11 i ∼ 2r1t1r2t2 e2iθ2 γ0|0i + γ2|2i + eiξ2 (r12 − t12)(r22 − t22)γ1|1i 01 as can be shown analogously to |φ10 i. 1±p truncation and teleportation to qutrit states 1 2 11 corresponds to |φ11 i assuming t22 = and θ2 = 0 or = π (show this!). √ √ 3) ≈ 0.21 or t12 = 61 (3 + 3)/6 ≈ 0.79 optimized 4 solutions for t12 = 16 (3 − and t22 = t12 if ξ4 = 0 or t22 = 1 − t12 if ξ4 = π. D2 BS1 0 in BS2 1 BS1 ’ D1’ 0 We cannot measure simultaneously q and p, thus we cannot measure directly the Wigner function W (q, p). 7. tomography of a nuclear spin I=3/2 212 problem this is the quantum tomography based on homodyne detection for reconstruction of Wigner function of optical fields optical homodyne tomography this is the tomography applied to quantum objects quantum tomography a method to reconstruct the shape of a hidden object from shadows (projections) at different angles tomography Can we reconstruct a hidden object (“real thing”) from its shadows? question [Πλατ ωνoς Πoλιτ ǫια - Plato’s Republic] We are like slaves in a dark cave watching only shadows on a wall. The shadows are projections of the “real things”. We think the shadows are real because we do not know better. Plato’s allegory 211 6. tomography of nuclear spins I=1/2 5. tomography of a single qudit 4. maximum-likelihood method 3. tomography of two qubits 2. tomography of a single qubit 1. optical-homodyne tomography of a single-mode field Outline How to reconstruct density matrix of a quantum state? 1 210 BS2 ’ Bob out Introduction to quantum tomography α D1 D2’ Alice generation and teleportation of qubit states 209 phase-space quasiprobability distributions in quantum mechanics uncertainty principle makes the concept of phase space in quantum mechanics problematic: 213 a particle cannot simultaneously have a well defined position and momentum, thus one cannot define a probability that a particle has a position q and a momentum p quasiprobability distribution functions = quasiprobabilities = quasidistributions functions which bear some resemblance to phase-space distribution functions useful not only as calculational tools examples: • Wigner (Wigner-Ville) W function Z ∞ i 1 W (q, p) = dxhq + x2 |ρ̂|q − x2 i exp px 2πh̄ −∞ h̄ – eigenstates of position operator q̂ What is the Wigner function? 214 • Cahill-Glauber s-parametrized W (s) function • Glauber-Sudarshan P function • Husimi (Husimi-Kano) Q function but can also reveal the connections between classical and quantum mechanics. where |q ± x i 2 • Wigner function can be negative • density matrix ρ can be calculated from Wigner function. Why is the Wigner function so important? Z ∞ ∞ −∞ Z −∞ W (q, p)dq W (q, p)dp marginal distributions of Wigner function correspond to classical distributions • position distribution pr(q) ≡ hq|ρ̂|qi = • momentum distribution pr(p) ≡ hp|ρ̂|pi = 10 formulations of quantum mechanics: 1. matrix formalism of Heisenberg (1925) 2. wave-function formalism of Schrödinger (1926) 3. density-matrix formalism of von Neumann (1927) 4. second-quantization formalism of Dirac, Jordan and Klein (1927) 5. variational formalism of Jordan and Klein (1927) 6. phase-space formalism of Wigner (1932) 7. path-integral formalism of Feynman (1948) 8. pilot-wave formalism of de Broglie and Bohm (1952) 9. many-worlds interpretation (MWI) of Everett (1957) W (α) = 2 exp[−2|α − α0|2] π simple examples of Wigner function coherent state |α0i: 10. action-angle quantum formalism of Hamilton and Jacobi (1983) ♣ Fock state |ni: where α = q + ip ♣ 2 W (α) = (−1)nLn(4|α|2) exp[−2|α|2] π W±(α = q + ip) = 215 216 f0 = 2 exp[−2q 2 − 2p2] f1 ± f0 cos(4pα0) π[1 + exp(−2α02)] Schrödinger cat states |ψ±i ∼ |α0i ± | − α0i: where Ln(x) is Laguerre polynomial ♣ where α0 ∈ R and f1 = exp[−2(q − α0)2 − 2p2] + exp[−2(q + α)2 − 2p2], 2 0 Imα 1 −1 −2 −3 −3 0 0.2 0.4 0.6 0.8 1 −2 −0.4 3 −0.2 0 0.5 −0.5 3 W 1 2 0 Reα 1 −1 −2 −3 −3 0 0.5 −2 −0.5 3 W 2 −1 0 Imα 1 Reα 0 −1 1 −2 2 −3 3 −3 −2 −0.8 3 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 −1 2 0 Reα 0 Imα 1 1 −1 2 −2 −3 3 −3 −2 −1 Reα 0 1 2 3 218 217 2 −1 0 Reα 1 Imα 0 −1 1 −2 2 −3 3 −3 −2 −1 3 −0.5 0 0.5 −1 2 0 Imα 0 Reα 1 1 −1 2 −2 −3 3 −3 −2 −1 Imα 0 1 2 3 Wigner function for Schrödinger cat states (a) |ψ+i, (b) |ψ−i and (c) mixture of coherent states ρ̂mix = |α0ihα0|+| − α0ih−α0| 0 3 0.1 0.2 0.3 W 0.4 0.5 0.6 0.7 W Wigner function for Fock states |0i, |1i, |2i W W 3 2 1 0 0.2 0.4 0.6 0.8 1 −1 −0.8 −0.6 −0.4 −0.2 0 3 −1 2 −2 1 −3 0 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 3 −1 2 −2 1 −3 pr(x) ≡ hx|ρ̂|xi = ∞ −∞ Z W (x, p)dp, 0 −1 pr(p) ≡ hp|ρ̂|pi = marginal distributions of Wigner function −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.6 0.4 1 0.8 1 0.8 ∞ −∞ Z −2 Wigner function for (a) |ψ+i, (b) |ψ−i and (c) ρ̂mix 220 W (x, p)dx −3 219 ♣ signal aˆ ' 2 I2 aˆ '1 detector I1 balanced homodyne detection detector aˆ s BS aˆ L ≅ α L laser (local oscillator) What is measured in homodyne detection? • intensity Ik (k = 1, 2) is proportional to the number of photons n̂k : I1 ∼ n̂1 = â1′†â1′ I2 ∼ n̂2 = â2′†â2′ ⇒ âL ≈ αL • parametric approximation for laser field: hn̂Li ≫ 1 • beam splitter (BS) is 50:50 (T = R = 1/2), thus it is called ‘balanced’ detection. • relations between input, âk , and output, âk′ , annihilation operators: 1 1 â1′ = √ (âS − âL) ≈ √ (âS − αL) 2 2 1 1 â2′ = √ (âS + âL) ≈ √ (âS + αL ) 2 2 221 222 • we measure difference of intensities, which is proportional to: 1 1 n̂2 − n̂1 = (âS† + αL∗ )(âS + αL) − (âS† − αL∗ )(âS − αL) 2 2 = αL∗ âS + αLâS† = |αL|(e−iθ âS + eiθ âS† ) = 2|αL|X̂(θ) where θ = arg(αL) and 1 X̂(θ) = (âS e−iθ + âS† eiθ ) 2 is the generalized quadrature operator X̂(0) = q̂ – momentum operator – position operator • special cases: X̂( π2 ) = p̂ Z −∞ ∞ W (q cos θ − p sin θ, q sin θ + p cos θ)dp 223 224 marginal distribution of Wigner function at angle θ pr(X, θ) = describes probability of measurement result of the quadrature X̂(θ) = q̂ cos θ − p̂ sin θ Remarks: • this is so-called Radon transformation • pr(X, θ) is directly measured in homodyne detection 1 (2π)2 Z ∞ −∞ Z 0 π Z −∞ ∞ pr(X, θ) exp[iξ(q cos θ +p sin θ −X)]|ξ|dXdθdξ Can we reconstruct Wigner function from its marginal distributions? W (q, p) = this is the inverse Radon transformation [K. Vogel and H. Risken, 1989] −∞ ∞ W (q cos θ − p sin θ, q sin θ + p cos θ)dp 226 225 3. first experimental reconstruction of single-photon state and its superposition with vacuum (qubit state) – A. Lvovsky and J. Mlynek (2002) 2. first experimental reconstruction of nonclassical state [i.e., squeezed vacuum] – D. Smithey, M. Raymer et al. (1993) 1. theoretical proposal – K. Vogel and H. Risken (1989) history e.g., in the first experiment, Smithey et al. performed 4000 measurements at 27 angles θ, so in total 108 000 measurements; in the second experiment they performed 160 000 measurements. number of measurements one repeats measurements on many copies i.e. identically prepared quantum objects Note: x ≡ q, y ≡ p, pr(X, 0) = hq|ρ̂|qi, pr(X, π2 ) = hp|ρ̂|pi pr(X, θ) = Z marginal distribution of Wigner function at angle θ (|Hi − i|V i) – right-circular polarization (|Hi + i|V i) – left-circular polarization √1 2 √1 2 |Ri = |Di+|Di √ , 2 √ − i|Di) = eiπ/4 |Di−i|Di , 2 = 1+i 2 (|Di |Ri+|Li √ 2 √ , |Di = eiπ/4 |Ri−i|Li 2 |Ri = |Hi = thus we readily have the inverse relations: 228 |Di−|Di √ 2 √ |Di = eiπ/4 |Li−i|Ri 2 √ |Li = eiπ/4 |Di−i|Di 2 √ = |V i = i |Ri−|Li 2 (|Hi + |V i) – linear-diagonal polarization at 1350 √1 2 |Di = |Li = (|Hi − |V i) – linear-diagonal polarization at 450 √1 2 |Di = |V i – vertical polarization |Hi – horizontal polarization single-qubit tomography by measuring Stokes parameters • notation: Optical-qubit tomography (part II) In principle, the same method can be applied, but there more effective methods, which will be described in the following. 4. How to reconstruct finite-dimensional states? hidden object ←→ Wigner function quantum “shadows” ←→ marginal distributions of Wigner function 3. By applying the inverse Radon transformation, we can reconstruct Wigner function and thus the density matrix. 2. But we can measure its marginal distributions pr(X, θ) at various angles θ 1. We cannot measure simultaneously q and p, thus we cannot measure directly Wigner function. Summary 227 |Li−i|ri hL|+ihR| √ √ 2 2 = 1 2 moreover hR|+hL| 1 √ √ |HihH| = |Ri+|Li = 2 |RihR| + |RihL| + |LihR| + |LihL| 2 2 hR|−hL| 1 √ √ = |V ihV | = |Ri−|Li 2 |RihR| − |RihL| − |LihR| + |LihL| 2 2 hR|+ihL| 1 √ √ |DihD| = |Ri−i|Li |RihR| + i|RihL| − i|LihR| + |LihL| = 2 2 2 |LihL| − i|RihL| + i|LihR| + |RihR| |DihD| = Pauli operators σ̂0 = |RihR| + |LihL| σ̂1 = |RihL| + |LihR| σ̂2 = i(|LihR| − |RihL|) σ̂3 = |RihR| − |LihL| four measurements of intensity, e.g.: (1) with a filter that transmits 50% of the incident radiation, regardless of its polarization: N N n0 = (hH|ρ̂|Hi + hV |ρ̂|V i) = (hR|ρ̂|Ri + hL|ρ̂|Li) 2 2 (2) with a polarizer that transmits only photons in state |Hi: n1 = N hH|ρ̂|Hi (3) with a polarizer that transmits only photons in state |Di: n2 = N hD|ρ̂|Di (4) with a polarizer that transmits only photons in state |Ri: n3 = N hR|ρ̂|Ri ≡ ≡ ≡ ≡ 2n0 2 (n1 − n0) 2 (n2 − n0) 2 (n3 − n0) Stokes parameters via photon counts S0 S1 S2 S3 229 230 3 1 X Si σ̂i, 2 i=0 S0 where 1 1 + hσ̂3i hσ̂1i − ihσ̂2i 2 hσ̂1i + ihσ̂2i 1 − hσ̂3i Si = N hσ̂ii = N Tr{σ̂iρ̂} ρ̂ = 231 Single-qubit density matrix via Stokes parameters ρ̂ = Why do we need S0? for renormalization of the count statistics to compensate experimental inefficiencies (e.g. of detectors) 232 single polarization-qubit measurement operators there are infinitely many sets of such projectors µ̂0′′ = µ̂0′ , µ̂0′ = = µ̂1′ = ∼ µ̂2′ = ∼ = µ̂3′ = ∼ µ̂1′′ = µ̂1′ , µ̂2′′ = |DihD|, µ̂3′′ = µ̂3′ |HihH| + |V ihV | |LihL| + |RihR| |HihH| |RihR| + |RihL| + |LihR| + |LihL| |DihD| |LihL| − i|RihL| + i|LihR| + |RihR| |HihH| − |HihV | − |V ihH| + |V ihV | |RihR| |HihH| − i|HihV | + i|V ihH| + |V ihV | the common ones are e.g. or µ̂0 µ̂1 µ̂2 µ̂3 = = = = N σ̂0 N σ̂1 N σ̂2 N σ̂3 = µ̂′0 = µ̂′1 − µ̂′0 = µ̂′2 − µ̂′0 = µ̂′3 − µ̂′0 (k = 0, ..., 3) µ̂00 = N σ̂0 ⊗ σ̂0, µ̂01 = N σ̂0 ⊗ σ̂1, ··· µ̂33 = N σ̂3 ⊗ σ̂3 µ^ 1 µ^ 2 µ^ 3 µ^ 0 µ^ 1 µ^ 2 µ^ 3 µ^ 0 µ^ 1 µ^ 0 µ^ 1 µ^ 2 µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 3 µ^ 3 234 233 ................................................................................................................................................................................................................................................ 3 qubits – 64 measurement operators: µ̂i ⊗ µ̂j ⊗ µ̂k 2 qubits – 16 measurement operators: µ̂i ⊗ µ̂j n qubits – 4n measurement operators: µ̂i1 ⊗ µ̂i2 ⊗ · · · ⊗ µ̂in ··· µ^ 2 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 2 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 1 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 1µ^ 2µ^ 3 µ^ 0 µ^ 0 1 qubit – 4 measurement operators: µ̂i •• • 3 qubits 2 qubits 1 qubit measurement operators required for tomography standard 16 projectors: two-qubit measurement operators N = 1 ⇒ µ̂k = σ̂k assuming loss-free detection (perfect detectors and wave plates) or Q2 2 H2 P D2 CCL H1, H2 – HWPs output 235 cos(2β) sin(2β) sin(2β) − cos(2β) cos(2β) − i sin(2β) sin(2β) − cos(2β) − i ′ Ŝλ/2 (β) = Ŝλ/2(−β), ′ Ŝλ/4 (β) = −Ŝλ/4(−β) slightly modified defs. Ŝλ/4(β) ∼ √1 2 quarter-wave plate (QWP) Ŝλ/2(0) = σ̂Z Ŝλ/2(22.50) = Ĥ Ŝλ/2(450) = σ̂X special cases of HWPs Ŝλ/2(β) = half-wave plate (HWP) wave plates - a reminder CCL – coincidence counter and logic 236 P1, P2 – polarizers transmits only photons of one polarization (say, vertical) Q1, Q2 – QWPs, black box – a source of photon pairs in an unknown state black box D1 P 1 Q1 H1 tomography of a polarization state of two photons |Hi − i|Vi √ 2 237 238 • rotation matrix 4 × 4 of 2 beams by 2 HWPs and 2 QWPs rotation[45, 0, 45, 0]; rotation[45, 0, 0, 0]; rotation[0, 0, 0, 0]; rotation[0, 0, 45, 0]; rotation[45/2, 0, 45, 0]; rotation[45/2, 0, 0, 0]; rotation[45/2, 45, 0, 0]; rotation[45/2, 45, 45, 0]; rotation[45/2, 45, 45/2, 0]; rotation[45/2, 45, 45/2, 45]; rotation[45/2, 0, 45/2, 45]; rotation[45, 0, 45/2, 45]; rotation[0, 0, 45/2, 45]; rotation[0, 0, 45/2, 90]; rotation[45, 0, 45/2, 90]; rotation[45/2, 0, 45/2, 90] 239 240 rotation[h1_, q1_, h2_, q2_] := Kron[HWP[h1 Degree].QWP[q1 Degree], HWP[h2 Degree].QWP[q2 Degree]] := := := := := := := := := := := := := := := := • our sequence of rotations R[1] R[2] R[3] R[4] R[5] R[6] R[7] R[8] R[9] R[10] R[11] R[12] R[13] R[14] R[15] R[16] • rotated ρ̂(n) = R̂(n)ρ̂(R̂(n))† for the nth measurement RhoRotated[n_] := R[n].rho .hc[R[n]] • examples MF 2,0 3,0 1,0 0,0 2,1 3,1 0,1 1,1 # % % % % 3,1 2,1 " |Ri ≡ 3,2 2,2 0,1 1,1 MF 2,0 3,0 0,0 1,0 ! ! |Hi + i|Vi √ , 2 RhoRotated 1 2,3 0,2 1,2 " " . 3,3 1,3 0,3 0,3 3,3 2,3 RhoRotated 2 2,2 3,2 0,2 ! ! |Li ≡ 1,3 tomography of polarization state of two photons No. Mode 1 Mode 2 HWP 1 QWP 1 HWP 2 QWP 2 1 |Hi |Hi 45o 0 45o 0 2 |Hi |Vi 45o 0 0 0 3 |Vi |Vi 0 0 0 0 4 |Vi |Hi 0 0 45o 0 5 |Ri |Hi 22.5o 0 45o 0 6 |Ri |Vi 22.5o 0 0 0 7 |Di |Vi 22.5o 45o 0 0 8 |Di |Hi 22.5o 45o 45o 0 9 |Di |Ri 22.5o 45o 22.5o 0 10 |Di |Di 22.5o 45o 22.5o 45o 11 |Ri |Di 22.5o 0 22.5o 45o 12 |Hi |Di 45o 0 22.5o 45o 13 |Vi |Di 0 0 22.5o 45o 14 |Vi |Li 0 0 22.5o 90o 15 |Hi |Li 45o 0 22.5o 90o 16 |Ri |Li 22.5o 0 22.5o 90o |Hi − |Vi √ , |Di = 2 Sin 2 t Cos 2 t " Mathematica program for the tomography scheme % % % % % % % % % $ Sin 2 t Cos 2 t 0,3 1,3 ; 1,2 " " • def. of Kronecker tensor product (Kron) and Hermitian conjugate (hc) Cos 2 t Sin 2 t 1,2 2,3 0,2 2,2 << LinearAlgebra‘MatrixManipulation‘ f[x_, y_] := x*y; Kron[matrix1_, matrix2_]:= BlockMatrix[Outer[f, matrix1, matrix2]] hc[x_] := Transpose[Conjugate[x]] Cos 2 t Sin 2 t 1 2 0,1 1,1 2,1 • def. of rotation by HWP and QWP and def. of [ρnm]4×4 ! ! " " " " 3,3 ! ! 3,2 3,1 HWP t_ : 0,0 1,0 # % % % % % % % % % % % % $ 2,0 3,0 ! ! ! ! " " " QWP t_ : rho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psi[1] psi[2] psi[3] psi[4] psi[5] psi[6] psi[7] psi[8] psi[9] psi[10] psi[11] psi[12] psi[13] psi[14] psi[15] psi[16] |ψ1i = |HHi, ..., |ψ16i = |RLi Kron[ketH, ketH]; Kron[ketH, ketV]; Kron[ketV, ketV]; Kron[ketV, ketH]; Kron[ketR, ketH]; Kron[ketR, ketV]; Kron[ketDbar, ketV]; Kron[ketDbar, ketH]; Kron[ketDbar, ketR]; Kron[ketDbar, ketDbar]; Kron[ketR, ketDbar]; Kron[ketH, ketDbar]; Kron[ketV, ketDbar]; Kron[ketV, ketL]; Kron[ketH, ketL]; Kron[ketR, ketL]; |ψ2i = |HVi, := := := := := := := := := := := := := := := := = (projection) measurement states = projectors our choice of tomographic states where |Hi=ketH, hH|=braH, etc. Z P 14 I Z 3,3 Z 3,2 Z c 2,3 c c 2,2 ^ 1 2 ^ P 13 ^ 1,1 ^ 1,0 ^ 0,1 ^ 0,0 ^ 1 2 ^ P 12 d 3,3 ^ d 3,2 ^ d 3,1 ^ ` 1 2 _` 3,3 ^ ` 3,2 ^ 3,0 _` 3,1 ^ 2,3 ` 3,0 ^ 2,2 ^ 2,3 ^ 2,1 ^ 2,2 ^ 2,0 ^ 2,1 ^ 1,3 ^ 2,0 ^ 1,2 ] 1,3 ` 1,1 ^ 1,2 ^ 1,1 ` 1,0 ` 1,0 ^ 0,3 ^ 0,3 A 3,3 ^ 0,2 3,2 Z ^ 0,2 3,1 ^ ^^ 0,1 3,0 [ b 0,1 2,3 ^ ^ 0,0 2,2 ^ ^ 0,0 2,1 ^ ^ 1 4 2,0 ^ ^ P 11 1,3 ^ ^^ 1,2 ^ ^ 1,1 ^ ^ 1,0 ^ ^ 0,3 ^ Z 0,2 ^ Z 0,1 ^ c 0,0 ^ Z 1 4 ^ c P 10 Z d P 9 ^ c 2,2 ^ d 2,0 ^ ^ 0,2 ^ ^ 0,0 ^^ ^ 1 4 ^ Z P 8 ^ ^ 3,3 ^ ^ 3,1 ^ ^ 1,3 ^ ^ 1,1 [ ^ 1 2 \ ^ P 7 Z ^ 3,3 ^ 3,1 ^ 1,3 ^ 1,1 ^ 1 2 ^ P 6 ^ 2,2 ^ 2,0 ` 0,2 ^ _` 0,0 ^ ` 1 2 ` 2,2 ^ 1 2 ^ P 5 ^ P 4 ^ 3,3 ^ P 3 ^ 1,1 ^ ^ P 2 ^ ^ , n, 1, 16 a ", P n ^ ketV ketV ^ 0,0 ^ d 2 2 hc ketH ; braV hc ketV ; hc ketD ; braDbar hc ketDbar ; hc ketR ; braL hc ketL ; ^ Do Print "P ", n, " ^ ^ P 1 \ _` all1 E 242 ^ ketH ^ b braH braD braR ] ` ; ketR : ` 2 ^ ketH ketH ^ ketV 1 0 ; ketV : ; 0 1 ketH ketV ; ketDbar : 2 ^ ketL : ketD : ketH : ^ 2,2 3,2 2,2 ^ 2,0 C ^ 0,2 B ^ 0,0 A ^ 1 2 C ^ P 5 B ^^ 1,1 C ^ P 2 B ^ 0,0 3,0 3,2 2,2 a FS A 0 1 B ^ ketVV : Kron ketV, ketV ; braVV : hc ketVV P n_ : braVV.RhoRotated n .ketVV P 1 A 2,0 D ` 0 ; 1 1,2 A ^ ketV : A 0,2 3,0 2,0 H ^ so 1,2 0,2 H (n) D 3,1 1,0 0,0 E 0,3 1,3 1 2 1 2 1 2 1 2 E H 0,1 1,1 D 3,2 3,3 2,3 D H 2,2 E 3,0 3,1 2,1 @ H 2,0 1,3 0,3 A 1,2 D 0,1 1,1 A H 0,2 1 2 1 2 1 2 1 2 @ 1,0 3,2 C H 0,0 A 2,2 A H 2,3 3,0 A H 3,3 A 2,0 E 2,1 1,2 H 3,1 A 0,2 H 0,3 @ H 1,3 1,0 @ 0,0 @ H 0,1 1 2 1 2 1 2 1 2 @ 2,3 3,3 @ 2,1 @ 0,3 MF 3,1 H 1,1 FS 1,3 F 0,1 1,1 equivalent approach using tomographic states Z measured probabilities Pn = hV V |ρ̂(n)|V V i = ρ33 where |V i = 1 2 1 2 1 2 1 2 ; RhoRotated 5 241 ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^^ E 8 8 8 7 7 8 8 7 / 6 / 6 7 / 6 / 4 8 / 4 / 4 / / 6 5 / 6 5 3 / / 6 / / 5 4 5 / / 6 / / 4 6 / 6 5 / / 6 5 / 4 / 4 / 6 / 4 5 / / 5 / 6 5 / / 4 / / 4 4 / 6 4 4 / 6 5 / / / 6 4 X 6 / 6 4 8 7 8 7 5 / / 6 / 6 4 7 5 / 4 5 / / 5 / 4 / 5 / 4 / 6 / 6 0 2 2 2 2 2 2 SS - 2 244 ; ; 243 p1 = hHH|ρ̂|HHi, p2 = hHV |ρ̂|HV i , etc.: Kron[braH, braH] .rho.Kron[ketH, ketH]; Kron[braH, braV] .rho.Kron[ketH, ketV]; Kron[braV, braV] .rho.Kron[ketV, ketV]; Kron[braV, braH] .rho.Kron[ketV, ketH]; Kron[braR, braH] .rho.Kron[ketR, ketH]; Kron[braR, braV] .rho.Kron[ketR, ketV]; Kron[braDbar, braV] .rho.Kron[ketDbar, ketV]; Kron[braDbar, braH] .rho.Kron[ketDbar, ketH]; Kron[braDbar, braR] .rho.Kron[ketDbar, ketR]; Kron[braDbar, braDbar].rho.Kron[ketDbar, ketDbar]; Kron[braR, braDbar] .rho.Kron[ketR, ketDbar]; Kron[braH, braDbar] .rho.Kron[ketH, ketDbar]; Kron[braV, braDbar] .rho.Kron[ketV, ketDbar]; Kron[braV, braL] .rho.Kron[ketV, ketL]; Kron[braH, braL] .rho.Kron[ketH, ketL]; Kron[braR, braL] .rho.Kron[ketR, ketL]; 245 p p p q q q 0 0 0 0 0 q q p p 0 0 0 0 0 0 0 0 0 q q f f g g e h FS , n, 1, 16 j p n g i i P (n) = p(n) g g k g possible methods: • method 1: ρ1 ρ5 ρ̂ = ρ9 ρ13 ρ3 ρ7 ρ11 ρ15 ρ4 ρ1 ρ00 ρ8 ρ.2 ≡ ρ.01 7→ . ρ12 . ρ16 ρ16 ρ33 • another labelling: x x + ix x + ix x + ix x1 1 2 11 3 12 4 13 x x + ix x + ix 5 6 14 7 15 x2 − ix11 x2 ρ̂ = 7→ .. x x x 3 − ix12 x6 − ix14 8 9 + ix16 x16 x4 − ix13 x7 − ix15 x9 − ix16 x10 ρ2 ρ6 ρ10 ρ14 247 248 represent operators of Hilbert space as superoperators in Liouville space convert 4x4 matrix into 16-dim column vector f ", P n f p ", n, " f h both methods give the same probabilities p p Do Print "P ", n, " p p test p p p 1 p 2 p 3 p 4 p 5 p 6 p 7 p 8 p 9 p p P 1 P 2 P 3 P 4 P 5 P 6 P 7 P 8 P 9 q q 0 q q 0 q q p 10 p 11 p 12 p 13 p 14 p 15 q q f p 16 q q P 10 P 11 P 12 P 13 P 14 P 15 P 16 q q p p reconstruction analysis q q e projection measurements := := := := := := := := := := := := := := := := 246 q q p p o enable determination of the probabilities p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15] p[16] all2 := Do[Print[n, " -> ", p[n] // FS], {n, 1, 16}] projection measurements hc[psi[1]] .rho.psi[1] ; hc[psi[2]] .rho.psi[2] ; hc[psi[3]] .rho.psi[3] ; hc[psi[4]] .rho.psi[4] ; hc[psi[5]] .rho.psi[5] ; hc[psi[6]] .rho.psi[6] ; hc[psi[7]] .rho.psi[7] ; hc[psi[8]] .rho.psi[8] ; hc[psi[9]] .rho.psi[9] ; hc[psi[10]].rho.psi[10]; hc[psi[11]].rho.psi[11]; hc[psi[12]].rho.psi[12]; hc[psi[13]].rho.psi[13]; hc[psi[14]].rho.psi[14]; hc[psi[15]].rho.psi[15]; hc[psi[16]].rho.psi[16]; enable determination of the probabilities := := := := := := := := := := := := := := := := p1 = hψ1|ρ̂|ψ1i = hHH|ρ̂|HHi, p2 = hψ2|ρ̂|ψ2i = hHV |ρ̂|HV i, etc.: p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15] p[16] all2 := Do[Print[n, " -> ", p[n] // FS], {n, 1, 16}] q p o q p q p p q p p q p p q p p q p p q p p l n m m m o o m m o o m m o o m m o o m n l l l l m m n n l l l l m m l n n m l l l l n m m l n n l l m l q p m m o o m m o o m m o o m l l l l l l l n m l l n m m l n n m l l n m m n n m l l ¡ ¡ ¡ ¡ ¡ ¡ ¡ 0 0 0 0 0 1 ¡ 0 ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ r s r r s r s s w z z z z z z z z z z z z z z z z z z z z z z { z z z z z z z z z z z z z t z z z z z z z z z z z z z z | {{ z {{ | { { { {{ | { { { { { { { { {{ { z z z z z z z z z z z 1 4 { { 0 {{ {{ z z z z z 1 4 0 {{ 0 v 0 {{ | { z z z z 0 0 z z { {{ | z z { { { {{ { {{ { { { { z z z z | z z { {{ z z z 1 2 1 2 1 2 1 0.487213 0.00454278 0.502019 0.00622529 0.228877 0.245661 0.188455 0.236968 0.251423 0.449062 0.212165 0.241693 0.18467 0.240739 0.234458 0.470907 0 0 0 0 0 0 0 0 1 2 0 0 {{ 0 { 0 {{ 0 { 1 {{ 0 0 0 0 { 0 | 0 s 0 0 { 1 2 { { 1 2 | {{ 0 | 0 { { 0 { { 0 { { 0 { { 1 2 1 4 {{ {{ {{ {{ {{ {{ {{ {{ {{ { { { { { { { { { { { { { { { { 0 { { 0 {{ 0 { { { 0 { { 0 {{ { 1 2 1 2 { { 1 { 1 2 1 2 | 0 v 0 | 0 {{ { { 0 {{ { 0 { { 0 { 0 { 0 | 0 { 0 { { 0 { 0 u { 0 {{ 0 {{ 0 {{ 0 0 1 2 0 1 2 0 { 0 { 0 1 2 0 0 { 1 2 0 0 {{ 0 0 {{ 0 1 {{ 1 2 1 2 0 1 2 0 1 2 0 0 1 2 1 2 1 2 0 {{ 1 2 1 4 { 0 0 { 0 {{ 1 2 1 2 | 1 2 1 4 1 4 1 4 0 0 1 2 1 2 0 0 0 1 4 1 4 1 4 1 2 0 0 { { 0 1 2 1 2 {{ 0 { { 0 0 0 1 2 1 4 1 4 1 4 0 0 Nexp : Transpose 34749, 324, 35805, 444, 16324, 17521, 13441, 16901, 17932, 32028, 15132, 17238, 13171, 17170, 16722, 33586 Norm Sum Nexp i , i, 1, 4 1 Pexp 1. Nexp Norm; MF Pexp ν=1 0 r 0 1 u 0 0 1 2 1 4 1 4 1 4 1 2 { 0 1 0 0 0 0 0 0 0 0 1 2 1 2 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 1 2 0 0 0 0 0 0 0 0 0 0 1 where nν = N hψν |ρ̂|ψν i so 4 X nν = N hHH|ρ̂|HHi+N hHV |ρ̂|HV i+N hV H|ρ̂|V Hi+N hV V |ρ̂|V V i = N 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 2 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 t 0 0 0 0 0 } 1 2 r 0 0 0 0 0 r 0 0 0 0 0 252 251 experimental estimations of probabilities Pexp = {Pexp1, ..., Pexp16} = { nN1 , ..., nN16 } £ 1 0 0 0 0 0 ¤ a n_, m_ : Coefficient P n , x m A : Table a n, m 1, 1 , n, 1, 16 , m, 1, 16 A MF 0 0 ¤ experimental coincidence counts Nexp = {n1, ..., n16} ¤ where P = [P1, P2, · · · , P16]T , x = [x1, x2, · · · , x16]T , A = [am,n]16×16 P P1 = x1 = 1 · x1 + 0 · x2 + · · · + 0 · x16 = [1, 0, · · · , 0] · x = n a1,nxn, etc. ¤ experimental data [James et al. 2002] 0 ¤ 0 0 ¤ 0 0 ¤ 0 0 ¤ 0 0 1 2 1 2 0 0 ¤ 0 ¤ 0 0 0 0 0 0 1 ¤ 1 ¡ ¤ 0 1 2 1 2 0 0 0 0 ¤ 0 ¤ 1 2 ¤ 0 ¤ 0 ¡ 1 1 2 1 2 0 0 1 2 1 2 1 0 1 0 0 0 1 2 1 2 1 2 1 2 1 0 0 0 0 0 ¡ ¤ 0 1 2 1 2 0 0 0 0 0 0 0 0 0 0 ¤ 0 1 2 1 2 1 0 0 1 0 ¤ 1 2 0 0 0 0 0 0 0 0 0 1 ¤ 1 2 1 2 1 2 0 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 0 ¤ 1 2 1 2 0 0 0 0 0 0 1 2 0 0 1 0 0 0 ¤ 1 2 0 0 1 2 0 ¤ set of equations P = Ax 250 1 2 1 2 ¤ 2 x 16 1 0 0 1 2 0 1 2 0 ¡ ¤ 2 x 15 0 1 2 1 ¤ 2 x 12 0 ¤ 2 x 11 x 15 ¤ x 10 x 14 ¤ x 8 ¤ 2x 6 ¤ x 5 ¤ 2x 4 ¡ ¤ x 1 1 2 1 2 1 2 0 0 0 0 ¤ 2 x 11 0 0 1 2 0 0 0 0 ¤ x 5 0 0 ¡ 0 1 1 2 1 2 0 0 0 0 ¤ x 1 x 13 0 ¤ 2 x 16 0 ¤ x 10 0 ¤ x 8 0 ¤ x 10 ¤ 2x 9 ¤ x 5 ¤ 2x 2 ¤ 2 x 12 ¤ x 10 ¤ 2x 9 ¤ x 8 ¤ x 5 ¤ 2x 2 x 10 2x 9 x 8 2x 7 2x 6 x 5 2x 4 ¡ 2x 3 0 ¡ 0 0 1 2 ¡ 1 ¤ 1 4 1 2 0 ¤ 1 2 1 2 ¡ 0 0 ¤ 1 2 ¤ x 8 ¤ 2x 2 0 1 2 0 0 ¡ ¤ x 1 0 1 0 0 0 1 2 1 2 0 0 ¤ x 1 0 1 2 0 0 1 0 0 1 2 0 0 0 0 0 ¡ ¤ x 1 0 1 2 0 1 0 0 0 0 ¡ x 16 0 0 0 0 x 14 1 2 0 0 0 x 13 0 0 0 0 2 x 11 ¤ P 16 ¡ P 15 ¤ 1 2 ¤ x 10 ¤ x 8 ¤ 2x 7 ¤ P 14 ¤ P 13 1 2 1 2 1 2 0 0 ¡ ¤ x 5 ¤ 1 2 1 0 0 0 0 0 1 2 0 ¤ 2x 3 1 2 1 2 1 2 1 ¤ 1 4 ¤ x 1 ¤ x 8 invA : Inverse A invA MF 0 0 x = A−1P where A−1 is equal to exists if A is non-singular and it is simply given by solution of the set of equations ¤ P 12 ¤ P 11 ¤ P 10 ¤ 1 4 ¤ 2x 3 , n, 1, 16 ¤ x 10 1, 1 ¡ ¤ 1 4 ¤ x 1 ¤ 1 2 ¤ 2x 7 ¤ x 5 x 15 x 10 x 12 x 5 x 8 ", P n ¤ 1 2 x 1 1 2 ¤ P 9 1 2 ¤ P 8 x 8 ¤ P 7 x 10 ¤ P 6 P 5 P 4 x 5 P 3 x 1 P 2 P 1 Do Print "P ", n, " In[81]:= all1 249 ¢ z z x z ~ {{ {{ {{ { { { | { {{ { {{ { { { { {{ | { {{ | { { {{ {{ {{ { { s s | {{ { {{ r { | s {{ {{ { {{ { { | {{ w | { { {{ y { { {{ § § § 5 ¦ ¦ ¦ ¦ § § § ¦ § 3 6 9 x ¦ ¦ § ¹ ¹ ¹ ¦ ¦ x x § ¦ ¦ ¦ ¦ ¦ 12 14 16 § § x x x 4 7 9 x x x 10 § © ¦ ¦ ¦ ¦ ¦ 13 15 16 ¸ ± ³ ³ ³ Tr {γ̂ν · γ̂µ} = δν,µ 16 X ¦ 16 X ¹ 0.519209 0.0380247 0.0648257 0.00762037 0.0694526 0.013383 0.502019 ¦ ® § x ¦ © γ̂ν rν , 253 254 1 0 ; 0 1 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 1 0 −i 0 0 0 0 0 −1 1 ¾ 0 0 0 0 1 2 0 0 0 0 2 ¾ 0 0 À 4m 0 1 ; 1 0 Do MF MF Å Æ ½ 2 n ¾ γ̂6 1 2 0 0 = 0 i 1 2 −i 0 0 0 0 0 −i 0 0 0 1 0 −i 0 0 0 −i 0 0 0 , γ̂ = 1 10 2 0 0 0 −1 1 0 , 0 0 0 i γ̂2 = 0 0 2 0 ¿ 0 0 0 i , γ̂ = 1 14 2 −1 0 0 0 ¿ ¾ 0 0 0 i 0 i 0 0 0 1 0 0 Á 1 0  2 0 À ; Kron ½ 3 ¾ m, ½ ½ n ¿ Ç 1 0 Á 0 1 À −i 0 , 0 0 γ̂7 1 2 0 0 0 −1 0 −1 0 0 0 −1 0 0 0 0 0 −i 0 0 = 1 0 1 2 0 1 0 0 , γ̂ = 1 15 2 i 0 0 0 0 0 1 0 1 0 0 0 ; Ê 0 0 , 0 1 γ̂8 1 2 Ç 255 1 0 0 0 −i 0 0 0 0 0 1 0 0 0 0 i 0 1 0 0 0 0 −1 0 1 0 = 0 0 1 2 0 0 0 1 256 0 0 = i 0 γ̂16 0 1 0 0 0 0 γ̂4 = 1 0 1 2 Ë 0 1 i 0 , γ̂ = 1 12 2 0 0 0 0 0 0 , 0 −1 0 0 −1 0 −i 0 0 0 0 −1 , 0 0 , m, 0, 3 , n, 0, 3 1 0 γ̂3 = 0 0 Å Â 0 0 , −i 0 È −1 0 0 0 , γ̂ = 1 11 2 0 i 0 0 0 0 0 −i É ÈÈ ½ exemplary generators of the Lie algebra SU (2) ⊗ SU (2) 1 2 1 2 1 0 0 0 0 0 i 0 0 0 = 0 1 1 2 0 1 = 0 0 1 2 Ê È 0 0 Ç Simplify Å 0 0 , 1 0 É È ½ GenerateGamma ¾ 0 1 2 16 : 1 0 0 Ñ Ä generators of the Lie algebra SU (2) ⊗ SU (2) Ð Ä Ã ¿ Ç Ì Ì Ä Simplify 1 2 Ó Ó Ó Ó Ó Ó ÐÐ 0 Õ 0 ÐÐ Ð 0 ÐÐ 2 Ð 0 Ð 0 Ð Ó Ó Ó Ó Ó Ó 1 2 0 0 2 0 2 0 0 Ð Å 0 1 γ̂1 = 0 0 ÐÐ Ó Ó Ó Ó Ó Ò Ð Ì Ì Ì Ì Ð Ì Ç Ì Ç Í Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Pexp 8 x ¸ ¦ Ä Ä x = A−1 11 x x x ¹ ν=1 Ô © 14 15 § ® § ® γ̂ν Tr {γ̂ν · ρ̂} ≡ Ñ x ¦ ® x x x § ν=1 ÐÐ 2 6 7 ¸ À Ð § § § ¯ ¯ ¯ § § § ¦ ¦ ¦ ¦ ¦ ¦ § § § § § § § ³ ³ ³ x x x © ÐÐ ÐÐ § ¦ ¦ ¦ ¦ ¦ ¦ ® § ® § § Å Å reconstructed ρ̂ by linear tomography § ¾ Ð Ð final solution for x assuming the analyzed experimental data: ¦ Ï Î x1 x1 x2 + ix11 x3 + ix12 x4 + ix13 x5 x6 + ix14 x7 + ix15 x2 x2 − ix11 . 7→ ρ̂ ≡ ρ̂exp = . x x x 3 − ix12 x6 − ix14 8 9 + ix16 x16 x4 − ix13 x7 − ix15 x9 − ix16 x10 11 12 13 ¸ ∀ρ̂ ρ̂ = Õ Flatten invA.Pexp ¦ x 1 ¦ § x x x ¦ x ¦ 0.487213, 0.00418524, 0.00975155, 0.519209, 0.00454278, 0.0271305, 0.0648257, 0.00622529, 0.0694526, 0.502019, 0.01142, 0.0178416, 0.0380247, 0.0145958, 0.00762037, 0.013383 rhoexp : 2 3 4 ¦ Ð Ð Ó Ó Ó Ó Ó Ó Ó Ó ¦ ° are generalized Pauli operators or their products chosen as, Õ x x x MF ¹ ¦ § § § 0.00975155 0.0178416 0.0271305 0.0145958 0.00622529 0.0694526 0.013383 ¹ ¯ ¯ ¯ ³ ³ ³ ³ ² Í 0 0 = 0 i γ̂5 Ð ¦ « ¥ ª e.g., generators of the Lie algebra SU (4) or just SU (2)⊗SU (2). Ð ¦ ¦ ¯ ¯ ¯ (m, n = 0, ..., 3) Ô ¦ ¦ § ° § ° § § § § ¦ ¦ ¦ ¦ ¦ ¦ ¯ ¯ ¯ § ° § ° § § § § § ° § Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ó Ó Ó Ó § § § 0.00418524 0.01142 0.00454278 0.0271305 0.0145958 0.0648257 0.00762037 ¹ § 1 γ̂4m+n = σ̂m ⊗ σ̂n, 2 γ̂9 Õ rhoexp © ¦ ¬ where σ̂n are the standard Pauli operators. ÐÐ ´ 0.487213 0.00418524 0.01142 0.00975155 0.0178416 0.519209 0.0380247 © © ´ • method 2: ρ̂ 7→ [rν ]16×1 © © Ï ¦ ¸ ¥ find a set of 16 linearly independent 4 × 4 matrices {γ̂ν } satisfying: ¸ ¨ for convention we denote γ̂16 = γ̂0 γ̂13 Ð Ó Ó Ó Ó Ó Ò ¹ ¹ Ï Ï Ï Ï Ï º ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ » Ï Î § ¹ ¹ µ · · · · · · · · · ¶ 0 1 , 0 0 0 −i , 0 0 0 0 , 0 −1 0 0 . 0 1 × Ö × Ö Ö × × Ö Ö × Ù Ø Ö Û Ú × Ü ß ß ß ß ß ß ß ß ß ß ß á ß àà àà àà à à à à ß ß ß ß ß à ß à à à à à ß ß ß ß ß ß ß ß ß ß ß ß á ß àà á àà ß ß ß ä × ß ä à ß ä á ß ä àà àà àà à à à à ß ä ß ä ß ä ß ä ß ä ß ä àà ß ä ß ä ß ä ß ä à Ø à ß ä ß ä àà ß ä ß ä ß ä ß ä à ß ä ß ä ß ä àà à à à à à à à à á á á á ß ä ß ä ß ä ß ä à ß ä á ß ä àà á àà ß ä ß ä ß ä ß ä à ß ä á ß ä àà ß ä ß ä ß ä ß ä ß ä ß ä á ß ä àà á à à à à à à à àà à ß ä ß ä ß ä à ß ä ß ä ß ä àà àà à à à á à ß ß 1 2 0 1 2 1 2 0 0 0 1 ò 2 2 2 0 0 1 0 0 M 16 î ä ß î ä ß î ä ß î ä ß î ä ä à à ß ß î ä ß 0 ä 0 ä î 0 1 0 0 MF 1 2 1 2 MF 0 0 2 1 1 0 0 0 2 2 2 õ î ä à ß 0 ä î 0 ä î 0 ä í õ à 0 à à ä î 0 á ä î 0 àà à ä î õ 1 2 àà à ä î 0 × ä î 0 àà ä î õ 0 á ä î õ 1 2 à ä î 0 àà ä î õ 0 à ä î 0 ä î õ 0 ä î 0 ä î õ 0 à à ä î õ 0 à à ä î àà à 0 àà ä î 0 ä î 0 ä ì 0 à ä M 2 í 0 à à ä î õ 0 àà ä î à 1 2 1 2 1 2 1 2 ó 0 à ä î 0 ä î 0 ä î 0 ä î æë 0 ä î ï 0 ä î ë 0 ä î ïï 0 ä î ð 0 â î 0 î 0 î ïï 1 2 1 2 1 2 î ï 0 î 0 î 0 î 0 ì 0 ü à 1 2 1 2 ü 1 2 1 2 1 2 1 2 ü 0 ü 0 ü 0 ü 0 ü å 0 ü 0 ü ï 0 ü 0 þ ï 0 ü 0 ü ïï 0 ü 0 ü ë 0 ü 0 ü 0 ü ò 0 ü ï 0 ü ïï 0 ü å ï 0 ü ï 0 ü ïï ï 0 ü æë 0 ü ï ï 0 ü ï ï 0 ü ïï ï 0 ü ïï 0 ü ïï ï 0 ü ð 0 ü ï ï ñ 0 ü ïï 0 ü ò 1 1 1 1 0 0 0 0 0 1 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 1 0 0 0 1 2 0 0 1 0 2 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 2 0 1 2 2 ï 1 2 1 2 0 0 0 0 0 1 2 m, n 2 õ 0 ü å Ú à àà á 0 æë õ 1 2 ë 0 ú å ï õ 1 2 æ ï ñ 0 ý ïï õ 0 þ ýý ï 0 ý ý ñ õ 0 ü ýý ï õ 0 ý ïï õ 1 2 ý ï ï õ 0 ýý ñ ñ õ 0 ý ý ò õ 0 ü ýý ï ð õ 0 ý ý ï ò 0 ý ý ï 0 ý ý ï õ 1 2 1 2 1 2 1 2 ý ý ïï 0 ý ý ïï õ 0 ý ý ï 0 ý ý ñ 0 ý ý ï 0 ý ý ñ õ 0 ù ïï ï 0 ü þ ïï ï õ 0 ý ï 0 ü þ ç M n_ : Sum invB M 1 MF ù ñ ï 0 ý ï õ 0 ý ýý ïï õ 0 ý ý ý ó 0 ü ýý õ 0 ý ý õ 0 ý ý ï õ 0 ý ý ñ õ 0 ý ý à àà 1 2 ý ý ï 0 ý ý ð 0 ý ý ñ 0 ý ïï 0 ü þ ï à 1 2 ü å õ 0 ý ý è 0 ü ýý ý ýý õ 0 ý ý ï õ 0 ýý ýý ñ õ 0 ý ý ò õ 0 ý ý å õ 0 ý ý õ 0 ý ý 0 ý ý 0 ý ý õ 0 ü þ å õ à 1 2 1 2 ü þ ýý 0 ý ý 0 ü ý õ 1 2 1 2 1 2 1 2 ü ýý ï õ 0 ü ýý ý ñ õ 0 ý ý õ 0 ý ý ï 0 ý ý ó 0 ý ý æ 0 ý ý æ 0 ý ý m , m, 1, 16 æ 0 ý ý 0 1 0 0 0 1 0 0 1 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 1 0 1 1 0 0 0 0 0 0 −1 ν=1 Bν,µ γ̂µ for our set of tomographic states ý ý P16 ý ý é 0 ü ýý ý 0 ü ýý 0 ü ýý þ 0 ü þ 0 ü ý þ 0 ü ýý ý ý 0 ü ýý ý ý 0 ü ýý ý ý 0 ü ýý ýý 0 ü ýý ý 0 þ þ 0 ü ýý þ 0 ý ý 0 ü þ ýý ý ý 0 ý ý ý ýý ýý 0 ü ýý ý 0 ü ýý 0 ýý þ 0 þ 0 ý 2 ýý à 1 2 1 2 ü ý 0 ü þ þ 0 ü ýý ýý ý 0 ü ýý ý 1 2 1 2 ü ýý 0 ü þ 0 ü ýý 0 ü þ 0 þ 0 ý 1 2 ýý 0 ý 0 ü þ 0 þ þ 0 ü ýý þ 0 ýý þ 0 ý þ 0 ý 0 þ ýý 0 ýý ý ý þ 0 ýý þ 0 ü ýý ý ý 0 ýý 0 þ 0 þ ý 0 ýý ýý ý ý 0 ü ýý 0 ü ýý ýý 0 þ þ 0 ýý ý 0 þ 0 ü ýý 0 ü þ 0 ý 0 ü þ 1 ýý 0 ý 0 ü þ 0 ü þ 0 ýý 0 ü þ 0 ü þ 0 ýý 0 ü ýý 0 ü þ 0 ýý þ 0 þ 0 ý ý 0 ýý 0 ýý ý ý 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 ýý M̂ν = ýý 1 2 1 2 1 2 1 2 þ 1 2 1 2 1 2 1 2 ý 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 ýý ý ý MF ýý ýý B ýý ý Ö Ö Ö b m_, n_ : hc psi m . n .psi m B : Table b m, n 1, 1 , m, 1, 16 , n, 1, 16 258 ýý where Bm,n = hψm|γ̂n|ψmi ü B = [Bm,n]16×16 ü non-singularity of B ↔ sensitivity of method ü with Bν,µ = hψν |γ̂µ|ψν i ü B = [Bν,µ]16×16 ü ν=1 ü γ̂ . ν,µ µ 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 þ 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 þ B −1 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 M̂ν = ü 16 X ü where ü reconstructed ρ P ρ̂ = (N )−1 16 ν=1 M̂ν nν 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 nν = N hψν |ρ̂|ψν i ö invB : Inverse B invB MF ø average number of coincidence counts inverse matrix B−1 ÷ µ̂ν = |ψν ihψν | projection measurement reconstruction analysis 257 ÿ æ ê å ô ï ïï ï ïï ô ï ïï ï ïï ï ïï ò ï ïï õ ô ïï ï ò ï ü û ì î í àà àà àà àà àà àà àà àà àà àà àà àà àà àà àà àà à à à à à à à à à à à à à à à à à à à à à à à à à à à à à à à ä ä Ý ß ã àà à àà à à á à àà à àà à á àà à á àà Û àà àà á à à à àà á à à àà à à à à à à à àà á à àà àà à àà à á × á à àà àà á àà à á àà àà àà à à à à à àà à Ü á Þ 260 259 another common set of tomographic states M 2 1 2 1 2 2 2 MF −1 ν=1 Bν,µ 2 0 γ̂µ for the new set of tomographic states 1 2 2 0 0 2 1 2 1 2 0 0 0 2 1 2 0 P16 1 2 0 2 0 2 1 2 2 1 1 2 0 1 0 0 0 2 2 2 1 2 1 2 0 1 2 M̂ν = No. Mode 1 Mode 2 1 |Hi |Hi 2 |Hi |Vi 3 |Vi |Vi 4 |Vi |Hi 5 |Ri |Hi 6 |Ri |Vi 7 |Di |Vi 8 |Di |Hi Note: we have just replaced D by D. 9 |Di |Ri 10 |Di |Di 11 |Ri |Di 12 |Hi |Di 13 |Vi |Di 14 |Vi |Li 15 |Hi |Li 16 |Ri |Li b m_, n_ : hc psi m . n .psi m B : Table b n, m 1, 1 , n, 1, 16 , m, 1, 16 invB : Inverse B M n_ : Sum invB m, n m , m, 1, 16 M 1 MF 261 262 2 −β −α i −α 0 = −β −i 0 1 0 0 1 2 0 0 α 0 1 0 , 0 0 2i −α β 0 , 0 0 0 0 0 0 0 1 i −α 0 0 = , 0 −i 0 −β 1 −β −α 2 1 2 0 −α 1 0 i 0 , −i 2 −β 0 −α 0 0 −β 0 1 −α 2 i −α = , 0 −i 0 0 1 −α 0 0 1 2 0 0 = −β 1 1 2 0 0 0 −α 0 β 2i 0 M̂6 = 12 , 0 α 0 0 −β −2i 0 0 M̂4 M̂2 16 matrices M̂n for the latter tomographic states M̂1 M̂3 0 0 M̂5 = 21 −2i −β i 0 , 0 0 i 0 , 0 0 0 0 1 0 0 1 0 0 1 0 , 0 0 0 2 0 −α 0 −α 0 2 M̂12 = 12 , 0 −β 0 0 −β 0 0 0 0 0 M̂10 = 0 1 0 0 0 −α 0 0 2 −α 0 −β 2 0 −β 0 0 0 M̂7 = 21 , M̂8 = 21 , 0 −α 0 0 2 −α 0 0 −β 2 0 0 −β 0 0 0 0 i 0 0 0 −i 0 0 0 0 −i 0 0 0 0 0 M̂9 = 0 i −i 0 0 0 M̂11 = 0 −i 0 −β β 0 0 0 0 0 , 0 0 M̂16 = 0 1 0 0 −1 0 0 −1 0 0 1 0 , 0 0 0 0 0 −α 0 0 0 −β 0 −α 0 0 −β 0 0 0 M̂13 = , M̂14 = 1 , 2 2 0 −β 0 0 −α 0 −2i −β 0 2 0 −α 0 2i 0 1 2 0 −2i 2i 0 M̂15 = α 0 −α 0 1 2 where α ≡ 1 + i; β ≡ 1 − i. Note: other M̂n’s are obtained for other sets of rotations. 263 264 with e.g. By the Schur decomposition, any normal matrix  (i.e., † = †) can be represented as  = Û T̂ Û † in terms of a triangular T̂ and a unitary Û . For simplicity, we neglect Û . of lower-triangular (or upper-triangular) form, which is easily invertible 0 0 , 0 t4 Tr{T̂ †T̂ } T̂ †T̂ t1 0 0 t5 + it6 t 0 2 T̂ = t11 + it12 t7 + it8 t3 t15 + it16 t13 + it14 t9 + it10 ρ̂phys ≡ ρ̂phys(t1, t2, . . . t16) = e.g. • positive ν=1 = 2 N min L(t) 2 t hψν |ρ̂p(t)|ψν i − nNν hψν |ρ̂p(t)|ψν i where 2 is a useful ‘likelihood’ function. L(t) = 16 X N hψν |ρ̂p(t)|ψν i − nν − max t 2N hψν |ρ̂p(t)|ψν i ν=1 16 X or, equivalently, analyze logarithm of P (n, t): for a given measured data n: " 2 # 16 Y N hψν |ρ̂p(t)|ψν i − nν exp − max . t 2N hψν |ρ̂p(t)|ψν i ν=1 find numerically a maximum of P (n, t) • normalized • Hermitian 3. optimize ρ̂phys(t): 1. construct explicitly a “physical” density matrix ρ̂phys Nnorm(t) = const – normalization assumed to be independent of t maximum-likelihood method 266 267 268 n̄ν (t) = N hψν |ρ̂phys (t) |ψν i – number of counts expected for νth measurement p σν (t) ≈ n̄ν (t) – standard deviation P N = 4ν=1 nν ≫ 1 – normalization, so n̄Nν corresponds to a probability where (nν − n̄ν (t))2 P (n, t) = exp − Nnorm(t) ν=1 2σν2(t) 1 16 Y is for given ρ̂phys(t) with t = (t1, t2, . . . t16) n = (n1, n2, . . . n16) the probability of obtaining a set of counts 2. construct a likelihood function assuming e.g. Gaussian noise of the measurements, maximum-likelihood method maximum-likelihood (MaxLik) method this is not a physical density matrix! Tr ρ̂2 = 1.053 eig ρ̂ = {1.02155, 0.0681238, −0.065274, −0.024396} problems with semi-definiteness and normalization 0.4872 0.0042 + i0.0114 0.0098 − i0.0178 0.5192 − i0.0380 0.0042 − i0.0114 0.0045 0.0271 + i0.0146 0.0648 − i0.0076 ρ̂ = 0.0098 + i0.0178 0.0271 − i0.0146 0.0062 0.0695 + i0.0134 0.5192 + i0.0380 0.0648 + i0.0076 0.0695 − i0.0134 0.5020 reconstructed ρ̂ by linear tomography n13 = 13171, n14 = 17170, n15 = 16722, n16 = 33586 n9 = 17932, n10 = 32028, n11 = 15132, n12 = 17238, n5 = 16324, n6 = 17521, n7 = 13441, n8 = 16901, n1 = 34749, n2 = 324, n3 = 35805, n4 = 444, counts for 16 projection measurements experimental example of James et al. 265 ‘likelihood’ function t5 t11 t15 $ % " ! & # ! ! ; thus, we need initial values tini. r M(0) (1) M00 (1) M01 q (1) (2) M00 M00,11 (2) M q01,12 √ (2) M00,11 ρ33 ρ √ 30 ρ33 s 0 (1) (2) M00 M00,11 (2) M q00,12 √ (2) M00,11 ρ33 ρ √ 31 ρ33 r 0 0 (2) M00,11 ρ33 ρ √ 32 ρ33 0 0 ρ33 0 √ How to estimate initial ‘physical’ matrix T̂ (tini)? T̂ (tini) = By calculating T̂ from our “unphysical” ρ̂: where M(0) = Det(ρ̂), (1) Mij = Det(ρ̂without row i & col j), (2) Mij,kl = Det(ρ̂without rows i,k & cols j,l) i, j, k, l = 0, ..., 3 269 270 FindMinimum[ L[t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,t11,t12,t13,t14,t15,t16], {t1,tini1},{t2,tini2},{t3,tini3},{t4,tini4},{t5,tini5},{t6,tini6}, {t7,tini7},{t8,tini8},{t9,tini9},{t10,tini10},{t11,tini11}, {t12,tini12},{t13,tini13},{t14,tini14},{t15,tini15},{t16,tini16}] t1 0 0 0 t6 t2 0 0 ; t12 t7 t8 t3 0 t16 t13 t14 t9 t10 t4 L t1_, t2_, t3_, t4_, t5_, t6_, t7_, t8_, t9_, t10_, t11_, t12_, t13_, t14_, t15_, t16_ : Module T, RhoPhys, sum, Nmean , T: RhoPhys : hc T .T Tr hc T .T ; sum 0; Do Nmean Re Norm hc psi m . phys .psi m ; sum sum Nmean Nexp m ^ 2 2 Nmean , m, 1, 16 Return Re sum ; numerical optimalization ' for our experimental data 271 0.5948i 0 0 0 0.8706 + 0.7282i 0.4060 0 0 T̂ (tini) = −0.0215 + 0.9933i −0.2827 − 0.2982i 0.0615i 0 0.7328 + 0.0536i 0.0915 + 0.0107i 0.0981 − 0.0189i 0.7085 0.7870 0.0325 − 0.0014i 0.0328 − 0.0051i 0.1289 − 0.0094i 0.0325 + 0.0014i 0.0850 −0.0024 − 0.0050i 0.0161 − 0.0019i ρ̂phys(tini) = 0.0328 + 0.0051i −0.0024 + 0.0050i 0.0034 0.0173 + 0.0033i 0.1289 + 0.0094i 0.0161 + 0.0019i 0.0173 − 0.0033i 0.1247 L(tini) = 3.0695 272 by applying the optimalization procedure we can diminish this value to L(topt) = 0.0104 for the following matrix ... our matrix reconstructed by MaxLik method 0.5028 0.0230 + 0.0117i 0.0279 − 0.0130i 0.4686 − 0.0346i 0.0051 0.0050 + 0.0001i 0.0333 − 0.0082i 0.0230 − 0.0117i ρ̂phys(topt) = 0.0279 + 0.0130i 0.0050 − 0.0001i 0.0066 0.0416 + 0.0102i 0.4686 + 0.0346i 0.0333 + 0.0082i 0.0416 − 0.0102i 0.4854 is a physical density matrix as it is • positive (semidefinite) eig ρ̂phys = {0.9687, 0.0313, 0, 0} • Hermitian † ρ̂phys = ρ̂phys • normalized Tr{ρ̂phys} = 1 2 } = 0.9394 ≤ 1 Tr{ρ̂phys 273 mn vh vh m hv m hv vh vh hh hh n hv n hv vh vh vh vh −0.05 hh 0 0.05 0 hh 0.5 vh vh m hv m hv vh vh hh hh 1 3 1+ √ √ 3 3hσ̂3i) 2 (hσ̂8i + 3 (hσ̂ i + ihσ̂ 1 2i) 2 3 (hσ̂4i + ihσ̂5i) 2 1+ 3 i) 2√(hσ̂1i − ihσ̂ √2 3 (hσ̂ i − 3hσ̂ 3 i) 8 2 3 (hσ̂6i + ihσ̂7i) 2 3 2 (hσ̂4 i − 3 − 2 (hσ̂6 i√ 1− vh vh vh vh 274 ihσ̂5i) ihσ̂7i) 3hσ̂8i n hv n hv 1 0 0 √1 0 1 0 . 3 0 0 −2 Analogously, a qudit density matrix can be reconstructed via measurement of generators of the Lie algebra SU (d) 0 0 −i 0 0 0 0 0 0 σ̂5 = 0 0 0 , σ̂6 = 0 0 1 , σ̂7 = 0 0 −i , σ̂8 = i 0 0 0 1 0 0 i 0 which can be chosen as generators of the Lie algebra SU (3) 0 1 0 0 −i 0 1 0 0 0 0 1 σ̂1 = 1 0 0 , σ̂2 = i 0 0 , σ̂3 = 0 −1 0 , σ̂4 = 0 0 0 , 0 0 0 0 0 0 0 0 0 1 0 0 via measurement of the generalized Pauli operators ρ̂ = reconstruction of a qutrit density matrix −0.05 hh 0 0.05 0 hh 0.5 ρ̂ by linear tomography (left) and ρ̂phys by MaxLik tomography (right figures) comparison of reconstructed matrices phys mn mn Re ρ mn Im ρ Re ρ Im ρphys simulation of gates by various circuits Toffoli and CCU gates reversible gates (part II) Quantum gates 276 Copies of ρ̂ are generated from the same (known) initial state(s) by applying the same transformations in the same experimental setup. No! The term “copies” refers to identically prepared quantum objects. Answer Thus, do we violate the no-cloning theorem? But unknown ρ̂ cannot be copied. For complete reconstruction of ρ̂ we need to repeat measurements on many copies of ρ̂. Quantum tomography and no-cloning theorem: Question 275 277 Ĥ(c0|+i + c1|−i) = c0|0i + c1|1i 1 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 |a〉 |a〉 |a〉 |b〉 |a〉 |b〉 |ab ⊕ c〉 |a ⊕ b〉 |c〉 |b〉 CNOT and Toffoli gates 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 = V V† V Note: Toffoli gate is a special case of CCU U Any CCU gate can be built from CNOT, CV, and CV† gates, where V2 = U theorem of Barenco,Bennett et al. (1995) replacement for CCU gate 0 0 0 0 1 2 3 3 3 0 0 0 reversible computing → + c3|11i 278 . / - a computational process that is (or almost is) time-invertible (time-reversible). Landauer’s principle - a computational process to be physically reversible, it must also be logically reversible. logically-reversible process - a discrete, deterministic computational process for which the transition function that maps input states to output states is a one-to-one function. examples of reversible quantum gates + c2|10i all unitary quantum gates are reversible, e.g.: + c1|01i Ĥ(c0|0i + c1|1i) = c0|+i + c1|−i = c0|00i XOR=CNOT ab | c -------00 | 0 01 | 1 10 | 1 11 | 0 R-XOR ab | abc ac bc -----------------00 | 000 00 00 01 | 011 01 11 10 | 101 11 01 11 | 110 10 10 - - - - - - - * * * * * * * + , |ψini ÛCNOT|ψini = c0|00i + c1|01i + c2|11i + c3|10i ≡ |ψouti ÛCNOT|ψouti = c0|00i + c1|01i + c2|1, 1 ⊕ 1i + c3|1, 0 ⊕ 1i = |ψini but measurement is irreversible examples of irreversible classical gates OR ab | c -------00 | 0 01 | 1 10 | 1 11 | 1 logic gates in a classical computer, other than NOT gate, are irreversible, e.g.: AND ab | c -------00 | 0 01 | 0 10 | 0 11 | 1 how to make these gates reversible? R-OR ab | abc ---------00 | 000 01 | 011 10 | 101 11 | 111 keep input(s) together with the standard output, e.g.: R-AND ab | abc ---------00 | 000 01 | 010 10 | 100 11 | 111 ( ) 279 280 281 5 V V† |x V 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 7 4 7 > @ = @ = @ = @ = @ = @ @ = 4 = @ @ = 4 4 4 4 4 = @ = @ ; ? 4 MF ket 0, 1 .bra 0, 1 v0,1 v1,1 ket 1, 0 .bra 1, 1 v0,0 v1,0 ket 1, 1 .bra 1, 0 CV gate – control unitary V = 0 0 1 0 9 : : = < 284 4 4 4 B C D B C D C B B C C C B B A B 4 4 4 4 4 E 4 4 4 0 0 0 1 0 0 0 v0,0 v0,1 0 v1,0 v1,1 B 1 0 0 0 cv : ket 0, 0 .bra 0, 0 ket 0, 1 .bra 0, 1 v0,0 ket 1, 0 .bra 1, 0 v0,1 ket 1, 0 .bra 1, 1 v1,0 ket 1, 1 .bra 1, 0 v1,1 ket 1, 1 .bra 1, 1 cv MF cv = |00ih00| + |01ih01| + v0,0|10ih10| + v0,1|10ih11| + v1,0|11ih10| + v1,1|11ih11| 0 0 0 1 8 0 1 0 0 6 1 0 0 0 cnot : ket 0, 0 .bra 0, 0 cnot MF 8 U T CNOT gate – cnot = |00ih00| + |01ih01| + |11ih10| + |10ih11| W C 4 U|x 5 U 4 V 5 X V|x 5 U 4 V† X B C V|x U V X C B |x U D B 4 4 4 U|x X C |x U |1 X |1 U |0 X |0 U |1 X |1 U ? X = U |1 X |1 U |1 X |1 U |1 X |1 S |1 5 V |1 L 9 |1 T 8 |1 W V U V† X V U U X |x U V† |x X |x U |x X |x U 9 |x X |0 U |0 X |1 U 7 |1 X 8 |0 U 7 |0 N ? M = M |0 Q |0 N |1 Q 7 |1 M 8 |1 M 7 |1 L 8 |1 L 7 |1 N 8 |1 M 7 |1 X 282 L |x M V|x R M |x R |x L R 0 0 0 1 N R ket 1, 1 L n2_ : Module state , ZeroMatrix 4, 1 ; 2 n1 n2 1, 1 1; state n2_ : Transpose ket n1, n2 MF M replacement for CCU gate (proof) U U |1 X |1 U |1 1 0 0 0 X |1 L ket n1_, state state Return bra n1_, ket 0, 0 U |1 |0 |x bra[n1, n2] = hn1, n2| → row vector X |1 |0 V |0 283 ket[n1, n2] = |n1, n2i → 4-element column vector with ‘1’ at pos. 2n1 + n2 + 1 M |x ? = |0 |x |0 |0 S |1 |0 V† |0 |0 V |1 |0 |x |0 |0 • two-qubit states implementing gates in M athematica L |0 5 L |0 V |0 |0 5 L |x |0 |0 5 M |x ? = |0 5 LO |0 U |0 |0 5 P |x 5 |0 |0 replacement for CCU gate (proof) 5 M 8 D C C B D I K H K K H H K H K K H 4 4 4 4 H K K H H K K H 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 H K H K F H J E G 0 0 0 0 1 0 0 0 Y [ 0, 1, 1 0, 1, 1 1, 1, 1 1, 1, 1 i f e e d e f dg d h e u u ~ 287 288 e 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 v0,0 v0,1 0 v1,0 v1,1 v 285 0 0 0 0 1 0 0 0 t d 3. CV23=kron(id,cv) t • three-qubit states v i CV23new : Kron id, cv CV23new MF s ‘1’ at position 4n1 + 2n2 + n3 + 1 u KET n1_, n2_, n3_ : Module state , state ZeroMatrix 8, 1 ; 1; state 4 n1 2 n2 n3 1, 1 Return state BRA n1_, n2_, n3_ : Transpose KET n1, n2, n3 KET 0, 0, 0 MF or explicitly t 1 0 0 0 0 0 0 0 MF v MF t f 0 0 0 1 0 0 0 v0,0 v0,1 0 v1,0 v1,1 0 0 0 0 0 0 0 0 0 0 0 0 u e 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 v0,0 v0,1 0 v1,0 v1,1 t v v u u 1 0 0 0 0 0 0 0 u i 0 1 0 0 0 0 0 0 286 u t t t 0 0 0 0 1 0 0 0 u t u u u u 1 0 t t t t t v v v v u u u u t CV23 : KET 0, 0, 0 .BRA 0, 0, 0 KET 0, 0, 1 .BRA 0, 0, 1 v0,0 KET 0, 1, 0 .BRA 0, 1, 0 v0,1 KET 0, 1, 0 .BRA v1,0 KET 0, 1, 1 .BRA 0, 1, 0 v1,1 KET 0, 1, 1 .BRA KET 1, 0, 0 .BRA 1, 0, 0 KET 1, 0, 1 .BRA 1, 0, 1 v KET 1, 1, 0 .BRA 1, 1, 0 v0,1 KET 1, 1, 0 .BRA 0,0 v1,0 KET 1, 1, 1 .BRA 1, 1, 0 v1,1 KET 1, 1, 1 .BRA t 0 0 0 1 0 0 0 v0,0 v0,1 0 v1,0 v1,1 0 0 0 0 0 0 0 0 0 0 0 0 CV23 1 0 0 0 0 0 0 0 u e KET 1, 0, 0 .BRA 1, 0, 0 KET 1, 1, 0 .BRA 1, 1, 1 t t t t v u u u u v t t t t BRA 0, 0, 1 \ e MF Z f 0 0 0 0 0 0 0 1 [ d BRA 1, 1, 1 Z e KET 0, 0, 1 .BRA 0, 0, 1 KET 0, 1, 0 .BRA 0, 1, 1 KET 1, 1, 1 .BRA 1, 1, 0 \ [ u u \ [ d d control gates for 3 qubits 0 0 0 0 0 0 1 0 [ \ [ [ \ [ \ +|100ih100| + |101ih101| + |111ih110| + |110ih111| Z 0 0 0 0 0 0 0 1 Z Z [ [ Z Z Z Z Z Z e j BRA[n1, n2, n3] = hn1, n2, n3| → row vector 0 0 0 0 0 1 0 0 w j j j 1. CNOT23=kron(id,cnot) where id = 0 1 MF 0 0 1 0 0 0 0 0 w KET[n1, n2, n3] = |n1, n2, n3i → 23-element column vector with 0 0 0 1 0 0 0 0 a \ [ [ r r n d d j e k m d d d j e p p p m m m m m p p p p p p m m m m m p p p p p m m m m m m p p p p p m m m m m p p p p p p m m m m m m p p p p p o m m l q q or CNOT23 = |000ih000| + |001ih001| + |011ih010| + |010ih011| Z CNOT23 : KET 0, 0, 0 .BRA 0, 0, 0 KET 0, 1, 1 .BRA 0, 1, 0 KET 1, 0, 1 .BRA 1, 0, 1 Z 0 1 0 0 0 0 0 0 ] [ [ Z CNOT23 1 0 0 0 0 0 0 0 { ] Z Z [ [ 2. CNOT12=kron(cnot,id) – analogously } } } } } } } } } } } } } } } } } } } } } } } } } } } } } } } | x z z z z z z z z z z z z z z z z z z z z z z z z z z z z z z z y c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c b ^ ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` ` _ 1 1 1 1 CV13= SWAP12 CV23 SWAP12 = kron(swap,id) CV23 kron(swap,id) tricky way to calculate CV13 0, 0, 1, 1, 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 v0,0 v0,1 0 0 0 v1,0 v1,1 0 0 0 0 0 v0,0 v0,1 0 0 0 v1,0 v1,1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 MF CV13 CV13 : KET 0, 0, 0 .BRA 0, 0, 0 KET 0, 0, 1 .BRA 0, 0, 1 KET 0, 1, 0 .BRA 0, 1, 0 KET 0, 1, 1 .BRA 0, 1, 1 v0,0 KET 1, 0, 0 .BRA 1, 0, 0 v0,1 KET 1, 0, 0 .BRA 1, v1,0 KET 1, 0, 1 .BRA 1, 0, 0 v1,1 KET 1, 0, 1 .BRA 1, v0,0 KET 1, 1, 0 .BRA 1, 1, 0 v0,1 KET 1, 1, 0 .BRA 1, v1,0 KET 1, 1, 1 .BRA 1, 1, 0 v1,1 KET 1, 1, 1 .BRA 1, ¡ 1 0 0 0 0 0 0 0 ¡ ¡ ¡ ¡ 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 u0,0 u0,1 0 u1,0 u1,1 ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¤ ¡ ¡ ¤ £ U • ¢ ¡ ¡ ¤ ¤ ¡ ¤ ¡ ¤ © ¡ ¥ £ ¢ ¥ ¤ ¡ ¡ ¤ £ ¨ ¨ v0,1 v1,0 v0,0 v0,1 £ v1,0 v1,1 ¢ ¡ v0,1 v1,0 MF v21,1 v0,1 v1,1 v0,0 v0,1 v0,0 v0,1 . ; v1,0 v1,1 v1,0 v1,1 0 0 0 1 0 0 0 0 ¤ u0,0 u0,1 u1,0 u1,1 0 1 0 0 0 0 0 0 ¤ ¬ © v0,0 v1,0 PQRS WVUT × × ¤ v20,0 • PQRS WVUT • = 5. CCU gate (U = V 2) = PQRS WVUT • × ¤ CV13=SWAP23 CV12 SWAP23 = · · · ¡ ¤ ª or ¡ ¤ £ 0 ¡ ¤ CV13new ¡ ¤ ¤ Max CV13 ¡ ¤ u0,0 u0,1 u1,0 u1,1 CCU : ¡ ¤ ¤ test U • ¤ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 v0,0 v0,1 0 0 0 v1,0 v1,1 0 0 0 0 0 v0,0 v0,1 0 0 0 v1,0 v1,1 ¡ ¤ 0 0 1 0 0 0 0 0 × × ¡ ¤ 0 1 0 0 0 0 0 0 × × ¤ 1 0 0 0 0 0 0 0 = × substitution circuit for SWAP gate U • substitution circuits for CU13 gate ¢ CV13new : Kron swap, id .CV23.Kron swap, id CV13new MF 290 +v0,0|110ih110| + v0,1|110ih111| + v1,0|111ih110| + v1,1|111ih111| 1 0 0 0 0 0 0 0 289 +v0,0|100ih100| + v0,1|100ih101| + v1,0|101ih100| + v1,1|101ih101| CV13 = |000ih000| + |001ih001| + |010ih010| + |011ih011| 4. CV13 gate ¢ ¬ ¬ ¨ ¬ « © © ¡ ¦ ¨ § × × 292 291 V 293 properties of unitary matrices V = thus for v0,0 v0,1 v1,0 v1,1 we have ∗ ∗ v0,0 v0,1 + v1,0 v1,1 = 0 v0,0 v0,0 v0,1 v1,0 v0,0 v0,0 v0,1 v1,0 etc. v0,0 v0,0 v0,1 v1,0 v0,1 v1,0 v0,0 v0,0 Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate ∗ ∗ v0,0 + v0,1 v0,1 = 1 v0,0 Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate Conjugate v1,0 v0,1 v1,1 v1,1 v1,0 v0,1 v1,1 v1,1 v1,0 v0,1 v1,1 v1,1 v1,1 v1,1 v1,0 v0,1 · 1 1 1 1 0 0 0 0 a matrix is unitary iff its row and column vectors are orthonormal r1 : r2 : r3 : r4 : r5 : r6 : r7 : r8 : · a substitution circuit for CCU gate V† ...finally · È È È È È È È È Ç Ç Ç Ç Ç Ç Ç Ç Å Å Å Å Å Å Å Å Ä Ä Ä Ä Ä Ä Ä Ä V gate1 gate2 gate3 gate4 gate5 294 · Ä 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 v0,1 0 0 0 0 0 0 v0,1 v1,0 v0,0 v1,1 2 v1,1 v0,1 295 296 CCU1 : CCU1temp . r1 . r2 . r3 . r4 . r5 . r6 . r7 . r8 CCU1 MF 0 1 0 0 0 0 0 ¶ substitution circuit for CCU gate 0 0 ; 1 0 1 0 0 0 0 0 0 v1,0 v1,0 v1,1 ¼ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ » 0 0 0 1 · Â Â Â Æ Æ Æ Æ Æ Æ Æ Æ Å Å Å Å Å Ã Ã Ã Ã Ä Ä Ä Ä Ä Ä Ä Å v0,1 » 0 1 0 0 · Å Å MF » 1 0 0 0 · 2 v0,0 · 0 0 ; cnot 0 1 · CCU1 · ±      à à à à Max CCU Á ¾ · 0 0 0 0 0 0 v0,0 v1,0 test 0 À ½ » · we have shown that our gate CCU1 simulates the original CCU gate ¶ ¸ º º º º º º º º º º º º º º º º º º º º 0 1 0 0 µ 0 0 1 0 ´ 1 0 0 0 µ º º º º º º º º º º º º ³ ³ ³ ³ ³ ³ IdentityMatrix 2 ; ´ ± 0 0 0 1 0 0 ; id 0 v0,0 v0,1 0 v1,0 v1,1 ´ ´ ´ ¹ ¯ ³ ° µ ³ ° ² ³ ° ° µ ³ ° ° µ ° ° ° ´ ² ³ ® ³ ° ³ ³ ³ ³ ³ ³ ³ ³ ³ ² swap : cv ³ ³ ³ ³ ± ³ ³ ³ ³ ³ µ ® ° ° ° ° ° ° ° ° ° ° ¯ ³ µ ´ 1 0 0 0 ¿ gate1 : Kron id, cv gate2 : Kron cnot, id gate3 : Kron id, hc cv gate4 : gate2 gate5 : Kron swap, id .gate1.Kron swap, id CCU1temp : gate5.gate4.gate3.gate2.gate1 ® ° ° ° ° ° ° ° ° ° ° ° ¯ Ê Ë Ê Ë Ê CCU2temp : CV13.Kron cnot, id .hc CV23 .Kron cnot, id .CV23 or more explicitly Ë 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 v20,0 Ï Ï Ï Ï Ï Ï Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ï Ó Ð Ð Ð Ì Ì Í Ï Î Ê Ò Ð Ì Ì Ë Ô É qubits cannot be measured without disturbing quantum information they carry, i.e., by measuring Steane’s ECC fault-tolerant gates without quantum ECCs construction of practical quantum computers would be impossible quantum information cannot be copied with perfect fidelity 3. no cloning we get either |0i with probability |a|2 or |1i with probability |b|2 a|0i + b|1i 2. measurement destroys quantum information Shor’s ECC a note there are infinitely many quantum errors 300 correction of bit-flip and phase-flip errors can we really correct quantum errors? thus we need a quantum ECC – an analog of classical ECC quantum states are very fragile to decoherence and dissipation, error-correcting code (ECC) there is a continuous set of quantum errors thus which can also correspond to removing the qubit and replacing it with garbage! 3. small errors √ √ a|0i + b|1i → a2 + ǫ|0i + b2 − ǫ|1i e.g. |+i → |−i & |−i → |+i (orthogonal) |0i → |0i & |1i → −|1i 2. phase-flip errors = phase errors |0i → |1i & |1i → |0i 1. bit-flip errors = amplitude errors errors in quantum computers 0→1&1→0 bit-flip errors errors in classical computers 299 1. errors are continuous 298 297 quantum vs classical errors Introduction to quantum error-correction codes MF Ó CCU2 Ó Max CCU Ó test Ó v21,1 v0,1 v1,1 Ó v0,1 v1,0 v0,0 v0,1 0 0 0 0 0 0 Ó v1,0 v1,1 0 0 0 0 0 0 v0,1 v1,0 Ó 0 0 0 1 0 0 0 0 Ñ 0 0 0 0 0 0 v0,0 v1,0 0 1 0 0 0 0 0 É 1 0 0 0 0 0 0 CCU2 : CCU2temp . r1 . r2 . r3 . r4 . r5 . r6 . r7 . r8 CCU2 MF É Ì Ì Ì Ì Ì Ì Ì Ì 301 bit-flip error |ψ2i = |ψ1i|00i = a|00100i + b|11000i 24 14 |ψ i = ÛCNOT ÛCNOT |ψ2i 3 = a|0, 0, 1, 0 ⊕ 0 ⊕ 0, 0i + b|1, 1, 0, 0 ⊕ 1 ⊕ 1, 0i = a|0, 0, 1, 0, 0i + b|1, 1, 0, 0, 0i = |ψ2i (sic!) 45hM ′ , M ′′|ψ4i = a|0, 0, 1i + b|1, 1, 0i so let us measure modes 4 & 5: 35 25 |ψ i = ÛCNOT ÛCNOT |ψ3i 4 = a|0, 0, 1, 0, 0 ⊕ 0 ⊕ 1i + b|1, 1, 0, 0, 0 ⊕ 0 ⊕ 1i = a|0, 0, 1, 0, 1i + b|1, 1, 0, 0, 1i a|0, 0, 1i + b|1, 1, 0i ⊗ |0, 1i = |ψ5i = in our case the measurement result (i.e. error syndrome) is: M ′ = 0, M ′′ = 1 error correction = recovery so we know that we have to flip the 3rd qubit |ψ6i = X̂3|ψ5i = a|0, 0, 0i + b|1, 1, 1i = |ψ0i M′ M′ M′ M′ = = = = so in general 0, M ′′ = 1 =⇒ |ψ0i = X̂3|ψ5i 1, M ′′ = 0 =⇒ |ψ0i = X̂2|ψ5i 1, M ′′ = 1 =⇒ |ψ0i = X̂1|ψ5i 0, M ′′ = 0 =⇒ |ψ0i = |ψ5i error syndrome =⇒ error correction 303 Analogously, we can find a proper action for other measurement results: 304 error detection = error syndrome diagnosis/measurement quantum repetition code in general +b 1 + b|1i)|0i⊗(N −1) = a|0i⊗N + b|1i⊗N a 00 + b 11 ⊗N a 000 + b 111 a0 302 |ψ0i = a|000i + b|111i → 3rd bit flip → |ψ1i = a|001i + b|110i ⊗N |ψi = a|0i + b|1i 12 12 ÛCNOT |ψi|0i = ÛCNOT (a|00i + b|10i) = a|0, 0 ⊕ 0i + b|1, 0 ⊕ 1i = a|00i + b|11i 1n ÛCNOT (a|0i 13 12 12 13 ÛCNOT ÛCNOT = ÛCNOT ÛCNOT note that 13 12 13 ÛCNOT Û CNOT|ψi|0i|0i = ÛCNOT(a|00i + b|11i)|0i = a|0, 0, 0 ⊕ 0i + b|1, 1, 0 ⊕ 1i = a|000i + b|111i i=2 N Y a 0 +b 1 0 a 0 +b 1 0 0 ⊗( N −1) a 0 +b 1 0 ˆ |ψ0i = X̂|3M ′′−2M ′||ψ5i, where X̂0 = I, |ψi |ψi |0i |0i |0i 45hM ′ • M M _ • • R Recovery |ψ0i 35 25 24 14 , M ′′|ÛCNOT ÛCNOT ÛCNOT ÛCNOT |ψ1i|00i • // syndrome detect error correct decode 89:; ?>=< • // 89:; ?>=< // // • |0i |0i 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • • GF @A M ′′BCED • MBCED′ GF @A R example: a quantum bit-flip ECC 306 89:; ?>=< • |ψi 89:; ?>=< • where −//− denotes quantum channel where an error can occur encode |0i • _ a typical error correction code (ECC) |ψ0i = X̂|3M ′′−2M ′| |ψ1i • _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ GF ED ?>=< 89:; ?>=< 89:; ′ @A BC GF ED 89:; ?>=< 89:; ?>=< @A ′′ BC _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Syndrome Measurement bit-flip error correction |ψ1i → |ψ0i |ψ1i = a|001i + b|110i, a|010i + b|101i, a|100i + b|011i or a|000i + b|111i |ψ0i = a|000i + b|111i bit-flip error |ψ0i → |ψ1i 305 |ψi initial encoding |−i → phase flip → |+i |0i + |1i |0i − |1i √ → phase flip → |−i = √ 2 2 13 12 |ψ0i = Ĥ ⊗3ÛCNOT ÛCNOT (a|0i + b|1i) = Ĥ1Ĥ2Ĥ3(a|000i + b|111i) = a| + ++i + b| − −−i or explicitly a|0i + b|1i → |ψ0i = a|000i + b|111i → |ψ0′ i = a| + ++i + b| − −−i |+i = so a|0i + b|1i → phase flip → a|0i − b|1i phase-flip error appears as an amplitude error (bit flip) and vice versa. a phase error (phase flip) in a rotated basis thus 1 1 1 0 1 1 1 1 1 0 √ Ĥ σ̂xĤ = √ = = σ̂z 1 0 0 −1 2 1 −1 2 1 −1 1 1 1 1 1 1 1 0 √ Ĥ σ̂z Ĥ = √ 1 −1 0 −1 2 1 −1 2 1 1 −1 1 1 = 1 −1 2 1 1 0 1 = = σ̂x 1 0 phase vs amplitude errors 308 307 encoded initial state |ψ0′ i = a| + ++i + b| − −−i phase-flip error |ψ0′ i → |ψ1′ i |ψ1′ i = a| + +−i + b| − −+i or a| + −+i + b| − +−i or a| − ++i + b| + −−i or a| + ++i + b| − −−i and analogously a| + −+i + b| − +−i → a|010i + b|101i a| − ++i + b| + −−i → a|100i + b|011i a| + ++i + b| − −−i → a|000i + b|111i so now our standard bit-flip ECC can be applied • Syndrome Measurement • 89:; ?>=< • • 89:; ?>=< • • R Recovery phase-flip error correction |ψ1′ i → |ψ0i H H H 89:; ?>=< 89:; ?>=< GF ′′ @A M BCED GF BC @A MED′ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |0i |0i |ψ0i ′ 35 25 24 14 , M ′′|ÛCNOT ÛCNOT ÛCNOT ÛCNOT Ĥ ⊗3|ψ1i|00i 309 310 apply Hadamard gates to |ψ1′ i → |ψ1i |ψ1i = Ĥ ⊗3|ψ1′ i = Ĥ1Ĥ2Ĥ3(a| + +−i + b| − −+i) = a|001i + b|110i |ψ1′ i 45 hM _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |ψ0i = X̂|3M ′′ −2M ′| • 89:; ?>=< • 89:; ?>=< a|000i + b|111i encoding qubit against bit-flip error a|0i + b|1i |0i |0i • 89:; ?>=< • 89:; ?>=< H H H a| + ++i + b| − −−i encoding qubit against phase-flip error a|0i + b|1i |0i |0i How to encode qubit against both errors? • 89:; ?>=< • 89:; ?>=< |0i |0i |0i |0i H H H 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • 89:; ?>=< • encoding code for Shor’s nine-qubit ECC a|0i + b|1i |0i |0i |0i |0i 311 312 first real error correction code against both bit-flip and phase-flip single error (|0i + |1i)(|0i + |1i)(|0i + |1i) 1 (|0i − |1i)(|0i − |1i)(|0i − |1i) a|0i + b|1i −→ a|0iL + b|1iL (|000i − |111i)(|000i − |111i)(|000i − |111i) ≡ |1iL 23/2 H H |0i |0i g g s |0i |0i a|0i + b|1i |0i H |0i g g g s g g g s g g g s 314 a|0iL + b|1iL encoding code for Steane’s seven-qubit ECC • there are other more concise ECCs • thus Shor’s code can correct combined bit and phase flips on a single qubit • the corrections of bit and phase flip errors are independent Note that: 23/2 1 → | − −−i = → 23/2 (|000i + |111i)(|000i + |111i)(|000i + |111i) ≡ |0iL 23/2 1 |1i → |111i → → | + ++i = 76 75 |ψ2i = ÛCN OT ÛCN OT |ψ1i = |ψ1i |ψ1i = Ĥ1Ĥ2Ĥ3|0i⊗6|ψ0i = |+, +, +, 0, 0, 0, 0i 34 35 36 |ψ5i = ÛCN OT ÛCN OT ÛCN OT |ψ4i 1 (|0, 0, +, 0, 0, 0, 0i + |0, 1, +, 1, 1, 0, 1i 2 +|1, 0, +, 1, 0, 1, 1i + |1, 1, +, 0, 1, 1, 0i) 1 = √ (|0, 0, 0, 0, 0, 0, 0i + |0, 0, 1, 0 ⊕ 1, 0 ⊕ 1, 0 ⊕ 1, 0i 8 +|0, 1, 0, 1, 1, 0, 1i + |0, 1, 1, 1 ⊕ 1, 1 ⊕ 1, 0 ⊕ 1, 1i +|1, 0, 0, 1, 0, 1, 1i + |1, 0, 1, 1 ⊕ 1, 0 ⊕ 1, 1 ⊕ 1, 1i +|1, 1, 0, 0, 1, 1, 0i + |1, 1, 1, 0 ⊕ 1, 1 ⊕ 1, 1 ⊕ 1, 0i) = 24 25 27 |ψ4i = ÛCN OT ÛCN OT ÛCN OT |ψ3i 1 = (|0, 0, +, 0, 0, 0, 0i + |0, 1, +, 0 ⊕ 1, 0 ⊕ 1, 0, 0 ⊕ 1i 2 +|1, 0, +, 1, 0, 1, 1i + |1, 1, +, 1 ⊕ 1, 0 ⊕ 1, 1, 1 ⊕ 1i) final step 5: step 4: 14 16 17 |ψ3i = ÛCN OT ÛCN OT ÛCN OT |ψ2i 1 = √ (|0, +, +, 0, 0, 0, 0i + |1, +, +, 0 ⊕ 1, 0, 0 ⊕ 1, 0 ⊕ 1i) 2 1 = √ (|0, +, +, 0, 0, 0, 0i + |1, +, +, 1, 0, 1, 1i) 2 step 3: step 2: step 1: |ψ0i = |0i⊗6|ψi |ψi = |0i let |0i → |000i 1 encoding code for Steane’s seven-qubit ECC encoding code for Shor’s nine-qubit ECC 313 316 315 so finally |ψi = |1i → |ψ5i ≡ |1iL 1 |0iL ≡ |ψ5i = √ (|0000000i + |0011110i + |0101101i + |0110011i 8 +|1001011i + |1010101i + |1100110i + |1111000i) X 1 = √ |νi 8 even ν∈Hamming Analogously where a|0i + b|1i→a|0iL + b|1iL 1 |1iL = √ (|1111111i + |1110000i + |1001100i + |1000011i 8 +|0101010i + |0100101i + |0011001i + |0010110i) X 1 = √ |νi 8 odd ν∈Hamming Thus, in general s s s s g g g g NM |0i s s s s |0i s s s 317 318 g g g g NM s bit-flip syndrome detection in Steane’s 7-qubit ECC |0i g g g g NM phase-flip syndrome detection - the same circuit but with extra Hadamard gates, as for Shor’s code fault-tolerant syndrome detection to make the circuit fault tolerant, each ancilla qubit must be replaced by four qubits in a suitable state. fault-tolerant device - a device that works effectively even when its elementary components are imperfect. s g s g g s g s ancilla data s g s g correct! g s fault-intolerant and fault-tolerant circuits data ancilla wrong! most general single-qubit error the most general single-qubit unitary error transformation (apart from irrelevant global phase factor) can be expanded to order ǫ as Ûerror = ǫiIˆ + Ô(ǫ) = ǫiIˆ + ǫxσ̂x + ǫy σ̂y + ǫz σ̂z discrete set of quantum errors fundamental result of quantum error correction theory: correcting just a discrete set of errors (bit flip, phase flip and combined bit-phase flip) a quantum error-correcting codes can correct a continuous set of errors. Shor’s and Steane’s ECCs • they protect not only against bit and phase flip errors but against arbitrary errors affecting only a single qubit • by measuring the error syndrome, the state collapses into one of the four states σ̂x|ψi, σ̂y |ψi, σ̂z |ψi, |ψi which are correctable with the codes. 319 s g 320 x ↔ Rθ g s g s ↔ i u |yi |xi |x ⊕ yi CNOT g s |xi |yi |xi 89:; ?>=< • • 89:; ?>=< • • 89:; ?>=< • • naive solution but not fault-tolerant |x ⊕ 1i |xi g s s |z ⊕ xyi |yi |xi 322 Toffoli gate Toffoli gate = controlled-controlled-NOT (CCNOT) NOT g g s 3. fault-tolerant Toffoli gate data data can also be applied bitwise with majority vote 2. fault-tolerant CNOT (XOR) = transversal CNOT Rθ Rθ Rθ can be applied bitwise with majority vote 1. fault-tolerant single qubit-gates fault-tolerant gates so far we have analyzed coding for fault-tolerant storing of quantum information, but we also need: fault-tolerant memory 321 e |z ⊕ xyi e |xi e r |yi NM NM NM 6 6 6 Z Z e r r Note: for convenience we neglect normalizations error-threshold theorem requirements for scalable QIP perfect ECC with the smallest number of ancillas (part II) Error correction codes 324 • each line represents a block of 7 qubits! • |ψ cati = √12 (|0i⊗7 + |1i⊗7), |0i = |0i⊗7, |1i = |1i⊗7, |xi = |xi⊗7, etc. • gates are implemented transversally • if a given measurement outcome is 1, the arrow points to the set of gates to be applied; no action is taken if the outcome is 0. r r r r |0i H r r r r |0i H r r e e |0i H r |ψcat i Z H e 6 NM |ψcat i Z H e NM |ψcat i Z H e NM e |xi e |yi r H |zi Shor’s construction of fault-tolerant Toffoli gate 323 Hilbert space for ECC subspaces corresponding to different errors should be orthogonal thus the total Hilbert space for ECC should be large enough to contain all the orthogonal subspaces. number of subspaces is 2(3n + 1) 325 Orthogonality requires a subspace for each of the three errors every qubit can suffer and another one for the unperturbed logical state. encoded qubits |0iL = |B1i|00i − |B3i|11i + |B5i|01i + |B7i|10i |1iL = −|B2i|11i − |B4i|00i − |B6i|10i + |B8i|01i in terms of the (unnormalized) 3-qubit Bell states: 2 |B 1 i = |000i ± |111i 4 |B 3 i = |100i ± |011i |B 5 i = |010i ± |101i 89:; ?>=< 326 • π @ABC GFED 8 @ABC GFED 7 • 6 |0iL = = = = |1iL = = = = |B1i|00i − |B3i|11i + |B5i|01i + |B7i|10i (|000i + |111i)|00i − (|100i + |011i)|11i +(|010i + |101i)|01i + (|110i + |001i)|10i |00000i + |11100i − |10011i − |01111i +|01001i + |10101i + |11010i + |00110i (lexicographic order) |00000i + |00110i + |01001i − |01111i −|10011i + |10101i + |11010i + |11100i encoded qubits explicitly Thus, all the allowed encodings have the same sign pattern: with two minus signs in one of the logical states and four in the other. Other allowed encodings can be found from those by permutations of bits and coordinated signs. 8 6 We must double this to have enough space to accommodate both logical states and their erroneous descendants. 26 ≤ / 16 for n = 4 & 32 ≤ 32 for n = 5 |B 7 i = |110i ± |001i ⇒ minimum number of encoded qubits is n = 5 for ECC 2(3n + 1) ≤ 2n 1. 5-bit encoder • • @ABC GFED • • 89:; ?>=< @ABC GFED 5 • it requires the minimum number of ancillas Los Alamos ECC ⇒ number of encoded qubits for ECC n = 9 – Shor ECC n = 7 – Steane ECC n = 5 – Los Alamos ECC H • π 4 @ABC GFED π 3 • 2 • [Laflamme, Miquel, Paz, Zurek (1996)] |ai H 1 H |bi |Qi |ci |di gates −|B2i|11i − |B4i|00i − |B6i|10i + |B8i|01i −(|000i − |111i)|00i − (|100i − |011i)|11i −(|010i − |101i)|01i + (|110i − |001i)|10i −|00011i + |11111i − |10000i + |01100i −|01010i + |10110i + |11001i − |00101i (lexicographic order) −|00011i − |00101i − |01010i + |01100i −|10000i + |10110i + |11001i + |11111i 327 328 8 7 5 4 @ABC GFED • syndrome |a′b′c′ d′ i none 0000 X3Z3 1101 X5Z5 1111 X2 0001 Z3 1010 Z5 1100 X2Z2 0101 X5 0011 Z1 1000 Z2 0100 Z4 0010 X1 0110 X3 0111 X4 1011 X1Z1 1110 X4Z4 1001 error −a|1i − b|0i −a|0i − b|1i a|0i − b|1i resulting state ′ |Q i a|0i + b|1i −a|1i + b|0i −a|0i + b|1i 2 π π 3 • • • 89:; ?>=< • errors and their syndromes 6 @ABC GFED @ABC GFED 89:; ?>=< 1 H H H input state |Qi = a|0i + b|1i, Xi – bit flip error, Zi – phase flip error gates @ABC GFED • • • @ABC GFED π • • (reverse of the encoder) Los Alamos ECC 2. decoder |b′i |a′i 330 |d′i |c′i |Q′i 329 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 R̂ = R11 R12 R21 R22 R • • double-controlled rotation: CCR 0 0 1 0 0 0 0 0 0 0 0 1 332 CCR= [ 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 R(1,1) R(2,1) 0 0 0 0 0 0 R(1,2) R(2,2)]; ˆ CCR = I+|110ih110|(R 11−1)+|110ih111|R12 +|111ih110|R21 +|111ih111|(R22 −1) CCNOT=[ 1 0 0 0 0 0 0 0 ]; CNOT=[ 1 0 0 0 ]; I=eye(2); % identity operator H=[1 1;1 -1]; % Hadamard gate - for simplicity we neglect 1/sqrt(2) R=-I; % pi-rotation Matlab program for Los Alamos ECC 331 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 R̂ = • • • R R11 R12 R21 R22 triple-controlled rotation: CCCR 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 333 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 R(1,1) R(1,2) R(2,1) R(2,2)]; 0])*(R(1,1)-1)+... 1])*R(1,2) +... 0])*R(2,1) +... 1])*(R(2,2)-1); 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 CCCR gate explicitly 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 334 ˆ CCCR = I+|1110ih1110|(R 11 −1)+|1110ih1111|R12+|1111ih1110|R21+|1111ih1111|(R22−1) [ 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 CCCR=eye(16)+... ket([1 1 1 0])*bra([1 ket([1 1 1 0])*bra([1 ket([1 1 1 1])*bra([1 ket([1 1 1 1])*bra([1 CCCR= 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 R̂ = 89:; ?>=< • 89:; ?>=< R R11 R12 R21 R22 another triple-controlled rotation: cCcR 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 R(1,2) R(2,2) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0])*(R(1,1)-1)+... 1])*R(1,2) +... 0])*R(2,1) +... 1])*(R(2,2)-1); 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 cCcR gate explicitly 0 0 0 0 R(1,1) R(2,1) 0 0 0 0 0 0 0 0 0 0 335 336 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; ˆ cCcR = I+|0100ih0100|(R 11 −1)+|0100ih0101|R12 +|0101ih0100|R21+|0101ih0101|(R22−1) [ 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 cCcR=eye(16)+... ket([0 1 0 0])*bra([0 ket([0 1 0 0])*bra([0 ket([0 1 0 1])*bra([0 ket([0 1 0 1])*bra([0 cCcR= 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 3 ]; SWAP=[ 89:; ?>=< • 1 0 0 0 = 0 0 1 0 0 1 0 0 0 0 0 1 89:; ?>=< × × × • × SWAP34=tensor_product(I,I,SWAP,I); CNOT45=tensor_product(I,I,I,CNOT); gate4=SWAP34*CNOT45*SWAP34; 5 qubit 1 2 3 4 • gate 4 gate1=tensor_product(H,H,I,H,I); gate2=tensor_product(I,CCCR); gate3=tensor_product(I,cCcR); 1 π π |di 89:; ?>=< • H • 89:; ?>=< |ci • • H |bi |Qi H |ai Los Alamos encoding circuit in Matlab • gates 1, 2, 3 338 337 5 4 3 2 1 89:; ?>=< 89:; ?>=< • = 89:; ?>=< • 89:; ?>=< • = × × 89:; ?>=< • × × ×× 89:; ?>=< • × × 339 6 7 8 • • • 89:; ?>=< π 89:; ?>=< • 340 † Ûdecoder = Ûencoder U_encoder=gate8*gate7*gate6*gate5*gate4*gate3*gate2*gate1; U_decoder=gate1*gate2*gate3*gate4*gate5*gate6*gate7*gate8; thus the encoding and decoding circuits are described by the operators gate=tensor_product(I,I,CCR); gate8=swap_qubits(gate,5,3); gate6=swap_qubits(CNOT45,5,3); gate7=swap_qubits(CNOT45,4,2); 5 4 3 1 2 • gates 6,7,8 CNOT45=tensor_product(I,I,I,CNOT); gate5 =swap_qubits(swap_qubits(CNOT45,5,3),4,1)*... swap_qubits(CNOT45,4,1); Alternatively, by defining a function swap_qubits(Û , n1, n2), which temporarily swaps n1th and n2th qubits in a matrix Û , we can simply write gate5=SWAP23*CNOT12*SWAP23*SWAP25*CNOT12*SWAP25; CNOT12=tensor_product(CNOT,I,I,I); SWAP23=tensor_product(I,SWAP,I,I); SWAP34=tensor_product(I,I,SWAP,I); SWAP45=tensor_product(I,I,I,SWAP); SWAP25=SWAP23*SWAP34*SWAP45*SWAP34*SWAP23; • gate 5 (z = −1) (z = −1) 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 U_encoder= z 0 0 0 z 0 1 0 0 0 0 1 0 1 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 U_decoder= 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 1 0 0 0 0 0 z 1 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 0 0 0 z 0 1 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 z 0 0 1 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 0 z 0 1 0 0 0 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 1 0 0 0 0 0 1 1 0 0 0 0 0 z 0 0 0 z 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 z 0 z 0 0 0 0 z 0 1 0 0 0 z 0 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 1 0 z 0 0 0 0 0 0 1 0 1 0 0 0 z 0 0 0 0 0 z 1 0 0 0 0 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 1 0 z 0 0 0 0 1 0 1 0 0 0 z 0 0 0 0 0 z 0 0 1 0 0 0 0 0 z 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 0 0 z 0 z 0 0 0 0 0 0 z 0 1 0 0 0 z 0 0 0 0 0 1 1 0 0 0 0 0 1 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 1 0 0 0 0 0 1 0 0 z 0 0 0 0 0 1 0 0 0 z 0 1 0 0 0 0 z 0 z 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 0 0 0 0 z z 0 0 0 0 0 z 0 0 0 1 0 1 0 0 0 0 0 0 1 0 z 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 0 0 0 0 z 0 0 z 0 0 0 0 0 z 0 0 0 1 0 1 0 0 0 0 1 0 z 0 0 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 1 0 0 0 0 0 1 z 0 0 0 0 0 1 0 0 0 z 0 1 0 0 0 0 0 0 z 0 z 0 0 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 z 0 z 0 0 0 0 1 0 z 0 0 0 z 0 0 0 0 0 1 0 0 z 0 0 0 0 0 z 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 1 0 z 0 0 0 0 0 0 z 0 z 0 0 0 z 0 0 0 0 0 z z 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 1 0 z 0 0 0 0 z 0 z 0 0 0 z 0 0 0 0 0 z 0 0 z 0 0 0 0 0 1 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 0 z 0 z 0 0 0 0 0 0 1 0 z 0 0 0 z 0 0 0 0 0 1 z 0 0 0 0 0 z 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 z 0 0 0 1 0 z 0 0 0 0 z 0 z 0 0 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 1 0 0 0 0 0 z 1 0 0 0 0 0 1 0 0 0 z 0 z 0 0 0 0 0 0 1 0 z 0 0 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 1 0 0 0 0 0 z 0 0 1 0 0 0 0 0 1 0 0 0 z 0 z 0 0 0 0 1 0 z 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 z 0 0 0 1 0 z 0 0 0 0 0 0 z 0 z 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 0 0 0 z 0 z 0 0 0 0 z 0 1 0 0 0 1 0 0 0 0 0 z 0 0 z 0 0 0 0 0 z 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 0 1 0 z 0 0 0 0 0 0 1 0 1 0 0 0 1 0 0 0 0 0 1 z 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 0 0 0 1 0 z 0 0 0 0 1 0 1 0 0 0 1 0 0 0 0 0 1 0 0 z 0 0 0 0 0 1 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 z 0 z 0 0 0 0 0 0 z 0 1 0 0 0 1 0 0 0 0 0 z z 0 0 0 0 0 z 0 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 1 0 0 0 0 0 z z 0 0 0 0 0 z 0 0 0 z 0 z 0 0 0 0 0 0 z 0 1 0 0 341 1 0 0 0 0 0 1 0 0 z 0 0 0 0 0 1 0 0 0 1 0 z 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 342 0 0 0 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 1 0 0 0 0 0 z 0 0 z 0 0 0 0 0 z 0 0 0 z 0 z 0 0 0 0 z 0 1 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 0 1 0 0 0 0 0 1 z 0 0 0 0 0 1 0 0 0 1 0 z 0 0 0 0 0 0 1 0 1 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 0 0 z 0 z 0 0 0 0 1 0 z 0 0 0 1 0 0 0 0 0 z 0 0 1 0 0 0 0 0 1 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 z 0 0 0 0 0 0 z 0 z 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 z 0 0 1 0 1 0 0 0 0 0 z 0 z 0 0 0 0 0 z 0 z 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 z 0 0 0 0 z 0 z 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 z 0 z 0 0 0 0 0 0 1 0 z 0 0 0 1 0 0 0 0 0 z 1 0 0 0 0 0 1 0 0 0 0 0 1 0 z 0 0 0 0 0 z 0 1 0 0 0 0 0 z 0 1 0 0 0 0 0 1 0 z 0 input state to encoder |ψini = a|00000i + b|00100i psi_in=a*ket([0 0 0 0 0])+b*ket([0 0 1 0 0]) or explicitly psi_in’ ans = [a000b000000000000000000000000000]; output state |ψouti = Ûencoder|ψini psi_out=U_encoder*psi_in 1. encoding qubit |0iL special case: a = 1, b = 0 psi_out’ ans =[100000100100000z000z010000101000] 343 344 show_state(psi_out,’|0>_L = ’) ans =|0>_L = +|00000>+|00110>+|01001>-|01111>... -|10011>+|10101>+|11010>+|11100> 2. encoding qubit |1iL input state to encoder: |ψini = |1i special case for a = 0, b = 1 psi_in’ ans = [00001000000000000000000000000000]; output state |ψouti = |1iL psi_out’ ans = [000z0z0000z01000z000001001000001] or explicitly show_state(psi_out,’|1>_L = ’) ans = |1>_L = -|00011>-|00101>-|01010>+|01100>... -|10000>+|10110>+|11001>+|11111> thus we have got |0iL and |1iL in agreement with our definition Q 345 decoder ′ |1i → −|0i |0i → −|1i A1→1 = −1 A0→1 = 0 ⇒ ⇒ |1i → −|1i |0i → |0i psi_in = ket([0 0 Q 0 0]); psi_syndrome = ket([0 1 Q_prime 1 0]); amplitude=psi_syndrome’*U_decoder*U_error*U_encoder*psi_in |Q′i → σ̂z → |Qi How to correct it? |Qi = a|0i + b|1i → |Q′i = a|0i − b|1i A1→0 = 0, A0→0 = 1, psi_syndrome=ket([1 0 Q_prime 1 0]); transition amplitudes |Qi → |Q′i syndrome sigma_z=[1,0;0,-1] U_error=tensor_product(I,I,sigma_z,I,I); error Ûerror = Iˆ ⊗ Iˆ ⊗ σ̂z ⊗ Iˆ ⊗ Iˆ example 2: phase flip of the 3rd qubit |Q′i → R̂(π)σ̂x → |Qi How to correct it? AQ→Q′ = hψsyndrome|ÛdecoderÛerrorÛencoder|ψini |ψsyndromei = |a b Q c d i = |01Q 10i ′ ′ ′ 346 ⇒ ⇒ so ⇒ ′ ′ Q so A1→1 = 0 A0→1 = −1 |Qi = a|0i + b|1i → |Q′i = −b|0i − a|1i A1→0 = −1, A0→0 = 0, AQ→Q′ transition amplitudes |Qi → |Q′i transition amplitude |a b c d i = |0110i ′ ′ ′ ′ syndrome |ψini = |abQcdi = |00Q00i input state to encoder U' U ' ≡ U a 'b 'c ' d ' c' d' sigma_x=[0,1;1,0]; U_error=tensor_product(sigma_x,I,I,I,I); error Ûerror = σ̂x ⊗ Iˆ ⊗ Iˆ ⊗ Iˆ ⊗ Iˆ Errors and their syndromes example 1: bit flip of the 1st qubit encoder Q' a' b' syndromes basic elements of the perfect ECC 1 error 348 347 Requirements for scalable QIP [Knill, Laflamme, Zurek et al., 2002] 1. Scalable physical systems: the ability to support any number of independent qubits. 2. State preparation: the ability to prepare any qubit (or at least large fraction of them) in the standard initial state |0i. 3. Measurement: the ability to measure any qubit (or at least large fraction of them) in the logical basis. 349 350 Note: sometimes the standard projective measurement can be replaced by weak measurements that return a noisy number whose expectation is the probability that a qubit is in the state |1i. 4. Errors: The error probability per gate must be below a threshold and satisfy independence and locality properties. • For the most pessimistic independent, local error models, the error threshold is above ∼ 10−6. For some special error models, the threshold is substantially higher. For example: • For the independent depolarizing error model, it is believed to be better than ∼ 10−4. • For the independent ‘erasure’ error model, where error events are always detected, the threshold is above .01. • The threshold is also above .01 when the goal is only to transmit quantum information through noisy quantum channels. 5. Quantum control: 351 the ability to implement a universal set of unitary quantum gates acting on a small number (usually at most two at a time) of qubits. • For most accuracy thresholds, it is necessary to be able to apply the quantum control in parallel to any number of disjoint pairs of qubits. This parallelism requirement can be weakened if a nearly noiseless quantum memory is available. 352 • The universality assumption can be substantially weakened by replacing some or all unitary quantum gates with operations to prepare special states or by having additional measurement capabilities. accuracy-threshold theorem Assuming the above requirements for scalable QIP: If the error per gate is less than a threshold, then it is possible to efficiently quantum compute arbitrarily accurately. This is one of the most important results in quantum ECC and fault-tolerant computation. Introduction to quantum algorithms (part I) quantum algorithms and classical cryptography public key cryptography (PKC) 354 f−1(out) = ? example but x mod 3 = 2 => x = 2, 5, 8, 11 ,14 , 17, 20, ... (not unique result) y = 17 mod 3 => y = 2 (unique result) f(in) = out, • without additional information, it is not always possible to decrypt by repeating encryption operations in reverse order 2. private, confidential key 1. public, open key two keys in PKC and independently by Diffie & Hellman (1976) invented by Ellis (1970, British CESG) public-key cryptography (PKC) = asymmetric cryptography = non-secret encryption 1997 Grover algorithm for searching databases: How to search the keys more effectively? 1994 Shor algorithm for finding discrete logarithms: How to break ElGamal cryptosystem? plaintext 1994 Shor algorithm for number factorization: How to break cryptosystems of RSA, Rabin, Williams, Blum-Goldwasser,…? cleartext plaintext ciphertext plaintext cleartext 355 Decryption using key k2 (private key) Encryption using key k1 (public key) 356 Decryption using the same key k key distribution Encryption using key k asymmetric algorithms = public-key algorithms cleartext plaintext ciphertext cleartext symmetric algorithms 1985 Deutsch (Deutsch-Jozsa / DJ) algorithm: How to see both sides of a coin simultaneously? quantum algorithms and cryptography 353 e c secure channel unsecured channel Eve 357 358 plaintext decryption d=e-1 Ed(c)=m m Bob Encryption with a secure channel for key exchange Alice key source e encryption Ee(m)=c m plaintext c unsecured channel unsecured channel Eve Why do we need PKC? Number of keys a problem of huge number of keys for symmetric algorithms. N (N −1) 2 359 360 • How many keys should be generated for N correspondents if everyone wants to communicate with all others using symmetric algorithms? nsym(N ) = (N2 ) = • How many keys are required for 6 or 1 milion correspondents? 1012 2 nsym(6) = 15 nsym(106) ≈ nasym(N ) = 2N • How many keys are required asymmetric algorithms? ElGamal computational problem integer factorization problem popular public-key encryption schemes based on integer factorization problem decryption d=e-1 Ed(c)=m m generalized ElGamal public-key encryption scheme RSA (Rivest-Shamir-Adleman & Cocks) Rabin Williams Blum-Goldwasser probabilistic Goldwasser-Micali probabilistic plaintext Bob key source Encryption with a unsecured channel for key exchange Alice e encryption Ee(m)=c m plaintext integer factorization problem integer factorization problem integer factorization problem quadratic residuosity problem or integer factorization problem discrete logarithm problem or integer factorization problem generalized discrete logarithm problem or integer factorization problem 361 p log n(log log n)) our estimations Nclas(n) = c3 exp[c(log n)1/3(log log n)2/3] by having only one key it is practically impossible to calculate the other key using practically available computers over practically long period of time. 364 363 for simplicity, we choose: c3 = 1 with c = 2 in the fastest version of number field sieve method due to Lenstra NShor(n) = c1(log n)2 log(log n) + c2 log n hard problems in PKC Thus, we estimate numbers of operations required to factorize n as: and log n steps of post processing on a classical computer. • practically, exceeding the specified bound on the number of operations or memory corresponding to a specified security parameter • require super-polynomial time or space ∼ (log n)2 log log n • optimized Shor algorithm ∼ (log n)3 • original Shor algorithm polynomial in time, poly(log n) hard = computationally infeasible – perhaps seconds or milliseconds. • practically, within a certain number of machine operations or time units • in polynomial time and space effectiveness of quantum factorization methods exponential in time, exp(log n) should be interpreted relative to a specified frame of reference. easy = computationally feasible effectiveness of classical factorization methods This is the fastest publicly available ;-) classical method for r ≥ 150. 5. number field sieve method p ∼ exp(c 3 log n(log log n)2) ∼ exp(c thus 3,4. Fermat method and quadratic sieve method √ ∼ exp(c r log r), where r ∼ log n is the number of bits ∼ n1/4 log3 n 2. Monte Carlo method (Pollard method) 1. direct method √ ∼ n = exp( 21 log n) What is the time or number of bit operations required to factorize n? Motivation for Shor’s Algorithm – effectiveness of classical integer factorization methods “easy” and “hard” computational problems 362 computational problem linear code decoding problem or error-correction-code problem Merkle-Hellman knapsack subset sum problem cracked by Shamir & Zippel! Graham-Shamir knapsack subset sum problem cracked! Lu-Lee knapsack subset sum problem cracked by Adleman & Rivest! Goodman-McAuley knapsack subset sum problem cracked! Powerline System knapsack subset sum problem (a simple version of Chor-Rivest) cracked by Lenstra! Chor-Rivest knapsack subset sum problem publicly not cracked! public-key encryption scheme McEliece popular public-key encryption schemes not based on integer factorization problem examples of factorization times MIPS = millions of instructions per second • 1-10 MIPS in a modest PC • hundreds-thousands of MIPS in a supercomputer /60 min /60 hours /24 days /365 years • say, N ′ = 106 MIPS are available in contemporary computers /N’ sec ⇒ Nclas ∼ 7.0 · 1020 MIPS → 22.5 mln years Nclas ∼ 3.5 · 1012 MIPS → 40.5 days ≈ 1 month Nclas ∼ 5.9 · 1013 MIPS → 687.5 days ≈ 2 years ⇒ ⇒ time required to factorize n t(n)= N(n) examples 1. n = 10130 2. n = 10150 3. n = 10300 How old is the universe? Tuniverse ∼ 1.25 · 1010 = 12.5 billion years What is the largest factorizable integer within Tuniverse? ⇒ T ≈ 3.17 · 1010 > Tuniverse 1. direct method N = 1060 ⇒ T ≈ 1.6 · 1010 > Tuniverse 2. Monte Carlo method (Pollard method) N = 1091 ⇒ T ≈ 1.36 · 1010 > Tuniverse 3,4. Fermat method and quadratic sieve method N = 1094 ⇒ ⇒ T ≈ 1.18 · 1010 < Tuniverse T ≈ 8.1 · 1010 > Tuniverse 5. number field sieve method N = 10400 N = 10375 365 366 Introduction to quantum algorithms (part II) classical RSA algorithm quantum Deutsch algorithm quantum Deutsch-Jozsa (DJ) algorithm quantum Hadamard transform quantum Fourier transform Rivest-Shamir-Adleman (RSA) algorithm (1978) I. generation of RSA keys (by Bob) and φ = (p − 1)(q − 1) 1. choose two large primes p 6= q 2. calculate n = pq 3. choose randomly an integer e (1 < e < φ) coprime to φ, i.e. their greatest common divisor is one, GCD(e, φ) = 1. n - modulus 367 368 4. using the extended Euclidean algorithm, calculate d (1 < d < φ) such that ed ≡ 1 mod φ 5. thus public key - (n, e) private key - d • cryptographic terms: e – encryption exponent d – decryption exponent 2= 32 10 2 0 116 mod 42 = 42 mod 32 = 32 mod 10 = 10 mod 2 = Answer: 2 a b x2 x1 y2 y1 372 a,b,x,y,d - integers GCD(116,42) = d = ax + by ? initial x1=y2=0; x2=y1=1 while b>0 q = Int(a/b) r =a-qb x1 = x2 -q x1 ; x2 = x1 y1 = y2 -q y1 ; y2 = y1 a=b b=r return (d,x,y)=(a, x2, y2) extended Euclidean algorithm GCD(a,b) = d =ax+by x,y,d=? 371 42 1 0 0 42 32 = 0 1= 1 [1 1-2 11 6 16 *0 mo /4 d 2] 42 116 initial values -2 = 0-2 *1 1 ------------------------------------- q GCD(116,42) = ? how to calculate the greatest common divisor (GCD)? Euclidean algorithm 370 It has been revealed only very recently that the RSA algorithm has been devised already in 1973 by Cocks for the British security agency CESG. A note m = cd mod n use your private key d to calculate III. RSA decryption (by Bob) 3. send a cipher c to Bob c = me mod n 2. calculate m ∈ h0, n − 1i 1. change cleartext into plaintext represented by an integer encrypt your message with the public key (n, e) II. RSA encryption (by Alice) 369 a b x2 x1 y2 y1 373 QCD(116,42) = d = ax + by ? q 42 0 1 1 0 -2 1 0 3 -2 1 374 ------------------------------------116 32 -1 42 1 2 10 -11 32 3 1 4 2 -1 10 58 3 -21 0 5 -11 || y 4 || x 2 || d answer: multiplicative or modular inverse of an integer problem: 4-1 mod 9 = ? 4y=1 mod 9 => 9x + 4y =1 mod 9 q a b y2 y1 -------------------9 4 0 1 2 4 1 1 -2 4 1 0 -2 . so 9*1 + 4*(-2)=1 & y=-2=7 mod 9 thus 7 is the inverse of 4 mod 9 4*7=1 mod 9 inverse of an integer modulo 9 problem: 8-1 mod 9 = ? 8y=1 mod 9 => 9x + 8y =1 mod 9 q a b y2 y1 -------------------9 8 0 1 1 8 1 1 -1 8 1 0 -1 . so 9*1 + 8*(-1)=1 & y=-1=8 mod 9 thus 8 is the inverse of itself mod 9 8=8-1 mod 9 8*8 = 1 mod 9 inverse of integers mod 9 2-1 mod 9 = 5 5-1 mod 9 = 2 1-1 mod 9 = 1 3-1 mod 9 = ? 3y=1 mod 9 9x + 3y =1 mod 9 3(3x + y) =1 mod 9 => no solution 6-1 mod 9 = ? => no solution all invertible elements modulo 9 Z*9={1,2,4,5,7,8} n is invertible GCD(n,9)=1 375 376 0 1 -12 37 y2 1 -12 37 . y1 2 2 2 2 2 5 GCD(φ, e) =GCD(160, 13) =GCD(25 · 5, 13) = 1 | | | | | | | 02 12 22 32 A˛ I P Y 03 13 23 33 B J q Z 04 14 24 34 C K R Ż 05 15 25 35 Ć L S Ź 06 16 26 36 D Ł Ś _ 07 17 27 37 E M T - E˛ N U ? 09 19 29 39 F Ń v , c = me mod n = 11713 mod 187 =? 2. Alice calculates cipher using the public key (n, e) = (187, 13): m = (01, 17) = 117 1. Alice converts her cleartext into the plaintext: A H Ó X 08 18 28 38 10 20 30 40 Assume that Alice wants to encrypt the following cleartext: “AM” :-) II. RSA encryption public key (n, e) = (187, 13) private key d = 37 or vice versa public key (n, d) = (187, 37) private key e = 13 5. so Bob has generated the keys: de = 13 · 37 = 481 ≡ 1 (mod 160) 01 11 21 31 OK 13 4 1 0 e 160 80 40 20 10 5 • inefficient direct method 1 ⇒ d = 37 160 13 4 1 φ 4. he double checks that d is the multiplicative inverse of e: ⇒ 12 3 4 q′ using, for example, the following public alphabet: 378 GCD(160, 13) = 1 3. Bob calculates d using the extended Euclidean algorithm ⇒ 160 mod 13 = 4 13 mod 4 = 1 4 mod #1 = 0 • efficient Euclidean algorithm 2. Bob checks whether φ are e are coprime: φ = (p − 1)(q − 1) = 10 · 16 = 160 n = pq = 11 · 17 = 187 1. Bob calculates (artificially small numbers so completely insecure) Bob chooses p=11, q=17, e=13 I. RSA key generation Example of RSA application • Mathematica: b=PowerMod[a,k,n] return(b) end end b = b · Aki mod n A = A2 mod n for i = 1 to j do begin b = Ak0 i i=0 ki 2 Pj if n ∈ I; a, k ∈ Zn convert k into binary form k = A=a else begin if k = 0 then b = 1 Algorithm b = ak mod n =? How to calculate powers modulo n? 377 G O W . 380 379 13 = (1101)2 i 0 1 2 3 ei 1 0 1 1 382 ≡135 1352 ≡86 (mod 187) A2 117 1172 ≡38 c 117 117 117*135≡87 87*86≡2 (mod 187) so the cipher is c = 2 • Mathematica: PowerMod[117, 13, 187]֒→ 2 III. RSA decryption 1. Bob calculates m = cd mod n = 237 mod 187 0 1 2 2 1 2 3 4 5 0 1 0 0 1 4 16 162 ≡69 692 ≡ 86 862 ≡ 103 (mod 187) 2 32 32 32 32*103 ≡ 117 (mod 187) d = 37 = (100101)2 i di A2 m 2. thus he finds Alice’s message to be m = 117 = (01, 17) = „AM” lengths of public keys 381 382 1. The above PKC examples were given for artificially small numbers and thus such cryptosystems are not secure at all. 2. To insure security of a PKC system, the lengths of public keys should be of hundreds of decimal digits. 3. To determine the required key length you should consider: • intended security • lifetime of the key • current state-of-the-art of factoring. 4. The wise cryptographer is ultra-conservative when choosing publickey key lengths as history teaches us a lot: • “I shall be surprised if anyone regularly factors numbers of size 1080 without special form during the present century” (R. Guy, 1976). • “Factoring a 125-digit number would take 40 quadrillion years”, tj. 4 · 1019 lat (R. Rivest, 1977). • ... but already in 1994 a 129-digit number was factorized. Recommended lengths of public keys length in bits (a) (b) (c) 768 1280 1536 1024 1280 1536 1280 1536 2048 1280 1536 2048 1536 2048 2048 => 383 length in decimal digits (a) (b) (c) 231 386 463 308 386 463 386 463 617 386 463 617 463 617 617 to be secure against attacks of (a) a single person, (b) private agencies, (c) national security agencies: year 1995 2000 2005 2010 2015 [Bruce Schneier: “Applied Cryptography”] 384 • American National Security Agency (NSA) recommends key lengths from 512 up to 1024 bits (so from 154 do 308 decimal digits) in their Digital Signature Standard (DSS). integer length (number of digits) length Lb of integer n with base b is given by Lb = [ logb n]] + 1 = [ ln n/ ln b]] + 1 How Eve can crack the RSA cryptosystem? By factorizing n: If Eve can factorize n = pq then she can calculate φ = (p − 1)(q − 1) and thus can calculate Bob’s private key d Why is the RSA believed to be secure for large integers? There is seemingly no efficient (i.e., polynomial-time and polynomial-space) classical algorithm for integer factorization. RSA hypothesis (1978) Any general method of cracking the RSA cryptosystem which enables finding private key d from public key (n, e) requires an efficient algorithm for integer factorization. Note There is still no proof of this hypothesis. Shor’s algorithm a Deutsch problem on a DNA computer. Adleman’s algorithm constant f1(0) = 0 f2(0) = 1 balanced f3(0) = 0 f4(0) = 1 f1(1) = 0 f2(1) = 1 f3(1) = 1 f4(1) = 0 balanced and constant functions Can you check whether a coin is real or fake by a single observation? or How many questions one should ask the oracle to check whether f is constant or balanced? Given an oracle (black box) that computes a Boolean function f (x) : {0, 1} → {0, 1} • or on a large quantum computer • 386 would be insecure if we could implement efficiently: RSA and majority of public-key cryptosystems 385 f • X H H f • H |−i |0i H H H H H f • (−1)f (x) Zk f • f • = Uf • = 388 |−i Uf |0i for constant f & |1i for balanced f |δk1i (k = 0, 1) |−i |0i for constant f & |1i for balanced f |0i − |1i different symbols for the same gate or |0i main idea |−i |0i or 1 query + 2 auxiliary operations |1i |0i H f (0) ⊕ f (1) f 387 |0i for constant f & |1i for balanced f 1 • 1 query + 3 auxiliary operations quantum solution 1 0 2 queries + 1 auxiliary operation classical solution 0 1 ψ1 H H x y Uf x y ⊕ f ( x) ψ3 Deutsch algorithm ψ2 Deutsch algorithm 1. prepare input state |ψ1i = |0i|1i H 2. apply Hadamard gates (create equal superposition) |0i + |1i |0i − |1i √ |ψ2i = Ĥ ⊗2|ψ1i = |+i|−i = √ 2 2 1 = (|00i − |01i + |10i − |11i) 2 |xi|yi → Ûf → |xi|y ⊕ f (x)i 3. apply Ûf gate |ψ3i = Ûf |ψ2i ψ4 389 390 ∼ Ûf |00i − Ûf |01i + Ûf |10i − Ûf |11i = |0, 0 ⊕ f (0)i − |0, 1 ⊕ f (0)i + |1, 0 ⊕ f (1)i − |1, 1 ⊕ f (1)i We neglect normalization thus sign ‘∼’ is used. |0, 0 ⊕ 0i − |0, 1 ⊕ 0i + |1, 0 ⊕ 0i − |1, 1 ⊕ 0i |0, 0i − |0, 1i + |1, 0i − |1, 1i |0i(|0i − |1i) + |1i(|0i − |1i) |+i|−i case I: f (0) = f (1) = 0 |ψ i ∼ 3 = = ∼ |ψ3i ∼ = = ∼ |0, 0 ⊕ 1i − |0, 1 ⊕ 1i + |1, 0 ⊕ 1i − |1, 1 ⊕ 1i |0, 1i − |0, 0i + |1, 1i − |1, 0i −|0i(|0i − |1i) − |1i(|0i − |1i) −|+i|−i case II: f (0) = f (1) = 1 so f (0) = f (1) ⇒ |ψ3i = ±|+i|−i 391 392 |0, 0 ⊕ f (0)i − |0, 1 ⊕ f (0)i + |1, 0 ⊕ f (1)i − |1, 1 ⊕ f (1)i |0, 0 ⊕ 0i − |0, 1 ⊕ 0i + |1, 0 ⊕ 1i − |1, 1 ⊕ 1i |0, 0i − |0, 1i + |1, 1i − |1, 0i |0i(|0i − |1i) + |1i(|1i − |0i) |0i|−i − |1i|−i |−i|−i case III: f (0) = 0, f (1) = 1 |ψ i ∼ 3 = = = ∼ ∼ |0, 0 ⊕ f (0)i − |0, 1 ⊕ f (0)i + |1, 0 ⊕ f (1)i − |1, 1 ⊕ f (1)i |0, 0 ⊕ 1i − |0, 1 ⊕ 1i + |1, 0 ⊕ 0i − |1, 1 ⊕ 0i |0, 1i − |0, 0i + |1, 0i − |1, 1i −|0i(|0i − |1i) − |1i(|1i − |0i) −|−i|−i case IV: f (0) = 1, f (1) = 0 |ψ i ∼ 3 = = = ∼ so f (0) 6= f (1) ⇒ |ψ3i = ±|−i|−i 393 H • H |0i or |1i Uf H is a common pattern in quantum algorithms H • lower qubit is an ancilla – it is not measured and can be discarded after function evaluation • a sequence of gates Hadamard - function evaluation - Hadamard f |−i |−i • relative phases are introduced by the function evaluation |0i • top qubit undergoes a single-qubit interference (−1)f (x) Some remarks on Deutsch algorithm QED x = 0 ⇒ f (0) = f (1) x = 1 ⇒ f (0) 6= f (1) 5. measure only the 1st register |ψ4i = ±|f (0) ⊕ f (1)i|−i ≡ |xi|yi thus we have for any case: 394 ˆ f (0) 6= f (1) ⇒ |ψ4i = ±Ĥ ⊗ I|−i|−i = ±|1i|−i = ±|f (0) ⊕ f (1)i|−i cases III & IV: ˆ f (0) = f (1) ⇒ |ψ4i = ±Ĥ ⊗ I|+i|−i = ±|0i|−i = ±|f (0) ⊕ f (1)i|−i cases I & II: ˆ 3i |ψ4i = Ĥ ⊗ I|ψ 4. apply Hadamard gates 0 ψ1 H H ⊗n ψ2 y x Uf y ⊕ f ( x) x ψ3 H ⊗n |ψ2i = Ĥ ⊗(2+1)|ψ1i = |+i⊗2|−i |0i + |1i |0i + |1i |0i − |1i √ √ √ = 2 2 2 1 |0i − |1i = (|00i + |01i + |10i + |11i) √ 2 2 ∼ |000i + |010i + |100i + |110i −(|001i + |011i + |101i + |111i) 2. apply Hadamard gates |ψ1i = |0i⊗2|1i = |001i 1. prepare input state Deutsch-Jozsa (DJ) algorithm for 3 qubits 1 ⊗n ψ4 Deutsch-Jozsa (DJ) algorithm 396 395 3. apply Ûf gate |ψ3i ∼ |00, 0 ⊕ f (00)i + |01, 0 ⊕ f (01)i +|10, 0 ⊕ f (10)i + |11, 0 ⊕ f (11)i − |00, 1 ⊕ f (00)i + |01, 1 ⊕ f (01)i +|10, 1 ⊕ f (10)i + |11, 1 ⊕ f (11)i 4. apply Hadamard gates ˆ 3i |ψ4i = Ĥ ⊗2 ⊗ I|ψ *. let’s analyze all cases for different functions f number of cases Ncases = C04 + C44 + C24 = 1 + 1 + 6 = 8 where |{00, 01, 10, 11}| = 4 Cnm – binomial coefficient case 1: (trivial) f (00) = f (01) = f (10) = f (11) = 0 |ψ4i = (Iˆ ⊗ Ĥ)|ψ1i = |00−i |ψ3i = Ûf |ψ2i = Ûf |00i + |10i + |01i + |11i |0i − |1i = |00i + |10i + |01i + |11i |1i − |0i = −|ψ2i f (00) = f (01) = f (10) = f (11) = 1 case 2: (trivial) so |ψ4i = −(Iˆ ⊗ Ĥ)|ψ1i = −|00−i 397 398 case 3: f (00) = 0, f (01) = 0, f (10) = 1, f (11) = 1 = |000i + |010i + |101i + |111i −(|001i + |011i + |100i + |110i) = (|00i − |10i + |01i − |11i)(|0i − |1i) = (|0i − |1i)(|0i + |1i)(|0i − |1i) ∼ |−, +, −i 399 −(|00, 1 ⊕ 0i + |01, 1 ⊕ 0i + |10, 1 ⊕ 1i + |11, 1 ⊕ 1i) |ψ3i ∼ |00, 0 ⊕ 0i + |01, 0 ⊕ 0i + |10, 0 ⊕ 1i + |11, 0 ⊕ 1i so 400 balanced answer constant ˆ 3i ∼ (Ĥ ⊗2 ⊗ I)|−, ˆ |ψ4i = (Ĥ ⊗2 ⊗ I)|ψ +, −i = |1, 0, −i |x, yi |0−i |0−i |2−i |1−i |3−i |3−i |1−i |2−i all cases in DJ algorithm for 3 qubits x ≡ (x1, x2) ∼ = 2x1 + x2 f (x) is balanced f (x) is constant case f(00) f(01) f(10) f(11) |x1, x2, yi 1. 0 0 0 0 |00−i 2. 1 1 1 1 |00−i 3. 0 0 1 1 |10−i 4. 0 1 0 1 |01−i 5. 0 1 1 0 |11−i 6. 1 0 0 1 |11−i 7. 1 0 1 0 |01−i 8. 1 1 0 0 |10−i ⇒ ⇒ 5. measure x if x = 0 if x > 0 x=0 |xi x X = x∈{0,1}n X = x1=0 x2=0 1 X 1 X ... xn =0 1 X = x=0 N −1 X 1 X Ĥ ⊗n|0i⊗n = Ĥ ⊗n|0i = √ |xi N x 1 2 where N = 2n and x = 2nx1 + 2n−1x2 + . . . xn in general ≡ x∈{0,1} 3 X |0i + |1i |0i + |1i √ √ Ĥ ⊗2|00i = 2 2 1 = (|00i + |01i + |10i + |11i) 2 1 X ≡ |xi 2 2 1. How to generate equal (equally-weighted) superposition quantum Hadamard transform end; 402 psi4=Norm*tensor_product(H,H,H)*psi3; % for clarity we add extra H gate vector2ket(psi4) psi3=ket([0,0,f00])+ket([0,1,f01])+ket([1,0,f10])+ket([1,1,f11])+... -(ket([0,0,1-f00])+ket([0,1,1-f01])+ket([1,0,1-f10])+ket([1,1,1-f11])); f00=f(n,1);f01=f(n,2);f10=f(n,3);f11=f(n,4); H=[1 1;1 -1];Norm=1/8; f=[ 0 0 0 0 1 1 1 1 0 0 1 1 0 1 0 1 0 1 1 0 1 0 0 1 1 0 1 0 1 1 0 0 ]; for n=1:8, Matlab program 401 1 403 1 1 X Ĥ|x1i = √ (−1)x1·z1 |z1i 2 z1=0 1 |x1x2...xni = √ N z1 ,z2,...,zn=0 (−1)x1·z1+x2·z2+...+xn·zn |z1z2...zni or compactly |xi = |x1, x2, ..., xni |zi = |z1, z2, ..., zn i hx|zi ≡ x · z = x1z1 + x2z2 + ... + xnzn is the scalar product. and x = (x1, x2, ..., xn), z = (z1, z2, ..., zn), 1 X Ĥ ⊗n|xi = √ (−1)hx|zi|zi N z where N = 2n is the Hilbert-space dimension Ĥ ⊗n 1 X 1 1 1 XX √ (−1)x1·z1+x2·z2 |z1z2i = 2 2 z1 =0 z2=0 • thus in general for n qubits where 1 |0i − |1i 1 X √ (−1)1·z |zi =√ 2 2 z=0 |0i + |1i 1 X √ (−1)0·z |zi =√ 2 2 z=0 1 1 X X 1 1 Ĥ ⊗2|x1x2i = √ (−1)x1·z1 |z1i √ (−1)x2·z2 |z2i 2 z1 =0 2 z2=0 • for two qubits combining together Ĥ|1i = Ĥ|0i = 404 2. How to compactly write the quantum Hadamard transform • for a single qubit DJ algorithm for n qubits 1. initialize (n + 1)-qubit state |ψ1i = |0i⊗n|1i |xi|−i 2. generate equal superposition by applying Hadamard gates X N x 1 |ψ2i = Ĥ ⊗(n+1)|ψ1i = (Ĥ ⊗n|0i⊗n)|−i = √ 3. calculate f by applying Ûf gate 1 X (−1)f (x)|xi|−i |ψ3i = Ûf |ψ2i = √ N x DJ algorithm for n qubits 4. apply Hadamard gate ˆ 3i |ψ4i = (Ĥ ⊗n ⊗ I)|ψ is constant is balanced 1 X Ĥ ⊗n|xi = √ (−1)f (x) Ĥ ⊗n|xi |−i N x 1 XX = (−1)hx|zi+f (x)|zi|−i N x z 5. measure |zi : z = 0 ⇒ f (x) z > 0 ⇒ f (x) 405 406 quantum Fourier transform (QFT) (N = 2n) • QFT on group (Z2)n = quantum Hadamard transform 1 X QF T ′|xi = Ĥ ⊗n|xi = √ (−1)hx|yi|yi, N y group Z2 = the set {0, 1} with addition modulo 2 (⊕) group (Z2)n = the set {0, 1}n with addition modulo 2 (⊕) bit by bit • QFT on group ZN 2πi 1 X exp xy |yi QF T |xi = √ N N y or just without the bold face: N −1 1 X 2πi xy |yi QF T |xi = √ exp N N y=0 group ZN = the set {0, 1, ..., N − 1}n with addition modulo N (⊕) Quantum Fourier Transform N −1 1 X 2πi |xi → QF T → √ exp xy |yi N N z=0 or compactly N −1 1 X xy β |yi |xi → QF T → √ N z=0 where 2πi β = exp N QFT of arbitrary pure state of n qubits N −1 N −1 N −1 X 1 X X |ψi = cx|xi → QF T → √ cxβ xy |yi N x=0 y=0 x=0 407 408 = exp 2πi N 2 1 1 1 0 1−i = √ QF T |ψi = QF T √ 0 0 2 2 2 1+i 1 eiπ/4 1 e−iπ/4 |01i + |11i = √ |00i + 2 2 2 |ψi = |00i + |11i 1 √ = √ [1001]T 2 2 π =i N = 22 = 4 ⇒ s = 3, β = exp i 2 1 1 1 1 1 1 1 1 1 1 β β2 β3 1 1 i −1 −i QF T = √ 2 4 6 = 2 1 β β β 1 −1 1 −1 4 1 β3 β6 β9 1 −i −1 i example: QFT of a Bell state s = N − 1 = 2n − 1 β where QF T |ψi = x=0 cx|xi 1 1 1 ... 1 c0 2 s 1 β β ... β c1 1 1 β 2 β 4 ... β 2s c2 √ s+1 .. .. .. . . . .. .. s 2s ss 1 β β ... β cs |ψi = N −1 X Quantum Fourier Transform in matrix representation 410 409 |x0i |x1i |x2i |x3i H H H H • R2 QFT for n=4 |x0i |x1i |x2i QFT for n=3 |x0i |x1i " • R2 # H • R3 • R2 • R2 • R4 • R3 H 1 1 1 , =√ 2 1 −1 H including SWAP where |x0i |x1i H QFT for n=2 |x0i QFT for n=1 H • R2 • R2 × |y0i H • R3 H H |y2i |y1i |y0i • R2 0 1 1 =√ i 2 0 exp 2π k 2 × |y1i Rk |y1i |y0i Circuits for QFT H 412 411 |y3i |y2i |y1i |y0i H S • T • H S • H |y2i |y1i |y0i Seemingly another implementation of QFT for n=3 |x2i |x1i |x0i # " R̂3 = T̂ 1 0 T̂ = 0 exp(iπ/4) # given in terms of the phase gate (S gate) and π/8 (sic!) gate (T gate) " 10 Ŝ = , 0 i but it is exactly our scheme as R̂2 = Ŝ, 413 414 problem for Grover’s algorithm (the needle in a haystack problem) assumptions • given an oracle which calculates f (x): f (x) = 1 if x = x0 f (x) = 0 if x 6= x0 • f (x) can be calculated using ordinary (reversible) computer code. problem find the element x0 in the least number of oracle queries. average number of evaluations of f classically – N/2 quantumly – ??? examples • searching a phone book for a name and phone number • searching a database for a key to crack a cryptosystem oracle queries 1. oracle query in DJ algorithm f Ûf = |xi|yi 7→ |xi|x ⊕ yi a note and f (x) is encoded in the sign of the control register |xi. Introduction to quantum algorithms (part III) Grover’s algorithm for searching database 2. oracle query in Grover’s algorithm 415 416 in fact, the content of the target register |yi is unchanged in the DJ algorithm Shor’s algorithm for integer factorization Ûf |xi = (−1)f (x)|xi or Ûf = 1 − 2|x0ihx0| it is just equivalent to the query in DJ algorithm ⇒ Ûf → Ûf′ 1 |ψII i = √ |xi. N x=0 N −1 X apply Hadamard gate Ĥ ⊗n to all qubits to get II. Generation of equally-weighted superposition |ψI i = |0i⊗n prepare n qubits in state I. Initialization ALGORITHM: find x0 among N = 2n elements encoded by n-qubits PROBLEM: n-Qubit Grover’s algorithm |yi = |−i Note: |xi → Uf′ → (−1)f (x)|xi 2. quantum phase-oracle Ûf′ |xi|yi → Uf → |xi|y ⊕ f (x)i f : {0, 1} → {0, 1} 1. quantum oracle Ûf of a Boolean function oracle = a black-box unitary operation two types of quantum oracles 418 III. Grover iteration √ repeat the following subroutine Int(π N /4) times: 419 |2> |3> |2> |3> −1 |1> −1 |0> −0.5 0 0.5 1 −0.5 0 0.5 1 1. oracle phase flip of |x0i −1 |1> −1 |0> −0.5 0 0.5 1 −0.5 0 0.5 1 initialization |1> |2> |0> |1> |2> 2. Hadamard gate |0> Hadamard gate |3> |3> two-qubit Grover’s search of |x0i = |2i 420 operations 2,3,4 are called the inversion about the average Note: 4. Apply Hadamard gate Ĥ ⊗n. f0 : |xi 7→ −(1 − 2δ0,x)|xi. 3. Flip the phase of all eigenstates |xi except |0i: 2. Apply Hadamard gate Ĥ ⊗n. where δy,x is Kronecker delta. fx0 : |xi 7→ (1 − 2δx0,x)|xi, 1. Call the oracle to flip the phase of eigenstate |x0i: <x|ψ> <x|ψ> 417 <x|ψ> <x|ψ> two-qubit Grover’s search of |x0i = |2i 4. Hadamard gate 3. phase flip of all |xi except |0i III.1 f Xo III.2 H H III.3 f0 421 III.4 H H 0 −0.5 −1 1 0.5 0 −0.5 −1 1 0.5 0 −0.5 −1 |0> |3> |4> |5> initialization |2> |6> |2> |3> |4> |5> |6> |7> |7> 0 −0.5 −1 1 0.5 0 −0.5 −1 |0> |0> |1> |2> |2> |3> |3> |4> |4> |5> |5> |6> |6> I.2 Hadamard gate |1> |7> |7> three-qubit Grover’s search of |x0i = |2i Hadamard gate |1> |1> |2> |3> |4> |5> |6> |7> |1> |2> |3> |4> |5> |6> |7> 1 0.5 0 −0.5 −1 1 0.5 0 −0.5 −1 |0> |0> |1> |2> |2> |3> |3> |4> |4> |5> |5> |6> |6> II.2 Hadamard gate |1> |7> |7> 423 424 three-qubit Grover’s search of |x0i = |2i I.4 Hadamard gate - end of 1st cycle I.3 phase flip of all |xi except |0i |0> I.1 oracle phase flip of |1> 1 |0> 1 |3> 1 |2> 1 |1> 0.5 0 −0.5 −1 |0> 0.5 |3> 0.5 |2> step II H H |x0i 0.5 0 −0.5 |1> - end of 1st Grover iteration √ Nqueries = Int(π N /4) where N = 2n = 4 so Nqueries = 1 422 recommended number of repetitions (or queries) −1 <x|ψ> quantum circuit for two-qubit Grover’s search qubit A qubit B |0> II.1 oracle phase flip of |x0i 1 0.5 0 −0.5 −1 <x|ψ> <x|ψ> <x|ψ> <x|ψ> |0> <x|ψ> <x|ψ> <x|ψ> <x|ψ> <x|ψ> |0> |1> |2> |3> |4> |5> |6> |7> |1> |2> |3> |4> |5> |6> |7> |0> |1> |2> |3> |4> |5> |0> |1> |2> |3> |4> |5> end of 5th cycle |6> |7> |6> |7> −1 −0.5 0 0.5 1 |0> |0> |2> |3> |4> |5> |1> |2> |3> |5> 426 |4> end of 6th cycle |1> c_0 + (k) c5 |101i + (k) c6 |110i (k) c_1 c_2 c_3 c_4 + c_5 (k) c7 |111i (k) + c2 |010i + c3 |011i after the kth Grover we get: (k) +c4 |100i + state (k) c1 |001i c_6 c_7 ] |7> |7> |ψ (6)i =[ -.01 |ψ (5)i =[ -.25 |ψ (4)i =[ -.38 |ψ i =[ -.31 -.01 -.25 -.38 -.31 -.09 .18 -.01 -.25 -.38 -.31 -.09 .18 (6) c5 -.74 -.11 .57 .97 .88 -.01 -1.00 -.25 -.38 -.31 -.09 .18 = h110|ψ (6)i = .99989 · · · ≈ 1 after the 6th iteration we get -.01 -.25 -.38 -.31 -.09 |ψ (2)i =[ -.09 (3) .18 .18 |ψ i =[ (1) -.01 -.25 -.38 -.31 -.09 .18 -.01 ] -.25 ] -.38 ] -.31 ] -.09 ] .18 ] __________________________________________________________ |ψ i =[ (k) |ψ (k)i = (k) c0 |000i |6> |6> end of 3rd cycle 425 numerical data for three-qubit Grover’s search of |x0i = |5i −1 |0> −1 0 0.5 1 −0.5 <x|ψ> −0.5 0 0.5 1 end of 4th cycle −1 −1 0 0.5 1 −1 0 0.5 1 II.4 gate Ĥ ends 2nd cycle −0.5 <x|ψ> −0.5 0 0.5 1 II.3 phase flip all |xi except |0i <x|ψ> −0.5 <x|ψ> three-qubit Grover’s search of |x0i = |2i <x|ψ> <x|ψ> O(N √ ) classical search algorithms O( N ) Grover’s algorithm How many oracle queries are required to search an N -element database? Mathematica: Sqrt[2ˆ56]/10.ˆ6/60 ֒→ 4.47 min • quantum computer needs about 4 minutes • classical computer needs 1,000 years assuming that 1 mln keys per second can be checked than What is the time required to find the correct DES key? 428 requires basically to search among N = 256 = 7.2 × 1016 possible keys. Attack on DES Grover’s algorithm and cryptography - it is the Digital Signature Standard (DSS) of US Federal Government - applied for digital signatures - asymmetric 3. DSA (Digital Signature Algorithm) ↔ Shor’s algorithm - applied for encryption/decryption and digital signatures - asymmetric 2. RSA (Rivest-Shamir-Adleman) ↔ Shor’s algorithm - used by US army - American and international cryptographic standard - applied for encryption and decryption - symmetric 427 1. DES (Data Encryption Standard) ↔ Grover’s algorithm The most popular cryptosystems and their cryptoanalysis Quantum speedup 429 Grover’s search is quadratically faster than classical search Note: It speeds up any kind of database search. However the maximum advantage is gained in unsorted databases. Can we find faster quantum-search algorithms? Shor’s algorithm is exponentially faster than classical ones, √ N iterations. so it possible to find also a search algorithm that fast? Optimality theorem: Grover’s algorithm is optimal! The search problem cannot be solved in less than O ⇒ 430 But can we find an algorithm that would run, say, twice faster? Possibly yes, but it is not the issue of the optimality theorem. quantum entanglement and quantum speed-up Q: Quantum entanglement is a key resource for QIP. But do we need it for quantum speed-up in e.g. Grover’s algorithm? A: "Entanglement is neither necessary for Grover’s algorithm itself, nor for its efficiency." [Bhattacharya et al., 2002] Q: Really? A: Inversion about the average amplitude is a classical process. Q: Can Grover’s algorithm be implemented classically? A: Grover’s algorithm has already been experimentally implemented using classical Fourier optics. quantum entanglement and database size Q: Is entanglement useful for Grover’s algorithm at all? A: Yes. Lack of entanglement limits the database size, which scales linearly with the beam diameter D (or D2 for a 2D version) ⇒ number of qubits scales only as ∝ log2 D assume D equal to the size of the universe, ∼ 1026m ⇒ it is equivalent to ∝ 86 qubits. 431 This limitation exists for any database containing classical information. Q: Anyway, it seems that entanglement is not necessary for quadratic speed-up. But do we need it for exponential speed-up? A: Most probably, yes. 432 How to implement Grover’s algorithm classically? via classical optical interference A classical implementation of Grover’s search [Amsterdam’s experiment of Bhattacharya et al. (2002)] • quantum probability amplitudes ֒→ a transverse laser beam profile = a complex electric field amplitude E(x) • quantum states, which label items of the database ֒→ continuous coordinate x • sought item x0 ֒→ narrow area around the “item position” x0 434 433 transmitted through one of the mirrors with transmission of 2% ֒→ after each roundtrip, a moving photodiode records light • readout Φ0(x′) = 0 elsewhere. Φ0(x′) = φ if x′ is in a narrow area around 0 E(x) → E(x) exp[iΦ0(x′)] (IAA = inversion about the average amplitude) which imprints phase profile in the Fourier plane ֒→ phase plate (called the IAA plate) like the oracle plate • gate Ûf (0) which marks |0i by phase flipping all |xi except |0i specifically: T=13.5ns ֒→ a roundtrip of a pulse between the cavity mirrors • each Grover’s iteration Φx0 (x) = 0 elsewhere. Φx0 (x) = φ if x is in a narrow area around x0 E(x) → E(x) exp[iΦx0 (x)] which imprints a phase profile Φx0 (x) on the beam ֒→ phase plate (called the oracle plate) • oracle Ûf (x0), which marks the item x0 by phase flipping all |xi except |x0i with focal lengths f1 = 40 cm, f2 = 60 cm specifically: spherical, achromatic doublet lenses ֒→ classical Fourier transform implemented by lenses • quantum Hadamard transform x f1 Ff F f1 IAA plate L1 f2 F L2 f2 M2 Iteration order 2 2 2 1 _5 _3 _ pulses out 0 1 ⊗n H H ⊗n y x Uf y ⊕ f ( x) x H H ⊗n used in: • Simon algorithm • Grover algorithm • Bernstein-Vazirani algorithm • and others quantum DJ algorithm is a common quantum subroutine F – Fourier transform (instead of Hadamard transform) via lenses L1,2 436 IAA – inversion about the average amplitude, M1,2 – mirrors, f1,2 – focal lengths E(x) Oracle plate pulse F in M o 1 Amsterdam’s experiment of Bhattacharya et al. (2002) classical implementation of Grover’s search 435 Simon’s problem (1994) it is an oracle problem closely related to Shor’s algorithm GIVEN: f : {0, 1}n → {0, 1}n – a 2-to-1 function such that ∀x 6= y : f (x) = f (y) ⇔ y = x ⊕ r, r – a fixed n-bit string called the function’s period TASK: find period r. How many oracle queries are required to find period r? exponential number classically polynomial number quantumly Simon’s algorithm - step by step • initial state |ψ0i = |0i⊗n|0i⊗n ≡ |0i|0i |xi|0i. |xi|f (x)i. x∈{0,1}n • apply n Hadamard gates to the first n qubits: X 1 |ψ1i = √ 2n X where |xi ≡ |x1, x2, · · · , xni • apply quantum oracle: 1 |ψ2i = √ 2n x∈{0,1}n 437 438 • measure the last n qubits and obtain a certain f (x′) ∈ {0, 1}n, which yields the (n-qubit) state: 1 |ψ3i = √ (|x′i + |x′ ⊕ ri). 2n • apply n Hadamard gates to the n qubits: X ′ ′ 1 |ψ4i = √ (−1)x ·y + (−1)(x ⊕r)·y |yi 2n+1 y∈{0,1}n X ′ 1 (−1)x ·y |yi. = √ 2n−1 r·y=0 • measure |ψ4i to get y such that r · y = 0. • repeat the above steps a polynomial number of times to get, with high probability, 439 440 n linearly-independent values y = {y1, y2, . . . , yn} such that y · r = r, which determines r. multiplicative order of an integer k = ord(x, n) ≡ ord(x) Mathematica: MultiplicativeOrder[x,n] 1 = xk mod n multiplicative order of x is the smallest integer k for which example ord(8, 21) =? 81 mod 21 = 8 ord(8, 21) = 2 82 mod 21 = 64 mod 21 = 64 − 3 · 21 = 1 ⇒ 441 a1 + 1 1 a2 + K+a N 1 ≡ {a0, a1, a2, K, aN } 442 4+ 2 = 1 + 1 1 ≡ {1, 4, 2} Mathematica: ContinuedFraction[11/9] ֒→ {1,4,2} 2 11 = 1 + 2 = 1 + 1 9 9 9 Example: CF for r=11/9 Note: The procedure will halt iff r is rational. 1. split r into integer part [[r]] and fractional part f = r − [[r]] 2. stop if f = 0 3. calculate 1/f and return to step 1. How to calculate CF for number r ? a0 + Continued fractions (CF) • polynomial time using quantum phase estimation (a part of Shor’s algorithm) • exponential time using known classical algorithms complexity of algorithms for order calculation orders of all elements modulo 21 x = 1 2 4 5 8 10 11 13 16 17 19 20 ord(x,21) = 1 6 3 6 2 6 6 2 3 6 6 2 25 mod 21 = 32 mod 21 = 11 26 mod 21 = 22 mod 21 = 1 24 mod 21 = 16 23 mod 21 = 8 22 mod 21 = 4 21 mod 21 = 2 another example 3.245 4.082 12.250 4.000 12+ 14 4+ 1 1 = {3, 4, 12, 4} = 0.245 1 / 0.245 = 4.082 = 0.082 1 / 0.082 = 12.250 = 0.250 1 / 0.250 = 4.000 = 0.000 3.245 = 3 + -3 -4 - 12 -4 443 ⊗m ⊗n H ⊗n y x U a x mod N 4. Measurement 3. Quantum Fourier transform 2. Modular exponentiation 1. Hadamard transform x QFT −1 444 p q is a convergent of CF and GCD(p, q) = 1 Basic steps of Shor’s algorithm: 1 0 ⇒ Shor's factorization algorithm p q − x < 2q12 Useful criterion for Shor’s algorithm is a truncated CF x = {a0, a1, ..., an , ..., aN }. The nth convergent, denoted by {a0, a1, ..., an }, A convergent of CF Mathematica: ContinuedFraction[3245/1000] ֒→ {3, 4, 12, 4} 3 4 12 4 Example: CF for 3.245 Shor’s algorithm to factorize an integer N 1. If N is even then return factor f = 2. 2. Test classically whether N 6= ab. 3. Choose randomly an integer a (1 < a < N ) and apply the Euclidean algorithm to check whether a and N are coprime, i.e. GCD(a, N ) = 1 If not then choose another a. 4. Prepare two registers: register #2 has n2 = ⌈log2 N ⌉ qubits (this is the number of qubits to store N ); register #1 has n1 = 2n2 qubits 445 446 (in optimized versions of the algorithm, n1 can be smaller). 5. Initialize both registers: |ψ1i = |0i⊗n1 |1i⊗n2 6. Apply Hadamard gates to register # 1: i.e. create an equally-weighted superposition x=0 N1−1 1 X |ψ2i = Ĥ ⊗n1 |ψ1i = √ |x, 15i N1 where N1 = 2n1 . 7. Apply modular exponential gate to register # 2: |ψ3i = Ûmod.exp|ψ2i aR1 mod 15 R1 |0i |1i |2i ... |N1 − 1i R2 |a0 mod N i |a1 mod N i |a2 mod N i ... |aN1−1 mod N i 8. Measure register #2 to get some state |x′i: |ψ4i = N 2hx′|ψ3i where N is an (unimportant) renormalization constant. 9. Apply QFT† on the register # 1: |ψ5i = QF T †|ψ4i and measure it. 10. Apply the classical method of continued fractions to find period r. 11. If r is even and r2 6= −1( mod N ) then calculate f =GCD(ar/2 ± 1, N ) If f 6= 1 or f 6= N then return f . Otherwise repeat the algorithm. How to factorize N=15? 15=3*5 447 448 1. choose x such that 1<x<N-1, GCD(x,N)=1 e.g. x=11 2. find multiplicative order r=ord(x): 111 112 113 114 115 116 11 1 11 1 11 1 (mod N) so r=2 3. find y: y2=1 mod N since xr=1 mod N then y=xr/2=11 4. calculate GCD(y+1,N)=GCD(12,15)=3 GCD(y-1,N) =GCD(10,15)=5 so ⇒ OK ⇒ OK ⇒ OK x=0 |y, 15i aR1 mod 15 R1 |0i |1i |2i ... |255i 0 1 2 255 R2 |7 mod N i |7 mod N i |7 mod N i ... |7 mod N i = |1i = |7i = |49i ≡ |4i ... = |13i |ψ3i = Ûmod.exp|ψ2i 7. apply modular exponential gate to register # 2: where N1 = 2n1 = 256 1 |ψ2i = Ĥ ⊗n1 |ψ1i = √ N1 NX 1−1 i.e. create equally-weighted superposition 6. apply Hadamard gates to register # 1: 450 Mathematica: 2∧∧1111 ֒→ 15 |ψ1i = |0i⊗n1 |1i⊗n2 = |00000000i|1111i ≡ |0i|15i 5. initialize both registers: Mathematica: Ceiling[Log[2, 15]] ֒→ 4 register # 2: n2 = ⌈log2 N ⌉ = ⌈3.906⌉ = 4 register # 1: n1 = 2n2 = 8 4. find the required dimension of registers (number qubits): GCD(7,15) = 1 Euclidean algorithm to check whether a and N are coprime: 3. let’s, e.g., choose a = 7 (unlucky choice) and apply 2. N 6= ab 1. N is odd Example 1: Factorize N = 15 449 451 |3i |13i |7i |13i |11i ... |13i ... r=4 ⇒ 12. final test 3 × 5 = 15 ⇒ OK these are the sought factors :-) GCD(13r/2 − 1, 15) = GCD(168, 15) = 3 = GCD(169 + 1, 15) = GCD(17 × 2 × 5, 3 × 5) = 5 GCD(13r/2 + 1, 15) = GCD(132 + 1, 15) ⇒ 11. r is even and r2 6= −1 mod N 64 1 256 = 4 (which reduces to a trivial case now) to find the period r 10. apply the classical continued fraction method |ψ5i = QF T †|ψ4i = 21 (|0i1 − |64i1 + |128i1 − |192i1) 9. apply QFT† on the register # 1: |ψ4i = N 2h13|ψ3i = N (|3i1 + |7i1 + |11i1 + ...) where N is a renormalization constant R1 R2 for example, we get the state |13i: 8. measure register #2: 452 so Mathematica: PowerMod[aˆx, 15] R1 |0i |1i |2i |3i |4i |5i |6i |7i |8i |9i |10i |11i ... R2 |1i |7i |4i |13i |1i |7i |4i |13i |1i |7i |4i |13i ... a^3 8 12 4 5 6 13 2 9 10 11 3 7 14 a^4 1 6 1 10 6 1 1 6 10 1 6 1 1 ... 2 4 3 9 4 1 5 10 6 6 7 4 8 4 9 6 10 10 11 1 12 9 13 4 14 1 8 1 12 6 4 1 5 10 6 6 13 1 2 1 9 6 10 10 11 1 3 6 7 1 14 1 unlucky choice lucky choice bad choice 2 4 8 1 3 9 12 6 4 1 4 1 5 10 5 10 6 6 6 6 7 4 13 1 8 4 2 1 9 6 9 6 10 10 10 10 11 1 11 1 12 9 3 6 13 4 7 1 14 1 14 1 ⇒ long period ⇒ short period ⇒ GCD(x, N ) 6= 1 ⇒ ⇒ ⇒ lucky, unlucky and bad choices of a modulo N=15 a^2 4 9 1 10 6 4 4 6 10 1 9 4 1 condition: GCD(a,N)=1 a 2 3 4 5 6 7 8 9 10 11 12 13 14 a=2, 7, 8, 13 a=4, 11, 14 a=3,5,6,9,10,12 2 3 4 5 6 7 8 9 10 11 12 13 14 ⇒ ⇒ ⇒ 453 a^14(mod N) 4 unlucky 9 wrong 1 lucky 10 wrong 6 wrong 4 unlucky 4 ... 6 10 1 9 4 1 harder factorization easier factorization excluded 454 Example 2: Factorize again N = 15 but for a = 11 1–2. ditto ⇒ OK 3. we choose a = 11 (lucky choice) GCD(11,15) = 1 R1 |0i |1i |2i |3i |4i |5i |6i |7i |8i ... R2 |1i |11i |1i |11i |1i |11i |1i |11i |1i ... 8. measure register #2: |1i |11i |3i |11i |5i ... |11i ... for example, we get state |11i: R1 R2 |ψ4i = N 2h11|ψ3i = N (|1i1 + |3i1 + |5i1 + ...) where N is a renormalization constant 2 9. apply QFT† on the register # 1: |ψ5i = QF T †|ψ4i = √1 (|0i1 + |128i1) r=2 ⇒ 10. apply the classical continued fraction method ⇒ to find period r 128 = 1 256 2 11. r is even and r2 6= −1 mod N 455 456 GCD(11r/2 + 1, 15) = GCD(11 + 1, 15) = GCD(3× 4, 3× 5) = 3 GCD(11r/2-1, 15) = GCD(10, 15) = 5 which are the sought factors. Actually, Shor has found his algorithm by generalizing Simon’s algorithm, i.e. by replacing Simon’s Hadamard transforms (Fourier transform over Z2n) by Fourier transform over ZN . Shor’s algorithm resembles Simon’s algorithm. Note 2: Note 1: = |11i 4–6. ditto = |121i ≡ |1i ... 7. apply modular exponential gate to register # 2: = |11i To find period r, one usually has to apply the complete classical continued fraction method. |ψ3i = Ûmod.exp|ψ2i aR1 mod 15 = |1i R1 |0i |1i |2i ... |255i R2 |110 mod N i |111 mod N i |112 mod N i ... |11255 mod N i so ÛÛ Û Û Û Ú Û Û Û 256 ; Ý Ü Û Û Ü ã á à Û ÛÛ Û Û Ú â â Ý ä åæ é é é é é é é éé é 1 3 FromContinuedFraction ⇒ Õ ÙÙ Ù Ù Ù criterion 1, 3, x Ø ⇒ PERIOD r 6= 3 False çè 0, 3 1 a 1 65 1 = 0.079 · · · < / 2= = 0.055 · · · − x = − b 3 256 2b 18 1 1 a = ≡ 3 3 b 458 ↑ |ψ1i / H |0i H .. H H |1i⊗n PERIOD r = 4 0, 3, 1 TRUE 460 459 ↑ |ψ2i Uf1 • Uf2 • Uf3 • ... ... ... Ufn • ↑ |ψ3i nm ↑ |ψ4i QF T † Quantum circuit for Shor’s factorization algorithm ⇒ True |0i |0i |0i î 1st convergent of the continued fraction î criterion 1, 4, x êë {0, 3} ≡ 0 + Û 65 256 1 4 FromContinuedFraction ⇒ 1 1 a = ≡ 3 + 11 4 b 1 a 1 65 1 1 = < = − x = − b 4 256 256 2b2 32 {0, 3, 1} ≡ 0 + 2nd convergent of the continued fraction î 0, 3, 1, 15, 4 = ; 457 ê 65 256 Û í FromContinuedFraction Û 15+ 14 Û îî 3 + 1+ 1 1 Û 1 Û {0, 3, 1, 15, 4} ≡ 0 + Û 2 b2 Û 0, 3, 1, 15, 4 1 ÛÛ ContinuedFraction x a 65 256 ß x b InputForm Þ criterion a_, b_, x_ : Abs x 65 Mathematica: Example: Find period r from Û ìí ×Ø ÕÖ ↑ |ψ5i nm nm nm .. nm 2: 1: H H H • 89:; ?>=< • 89:; ?>=< B • 89:; ?>=< C • 89:; ?>=< • D • 89:; ?>=< E • 89:; ?>=< B 89:; ?>=< • F • 89:; ?>=< • D • • 89:; ?>=< 89:; ?>=< • G H • • 89:; ?>=< G QF T † QF T † 461 nm nm nm 462 nm nm nm 7-qubit circuit for Shor’s factorization of N = 15 3: 4: 5: 6: 7: A 89:; ?>=< • optimized version of the 7-qubit circuit for Shor’s factorization of N = 15 H (gates C, E, F, H are removed) 1: H H 2: 3: 4: 5: 6: 7: A H R2 • • R2 R3 • H H • R3 R2 • • R2 H H |x2i |x1i |x0i |y2i |y1i |y0i Quantum Fourier transform and its inverse for 3 qubits QFT |x2i |x1i |x0i H QFT−1=QFT† |y2i |y1i |y0i where the rotations are R2 = R(900), R3 = R(450 ) Outline 1. Generalized projective measurements 2. Positive operator valued measure (POVM) 3. Kraus representation 4. Damping channels for a single qubit 5. Imperfect photocount detectors 6. Bell-state and GHZ-state analyzers 463 464 7. Fidelity and other measures of quality of the state generation 8. Entanglement measures 466 where m,n,M,N X (|µihµ|)mM,nN (ρsys)mn(ρaux)MN M,N X (|µihµ|)mM,nN (ρaux)MN Pµ = Tr(Aµρsys) (Aµ)mn = which be rewritten as Pµ = Tr[|µihµ|(ρsys ⊗ ρaux)] = • probability of outcome µ: Different outcomes correspond to orthogonal and complete projectors |µihµ| A maximal test is then performed in the combined Hilbert space K. The combined, uncorrelated state of the original quantum system and the ancilla is (ρsys ⊗ ρaux)mM,nN = (ρsys)mn (ρaux)MN • Let’s prepare an auxiliary quantum system (ancilla) in a known state ρ̂aux Generalized projective measurement do not commute and are therefore not simultaneously observable • measurement operators, corresponding to nonorthogonal states, also called the projection valued (PV) measure |µihµ| – measurement operator or projection operator = probability of the measurement outcome µ Pµ – probability that the system is in state |µi where Pµ = Tr{|µihµ|ρsys} is represented by complete, orthonormal set of states |µi von Neumann-type projective measurement 465 |ψi = α|ui + β|vi general state Can you distinguish whether qubit is in state |ui or |vi? where θ the angle between polarization vectors hu|vi = cos θ non-orthogonal linear polarization states |ui and |vi Cryptographic example of POVM (at leat sometimes) How to distinguish conclusively between two non-orthogonal states? 468 • POVM allow extraction of more mutual information that can the usual von Neumann-type projective measurement nonorthogonal states. • POVM allow the possibility of measurement outcomes associated with • the number of available outcomes may differ from the number of available preparations (of a given state) and the dimension of the Hilbert space. Advantages of POVM over PV measures Pµ = Tr(Aµρsys) • probability that a quantum system is in a particular state is given by the expectation value of the POVM operator corresponding to that state µ which are positive Hermitian operators acting on original Hilbert space X Aµ = 1 positive operator valued measures (POVM) = set of Aµ’s 467 ⇒ 471 Neumark’s theorem 469 POVM operators 1 − |vihv| 1 + hu|vi inconclusive measurement Av = are not projective measurements 1 − |uihu| , 1 + hu|vi ⇒ detector Dv will not click 470 = |α + β|2 cos θ ⇒ detector Du will not click Du M Dv analogy a pure state of the bipartite system AB may behave like a mixed states when we observe subsystem A alone, similarly an orthonormal measurement of the system AB may be a nonorthonormal POVM on A alone Kraus representation = operator sum representation basic idea an elegant representation to describe open system dynamics = TrE {ρSE (t)} • reduced density matrix ρS (t) X h{ki}|ρSE (t)|{ki}i {ki } where |{ki}i ≡ |k1 · · · ki · · · i = Πi ⊗ |kii = ρ(0) = ρS (0) ⊗ ρE (0) −→ ρSE (t) = U (t)ρS (0) ⊗ ρE (0)U †(t) • So let’s analyze evolution of both the system and environment: • The final state of an open system cannot be described by a unitary transformation of the initial state. 472 It is also generally thought that a POVM belongs to the most general test to which a quantum system may be subjected. most general measurement operator There always exists an experimentally realizable procedure generating any desired POVM represented by given matrices Aµ physical implication One can extend the Hilbert space H of states, in which the Aµ are defined, in such a way that there exists, in the extended space K, a set of orthogonal and complete projectors |µihµ| such that Aµ is the result of projecting |µihµ| from K into H. Au = A? = 1 − Au − Av Pi = hψ|Ai|ψi = |β|2(1 − cos θ), Pv = 0 ⇒ P? probabilities of measurements = |α|2(1 − cos θ), Pv ⇒ Pu = 0 Pu |ψi = |ui ⇒ special input states |ψi = |vi an optical implementation of POVM D? BS HWP 4 (NOT) π optical implementation of POVM BS PBS π HWP 8 (H) ψ is an orthonormal basis of the environment Hilbert space {ki } † ′ i ki =k USE p A0 = 1 − p, Aµ = E hµ|USE |0iE A1 = A2 = A3 = so 474 473 p 3 r h i p σx|ψiS ⊗|1iE +σy |ψiS ⊗|2iE +σz |ψiS ⊗|3iE 3 Kraus operators p : |ψiS ⊗|0iE → 1 − p|ψiS ⊗|0iE + evolution of single qubit in depolarizing channel and error (bit flip and/or phase flip) occurs with probability p i.e. the qubit remains intact with probability 1 − p depolarizing channel Example of Kraus representation it seems extremely difficult to use K.r. if the environment is at T > 0 • drawback k A†k (t)Ak (t) = I P h{ki}|U (t)|{0}i Ak (t)ρS (0)Ak (t) stands for summation under the condition • completeness relation X where P′ Ak (t) = k X k=0 in terms of Kraus operators ρS (t) = ∞ X • alternatively, the evolution can be given in the so-called Kraus representation |1i → −|1i σ̂z ≡ |1i → −i|0i |ψi → σ̂y |ψi |0i → i|1i, |ψi → σ̂z |ψi |0i → |0i, |1i → |0i |ψi → σ̂x|ψi |0i → |1i, where the σk are the Pauli operators 0 1 0 −i σ̂x ≡ , σ̂y ≡ , 1 0 i 0 3. both errors † p Fµ 1 0 0 −1 and Aµ are Kraus operators 2. phase flip error 1. bit flip error Fµρs where µ Xp Fµ = A µ A µ ρs → modifies a density matrix as follows three types of errors POVM POVM from Kraus representation 476 475 + √ p|0iS |1iE damping channels for a single qubit 1. amplitude-damping channel → p|1iS |0iE |0iS |0iE → |0iS |0iE p 1− |1iS |0iE 2. phase-damping channel p √ |0iS |0iE → 1 − p|0iS |0iE + p|0iS |1iE p √ 1 − p|1iS |0iE + p|1iS |2iE |1iS |0iE → ^ ^ b2 r b^ 3 b^ 4 N4 photons N3 photons N2 photons Φ b̂1 b̂ 2 b̂3 b̂4 ρˆ 1 N4 477 photons N3 photons N2 photons 478 i ph σx|ψiS |1iE + σy |ψiS |2iE + σz |ψiS |3iE 3 • it is a „caricature” model of decoherence in real systems • no bit flip in system! p 1 − p|ψiS |0iE + 3. depolarizing channel |ψiS |0iE → a^ 2 a^3 b1 ρˆ 1 2) (b3) (b4 ) Tr(b2,b3,b4) Π̂(b N2 Π̂N3 Π̂N4 |ΦihΦ| POVM for the kth imperfect photocounter detecting Nk photons four-mode state before the measurements single-mode state after the measurements renormalization constant ρ̂1 = N a^ 4 conditional measurements using imperfect photocounters a^ 1 (b ) Π̂Nkk |Φi ρ̂1 N • 479 POVM for photocount detectors (I) Perfectly-Resolving Photon Counter m=0 n=0 ( No clicks ) 480 η − inefficiency ν − dark count rate ( N−n) ∞ N −ν ν ˆ η n (1−η )m−n C mn m m Π = e ∑∑ (N − n)! N ∞ m =0 ˆ c = 1− Π ˆc Π 1 0 single-photon counter ( 1 click ) POVM for photocount detectors (II) • Dark count rate ~ 100 − 1000 s − 1 ( Click ) ˆ c = ∑ e −ν (1 − η ) m m m Π 0 • Conventional Photon Counter (CPC) • m=0 ∞ ˆ vl = ∑ e −ν (1 − η ) m m m ( no clicks ) Π 0 ( 2 clicks ) 10 4 s −1 (1−n) ∞ 1 −ν ν ˆ vl Π ηn (1−η )m−n C mn m m 1 = ∑∑ e (1− n)! m=0 n=0 ˆ vl = 1 − Π ˆ vl − Π ˆ vl Π 2 0 1 • high dark count rate ~ for detection of single photon η < 25% ⇒ ⇒ basically noise free useless for QC for detection of single photon ⇒ small resolution time (for detection of two photons) 482 • • • 2 |00i ± |11i √ |Φ±i = How to distinguish experimentally the GHZ states? Is it possible to unambiguously discriminate between all the Bell states using only linear optics? Do they exist perfect linear Bell-state analyzers? 2 |01i ± |10i √ |Ψ±i = How to discriminate between the optical Bell states? Bell-state and GHZ-state analyzers τ ∼ 2ns η ∼ 47% for detection of two photons η ∼ 88% with noise free avalanche photomultiplication • VLPC = visible light photon counters Fnoise = 1 η ∼ 70 − 80% • SSPM = solid state photomultiplier η ∼ 6% for detection of two photons • PMT = photomultiplier tube relatively high noise ⇒ Fnoise ≥ 2 η ∼ 70 − 80% • Si APD = Si avalanche photodiode (working in Geiger mode) Photocounters (photon count detectors) 481 A A PBSAB HWP PBS 1 BS PBS 1 B B D3 D4 Ψ + ← (1, 3) o r ( 2 , 4 ) D1 PBS 2 D3 D4 Φ − ← (1, 3) o r ( 2 , 4 ) Ψ ± ← (1,1), ( 2, 2 ), (3, 3), o r ( 4 , 4 ) HWP D2 Φ + ← (1, 4 ) o r ( 2 , 3) 484 Φ ± ← (1,1), ( 2 , 2 ), (3, 3), o r ( 4 , 4 ) PBS 2 D2 Ψ − ← (1, 4 ) o r ( 2 , 3) Bell-state analyzer (2) D1 Bell-state analyzer (1) 483 485 What Bell states can uniquely be distinguished in the (Pan-Zeilinger) analyzer? notation: particles A,B; modes 1,2 1. general input state 2. state after PBSAB if horizontal (vertical) polarization component is transmitted (reflected) then PBS1 |ψini = α|HAi|HB i + β|HAi|VB i + γ|VA i|HB i + δ|VAi|VB i |ψi = α|HA2i|HB1i + β|HA2i|VB2i + γ|VA1 i|HB1i + δ|VA1i|VB2i A B D3 HWP D4 D5 PBS3 GHZ-state analyzer D1 D2 PBS2 HWP PBSBC C polarization GHZ states 1 |Φ±i = √ |Hi|Hi|Hi + |V i|V i|V i 2 1 |Ψ1±i = √ |V i|Hi|Hi + |Hi|V i|V i 2 1 |Ψ2±i = √ |Hi|V i|Hi + |V i|Hi|V i 2 1 |Ψ3±i = √ |Hi|Hi|V i + |V i|V i|Hi 2 • the Pan-Zeilinger analyzer discriminates between only two (|Φ±i) among 2N GHZ states PBSAB HWP 3. indistinguishability of photons implies that we can omit subscripts A, B: D4 ≡ DH2 486 |ψi = α|H1i|H2i + β|H2i|V2i + γ|V1i|H1i + δ|V1i|V2i 1 1 1 1 + − + − = √ (α + δ)|Φtwo i + √ (α − δ)|Φtwo i + √ (β + γ)|Ψone i + √ (β − γ)|Ψone i 2 2 2 2 where 1 1 ± ± i = √ |H1i|V1i ± |V2i|H2i |Φtwo i = √ |H1i|H2i ± |V1i|V2i , |Ψone 2 2 4. state after quarter-wave plates and detection D3 ≡ DV 2 , 1 |Vii → √ (|Hii − |Vii) 2 • quarter-wave plates change polarizations as follows D2 ≡ DV 1 , 1 |Hii → √ (|Hii + |Vii), 2 • notation: D1 ≡ DH1, • case 1: 1 + + |Φtwo i → |Φtwo i = √ |H1i|H2i + |V1i|V2i 2 then photons are detected in D1 and D4 or D2 and D3 1 |H1i|H1i ± |H2i|H2i − |V1i|V1i ∓ |V2i|V2i 2 • case 2: 1 − + |Φtwo i → |Ψ |H1i|V2i + |V1i|H2i twoi = √ 2 then photons are detected in D1 and D3 or D2 and D4 • cases 3,4: + |Ψone i→ then both photons are detected in either D1, or D2, or D3, or D4 487 488 D6 Dn … 490 Linear optical elements can provide any arbitrary unitary mapping only over creation operators, not over a general input state. Why no-go? Within this subclass it is not possible to discriminate unambiguously four equiprobable Bell states with a probability higher than 50%. Calsamiglia-Lütkenhaus theorem: There is no perfect Bell state linear analyzer on two qubits in polarization entanglement without conditional measurements. No-go theorem: Do they exist perfect Bell-state linear analyzers? [Calsamiglia i Lütkenhaus, 2000] Uˆ2(N1) N1 Uˆ3(N2) N2 … … … N3 QND HWP on (P)BS – (polarization) beam splitter. HWP – half-wave plate, CCL – classical coincidence & logic, Key: QND – quantum non-demolition detector, PBS off BS 492 off on Bell state QND CCL |Ψ+i off off |Ψ−i off on |Φ+i on off |Φ−i on on CCL nonlinear Mach-Zehnder interferometer based on [Paris et al. (2000)] Complete nonlinear Bell-state analyzer Bell state source 491 1 →1 1 + f (Nph, Ncm) where Ncm – number of conditional measurements Nph – number of entangled photons in auxiliary modes auxiliary modes Psuccess = 1 − … 1 2 … auxiliary modes Û1 … … Psuccess ≤ auxiliary modes input modes - based on conditional measurements Complete linear Bell-state analyzer … probability to discriminate unambiguously between all the Bell states is bˆn … vacuum auxiliary aˆn modes Not complete linear Bell-state analyzer â1 b̂1 D1 â2 input b̂ 2 modes â3 D2 â4 Û â5 489 … signal α 0 BS QND – Fock filter Kerr PS BS Key: Kerr nonlinear medium described by 3rd order susceptibility χ(3) |αi – strong coherent field PS – phase shifter ' %6 0 ' 493 494 QND measurement does not destroy coherence of the signal state, but only adds extra phase, which can easily be corrected. .HUU PHGLXP Kerr effect and QND %6 0 θ YDFXXP SUREH VLJQDO on off Measures of quality of the state generation 1. Fidelity = Uhlmann’s transition probability for mixed states q 2 p p Tr ρ̂thρ̂exp ρ̂th F (ρ̂exp, ρ̂th) = ρ̂exp – density matrix for the experimentally generated state ρ̂th – theoretically predicted density matrix 2. Bures distance a measure of discrepancy between ρ̂exp and ρ̂th: q DB (ρ̂exp||ρ̂th) = 2 − 2 F (ρ̂exp, ρ̂th) • it satisfies the usual metric properties including symmetry DB (ρ̂exp||ρ̂th) = DB (ρ̂th||ρ̂exp) 495 496 3. Quantum relative entropy = Kullback-Leibler ‘distance’ S(σ||ρ) = Tr (σ lg σ − σ lg ρ) it is not a true metric since S(σ||ρ) 6= S(ρ||σ) 4. MaxLik parameters - used by us in the tomographic reconstruction of physical density matrices 5. Relative entropy of entanglement minimum of the quantum relative entropy over set D of all separable states ρ: E (σ ) = minρ∈D S (σ||ρ) = S (σ||ρ̄) ρ̄ is the separable state closest to σ. i pi E(ρi ) ≥ E( i p i ρi ) P entanglement of distillation ED • = EC = ER = ED EF ≥ EC ≥ ER ≥ ED for mixed states EF for pure states • ··· relative entropy of entanglement ER entanglement cost EC = limn→∞ EF (ρn entanglement of formation EF • • • ⊗n Measures of entanglement ) 498 C5.∗ E(ρ) should reduce to the entropy of entanglement for a pure state i.e., mixing cannot increase E(ρ) P C4. Entanglement is convex under discarding information: C3. LOCC operations cannot increase E(ρ) C2. local unitary transformations UA ⊗ UB do not change E(ρ) E(ρ) = 1 for a Bell state E(ρ) = 0 for an unentangled state C1. E(ρ) ≥ 0 ρ – the density matrix of a given state Criteria for a good entanglement measure 497 1 2 [1 + √ λ1 − λ2 − maximally entangled pairs distillation LOCC LOCC formation partially entangled pairs asymptotic conversion ratios for entanglement of formation and distillation 500 “C(ρ) is a measure of the entanglement of formation in its own right.” [Wootters] where λi’s are in nonincreasing order the eigenvalues of ρ(σy ⊗ σy )ρ∗(σy ⊗ σy ) C(ρ) = max{0, √ p 2 1 − C (ρ)] in terms of the concurrence EF (ρ) = H √ √ λ3 − λ4 } i piE(|ψi ihψi |) P For two qubits [Wootters, PRL’98] Special case EF (ρ) = minpi,ψi It is the minimized P average entanglement of any ensemble of pure states |ψii realizing ρ = i pi|ψiihψi|: [Bennett, DiVincenzo, Smolin, and Wootters, PRA’96] Entanglement of formation 499 Negativity 501 [Życzkowski et al. PRA’98, Eisert, Plenio JMO’99, Vidal, Werner, PRA’02] a quantitative version of the Peres-Horodecki entanglement criterion [Peres, PRL’96, Horodecki et al., PLA’96] N (ρ) = max{0, −2 mini µi} P N (ρ) = max{0, −2 µi} where µi eigenvalues of the partial transpose of ρ Logarithmic negativity EN (ρ) = log2[N (ρ) + 1] a measure of the PPT entanglement cost [Audenaert et al. PRL’03, Ishizaka PRA’04] 502 EN gives upper bounds on the teleportation capacity and the entanglement of distillation ED [Vidal, Werner PRA’02] σ * S (σ || ρ sep ) geometric measure of entanglement entangled states ρ * sep separable states Relative entropy of entanglement (REE) [Vedral, Plenio, Jacobs, Knight, PRA’1997] E (σ ) = inf ρ∈D S (σ||ρ) = S (σ||ρ∗) ρ∗ – the closest separable state to σ quantum relative entropy or a quantum Kullback-Leibler distance S (σ||ρ) = Tr (σ log σ − σ log ρ) NOTE: S(σ||ρ) is a “distance” between σ and ρ • but it is not a true metric: it is neither symmetric nor satisfies the triangle inequality • it is not a unique measure of the distance: 503 504 see also Bures measure with the Uhlmann transition probability (or fidelity) p √ √ S ′(σ||ρ) = 2 − 2 F (σ, ρ) with F (σ, ρ) = [T r( ρσ ρ)1/2]2 Bell-inequality violation 3 P n,m=1 I⊗ I + r · σ ⊗ I + I ⊗ s · σ + tnm σn ⊗ σm ! For two qubits the Bell inequality due to Clauser, Horne, Shimony and Holt (CHSH) [PRL’69]: |Tr (ρ BCHSH)| ≤ 2 where Bell operator is BCHSH = a · σ ⊗ (b + b′) · σ + a′ · σ ⊗ (b − b′) · σ 1 4 and arbitrary ρ in Hilbert-Schmidt basis is ρ= with tnm = Tr (ρ σn ⊗ σm) Horodecki et al. [PLA’95] showed that p M (ρ) maxBCHSH Tr (ρ BCHSH) = 2 where M (ρ) = maxj<k {uj + uk }; uj are eigenvalues of Uρ = TρT Tρ; Tρ = [tnm]3×3 the maximal violation of Bell inequality state ρ admits local hidden variable model max {0, M (ρ) − 1 } ⇒ ⇒ p − 506 for ⇒ Entanglement without Bell inequality violation for p ∈ states are entangled iff 1/3 < p ≤ 1. 1 √1 3, 2 E 0≤p≤1 n o C(ρW ) = N (ρW ) = max 0, 12 (3p − 1) 3. Concurrence & negativity thus the √ Werner state violates the Bell inequality iff 1/ 2 < p ≤ 1 B(ρW ) = max{0, 2p2 − 1}1/2 2. Degree of Bell-inequality violation ρW = p|Ψ−ihΨ−| + 1−p 4 I ⊗I Mixture of the maximally entangled state (singlet) |Ψ i and the (separable) maximally mixed state [Werner, PRA’89]: 1. Definition Two-qubit Werner states C (Ψ) = N (Ψ) = B (Ψ) = 2|c00c11 − c01c10| Entanglement measures for two-qubit pure states |Ψi = c00|00i + c01|01i + c10|10i + c11|11i B(ρ) = 1 B(ρ) = 0 then B(ρ) ≡ • useful parameter Horodecki theorem: the Bell inequality is violated iff M (ρ) > 1 Degree of the Bell-inequality violation (BIV) 505 508 No, if we want to define entanglement measures for examining the problems of how to prepare and use the entanglement. Can we avoid the state-ordering ambiguity? Yes, as these incomparable states cannot be transformed to each other with unit efficiency by LOCC. Is it physically reasonable? It is implied by the requirements of equivalence and continuity of the measures on pure states Why is it so? all good asymptotic entanglement measures, which reduce to the entropy of entanglement for pure states, are either equivalent or do not impose the same state ordering Virmani-Plenio theorem: ordering of states can depend on the applied measures of entanglement Eisert-Plenio conclusion from Monte Carlo simulations: Is it a necessary condition for consistency of entanglement measures? Strange, but no. Eisert and Plenio [J. Mod. Opt. 1999] - numerical example Can this condition be violated? Yes. QUESTIONS: E ′(ρ1) < E ′(ρ2) ⇔ E ′′(ρ1) < E ′′(ρ2) Entanglement measures should impose the same ordering of states CONJECTURE 507 Questions 509 1. Can we find analytical examples of two-qubit states for which entanglement measures impose different orderings ? 2. Are there mixed states more entangled than pure states ??? 3 Bob ABCD U EPR source out 510 3. Can we find a physical process manifesting the different orderings ? ideal teleportation Alice ABCD 2 Bennett, Brassard, Crepeau, Jozsa, Peres, Wootters (1993) BSM 1 in teleportation fidelity F = 〈 in | ρ out | in 〉 = 1 imperfect BSM 1 in Bob ABCD out 511 512 ρ imperfect U imperfect EPR state 3 real teleportation Alice ABCD 2 ? E (ρ ') > E (ρ '') ⇒ F (ρ ') > F (ρ '') 1 0.8 0.6 0.4 0.2 0 0 0.8 pure & Werner states Horodecki states 0.4 0.6 concurrence Numerical simulations of 5 × 104 states. 0.2 1 Upper & lower bounds for negativity vs concurrence negativity = C (ρ) ⇐⇒ Examples of the minimum-negativity states states have the minimum negativity for a given concurrence ⇒ N (ρW ) = C (ρW ) = max{0, 3p−1 2 } 2(a) Werner state ρW (p) 2. Bell diagonal states = p|ψ−ihψ−| + 1−p 4 I ⊗I Examples of the maximum-negativity states 1. pure states • concurrence C(ρH ) = p p negativity N (ρH ) = (1 − p)2 + p2 − (1 − p) ⇒ 1. Horodecki state ρH (p) = p|ψ−ihψ−| + (1 − p)|00ih00| 2. Horodecki-like state ρ′H (p) = p|ψ−ihψ−| + (1 − p)|ψ ′ihψ ′ | where |ψ ′i = 12 (|00i + |01i + |10i + |11i) • ⇒ two eigenvalues are vanishing and the other two correspond to eigenvectors, which are a Bell state and separable state orthogonal to it = fC (ρ) ⇐⇒ states have the maximum negativity for a given concurrence 2. N (ρ) ⇒ the eigenvector corresponds to the negative eigenvalue of ρTA is a Bell state 1. N (ρ) 514 [1 − C (ρ)]2 + C 2(ρ) + C (ρ) − 1 ≡ fC (ρ) Structure of the extremal states: C (ρ) ≥ N (ρ) ≥ q [Verstraete et al. JPA’01] Upper & lower bounds for negativity vs concurrence 513 0.2 Y 0.4 X 0.6 concurrence V Z P R 0.8 T O S 1 515 and the Werner states: √ ρY = ρW ( 2/3), ρZ = ρW (2/3), √ ρV = ρW (1/3 + 2/6) √ √ or pure state |Ψ(p)i = p |01i + 1 − p |10i: p √ ρY = |Ψ(p)ihΨ(p)| for p = 1/2 ± 1 + 2 2/4, √ ρZ for p = 1/2 ± 3/4, √ ρV for p = 1/2 ± 14/8 Let us choose the Horodecki state: ρX = ρH (1/2) Explicit examples of states extremely violating the ordering condition 516 Two states, when one corresponds to a point O (or X) and the other to any other point in yellow regions, exhibit different state orderings for N and C. 0 0 0.2 0.4 0.6 0.8 1 Regions of different state orderings for negativity N and concurrence C negativity 0.6 0.8 1 1 0.8 0.6 0.4 0.2 0 0 (b) 0.2 0.6 concurrence 0.4 0.8 1 1 0.8 0.6 0.4 0.2 0 0 (c) 0.2 0.6 concurrence 0.4 0.8 517 518 1 States specifically violating the ordering condition 1 (a) 0.4 (a) states with constant negativity (b) states with constant concurrence (c) states for which C(ρ1) − C(ρ2) = −[N (ρ1) − N (ρ2)] concurrence negativity 0.8 0.6 0.2 negativity ρ′ = ρ̄(p, q ′) for q ′ = ρ′′ = ρ′′′ = ρ̄(p, q ′′′) for q ′′′ = 1 + [N (ρ)+C(ρ)+1−2p]2−1 2p(1−p) 3. states giving exactly opposite predictions: ρ̄(C0, q) 2. states with the same concurrence C0: N0 [N0+2(1−p)]−p2 2p(1−p) • Three classes of states: 1. states with the same negativity N0: then p N (ρ̄(p, q)) = 1 − 2p(1 − p)(1 − q) − (1 − p) C(ρ̄(p, q)) = p States specifically violating the ordering condition ρ̄(p, q ) = p|ψ ihψ | + (1 − p)|ψq ihψq | − − p √ 1 − q|00i + q|01i with |ψq i = 0 0 0.2 0.4 negativity 519 REE vs concurrence and negativity REE for Bell diagonal (including Werner) states pi|ψAiihψAi| ⊗ |ψBiihψBi| 520 EW (C) = EW (N ) = 12 [(1 + C) log(1 + C) + (1 − C) log(1 − C) ] REE for pure states √ EP (C) = EP (N ) = H 12 [1 + 1 − C 2] REE for Horodecki states σH = C|ψ−ihψ−| + (1 − C)|00ih00| EH (C) = (C − 2) log(1 − C/2) + (1 − C) log(1 − C) p 2N (1 + N ) − 1) EH (C = Numerics for two-qubit REE: [Vedral and Plenio, PRA’1998] Caratheodory’s theorem: Any state in D can be decomposed into a sum of at most (dim(HA) × dim(HB ))2 products of pure states. 16 X i=1 there are at most 15 + 16 × 4 = 79 real parameters ρ= Thus, any disentangled 2 qubit state can be given by ⇒ How to minimize S (σ||ρ) over 79 parameters? S(σ||ρ) is a convex function D is a convex set (convex hull) of its pure states a convex function over a convex set can only have a global minimum 0.8 1 0.2 0.4 0.6 concurrence Horodecki states pure states Werner states 0.8 1 Relative entropy vs concurrence: Numerical simulations of 5 × 104 states 0 0 0.2 0.4 0.6 0.8 1 522 0.2 0.4 0.6 0.8 entanglement of formation Horodecki states pure states Werner states 1 524 523 0 0 0.5 1 0.2 0.4 N 0.6 0.8 1 1 0.8 0.6 0.4 C 0.2 0 How to find different state orderings imposed by E, C & N Relative entropy vs entropy of formation: Numerical simulations of 5 × 104 states 0 0 0.4 0.6 concurrence 0 0 0.4 0.6 0.8 1 0.2 0.2 Horodecki states pure states Werner states 521 0.2 0.4 0.6 0.8 1 Relative entropy vs concurrence: Epure(C ) > EW(C ) > EH(C ) for C ∈ (0, 1) relative entropy relative entropy relative entropy E E’’-E’ N’’-N’ 1. 2. All cases of different state orderings C’’-C’ 5. 8. 4. 7. 3. 6. 11. 14. 10. 13. 9. 12. Concurrences C(σ ′) < C(σ ′′) C(σ ′) < C(σ ′′) C(σ ′) > C(σ ′′) C(σ ′) < C(σ ′′) C(σ ′) = C(σ ′′) C(σ ′) < C(σ ′′) C(σ ′) = C(σ ′′) C(σ ′) < C(σ ′′) C(σ ′) = C(σ ′′) C(σ ′) < C(σ ′′) C(σ ′) = C(σ ′′) C(σ ′) > C(σ ′′) C(σ ′) = C(σ ′′) C(σ ′) < C(σ ′′) Negativities N (σ ′) < N (σ ′′) N (σ ′) > N (σ ′′) N (σ ′) < N (σ ′′) N (σ ′) < N (σ ′′) N (σ ′) = N (σ ′′) N (σ ′) = N (σ ′′) N (σ ′) < N (σ ′′) N (σ ′) < N (σ ′′) N (σ ′) = N (σ ′′) N (σ ′) = N (σ ′′) N (σ ′) < N (σ ′′) N (σ ′) = N (σ ′′) N (σ ′) > N (σ ′′) N (σ ′) > N (σ ′′) REEs E(σ ′) < E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) > E(σ ′′) E(σ ′) = E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) = E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) = E(σ ′′) E(σ ′) = E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) < E(σ ′′) E(σ ′) = E(σ ′′) Table 1: All cases of different state orderings by E, C & N. Class 1 2 3 4 5 6 7 8 9 10 11 12 13 14 525 526 527 528 Quantum information processing with quantum dots 1. based on electron spins of quantum dots using optical methods 2. based on nuclear spins of quantum dots using NMR methods What are the quantum dots? Quantum dots (q-dots or artificial atoms) are small metal or semiconductor boxes that hold a well-defined number of electrons useful property number of electrons in a q-dot can be adjusted by changing the dot electrostatic environment parameters number of electrons: from 0 to hundreds size of q-dots: from 30 nm to 1 micron energy energy 1D 0D 530 energy microdisk lasers microcavity Quantum-information processing based on q-dots and cavity QED [Imamoǧlu et al., PRL’99] energy 529 quantum well quantum wire quantum dot 2D DOS bulk system DOS 3D DOS Energy spectra of q-dots vs other nano-structures DOS 531 microdisk microcavity laser Identical quantum dots in microcavity microdisk microcavity lasers 532 Scalable system of quantum dot interactions and cavity QED |1>n ωcav (1) ∆ωn |v>n |0>n (0) ∆ωn (L) ωn conduction band ∆n valence band 533 Hamiltonian for three-level q-dots interacting with N + 1 fields 534 Level structure of quantum dots in V configuration N n X X n (L) † (L) † (L) ) ân , h̄gnv1(âcav σ̂n1v + âcavσ̂nv1), (L) (L) h̄gnv0[ân σ̂n0v + (ân )†σ̂nv0] † âcav + = h̄ωcavâcav = h̄ωn (ân Ĥ = ĤQD + ĤF + Ĥint, X = (En(0)σ̂n00 + En(1)σ̂n11 + En(v)σ̂nvv ), ĤQD ĤF Ĥint nX + n where the nth dot operator is σ̂nxy = |xinnhy| Note: Q-dots are coupled only indirectly via the cavity and laser fields Derivation of q-dot interaction Hamiltonian (L) 3-level Hamiltonian for |gn >, |en >, |vn >, an , acav (L) effective 2-level Hamiltonian for |gn >, |en >, an , acav 535 536 effective spin-spin interaction Hamiltonian for |gn >, |en > equivalent-neighbor (SVW) Hamiltonian Hamiltonian after adiabatic elimination where gn(t)gm(t) ∆ n P − ei(∆n−∆m)t + σ̂ −σ̂ + e−i(∆n−∆m)t] Ĥeff = h̄2 n6=m κnm(t)[σ̂n+σ̂m n m κnm(t) = 1 (0) ∆ωn + 1 (1) ∆ωn gn(t) = gnv0gnv1|En (k = e, g) (L) (t)| ∆ωk(n) = ωk(n) − ωV(n)B − ωcav ∆n = ωe(n) − ωg(n) + ωL(n) − ωcav = ∆ωe(n) − ∆ωg(n) Adiabatic elimination requires (1) (0) 1. coupling strength, cavity decay rate, and thermal fluctuations (x) ≪ h̄∆n, h̄∆ωn , En − En 2. valence-band levels |vin are far off resonance. i h +σ̂ − + σ̂ −σ̂ + + γ(σ̂ +σ̂ + + σ̂ −σ̂ − ) σ̂ T n m n m n m n6=m nm n m P x x n6=m σ̂n σ̂m P 538 [Koashi,Bužek,Imoto, PRA’00] Tight bound for symmetric sharing of entanglement Cij ≤ 2/N Entangled webs in spin van der Waals (SVW) model Ĥint = 4κ SVW model for γ = 1 Non-conservative SVW model (NCE model) SVW model for γ = 0 P − + σ̂ −σ̂ + Ĥint = κ n6=m σ̂n+σ̂m n m Conservative SVW model (CE model) Frenkel model for Tnm=const P − + σ̂ −σ̂ + + γ(σ̂ +σ̂ + + σ̂ −σ̂ − )] Ĥint = κ n6=m[σ̂n+σ̂m n m n m n m Spin van der Waals (SVW) model Ĥint = h̄ Frenkel-type model or Heisenberg model Models 537 6. tomography of nuclear spins 5. quantum algorithms 4. quantum gates 3. pseudo-pure states 2. control of nuclear spins 1. nuclear qubits and qudits 540 N-M Quantum Computing Based on Nuclear Magnetic Resonance (NMR) M Q-dot bipartite entanglement in SVW model 539 541 Quantum computer in a cup of coffee? (3) (1) (2) (1) (3) 542 3,7-Dihydro-1,3,7-trimethyl-1H-purine-2,6-dione C8H10N4O2 caffeine molecule - qubits encoded in nonequivalent C-nuclei CHEMICAL NAME : (2) (2) (3) three qubits encoded in alanine molecule 13C-labeled (1) J-coupling Hamiltonian for carbon-13 nuclei Ĥint = ν1σ̂z + ν2σ̂z + ν3σ̂z + 21 (J12σ̂z σ̂z + J23σ̂z σ̂z + J13σ̂z σ̂z ) antenna gate 543 544 3 1 ,− 2 2 3 1 , 2 2 3 3 , 2 2 3 3 ,− 2 2 NMR quantum computing in nanostructures oscillating magnetic-field high-frequency current insulator ω / 01 ω12/ ω 23/ point contact channel split Schottky gates barrier GaAs barrier GaAs backgate ω12 = ω 0 0 ω = ω −ω 01 quadrupole interaction q ω 23 = ω 0 + ω q energy levels of spin I=3/2 ω0 ω0 ω0 Zeeman interaction 2nd order shifts (usually neglected) ω 01 ω12 ω 23 ω 02 ω13 ω 03 Mz – longitudinal component of M static along B0 0 ≡ 00 3 1 ,− 2 2 3 3 ,− 2 2 546 3 1 , 2 2 ≡ 3,3 2 2 1 ≡ 01 ≡ 2 ≡ 10 ≡ 3 ≡ 11 ≡ Mxy – transverse component of M precessing in the x–y plane (Larmor precession) B1 – magnetic field component of the r.f. field B0 – static magnetic field M – magnetization vector of an ensemble of nuclear spins Larmor precession quartit = ququart = 4-level qudit ω0 ω0 ω0 quartit is equivalent to 2 virtual qubits 545 Ĥ0 = 21h̄ω0σ̂z ih̄ h1 e−iω0tp/2 0 , 0 eiω0tp/2 h̄ωnut (σ̂x cos φp + σ̂y sin φp) 2 where ωnut = γBrf is the nutation frequency ′ Ĥrf,rot = • Hamiltonian in the rotating frame ωLarmor = ωref spin Hamiltonian during the r.f. pulse set on resonance – ωref is the spectrometer reference frequency – Brf is the amplitude of the oscillating r.f. field – φp is the phase of the pulse Brf (t) = Brf cos(ωref + φp)ex coil is along x-axis generating an r.f. pulse field • assumptions X - and Y -rotations i Ĥ0 tp = which is equal to Ẑ(θ) for θ = ω0tp. R̂z (ω0tp) = exp in terms of the propagator |ψ(tp)i = R̂z (ω0tp)|ψ(0)i • solution of Schödinger equation • Hamiltonian corresponds to free evolution of a spin-1/2 system without r.f. fields Z -rotation rotations via NMR techniques 548 547 • solution of Schödinger equation 549 550 for a pulse at resonant frequency ω of duration tp, which corresponds to the nutation angle θp = ωnuttp, and general phase φp can be given as |ψ(tp)i = χ̂(φp, θp)|ψ(0)i where the propagator is X̂(θ) = χ̂(0, θ) = Ẑ(φp)X̂(θp)Ẑ(−φp) i h1 ′ Ĥrf,rot t χ̂(φ , θ ) = exp p p p i h ih̄ i = exp − θp(σ̂x cos φp + σ̂y sin φp) 2 " # θ θ cos p −i exp(−iφp) sin 2p 2 = θ θ −i exp(iφp) sin 2p cos 2p • special cases Ŷ (θ) = χ̂(π/2, θ) selective rotations in a quartit √1 1 0 0 0 − √i2 0 0 i 1 2 0 0 − √i √1 0 √2 − √2 0 2 2 X̂01( π2 ) = , , X̂12( π2 ) = 0 0 − √i2 √12 0 0 1 0 0 0 0 1 0 0 0 1 √1 √1 − √1 0 0 0 − √i2 0 2 2 2 √1 √1 1 0 0 0 0 0 X̂02( π2 ) = √i , Ŷ01( π2 ) = 2 2 , − 2 0 √12 0 0 0 1 0 0 0 0 1 0 0 0 1 0 −i 0 0 0 −1 0 0 −i 0 0 0 , 1 0 0 0 , ··· X̂01(π) = Ŷ01(π) = 0 0 1 0 0 0 1 0 0 0 0 1 0 0 0 1 Who has introduced NMR quantum computing? the method has been developed independently by: 551 1. Cory, Fahmy, Havel in the article on "NMR spectroscopy: an experimentally accessible paradigm for quantum computing" (1996) 552 2. Gershenfeld and Chuang in "Bulk quantum computation" (1996). How to generate pseudo-pure states? pseudo-pure states (PPS) or effective pure states can be obtained via 1. spatial averaging 2. temporal averaging 3. logical labeling Here, we describe only a version of spatial averaging. How to obtain pseudo-pure states? population of the spin energy level |ki Nk = A + (4 − k)∆ where N 5N h̄ω N h̄ω0 0 A = − , ∆= , 4 8kB T 4kB T N – total umber of spins ω0 – Larmor frequency Note: A does not contribute to the observed NMR signal Assumption: quadrupole interaction ≪ Zeeman interaction ••↔ ∆ – relative occupation of the corresponding quantum level |3i •• |2i •••• |1i •••••• |0i •••••••• thermal or equivalently •• ••••’ ••••••’ ••••’ X̂23( π2 ) −→ •••• ••••’ •••••••• •••••••• |0i •••• ••••••••’ X̂23(π) ••••’ −→ •••• |2i •• ••••••’ X̂23(π) •••••• −→ ••••••’ −|3i ••••••’ ••’ X̂12(π) •••••• −→ •••••• −|2i • states (−|ki) •••• •••• X̂12(π) ••••••••’ −→ ••••’ |1i |3i •• |2i •••• X̂02( π2 ) |1i •••••• −→ |0i •••••••• thermal |3i •••• |2i •••• X̂01(π) |1i •••• −→ |0i •••••••• |0i • states |ki (k = 1, 2, 3): 554 •••••• ••••••’ ••’ •••••• −|1i ••••••••’ ••••’ •••• •••• |3i How to obtain pseudo-pure states in a quartit |3i •• ••••’ |2i •••• •••• X̂13( π2 ) −→ ••••’ |1i •••••• |0i •••••••• •••••••• thermal |0i thermal |3i •• |2i •••• X̂12(π) −→ |1i •••••• |0i •••••••• How to obtain pseudo-pure state |0i in a quartit 553 556 1 2 P k |ki •• •••••’ ••••• •••• •••• •••••’ X̂23( π2 ) X̂03( π2 ) −→ •••••• −→ •••••’ •••••• •••••••• •••••’ ••••• 1 P |ki thermal k 2 |3i |2i |1i |0i Ĥ A ⊗ Ĥ B |0i = Apply gate equivalent to two-qubit Hadamard gate. How to prepare equally-weighted superposition of pseudo-pure states in a quartit? Whether |11i or |00i is the most populated depends on our labeling and direction of external magnetic field. Note: here |11i – is the most populated state in equilibrium. Initial state Interchanging pseudo-pure states in a quartit 555 (a) (b) (c) (d) (e) 1 2 3 4 1 01 2/3 01 2 01 2/3 01 2/3 01 PPS 0 1 1 4 0 11 2 00 2/3 11 2 00 2/3 11 2/3 00 2/3 11 2/3 00 2 11 2/3 00 1 10 3 2 2/3 10 2/3 10 2 10 2/3 10 Eq j00i pps j01i pps j10i pps j11i pps PPS − 1 0 557 PPS − 1 1 ω 01 ω12 ω 23 ω 23 ω 01 ω12 558 PPS - pseudo-pure states, Eq - equilibrium [Mahesh et al.’03] ω 01 ω12 ω 23 PPS 0 0 NMR spectra for a quartit population , energy level NMR spectra for pseudo-pure states of two spin-1/2 systems Key: equilibrium ω 01 ω12 ω 23 ω 01 ω12 ω 23 ω 01 ω12 ω 23 3 2 1 0 [Hirayama et al.’06] ≡ 3 !ω 23 Rk (θ ) 2 1 0 ≡ Rk (θ ) Rk (θ ) experimental generation of pseudo-pure states in solid-state systems of 69Ga nuclei Rk (θ ) How to rotate “qubit” B in a quartit? qubit A qubit B !ω 01 Rk (θ ) 559 560 ω 01 ω12 ω 23 Rk (θ ) ≡ 0 1 !ω 02 Rk (θ ) 2 !ω13 Rk (θ ) ≡ ≡ Y (−π ) Rk (θ k ) Rk (θ ) Y (π ) 0 1 2 3 ≡ Y (−π ) Rk (θ k ) Rk (θ ) ω 01 ω13 ω 02 ω 23 ω12 ω 01 ω 23 Y (π ) ω12 562 Rk (θ ) Rk (θ ) Another method to rotate “qubit” A in a quartit qubit B qubit A 3 How to rotate “qubit” A in a quartit? 561 01 00 01 00 pulses are applied simultaneously 10 11 10 11 RB 564 RC 000 000 RA simultaneous application of pulses 001 001 110 101 111 100 011 010 RB 100 011 010 110 101 111 000 001 100 011 010 101 110 111 rotations of a virtual qubit in 8D qudit RA rotations of virtual qubit in quartit 563 1 0 0 0 0 0 0 1 = iX̂(π) 0 1 A = ÛNOT = iX̂02(π)X̂13(π) 0 0 0 1 1 0 NOT gates via rotations • NOT gate for a qubit ÛNOT = σ̂x = 0 0 0 1 • NOT gate for qubit A in a quartit 0 0 A ÛNOT = ÛNOT ⊗ Iˆ = 1 0 1 0 0 0 0 0 B = ÛNOT = iX̂01(π)X̂23(π) 1 0 • NOT gate for qubit B in a quartit 0 B 1 ÛNOT = Iˆ ⊗ ÛNOT = 0 0 (a) NOP (t) NOT(I2)+ XOR1 (n) SWAP+ XNOR2 (h) XNOR2 (b) NOT(I1) (u) NOT(I1)+ XNOR2 (d) NOT(I1,I2) (e) XOR1 (f) XOR2 (j) SWAP+NOT (k) SWAP+XOR1(l) SWAP+XOR2 (v) NOT(I2)+ XNOR1 (w) SWAP+ NOT(I1) (x) SWAP+ NOT(I2) (o) SWAP+NOT (p) SWAP+NOT (q) SWAP+NOT (r) SWAP+NOT +XOR1 +XOR2 +XNOR1 +XNOR2 (i) SWAP (c) NOT(I2) NMR spectra for classical gates in a spin-1/2 system (g) XNOR1 (m) SWAP+ XNOR1 (s) NOT(I1)+ XOR2 [Mahesh et al.’03] 565 566 Hadamard gates via rotations • Hadamard gate for a qubit 0 1 0 1 1 0 −1 0 0 0 1 1 0 0 = iX̂01(π)Ŷ01( π2 )X̂23(π)Ŷ23( π2 ) 1 −1 567 568 0 1 = iŶ12(π)X̂01(π)Ŷ01( π2 )X̂23(−π)Ŷ23(− π2 )Ŷ12(−π) 0 −1 for qubit A in a quartit 1 1 1 = iX̂(π)Ŷ ( π2 ) = iŶ ( π2 )Ẑ(π) Ĥ = √ 2 1 −1 • Hadamard gate 1 1 0 Ĥ A = Ĥ⊗Iˆ = √ 2 1 0 1 −1 0 0 • Hadamard gate for qubit B in a quartit 1 1 1 Ĥ B = Iˆ ⊗ Ĥ = √ 2 0 0 X (π ) X (π ) H ≡ ≡ ω 01 ω12 Z (π ) Y ( π2 ) Y ( π2 ) ω 23 Z (π ) HB Equivalent realizations of Hadamard gate H B (up to factor i) qubit A Y ( π2 ) qubit B ≅ Y ( π2 ) ≡ 0 1 0 0 0 0 0 1 CNOT1 in | out -------00 | 00 01 | 01 10 | 11 11 | 10 ÛCNOT2 ÛCNOT1 Y (− π2 ) Y ( π2 ) ≅ Y (π ) Y (−π ) 0 0 0 1 0 0 1 0 ≡ qubit B qubit A H CNOT2 in | out -------00 | 00 01 | 11 10 | 10 11 | 01 SWAP in | out -------00 | 00 01 | 10 10 | 01 11 | 11 570 Y (π ) 0 0 = diag([1, 1, −1, 1]) · Ŷ23(π), 1 0 0 1 = diag([1, 1, 1, −1]) · Y12(π)Ŷ23(π)Ŷ12(−π) 0 0 truth tables X (π ) Y (− π2 ) X (−π ) Y ( π2 ) CNOT gates via rotations Z (π ) Z (π ) HA 1 0 = 0 0 1 0 = 0 0 Y (−π ) ω 23 ω12 ω 01 569 A Equivalent realizations of Hadamard gate H (up to factor i) 00 00 01 10 10 01 11 11 CNOT21 572 100 011 010 001 000 001 000 110 101 111 100 011 010 110 101 111 CNOT32 pulses are applied simultaneously CNOT23 CNOT gates of 3 virtual qubits #1 CNOT12 CNOT gates of 2 virtual qubits 571 CNOT12 573 CNOT gates of 3 virtual qubits #2 111 110 101 111 010 100 011 110 101 111 010 100 011 110 101 111 CNOT31 110 101 100 011 010 001 CNOT13 100 011 010 001 001 000 CNOT21 001 000 0 1 0 0 0 0 = ÛCNOT1ÛCNOT2ÛCNOT1 0 1 574 000 000 pulses are applied simultaneously 0 0 1 0 SWAP gates via rotations 1 0 ÛSWAP = 0 0 0 0 1 0 0 −1 0 0 1 ′′ 0 ÛSWAP = 0 0 similar gates 0 0 = Ŷ12(π), 0 1 relation between them 0 0 −i 0 0 −i 0 0 0 0 = X̂12(π) 0 1 which acts as follows ÛSWAP c0|00i + c1|01i + c2|10i + c3|11i = c0|00i + c2|01i + c1|10i + c3|11i 1 ′ 0 ÛSWAP = 0 0 ÛSWAP ′ ′ = ÛSWAP · diag([1, 1, −1, 1]) = diag([1, −1, 1, 1]) · ÛSWAP ′′ ′′ = ÛSWAP · diag([1, i, i, 1]) = diag([1, i, i, 1]) · ÛSWAP NMR implementations of two-qubit algorithms 575 [Leung et al.’99] [Fu et al.’99] [Jones et al.’98] 1. Deutsch-Jozsa algorithm [Chuang et al.’98, Jones et al.’98] 2. Grover search algorithm 3. quantum Fourier transform 4. quantum error detection [Fang et al.’99] [Somaroo et al.’99] 6. dense coding [Zhu et al.’01] 5. quantum simulations 7. Hogg algorithm [Teklemarian et al.’02] 576 8. quantum erasers Hogg algorithm – a highly structured search algorithm What can be done with two-virtual qubits? NMR implementations of two-qubit algorithms on spin-3/2 nuclei: 1. demonstration of classical gates [Khitrin et al.’00, Sinha et al.’01, Kumar et al.’02] 2. demonstration of quantum gates [Sarthour et al.’03, Kampermann et al.’02, Steffen’03, 3. generation of Bell states [Sarthour et al.’03, Kampermann et al.’02] 4. quantum tomography [Bonk et al.’04, Steffen’03, Kampermann et al.’05] 577 (∗) - not implemented yet ω 23 ω12 ω 01 H qubit B 1 y x Uf y ⊕ f ( x) x H HA HB Uf HA Deutsch's algorithm for a quartit H qubit A 0 Deutsch's algorithm for two qubits 578 12. ∗ quantum erasers 11. ∗ Hogg algorithm 10. ∗ dense coding 9. ∗ quantum simulations 8. ∗ quantum error detection [Kampermann et al.’05] 7. quantum Fourier transform [Das et al.’03, Kampermann et al.’05] 6. Deutsch-Jozsa algorithm [Ermakov et al.’02, Steffen’03, Kampermann et al.’05] 5. Grover search algorithm 2 π Uf 2 π Uf 580 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 1 0 0 1 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 1 0 1 f3 0 0 1 0 Pulse no pulse X̂01(π)X̂23(π) X̂23(π) Uf 0 1 f2 f1 1 0 0 0 0 0 1 0 3 1 , 2 2 3 3 , 2 2 1 ≡ 01 ≡ 0 ≡ 00 ≡ 3 1 ,− 2 2 3 3 ,− 2 2 2 ≡ 10 ≡ 3 ≡ 11 ≡ 0 0 0 1 X̂01(π ) 0 1 0 0 1 0 f4 NMR pulses for the oracle in Deutsch’s algorithm Constant Balanced ω 23 ω12 ω 01 experimental Deutsch's algorithm for a quartit qubit B 0 qubit A 0 579 experimental Deutsch's algorithm for two qubits ω ω 1 Û1 2 4 Û 2 3 ω ω ω 1 4 Û 3 ω ω2 3 |10〉 ω1 ω2 ω3 ω4 |11〉 | 00〉 Û 3 581 Û 4 Û 4 582 ω1 ω2 ω3 ω4 exemplary NMR spectra for Deutsch algorithm in 2 qubits ω ω 3 ω1 ω2 | 01〉 ω4 Û 2 ω01 ω12 ω23 ω01 ω12 ω23 exemplary NMR spectra for Deutsch algorithm in a quartit Û1 ω01 ω12 ω23 |1〉 | 2〉 | 3〉 ω01 ω12 ω23 ω 23 ω12 ω01 | 0〉 HA HB step II Z ( π2 ) Z (−π ) Z ( 32π ) III.1 HA HB III.2 a circuit for Grover’s search in a quartit ω 01 ω12 ω 23 H H step II f Xo III.1 H H III.2 a circuit for Grover search in two qubits qubit A qubit B III.3 Z ( π2 ) Z (−π ) Z ( 32π ) III.3 f0 NMR detections of magnetization of a spin-3/2 III.4 HA HB III.4 H H 583 584 • Mxy -detection of a two spin-1/2 system enables determination of the offdiagonal elements marked in boxes: ρ00 ρ01 ρ02 ρ03 ρ ρ 10 ρ11 ρ12 13 ρ̂ = ρ20 ρ21 ρ22 ρ23 ρ30 ρ31 ρ32 ρ33 ρ03 ρ13 ρ23 ρ33 • Mxy -detection of a spin-3/2 system gives the other off-diagonal elements: ρ00 ρ01 ρ02 ρ03 10 ρ11 ρ12 ρ13 ρ ρ̂ = ρ20 ρ21 ρ22 ρ23 ρ30 ρ31 ρ32 ρ33 • Mz -detection of a spin-3/2 system gives: ρ00 ρ01 ρ02 ρ ρ10 11 ρ12 ρ̂ = ρ ρ 20 ρ21 22 ρ30 ρ31 ρ32 ≡ Ŷ ( π2 ) R̂3 = Iˆ ⊗ Ŷ , R̂6 = X̂ ⊗ Ŷ , R̂9 = Ŷ ⊗ Ŷ 586 ⇒ XXXXXX, IIIIII, YYYYYY, IIXXYY, XXYYII, YYIIXX, IXIYXY, XYXIYI, YIYXIX, IIXYYX, XXYIIY, YYIXXI, IIYIXI, XXIXYX, YYXYIY, IYYXIX, XIIYXY, YXXIYI, IXIIIY, IYYXXI, XIIYYX, YIXXIY, YXXIXX, XYXYII, XYYIYY, YXYYYI, IIIXYI, IYIYXX, XIYXXX, XYXIIY, IIXIXI, IXXYYX, XXIXIX, YIIYXY, YXYIYY, IYYIIX, YYIXXY, IXYYXY, YXYYII, YYIIYX, IIIXII, XIXXXX, IYXXYY, XXXXXI, XYYYYY, IIXYIX, YIYXIY, IXIIXY, XIIIII, IXXXYY, IYIYXI, XYYIYI, YXXYXX, XXXIIY, YIIIII, YIYXYX, XYIIXX, YYXYII, YYYXXI, IXXYIX, YXIXIX, YXIYYI, IIIXYY, XIXYXY, XYXXXX, IIYIXY, IXYIYX, XIYYII, XXIXYY, XYYYYX, YIXIYI, YYYIIX, IXIYIY, IYIIYX, XXYXXX, XYIXIY, YIXIYY, IXIIXI, IXXXII, IXYXII, IYIYYI, IYXIXI, XIIIYX, XIIXYI, YIYYXX, YXYIXY, YYXXYX, IIIXXX, IIXIIX, IIYYII, XXIIXI, XXXYYX, XYYXIY, IIXYXI, IIYXYY, IYIYIX, IYXIXY, XIXXXI, YIIYXX, YXXXYY, IIIIXX, IIXYYI, IIYIIX, IIYIYX, IIYYIY, IXIYXX, IXYXYI, IYIIIY, IYXXYX, IYYYXI n=6 ⇒ XXXXX, IIIII, YYYYY, IIIXX, XXYYY, YYXII, IIXIY, XYXXI, YXIYY, XXYIX, IYIXX, IIXYI, YIYXI, XXIYY, YXYXI, YYXIX, IYYYY, XIXIY, IIIYI, IXYIX, XIYYX, IXXXY, XYIXI, YXIIY, XIYIX, XYYXI, YIIXX, YYXYX, IIXXY, IXXYI, XYIII, IYYIY, YYXYY, IXIXX, XIXXY, XXXIX, XYIYI, YXYXX, IIYIY, IYIXY n=5 YYXY, XYXI, IIYX, IXIY, IIXI, IYIY 2. other sets of rotations for tomography of n-qubits n=1 ⇒ X, I n=2 ⇒ XX, II, IX, IY; n=3 ⇒ XXX, III, IIY, XYX, YII, XXY, IYY n=4 ⇒ XXXX, IIII, IIYY, YYXX, IIIY, XYXX, YXYI, IXYI, IIIX, XIYY, YXII, X̂( π2 ), Ŷ R̂2 = Iˆ ⊗ X̂, R̂5 = X̂ ⊗ X̂, R̂8 = Ŷ ⊗ X̂, where X̂ ≡ ˆ R̂1 = Iˆ ⊗ I, ˆ R̂4 = X̂ ⊗ I, ˆ R̂7 = Ŷ ⊗ I, 1. Chuang’s set of rotations for Mxy tomography of 2 qubits ρ̂(k) = R̂k ρ̂R̂k† • The remaining matrix elements can be obtained by rotating the original density matrix ρ̂ through properly chosen rotational operations R̂k : • Mxy and Mz detections give only some elements of ρ̂ is a method for complete reconstruction of a given density matrix ρ̂ in a serious of NMR measurements. Principle ideas NMR quantum state tomography 585 X 02 X 13 X 03 00 01 10 • • • R̂11 = X̂03( π2 ), three-photon transitions = X̂02( π2 ), R̂9 = X̂13( π2 ), R̂7 two-photon transitions R̂1 = X̂01( π2 ), R̂3 = X̂12( π2 ), R̂5 = X̂23( π2 ), single-photon transitions R̂12 = Ŷ03( π2 ) R̂10 = Ŷ13( π2 ) R̂8 = Ŷ02( π2 ) = Ŷ01( π2 ) R̂4 = Ŷ12( π2 ) R̂6 = Ŷ23( π2 ) R̂2 a natural set of rotations for tomography of a quartit state (3+2+1)=6 so 12 readouts (including Y-rotations) X 01 X 12 X 23 11 588 tomography of 4-level system 587 rotations in Mz -based tomography e.g. ρ̂ after R̂1 = X̂01( π2 ) ... ... ρ22 ... ... ... ... ρ33 ρ̂ after R̂5 = X̂23( π2 ) ... ... ... ρ11 ... ... ... 1 (ρ − ρ23 − ρ32 + ρ33) ... 2 22 1 ... ... 2 (ρ22 + ρ23 + ρ32 + ρ33 ) 1 ... 2 (ρ00 − ρ01 − ρ10 + ρ11) 1 ... 2 (ρ00 + ρ01 + ρ10 + ρ11) ρ̂′ = ... ... ... ... ρ00 ... ρ̂′′ = ... ... so let us apply both rotation before read-outs Ŷ13(π)X̂01(θ)Ŷ13(−π) Ŷ02(π)X̂23(−θ)Ŷ02(−π) Ŷ01(π)X̂13(−θ)Ŷ01(−π) Ŷ23(π)X̂02(θ)Ŷ23(−π) Ŷ01(π)Ŷ23(π)X̂12(−θ)Ŷ23(−π)Ŷ01(−π) ··· tomography without three-photon rotations X̂03(θ) = = = = = = Ŷ23(π)X̂12(θ)Ŷ23(−π) Ŷ12(π)X̂23(−θ)Ŷ12(−π), Ŷ23(π)Ŷ12(θ)Ŷ23(−π) Ŷ12(π)Ŷ23(−θ)Ŷ12(−π), Ŷ12(π)X̂01(θ)Ŷ12(−π) Ŷ01(π)X̂12(−θ)Ŷ01(−π), Ŷ12(π)Ŷ01(θ)Ŷ12(−π) Ŷ01(π)Ŷ12(−θ)Ŷ01(−π). tomography without two-photon rotations X̂13(θ) Ŷ13(θ) X̂02(θ) Ŷ02(θ) = = = = = = = = 589 590 7 591 +5 +4 +3 +2 +1 = 28 so 56 readouts (including X and Y-rotations) [Tseng et al.’00] 000 001 100 011 010 110 101 111 tomography of 8-level system +6 592 [Laflamme et al.’97] NMR implementations of 3-qubit algorithms on spin-1/2 nuclei 1. generation of GHZ states [Cory et al.’98] [Nielsen et al.’98] 2. quantum error correction 3. quantum teleportation [Linden et al.’98] [Kim et al.’00] 4. Deutsch-Jozsa algorithm 5. refined Deutsch-Jozsa algorithm for entangled qubits (a meaningful test of quantum parallelism) 6. Grover algorithm (with cancellation of systematic errors) 7. quantum simulation [Chang et al.’01] [Vandersypen et al.’00] 8. Schulman-Vazirani algorithm (a cooling scheme) [Viola et al.’01] [Weinstein et al.’01] 9. noiseless subsystems 10. quantum Fourier transform [Peng et al.’02] 14. Hogg algorithm [Vandersypen et al.’04] [Steffen et al.’03] [Khitrin et al.’01] 10. logical labeling 9. quantum teleportation 8. quantum error correction 7. generation of GHZ states 6. quantum tomography • not implemented yet 3. test of phase coherence in EIT [Murali et al.’04] 2. half-adder and subtractor operations [Murali et al.’02, Kumar et al.’02] 1. quantum simulation • implemented (in liquid NMR systems) 596 595 2018 - estimated year for reaching the fundamental limits • the end of Moore’s law? transistors 2007 - Dual-Core Intel Itanium 2 chip contains 1.7 billion 1965 - approximately 60 devices on a chip. The number doubles every 24 months. • modified Moore’s law (1975) The number of transistors and resistors on a chip doubles every 18 months. • original Moore’s law (1965) Gordon E. Moore - a co-founder of Intel. possible QIP applications of spin-7/2 nuclei 21. adiabatic quantum optimization algorithm 20. Hogg algorithm empirical Moore’s law 594 [Vandersypen et al.’01] 19. phase estimation algorithm 18. quantum baker’s map 17. quantum erasers 16. quantum Fourier transform 15. noiseless subsystems 14. Schulman-Vazirani algorithm (a cooling scheme) 13. Grover algorithm (with cancellation of systematic errors) (a meaningful test of quantum parallelism) 12. refined Deutsch-Jozsa algorithm for entangled qubits 11. Deutsch-Jozsa algorithm What can be done with three virtual qubits? Shor’s factorization algorithm NMR implementation of a seven-qubit algorithm [Murali et al.’04] 18. test of phase coherence in electromagnetically induced transparency (EIT) 17. quantum state and process tomography 16. adiabatic quantum optimization algorithm 15. half-adder and subtractor operations [Murali et al.’02, Kumar et al.’02] [Lee et al.’02] [Weinstein et al.’02] 593 [Teklemarian et al.’01] 13. phase estimation algorithm 12. quantum chaotic map (baker’s map) 11. quantum erasers [source: Wikipedia] end of Moore’s Law? • how to provide energy to a chip? • how to cool down a chip? As power-driven heat can cause major malfunctions. 597 598 “Chip with 3nm-length gates would overheat itself.” [Gargini] • when the gate length < 5 nm quantum effects become important gate lengths 37 nm in 2007 ∼5 nm in 2015-2018 quantum tunneling source & drain are so close that the electrons will tunnel even if voltage is not applied to the gate ⇒ Heisenberg uncertainty becomes important ⇒ transistor becomes unreliable Physical principles of quantum computing 1. Superposition 2. Interference 3. Entanglement 4. Non-cloning 5. Uncertainty 599 600 purely quantum applications of QC 1. Quantum cryptography 2. Quantum teleportation advantages of QC • superfast (Shor) and fast (Grover) algorithms • understanding new aspects of measurement theory • improvement of precision spectroscopy • understanding dissipation in mesoscopic system • partial control of decoherence • quantum state engineering • quantum simulations "Quantum computers must be a component of any world view that seeks to be fundamental" David Deutsch • Information Science • Cryptography • Quantum physics including • Quantum Optics • Nanotechnology Quantum Computing is the frontier of 601