Simulation of a proposed new high-purity germanium detector for LENA
Grayson Rich
Advisor: Arthur Champagne
PHYS482L, UNC-Chapel Hill, Spring 2009
For nuclear physics experiments where the expected experimental count rate is
very low, minimization of background noise in measurements is of great
importance. At the low energy nuclear astrophysics (LENA) lab at Triangle
Universities Nuclear Laboratory (TUNL), investigations are made into nuclear
reactions important in stellar evolution; such experiments typically have much
lower count rates than some of the others conducted in the field of nuclear physics,
and as such, it is prudent for LENA to make efforts towards background reduction.
To this end, I have been working on initial simulations of a transversely segmented
p-type point-contact high-purity germanium detector that has been considered for
fabrication for LENA's use. The goal of these simulations is to predict if the detector
would warrant its cost (estimated to be near $200,000) through efficient detection
and reduction of background in spectra. While more simulations need be run, and a
serious cost/benefit analysis performed, the initial results do show that such a
detector can remove a non-trivial amount of background from simulated spectra.
The simulation software used for these tests was based on GEANT4 by CERN, and
the geometry was a very simple approximation of what an experimental
environment for the detector might look like: an isotropic circular plane source of
gamma rays, 1 inch in diameter, parallel to the face of the detector, was located 1.6
cm away from the front face of the detector and background radiation sources
(Thallium-208 and Potassium-40) were emulated as isotropic point sources located
6 cm directly above the midline of the detector volume. While the target gamma
approximation is fairly accurate, the background sources are geometrically
unrealistic and should be more carefully approximated in future simulations: nonethe-less, their current incarnation does serve to demonstrate the potential for
background elimination with this detector.
In theory, there are several ways that the output from this proposed detector could
be used to suppress background contamination of spectra. As the primary interest
at LENA is to measure photoabsorbed gammas from the target, one might look at
how many scattering events take place for photoabsorbed target gammas and for all
target gammas: presumably,
there will be some kind of a
threshold for fully-absorbed
gammas, so eliminating all
events which show fewer than
X interactions within the crystal
should cut out some nonphotoabsorbed target gammas.
Such a technique requires the
ability to count interactions
with the germanium, a feature
present with a point-contact
type detector but lacked with
meaningful resolution by
coaxial detectors. Above is a histogram depicting the number of interactions target
gammas undergo in the detector volume: the solid, green area are photoabsorbed
events, so we can see that eliminating events with fewer than 5 interactions would
remove some background without having a large effect on the number of photopeak
events counted. Such a cut would also be applied to room and cosmic background
events, and would have positive results for noise suppression: many background
gammas may be incident with the detector at relatively high angles or near edges,
increasing the probability that they will scatter out of the crystal volume with few
interactions and thus be
ignored. If we examine the
Tl-208 simulation
background spectrum
(shown at left), we see that
applying this cut reduces the
counts by roughly .28 (the
blue area are counts that
would be removed).
Another way to cut out
background noise is to
examine where the first
interaction with the crystal volume occurs:
gammas from the target should interact first in
the front half of the crystal, while room and
cosmic background gammas should have a
relatively higher probability of showing a first
interaction in the latter half of the volume. If
we overlay the locations along the detector’s
axis of the first interactions of
photoabsorption events and Tl-208
background events, we get the histogram
shown at right (the detector is centered at z=0,
and higher z values are towards the front of
the detector). The “target gammas” curve
represents photoabsorbed gammas from the
target (those in which we are interested). Visually, we can see that cutting counts
occurring below z=0 will eliminate many background counts from Tl-208 without
affecting too dramatically photopeak counts.
We can now examine the
effect the cuts detailed
above have on the
spectrum we have from
simulations of target and
background gammas. The
total spectrum is shown
here at left: the green
highlighted section is the
contribution from target gammas, and the solid red section is the focus of the
experiment (photoabsorbed gammas). The remaining parts of the spectrum are due
to our two background sources: 40K and 208Tl. In addition to the coarse geometric
approximations of background, the relative intensity of the background sources to
each other, and to the target gammas, is entirely artificial and based on no physical
considerations whatsoever; this being said, this spectrum can still be used to
demonstrate the effects of background reduction. Graphical representations of the
effects of the cuts will follow the conclusion of this paper in the appendix (so as to
not clutter the text unnecessarily), but a quantitative explanation is offered here for
each spectrum independently (target gammas, 208Tl, and 40K).
By applying the cuts determined above (first interaction in first half of detector and
more than 5 interactions), we are able to reduce the number of counts from 208Tl
from 14225 to 5110: 0.36 of its initial contribution. We’re able to reduce 40K’s
contribution even more significantly: 15227 counts drop to 6767, 0.44 of the initial
amount. When we apply our cuts to our target gammas, we drop the total counts
from 30312 to 13368. There is, of course, concern about how these cuts affect the
height of the photopeak: we reduce its height from 11984 to 8207 counts
(considering here only the contribution from the target gammas).
Conclusions, future work
We can see that a detector such as the one proposed and simulated here shows
promise for background reduction. To determine a more realistic estimate of its
potential capability, more in depth quantitative analysis must be performed and in
order for this analysis to be of any real meaning, more realistic simulations must be
performed. Much of the artificiality of the simulations whose results are shown here
comes from the simple, unrealistic geometry and treatment of background sources:
fully modeling LENA’s current target and detector configuration (swapping the
current detector for our new, segmented detector) could be done without much
difficulty, but would require careful measurement of the existing setup. Giving
realistic treatment to background sources would require careful thought and
consideration of the physical realities of these sources: these realities include from
what materials they originate (e.g., lead bricks, concrete walls) and how, in these
materials, the sources are distributed.
Another degree of realism can be added to the simulations by considering an
experiment that has already been run at LENA. By normalizing the peaks observed
in the experimental spectra and considering their sources, a near-experimental
spectrum could be simulated in GEANT4 and then analyzed in depth, allowing more
tangible comparisons to real-world data and a more realistic demonstration of how
effective this detector might be at suppressing background in an experimental
situation.
This entire endeavor has operated under the assumption that the output of the
detector can be interpreted such that each segment has its own independent output
free of influence from other segments; in reality, the output of each segment would
be convoluted with information originating in each of the other segments. Many
other groups have dealt with interpretation of such outputs, and deconvolution of
these signals is a science that has been shown to be manageable. Owing to these
complications, a highly realistic simulation and characterization of this detector
quickly becomes the territory of a Masters, or even Doctoral, project.
Given the initial, promising results shown here, further exploration of the
possibilities of such a detector is well advised, but fair treatment of such an
endeavor requires full-time dedication.
Appendix A: Background reduction results
For each histogram, the smaller curve represents the spectrum resulting from
application of background reduction cuts and the larger area is the original
spectrum, with no cuts
applied.
Figure 1 - 40K
Figure 2 - 208Tl
Figure 3 - Target gammas