Characteristic study of silicon sensor for ILD ECAL
Shusuke Takada, Hiroto Hirai, Kiyotomo Kawagoe, Yohei Miyazaki, Yuji Sudo, Taikan Suehara, Hiroki Sumida, Tatsuhiko Tomita, Hiraku Ueno, Tamaki Yoshioka
CCharacteristic study of silicon sensor for ILD ECAL
Shusuke Takada, Hiroto Hirai, Kiyotomo Kawagoe, Yohei Miyazaki, Yuji Sudo,Taikan Suehara, Hiroki Sumida, Tatsuhiko Tomita, Hiraku Ueno, Tamaki YoshiokaKyushu University
Abstract
Excellent jet energy measurement is important at the International Linear Collider (ILC) becausemost of interesting physics processes decay into multi-jet final states. We employ a particle flow methodto reconstruct particles, hence International Large Detector (ILD) needs high spatial resolution which canseparate each particle in jets. We study pixelized silicon sensors as active material of ILD Silicon electro-magnetic calorimeter (SiECAL). This paper reports studies of temperature and humidity dependence ondark current and response of laser injection.
The International Large Detector(ILD) is a proposed detector for the International Linear Collider (ILC) [1,2].Figure 1 shows the ILD and its electromagnetic calorimeter (ECAL). The ILD ECAL is a sampling calorimeterconsisting of tungsten absorber and subdivided sensors. As candidates for the sensor technologies three typesof sensors are designed, which are silicon detector, scintillation detector and hybrid of the two. ParticleFlow Algorithm (PFA) [3] is planned to used as analytical method in ILC. In PFA, particles are detected atoptimal detectors. Charged particles, photons and neutral hadrons are detected in a tracking detector, anelectromagnetic calorimeter and a hadron calorimeter, respectively. To improve the performance of PFA, thesensors are finely segmented to achieve excellent particle separation in jets. The pixel size of sensors has tobe less than 1 cm to satisfy a requirement for jet energy resolution. The details of current ECAL design ofILD are described in the ILD-DBD. Electromagnetic Calorimeter(ECAL)
Figure 1: ILD detector and ECAL
Talk presented at the International Workshop on Future Linear Colliders (LCWS14), Belgrade, Serbia, 6-10 October 2014. a r X i v : . [ h e p - e x ] M a r Pixelized silicon detector
Figure 2 shows a sample of pixelized silicon sensor (Si-pad) which was made by Hamamatsu Photonics. Thespecifications of Si-pad are the following: 5.5 × pixel size, 320 µ m thickness and 16 ×
16 pixels.A silicon sensor usually has “Guard-ring” which is located at edge of the sensor. Advantages of guard-ring are to collect surface current and to prevent an electric discharge at the sensor edge. The guard-ring,however, decreases sensitive area and makes crosstalk signal along the edge of the Si-pad. Our motivationis to optimize Si-pad design by studying crosstalk caused by the guard-ring, crosstalk between pixels. Darkcurrent is measured as a basic property of the Si-pad and compared among the guard-ring types.Figure 3 shows small Si-pad samples (baby chips) which are measured to compare the effects of differentguard-ring designs. Pixel size and thickness of the baby chips are the same as those of 16 ×
16 pixels type.Two types of baby chips were used. One is for comparison of guard-ring effects and the other is for measuringinter-pixel crosstalk. For the guard-ring comparison we used 4 types of guard-ring(s), which are 0, 1, 2 and 4guard-ring(s). Figure 3 and 4 show the structure of each guard-ring. The guard-ring of the 1 guard-ring typeis a continuous line without any divisions and those of the 2 and 4 guard-rings types are alternately divided.Figure 5 shows a sample for inter-pixel crosstalk study. Electrodes covering the chip is partly meshed inorder to pass the laser photons though the mesh to the bulk of the silicon sensor.Figure 2: Picture of the 16 ×
16 type Si-pad. Figure 3: Picture of the baby chips. The right picturesshow magnification images around the edge regions ofthe chips. . mm Guard ring
Figure 4: Schematics of the guard-rings of baby chips. Structure of the 4 guard-rings type is similar to the2 guard-ring type. 2igure 5: Pictures of the baby chip of meshed electrodes. The right picture is a magnification of the edge ofone open pixel.
We performed two measurements to optimize the Si-pad design. One is the measurement of temperatureand humidity dependence on dark current of Si-pad. The other is the measurement of the response to laserinjection to the edge of the baby chips and inside pixels of the chip.
Figure 6 and 7 show the diagram and picture of the I-V measurement system. Si-pad is put in a plastic box andreadout pins touch each pixel. Signal of readout pins are collected to a copper sheet and we measured summedcurrent of Si-pad. We put the measurement box in thermohygrostat to stabilize temperature and humidity.During the measurement, we recorded temperature and humidity around Si-pad with a thermocouple and ahumidity sensor.
Si-pad BOX
Keithley 6517B Source / Am meter
Thermohygrostat Humidity Sensor PC USB current thermocouple
GPIB
Figure 6: Schematic setup of I-V measurement. Lab-VIEW is used by data taking. Figure 7: A picture of the setup of I-V measurement.
Figure 8 shows the setup of the measurement of laser injection. Wavelength of the infrared laser is 1064nm, which corresponds to 1.16 eV, slightly above the Silicon’s band-gap energy of 1.12 eV so that one laserphoton can produce an electron-hole pair. The peak power of the laser is 13 kW and we can adjust the laserpower by filters. Reputation rate is 1 kHz. Laser spot size is less than 20 µ m. Figure 9 shows a box for thelaser injection. The readout pins touch each pixel and we can measure the signal from individual pixels. We3igure 8: Picture of setup of Laser injection. Figure 9: Picture of measurement box for laser injec-tion. Length of the pins can be adjusted since the pinhas a spring in it.fix the readout pins by a thin acrylic plate, which has holes to inject laser light into pixels directly. Injectionpoint of the laser light is the gap region between the pixel edge and the guard-ring, because pixels are coveredwith aluminum electrode. We prepared small Si-pads with meshed electrodes shown in Figure 10 to studycrosstalk effect between pixels.Figure 10: Points of laser injection for the comparison of guard-ring effect (left) and for the inter-pixelcrosstalk study with the meshed chip (right). Blue dot shows the injection point. Figure 11 shows the result of temperature dependence of dark current. There are no big differences amongthe guard-ring types. Fitting function is expressed as I ( T ) = AT exp( − E g ( T )2 k B T ) (1)4here A is a constant factor, T is temperature and k B is Boltzmann constant. Silicon’s band-gap energy E g also depends on temperature which is described by Varshni’s empirical expression, E g ( T ) = E g (0) − αT β + T (2)where parameters E g (0), α and β are approximately 1.1557 [eV], 7 . × − [eV / K] and 1108 [K], respec-tively. The function E g ( T ) has been experimentally determined [4]. Current of edge region is larger thanthat of bulk region because electric field of edge of Si-pad differs from bulk region of it. The fitting function,however, does not include the edge current. Fitting result of silicon’s band-gap energy is around 1.7 eV forall types of the guard-ring. The big deviation from theoretical value of 1.1557 eV can be caused by edgecurrent which is not considered in the current fitting. This issue is under investigation.However, we have a problem for reproducibility. Figure 12 shows the result of temperature dependence ofthe 1 guard-ring type in multiple sets of measurements. If once we take out the baby chip from measurementbox and set it again, the dark current is significantly changed and the maximum difference is factor of twobetween different measurement sets. The shift can be occurred by surface resistance of copper sheet whosecondition is not good. Figure 13 shows the result that the shift is reduced by clipping same area on thecopper sheet. It also have the shift, but it is much smaller than before. Dark current of the baby chips,which is vertical axis, are normalized by size of the chips and constant term is added in fitting function. Atroom temperature region, no significant difference was seen among different guard-ring types. The band-gapenergy is improved to 1.65 eV level. At high temperature region, we can see a difference of chip size. Thisshould be also caused by the edge effect. Temperature[K]285 290 295 300 305 310 315 D a r k C u rr en t[ n A ]/ p i x e l -1 / ndf χ ± ± χ ± ± Temperature Dependence / ndf χ ± ± χ ± ± / ndf χ ± ± χ ± ± χ ± ± χ ± ± Temperature[K]285 290 295 300 305 310 315 D a r k C u rr en t[ n A ]/ p i x e l -1 / ndf χ ± ± χ ± ± Temperature Dependence / ndf χ ± ± χ ± ± / ndf χ ± ± χ ± ± χ ± ± χ ± ± Voltage 120 V, Humidity 50 % D a r k C u rr en t[ n A ] / p i x e l Temperature[K] band-gap energy
Figure 11: Temperature dependence. Horizontal axisshows temperature, and vertical axis shows dark cur-rent normalized by the number of pixels. Fitting pa-rameter of p1 means silicon’s band-gap energy.
Temperature[K]285 290 295 300 305 310 315 D a r k C u rr en t[ n A ]/ p i x e l -2 -1 / ndf χ ± ± χ ± ± Temperature Dependence / ndf χ ± ± χ ± ± χ ± ± χ ± ± χ ± ± χ ± ± D a r k C u rr en t[ n A ] / p i x e l Temperature[K]
Figure 12: Reproducibility of I-T correlation for 1guard-ring type. Other types also cannot get repro-ducibility.Figure 14 shows the result of the measurement of laser injection for 0, 1, 2 and 4 guard-ring(s) types.Percentage shown on the graphs means a slope on the graph for each pixel. For 1 guard-ring type, ring-formedcrosstalk was seen because current flows along the guard-ring. Uncertainty of the slope is less than 3% for thesame chip. On the other hand, for the other types of guard-ring samples, there are no significant crosstalks.For 2 and 4 guard-ring types, guard-rings are divided so that current cannot flow to other channels. For 0guard-ring type, current also cannot flow.Figure 15 shows the result of laser injection to the meshed electrode. Laser point is shown in Figure 10.For all channels, crosstalks are less than 0.4 %. The difference of slope can be due to the small difference ofgaps caused by a small variation at the production. We plan to compare between 0 and 1 guard-ring types.5 emperature[K]285 290 295 300 305 310 315 D a r k C u rr en t[ n A ]/ s i z e [ mm * mm ] -3 -2 -1 / ndf χ ± ± ± χ ± ± ± Temperature Dependence / ndf χ ± ± ± χ ± ± ± χ ± ± ± χ ± ± ± χ ± ± ± χ ± ± ± Voltage 120 V, Humidity 50 %
Temperature[K] band-gap energy D a r k C u rr en t[ n A ] / s i z e ( mm ) Figure 13: Improved result of temperature dependence on dark current of Si-pad ch1 ch1 ch2 ch2 ch3 -0.09 +/- 0.01 % ch3 ch4 ch4 ch5 -0.03 +/- 0.00 % ch5 ch6 -0.09 +/- 0.00 % ch6 ch7 -0.16 +/- 0.01 % ch7 ch8 -0.25 +/- 0.01 % ch8 ch9 -0.05 +/- 0.00 % ch9 ch10 -0.15 +/- 0.01 % ch10 ch11 -0.09 +/- 0.00 % ch11 ch12 -0.05 +/- 0.00 % ch12 ch1 ch1 ch2 -0.48 +/- 0.02 % ch2 ch3 -0.78 +/- 0.03 % ch3 ch4 ch4 ch5 -0.82 +/- 0.03 % ch5 ch6 -0.75 +/- 0.03 % ch6 ch7 -0.69 +/- 0.03 % ch7 ch8 -0.56 +/- 0.02 % ch8 ch9 -0.73 +/- 0.03 % ch9 ch10 -0.72 +/- 0.03 % ch10 ch11 -0.59 +/- 0.02 % ch11 ch12 -0.91 +/- 0.04 % ch12 ch1 ch1 ch2 ch2 ch3 ch3 ch4 ch4 ch5 ch5 ch6 ch6 ch7 ch7 ch8 ch8 ch9 ch9 ch1 ch1 ch2 ch2 ch3 ch3 ch4 ch4 ch5 ch5 ch6 ch6 ch7 ch7 ch8 ch8 ch9 ch9
Injection point
Figure 14: Laser injection for 0, 1, 2 and 4 guard-ring(s) types. Horizontal axis shows response of the nearestpixel to the injection point, so the axis is proportional to laser power. Vertical axis shows the response ofeach pixel. Value of the slope at each pixel varies chip by chip, but pattern of the slopes was consistent withall chips of the identical specification. 6
000 2000 3000-2002040 ch1 ch1 ch2 ch2 ch3 ch3 ch4 ch4 ch5 ch5 ch6 ch6 ch7 ch7 ch8 ch8 ch9 ch9 ch 1 ch 2 ch 3 ch 5 ch 6 ch 7 ch 8 ch 9 ch 4
Injection point
Figure 15: Result of laser injection inside pixel with a meshed chip shown in similar format with Figure 14.
We measured temperature dependence of dark current and responses of laser injection for Si-pad. Fortemperature and humidity dependence of dark current, no significant difference appeared among differentguard-ring types. For laser injection between guard-ring and edge of a pixel, crosstalk is seen at the 1 guard-ring type, but crosstalk is not seen at 0, 2 or 4 guard-ring(s). For the laser injection inside a pixel, crosstalkbetween pixels is at around 0.4 percent. In conclusion, currently we do not see any disadvantages in noguard-ring sensors. Crosstalk from the guard-ring is only seen in the guard-ring type.For the next step, we will improve the setup of I-V measurement to reject the edge effect. We also planto move to thicker sensors of for example, 500 µ m thickness to achieve better energy resolution. We have toinspect the characteristics of 500 µ m sensors to check the no guard-ring sensor does not have disadvantages.At last, we are preparing to establish measurement procedure for quality control in mass Si-pad production. Acknowledgement
We thank colleagues in LLR and CALICE-Asia groups, especially Dr. Vladislav Balagura for useful discus-sion. This work was supported by MEXT / JSPS KAKENHI Grant Numbers 23000002 and 23104007.