OLD AND NEW RESULTS FOR HADRONIC-LIGHT-BY-LIGHT

Transcription

1 1/64 OLD AND NEW RESULTS FOR HADRONIC-LIGHT-BY-LIGHT Lund University bijnens bijnens/chpt.html Hadronic Probes of Fundamental Symmetries, Amherst, 6-8 March 2014

2 1 2 QED HO hadronic 3 : new stuff is here 4 Theory Experiment 5 2/64

3 Literature 3/64 Final experimental paper: G. W. Bennett et al. [Muon G-2 Collaboration], Final Report of the Muon E821 Anomalous Magnetic Moment Measurement at BNL, Phys. Rev. D 73 (2006) [hep-ex/ ]. Review 1: F. J. M. Farley and Y. K. Semertzidis, The 47 years of muon g-2, Prog. Part. Nucl. Phys. 52 (2004). Review 2: J. P. Miller, E. de Rafael and B. L. Roberts, Muon (g-2): Experiment and theory, Rept. Prog. Phys. 70 (2007) 795 [hep-ph/ ]. Review 3: F. Jegerlehner and A. Nyffeler, The Muon g-2, Phys. Rept. 477 (2009) 1 [arxiv: [hep-ph]].

4 Literature 4/64 Lectures: M. Knecht, Lect. Notes Phys. 629 (2004) 37 [hep-ph/ ]. Final HLBL number: J. Bijnens and J. Prades, Mod. Phys. Lett. A 22 (2007) 767 [hep-ph/ ]. J. Prades, E. de Rafael and A. Vainshtein, Hadronic Light-by-Light Scattering Contribution to the Muon Anomalous Magnetic Moment, (Advanced series on directions in high energy physics. 20) [arxiv: [hep-ph]]. New stuff here: JB, Mehran Zahiri Abyaneh, Johan Relefors pion loop contribution arxiv: , arxiv: , arxiv: and to be published

5 Muon g 2: measurement 5/64

6 Muon g 2: measurement 6/64

7 Muon g 2: measurement 7/64 Phys.Rev. D73 (2006) [hep-ex/ ]

8 Muon g 2: measurement 8/64 Phys.Rev. D73 (2006) [hep-ex/ ] 2001: µ, others µ +

9 Muon g 2: overview 9/64 in terms of the anomaly a µ = (g 2)/2 Data dominated by BNL E821 (statistics)(systematic) a exp µ = (6)(5) a exp = (8)(3) µ aµ exp = (5.4)(3.3) Theory is off somewhat (electroweak)(lo had)(ho had) aµ SM = (0.2)(4.2)(2.6) a µ = aµ exp aµ SM = 28.7(6.3)(4.9) (PDG) E821 goes to Fermilab, expect factor of four in precision Note: g agrees to with theory Many BSM models CAN predict a value in this range (often a lot more or a lot less) Numbers taken from PDG2012, see references there

10 of Muon g 2 10/ a µ exp theory QED EW LO Had HO HVP difference Error on LO had all e + e based OK τ based 2 σ Error on Errors added quadratically 3.5 σ Generic SUSY: ( 100 GeV M SUSY ) 2tanβ M SUSY 66 GeV tanβ Difference: 4% of LO Had 270% of 1% of leptonic LbL

11 Muon g 2: QED 11/64 a QED µ = α 2π (27)( α π) (43) ( α π (80) ( α π) (20) ( α π) 5 + First three loops known analytically four-loops fully done numerically Five loops estimate Kinoshita, Laporta, Remiddi, Schwinger,... α fixed from the electron g 2: α = 1/ (51) a QED µ = (0.015) Light-by-light surprisingly large: ) 3 QED HO hadronic e = 20.95,µ = 0.37,τ = 0.002

12 Muon g 2: HO hadronic 12/64 Two main types of QED HO hadronic HO HVP HO HVP is like LO Had, can be derived from e + e hadrons. aho HVP µ = 9.84(0.06) is the real problem: best estimate now: aµ = 10.5(2.6) Note that the sum is very small: but not an indication of the error

13 : the main object to calculate 13/64 p 3 β p 2 α p 1 ν q ρ p p p 4 p 5 Muon line and photons: well known The blob: fill in with hadrons/qcd Trouble: low and high energy very mixed Double counting needs to be avoided: hadron exchanges versus quarks

14 A separation proposal: a start E. de Rafael, Hadronic to the muon g-2 and low-energy QCD, Phys. Lett. B322 (1994) [hep-ph/ ]. Use ChPT p counting and large N c p 4, order 1: pion-loop p 8, order N c : quark-loop and heavier meson exchanges p 6, order N c : pion exchange Does not fully solve the problem only short-distance part of quark-loop is really p 8 but it s a start 14/64

15 A separation proposal: a start E. de Rafael, Hadronic to the muon g-2 and low-energy QCD, Phys. Lett. B322 (1994) [hep-ph/ ]. Use ChPT p counting and large N c p 4, order 1: pion-loop p 8, order N c : quark-loop and heavier meson exchanges p 6, order N c : pion exchange Implemented by two groups in the 1990s: Hayakawa, Kinoshita, Sanda: meson models, pion loop using hidden local symmetry, quark-loop with VMD, calculation in Minkowski space JB, Pallante, Prades: Try using as much as possible a consistent model-approach, ENJL, calculation in Euclidean space 14/64

16 15/64 Π ρναβ (p 1,p 2,p 3 ) = p 1 Actually we really need δπρναβ (p 1,p 2,p 3 ) δp 3λ q p3 =0 p 2 p 3

17 16/64 Π ρναβ (p 1,p 2,p 3 ): In general 138 Lorentz structures (but only 28 contribute to g 2) Using q ρ Π ρναβ = p 1ν Π ρναβ = p 2α Π ρναβ = p 3β Π ρναβ = 0 43 gauge invariant structures Bose symmetry relates some of them All depend on p 2 1, p2 2 and q2, but before derivative and p 3 0 also p 2 3,p 1 p 2,p 1 p 3 Compare HVP: one function, one variable calculation from experiment difficult to see how In four photon measurement: lepton contribution

18 17/64 d 4 p 1 d 4 p 2 plus loops inside the hadronic part (2π) 4 (2π) 4 8 dimensional integral, three trivial, 5 remain: p 2 1,p2 2,p 1 p 2,p 1 p µ,p 2 p µ Rotate to Euclidean space: Easier separation of long and short-distance Artefacts (confinement) in models smeared out. More recent: can do two more using Gegenbauer techniques Knecht-Nyffeler, Jegerlehner-Nyffeler,JB Zahiri-Abyaneh Relefors P 2 1, P2 2 and Q2 remain study aµ X = dl P1 dl P2 aµ XLL = dl P1 dl P2 dl Q aµ XLLQ l P = ln(p/gev), to see where the are Study the dependence on the cut-off for the photons

19 17/64 d 4 p 1 d 4 p 2 plus loops inside the hadronic part (2π) 4 (2π) 4 8 dimensional integral, three trivial, 5 remain: p 2 1,p2 2,p 1 p 2,p 1 p µ,p 2 p µ Rotate to Euclidean space: Easier separation of long and short-distance Artefacts (confinement) in models smeared out. More recent: can do two more using Gegenbauer techniques Knecht-Nyffeler, Jegerlehner-Nyffeler,JB Zahiri-Abyaneh Relefors P 2 1, P2 2 and Q2 remain study aµ X = dl P1 dl P2 aµ XLL = dl P1 dl P2 dl Q aµ XLLQ l P = ln(p/gev), to see where the are Study the dependence on the cut-off for the photons

20 18/64 π 0 exchange π 0 π 0 = 1/(p 2 m 2 π ) The blobs need to be modelled, and in e.g. ENJL contain corrections also to the 1/(p 2 m 2 π) Pointlike has a logarithmic divergence Numbers π 0, but also η,η

21 π 0 exchange 19/64 BPP: Nonlocal quark model: a π0 µ = 5.9(0.9) a π0 µ = A. E. Dorokhov, W. Broniowski, Phys.Rev.D78 (2008) [ ] DSE model: a π0 µ = Goecke, Fischer and Williams, Phys.Rev.D83(2011)094006[ ] LMD+V: a π0 µ = ( ) M. Knecht, A. Nyffeler, Phys. Rev. D65(2002)073034, [hep-ph/ ] Formfactor inspired by AdS/QCD: a π0 µ = Cappiello, Cata and D Ambrosio, Phys.Rev.D83(2011) [ ] Chiral Quark Model: a π0 µ = D. Greynat and E. de Rafael, JHEP 1207 (2012) 020 [ ]. Constraint via magnetic susceptibility: a π0 µ = A. Nyffeler, Phys. Rev. D 79 (2009) [ ]. All in reasonable agreement

22 20/64 MV short-distance: π 0 exchange K. Melnikov, A. Vainshtein, Hadronic light-by-light scattering contribution to the muon anomalous magnetic moment revisited, Phys. Rev. D70 (2004) [hep-ph/ ] take P 2 1 P2 2 Q2 : Leading term in OPE of two vector currents is proportional to axial current Π ρναβ Pρ 0 T (J P1 2 Aν J Vα J Vβ ) 0 J A comes from + AVV triangle anomaly: extra info Implemented via setting one blob = 1 π 0 = π 0 a π0 µ =

23 21/64 π 0 exchange The pointlike vertex implements shortdistance part, not only + Are these part of the quark-loop? See also in Dorokhov,Broniowski, Phys.Rev. D78(2008)07301 BPP quarkloop + MV

24 π 0 exchange 22/64 LL a µ (VMD) 7e-10 6e-10 5e-10 4e-10 3e-10 2e-10 1e Which momentum regimes important studied: JB and J. Prades, Mod. Phys. Lett. A 22 (2007) 767 [hep-ph/ ] a µ = 1 P 2 dl 1 dl 2 a LL µ with l i = log(p i /GeV) P 1 LL a µ (KN) 7e-10 6e-10 5e-10 4e-10 3e-10 2e-10 1e Which momentum regions do what: volume under the plot a µ 1 P P 1

25 Pseudoscalar exchange 23/64 Point-like VMD: π 0 η and η give 5.58, 1.38, Models that include U(1) A breaking give similar ratios Pure large N c models use this ratio The MV argument should give some enhancement over the full VMD like models Total pseudo-scalar exchange is about a PS µ = AdS/QCD estimate (includes excited pseudo-scalars) a PS µ = D. K. Hong and D. Kim, Phys. Lett. B 680 (2009) 480 [ ]

26 24/64 A bare (sqed) give about The ππγ vertex is always done using VMD ππγ γ vertex two choices: Hidden local symmetry model: only one γ has VMD Full VMD Both are chirally symmetric The HLS model used has problems with π + -π 0 mass difference (due to not having an a 1 ) Final numbers quite different: 0.45 and 1.9 ( ) For BPP stopped at 1 GeV but within 10% of higher Λ

27 π loop: Bare vs VMD 25/64 2e-10 π loop 1.5e-10 1e-10 LLQ -a µ 5e VMD bare Q P 1 = P 2 plotted a LLQ µ for P 1 = P 2 a µ = dl P1 dl P2 dl Q a LLQ µ l Q = log(q/1 GeV)

28 π loop: VMD vs HLS 26/64 LLQ -a µ 1e-10 8e-11 6e-11 4e-11 2e e-11-4e-11 π loop VMD HLS a=2 Q P 1 = P 2 Usual HLS, a = 2

29 π loop: VMD vs HLS 27/64 1.2e-10 π loop 1e-10 8e-11 6e-11 LLQ -a 4e-11 µ 2e-11 0 VMD HLS a=1 Q P 1 = P 2 HLS with a = 1, satisfies more short-distance constraints

30 π loop 28/64 ππγ γ for q1 2 = q2 2 has a short-distance constraint from the OPE as well. HLS does not satisfy it full VMD does: so probably better estimate Ramsey-Musolf suggested to do pure ChPT for the π loop K. T. Engel, H. H. Patel and M. J. Ramsey-Musolf, Hadronic light-by-light scattering and the pion polarizability, Phys. Rev. D 86 (2012) [arxiv: [hep-ph]]. So far ChPT at p 4 done for four-point function in limit p 1,p 2,q m π (Euler-Heisenberg plus next order) Polarizability (L 9 +L 10 ) up to 10%, charge radius 30% Both HLS and VMD have charge radius effect but not polarizability

31 π loop 28/64 ππγ γ for q1 2 = q2 2 has a short-distance constraint from the OPE as well. HLS does not satisfy it full VMD does: so probably better estimate Ramsey-Musolf suggested to do pure ChPT for the π loop K. T. Engel, H. H. Patel and M. J. Ramsey-Musolf, Hadronic light-by-light scattering and the pion polarizability, Phys. Rev. D 86 (2012) [arxiv: [hep-ph]]. So far ChPT at p 4 done for four-point function in limit p 1,p 2,q m π (Euler-Heisenberg plus next order) Polarizability (L 9 +L 10 ) up to 10%, charge radius 30% Both HLS and VMD have charge radius effect but not polarizability

32 π loop 28/64 ππγ γ for q1 2 = q2 2 has a short-distance constraint from the OPE as well. HLS does not satisfy it full VMD does: so probably better estimate Ramsey-Musolf suggested to do pure ChPT for the π loop K. T. Engel, H. H. Patel and M. J. Ramsey-Musolf, Hadronic light-by-light scattering and the pion polarizability, Phys. Rev. D 86 (2012) [arxiv: [hep-ph]]. So far ChPT at p 4 done for four-point function in limit p 1,p 2,q m π (Euler-Heisenberg plus next order) Polarizability (L 9 +L 10 ) up to 10%, charge radius 30% Both HLS and VMD have charge radius effect but not polarizability

33 π loop: L 9,L 10 ChPT for muon g 2 at order p 6 is not powercounting finite so no prediction for a µ exists. But can be used to study the low momentum end of the integral over P 1,P 2,Q The four-photon amplitude is finite still at two-loop order (counterterms start at order p 8 ) Add L 9 and L 10 vertices to the bare pion loop JB-Zahiri-Abyaneh Program the Euler-Heisenberg plus NLO result of Ramsey-Musolf et al. into our programs for a µ Bare pion-loop and L 9, L 10 part in limit p 1,p 2,q m π agree with Euler-Heisenberg plus next order analytically 29/64

34 π loop: L 9,L 10 ChPT for muon g 2 at order p 6 is not powercounting finite so no prediction for a µ exists. But can be used to study the low momentum end of the integral over P 1,P 2,Q The four-photon amplitude is finite still at two-loop order (counterterms start at order p 8 ) Add L 9 and L 10 vertices to the bare pion loop JB-Zahiri-Abyaneh Program the Euler-Heisenberg plus NLO result of Ramsey-Musolf et al. into our programs for a µ Bare pion-loop and L 9, L 10 part in limit p 1,p 2,q m π agree with Euler-Heisenberg plus next order analytically 29/64

35 π loop: VMD vs charge radius 30/64 1e-10 8e-11 VMD L 9 =-L 10 π loop 6e-11 LLQ -a µ 4e-11 2e e-11-4e Q P 1 = P 2 low scale, charge radius effect well reproduced

36 31/64 π loop: VMD vs L 9 and L e-10 π loop LLQ -a µ 1e-10 8e-11 6e-11 4e-11 2e VMD L 10,L 9 Q 0.2 L 9 +L 10 0 gives an enhancement of 10-15% P 1 = P 2 To do it fully need to get a model: include a 1

37 32/64 Include a 1 L 9 +L 10 effect is from a 1 But to get gauge invariance correctly need a 1 a 1

38 33/64 Include a 1 Consistency problem: full a 1 -loop? Treat a 1 and ρ classical and π quantum: there must be a π that closes the loop Argument: integrate out ρ and a 1 classically, then do pion loops with the resulting Lagrangian To avoid problems: representation without a 1 -π mixing Check for curiosity what happens if we add a 1 -loop

39 34/64 Include a 1 Use antisymmetric vector representation for a 1 and ρ Fields A µν, V µν (nonets) Kinetic terms: 1 2 λ V λµ ν V νµ 1 2 V µνv µν λ A λµ ν A νµ 1 2 A µνa µν 1 2 Terms that give to the L r i : F V 2 f 2 +µνv µν + ig V 2 V µν u µ u ν + F A 2 f 2 µνa µν L 9 = F VG V, L 2MV 2 10 = F2 V + F2 4MV 2 A 4MA 2 Weinberg sum rules: (Chiral limit) F 2 V = F2 A +F2 π VMD for ππγ: F 2 V M2 V = F2 A M2 A F V G V = F 2 π

40 V µν only 35/64 Π ρναβ (p 1,p 2,p 3 ) is not finite (but was also not finite for HLS) But δπρναβ (p 1,p 2,p 3 ) also not finite δp 3λ (but was finite for HLS) p3 =0 Derivative one finite for G V = F V /2 Surprise: g 2 identical to HLS with a = F2 V F 2 π Yes I know, different representations are identical BUT they do differ in higher order terms and even in what is higher order Same comments as for HLS numerics

41 V µν only 35/64 Π ρναβ (p 1,p 2,p 3 ) is not finite (but was also not finite for HLS) But δπρναβ (p 1,p 2,p 3 ) also not finite δp 3λ (but was finite for HLS) p3 =0 Derivative one finite for G V = F V /2 Surprise: g 2 identical to HLS with a = F2 V F 2 π Yes I know, different representations are identical BUT they do differ in higher order terms and even in what is higher order Same comments as for HLS numerics

42 36/64 V µν and A µν Add a 1 Calculate a lot δπ ρναβ (p 1,p 2,p 3 ) δp 3λ p3 =0 finite for: G V = F V = 0 and F 2 A = 2F2 π If adding full a 1 -loop G V = F V = 0 and F 2 A = F2 π Clearly unphysical (but will show some numerics anyway)

43 36/64 V µν and A µν Add a 1 Calculate a lot δπ ρναβ (p 1,p 2,p 3 ) δp 3λ p3 =0 finite for: G V = F V = 0 and F 2 A = 2F2 π If adding full a 1 -loop G V = F V = 0 and F 2 A = F2 π Clearly unphysical (but will show some numerics anyway)

44 37/64 V µν and A µν Start by adding ρa 1 π vertices λ 1 [V µν,a µν ]χ +λ 2 [V µν,a να ]h µ ν +λ 3 i [ µ V µν,a να ]u α +λ 4 i [ α V µν,a αν ]u µ +λ 5 i [ α V µν,a µν ]u α +λ 6 i [V µν,a µν ]f α ν +λ 7 iv µν A µρ A ν ρ All lowest dimensional vertices of their respective type Not all independent, there are three relations Follow from the constraints on V µν and A µν (thanks to Stefan Leupold)

45 V µν and A µν : big disappointment 38/64 Work a whole lot δπ ρναβ (p 1,p 2,p 3 ) δp 3λ not obviously finite p3 =0 Work a lot more Prove that δπρναβ (p 1,p 2,p 3 ) finite, only same solutions as before δp 3λ p3 =0 Try the combination that show up in g 2 only Work a lot Again, only same solutions as before Small loophole left: after the integration for g 2 could be finite but many funny functions of m π,m µ,m V and M A show up.

46 39/64 π loop: add a 1 and F 2 A = 2F2 π π loop 2e e-10 1e-10 LLQ -a µ 5e bare F A 2 = -2 F 2 Q P 1 = P 2 Lowers at low energies, L 9 +L 10 < 0 here funny peak at a 1 mass

47 40/64 π loop: add a 1 and F 2 A = F2 π plus a 1 -loop 2e e-10 bare F A 2 = -F 2 π loop 1e-10 LLQ -a µ 5e Q P 1 = P 2 Lowers at low energies, L 9 +L 10 < 0 here funny peak at a 1 mass canceled Still unphysical case

48 41/64 a 1 -loop: cases with good L 9 and L 10 2e e-10 1e-10 π loop bare Weinberg no a 1 -loop 5e-11 0 LLQ -5e-11 -a µ -1e e-10-2e Q P 1 = P 2 Add F V, G V and F A Fix values by Weinberg sum rules and VMD in γ ππ no a 1 -loop

49 42/64 a 1 -loop: cases with good L 9 and L 10 2e e-10 π loop bare Weinberg with a 1 -loop 1e-10 5e-11 0 LLQ -5e-11 -a µ -1e e-10-2e Q P 1 = P 2 Add F V, G V and F A Fix values by Weinberg sum rules and VMD in γ ππ With a 1 -loop (is different plot!!)

50 43/64 a 1 -loop: cases with good L 9 and L 10 2e-10 π loop bare a 1 no a 1 -loop,vmd 1.5e-10 1e-10 LLQ -a µ 5e Q P 1 = P 2 Add a 1 with F 2 A = +F2 π Add the full VMD as done earlier for the bare pion loop

51 44/64 a 1 -loop: cases with good L 9 and L 10 2e-10 π loop bare a 1 with a 1 -loop,vmd 1.5e-10 1e-10 LLQ -a µ 5e Q P 1 = P 2 Add a 1 with F 2 A = +F2 π and a 1-loop Add the full VMD as done earlier for the bare pion loop

52 Integration 45/64 -a µ Λ 5e e-10 4e e-10 3e e-10 2e e-10 sqed sqed π 0 VMD HLS HLS a=1 ENJL 1e-10 5e Λ P 1,P 2,Q Λ

53 Integration 46/64 -a µ Λ 4e e-10 3e e-10 2e e-10 1e-10 a 1 F A 2 = -2F 2 a 1 F A 2 = -1 a1 -loop HLS HLS a=1 VMD a 1 VMD a 1 Weinberg 5e Λ P 1,P 2,Q Λ

54 47/64 Integration with a 1 Problem: get high energy behaviour good enough But all models with reasonable L 9 and L 10 fall way inside the error quoted earlier ( 1.9±1.3) Tentative conclusion: Use hadrons only below about 1 GeV: a π loop µ = ( 2.0±0.5) Note that Engel and Ramsey-Musolf, arxiv: is a bit more pessimistic quoting numbers from ( 1.1 to 7.1) 10 10

55 Pure quark loop 48/64 Cut-off a µ 10 7 a µ 10 9 a µ 10 9 Λ Electron Muon Constituent Quark (GeV) Loop Loop Loop (8) 2.41(3) 0.395(4) (10) 3.09(7) 0.705(9) (7) 3.76(9) 1.10(2) (6) 4.54(9) 1.81(5) (9) 4.60(11) 2.27(7) (6) 4.84(13) 2.58(7) Known Results (4) (16) M Q : 300 MeV now known fully analytically Us: 5+(3-1) integrals extra are Feynman parameters Slow convergence: electron: all at 500 MeV Muon: only half at 500 MeV, at 1 GeV still 20% missing 300 MeV quark: at 2 GeV still 25% missing

56 Pure quark loop: momentum area 49/64 quark loop m Q = 0.3 GeV P 2 = P 1 P 2 = P 1 /2 P 2 = P 1 /4 P 2 = P 1 /8 7e-10 6e-10 5e-10 4e-10 3e-10 2e-10 1e P 1 1 Q Most from P 1 P 2 Q, sizable large momentum part 10 10

57 ENJL quark-loop 50/64 Cut-off a µ a µ a µ a µ Λ sum GeV VMD ENJL masscut ENJL+masscut Very stable ENJL cuts off slower than pure VMD masscut: M Q = Λ to have short-distance and no problem with momentum regions Quite stable in region 1-4 GeV

58 ENJL: scalar 51/64 α p 2 p 3 β r qρ Π ρναβ p 1 ν = Π VVS ab (p 1,r)g S ( 1+gS Π S (r) ) Π SVV cd (p 2,p 3 )V abcdρναβ +permutations g S (1+g S Π S ) = g A(r 2 )(2M Q ) 2 2f 2 (r 2 ) V abcdρναβ : ENJL VMD legs 1 M 2 S (r2 ) r 2 In ENJL only scalar+quark-loop properly chiral invariant

59 ENJL: scalar/ql 52/64 Cut-off a µ a µ a µ Λ GeV VMD ENJL Exchange ENJL only scalar+quark-loop properly chiral invariant Note: ENJL+scalar (BPP) VMD (HKS) M S 620 MeV certainly an overestimate for real scalars If scalar is σ: related to pion loop part? quark-loop: a ql µ bare a ql µ =

60 Quark loop DSE 53/64 DSE model: a ql µ = 13.6(5.9) T. Goecke, C. S. Fischer and R. Williams, Phys. Rev. D 83 (2011) [arxiv: [hep-ph]] Not a full calculation (yet) but includes an estimate of some of the missing parts a lot larger than bare quark loop with constituent mass I am puzzled: this DSE model (Maris-Roberts) does reproduces a lot of low-energy phenomenology. I would have guessed that it would give numbers very similar to ENJL. Can one find something in between full DSE and ENJL that is easier to handle? Error found in calculation, still not finalized: preliminary a ql µ = 10.7(0.2) T. Goecke, C. S. Fischer and R. Williams, arxiv:

61 Other quark loop 54/64 de Rafael-Greynat Boughezal-Melnikov Masjuan-Vanderhaeghen ( ) ( ) ( ) Various interpretations: the full calculation or not All (even DSE) have in common that a low quark mass is used for a large part of the integration range

62 Axial-vector exchange exchange 55/64 Cut-off a µ from Λ Axial-Vector (GeV) Exchange O(N c ) (0.01) (0.01) (0.01) (0.02) (0.07) There is some pseudo-scalar exchange piece here as well, off-shell not quite clear what is what. a axial µ = MV: short distance enhancement + mixing (both enhance about the same) a axial µ =

63 : ENJL vc PdRV 56/64 BPP PdRV arxiv: quark-loop (2.1±0.3) pseudo-scalar (8.5±1.3) (11.4±1.3) axial-vector (0.25±0.1) (1.5±1.0) scalar ( 0.68±0.2) ( 0.7±0.7) πk-loop ( 1.9±1.3) ( 1.9±1.9) errors linearly quadratically sum (8.3±3.2) (10.5±2.6) 10 10

64 What can we do more? 57/64 The ENJL model can certainly be improved: Chiral nonlocal quark-model (like nonlocal ENJL): so far only done DSE: similar to everyone else, quark-loop very different, looking forward to final More resonances models should be tried, AdS/QCD is one approach, RχT (Valencia et al.) possible,... Note short-distance matching must be done in many channels, there are theorems JB,Gamiz,Lipartia,Prades that with only a few resonances this requires compromises : HLS smaller than double VMD (understood) models with ρ and a 1 : difficulties with infinities Theory Experiment

65 What can we do more? Constraints from experiment: J. Bijnens and F. Persson, hep-ph/hep-ph/ Studying three formfactors Pγ γ in P l + l l + l, e + e e + e P exact tree level and for g 2 (but beware sign): Conclusion: possible but VERY difficult Two γ off-shell not so important for our choice of form-factor All information on hadrons and off-shell photons is welcome: constrain the models More short-distance constraints: MV, Nyffeler integrate with all, not just Need a new overall evaluation with consistent approach. Lattice has done first steps Some tentative steps from dispersion theory Pauk-Vanderhaeghen Theory Experiment 58/64

66 What can we do more? Constraints from experiment: J. Bijnens and F. Persson, hep-ph/hep-ph/ Studying three formfactors Pγ γ in P l + l l + l, e + e e + e P exact tree level and for g 2 (but beware sign): Conclusion: possible but VERY difficult Two γ off-shell not so important for our choice of form-factor All information on hadrons and off-shell photons is welcome: constrain the models More short-distance constraints: MV, Nyffeler integrate with all, not just Need a new overall evaluation with consistent approach. Lattice has done first steps Some tentative steps from dispersion theory Pauk-Vanderhaeghen Theory Experiment 58/64

67 BNL magnet has moved to Fermilab 59/64 Goal ± Theory Experiment Credit: Brookhaven National Laboratory

68 BNL magnet has moved to Fermilab 60/64 Goal ± Theory Experiment Credit: Brookhaven National Laboratory

69 BNL magnet has moved to Fermilab 61/64 Credit: Theory Experiment Fermilab

70 BNL magnet has moved to Fermilab 62/64 Goal ± Theory Experiment Credit: Fermilab

71 JPARC with a very different method Ultracold muons at low energy (Credit: JPARC) 3 GeV proton beam ( 333 ua) Silicon Tracker Graphite target (20 mm) Surface muon beam (28 MeV/c, 4x108/s) Muonium Production (300 K ~ 25 mev) Resonant Laser Ionization of Muonium (~106 µ+/s) 66 cm diameter Super Precision Magnetic Field (3T, ~1ppm local precision) Theory Experiment Muon LINAC (300 MeV/c) 19 63/64

72 of Muon g 2 64/ a µ exp theory QED EW LO Had HO HVP difference Error on LO had all e + e based OK τ based 2 σ Error on Errors added quadratically 3.5 σ Generic SUSY: ( 100 GeV M SUSY ) 2tanβ M SUSY 66 GeV tanβ Difference: 4% of LO Had 270% of 1% of leptonic LbL