Experiment IX: Angular Correlation of Gamma Rays

Curran D. Muhlberger
University of Maryland, College Park
(Dated: May 9, 2008)
Using a pair of scintillation detectors and a coincidence circuit, we measured the angular correlation functions
for g-rays emitted from sodium and cobalt nuclei. Theoretical arguments predict that successive g-rays emitted
from the same spinning nucleus should be distributed anisotropically, and this is indeed what we obseve. Photons
emitted via pair annihilation in sodium must exhibit a d -function distribution to conserve momentum, and our
measurements indicate that coincidence rates are strongly peaked at q = 180° with a dropoff consistent with the
finite size of the detectors. Furthermore, our Legendre polynomial coefficients of A22 = 0.0977 ± 0.0099 and
A44 = 0.0011 ± 0.0125 are consistent with their theoretical values for 4-2-0 quadrupole-quadrupole transitions
in nickel after adjusting for finite detector size. We also compared our data with the results of Monte Carlo
simulations and found them to be in agreement.

I. INTRODUCTION

When two g-rays are emitted in succession from an atomic

nucleus, their directions are often correlated due to the physics
of the emission process. In particular, when an excited nu-
clear state decays to the ground state through one or more
intermediate states, the spin of the nucleus affects the angu-
lar distribution of the photons emitted during each transition.
The relative probability that a photon will be emitted at an an-
gle q with respect to a previously emitted photon is denoted
W (q ) and depends both on the angular momenta of the states
involved in the transitions and on the multipole order of the
emitted radiation. Dr. D. R. Hamilton derived the theoretical
forms of W for dipole and quadrupole radiation and computed
the relevant coefficients for all possible angular momenta [1]. FIG. 1: Energy level diagram for the primary decay mechanism of
Here we experimentally verify these results for 60 Co . sodium-22. The half-life of 22 Na is 2.6 yr and the initial b + decay
Other nuclear processes, including electron-positron pair occurs with 90% probability.
annihilation, can also produce correlated g-rays. Sodium-22
in particular exhibits a simple angular distribution suitable for
evaluating our detection apparatus. After undergoing b + de-
cay (a 90% scenario), the resulting positron is captured by an or equivalently
electron, and the pair annihilates to produce a pair of 511 keV
g-rays as diagrammed in Figure 1. As the momentum of the
positronium system must be conserved, the photons must be W (q ) = 1 + A22 P2 (cos(q )) + A44 P4 (cos(q )) (3)
emitted in opposite directions. Therefore, the angular distri-
bution function for 22 Na is given by where A22 = 0.1020, A44 = 0.0091, and Pn represent the Leg-
endre polynomials. Note that W (q ) here gives the ratio of the
W (q ) = d (q p) (1) likelihood of detecting a photon at an angle q to the likelihood
of detecting one at an angle of 90°. By experimentally mea-
Once we have calibrated the detection apparatus and co- suring these angular correlation functions, we not only test
incidence circuit using the 22 Na source, whose physics are the validity of the theory of correlated emission but also gain
well-understood, we can consider the more complicated case insight into the angular momenta of these decay processes.
of cobalt-60. Through b emission, the cobalt nucleus de-
cays into an excited state of nickel. This state then decays to We wish to detect not only the presence of g-rays (with
the ground state though an intermediate state with a lifetime near-perfect efficiency), but also their energies. To do so we
of 8 ⇥ 10 13 s, shown in Figure 2. The g-rays emitted dur- employ two NaI(Tl) scintillation detectors. When a high-
ing these two transitions are thus effectively in coincidence. energy g-ray interacts with the sodium iodide crystal it pro-
Assuming spin states of J1 , J2 , J3 = 4, 2, 0 and quadrupole- duces a low-energy (visible) photon that is reflected towards
quadrupole transitions [2], the angular correlation function for an attached photomultiplier tube. Here it releases an electron
60
Co is described by from the cathode via the photoelectric effect, causing a cas-
cade that is multiplied at each dynode in the tube to produce a
1 1 voltage pulse at the anode whose height is proportional to the
W (q ) = 1 + cos2 (q ) + cos4 (q ) (2) energy of the initial g-ray.
8 24
 2

Scintillator A

Scintillator B

LeCroy 623B Tennelec TC412A

LeCroy 428F Octal Discriminator Variable Delay
Quad Linear
Fan-In/Fan-Out

Ortec 570
LeCroy 428F Amplifier

MCA

LeCroy 688AL
LeCroy 365AL Level Adapter Ortec 426
4-Fold Logic Unit Linear Gate

Ortec 871
Timer and Counter

FIG. 2: Energy level diagram for the decay of cobalt-60. Half-life is FIG. 4: Schematic of coincidence circuit indicating specific models
5.2 yr, and the intermediate state lives for 8 ⇥ 10 13 s. of relevant NIM modules and their connections.

   





 that would only pass a sum of signals detected simultaneously
by both detectors. A discriminator set to its minimum thresh-


old of 30 mV was used to generate logic pulses from de-
tected photons. These were passed through a variable delay
and then to a 4-fold logic unit triggering a gate between a lin-


ear fan-in and a multichannel analyzer (MCA). In this way, the
sums of pulses caused by correlated g-rays were amplified and
 delivered to the MCA while extraneous signals were filtered
out. We also kept a count of coincidence events to verify the
results of the MCA. Figure 4 describes the exact components
and connections used in the construction of this circuit.

To account for any relative delays between the two de-

FIG. 3: Schematic of detector array (top-down view). Components
tectors, we measured the coincidence rate of a 22 Na source
are drawn to scale.
(q = 180°) for various imposed delays in the two channels.
The results are recorded in Table I and plotted in Figure 5,
and from them we conclude that a delay of 16 ns in the sig-
II. PROCEDURE
nal from the first detector is necessary to synchronize the two
channels. Furthermore, we see that the resolution of this co-
In order to measure the angular correlation of various ra- incidence circuit, measured as half the full width at half max-
dioactive sources, we constructed a detector array composed imum of the plot, is 28 ns. The required imposed delay was
of two NaI scintillation counters doped with thallium. The verified and monitored throughout the experiment using a fast
surface on which the detectors were placed was marked with oscilloscope.
angles in 5° increments between 90° and 195° with respect to
a reference direction facing the first detector. A 10 in arc en- Once the circuit was confirmed to be in-time, we could con-
sured that the detector faces were always a constant distance fidently measure the emission spectra of both the sodium and
from the source, which was placed at the center. We were able cobalt sources using the MCA. Placing the gate in “Pulse In-
to position the second detector at any of the marked angles hibit” mode allowed us to measure the non-coincidence spec-
along the arc to within sq = 0.45° and sr = 1 mm. Figure 3 trum of sodium and thus perform a 2-point calibration of the
presents a sketch of this detection apparatus. MCA using its peaks at 511 keV and 1277 keV, as shown in
Not all photons produced by the radioactive sources repre- Figure 6. We could further refine this calibration with mea-
sent pairs of emissions from the same nucleus. To limit our surements of the cobalt spectrum, plotted in Figure 7, which
analysis to such pairs, we constructed a coincidence circuit clearly show peaks at 1172 keV and 1333 keV.
 3

TABLE I: Measurements of the coincidence rate at various imposed Sodium-22 Multichannel Analysis
delays between the two detector channels. Positive delays corre- 1400
Detector A (Pulse Inhibit)
spond to a delay in the first detector’s signal, while negative values Detector B (Pulse Inhibit)
1200
indicate a delay for the second detector. Coincidence (NORM)
1000
Delay [ns] Counts per Minute Delay [ns] Counts per Minute
-32 402 16 19255 800

Counts
-28 595 20 19534
600
-24 1079 24 19295
-20 3999 28 19559 400
-16 12082 32 19276
-12 17621 36 18274 200
-8 18857 40 13478 0
-4 18928 44 5371 0 200 400 600 800 1000 1200
0 19057 48 1899 Channel #
4 19222 52 1320
8 19040 56 1209 FIG. 6: Emission spectra of sodium-22 for both coincidence and non-
12 19525 60 1100 coincidence events. Non-coincidence spectra are plotted for both
channels to check for equal gain settings.

Coincidence Timings Cobalt-60 Multichannel Analysis

300
20000

● ● ●
Detector A (Pulse Inhibit)
● ●
● ● ● ●
●
●
Coincidence at 180° (NORM)
● 250 Coincidence at 90° (NORM)
●

200
15000

Counts

● 150
Counts per Minute

100
10000

50

0
0 200 400 600 800 1000 1200
5000

● Channel #

FIG. 7: Emission spectra of cobalt-60 for both coincidence and non-

●
● ●
● ●

coincidence events.
●
●
0

−20 0 20 40 60

Channel A Delay [ns] time required, several of these measurements had to be taken
on different occasions. In such cases, a repeat measurement
was made at q = 180° so that the results could be scaled to
FIG. 5: Plot of coincidence rate versus delay between the two detec- match previously-taken data. Both the raw and scaled mea-
tor channels. The full width at half maximum determines the resolv-
surements are recorded in Table III. Due to potential satura-
ing time of the coincidence circuit.
tion of the linear fan-in unit (indicated by the sharp edge at
the right of Figure 7), the coincidence peak in the cobalt spec-
trum was difficult to define in a consistent manner. Therefore,
III. DATA

With the coincidence circuit and MCA properly calibrated,

TABLE II: Coincidence rates for 22 Na in 5° increments. Counts were
we proceeded to measure the angular correlation of coinci- recorded over 60 s. Net areas (and uncertainties) were computed us-
dence rates for both 22 Na and 60 Co . For sodium, we counted ing the MAESTRO MCA emulator.
the number of events in 1 min starting at q = 180° and decre-
menting q by 5° until the rate fell to zero. These measure- q [°] Counts Net Area sNet Area
ments are tallied in Table II. 180 19460 16450 200
175 12736 10454 168
For cobalt, we took measurements over 30 min intervals 170 4443 3272 107
traveling from q = 180° to q = 90° at 30° increments, then 165 443 139 31
from q = 105° to q = 165° in order to help compensate for 160 304 0 1
temporal effects. Unfortunately, due to the long integration
 4

TABLE III: Coincidence rates for 60 Co in 15° increments. Counts Angular Correlation for Sodium−22
were recorded over 30 min. Data taken on different occasions are
scaled based on overlapping measurements (not shown). ●
●
● Data

15000
q [°] Raw Counts Corrected Counts sCounts Monte Carlo
●

180 8657 — 93
165 8687 8542 171
150 8501 — 92
135 8434 8294 167

10000
Photopeak Counts
●
●

120 7695 — 88
105 7329 7611 121
90 7598 — 87

5000
●

●


 ☢ ● ●

0
●



160 165 170 175 180

Angle [deg]


FIG. 9: Measurements and uncertainties of the angular correlation
 ☢
 function of sodium-22 based on net photopeak area. Monte Carlo
simulation results provide an indication of the effects of finite detec-
tor resolution.
FIG. 8: Analysis in 2 dimensions of the effects of finite detector
resolution on the ability to detect g-rays emitted by pair annihilation.
In particular, given the size of our detectors, the coincidence rate
should drop to zero around q = 15°. Figure 9. The two results agree with each other at the ex-
tremes, but diverge for intermediate angles. This is most likely
explained by the dependence of the real detectors’ efficiency
we based our analysis on the total count of coincidence events, on the incidence angles of the incoming g-rays. Non-normal
using Poisson statistics for the error. As the events in the peak g-rays are less likely to interact with the NaI crystal, which is
clearly dominate the total spectrum, the effects of this choice consistent with the more rapid decrease in coincidence rates
on the results should be negligible. compared with the simulation.
As the theoretical angular correlation function for 60 60 is
finite and smooth, we can fit our measurements to a func-
IV. RESULTS tion of similar form (specifically, a scaled version of Equa-
tion 3) and evaluate the agreement of the resulting coeffi-
Ideally, the angular correlation function for 22 Na would be cients with theory. Using the Levenberg-Marquardt method,
a d -function centered at q = 180° (Equation 1). Due to the fi- we find our data to be described by function parametrized by
nite angular resolution of the detectors, however, this function A22 = 0.0977 ± 0.0099 and A44 = 0.0011 ± 0.0125. The c 2
is smeared over a range of detector angles. A 2-dimensional for the fit is 5.3, which, on n = 4 degrees of freedom, implies
analysis conducted in the central plane of the detectors, pic- that deviations from the fit as large as those exhibited by our
tured in Figure 8, indicates that the coincidence rate of pho- data set would be expected 26% of the time. Our resulting fit,
tons emitted in opposite directions should rapidly approach along with the scaled data, is plotted in Figure 10.
zero between q = 13° and q = 17° given the size and separa- These coefficients are not directly comparable with those
tion of our detectors. of the theoretical correlation function, however. The finite
Our observations of the angular dependence of coincidence angular acceptance of both detectors smears the correlation
events from sodium, plotted in Figure 9, agree with this analy- function, requiring an adjustment to the theory before the
sis. The data is strongly peaked at q = 180° and drops to zero fit can be evaluated for consistency. In the case of cylin-
between q = 15° and q = 20°. drically symmetric detectors, the correction factors can be
To further test the extent to which our measurements agree computed directly via numerical integration if the efficiency
with the theory, we developed a Monte Carlo simulation pack- of the detectors with respect to the incident angle is known.
age able to accommodate custom detector arrays and radioac- For these 3 in ⇥ 3 in detectors separated from the source by
tive sources. Approximating our detectors as circular discs 10 in, the whole-spectrum correction factors are Q2 = 0.986
with 100% efficiency, we computed the expected relative co- and Q4 = 0.954, resulting in expected theoretical coefficients
incidence rates for 22 Na and plotted them along our data in of A22 = 0.1020 and A44 = 0.0091. These agree with our fit-
 5

V. CONCLUSION
Angular Correlation for Cobalt−60
8800

Data
Our results demonstrate that our theoretical understanding
●

●
● Monte Carlo
● ●
of the angular correlation of g-rays produced in nuclear de-
8600

● cay is accurate to within the uncertainties of our experimental

apparatus. In particular, our measurements of sodium-22 are
●
8400

● consistent with the d -function distribution expected from pair

●
annihilation, and our measurements of cobalt-60 agree with
Total Counts

8200

those expected from sequential 4-2-0 quadrupole-quadrupole

● transitions.
8000

While our experimental results were consistent with estab-

7800

● lished theory, uncertainties in our measurements prevent us

●
from extracting strict constraints on that theory. In particu-
7600

●
●
● lar, the long integration times required to minimize statistical
●
uncertainties in the cobalt spectrum forced the data to be gath-
ered on several different occasions, allowing temporal effects
100 120 140 160 180
in the detection apparatus to accumulate significantly. Addi-
Angle [deg] tionally, time constraints limited the extent to which we could
take overlapping measurements to compensate for these ef-
fects. The limited output range of the linear fan-in also dete-
FIG. 10: Measurements and uncertainties of the angular correla- riorated the quality of our measurements by making analysis
tion function of cobalt-60 based on total counts. The least squares of the photopeak impractical.
quadrupole-quadrupole fit is plotted as well. Results from Monte
Carlo simulations provide a second means of assessing the effects of
finite detector resolution. Future work could improve on our results by providing an
environment in which a large number of partially overlap-
ping measurements could be taken in close temporal proxim-
ted values given our propagated uncertainties. The precision ity with a highly stable high voltage power supply. Extend-
of our measurements is not great, however, as A22 is only de- ing integration times to 45 min or longer would reduce the
termined to within 36%. moderately-sized statistical uncertainties, and scaling uncer-
As an additional check of the correction factors, we con- tainties would be removed entirely. These uncertainties cur-
ducted a 2-dimensional Monte Carlo simulation in the cen- rently dominate the final error and will continue to do so until
tral plane of the detectors, which were again represented as the separation between data points approaches 1°. The Monte
3 in disks with perfect efficiency at all wavelengths and inci- Carlo representation of the cobalt source could also be ex-
dent angles. We fitted the same form of angular correlation tended to 3 dimensions to get a better handle on the smear-
function to the results, plotted in Figure 10, and obtained co- ing effects of the detectors’ finite angular resolution. Finally,
efficients statistically consistent with the corrected theoretical smaller detectors would improve our ability to measure the
coefficients with a c 2 probability of 19%. angular correlation for 22 Na .

[1] D. R. Hamilton, Physical Review 58, 122 (1940).

[2] E. L. Brady and M. Deutsch, Physical Review 78, 558 (1950).