mario giuliani
play

Mario Giuliani Christof Gattringer, Pascal Trek Karl Franzens - PowerPoint PPT Presentation

Functional Fit Approach (FFA) for Density of States method: SU(3) spin system and SU(3) gauge theory with static quarks Mario Giuliani Christof Gattringer, Pascal Trek Karl Franzens Universitt Graz Mario Giuliani (Universitt Graz)


  1. Functional Fit Approach (FFA) for Density of States method: SU(3) spin system and SU(3) gauge theory with static quarks Mario Giuliani Christof Gattringer, Pascal Törek Karl Franzens Universität Graz Mario Giuliani (Universität Graz) Southampton, 26th July 2015 1 / 22

  2. Sign Problem: a brief review Different approaches to solve the sign problem: – Reweighting – Expansion methods – Stochastic differential equations – Mapping to dual variables – Et cetera... Density of states approach: – Method used FFA Functional Fit Approach ( ARXIV: 1503.04947, 1607.07340 ) – See also LLR Linear Logarithmic Relaxation by K. Langfeld, B. Lucini and A. Rago ( ARXIV:1204.3243, 1509.08391 ) Mario Giuliani (Universität Graz) Southampton, 26th July 2015 2 / 22

  3. Density of States Method In QFT we want to compute for our theory: � �O� = 1 � D [ ψ ] e − S [ ψ ] D [ ψ ] O [ ψ ] e − S [ ψ ] Z = Z In the density of states approach we divide the action into two parts: S [ ψ ] = S ρ [ ψ ] + c X [ ψ ] * S ρ [ ψ ] and X [ ψ ] are real functionals of the fields ψ * S ρ [ ψ ] is the part of the action that we include in the weighted density ρ * Here c is purely imaginary: c = i ξ Mario Giuliani (Universität Graz) Southampton, 26th July 2015 3 / 22

  4. Density of States Method The weighted density is defined as: � D [ ψ ] e − S ρ [ ψ ] δ ( X [ ψ ] − x ) ρ ( x ) = Using ρ ( x ) we can write Z and �O� s: � x max � x max �O� = 1 dx ρ ( x ) e − i ξ x dx ρ ( x ) e − i ξ x O [ x ] Z = Z x min x min → ψ ′ such that we can write: Usually there is a symmetry ψ − � x max Z = 2 dx ρ ( x ) cos ( ξ x ) 0 ... and for the observables O : � x max �O� = 2 dx ρ ( x ) { cos ( ξ x ) O even ( x ) − i sin ( ξ x ) O odd ( x ) } Z 0 Where O even = O ( x )+ O ( − x ) and O odd = O ( x ) −O ( − x ) 2 2 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 4 / 22

  5. First example: SU(3)-spin model Arxiv:1607.07340 SU ( 3 ) spin model is a 3D effective theory for heavy dense QCD Relevant d.o.f. is the Polyakov loop P ( n ) ∈ SU ( 3 ) (static quark source at n ) The model has a real and positive dual representation ⇒ reference data We have an action: 3 � TrP ( n ) TrP ( n + ν ) † + c . c . � [ e µ TrP ( n )+ e − µ TrP ( n ) † ] � � � S [ P ] = − τ − κ n ν = 1 n The action depends only on the trace ⇒ simple parametrization  e i θ 1 ( n )  0 0 e i θ 2 ( n ) P ( n ) = 0 0   e − i ( θ 1 ( n )+ θ 2 ( n )) 0 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 5 / 22

  6. Definition of density of states We define the weighted density of states with S ρ [ P ] = Re [ S [ P ]] and Im [ S [ P ]] = 2 κ sinh ( µ ) X [ P ] : � D [ P ] e − S ρ [ P ] δ ( x − X [ P ]) ρ ( x ) = x ∈ [ − x max , x max ] Symmetry P ( n ) → P ( n ) ∗ implies ρ ( − x ) = ρ ( x ) This simplifies the partition function: x max x max � � Z = dx ρ ( x ) cos ( 2 κ sinh ( µ ) x ) = 2 dx ρ ( x ) cos ( 2 κ sinh ( µ ) x ) − x max 0 x max � = 2 � � � � � �O X dx ρ ( x ) O E ( x ) cos ( 2 κ sinh ( µ ) x ) + i O O ( x ) sin ( 2 κ sinh ( µ ) x ) Z 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 6 / 22

  7. Parametrization of the density ρ ( x ) Ansatz for the density: ρ ( x ) = e − L ( x ) , normalization ρ ( 0 ) = 1 ⇒ L ( 0 ) = 0 We divide the interval [ 0 , x max ] into N intervals n = 0 , 1 , . . . , N − 1. L ( x ) is continuous and linear on each of the intervals, with a slope k n : 5 k N − 1 4 k N − 2 3 k 6 k 5 k 4 2 k 3 k 1 k 2 1 k 0 0 ∆ N − 2 ∆ N − 1 ∆ 0 ∆ 1 ∆ 2 ∆ 3 ∆ 4 ∆ 5 ∆ 6 l max Mario Giuliani (Universität Graz) Southampton, 26th July 2015 7 / 22

  8. Determination of the slopes k n How do we find the slopes k n ? Restricted expectation values which depend on a parameter λ ∈ R : 1 � D [ P ] e − S ρ [ P ]+ λ X [ P ] O � � � � ��O�� n ( λ ) = X [ P ] θ n X [ P ] Z n ( λ ) � D [ P ] e − S ρ [ P ]+ λ X [ P ] θ n � � Z n ( λ ) = X [ P ] � 1 for x ∈ [ x n , x n + 1 ] � � θ n x = 0 otherwise Update with a restricted Monte Carlo Vary the parameter λ to fully explore the density Mario Giuliani (Universität Graz) Southampton, 26th July 2015 8 / 22

  9. Functional Fit Approach FFA Expressed in terms of the density: x max x n + 1 x n + 1 � � � dx ρ ( x ) e λ x = c dx ρ ( x ) e λ x θ n � � dx e ( − k n + λ ) x Z n ( λ ) = = x − x max x n x n = c e ( λ x − k n ) x n + 1 − e ( λ x − k n ) x n λ − k n For computing the slopes we use as observable X [ P ] : x n + 1 1 � dx ρ ( x ) e λ x x = ∂ � � �� X [ P ] �� n ( λ ) = Z n ( λ ) ∂λ ln Z n ( λ ) x n Mario Giuliani (Universität Graz) Southampton, 26th July 2015 9 / 22

  10. Functional Fit Approach FFA Explicit expression for restricted expectation values: n − 1 1 � � − 1 � � � �� X [ P ] �� n ( λ ) − ∆ j 2 = h ( λ − k n )∆ n ∆ n j = 0 1 − e − r − 1 1 r − 1 h ( r ) = 2 Strategy to find k n : Evaluate �� X [ P ] �� n ( λ ) for different values of λ 1 Fit these Monte Carlo data h (( λ − k n )∆ n ) 2 k n are obtained from simple one parameter fits 3 The quality of the fit provide a self-consistent check of our simulation Mario Giuliani (Universität Graz) Southampton, 26th July 2015 10 / 22

  11. Fit of slopes ⇒ density ρ ( l ) Example: 8 3 , κ = 0 . 005 , µ = 0 . 0 τ = 0 . 075 1 2 - 0 0.5 ) 0 j 20 ∆ ln( ρ (x)) n-1 ∑ j=0 0.4 40 )- -2000 λ 0.3 ( 60 n 〉 〉 F[P] 80 0.2 -4000 〈 100 τ =0.025 〈 ( 0.1 n 1 120 τ =0.050 ∆ -6000 0 140 τ =0.075 τ =0.100 − 160 0.1 -8000 τ =0.125 180 − 0.2 τ =0.150 − 0.3 -10000 − 0.4 -12000 − 0.5 − − 4 2 0 2 4 λ 0 200 400 600 800 1000 1200 1400 x ρ ( x ) = e − L ( x ) L ( x ) k n Mario Giuliani (Universität Graz) Southampton, 26th July 2015 11 / 22

  12. Observables 1 Particle number density n : x max n = 1 1 ∂ sinh ( µ ) ln Z = 1 2 ∂ � dx ρ ( x ) sin ( 2 κ sinh ( µ ) x ) x V 2 κ V Z 0 2 ... and the corresponding susceptibility χ n : χ n = 1 ∂ ∂ sinh ( µ ) n 2 κ � 2 x max x max � 2 = 1 � 2 � � � dx ρ ( x ) cos ( 2 κ sinh ( µ ) x ) x 2 + dx ρ ( x ) sin ( 2 κ sinh ( µ ) x V Z Z 0 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 12 / 22

  13. Comparison with dual approach With large statistic and small intervals we are able to explore results up to µ = 4: Particle number density n Lattice 8 3 , τ = 0 . 130 and κ = 0 . 005: 0.07 n Density of states Dual formulation 0.06 0.05 0.04 0.03 0.02 0.01 0.00 µ 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 We find a good agreement for chemical potential up to µ ≈ 4 . 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 13 / 22

  14. Comparison with dual approach Susceptibility χ n Lattice 8 3 , τ = 0 . 130 and κ = 0 . 005: 3.5 χ n 3.0 2.5 2.0 1.5 1.0 Density of states Dual formulation 0.5 0.0 µ 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 The density performs fine up to µ ≈ 4 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 14 / 22

  15. Comparison with dual approach We can also go to bigger lattice size with the same parameters: Particle number density n Lattice 12 3 , τ = 0 . 130 and κ = 0 . 005: 0.07 n Density of states Dual formulation 0.06 0.05 0.04 0.03 0.02 0.01 0.00 µ 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 We find a good agreement for chemical potential up to µ ≈ 4 . 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 15 / 22

  16. Comparison with dual approach Susceptibility χ n Lattice 12 3 , τ = 0 . 130 and κ = 0 . 005: 3.5 χ n 3.0 2.5 2.0 1.5 1.0 Density of states Dual formulation 0.5 0.0 µ 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 The density performs fine up to µ = 4 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 16 / 22

  17. We find smaller error bars for larger µ 2 π The oscillating factor is bigger, but we still have: ∆ n ≪ 2 κ sinh µ This can be explained looking at the different densities: ln( ρ (x)) 0 -50 µ =0.1 -100 µ =0.5 µ =1.0 µ =1.5 -150 µ =2.0 µ =2.5 µ =3.0 -200 µ =3.5 µ =4.0 -250 0 100 200 300 400 500 600 x The changed shape of the density above µ ≈ 2 . 25 weakens the piling up of the errors on the singles k n Mario Giuliani (Universität Graz) Southampton, 26th July 2015 17 / 22

  18. Second example: SU(3) static quarks Further step towards a real QCD system SU ( 3 ) static quarks is a 4D effective theory for heavy dense QCD A SU ( 3 ) gauge theory plus the static quarks represented by Polyakov loops We have the following action: S [ U ] = − β � � � n ) † � � � e µ N T � n ) + e − µ N t � TrU µν ( n ) − κ P ( � P ( � Re 3 n µ<ν n � n � Where the Polyakov loops are: N T − 1 n ) = 1 � P ( � 3 Tr U 4 ( � n , n 4 ) n 4 = 0 Mario Giuliani (Universität Graz) Southampton, 26th July 2015 18 / 22

Recommend


More recommend