Abstract: Electrical characteristics of the baby detectors (extra detectors etched on the same wafers as the GLAST detectors) were measured under different conditions. The effect of voltage applied to the guard ring around the detector array was investigated. Finally, the detectors were checked again after being diced off the main wafer section they were part of.
This summer, under the guidance of Professor Hartmut Sadrozinski, and working with graduate student Grant Winker and fellow RET teacher Lisa Aguerria, I assisted in characterizing the parameters of baby silicon strip detectors. These detectors are smaller versions of those manufactured for GLAST by the Hamamatsu company. They are etched onto the same wafers as the main detectors but are placed along 2 chords on opposite sides of the disk, with the main detector being centered between them (Fig. 1). There are approximately 50 of these "baby" detectors, each having 32 actual strips detectors (SSD's) for registering the passage of ionizing radiation. Since these small detector arrays have no direct function in GLAST they are free to be used for other purposes. The plan is to use them to construct a small GLAST ladder stack to be tested under thermal expansion, radiation exposure, and vibration conditions.
As with the GLAST strip detectors, the baby detectors share the feature that the array of parallel strips diodes tied together electrically through parallel resistors to a bias ring. This in turn is surrounded by a closed ring of P+ type doped semiconductor, approximately 20 microns in width and overlain by a trace of aluminum, called the guard ring (Fig. 2). A bias voltage of about 100V would normally be used to operate the detectors, this being applied across the thickness of the silicon wafer. The entire back plane of the wafer comprises the positive electrode. Since the silicon matrix is N doped, the P+ strips on the top surface form linear diodes, the applied voltage reverse biases these diodes. It is only the passage of ionizing radiation that produces free electrons and holes in the bulk of the silicon to allow a pulse of electric current (32,000 electron / hole pairs per event ref. 2) to flow through the silicon bulk to be collected at the electrodes as a current signal. Absent one of these signal producing events, there is a slight leakage current (typically about 12 to 18 nA) which manages to flow through the reverse biased strip detector diodes. Our team investigated the magnitude of this leakage current as a function of the bias voltage applied between the back plane (+) and bias ring (-) on the top of the wafer. Additionally we measured the leakage current entering the backplane and that entering the bias ring (through the strips of the detector) with and without a defined voltage applied to the guard ring.
The issue of leakage current into the backplane and out of the bias ring on the top of the chip is such that there is a transitory imbalance in these two when voltage is first applied. The difference (about 1 nano amp) disappears after a few seconds. It is the result of internal currents inside the chip as electrons and holes are redistributed inside the semiconductor in creating the depletion zone (region inside the bulk of the semiconductor wafer which is devoid of free charge carriers). This is the process whereby the applied voltage sweeps the electrons from the N- doped bulk into the positive holes of the dopant atoms in the P+ type semiconductor thereby establishing an internal electric field opposing the external field. When equilibrium is established there is a continuous leakage current through the detector of approximately 12 to 18 nA with an applied voltage of lOOV. It was found this leakage current increased by several times if the applied voltage stepping process was repeated more than once.
The testing process involved using the HP 4145 Semiconductor Parameter Analyzer with probes set at the detector backplane and the other at the contact pad of the bias ring. The bias ring probe was held at a constant -1OOV while the backplane voltage was stepped from -1OOV to +1OOV in 2 volt increments over a time period of about 1 minute. Current vs. Voltage (I/V) curves were produced, the output consisted of the backplane voltage level graphed against the leakage currents at both the backplane and at the bias ring.
Figure 3 shows a typical output graph for one of the detectors, the relevant curves being Ib for backplane leakage and It for the bias ring (top) leakage current. A second test was done on each of the baby detectors, this was the same as that just described with an additional probe being placed on the guard ring. This extra probe was held at the same potential as the bias ring (lOOV) relative to the backplane. The outputs measured as a function of the backplane voltage were l)the backplane leakage current (Ibg), 2)the bias ring leakage current (Itg), and 3) the current leaking through the guard ring (Ig). At equilibrium item 1) should equal the sum of 2) and 3) which it did. This data was graphed using Excel after exporting it from the HP analyzer; a typical result is also present in Fig. 3.
The significant difference between the two setups (with and without an applied voltage to the guard ring) was that with the guard ring held at -1OOV, the total leakage current was approximately 17 nA at a lOOV potential. When the guard ring voltage was allowed to float, the total leakage current was typically only about 14 nA. When the guard ring was grounded along with the bias ring, this essentially created a 2nd negative electrode on the top of the detector at a larger voltage difference than it would be if left to float. This in turn allowed more current to flow from the backplane. The guard ring's purpose is as a passive structure, to establish an electric field across its P/N junction with the bulk silicon substrate it shares with the detector strips. Since the guard ring is situated near the edge of the strip detector, it serves to mitigate the non-uniform nature of the electric field at the perimeter. In other words it adds its field to gracefully shape the larger field at the edge so the strip detectors near there are not affected by field edge effects.
Attempts to measure the voltage on the guard ring when it was allowed to float proved difficult. The voltage probe from the HP instrument proved to have too low impedance to allow the rest of the chip to behave normally when it was connected. I had the best results using the Keithly electrometer voltage probe. The readings showed the guard ring's voltage to be approximately 55V when the overall voltage (backplane to bias ring) was at lOOV. Its voltage climbed to a maximum of 76V when the total voltage was 200V. The effect on the I-V curve from the back plane to the bias ring was to raise the total leakage current to around 40 nA when the guard ring's voltage was being measured. The guard ring reflected its proximity to the bias ring since its floating voltage was closer to that of the bias ring than that of the backplane.
Of the 12 baby detectors initially tested we found one which was electrically defective (shorted). The IV curves for the rest were consistent in rising to a leakage current level of 12 to 18 nA when the applied voltage was lOOV. This value could rise to over 40 nA if the detectors had voltage applied repeatedly. The maximum initial leakage current when the applied bias voltage was 200v was around 25 nA. It was found that grounding the guard ring in addition to the bias ring would increase the total leakage current by about 15%. We tested several of the baby detectors after they were diced from the surrounding chord section of silicon wafer they were originally part of. This only involved cutting one edge of the detector array. Their I-V curves (Fig. 4) were basically unchanged in nature although had a slightly higher level of leakage current.
l)Andreas Schwarz; Silicon Strip Detectors at ZO Factories. 1992 Proceedings of B Factories BFAC92-055
2)Alan Litke & Andreas Schwarz; The Silicon Microstrip Detector. Scientific American (5/95)
Another project which I was initially involved in this summer was that of Professor Bruce Schumm. Although I didn't contribute to the final result I did keep abreast of its progress and am including a summary of the ideas behind it. At the initial presentation of projects this summer, Professor Bruce Schumm told us about his work developing simulations for the NLC (Next Linear Collider). He mentioned the fact that as the energy of particle collisions increases, that the strength of the electromagnetic force and the strength of the strong force converge. In High School physics classes we have only taught that the strong force is much stronger that the electromagnetic (since it is able to predominate inside the atomic nucleus) but that it works only over very short distances. I had never encountered a rational for this. My understanding from Dr. Schumm of the reality of the energy dependent coupling constants of these two forces is that the actual force field surrounding an electric charge in space is much stronger than is normally experienced by some other charge. However, an electric field should actually be thought of as a region in which virtual charge carriers (photons in the case of an electrically charged object) are being continuously created and annihilated. These bosons act to add to the field of their parent particle. In the case of the virtual photons, these can pair produce to make electron/positron pairs. The electrons and positrons in turn tend to polarize in their positions around the main charged body at the center so as to shield (decrease) the field of the main central charge. By colliding electrically charged particles with increasing energy (which correlates to smaller wavelength) allows them to penetrate each other's fields more deeply as well as at a more specific point (due to the smaller resulting size). This serves to move the probe particle into regions inside the virtual boson sea, past the area of shielded electric field to the area close to the field of the naked particle. Consequently, these high energy collisions show the electric field to be stronger than that experienced in lower energy collisions. The strong force shares in this characteristic of having the magnitude of its field near strong charged particles changed by the addition of fields associated with the virtual bosons it produces. Unlike the photons that surround an electrically charged particle, particles with color charge (quarks) have force fields composed of gluons. Gluons themselves carry color charge (unlike the neutral photons) so they contribute to the field of the central particle by adding to its strength.
This acts as a sort of negative shielding (anti-shielding), making the overall strong force field experienced by any other external color charged particle much stronger than that of the central particle alone. Again, colliding color charged particles together at higher energies allows them to penetrate more deeply into each other's fields, bypassing the shielding effect of the surrounding bosons.
With color charge, the field coupling constant gets progressively weaker with higher collision energy. With electromagnetic charge, it gets stronger due to bypassing the intervening virtual photons. At a certain level of collision energy, the coupling constants of the two forces converge.
Inherent in this explanation is the idea that the strength of a field is not defined until an event (interaction with another charged body) occurs. The strength of the interaction is determined by the kinetic energy possessed by the bodies and the resulting momentum change.
The work of undergraduates Alex Merchant and Andrew Truitt was to write changes to the Java jetfinder program which would allow the detection of simulated collisions resulting in 3 jet events (specifically 2 up, down, strange, charm or bottom quarks and a gluon). This in turn was a function of finding groups of particle tracks which added to make jets of different specified percentages of the total collision energy (referred to as a "Y" cut). These 3 jet collision results themselves form a percentage of the total number of results. As they are interpreted to be caused by strong interactions resulting from the original electron/positron collision, this percentage reflects the strength of the strong force coupling constant at a given energy.
Alex presented the results of the cuts needed to discriminate the desired strong force interactions from the background data at a conference at SLAC this August. These simulation results and their analysis provide direction in the design of hardware and experiments in the NLC.
Back to RET Home Page