Single (transverse) Spin Asymmetry & QCD Factorization

1. Single (transverse) Spin Asymmetry & QCD Factorization Xiangdong Ji University of Maryland — Workshop on SSA, BNL, June 1, 2005—

2. Outline • General Remarks • DIS/Drell Yan processes • pp  πX & friends • Summary

3. Single (transverse) Spin Asymmetry • SSA is a general phenomenon in physics, and it exists so long as there are • A single transverse spin • A mechanism for helicity flip • Initial and/or state interactions • “Ok, SSA is an interesting phenomenon, but what do you learn about QCD from it?” or “Why do we have to spend $ & time to measure it?” • We have some models that fit the data(who cares about models? we have QCD) • We learn something about the nucleon spin structure(what exactly do you learn? And why that is interesting? can you check it in lattice QCD? What is it missing if we don’t measure it?) • ….

4. Pertubative & Nonperturbative Mechanisms • In general, however, the physics mechanism for SSA in strong interactions can be either be perturbative & non-perturbative, • pp to pp at low energy: non-perturbative • What one would like to understand is the SSA in perturbative region=> we hope to learn something simple, maybe! There must be some hard momentum: Perturbative description of the cross section must be valid  • Factorization • A good description of spin-averaged cross sections

5. SSA & processes pp -> πX & friends DIS & Drell-Yan Hard scale Q2 PT QCD factorization In TMD’s Small PT~ΛQCD Non-perturbative QCD factorization In TMD’s ? Twist-3 effects QCD factorization In TMD’s Twist-3 effects Q2,s» PT»ΛQCD

6. k k’ X P SIDIS at low pT • Single-jet production • If the target is transversely polarized, the current jet with a transverse momentum kT has a SSA which allows a QCD factorization theorem even when kT is on the order ΛQCD • The SSA is of order 1 in the scaling limit, i.e. a twist-2 effect! q

7. Factorization for SIDIS with P┴ • Must consider generic Feynman diagrams with partons having transverse momentum, and gluon loops. • We have two observable scales, Q and P┴(soft). We consider leading order effects in P┴/Q. • The gluons can be hard, soft and collinear. Can one absorb these contributions into different factors in the cross sections. • X. Ji, F. Yuan, and J. P. Ma, PRD71:034005,2005

8. Example at one-loop • Vertex corrections q p′ k p Four possible regions of gluon momentum k: 1) k is collinear to p (parton dis) 2) k is collinear to p′ (fragmentation) 3) k is soft (wilson lines) 4) k is hard (pQCD correction)

9. A general reduced diagram • Leading contribution in p┴/Q.

10. Factorization • Factoring into parton distribution, fragmentation function, and soft factor:

11. TMD parton distributions • The unintegrated parton distributions is defined as where the “light-cone” gauge link is the usual parton distribution may be regarded as

12. Classification • The leading-twist ones are classified by Boer, Mulders, and Tangerman (1996,1998) • There are 8 of them, corresponding to the number of quark-quark scattering amplitudes without T-constraint q(x, k┴), qT(x, k┴) (sivers), ΔqL(x, k┴), ΔqT(x, k┴), δq(x, k┴),δLq(x, k┴),δTq(x, k┴), δT’q(x, k┴) • Similarly, one can define fragmentation functions

13. Sivers’ Function • A transverse-momentum-dependent parton distribution which builds in the physics of SSA! k P S The distribution of the parton transverse momentum is not symmetric in azimuth, it has a distribution in S ·(p × k). Since kT is small, the distribution comes from non-perturbative structure physics.

14. Physics of a Sivers Function • Hadron helicity flip • This can be accomplished through non-perturbative mechanics (chiral symmetric breaking) in hadron structure. • The quarks can be in both s and p waves in relativistic quark models (MIT bag). • FSI (phase) • The hadron structure has no FSI phase, therefore Sivers function vanish by time-reversal (Collins, 1993) • FSI can arise from the scattering of jet with background gluon field in the nucleon (collins, 2002) • The resulting gauge link is part of the parton dis.

15. Light-Cone Gauge Pitfalls • It seems that if one choose the light-cone gauge, the gauge link effect disappears. • FSI can be shifted ENTIRELY to the initial state (advanced boundary condition). Hence the FSI effects must come from the LC wave functions. • LCWF components are not real, they have nontrivial phase factors! • A complete gauge-independent TMD PD contains a additional FSI gauge link at ξ± = ∞ which does not vanish in the light-cone gauge • Conjectured by Ji & Yuan (2002) • Proved by Belitsky, Ji & Yuan (2002)

16. The extra FSI gauge link • Through an explicit calculation, one can show that the standard definition of TMD PD is modified by an additional gauge link • Gauge link arises from the eikonal phase accumulation of final state particle traveling in its trajectory. Although the dominant phase accumulation is in the light-cone direction, however, the phase accumulation happens also in the transverse direction.

17. SSA in A Simple Model • A proton consists of a scalar diquark and a quark, interacting through U(1) gauge boson (Brodsky, Hwang, and Schmidt, PLB, 2002). • The parton distribution asymmetry can be obtained from calculating Sivers’ function in light-cone gauge (Ji & Yuan)

18. Factorization theorem • For semi-inclusive DIS with small pT ~ • Hadron transverse-momentum is generated from • multiple sources. • The soft factor is universal matrix elements of Wilson • lines and spin-independent. • One-loop corrections to the hard-factor has been • calculated

19. Spin-Dependent processes • Ji, Ma, Yuan, PLB597, 299 (2004); PRD70:074021(2004)

20. Additional Structure Functions Sivers effect Collins effect

21. As PT becomes large… • If PT become hard (PT» ΛQCD), so long as Q» PT the above factorization formula still works! • On the other hand, in this region one can calculate the PT dependence perturbatively, • The pT dependence in the soft factor is easily to calculate.. • Expanding in parton momentum, one leads to the following

22. As PT becomes large… • The pT dependence in the TMDs can also be calculated through one-gluon exchange… • The soft matrix element is the twist-3 matrix elements TD

23. Putting all together • One should obtain a SSA calculated in Qiu-Sterman approach (H. Eguchi & Y. Koike) Therefore, SSA becomes twist-3, JI, Ma, Yuan (to be published)

24. Relation between TMDs & Twist-3? • The TMD approach for DIS/DY works for both small and perturbative, but moderate PT. • At small PT, it is a twist-two effect • At moderate PT, it is a twist-three effect. • The TMD approach is more general, but not necessary at moderate PT • The twist-3 approach works only at large PT, but is the most economical there!

25. SSA & processes pp -> πX & friends DIS & Drell-Yan Hard scale Q2 PT QCD factorization In TMD’s Small PT~ΛQCD Non-perturbative QCD factorization In TMD’s ? Twist-3 effects QCD factorization In TMD’s Twist-3 effects Q2,s» PT»ΛQCD

26. pp  πX & friends • PT must be large so that perturbative QCD works. • In this region, it is not need to use the TMD formalism. The twist-3 approach is sufficient. • Phases are generated perturbatively.

27. Perturbative Way to Generate Phase Coulomb gluon Some propagators in the tree diagrams go on-shell No loop is needed to generate the phase! Efremov & Teryaev: 1982 & 1984 Qiu & Sterman: 1991 & 1999

28. A possible exception • Is it possible that at moderate pT, the intrinsic transverse-momentum effect is so large that it cannot be expanded? • Soft function is still perturbative... • One could include the Sudakov form factors • I don’t know yet an argument to rule this out. However, I don’t know an example where this is true. • Difficulty: • No proof of factorization (may be it will work!) • The gauge links on the TMDs might be very complicated (both initial and final state interactions are present).

29. Conclusion • For SIDIS/DY with small and moderate transverse momentum, there is a QCD factorization theorems involving TMDs. • At moderate P┴, one recovers the twist-3 mechanism (ETQS). • For pp->πX at perturbative P┴, twist-3 mechanism seems to be complete. • One has yet to find a TMD type of factorization for pp->πX at perturbative P┴; and the TMD distributions might not be related to those in SIDIS/DY.