combining tensor networks and monte carlo for lattice
play

Combining Tensor Networks and Monte Carlo for Lattice Gauge - PowerPoint PPT Presentation

Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories Entanglement in Strongly Correlated Systems, Benasque 21st of February 2020 Patrick Emonts , E. Zohar, M. C. Banuls, I. Cirac MPI of Quantum Optics Why do we need Lattice


  1. Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories Entanglement in Strongly Correlated Systems, Benasque 21st of February 2020 Patrick Emonts , E. Zohar, M. C. Banuls, I. Cirac MPI of Quantum Optics

  2. Why do we need Lattice Gauge Theories? QED QCD L QED = iΨγ μ 𝜖 μ Ψ − eΨγ μ A μ Ψ − mΨΨ − 1 4 F μν F μν e − e − γ e + e + Small coupling α QED = e 2 1 4 π ≈ 137 Image adapted from Alexandre Deur, Stanley J. Brodsky, and Guy F. de Téramond, 2016, Progress in Particle and Nuclear Physics Slide 2 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  3. Path integral formalism in QFT pure QED S QED [ A μ ] = − 1 4 ∫ d x α F μν ( x α ) F μν ( x α ) = ∫ d x α 𝜖 μ A ν ( x α )𝜖 ν A μ ( x α ) vacuum expectation value Problems iSQED [ Aμ ]  Numerator oscillating ∫ D AO [ A μ ] e ⟨ Ω | O [ A μ ]| Ω ⟩ = iSQED [ Aμ ]  Integration measure ill-defined ∫ D Ae Slide 3 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  4. Wick rotation Shift to imaginary time t → − iτ Change of metric from Minkowski to Euclidean e iS M = e i ∫ d x α M L ( x α M ) ⟶ e − ∫ d x α E L ( x α E ) = e − S E Problems  Numerator converging  Integration measure ill-defined Slide 4 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  5. ̃  ̃ Discretization: Lattice Gauge Theory x α a A μ → U μ = e iaA μ S E that agrees with S E in the continuum limit of vanishing a Find the lattice action S E [ U ] → S E [ A ]( a → 0 ) Kenneth G. Wilson, 1974, Physical Review D Slide 5 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  6. Vacuum expectation value in the action formalism Vacuum expectation value ∫ D UO [ U ] e − SE [ U ] ⟨ O [ U ]⟩ = with D U = ∏ x α d U μ ( x α ) ∫ D Ue − SE [ U ] Problems  Numerator converging  Integration with the Haar measure Slide 6 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  7. Lattice Systems Hilbert space H ⊂ H gauge fields ⊗ H fermions A general state | Ψ ⟩ = ∫ DG | G ⟩ ∣ Ψ F ( G )⟩ with DG = ∏ x , k dg ( x , k ) Erez Zohar and J. Ignacio Cirac, 2018, Physical Review D Patrick Emonts and Erez Zohar, 2020, SciPost Physics Lecture Notes Slide 7 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  8. Gauss law Gauss law ∑ k ( E k ( x ) − E k ( x − e i )) | phys ⟩ = 0 ∀ x E 2 ( x ) E 1 ( x − e 1 ) E 1 ( x ) Classical analogue in (cont.) E 2 ( x − e 2 ) electrodynamics ∇ ⋅ E = 0 Slide 8 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  9. Expectation value of an Observable Assume that O acts only on the gauge field and is diagonal in the group element basis: ⟨ O ⟩ = ⟨ Ψ | O | Ψ ⟩ ⟨ Ψ | Ψ ⟩ = ∫ DG ⟨ G | O | G ⟩ ⟨ Ψ F ( G )∣ Ψ F ( G )⟩ ∫ DG ′ ⟨ Ψ F ( G ′ )∣ Ψ F ( G ′ )⟩ = ∫ DGF O ( G ) p ( G ) ⟨ Ψ F ( G )∣ Ψ F ( G )⟩ ⟨ Ψ F ( G )∣ Ψ F ( G )⟩ with p ( G ) = Z ∫ DG ′ ⟨ Ψ F ( G ′ )∣ Ψ F ( G ′ )⟩ = Slide 9 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  10. The rest of this talk Expectation value ⟨ O ⟩ = ∫ DGF O ( G ) p ( G ) ⟨ Ψ F ( G )∣ Ψ F ( G )⟩ with p ( G ) = Z TODO List 1 How do we construct ∣ Ψ F ( G )⟩ ? 2 How do we efficiently calculate p ( G ) ? 3 Are those states useful? Slide 10 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  11. Creation of the fermionic state Desirable properties | Ψ ⟩ fulfills the Gauss law Definition of Ψ ∣ Ψ F ( G )⟩ allows efficient calculations of | Ψ ⟩ = ∫ DG | G ⟩ ∣ Ψ F ( G )⟩ the norm expectation values Choice for ∣ Ψ F ( G )⟩ We construct ∣ Ψ F ( G )⟩ with a tensor network. Patrick Emonts and Erez Zohar, 2020, SciPost Physics Lecture Notes Slide 11 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  12. Definition of Modes Gauss law in terms of our modes G 0 = E r − E l + E u − E d = r † + r + − r † − r − − l † + l + + l † − l − + u † + u + − u † − u − − d † + d + + d † − d − u − u + Definition of positive and negative l + r + modes Ψ r − l − a : { l + , r − , u − , d + } (neg. modes) b : { l − , r + , u + , d − } (pos. modes) d − d + Erez Zohar et al., 2015, Annals of Physics Slide 12 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  13. ω ( x , k ) ∏ A ( x ) x x , k T ij a † i ( x ) b † A ( x ) = exp ⎛ j ( x )⎞ ij ω ( x , k ) = ω k ( x ) Ω k ( x ) ω † k ( x ) ω 0 ( x ) = exp ( l † + ( x + e 1 ) r † − ( x )) exp ( l † − ( x + e 1 ) r † + ( x )) ∑ ⎜ ⎟ ⎠ ∏ ⎝ Creating a fermionic state The state ∣ ψ 0 ⟩ = ⟨ Ω V ∣ | Ω ⟩ Slide 13 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  14. ω ( x , k ) x , k ω ( x , k ) = ω k ( x ) Ω k ( x ) ω † k ( x ) ω 0 ( x ) = exp ( l † + ( x + e 1 ) r † − ( x )) exp ( l † − ( x + e 1 ) r † + ( x )) ⎟ ⎠ ⎝ ∑ ∏ ∏ ⎜ Creating a fermionic state The state ∣ ψ 0 ⟩ = ⟨ Ω V ∣ A ( x ) | Ω ⟩ x T ij a † i ( x ) b † A ( x ) = exp ⎛ j ( x )⎞ x 01 x 11 x 21 ij x 00 x 10 x 20 Slide 13 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  15. ⎟ ⎠ ∑ ⎝ ⎜ Creating a fermionic state The state ∣ ψ 0 ⟩ = ⟨ Ω V ∣ ∏ ω ( x , k ) ∏ A ( x ) | Ω ⟩ x x , k ω 01 , h ω 11 , h T ij a † i ( x ) b † A ( x ) = exp ⎛ j ( x )⎞ x 01 x 11 x 21 ij ω 00 , v ω 10 , v ω 20 , v ω ( x , k ) = ω k ( x ) Ω k ( x ) ω † k ( x ) ω 00 , h ω 10 , h ω 0 ( x ) = exp ( l † + ( x + e 1 ) r † − ( x )) x 00 x 10 x 20 exp ( l † − ( x + e 1 ) r † + ( x )) Slide 13 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  16. Moving towards local symmetry Lattice Gauge theory We demand a local symmetry ∑ x G ( x ) | Ψ ⟩ = 0 → G ( x ) | Ψ ⟩ = 0 x 01 x 11 x 21 x 00 x 10 x 20 Erez Zohar et al., 2015, Annals of Physics Erez Zohar and Michele Burrello, 2016, New Journal of Physics Slide 14 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  17. Local symmetry – The state Substitution ± ( x ) → e ± iθ ( x ) r † r † ± ( x ) ± ( x ) → e ± iθ ( x ) u † u † ± ( x ) x 01 x 11 x 21 x 00 x 10 x 20 Slide 15 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  18. Fermionic state Fermionic state ∣ ψ ( G )⟩ = ⟨ Ω v ∣ ∏ ω ( x ) ∏ A ( x ) | Ω ⟩ U Φ ( x ) ∏ x x x  Gauge invariance of | Ψ ⟩ by constructing Ψ ( G )  Obeys all demanded symmetries ? Efficient to calculate with Slide 16 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  19. ⎟ ⎜ ⎠ ∑ ⎝ Is ∣ Ψ F ( G )⟩ special? The fermionic state ∣ Ψ F ( G )⟩ ∣ Ψ F ( G )⟩ = ⟨ Ω v ∣ ∏ ω ( x ) ∏ A ( x ) | Ω ⟩ U Φ ( x ) ∏ x x x T ij a † i ( x ) b † A ( x ) = exp ⎛ j ( x )⎞ ij ω ( x ) = ω 0 ( x ) ω 1 ( x ) Ω ( x ) ω † 1 ( x ) ω † 0 ( x ) ω 0 ( x ) = exp ( l † + ( x + e 1 ) r † − ( x )) exp ( l † − ( x + e 1 ) r † + ( x )) ω 1 ( x ) = exp ( d † + ( x + e 2 ) u † − ( x )) exp ( d † − ( x + e 2 ) u † + ( x )) Slide 17 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  20. Gaussian States Definition Fermionic Gaussian states are represented by density operators that are exponentials of a quadratic form in Majorana operators. ρ = K exp (− i 4 γ T Gγ ) Covariance matrix Covariance matrix for a state Φ : ⟨ Φ |[ γ a , γ b ]| Φ ⟩ Γ ab = i 2 ⟨[ γ a , γ b ]⟩ = i 2 ⟨ Φ | Φ ⟩ Sergey Bravyi, 2005, Quantum Inf. and Comp. Slide 18 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  21. ⏟ ∏ ) ⏟ ⏟ ⏟ ⏟⏟ ⏟ ⏟⏟⏟ ⏟ ⏟⏟ ⏟ Calculating the Norm and the Observables ∣ ψ ( G )⟩ = ⟨ Ω v ∣ ∏ ω ( x ) ∏ A ( x ) | Ω ⟩ U Φ ( x ) x x x ❀ Γ in ( G ) ❀ Γ M A Physical-Physical correlations i , j = ( A B Γ M − B T B Physical-Virtual correlations D ) C Virtual-Virtual correlations Norm ⟨ ψ ( G )∣ ψ ( G )⟩ = √ det ( 1 − Γ in ( G ) M D 2 Slide 19 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

  22. The whole framework Build the state ∣ Ψ ( G ′ )⟩ Draw new gauge field configuration ∣ G ′ ⟩ Calculate the acceptance probability Accept or decline [Measure observables] by computing ⟨ Ψ ( G ′ )∣ Ψ ( G ′ )⟩ the new configuration G ′ Slide 20 Combining Tensor Networks and Monte Carlo for Lattice Gauge Theories | 21st of February 2020 | PE, EZ, MCB, IC

Recommend


More recommend