On S-duality and 6d CFTs Seok Kim (Seoul National University) Autumn symposium on string theory, KIAS Sep 12, 2017 Talk based on: SK, June Nahmgoong, “Asymptotic M5-brane entropy from S-duality” 1702.04058. + more works in (slow) progress Some related works: Billo, Frau, Fucito, Lerda, Morales, “S-duality and the prepotential in N=2* theories (I): the ADE algebras” 1507.07709. Haghighat, Iqbal, Kozcaz, Lockhart, Vafa, “M-strings” 1305.6322. Di Pietro, Komargodski, “Cardy formulae for SUSY theories in d=4 and d=6” 1407.6061. Galakhov, Mironov, Morozov, “S-duality & modular transformation as a non-perturbative deformation of the ordinary pq-duality” 1311.7069. S-duality • Relation between two interacting QFTs w/ coupling constants inversed. • 3+1d gauge theories: electric-magnetic duality • Simple example: maximal super-Yang-Mills [Montonen, Olive] [Osborn] - Evidences: dyon spectrum [Sen], various partition functions [Vafa, Witten] … , etc. • Accounted for by embedding into bigger systems - D3’s on type IIB: consequence of S-duality of IIB string theory - A smaller embedding, in QFT: 6d N=(2,0) QFT (e.g. on M5’s) on 푇2 W-boson 휃 4휋푖 complex structure 휏 = + ~ 4d coupling constant 2휋 푔2 푌푀 monopole - Tied to subtle properties of 6d QFTs • 6d (2,0) theory has been a guidance to S-duality. Can it be true the other way? 3 The observable • N M5’s on S1 : an index for BPS states in the tensor branch. 푍[푅4 × 푇2] - bound states of wrapped self-dual strings & momenta • IIA picture: k D0’s bound to N D4’s & F1’s [Nekrasov] (2002) [Bruzzo, Fucito, Morales, Tanzini] (2002) [Nekrasov, Okounkov] (2003) [H.-C. Kim, SK, E. Koh, K. Lee, S. Lee] (2011) • Chemical potentials: - 휏: for left-moving momentum. (more on it on next slide) - 푣푖: for winding numbers. ~ tensor/Coulomb VEV - 푚: for 푆푈 2 퐿 ⊂ 푆푂 5 푅. ~ mass deformation parameter for MSYM - 휖1,2: 푆푂(4) spatial rotations (locked w/ 푆푈 2 푅 ⊂ 푆푂(5)). “Ω deformation” (partly IR regulator) 4 Using S-duality in 6d? • A partition function on R4 x T2 : (spatial S1) x (temporal S1) 푅 푖 - 휏 ~ 푖 푡 ~ : inverse-temperature (in KK unit). coupling const. in 4d limit 푅푠 푇 푅푠 2 - finite T : Is there S-duality 휏 → −1/휏 (or 푅푠 ↔ 푅푡) in the partition function? • Connects “low & high T”: useful tool to study “high T” or “decompactifying” regime - But if log 푍 ∼ 푁3 & log 푍 ∼ 푁2 at high/low T, hard to expect exact S-duality. - So, natural to expect certain ‘anomaly’ of S-duality. • Goal: Explore the 6d extension of 4d S-duality, and use it to study 6d physics. 푓(휏,푣,푚) • Will mostly focus on “prepotential” : 푍(휏, 푣, 푚, 휖1,2) ∼ exp[− ] at 휖1,2 → 0 . 휖1휖2 - Can be viewed either as Seiberg-Witten action, or free energy at “large volume” - Technically easier to study than the full partition function 5 4d limit: S-duality of 푁 = 2∗ prepotential • S-duality in SW theory: “magnetic prepotential = same function as electric one” - Expect S-duality after a suitable a decomposition independent of Coulomb VEV: ambiguous in SW theory • Tested in small 푀 expansion by summing over 푞 = 푒2휋푖휏 4푑 2 2 2푛 - 퐹푆 푑푢푎푙 휏, 푎, 푀 = 휋푖휏푎 + 푓 ≡ 휋푖휏푎 + σ푛=1 푀 푓푛(휏, 푎) “quantum” (=1-loop+instanton) part can write coefficients exactly in 푞, using “quasi-modular forms” [Billo, Frau, Fucito, Lerda, Morales] (2015) 6 6d extension • For simplicity, first consider 푀 ≪ 푎: classical S-duality - Good in 4d: Scalar eigenvalues classically live on 퐶 ∼ 푅2. 6d tensor VEV (noncompact) - 6d: Scalars live on 푅 × 푆1. Unnatural to rotate w/ complex 휏. 6d 2-form on 푇2 (compact) • Natural dimensionless variables (~ chemical potentials) 2 redefine to dimensionless 푓 by absorbing 푅푡 factor 푚 = 푀 푅푡 , v = 푎 푅푡 • 6d version of S-duality: using dimensionless variables & 푓 (correctly reduces to 4d ones) • Checked in same way, expanding in small 푚: curious ‘anomaly’ in S-duality nd After restoring 푅푡, the 2 term 2 4 3 ∝ 푅푡 푀 푁 − 푁 → 0 disappears in 4d limit - 퐹푎푛표푚(휏, 푚) meaningless in SW theory. But in free energy, neutral states’ contribution. [Nahmgoong, SK] (2017) 7 Alternative derivation • Expand 푍 w/ tensor VEVs, or fugacities 푒−푣푖 for string winding #’s simple roots - 푍 푛푖 (휏, 푚, 휖1,2): elliptic genus of self-dual strings [Haghighat, Iqbal, Kozcaz, Lockhart, Vafa] (2013) [HIKLV] (2013) • Modular anomaly of elliptic genus 푍 푛푖 (휏, 푚, 휖1,2): 푖푗 Ω : Cartan matrix for 퐴푁−1 Lie algebra 1 휌 = σ 훼: Weyl vector for 퐴 2 훼>0 푁−1 • This determines the S-duality anomaly of 푍(휏, 푣, 푚, 휖1,2), & 푓(휏, 푣, 푚). 8 Modular anomaly equation • 휏 dependence via quasi-modular forms: - 3 generators: causes modular anomaly of 푍 푛푖 (휏, 푚, 휖1,2) • modular anomaly equation: • some manipulations: complete-square RHS 푖 푖푗 퐼푖 푣 Ω 퐼푖 퐼푗 푍푆−푑푢푎푙 ≡ 푍መ exp + 퐸2(휏) : heat equation 휖1휖2 24 휖1휖2 푐 |퐺| 푁3−푁 휌2 = 2 → for 퐺 = 퐴 12 12 푁−1 • modular anomaly: “standard” anomaly + anomaly of “standard” anomaly [Nahmgoong, SK] (2017) 9 푍 and 푓 • S-dual of 푍푆 푑푢푎푙 (satisfying ‘heat equation’) given by convoluting w/ Gaussian heat 6 kernel (훿 ≡ ) 휋푖휏 • However, for many reasons, this expression is subtle: - Firstly, this expression is “wrong”. - 4d limit: 퐾(푣, 푣′) = 푍[푆3] of 푇[푆푈(푁)] QFT (S-duality domain wall) … motivated by AGT [Hosomichi, Lee, Park] [Drukker, Gaiotto, Gomis] [Teschner] - Why wrong? Used ‘wrong’ (~Weyl asymmetric) perturbative part [Galakhov, Mironov, Morozov] - Correction to Gaussian kernel is non-perturbative in 휖1,2 ≪ 1: prepotential OK [→ Left-over question: 6d uplift of exact S-duality kernel & defect interpretation?] - Also, often interested in regimes where other parameters get comparable to / larger than 푣, where we encounter phase transitions. Easier to handle in prepotential 10 Prepotential revisited • ‘S-dual’ and ‘anomalous’ parts: Recall that 푓 (휏,푣,푚) • Saddle point approximation of S-duality relation for 푍푆−푑푢푎푙 ∼ exp[ − ] : 휖1휖2 퐹 (휏,푚) • ‘anomalous parts’ in S-duality: remaining factors ∼ exp[− 푎푛표푚 ] 휖1휖2 - Both terms can be separately S-dualized easily, using and 11 Asymptotic free energy of 6d (2,0) theories • Strategy: use “dual weak-coupling setting”. anomalous part + 5d perturbative part 휏 → 푖0+ 푚 =fixed 퐷 푣퐷=fixed 퐷 2 - Want to keep 6d scalar small, Φ퐷 ≪ 푇 : works if we keep 푣퐷 = finite - 푚퐷 → 0 is maximal SUSY enhancement point, 푓 = 0: keep it finite to obstruct cancelation - S-duality makes calculus easy if 푓 on RHS can be approximated by 푓푝푒푟푡 (not always true) • Results in “easy regions”: from 퐹푎푛표푚 from “low T” dual perturbative part 3 4 - real 푚퐷: Take high T & then large N. − log 푍 ∝ 푁 푚퐷. (Physics not very clear to me…) 12 Index → “partition function”…? 2휋 • Non-perturbative phase transitions on RHS at Im 푚 → ± 퐷 푁 • Bose/Fermi cancelation maximally obstructed at 푚퐷 = 휋푖(표푑푑 푖푛푡푒푔푒푟) all states in the index counted w/ ‘+’ signs: would maximize the “index entropy” 2휋푖 4휋푖 • A sequence of phase transitions: 푚 = 0, − , − , … , −2휋푖 퐷 푁 푁 - 2 M5’s on 푆1 ~ 5d SU(2) 푁 = 1∗ (w/ instanton corrections) ~ 2 D5’s forming web w/ NS5: NS5 ~ (0,1) 푚 = 푚퐷휏 2 D5’s ~ (2,0) 5d SU(2) SYM strong coupling (0,1) compactified on 푆1 (0,1) (0,1) (2,1) (2,0) (2,2) 푚 = 0: perturbative singularity 푚 = −휋푖: 5d SCFT 푚 = −2휋푖 ∼ 0 퐷 퐷 퐷 13 Phase transitions ~ 5d SCFTs • More phase transition sequences given by 5d SCFTs: - 3 M5’s 5d SU(3) SYM at level 0 strong coupling (0,1) (0,1) (0,1) (0,1) (3,1) (3,0) (3,2) (3,3) 2휋푖 4휋푖 m=0: perturbative singularity 푚 = − : 5d SCFT 푚 = − : 5d SCFT 푚 = −2휋푖 ∼ 0 3 3 - Similar (but more nontrivial) sequences of 5d SCFTs for 푁 ≥ 4 M5’s It apparently appears challenging to penetrate through the non-perturbative phase transitions to arrive at 푚 = −휋푖, esp. at large N, but we are getting some progress… (and puzzles…) 14 A check • Despite all the possible phase transitions, one can show that the anomalous part 3 4 ∝ 푁 푚퐷 is robust and is a direct consequence of 6d ‘t Hooft anomaly • A check: [Di Pietro, Komargodski] [Banerjee, Bhattacharya, Bhattacharyya, Jain, Minwalla, Sharma] Some physics for “high T ~ small temporal 푆1” admits 5d background fields’ EFT - 5d reduction of 6d Omega background (푚, 휏 below mean 푚퐷, 휏퐷… sorry…!) → 0 - gauge field in 푈 1 ⊂ 푆푈 2 퐿 ⊂ 푆푂 5 푅: - 5d CS terms determined by 6d U(1) anomaly 푁3 - One can argue 푂(푚4) of 퐼푚(− log 푍) is given solely from this CS term. - Actual calculus completely agrees w/ S-duality-based studies (robust against phase transitions) 15 Generalization to 6d (1,0) SCFTs • A large part of the analysis for 퐴푁−1 (2,0) theory is extended to all (1,0) SCFTs. - S-duality anomaly ← modular anomaly of elliptic genus for wrapped strings - Modular anomalies ← 2d ‘t Hooft anomalies on 6d self-dual strings - 2d anomaly ← 6d anomaly: inflow [H.-C.Kim, SK, Park] [Shimizu, Tachikawa] (2016) - 6d anomaly in tensor branch [Ohmori, Shimizu, Tachikawa, Yonekura] [Intriligator] Green-Schwarz 4-form for classical anomaly in tensor branch (quadratic in field strengths {퐹} ) - Result (prepotential): GS 4-form w/ {퐹} replaced by corresponding masses 푟: number of tensor multiplets Ω푖푗: tensor multiplets’ kinetic term matrix (푖, 푗 = 1, … , 푟) • Can we use it to study high T behaviors of (1,0) theories? - If there are 5d “dual weakly-coupled regimes” at small 푆1.