Chapter V: The future of UCN π
5.8 Conclusion
This section discussed upgrades to the UCNπ experiment that could enable the measurement of the free neutron lifetime to a total uncertainty of less than 0.10 s.
Section 5.2 discussed how increasing the number of UCN produced by the source would decrease the statistical uncertainty of an extracted lifetime, and Sections 5.3 and 5.4 discussed how to make better use of however many UCN may be available.
Section 5.5 discussed the need to upgrade the primary detector to accommodate an increase in the number of UCN loaded into the trap. Sections 5.6 and 5.7 discussed how to decrease some of the largest systematic uncertainties of this experiment. If most of the potential upgrades presented in this section can be realized, then a future iteration of the UCNπexperiment will be able to achieve a measurement of the free neutron lifetime to a total uncertainty of less than 0.10 s.
A p p e n d i x A
OPTIMIZING THE HOLDING LENGTH
Consider a simplified version of a UCNπwith the following assumptions:
1. The number of UCN loaded into the trap is a random variable πβΌ Integer[Gaussian(π , π π)],
where π β₯ 1 is fixed and known andπ is fixed but unknown
2. The number of background events during the unloading process is π βΌ Poisson(π΅),whereπ΅is fixed and known
3. Other than background, there are no systematic effects
4. All UCN are counted exactly 70 s after the beginning of the unloading process, and the counting of UCN is perfectly efficient
For a holding lengthπ‘ ,the proportion of UCN that are expected not to decay between the end of the filling process and when the UCN are counted is π=πβ(120+π‘)/π.The 120 s in the exponent comes from two sources: 50 s of cleaning, as well as 70 s of delay between the beginning of the unloading process and when the UCN are counted.
As explained in Section 3.2.1, there is benefit to having multiple short holding lengths. Anoctetis a set of eight runs: four short runs with holding lengths of 20 s, 50 s, 100 s, and 200s; and four long runs with a holding length ofπ‘long.Starting from Equation 3.14 and making the simplifying assumption thatπ =1 andππ=ππΆ =0, the likelihood function simplifies to
L (π, π|π’π, π‘π) =βΓ
π
ln Gaussian π’π
π ππ+π΅, π π π2
π +π ππ(1βππ) +π΅
, (A.1) whereπsums over all runs in the analysis.
The maximum likelihood ofπ andπ,as well as the uncertainties of those estimates (ππ, ππ), was found using the methodology described in Section 3.9. ππis inversely proportional to the square of the amount of data gathered, so maximizing the data gathered per unit of time (Ξπ‘) is equivalent to maximizingπβ2
π /Ξπ‘ .
Figure A.1: Relative amounts of data gathered per hour as a function ofπ‘πΏ. Each run takes a total timeπ‘ +πΏ to complete, whereπ‘ is the holding length of this run andπΏis some constant amount of overhead time per run. An octet takes a total time of 370 s+4π‘πΏ + 8πΏ to complete, where π‘πΏ is the holding length of the long runs and 370 s= (20+50+100+200)s. Monte Carlo simulations were modelled after the simplified version of the UCNπexperiment described at the beginning of this appendix. These simulation were repeated with different choices for the long holding lengthπ‘πΏ. Figure A.1 shows the rate of data gathered per hour as a function of long holding length for reasonable choices ofπ =10,000, π΅=50,andπΏ=630 s, and for two options for π . Larger values for π (larger scales of the uncertainty of the normalization estimateππ relative to the ideal case ofππ =β
π) led to a larger value of the optimum holding length and decreased the statistical precision of a measurement made from a fixed number of runs.
A p p e n d i x B
UCN EVENT RECONSTRUCTION ALGORITHMS
In Algorithm 1:
β’ π‘πis the time that theπthphotoelectron was observed
β’ chπ β {1,2} is which photomultiplier tube of the primary detector the πth photoelectron was observed in
β’ ππΆ is the coincidence window
β’ ππ is the tail window
β’ πPEis the photoelectron threshold
Algorithm 1UCN reconstruction without a prompt window INPUTS:{(π‘π,chπ)}ππΏ=1whereπ‘π β€ π‘π+1
PARAMETERS:ππΆ, ππ, πPE πβ 1
whileπ β€ πΏdo
π , π,stateβπ+1,1,0 while π β€ πΏdo
ifstate=0then ifπ‘π βπ‘π > ππΆ then
breakloop over π else
πβ π+1 if chπ β chπ then
stateβ 1 end if end if else
ifπ‘π βπ‘πβ1> ππ then breakloop over π else
πβ π+1 end if end if
π β π+1 end while
if state=1 andπ β₯ πPEthen
push backreconstructed UCN event with start timeπ‘π, end timeπ‘πβ1+ππ, andπphotoelectron
πβ π else
πβπ+1 end if end while
return reconstructed UCN events
In Algorithm 2:
β’ π‘πis the time that theπthphotoelectron was observed
β’ chπ β {1,2} is which photomultiplier tube of the primary detector the πth photoelectron was observed in
β’ ππΆ is the coincidence window
β’ ππis the prompt window
β’ ππ is the tail window
β’ πPEis the photoelectron threshold
Algorithm 2UCN reconstruction with a prompt window INPUTS:{(π‘π,chπ)}ππΏ=1whereπ‘π β€ π‘π+1
PARAMETERS:ππΆ, ππ, ππ, πPE πβ 1
whileπ β€ πΏdo
ππ, ππ,state, π β1,1,0, π+1 while π β€ πΏdo
ifstate=0then ifπ‘π βπ‘π > ππΆ then
breakloop over π else
ππ, ππ βππ+1, ππ +1 if chπ β chπ then
stateβ 1 end if end if
else if state=1then ifπ‘π βπ‘π > ππ then
if ππ β₯ πPEthen stateβ 2 else
breakloop over π end if
else
ππ, ππ βππ+1, ππ +1 end if
end if
ifstate=2then
ifπ‘π βπ‘πβ1> ππ then breakloop over π else
ππ β ππ +1 end if
end if π β π+1 end while if state=2then
push backreconstructed UCN event with start timeπ‘π, end timeπ‘πβ1+ππ, ππprompt photoelectrons, andππ total photoelectrons
πβ π else
πβπ+1 end if end while
return reconstructed UCN events
In Algorithm 3:
β’ π‘πis the start time of the reconstructed UCN event
β’ πΏπis the length of each reconstructed UCN event
β’ ππ
π is the number of photoelectrons found in the prompt window of each reconstructed UCN event
β’ ππ
π is the total number of photoelectron found in each reconstructed UCN event
β’ ππis the prompt window from the reconstruction algorithm
β’ ππ is the tail window from the reconstruction algorithm
β’ πPEis the photoelectron threshold from the reconstruction algorithm
β’ cdf is a cumulative distribution function of the measured time of detection of photoelectron in a reconstructed UCN event, relative to the time of the first photoelectron in a reconstructed UCN event
Algorithm 3UCN-event-tail correction INPUTS:
π‘π, πΏπ, ππ
π , ππ
π
πΏ
π=1whereπ‘π β€ π‘π+1, πΏπ β₯ ππ,andπPE β€ ππ
π β€ ππ
π
PARAMETERS:ππ, ππ, πPE,cdf {πΉπ}πΏπ=1,{ππ}ππΏ=1β0,1
forπ β1 to πΏdo π΄πβ ππ
π /cdf(πΏπ) for π βπ+1 to πΏdo
πΉπ β πΉπ+ π΄π
cdf(π‘π βπ‘π+ππ) βcdf(π‘π βπ‘π) ππ
π β ππ
π β π΄π
cdf(π‘π βπ‘π+πΏπ) βcdf(π‘πβπ‘π) end for
for π β ππ
π βπPE+1 toππ
π do ππ βππβPoisson(π|πΉπ) end for
end for
return {ππ}ππΏ=1
A p p e n d i x C
THE THINNED POISSON DISTRIBUTION
Denote the probability of observing a random valueπ₯ drawn from a Poisson dis- tribution with mean π as ππ(π₯|π). Denote the probability of observing a random value π¦from a Binomial distribution withπ trials, each of which has a probability of success π,asππ΅(π¦|π , π).
Letπ βΌPoisson(π), and then letπ₯ βΌBinomial(π, π).Then π(π₯|π , π) =Γ
π
ππ΅(π₯|π , π) Β·ππ(π|π)
=
β
Γ
π=π₯
π!
π₯!(π βπ₯)!ππ₯(1β π)πβπ₯Β· πππβπ π!
= ππ₯πβπ π₯!
β
Γ
π=π₯
ππ(1βπ)πβπ₯ (π βπ₯)!
= (π π)π₯πβπ π₯!
β
Γ
π=π₯
[π(1β π)]πβπ₯ (π βπ₯)!
= (π π)π₯πβπ π π₯!
β
Γ
π=π₯
[π(1β π)]πβπ₯πβπ(1βπ) (π βπ₯)!
= (π π)π₯πβπ π π₯!
β
Γ
π=0
[π(1β π)]ππβπ(1βπ) π!
= (π π)π₯πβπ π π₯! Β·1
= (π π)π₯πβπ π π₯!
=ππ(π₯|π π).
Thereforeπ₯ βΌ Poisson(π π).