Radiation Characteristics of Hamamatsu Layer 2 Sensors for the D0 Run IIB Upgrade
(revised, July 14, 2003)
M. Ahsan(a), T. Bolton(a), K. Carnes(b), R. Demina(a), T. Gray(b), S. Korjenevski(a),
A. Rankin(b), M. Shamin(a)
(a) Kansas State University, HEP group, Manhattan, KS
(b) Kansas State University, James R. Macdonald Laboratory, Manhattan, KS
Revised 4/10/03
1. Summary
Four Hamamatsu (HPK) layer 2 (L2) sensors were irradiated with 10 MeV proton beams to accumulated exposures exceeding 1.3×1013 p/cm2. Type inversion was observed for all sensors. In all cases, depletion voltage remained well below 200 V (Figure 1), and breakdown voltages exceeded 350 V (Figure 2).
Figure 1 Whole sensor inverse squared capacitance vs. bias voltage for the four irradiated HPK L2 sensors at different cumulative radiation exposures.
Figure 2 Whole sensor leakage current vs. bias voltage for the four irradiated HPK L2 sensors at a series of radiation exposures.
The following sections provide details of the irradiation procedure and analysis.
2. Setup[1]
To irradiate the detectors we used the 7 MV tandem Van de Graaf accelerator located at the Kansas State University (KSU) James R. Macdonald Laboratory(JRML)[2] in Manhattan, KS. In our tests, the beam energy was set at 10 MeV, and beam size at the target was approximately 3 mm in diameter. Irradiation was performed in steps, with beam currents set between 20 nA at low dose points and 55 nA for the highest dose points. Figure 3 shows a schematic of the beam line.
To provide uniform irradiation of the 320 mm thick XY=9.8×3.8 cm2 Si sensors, the beam was independently rastered in the X- and Y- planes with oscillating electric fields produced by a specially built electrostatic deflector. Raster frequencies were 80 Hz and 2.4 kHz in the horizontal and vertical directions, respectively. The beam was actually moved in discrete steps by a DAC output card controlled by LabView. The overlap in beam coverage between successive steps was 80%. Additional beam averaging in the horizontal direction was produced by the few percent jitter present in the beam energy.
Figure 3 Schematic of irradiation test beam line in the KSU James R. Macdonald Laboratory.
Sweeping performance was verified by placing G10 dummy targets in the 10 MeV proton beam. Radiation produced a discoloration over the region scanned by the beam, as can be seen in Figure 4 for the configurations for sensors for L2, as well as L0 and L1. Beam illumination is quite uniform for all layers. For L2, an upstream aperture cuts the beam off at the extreme horizontal limits. This feature reflects the fact that the beam line was originally assembled to test L0 and L1 assemblies; and most of the unexposed part of the L2 sensor is inactive material. The slight tilt in the three rectangular intensity patterns reflects a small misalignment in the electric field deflector plates and is of no consequence.
Figure 4 Photographic negatives of G10 pieces exposed to 10 MeV proton beam with electric field sweeping on. Rectangular outlines show sensor size, and lighter areas show beam profiles. The G10 pieces are shown rotated by 90 degrees relative to their actual orientation in the beam.
The silicon detectors were mounted on 3.2 mm thick aluminum holder and placed into an evacuated target chamber (typical vacuum ~ 10-5 mtorr). The sensors were held in place by Teflon clamps to a rectangular surface approximately 1 mm in width around the perimeter of the sensor and recessed into the Al holder by the 320 mm sensor thickness. This contact provided adequate sinking for the beam heating that occurred during the runs, maintaining the temperature at an estimated 22±2 °C.
Figure 5 Aluminum sensor holder with sensor mounted. The holder is inserted into the target vacuum chamber at left and held in place with a vacuum flange (not shown).
Radiation exposure was monitored by the beam current, and more accurately, by a current integrating Faraday cup with 10 pC least count resolution placed downstream of the sensors. The Al target holder stopped all beam flux not incident upon the Si sensor, so the Faraday cup provided a direct measurement of the flux. The HPK-L2-52 and HPK-L2-54 detectors were mounted on opposite sides of the same holder and exposed simultaneously. This turned out to be a relatively poor decision as large multiple Coulomb scattering necessitated significant acceptance corrections. As a consequence, HPK-L2-54 was effectively exposed to a 7 MeV p beam, accounting for energy loss in the upstream sensor. Sensors HPK-L2-59 and HPK-L2-62 were run one at a time in the beam, and the Faraday cup was moved upstream from 30 cm to 12 cm separation from the sensors. Acceptance corrections used in the final flux determinations are described in Section 5.
Figure 6 Original Faraday cup geometry (left) and improved higher acceptance setup (right).
After irradiation, sensors were annealed for 80 minutes at 60±2 °C, and then transported to a freezer held at -20 °C for storage. They were then taken to the KSU clean room, where leakage current and capacitance measurements were performed using our Alessi R61 semiautomatic probe station via a LabView driven acquisition program. Measurements were made at 0±1°C on the probe station’s cold chuck.
3. Data
Table 1 summarizes exposure information for each data point. Integrated flux is computed from the measured integrated charge and the known area of the sensors. Acceptance corrections are applied to account for protons scattered out of the Faraday cup. The factor that takes protons to 1 MeV neutrons for non-ionizing energy loss calculations[3] is taken to be 4.58 for the majority of runs that see 10 MeV beam and 8.27 for a few runs in which the second sensor in the beam line sees a 7 MeV proton beam due to energy loss in the upstream sensor. The equivalent neutron flux assumes the NIEL scaling model, upon which further comment will be given.
Table 1 Summary of HPK L2 exposure points. Note that sensor HPK 54 effectively saw a 7 MeV proton beam.
Figure 1 and Figure 2 summarize CV and IV scans. Our raw data is accessible at unix.phys.ksu.edu/~d0server.
4. Analysis
Figure 7 Representative two-line intersection procedure used to estimate depletion voltage from 1/C2 vs. V plot. (These data correspond to the eighth entry in Table 2).
Depletion voltages were obtained from the intersection of two lines fit to the 1/C2 vs V curves (Figure 1) for voltages well below and well above full depletion. A representative fit is shown in Figure 7. This technique contains ambiguities having mainly to with the selection of points used in determining the lines. The resulting systematic uncertainty ranges from a few volts at low exposure points to 10 V at the highest irradiation points where the 1/C2 vs V curve becomes more complex. A standard leakage current for a sensor at a given irradiation was then defined as the current measured at the depletion voltage plus 50 V corrected to a temperature of 20C. For measurements taken at T=0C, this factor was 5.3. Data are summarized in Table 2, Figure 8, and Figure 9.
Table 2 Depletion voltage and leakage current for HPK L2 sensors at different exposures.
Figure 8 Depletion voltage vs. proton fluence for HPK L2 sensors. HPK-L2-54 was exposed to a 7 MeV proton beam. For all other sensors, the beam energy was 10 MeV at the upstream end of the sensor.
From Figure 8 one can see that the point of inversion occurs at an integrated 10 MeV proton fluence of ~1×1013 cm-2.
Figure 9 Leakage current vs. proton fluence fluence for HPK L2 sensors. HPK-L2-54 was exposed to a 7 MeV proton beam. For all other sensors, the beam energy was 10 MeV at the upstream end of the sensor.
The leakage current is observed to increase linearly with flux as expected. If we correct the data to correspond to measurements at a temperature of 20°C, and use only data from HPK-L2-59 and HPK-L2-62, which have more reliable fluence measurements, we obtain aP=6.8±0.2×10-17A/cm. This is approximately two times lower than the value expected from NIEL scaling from neutrons or high energy protons. However, the slope is in accord with other measurements that use 10 MeV protons.[4]
5. Acceptance Corrections
Our original setup for the target holder and Faraday cup (used to measure the flux) is shown in Figure 6. The entrance to the Faraday cup was 30 cm downstream from the sensor location. This position was fine for running 10 MeV protons through single sensors. However; its acceptance for the same energy through double sensors was very poor, essentially because a double sensor nearly stops the beam and dramatically increases multiple Coulomb scattering (MCS). We realized this point in the original setup of the assembly; double sensors were to be handled by increasing the beam energy to 14 MeV. However, we failed to raise the beam energy in our initial December 2002 running. For our January running, we never put more than one sensor at a time in the beam line and also moved the Faraday cup closer to the target.
In order to use the December data, we measured acceptance corrections for the double sensor configuration. To understand these effects further, we calculated acceptances with a Geant-3 simulation. The much smaller corrections for the improved January setup were also measured.
Figure 10 Comparison of calculated radial distributions of protons transmitted through two L1(top row) or L2 (bottom row) sensors at the Faraday cup for December running. The effective sensor thicknesses in the calculation that have been tuned to match the data are shown. The right column blows up the radial distributions in the left column to the 0 to 5 cm interval accepted by the Faraday cup.
Acceptances for single and double sensors were measured in a special session in which the number of counts in five minute runs with sensors mounted on the target was compared to the number with no sensors mounted. The beam sweeping profile was kept the same for the two runs; hence, the ratio of counts yields the correction for MCS or range-out losses. Because layer 0 (L0) and layer 1 (L1) sensors were also being irradiated, acceptance corrections were measured for these setups as well; the different geometries proved useful in understanding the mechanism for losses in the Faraday cup. Acceptance corrections are given in Table 3. The major uncertainty in this method was due to drifts in the beam intensity between target-in and target-out runs. This effect is estimated to contribute a 30% error to the measurement.
Calculations with Geant3 matched the measurements poorly unless energy loss fluctuations were incorporated into the Geant calculation, and the effective sensor thicknesses used in the Geant model were increased to 330 (vs 290 actual) mm and 355 (vs. 320 actual) mm for L1 and L2, respectively. The need for these adjustments is perhaps not surprising given that the sensors were modeled by simple silicon slabs, and that the Geant energy loss/MCS algorithms are model dependent and complicated for protons that are nearly stopped. With adjusted effective thicknesses, the acceptance calculations satisfactorily describe the data. Calculated acceptances were used to correct the flux in the data with an uncertainty, estimated by comparing to the measurement, of 30% assigned.
configuration / Acceptance-measured / Acceptance-calculated1×L1 / 1.15±0.09±0.35 / 1.10(teff.=330 mm)
2×L1 / 1.92±0.12±0.58 / 1.97(teff =330 mm)
1×L2 / 1.30±0.16±0.39 / 1.28(teff =355 mm)
2×L2 / 12.5±1.4±3.8 / 7.00(teff =355 mm)
Table 3 Measured and calculated acceptance corrections for single and double L1 and L2 sensors in the December 30 cm Faraday cup separation configuration. The effective sensor thicknesses used in the Geant calculation are shown.
In addition to the apparatus improvements, a better procedure was also used to measure the (much smaller) acceptance corrections for the January runs. A series of 10 alternating one minute runs were performed with a dummy sensor mounted in place. The odd numbered runs had the beam sweeping over the sensor, while the even numbered ones kept the beam focused at the center. In the centered configuration the acceptance is nearly 100%, so these runs effectively tracked changes in beam intensity. The efficiency determination is described in detail in an appendix (Section 6). The same effective sensors determined from the December data were used in Geant calculations, although in this case there was very little sensitivity to detector thickness. Calculated acceptances were used to correct the flux in the data with an uncertainty, estimated by comparing to the measurement, of 10% assigned.
configuration / Acceptance-measured / Acceptance-calculated1×L1 / 1.06±0.10 / 1.01(teff =330 mm)
2×L1(not used) / 1.03±0.10 / 1.12(teff =330 mm)
1×L2 / 1.00±0.10 / 1.11(teff =355 mm)
2×L2(not used) / 2.27±0.23 / 2.71(teff =355 mm)
Table 4 Measured and calculated acceptance corrections for single and double L1 and L2 sensors in the January 12 cm Faraday cup separation configuration. The effective sensor thicknesses used in the Geant calculation are shown.
Figure 11 Comparison of calculated radial distributions of protons transmitted through one L1(top row) or L2 (bottom row) sensor at the Faraday cup for January running, for which the Faraday cup is 18 cm closer to the target and only one sensor at a time is in the beam. The effective sensor thicknesses in the calculation that have been tuned to match the data are shown. The right column blows up the radial distributions in the left column to the 0 to 5 cm interval accepted by the Faraday cup.
6. Appendix: Acceptance Procedure for New Faraday Cup Configuration
Define Snmi as the number of counts in a one minute run taken at time ti with the beam in full sweep mode with m layer n type sensors mounted in the holder (m=0,1,2; n=0,1,2). Define Nnmi as the corresponding number of counts with the same sensor configuration and the beam centered at the center of the target. (Only dummy or damaged sensors. were used for this measurement). Our model is