Charge Loss Between Contacts Of CdZnTe Pixel Detectors

Charge Loss Between Contacts Of CdZnTe Pixel Detectors A. E. Bolotnikov 1, W. R. Cook, F. A. Harrison, A.-S. Wong, S. M. Schindler, A. C. Eichelberger Space Radiation Laboratory, California Institute of Technology Pasadena, CA 91125 Abstract The surface of Cd 1-x Zn x Te (CZT) material has high resistivity but is not a perfect dielectric. Even a small surface conductivity can affect the electric field distribution, and therefore, the charge collection efficiency of a CZT pixel detector. The paper describes studies of this phenomenon for several contact configurations made on a single CZT detector. We have determined the maximum inter-contact separation at which the surface inter-pixel charge loss can be neglected. 1. Introduction A great deal of experimental work has been undertaken in order to study the physics and properties of semiconductor Cd 1-x Zn x Te as a detecting medium for X-rays and low-energy gamma-rays. The current status of CZT detector development can be found in James et al. [1], which also contains detailed consideration of the relevant semiconductor physics. The measurements described in this paper deal with the effect of enhanced surface conductivity on the field line distribution and charge collection in CZT pixel detectors. Although surface effects are known to play an important role in Si and Ge devices, similar effects have not yet been considered for CZT detectors. Specifically, the existence of surface states as well as a space charge layer just below the semiconductor-oxide interface in Si and Ge semiconductors have been known for some time (see for instance Ref. [2]). The conductivity of this layer is different from the bulk conductivity (typically it is significantly higher), thus if such a layer also exists in CZT material, it will affect the electric field distribution inside CZT detectors. The basic design requirement for pixel detectors is that all field lines, starting on the planar contact (cathode) of the detector, end on pixel contacts on the opposite side (anode) independent of the fraction of the surface covered by the pixel contacts. As a result, all electrons produced by an incident photon will be collected by the pixel contacts with 100% efficiency. This idealistic situation is possible if the semiconductor surface were a good dielectric. However, if the conductivity of the surface, or more correctly, the conductivity of the thin layer of the material near the surface is different from the bulk, this will change the field distribution inside. In this case, some fraction of the lines which would otherwise terminate on pixel contacts instead intersect the surface between the pixels. Signal electrons moving along the field lines may become trapped in the surface layer. It is clear that if the charge indeed gets trapped at the surface, the fraction of trapped charge should depend on the contact size and surface preparation. To avoid this effect the contacts should be sufficiently large. On the other hand, this contradicts the requirement that contacts must be small to minimize spectral tailing due to signals induced by uncollected holes [3], as well as to minimize input capacitance to the readout electronics. One of the goals of this work is to determine the maximum contact separation at which the charge loss at the surface between pixel contacts can be neglected. 2. Experimental setup A 10x10x2 mm 3 CZT detector with nine different groups of gold contacts on one side and a single monolithic contact on the other side was used in these 1 Corresponding author. E-mail: bolotnik@srl.caltech.edu

second stage amplifier with a 1 µs shaping time. During the measurements, four central pixels from each of the patterns were read out, and all the surrounding pixels were kept at ground potential. It should be mentioned here, there was an additional grounded plate (shielding plate) on the surface of the fan-out just underneath of the contact side of the detector. Fig. 1 Pixel contact patterns used in these measurements. All the pixels have the same dimensions of 500x500 µm, while the contact sizes vary between 150 and 450 µm. measurements. Each group consisted of a 4x4 array of identical contacts evenly spaced with a 500 µm grid pitch and enclosed in a guard ring (Fig. 1). The contact An X-ray beam from a Philips X-Ray Generator 2 was collimated to a 50 µm spot size on the detector plane and used to create ionization events near the cathode inside the CZT detector. We employed the asymmetric Bragg scattering technique to select the K α doublet (~17.4 kev) from the continuum spectrum, generated by a Mo X-ray tube, and simultaneously to compress the scattered beam in one direction, allowing us to maintain a high intensity flux. We used two linear actuators controlled by a computer to position the detector with respect to the X-ray beam. The detector response was measured over a two-dimensional geometric grid of beam locations with a 50 µm pitch. For each triggered event, i. e. when any among the four signals was above the threshold level, we recorded the pulse-heights from the four central pixels of each Fig. 2 Typical pulse-height spectra obtained when the 17.4 kev X-ray beam was scanned across the pixel: (a) the X-ray beam is pointed to the center of the pixel, (b) the X-ray is pointed between the pixels, (c) the X-ray beam is close to the edge of the pixel. The pixel contact size is 400 µm. sizes vary from group to group over the range 100 to 450 µm. The CZT crystal was coupled to the readout electronics using indium bump bonding technology. Each electronic channel included a charge-sensitive preamplifier with a 0.5 ms decay time constant, and a group. These data were used to evaluate the pulseheight spectra and correlation of pulse-heights from adjacent pixels. 2 Model XRG-3100, Philips Electronic Instruments, Inc.

3. Results and discussion Fig. 2 (a) shows the typical pulse-height spectrum collected for a single pixel when the X-ray beam was pointed at the center of the pixel. It can be seen that the FWHM of the 17.4 kev line (left peak in the spectrum) is determined by the electronic noise (pulser peak on the right side) with a FWHM of ~2 kev in this case. The changes in the spectra, as the X-ray beam was scanned across the pixel, are shown in Fig. 2 (b,c). The spectrum in Fig. 2 (b) is the superposition of the 17.4 kev line (all the electrons are collected on one contact) and continuum, representing charge sharing between two contacts. The spectrum in Fig 2 (c) was obtained when the X-ray beam was moved closer to the adjacent pixel. Fig. 3 shows the variation of the peak position on pulse-height spectra measured from adjacent pixels as the X-ray beam was scanned from one pixel to the other. As is seen, signal is shared when the beam is between the pixel contacts. Fig. 3 Variation of the 17.4 kev peak position on pulse-height spectra measured from adjacent pixels as the X-ray beam was scanned from one pixel to the other. The pixel contact size is 400 µm. Charge splitting between pixels is due to diffusion of the electron cloud produced by an ionization event. It takes about 200 ns for the charge produced by an X-ray absorbed near the cathode to travel across the crystal to the pixel contacts. Taking the diffusion coefficient to be 25 cm 2 /s (for a rough estimate we use the Einstein relationship to calculate the diffusion coefficient) one can find that during this time, the electron cloud increases in size up to several tens of microns. As a result, for interactions occurring above pixel boundaries, electrons from the cloud will be swept by the electric field toward more than one pixel. In addition, some electrons may travel close enough to the surface (in the space-charge layer) to become trapped for a time that is long compared to the amplifier shaping time (1 µs). In order to prove that the latter effect did indeed take place we plot the correlation between the signals read out from two adjacent pixels, shown in Fig. 4. Here the signal amplitude from one pixel is plotted versus the amplitude for the adjacent one as the X-ray beam was scanned across the inter-pixel gap. Separate plots are shown for 200, 350, 400, and 450 µm contacts. The points on this plot should lie on a straight line, unless there is signal loss that will result in curvature. For the 450 µm contact case (50 µm gap between contacts) charge splitting occurs between pixels with no signal loss. All events fall along a straight line connecting the groups of events for which the signal is collected on one or the other pixel. For the case of 400 and 350 µm contact sizes (100 and 150 µm gap), signal loss is apparent. For the smaller contact sizes signal loss becomes more significant. For a contact size of 200 µm or less, the dead regions between pixels where signals are entirely lost are clearly seen. It should be mentioned that our results for the 450 µm contact size are consistent with earlier measurements reported in Ref. [4], which indicated charge splitting occurred with no signal loss for X-rays absorbed in a 50 µm gap between the strips. Likewise, our results indicating charge loss for larger intercontact separation are in agreement with those obtained in Ref. [5]. In order to model the distribution of electric field lines inside the CZT pixel detector, we used the approach developed in Ref [6]. An infinitesimal layer with enhanced conductivity was assumed at the surface of CZT, and a simple numerical algorithm was applied to solve the 3-dimentional Dirichlet s problem with the condition of current continuity at the bulk-layer interface. The free parameters used in the model were surface and bulk resistivities. This algorithm also allowed us to calculate (taking the surface and bulk resistivities to be known parameters) currents flowing through the detector when specific voltages are applied to the contacts, or specific resistances are assumed between different contacts, and to make comparison with measured current or resistance values. One can also try to solve the inverse problem, i.e. estimate bulk and surface resistivities based on measurements of currents. In order to distinguish effects of bulk and surface resistivities, we chose two

Fig. 4 Correlation between the amplitudes of the signals from two adjacent pixels as the X-ray beam was scanned across the inter-pixel gap (the amplitude of the signal from one pixel is plotted versus the amplitude from the other). The contact sizes in microns: 450 (upper left), 400 (upper right), 350 (lower left), and 150 (lower right).

measured between the central pixel and the surrounding contacts. As an example, for the 200 µm contact size, we measured the total leakage current in the first experiment and the total resistance in the second experiment to be 36 na and 190 Gohm, respectively. Fig. 5 Two sets (solid and dash lines) of the surface and bulk resistivities that gave correct values for the current measured in two experiments (see text). The set satisfying both the measurements (intersection of the two lines) gives estimates for the bulk and surface resistivities of the CZT sample used in this measurements: ~2x10 11 Ω cm and ~2x10 12 Ω/square, respectively. particular setups. For both the experiments, the sample was placed in a probe station and five probes were positioned on the pixel contacts and one on the guard ring. In one experiment, the total current was measured through the central pixel while all the other pixel contacts and guard ring were grounded, and 300 V was applied to the cathode. In the second experiment, the cathode was grounded, and the total resistance was By varying the bulk and surface resistivities as free parameters, we found the sets of surface and bulk resistivities that give correct values for the measured current and resistance. The sets found from modeling the first and second experiments are represented in Fig. 5 by the solid and dash lines, respectively. As is seen, in the first experiment the total current is not sensitive to variation of the surface resistivity around the intersection point, likewise the resistance in the second experiment is not sensitive to the bulk conductivity. The set satisfying both measurements (intersection of the two lines) gives the estimates of ~2x10 11 Ω cm and ~2x10 12 Ω/square for the bulk and surface resistivities, respectively. Fig. 6 (b) shows the field line distribution calculated for a 200 µm contact size, employing the values of the bulk and surface resistivities listed above. For comparison, the distribution calculated for a nonconducting surface is also shown in Fig. 6 (a). It should be pointed out that the above estimates are based on the field line distribution generated by our algorithm, which, of course, is not precise, and should be considered as order of magnitude estimates. The main purposes of this calculation is to show that the volume resistivity inside the thin layer at the surface of CZT detectors is significantly less than bulk resistivity. To illustrate this, one should take into account that the thickness of space-charge layers in semiconductors can Fig. 6 The field line distributions calculated for a 200 µm contact size: (a) the surface is ideal dielectric; (b) the surface is slightly conducting (bulk and surface conductance were assumed to be 2x10 11 Ω cm and 2x10 12 Ω/square respectively). For comparison, the pattern calculated for the same values of bulk and surface conductance as above but for a 450 µm contact size is shown in (c).

exceed several hundreds angstroms [2]. Even assuming an exaggerated value of 1 µm, one can easily see that the volume resistivity inside the space-charge layer at the surface of the CZT detector should be 3 orders of magnitude less than the bulk resistivity. Fig. 6 (c) shows the result of similar calculations as in Fig 6 (a,b) but for a 450 µm contact size, again using the values of the bulk and surface resistivities listed above. It can be seen, that in spite of the fact that field lines enter the surface between the contacts, the charge trapping effect was not observed in our measurements. This suggests that signal electrons are still capable of drifting some distance along the surface before being trapped. However, to make a quantitative analysis of this effect, more accurate measurements are needed. 4. Conclusion A CZT pixel detector with various contact sizes has been tested in order to determine the maximum contact separation at which signal loss due to charge trapping at the surface between pixels can be avoided or neglected. No signal loss was observed for a contact size of 450 µm and a 50 µm distance between the contact edges. Thus, simply by adding the signals from neighboring pixels the total charge produced by an incident X-ray can be found. For smaller contacts, 400 and 350 µm, corresponding to larger gaps between contacts, 100 and 150 µm, respectively, signal loss becomes apparent for X-rays interacting above the inter-pixel area; however, the correlation between pixel signal amplitudes could be used to correct for the signal loss. For increasing inter-pixel gap (>300 µm), signal loss increases in a non-correlated manner, leading to dead regions between pixel. Acknowledgements The authors wish to thank Dr. Carl Stahle from GSFC, J.A. Burnham and B. Matthews from CIT, and Dr. J. Liu and M. Fitzsimmons from JPL for helping to fabricate CZT pixel detectors; Dr. Finn E. Christensen from DSRI for his contribution in the design of the X- ray calibration system, and LLNL for loan of the X-ray generator. References 1. R.B. James, B.Brunett, J.Heffelfinger et al., J. Electron. Mater. 27 (1998) 788-799. 2. In Semiconductor surface physics, edited by R.H. Kingston, University of Pennsylvania Press, 1957. 3. H.H.Barrett, J.D.Eskin, and H.B.Barber, Phys. Rev. Lett. 75 (1995) 156-159. 4. P. Kurczynski, J.F. Krizmanic, C.M. Stahle et al., IEEE Trans. Nucl. Sci. 44 (1997) 1011-1015. 5. P.R. Bennett, K.S. Shah, L.J. Cirignano et all., IEEE Trans. Nucl. Sci., 45 (1998) 417-420. 6. J.J. Florent, J. Gaudaen, L. Ropelewski and F. Sauli, Nucl. Instr. And Meth. A329 (1993) 125-132.