Partial Wave Analysis for Dummies

Jim Napolitano
RPI and Cornell

Outline
• The Basic Quantum Mechanics
• Partial Wave Analysis in Textbooks
• An Example for CLEO-c: J/ψ → γππ, γKK
• Extensions, Complications, Technical Details
• Advanced Example: b1(1235) → ωπ
• Advanced Example: γp → pπ +π −

CLEO Lunch Seminar

September 9, 2003
1
 Disclaimers

Jim Napolitano is one of the “dummies”!

The term “Partial Wave Analysis” is poorly defined and

over used. Always keep your own physics in mind, and use
what is relevant for your problem.

Other terms used include “Amplitude Analysis” and

“Partial Wave Decomposition”. These terms have different
meanings for different people!

Important distinction for many analyses:

Energy-Independent versus Energy-Dependent.

You are getting my own slant on all of this!

2
 The Basic Quantum Mechanics

For a complete set of states |`i, write an arbitrary state

|αi = c`|`i
X

For orthonormal |`i, h`0|`i = δ``0 , so

h`|αi = c`

Therefore, we can write

|αi = I|αi
where
I = |`ih`|
X

This is called “inserting a complete set of states” and is

the essence of “Partial Wave Analysis”
3
 Partial Wave Analysis in Textbooks

Perhaps you learned that Partial Wave Analysis meant

dσ
= |f (θ)|2
dΩ
where
1 ∞
f (θ) = (2` + 1)P`(cos θ)eiδ` sin δ`
X

k `=0

This is a special case, namely elastic scattering of a

nonrelativistic spinless particle from a static, central
potential.

Nevertheless, it is obtained by inserting a complete set of

states in the scattering matrix.

4
 The Details: Sakurai “Modern QM” §7.6

k0, ~
f (~ k0|T |~
k) ∝ h~ ki
∝ k0|`0m0ih`0m0T |`mih`m|~
h~ ki
X X

`0m0 `m
∝ T`(E)Y`m(k̂0)Y`m?(k̂)
X

`m
∝ T`(E)(2` + 1)P`(cos θ)δm0
X

`m
∝ T`(E)(2` + 1)P`(cos θ)
X

The rest is a parameterization of T`(E) in terms of “phase

shifts” δ` that makes use of the behavior of the scattered
wave at distances far from the potential.

This is mostly irrelevant for today’s discussion!

5
 An Example for CLEO-c: J/ψ → γππ, γKK

Data: Mark III at SPEAR

Recall: J/ψ → γ X
The search for glueballs X → K +K −

J/ψ Radiative Decay: J ⁄ ψ → γG

γ
c

c Two Gluons ⇒ G

Need to know the

quantum numbers of the
glueball candidate G

Illustrates most crucial concepts of Partial Wave Analysis.

6
 Step One: Some References

To really learn this, you need to review basic formalisms of

quantum mechanics, especially the theory of angular
momentum and rotational symmetry. In addition to a
good quantum mechanics textbook I recommend the
following, all available electronically from SPIRES:

• L. P. Chen, “An amplitude analysis of the K K̄ system

(M ≤ 2 GeV/c2) produced in J/ψ radiative decay,”
SLAC-0386
• J. D. Richman, “An Experimenter’s Guide To The
Helicity Formalism,” CALT-68-1148
• B. S. Zou and D. V. Bugg, “Covariant tensor formalism
for partial wave analyses of ψ decay to mesons,” Eur.
Phys. J. A 16, 537 (2003)

7
 Step Two: Formal Beginning

For the reaction e+e− → J/ψ → γX → γmm̄, write

dN = |M(e+e− → γmm̄)|2dMX dΩX dΩ?m
where MX , ΩX , and Ω?m specify the kinematics.

M is a matrix element of some “decay operator” U :

M = hf |U |ii = hf |UAUB · · · |ii

We make the assumption that

M(e+e− → γmm̄) = M(J/ψ → γX) × M(X → mm̄)
X

X
(??Are we inserting a complete set |XihX|??)

Our goal is to learn about U . We do this by

parameterizing M, and fitting the parameters to our data.
These parameters are coefficients of “partial waves”.

8
 Step Three: The Helicity Basis

Consider the two-body decay α → 1 + 2 where particle α

has angular momentum quantum numbers J and M . (Of
course, we say α has “spin J ”.)

The helicity operator h ≡ S~·p p| (eigenvalues λ) is

~/|~
invariant under rotations, so it commutes with the total
angular momentum operator J~.

Therefore, we can form basis states |jmλ1λ2i that are

simultaneous eigenstates of operators J~2, Jz , h1, and h2.

We will be inserting this complete set of states.

The tricky part is to properly rotate the “axis of

quantization”. For this, we use Wigner D-functions.

9
 M = hf |U |ii = h1 + 2|U |αi = hθ, φ, λ1, λ2|U |J M i

Now insert |jmλ01λ02i where the |jmi quantization axis is

the same as for |J M i. Use the rotational symmetry of U .

M = hθ, φ, λ1, λ2|jmλ01λ02ihjmλ01λ02|U |J M i

X

jmλ01λ02
= hθ, φ, λ1, λ2|J M λ01λ02ihJ M λ01λ02|U |J M i
X

λ01λ02
(J )
= hθ, φ, λ1, λ2|J M λ01λ02iAλ0 λ0
X

λ01λ02 1 2

(J )
We have now parameterized the interaction U by Aλ0 λ0
1 2
for the decay of a particle of spin J into daughter particles
with helicities λ01 and λ02. The angular distribution of this
decay is governed by the factor(s) hθ, φ, λ1, λ2|J M λ01λ02i
which are independent of the dynamics. However,. . .
10
 v

2J + 1 J ?
u
u
0 0
u

hθ, φ, λ1, λ2|J M λ1λ2i = δλ1,λ0 δλ2,λ0 DM,λ1−λ2 (φ, θ, 0)

u
u
u
t
1 2 4π
See Richman for details. Recall that D functions are
matrix elements of the rotation operator R(α, β, γ) for
Euler angles α, β, γ, i.e.
j
Dm0m(α, β, γ) = hjm0|R(α, β, γ)|jmi
There is an arbitrary choice of phase because φ is
undefined at θ = 0. Our choice of γ = 0 defines the phase
convention. In any case, this leaves us with. . .

(J )
M = hθ, φ, λ1, λ2|J M λ01λ02iAλ0 λ0
X

λ01λ02 1 2
v

(J ) 2J + 1 J ?
u
u
u

= Aλ1λ2 DM,λ1−λ2 (φ, θ, 0)

u
u
u
t
4π

11
 Step Four: Apply Formalism to our Sequential Decay

M(J/ψ → γX)
For e+e− → J/ψ, have J = 1 and M = ±1
Since photons are massless, λγ = ±1
For spin JX , λX = 0, ±1, . . . , ±JX
Need CJ/ψ = Cγ CX so must have CX = +1

M(X → mm̄)
For pseudoscalar m, m̄, need λm = 0
Have J (mm̄) = ` and P = C = (−1)`
⇒ Only JX = 0, 2, . . . allowed

12
Putting it together:
v

2JX + 1 1? ?
u
λγ λγ JX
u

Dλψ ,λX −λγ (ΩX )DλX ,0(Ω?m)

u

MJX ,λX = aJX ,λX

u
u
u
t
4π
3



 2
λγ
dMX dΩX dΩ?m
 

MJX ,λX
 
dN =
X  X 
 
 
16π λψ ,λγ JX ,λX
 
 
 

λγ (1) (J )
where aJX ,λX ∝ AλX ,λγ A0,0X

−λγ λγ
Parity conservation: aJX ,−λX = aJX ,λX

So, the “joint” angular distribution is described by the

complex parameters
a0,0
a2,0 a2,1 a2,2
···
Our task is now to fit these parameters to the data.

13
 Predicted angular distributions

The results are far from obvious!

a0,0 6= 0

a2,0 6= 0

a2,1 6= 0

a2,2 6= 0

b+c+d

a2,0 = a2,1 = a2,2

Interferences are important!

14
Step Five: Fit to the Data Independently in Bins of ∆MX

Mark III Technique: Formulate the “accepted moments”

and fit these to the data by minimizing χ2. Then, extract
amplitudes aJ,λ by solving nonlinear algebraic equations.

Unbinned Extended Maximum Likelihood: Fit for aJ,λ

directly, including the Poisson fluctation probability in the
number of events. Includes the acceptance specifically for
each term in the expansion.

Important: The “acceptance” (geometric, plus analysis

cuts) is determined with a “phase space” Monte Carlo
simulation. This simulation contains no “physics”, but
populates each element of ∆ΩX , ∆Ω?m evenly.

15
The Result: Evidence for f0(1380), f0(1710), f2(1270), f2(1525)

Recall J/ψ → γX, where X → . . .

|a0,0|2 |a2,0|2 |a2,1|2 |a2,2|2
3000
(a) (b) (c)
(a)
2000 1000

1000
KS KS
0 0

Events/(0.1 GeV)
(d) (e) (f)
Events/(0.1 GeV)

(b)
2000
K +K −
1000

1000

0 0
(g) (h) (i)
(c)
1000
π +π − 1000
500

0 0
1.0 1.5 2.0 1.2 1.6 2.0 1.2 1.6 2.0 1.2 1.6 2.0
6-96 Mass (GeV) 7236A5 6-96 Mass (GeV) 7236A6

Solid lines: An Energy (i.e Mass) Dependent Fit

16
 Extensions, Complications, Technical Details

Lots of data and many amplitudes to fit, means you need

lots of disk space and lots of CPU time.

Mathematically ambiguous solutions. If not, maybe

statistically ambiguous solutions.

The sum over intermediate states X is really infinite. Did

you truncate the the set of basis states too early?

What’s the right way to handle multi-hadron final states?

Are you summing over non-orthogonal states? Is the

“double counting” a big problem for your physics goals?

17
 A General Use Program: PWA2000

Authors:
John Cummings (RPI) and Dennis Weygand (JLab)

Various features, including

• Extended maximum likelihood fitting
• Built-in options for quasi-two-body decays
• Acceptance incorporated into fit function
• Tools for display, comparison, combining amplitudes
• Object-oriented approach

Program available now. Paper in preparation.

18
 Advanced Example: b1(1235) → ωπ

Two principle decay modes: b1 → (ωπ)S and b1 → (ωπ)D

X t
/ 0
1
0
1 1
0
0
1

L /
Ex
p 1
0
0
1 p

M. Nozar et al., Phys. Lett. B 541, 35 (2002)

Result of partial wave analysis:

|D/S| = 0.269 ± 0.020 and φ(D − S) = 10.54◦ ± 2.4◦ ± 3.9◦

19
b1(1235) → ωπ: Separate quantum numbers

Dominated by b1(1235) and ρ3(1670)

20
 b1(1235) → ωπ: Adjusting D/S

Constrain magnitude or phase and vary the other

21
 Advanced Example: γp → pπ +π −

Goal: Baryon spectroscopy

π+
π
γ γ ∗ p
_ P ∆,Ν
J
     
















π M

















L
p p p π

π
γ
P
ρ,σ π
J M
L
p p

22
 γp → pπ +π −: Some Preliminary Results

J P = 3/2+ J P = 5/2+
+ +
6 4) 3 (M= 1 ) → ∆ ++ π - (l=1) 7) 5 (M= 1 ) → ∆ 0 π+ (l=1)
2 2 2 2
d σ ( µ b) / ( ≈ 40 MeV/c )

d σ ( µ b) / ( ≈ 40 MeV/c )
8
2

2
5 7

6
4
5
3
4

2 3

2
1
1

0 0
1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4
2 2
W GeV/c W GeV/c
+ +
15) 3 (M= 1 ) → p ρ (l=1 s=1) 23) 5 (M= 3 ) → ∆ 0 π+ (l=1)
2 2 2 2
d σ ( µ b) / ( ≈ 40 MeV/c )

d σ ( µ b) / ( ≈ 40 MeV/c )
2.5
2

2
5

2
4

3 1.5

2 1

1 0.5

0 0
1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4
2 2
W GeV/c W GeV/c

23
 Conclusions

Partial Wave Analysis (whatever it is) is a useful tool for

extracting physics, especially when the effects of
interference are important.

The technique is not a panacea. Always keep your physics

goals in mind, and use these to make decisions for your
particular analysis.

Multihadron final states are a particularly thorny problem.

24