Two devices G101a and G102a have been tested so far. The crystals are 100mm diameter and 33.3 thick. Fig. A1 shows the ionization electrode design on one side of the crystal. A ground plane is deposited on the other side of the crystal.

Fig. A1. 100mm ionization testing device electrode design pattern

More complete run summary can be found at here and here.

• G101 had a scratch across Q2 to Q4 on the electrode side.
• Charge return wire was only bonded to Phonon A on DIB 1.
• 4 collimated Am-241 sources (~20 Hz each) were placed ~2mm above the 4 electrodes.

• 60 keV peaks observed on Q1, Q2, and Q4.
• Q1 and Q3 were much weaker than Q2 and Q4.
• Charge collection efficiency was calculated based on Q2.

Fig. A2. Charge spectrum of Q1-4 at -4V bias	Fig. A3. Charge collection efficiency based on Qsum(cQ2)

• The detector could hold neutralization for ~0.4 hours at -4V bias.
• Neutralization time increased with voltage bias magnitude.
• This provides a way to study charge trapping in the crystal.

Fig. A4. Q2 vs Eventtime at -4V	Fig. A5. Neutralization time vs. bias

More complete summary can be found at here

• Charge return wires are bonded to Phonon B on DIB 1 and Phonon C and D on DIB 2.
• 4 collimated Am-241 sources (~20 Hz each) were placed ~2mm above the 4 electrodes.

• 60keV peak observed on all 4 channels.
• 4 electrodes were better balanced than G101a.
• Charge collection efficiency calculated for 3 inner channels.

Fig. A7. G102a charge spectrum at -4V	Fig. A8. Charge collection efficiency based on Qsum(cQ#)

• G102a could hold neutralization ~15min.
• Good neutralization time increased with bias.

Fig. A9. Q2 vs. Eventtime at -4V	Fig. A10. Q2 vs. Eventtime at -8V

• Share events were still observed between Q1 and Q3, Q2 and Q4.

Fig. A11. Events shared by Q1 and Q3 at -4V	Fig. A12. Events shared by Q2 and Q4 at -4V

The justification for this run lays in the discrepancy found between R16 and R20's results with G102a. Please refer to Hassan's preliminary comparison between the two runs for details (http://www.hep.umn.edu/cdms/cdms_restricted/K100/runs/run020/compareRuns/k100Run16Run20ComparisonPrelim.pdf). Essentially, in R20 we found that there were a large abundance of low energy events which were not present in R16. Note that in R16 Q4 was not working.

The source configuration for this run is different from the previous because the source has been placed next to the opposite face of the detector. The first round of taking data included series with the detector biased at -4V, -6V, -8V, and -10V and exposed to the internal Am-241 source. In these data series, the spectra looked relatively normal (https://zzz.physics.umn.edu/cdms/doku.php?id=cdms:k100:run_diary:20111204). There were a couple of positive things to take from this: all four channels seemed to be working properly, and no low energy event peaks were found similar to those in R20. The chi^2 vs. charge plots for -8V and -10V, however, featured a double band. An investigation was made into this and will be mentioned a bit later.

Following the promising results from the first 4 series, the next 3 data series were taken with the detector biased at -4V, -8V, and -13V and exposed to an external Ba-133 source. Surprisingly, the spectra for these biases revealed significant low energy event peaks (https://zzz.physics.umn.edu/cdms/doku.php?id=cdms:k100:run_diary:20111205). This is interesting to note because in R20 the spectra for -13V bias exposed to Ba-133 seemed to make the low energy events disappear. It seems from the data taken in this run, though, that this disputes that. The low energy event peak seems to be source-independent.

It is important to note that after the first day of taking data, each data-taking sequence afterwards experienced an interruption with the DAQ crashing. The implications of this are significant. When the DAQ crashed overnight, it left the detector biased for hours. Usually we flash between data series to reverse the effects of degradation. In these cases, though, the detector is usually only biased for an hour at a time. Perhaps it would be necessary to flash longer due to the extensive bias time from the DAQ crashing. We flashed the detector overnight for a total of 10 hours of flashing. The detector was biased at -13V with the Am-241 source. The chi^2 vs. charge and charge spectra for this data set can be seen in figures B1 and B2. Below Fig. B1 and B2 are the plots from data with the detector taken under the same conditions except for the extensive flashing (and post-DAQ crash).

Fig. B3 - Chi2 vs. Charge (1000 a.u.) at -13V Bias

Fig. B4 - Charge Spectra (1000 a.u.) at -13V Bias

Comparing the two -13V/Am-241 data sets, it makes a strong support for the argument that there is a lot of degradation to overcome in the detector when the DAQ crashed. The normal spectra is achievable, though, as seen in Fig's B1 and B2. This speaks well for the functioning of the detector.

Following in this investigation are long series of taking data to see the behavior of the charge spectra, etc, over time. The first set is a 12 hour run of taking data. The subsequent set after that is an ongoing data series that is being analyzed periodically to see if/when the low energy events appear.

Fig. B9b - Charge Spectra (1000 a.u.) at -10V Bias

Here begins the continous long series, referred to as Long Series #2.

Recall from the -4V → -10V data set that a double band feature appeared in the chi^2 vs. charge plot in Q3 for -8V and -10V. In all subsequent data sets with the large low energy event peak, a significant spread in the chi^2 vs. charge is present. Below are some quick investigations into the nature of the bands/spread.

In Figs. C1a and C2a, we make a distinction between the 3 series of data taken at each bias. Clearly, the spread is occurring because the chi^2 values between the series behave differently. It is not in a consistent manner, though, between different biases. For -8V the third series has much higher chi^2 values than the first two series, while in the -10V set its chi^2 values are lower. We looked at the PSD's for these biases to see if there was a clear correlation between chi^2 and the PSD, and for Q3 we do see a slightly wild behavior in the PSD.

Fig. C1a - Chi2 vs. Q3 (1000 a.u.) at -8V Bias

Fig. C1b - PSD (V/Sqrt[Hz] vs. Hz) at -8V Bias

Fig. C2a - Chi2 vs. Q3 (1000 a.u.) at -10V Bias

Fig. C2b - PSD (V/Sqrt[Hz] vs. Hz) at -10V Bias

We decided to also compare the PSD's for the -13V sets presented earlier (Figs. B1-B4) for their low event characteristics.

Fig. C3 - PSD (V/Sqrt[Hz] vs. Hz) at -13V Bias - Long Flashing Period

Fig. C4 - PSD (V/Sqrt[Hz] vs. Hz) at -13V Bias

It's interesting to see that in the long flashing spectra, the first series has only one peak in the beginning. One thought is that having double peaks at the beginning may be due to the switching on/off of lights in the lab.

Below is a list of the acceptable and bad data sets at various biases. There were several biases that had multiple retakes, so if at one bias, say, -13V with Ba-133, had good and bad data, I've only listed the good data below. However, if a bias has only bad data, say +3V with Ba-133, then I've listed it below.