New Examples of Perfect State Transfer - IME-USP
Transcription
New Examples of Perfect State Transfer - IME-USP
New Examples of Perfect State Transfer Gabriel Coutinho University of Waterloo A conference to celebrate the work of Chris Godsil. 2 2 2 3 Perfect state transfer: I Motivation from quantum computing. I Example, definition, some known results. I Our methods to find new examples. 3 Model: To each vertex of a graph, assign a 2-dim vector space. 4 Model: To each vertex of a graph, assign a 2-dim vector space. A 2-dim space is a basic unit of quantum information, or a qubit. 4 Model: To each vertex of a graph, assign a 2-dim vector space. A 2-dim space is a basic unit of quantum information, or a qubit. Graph is associated to a vector space of dimension 2|V | . 4 Model: To each vertex of a graph, assign a 2-dim vector space. A 2-dim space is a basic unit of quantum information, or a qubit. Graph is associated to a vector space of dimension 2|V | . State: corresponding to 1 0 0 0 ⊗ ⊗ ⊗ 0 1 1 1 4 Next step: define a Hamiltonian to govern the time-evolution of this quantum system. 5 Next step: define a Hamiltonian to govern the time-evolution of this quantum system. That is, a Hermitian matrix H of dimension 24 . 5 Next step: define a Hamiltonian to govern the time-evolution of this quantum system. That is, a Hermitian matrix H of dimension 24 . Choice: a time independent Hamiltonian that encodes precisely the adjacency of the graph. 5 We are interested in a perfect transfer of state, ie, a time t such that evolves to (up to a factor λ ∈ C) 6 We are interested in a perfect transfer of state, ie, a time t such that evolves to (up to a factor λ ∈ C) 6 This is related to the problem of transferring information in certain models of quantum computing. The action of our Hamiltonian is equivalent to the action of the adjacency matrix A = A(X ). 7 The action of our Hamiltonian is equivalent to the action of the adjacency matrix A = A(X ). By Schrödinger’s Equation, at time t ∈ R+ , this means that the evolution will be determined by exp(itA) = X (itA)k k≥0 7 k! . The action of our Hamiltonian is equivalent to the action of the adjacency matrix A = A(X ). By Schrödinger’s Equation, at time t ∈ R+ , this means that the evolution will be determined by exp(itA) = X (itA)k k≥0 k! . So we are looking for a time t such that 1 0 0 0 exp(itA) 0 = λ 0 0 where |λ| = 1. 7 Formalization: We say that X admits perfect state transfer from vertex u to vertex v if there is a time t ∈ R+ and a phase λ ∈ C such that exp(itA)eu = λev or equivalently | exp(itA)u,v | = 1. We say that u is periodic if u = v above. The existence of periodic vertices depends only on the eigenvalues, whereas state transfer depends also on the structure of the graph. 8 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 9 (even) A 1 0 = 0 1 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 k! = cos(t) Hence exp(itA) 1 0 0 1 + i sin(t) 0 1 1 0 1 cos(t) = 0 i sin(t) 9 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t=0 9 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 2π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 3π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 4π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 π 2 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 6π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 7π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 8π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t= 9 9π 10 Suppose X = K2 . In this case, 0 1 0 1 (odd) A= A = 1 0 1 0 (even) A 1 0 = 0 1 So it is easy to see that exp(itA) = X (itA)k k≥0 Hence k! 1 0 0 1 = cos(t) + i sin(t) 0 1 1 0 1 cos(t) exp(itA) = 0 i sin(t) t=π 9 A taste of how things are in general.... Suppose A = A(X ) has distinct eigenvalues θ0 > ... > θd . 10 A taste of how things are in general.... Suppose A = A(X ) has distinct eigenvalues θ0 > ... > θd . Spectral decomposition: A= d X θr E r r =0 Function on A ⇒ function on the eigenvalues. 10 A taste of how things are in general.... Suppose A = A(X ) has distinct eigenvalues θ0 > ... > θd . Spectral decomposition: A= d X θr E r r =0 Function on A ⇒ function on the eigenvalues. Thus: ! ! d d X X PST itθr exp(itA)eu = e Er eu = λev = λ Er ev r =0 r =0 10 A taste of how things are in general.... Suppose A = A(X ) has distinct eigenvalues θ0 > ... > θd . Spectral decomposition: A= d X θr E r r =0 Function on A ⇒ function on the eigenvalues. Thus: ! ! d d X X PST itθr exp(itA)eu = e Er eu = λev = λ Er ev r =0 r =0 Necessary condition for PST: Er eu = ±Er ev for all r . 10 Known results: Problem was introduced by Bose in 2003 and subsequently studied in more detail by Christandl and others in 2005. 11 Known results: Problem was introduced by Bose in 2003 and subsequently studied in more detail by Christandl and others in 2005. Perfect state transfer happens in: Paths P2 and P3 Iterated cartesian powers - hypercubes. 11 Known results: Problem was introduced by Bose in 2003 and subsequently studied in more detail by Christandl and others in 2005. Perfect state transfer happens in: Paths P2 and P3 Iterated cartesian powers - hypercubes. Does not happen in: Paths Pn , n > 3. 11 Known results: Problem was introduced by Bose in 2003 and subsequently studied in more detail by Christandl and others in 2005. Perfect state transfer happens in: Paths P2 and P3 Iterated cartesian powers - hypercubes. Does not happen in: Paths Pn , n > 3. It doesn’t happen in complete graphs Kn with n > 2. 11 Known results: Problem was introduced by Bose in 2003 and subsequently studied in more detail by Christandl and others in 2005. Perfect state transfer happens in: Paths P2 and P3 Iterated cartesian powers - hypercubes. Does not happen in: Paths Pn , n > 3. It doesn’t happen in complete graphs Kn with n > 2. Theorem (Godsil ’12) There are only finitely many graph with maximum valency k in which perfect state transfer can occur. 11 Perfect state transfer in circulants (Cayley graphs for Zn ) was characterized by Bašić and others 2011 and 2013. 12 Perfect state transfer in circulants (Cayley graphs for Zn ) was characterized by Bašić and others 2011 and 2013. Perfect state transfer in cubelike graphs (Cayley graphs for Z2d ) was characterized by Godsil and others in 2008 and 2011. 12 Perfect state transfer in circulants (Cayley graphs for Zn ) was characterized by Bašić and others 2011 and 2013. Perfect state transfer in cubelike graphs (Cayley graphs for Z2d ) was characterized by Godsil and others in 2008 and 2011. The effect of certain graph operations (joins, products) was studied by Tamon and others 2009-2012. 12 Cartesian product: Consider graphs X and Y and their cartesian product X Y . 13 Cartesian product: Consider graphs X and Y and their cartesian product X Y . We have A(X Y ) = A(X ) ⊗ I + I ⊗ A(Y ) As a consequence (Christandl et al. ’05) exp itA(X Y ) = exp itA(X ) ⊗ exp itA(Y ) 13 Cartesian product: Consider graphs X and Y and their cartesian product X Y . We have A(X Y ) = A(X ) ⊗ I + I ⊗ A(Y ) As a consequence (Christandl et al. ’05) exp itA(X Y ) = exp itA(X ) ⊗ exp itA(Y ) Thus X Y admits perfect state transfer at time τ if and only if I X admits perfect state transfer at time τ , and I Y admits perfect state transfer at time τ or contains a periodic vertex at time τ . 13 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 Example: P3 P3 admits perfect state transfer between the pairs of π vertices below at time √ . 2 Similarly, P3 K1,8 admits perfect state transfer. 14 I Our goal: find more examples of perfect state transfer. 15 I Our goal: find more examples of perfect state transfer. I Our method: use a general form of graph products. 15 I Our goal: find more examples of perfect state transfer. I Our method: use a general form of graph products. I Suppose X is a graph such that A(X ) = B1 ⊗ B2 + C1 ⊗ C2 , where B1 , B2 , C1 , C2 are 01-symmetric matrices (NEPS). 15 I Our goal: find more examples of perfect state transfer. I Our method: use a general form of graph products. I Suppose X is a graph such that A(X ) = B1 ⊗ B2 + C1 ⊗ C2 , where B1 , B2 , C1 , C2 are 01-symmetric matrices (NEPS). I This generalizes: cartesian, direct, lexicographic products. I Double covers can also be expressed in this way. 15 Direct product: I I I Defined as A(X × Y ) = A(X ) ⊗ A(Y ). If perfect state transfer happens in the direct product X × Y , it must happen in at least one of the factors. On the other hand, the conditions required in the other factor are fairly weak. 16 Direct product: I I I Defined as A(X × Y ) = A(X ) ⊗ A(Y ). If perfect state transfer happens in the direct product X × Y , it must happen in at least one of the factors. On the other hand, the conditions required in the other factor are fairly weak. Theorem (Coutinho and Godsil ’13) If ∗ Y admits perfect state transfer, and ∗ eigenvalues of X and Y are integers or integer multiples of a square root, and ∗ integer parts of eigenvalues of X are all divisible by same powers of 2, then there is a k0 ∈ Z+ such that X × Y m.k0 admits perfect state transfer for all m ∈ Z. 16 Two families of examples: If Y admits perfect state transfer at time π 2, then I K1,n × Y 2m for all n and m. I if X is one of the 32548 non-isomorphic strongly regular graphs with parameters (36, 15, 6, 6) and eigenvalues {15, 3, −3}, then X × Y 2m admits perfect state transfer. 17 Double covers: Given graphs X and Y on the same set of vertices, we define the graph X n Y by A(X ) A(Y ) A(X n Y ) = A(Y ) A(X ) I If A(X ) ◦ A(Y ) = 0, this is a double cover of the graph with adjacency matrix A(X ) + A(Y ). I If Y = X , this is a double cover of the complete graph and is known as the switching graph of X . 18 Theorem (Coutinho and Godsil ’14) If A(X ) and A(Y ) commute, then perfect state transfer happens in X n Y if and only if both X and Y are periodic at the same vertex at the same time with certain phase factors. 19 Theorem (Coutinho and Godsil ’14) If A(X ) and A(Y ) commute, then perfect state transfer happens in X n Y if and only if both X and Y are periodic at the same vertex at the same time with certain phase factors. I If X = Kn Kn , then X n X admits perfect state transfer for all n divisible by 4. 19 Open questions: I Find examples that do not depend on previous constructions. I Characterize perfect state transfer in Cayley graphs for abelian groups. I Characterize perfect state transfer in trees (possibly the only trees admitting pst are P2 and P3 ...) 20 21 22 23 24 25