QCD spin physics - a theoretical overview

Daniël Boer
Free University Amsterdam
(at RBRC: 10/1998-06/2001)

QCD spin physics is about defining and extracting universal nonperturbative quantities
that capture aspects of the proton spin

polarized DIS → quark helicity distributions → polarized gluon distribution → orbital

angular momentum → GPDs

single spin asymmetries → TMDs → Wigner distributions → GPDs

Nick is right

RHIC Physics in the Context of the Standard Model, June 23, 2006 1
 QCD spin physics - a theoretical overview
Daniël Boer
Free University Amsterdam
(at RBRC: 10/1998-06/2001)

QCD spin physics is about defining and extracting universal nonperturbative quantities
that capture aspects of the proton spin

polarized DIS → quark helicity distributions → polarized gluon distribution → orbital

angular momentum → GPDs

single spin asymmetries → TMDs → Wigner distributions → GPDs

RHIC Physics in the Context of the Standard Model, June 23, 2006 2
 QCD spin physics
QCD spin physics = (to a large extent) the study of the spin of the proton expressed
in terms of quark and gluon d.o.f.

How does a dynamical system of quarks and gluons yield a spin-1/2 proton?

Quarks are confined inside hadrons, so one needs to scatter with high energy particles
off hadrons to probe quark properties

As probes one can use electrons, muons, neutrinos, hadrons, ...

RHIC Physics in the Context of the Standard Model, June 23, 2006 3
 Elastic and inelastic Scattering

e (k) e’ (k’) e e’

γ (q) γ
increasing Q2
−→















 



  


















   



   













 


 
 














 



  

 



  














 

  

 



















 





 















 
 

X

p p’ p

Q2 = −q 2 = −(k 0 − k)2

Elastic scattering (Q2 small) probes form factors

Deep inelastic scattering (Q2 large) probes structure functions and parton densities

RHIC Physics in the Context of the Standard Model, June 23, 2006 4
 QCD spin physics
Parton densities are hadronic matrix elements of some appropriate operator:

hP | operator(ψ, ψ, A) | P i

Goal: define and extract those operators that capture aspects of the proton spin

For example, ∆Σ, the sum of quark contributions to the proton spin:

∆Σ ≡ ∆u + ∆d + ∆s

∆q = (q+ − q−) + (q̄+ − q̄−)

The operator matrix element (OME) definition of ∆q is:

hP, S|ψ q γ µγ5ψq (0)|P, Si ∼ ∆q S µ

And it can be extracted from deep inelastic ~e p~ scattering (GRSV, AAC, ...)

RHIC Physics in the Context of the Standard Model, June 23, 2006 5
 Factorization in deep inelastic scattering
e

γ
σ~ = + ...




 


 


 



























 


 




































 

 

     


 


 


 


 




 

 




















 





  


  


      





 


 


 





  


  


    



 

 

 
p p

If Q2 is large enough, small QCD coupling allows for factorization of the cross section
q q

H
∗
R  2 2
 2


σ(γ p → X) ∝ xP xP = dx Tr H(x; Q )Φ(x; M ) + O 1/Q

Φ
P P

x = p+/P + is the light-cone momentum fraction of a quark (p) inside a hadron (P )

RHIC Physics in the Context of the Standard Model, June 23, 2006 6
 Gauge invariant definition of OME

In DIS one encounters functions of lightcone momentum fractions

Z
dλ iλx
Φij (x) = e hP, S| ψ j (0) L[0, λ] ψi(λn−) |P, Si (n−2 = 0)
2π
Φ(x): operator matrix element (OME) of a nonlocal lightcone operator

!
Z λ
+ A+ =0
L[0, λ] = P exp −ig dη A (ηn−) −→ 1
0

Such a path-ordered exponential (link) not just inserted, but can be derived
Efremov & Radyushkin, Theor. Math. Phys. 44 (1981) 774

Z
Mellin moments: dx xN Φ(x) −→ local operators

RHIC Physics in the Context of the Standard Model, June 23, 2006 7
 Parton distribution functions
In general, one has

1
Φ(x) = [f1(x)6 P + g1(x) λγ56 P + h1(x) γ56 S T 6 P ] + higher twist functions
2

Often one also writes f1q (x) = q(x), g1q (x) = ∆q(x), hq1(x) = δq(x)
Z Z
dx g1q (x) dxTr Φ(x)γ γ5 ∼ hP, S | ψ q γ +γ5ψq (0) | P, Si
+


∆q = ∼

= density of righthanded quarks - lefthanded

Very nice!
We have a definition of the quark helicity distribution ∆q and a method of extracting it
experimentally

However...

RHIC Physics in the Context of the Standard Model, June 23, 2006 8
 ∆Σ

∆Σ as extracted from experiment turns out to be much smaller than expected

European Muon Collaboration [1988] (hQ2i = 10.75 GeV2): ∆Σ = 0.00 ± 0.24

Spin Muon Collaboration [1998] (Q2 = 5 GeV2): ∆Σ = 0.13 ± 0.17

It means that the quarks and antiquarks contribute almost nothing to the proton spin!

This was dubbed the “spin crisis” or “spin puzzle”

RHIC Physics in the Context of the Standard Model, June 23, 2006 9
 Spin sum rule
In general, one expects the following “spin sum rule” to hold
1 1
proton spin = = ∆Σ + ∆g + Lz
2 2

Z Z
∆g = dx ∆g(x) = dx [g+ − g−]

The unintegrated polarized gluon distribution is defined as

i +α +
∆g(x) = F.T. hP, S L | F (0) L[0, λ] F̃ α (λn−) | P, SLi
4πx(P +)2
R
This means that dx ∆g(x) is intrinsically nonlocal along the lightcone
It becomes local (in A+ = 0 gauge) only after a tricky partial integration
Nonlocal operators are difficult to model or to calculate using the lattice

RHIC Physics in the Context of the Standard Model, June 23, 2006 10
 ∆g and Lz
At present ∆g(x) is under active experimental investigation:
√
• polarized p p collisions at RHIC (2005 data at s = 200 GeV indicates ∆g(x)  g(x))
• polarized DIS at CERN (COMPASS)

Measuring ∆g and using the spin sum rule, yields information about the size of Lz

Natural question: is Lz = Lqz + Lgz ?

Appears not to be the case, unless one chooses a specific gauge or a specific frame

Bashinsky, Jaffe, NPB 536 (1998) 303: gauge invariant, frame dependent decomposition

1 1
= ∆Σ + ∆g + Lq + Lg
2 2

No known way of accessing these Lq , Lg experimentally

RHIC Physics in the Context of the Standard Model, June 23, 2006 11
 Orbital angular momentum of quarks
Another gauge invariant decomposition has been considered:

1 1
= ∆Σ + Lq + Jg
2 2
" #
Z ~
†Σ
J~q = ~
d x ψ ψ + ψ †~x × (−iD)ψ
3
2
Z h i
J~g = 3 ~ × B)
d x ~x × (E ~

hP, S|J~q,g |P, Si ∝ Jq,g (Q2)S

~

Leads to a similar question: Jg = ∆g + Lg ?

To get a handle on orbital angular momentum of quarks Lq (x), Xiangdong Ji proposed

(PRL 78 (1997) 610) to use Deeply Virtual Compton Scattering: γ ∗ + p → γ + p0

RHIC Physics in the Context of the Standard Model, June 23, 2006 12
 DVCS
γ∗ γ

This involves off-forward, lightcone OMEs

Generalized Parton Distributions
GPD
P P’

Z
dλ iλx 0
e hP | ψ(−λ/2) γ +L[−λ/2, λ/2] ψ(λ/2) |P i =
2π
0 + iσ +ν ∆ν
0
Hq (x, ξ, t) ū(P )γ u(P ) + Eq (x, ξ, t) ū(P ) u(P )
2MN
+
with ∆ = P 0 − P , ξ = −∆+/(P 0 + P +) and t = ∆2

Forward limit: Hq (x, 0, 0) = q(x), Eq (x, 0, 0) ≡ Eq (x)

Eq (x) is not accessible in DIS

RHIC Physics in the Context of the Standard Model, June 23, 2006 13
 Generalized parton distributions

Orbital angular momentum of quarks defined as (Hoodbhoy, Ji, Lu, PRD 59 (1999) 014013):

1
Lq (x) = [xq(x) + xEq (x) − ∆q(x)]
2

Going from GPDs to Lq is very hard (extrapolations)

Hence, OAM is not straightforward to define and measuring it is even less trivial

But GPDs have become of interest in their own right, beyond OAM

GPDs yield a more complete picture of momentum and spatial distributions of partons
Burkardt, PRD 62 (2000) 071503; hep-ph/0207047
Belitsky & Müller, hep-ph/0206306
Diehl, EPJC 25 (2002) 223

RHIC Physics in the Context of the Standard Model, June 23, 2006 14
 Transverse polarization
For a transversely polarized proton a tensor charge can be defined:

hP, S|ψ q σ µν γ5ψq (0)|P, Si ∼ δq [P µS ν − P ν S µ]

Z
δq = dx hq1(x) h1(x) is called transversity

Difficult, but not impossible, to measure; many proposals have been put forward

Not measurable in elastic scattering or in inclusive deep inelastic scattering

However, h1(x) can be probed by using other hadrons: e.g. in e p↑ → e0 π X or p↑ p↑

h1 contains completely new information on the proton spin structure

Realistic extractions can come from BELLE+RHIC data (using interference

fragmentation functions) or from future GSI/FAIR p↑ p̄↑ data

RHIC Physics in the Context of the Standard Model, June 23, 2006 15
 What is known about the size of δq?
Thus far δq has only been studied in models and using lattice QCD
First (quenched) lattice QCD determination yielded (at µ2 = 2 GeV2):

δu = +0.839(60), δd = −0.321(55), δs = −0.046(34)

S. Aoki et al., PRD 56 (1997) 433

See also: Orginos, Blum, Ohta, hep-lat/0505023

Most recent lattice determination (with dynamical quarks) obtained (at µ2 = 4 GeV2):

δu = 0.857 ± 0.013, δd = −0.212 ± 0.005

QCDSF and UKQCD Collab., Göckeler et al., PLB 627 (2005) 113

Most models find results roughly in the range:

δu = +1.0 ± 0.2, δd = −0.2 ± 0.2

RHIC Physics in the Context of the Standard Model, June 23, 2006 16
 Left-right asymmetries
Transverse polarization has another trick up its sleeve...

Large single spin asymmetries in p p↑ → π X have been observed

E704 Collab. (’91); AGS (’99); STAR (’02); BRAHMS (’05); ...

A left-right asymmetry
Pion distribution is asymmetric
depending on transverse spin
direction and on pion charge

What is the explanation at the quark-gluon level?

Clearly a spin-orbit coupling, but how to describe such effects?

RHIC Physics in the Context of the Standard Model, June 23, 2006 17
 Transverse momentum of quarks
OAM requires transverse momentum of the quarks inside a hadron

Natural, but highly nontrivial, extension of Φ(x) → Φ(x, kT )

Soper ’77; Ralston & Soper ’79; many others

Φ(x, kT ) can be probed in experiments, for instance in semi-inclusive DIS

Ph ⊥ ^h φ
SIDIS k’ Ph
e p → e0 π X
q ^z
k

lepton scattering plane

A multiscale process: M, |P⊥π | and Q with |P⊥π |2  Q2

RHIC Physics in the Context of the Standard Model, June 23, 2006 18
 Spin dependent TMDs
TMD = transverse momentum dependent parton distribution function

1 2 P ·(kT × ST ) ⊥
Φ(x, kT ) = f1(x, kT ) 6 P + f1T (x, kT2 ) 6 P + ...
2 2M

⊥
Upon integration over transverse momentum the Sivers function f1T drops out

q kT q kT
⊥
sT
f 1T
= P −
sT

Sivers, PRD 41 (1990) 83

The Sivers effect leads to various single spin asymmetries, which are being tested at
several labs (BNL, CERN, DESY, Jefferson Lab)

RHIC Physics in the Context of the Standard Model, June 23, 2006 19
 Measuring the Sivers effect: two examples
One can probe the kT -dependence of the Sivers function directly in “jet SIDIS”

dσ(e p↑ → e0 jet X) e e kT ⊥ 2 jet

∝ |ST | sin(φjet − φS ) f (x,kT ), kT2 = |P⊥ |2
dφejet dkT2 M 1T

D.B. & Mulders, PRD 57 (1998) 5780

Upon integration over the transverse momentum of the jet: no SSA remains
Christ & Lee, PR 143 (1966) 1310

Asymmetric jet or hadron correlations in p↑ p → h1 h2 X 

  
D.B. & Vogelsang, PRD 69 (2004) 094025
Bacchetta et al., PRD 72 (2005) 034030

RHIC Physics in the Context of the Standard Model, June 23, 2006 20
 Link structure of TMDs
At first sight a P ·(kT × ST ) correlation appears to violate time reversal invariance
Collins, NPB 396 (1993) 161

But a model calculation by Brodsky, Hwang, Schmidt (PLB 530 (2002) 99) implied otherwise

Φ(x, kT ) is a matrix element of operators that are nonlocal off the lightcone

Φ(x, kT ) = F.T. hP | ψ(0) L[0, ξ] ψ(ξ) |P iξ=(ξ−,0+,ξ
T)

In SIDIS one has a future pointing Wilson line, whereas in Drell-Yan a past pointing one

ξT ξT

− −
ξ ξ

In lightcone gauge A+ = 0 the piece at infinity is crucial

Belitsky, Ji & Yuan, NPB 656 (2003) 165

RHIC Physics in the Context of the Standard Model, June 23, 2006 21
 Process dependence of TMDs

As a consequence there is a calculable process dependence (Collins, PLB 536 (2002) 43):

⊥ ⊥
(f1T )DIS = −(f1T )DY

The color flow of a process is crucial for the link structure

The more hadrons are observed, the more complicated the link structure
Bomhof, Mulders & Pijlman, PLB 596 (2004) 277; hep-ph/0601171

This is a new feature and factorization theorem proofs need to be revisited

Factorization theorems for processes involving a small transverse momentum

Collins & Soper, NPB 193 (1981) 381
Ji, Ma & Yuan, PRD 71 (2005) 034005; PLB 597 (2004) 299

RHIC Physics in the Context of the Standard Model, June 23, 2006 22
 Other consequences of the link structure

The link structure leads to intrinsically nonlocal lightcone OME

Taking A+ = 0 or taking Mellin moments does not yield local OMEs
For example,

kT2
Z +
Z
⊥(1) ⊥ A =0
f1T (x) ≡ d2kT f
2 1T
(x, k 2
T ) ∝ h ψ(0) dη − F +α(η −) γ + ψ(ξ −) i
2M

Hard to interpret or to evaluate using models or the lattice

Theoretically the energy scale (Q2) dependence poses questions still

Very involved; not leading twist operators in the OPE sense
Henneman, D.B. & Mulders, NPB 620 (2002) 331

RHIC Physics in the Context of the Standard Model, June 23, 2006 23
 Wigner distributions
⊥(1)
A relation between f1T and the GPD E has been put forward
Burkardt, hep-ph/0302144; Burkardt & Hwang, hep-ph/0309072

Z
⊥(1) ∂
f1T (x) ∝ ij
T ST i
2
d b⊥I(b⊥) E(x, b⊥)
∂bj⊥

The factor I(b⊥) not analytically calculable, has to be modeled, limits the use

GPDs and TMDs viewed as reductions of quantum phase-space (Wigner) distributions

d3q −i~q·~r
Z
1
WΓ(~r, k) ≡ 3
e h~q/2| ŴΓ(k) | − ~q/2i
2MN (2π)
Z
ŴΓ(k) = d4η eiη·k Ψ(−η/2)ΓΨ(η/2)

Ji, PRL 91 (2003) 062001; Belitsky, Ji & Yuan, PRD 69 (2004) 074014

RHIC Physics in the Context of the Standard Model, June 23, 2006 24
 Reductions of Wigner distributions
Fourier transforms of GPD’s are obtained as follows:
Z 2 −
Z 
d kT dk
W + (~r, k)
(2π)2 2π γ
+i
 
iσ q i
∝ F.T. Hq (x, ξ, t) ū(~q/2)γ +u(−~q/2) + Eq (x, ξ, t) ū(~q/2) u(−~q/2)
2M

TMDs can also be seen as reductions of Wigner distributions

3 −
Z Z 
d r dk
q(x, kT ) = 3
Wγ + (~r, k)
(2π) 2π

The six-dimensional phase space quantity: just a definition or measurable?

dk −
Z
fγ + (~r, x, kT ) ≡ W + (~r, k)
2π γ

Note: kT and rT are not each other’s Fourier conjugates

RHIC Physics in the Context of the Standard Model, June 23, 2006 25
 Λ polarization from unpolarized scattering
Large asymmetries have been observed in p + p → Λ↑ + X
G. Bunce et al., PRL 36 (1976) 1113; and by many others afterwards

PΛ > 0
^y

x^
Λ
^z
p p

The observed transverse polarization is negative (so opposite to the blue arrows)

The fragmentation analogue of the Sivers effect can describe such data (pT > 1 GeV)
Anselmino, D.B., D’Alesio & Murgia, PRD 63 (2001) 054029

RHIC Physics in the Context of the Standard Model, June 23, 2006 26
 * How to measure Λ polarization *
“Possible detection of parity nonconservation in hyperon decay”
T.D. Lee et al., PR 106 (1957) 1367
“Possible Determination of the Spin of Λ0 from its large Decay Angular Asymmetry”
T.D. Lee & C.N. Yang, PR 109 (1958) 1755

Λ polarization is “self-analyzing” due to parity violation

θ p
Proton distribution in Λ rest frame:

PΛ
dN (p) N0
= (1 + αΛPΛ cos θ)
d cos θ 2
Λ
PΛ = Λ polarization
αΛ = 0.64
π−

RHIC Physics in the Context of the Standard Model, June 23, 2006 27
 * Theoretical considerations *
√
Perturbative QCD conserves helicity, which leads to PΛ ∼ αsmq / ŝ (= small)
Kane, Pumplin & Repko, PRL 41 (1978) 1689

Many QCD-inspired models have been proposed, mostly based on

recombination of a fast ud diquark from the proton and a slower s quark from the sea
Spin-orbit coupling creates the polarization

The DeGrand-Miettinen model

PRD 23 (1981) 1227 & 24 (1981) 2419

RHIC Physics in the Context of the Standard Model, June 23, 2006 28
 * Collinear factorization *
√
For large s and pT factorization should apply; the process p + p → Λ + X should be
described in terms of parton densities and a fragmentation function

P2
σ ∼ q(x1) ⊗ g(x2) ⊗ DΛ/q (z)

q(x1) = quark density k Ph Ph k

g(x2) = gluon density

DΛ/q (z) = Λ fragmentation function
p p

PΛ ∼ q(x1) ⊗ g(x2) ⊗ ?
P1

Problem: no fragmentation function exists for q → Λ↑X

(due to symmetry reasons)

RHIC Physics in the Context of the Standard Model, June 23, 2006 29
 Sivers fragmentation function
ST
Λ Λ
⊥ kT kT
D 1T = -

Mulders & Tangerman, NPB 461 (1996) 197

• A nonperturbative kT × ST dependence in the fragmentation process

• Allowed by the symmetries (parity and time reversal)

Λ polarization arises in the fragmentation of an unpolarized quark

⊥
There are strong arguments to believe that D1T is universal (process independent)
Collins & Metz, PRL 93 (2004) 252001
⊥
Fits of D1T to data seem reasonable, predictions for SIDIS not tested yet
Anselmino, D.B., D’Alesio & Murgia, PRD 63 (2001) 054029; PRD 65 (2002) 114014

RHIC Physics in the Context of the Standard Model, June 23, 2006 30
 Saturation effects in p + A → Λ + X

Asymmetries can also be used to indicate changes in underlying physics

500

At high pT , leading twist pQCD predicts:

400
Qs/Λ=10 dσ(q A → q X) 1
Qs/Λ=5 ∼
dp2T p4T
pt dσ/dpt (qA−>qX)

300

For pT <∼ Qs saturation effects modify the

2

200
cross section
Figure: cross section (times p4T ) in √the
4

100
McLerran-Venugopalan model (large A & s)

⊥
Since D1T is kT -odd, it essentially probes the
0
0 10 20 30 40 50
derivative of the partonic cross section
pt/Λ

RHIC Physics in the Context of the Standard Model, June 23, 2006 31
 Λ polarization in p + A → Λ↑ + X
↑
√
p A → Λ X: at large A & s the asymmetry is sensitive to gluon saturation (Qs)

0 0

−0.02 −0.02

−0.04 −0.04
PΛ

PΛ
−0.06 −0.06

−0.08 −0.08
ξ = 0.1
Qs = 2 GeV ξ = 0.2
−0.1 Qs = 3 GeV −0.1 ξ = 0.3
ξ = 0.5

−0.12 −0.12
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

lt (GeV) lt (GeV)

D.B. & Dumitru, PLB 556 (2003) 33 [going beyond classical saturation is in progress]

RHIC Physics in the Context of the Standard Model, June 23, 2006 32
 QGP influence on A A → Λ↑ X

If a quark-gluon plasma is formed, then less recombination, so smaller asymmetry

Ayala et al. (PRC 65 (2002) 024902): polarization dependence on centrality

Hence, a full understanding of SSA may turn them into useful tools

RHIC Physics in the Context of the Standard Model, June 23, 2006 33
 Chiral-odd TMDs

⊥ ⊥
Besides f1T and D1T , there are two other, equally interesting kT -odd functions

One is called the Collins fragmentation function:

π kT π kT
sT
⊥
H1 = -
sT

Collins, NPB 396 (1993) 161

It can be extracted from e+ e− → π + π − X: hcos(2φ)i ∝ (H1⊥)2

D.B., Jakob & Mulders, NPB 504 (1997) 345

Matthias Grosse Perdekamp (around 2000): use off-resonance BELLE data

RHIC Physics in the Context of the Standard Model, June 23, 2006 34
 Collins asymmetry from BELLE

K. Abe et al., BELLE Collab., hep-ex/0507063, to appear in PRL

RHIC Physics in the Context of the Standard Model, June 23, 2006 35
 Anomalous asymmetry in Drell-Yan
sT
q q
⊥
kT kT
h = P − P
1

D.B. & Mulders, PRD 57 (1998) 5780

NA10 Collab. (’86/’88) & E615 Collab. (’89) measured the angular distribution of lepton pairs

1 dσ  2 2 ν 2

∝ 1 + λ cos θ + µ sin θ cos φ + sin θ cos 2φ
σ dΩ 2

Perturbative QCD relation (O(αs)): 1 − λ − 2ν = 0 Lam-Tung relation

Data shows a large deviation from the Lam-Tung relation

Nonzero h⊥
1 offers an explanation of this anomalous Drell-Yan data
D.B., PRD 60 (1999) 014012

RHIC Physics in the Context of the Standard Model, June 23, 2006 36
 * Violation of Lam-Tung relation in Drell-Yan *

Data from NA10 Collab. (’86/’88) & E615 Collab. (’89)

Data for π −N → µ+µ−X, with N = D, W

√
s ≈ 20 ± 3 GeV
lepton pair invariant mass Q ∼ 4 − 12 GeV

NA10: (1 − λ − 2ν) ≈ −0.6 at PT ∼ 2 − 3 GeV

E615: see figure

Large deviation from Lam-Tung relation

Order of magnitude larger & opposite sign

w.r.t. O(αs2) pQCD result

RHIC Physics in the Context of the Standard Model, June 23, 2006 37
 Very general conclusions

QCD spin physics is about defining and extracting the appropriate universal
nonperturbative quantities
Most of them concern operators nonlocal on and off the lightcone

• For longitudinal proton polarization orbital angular momentum needs to be understood

Generated the field of GPDs

• For transverse proton polarization spin-orbit coupling effects need to be understood

Generated the field of TMDs

Relations among the various quantities need to be explored further

RHIC Physics in the Context of the Standard Model, June 23, 2006 38
 Enjoy the rest of the workshop and the World Cup!

RHIC Physics in the Context of the Standard Model, June 23, 2006 39