I. Introduction

The preliminary choice for a glue to affix the silicon microstrip detectors to the Rohacell C-cages is Araldite [1], an epoxy that has been used extensively by CERN experiments, including L3, with no reported problems. However, the detector vendor, Micron Semiconductor [2], considers Araldite to be "dirty" and recommends against its use. Because of this recommendation, we decided that it would be prudent to run a series of experiments to determine whether or not Araldite is suitable for our use, and to determine what, if any, problems we might encounter in doing so. As Micron has supplied detectors with two different passivation coatings, polyimide and Si0₂, a minimum of two experiments were required to investigate possible reactions between Araldite and the coatings.

II. Test Series 1

Upon reception of the silicon microstrip detectors, both the long outer barrel and short inner barrel versions are subjected to an acceptance battery of electrical measurements to evaluate their suitability for use in the operational MVD, as spares, or for use in tests of other subsystem components. At the time of these tests, we had several tens of detectors already tested and categorized as MVD grade (A grade), electrical test grade (B grade), mechanical test grade (C grade), or junk. For the first series of tests, B grade polyimide coated long detector 1451_11 was selected. This detector was MVD quality except that its polysilicon test resistors failed to meet our resistance specifications. Since these structures are not physically part of the detector itself, 1451_11 was an ideal choice for these glue tests.

For this test, 1451_11 was glued to a piece of Rohacell cut off of an actual C-cage milled to the latest specifications. It ended up being approximately the same length and width as the detector, but only ~2 mm thick. This cage was not coated with the Parylene environmental coating. The reliefs around the edge of the Rohacell provide room for excess glue to flow out without touching the edge or upper surface of the detector. Since it was cut from a spec. C-cage, the contact surface is unchanged from the design specifications. After gluing, the detector and Rohacell was left to cure under a cardboard box to shield against dust.

Figure 1: Top view of the microstrip detector and Rohacell test piece. Side view of the two glued together showing the cut out reliefs around the edge.

After 24 hours, the detector and Rohacell composite was mounted on a jig, wire bonded, and ready for testing. Close visual inspection revealed that despite careful efforts, the Araldite did reach the edge of the detector in the center of the top (spy pad) edge in a region ~ 6 mm wide. Also, there was a faint "bathtub ring" structure, visible only under high magnification, about 2 mm wide around the outer edge of the top surface of the detector. It is believed that since the glue curing took place in static air, this structure was formed by outgassing products of the epoxy settling onto the edge of the detector.

Figure 2: Original individual channel IV data from detector 1451_11.

Electrical testing consisted of repeating the individual channel IV measurements and comparing to the original data shown in figure 2 above. The individual channel IV data taken after gluing to the Rohacell with Araldite is shown in figure 3 below. As every measurement of every channel at every bias voltage is negative, we can safely say that this detector is no longer functional. However, the low negative currents and the five pin structure (every fifth channel shows a consistently more negative current) resembles structures seen in detectors with low interstrip resistance. Possibly the "bathtub ring" structure shorted out every strip to the bias and/or guard rings. The effects of the slight glue slop on the top edge of the detector were not determined in this test.

Figure 3: The individual channel IV data for detector 1451_11 showing negative currents for all channels at all bias voltages after gluing under static air.

Unfortunately there was no way to conclusively prove or disprove that all the leakage current was shorting through the "bathtub ring" to the bias or guard rings. It was however suggested that the "bathtub ring" be cleaned off with isopropyl alcohol and the detector retested. This had the effect of restoring the detector to near its original condition. If anything, the detector's performance was improved slightly. Figure 4 below shows the results of this retesting.

Figure 4: The individual channel IV data showing restoration of detector performance after cleaning off the "bathtub ring."

III. Test Series 2

The second series of glue tests was undertaken to determine a way of gluing the detectors without the "bathtub ring" forming, and to determine the effects of glue slopping onto the edge or even the top surface of the detector. For this series of tests, B grade polyimide coated short detector 1451_17 was selected. This detector also was MVD quality except that its polysilicon test resistors failed to meet our specifications. The initial individual channel IV data is shown in figure 5.

Figure 5: Initial individual channel IV data for detector 1451_17.

In this test series, all gluing took place on a fan bench to keep a constant flow of air over the curing glue to sweep away the outgassing products and keep them from settling on the detector in the "bathtub ring" structure. The Rohacell piece for this test was not taken directly off of a spec. C cage, but was machined to close approximation. The initial gluing came out perfectly; no glue slopped onto the detector edges, and there was no sign of a "bathtub ring." Subsequent testing showed no appreciable change in detector performance; as shown in figure 6. The minute changes in this and following tests are likely due to slight changes in humidity and temperature in the laboratory.

Figure 6: Individual channel IV data for detector 1451_17 after first gluing under moving air.

For the second glue test in this series, glue was deliberately added to the entire top edge of the detector, and a little down both sides. The top surface of the detector was left clean. To prevent the deposition of outgassing products, the detector was left to cure under moving air in the same manner as the initial gluing. The results of the subsequent individual channel IV test, again showing no appreciable affect, are shown in figure 7.

Figure 7: Individual channel IV data for detector 1451_17 after adding glue to the top edge of the detector leaving the top surface clean.

For the final test in this series, glue was deliberately added to the top surface of the detector in a worst case gluing scenario. A finger of glue was added that reached around the edge of the detector from the backplane, across both guard rings, and about 1cm into the active area of the detector crossing many microstrips. A separate isolated dot of glue ~0.5 cm in diameter was also added to the central active area of the detector. All spy and bond pads were left clean, so the glue only had physical contact with the polyimide coating on the top surface of the detector, the aluminum on the back plane, and the raw silicon edge of the detector. Again, curing took place under moving air. And again, this final test showed no appreciable affect on the detector's performance, as shown by figure 8 below.

Figure 8: Individual channel IV data for detector 1451_17 after adding glue to top surface of detector.

IV. Test Series 3

Test series three was undertaken to determine the suitability of Araldite for use with silox (SiO2 ) coated detectors (both of the detectors in the previous two test series were polyimide coated). Micron, the detector vendor, prefers the polyimide coating as it is thicker and more protective. However, our tests detected the presence of an electrical short caused by the polyimide coating, so it was nominally decided to specify silox coatings for all detectors in the MVD. This necessitated a compatibility study of Araldite with the silox coated detectors to determine whether the glue or its outgassing products react with the coating and affect detector performance.

For this test series, detector 1364_19 was selected. This was the worst of the first five silox detectors delivered to us to determine whether the coating was in fact causing the electrical shorts we detected and not some other step in the manufacturing process. None of these detectors have yet been rated as A, B, or C class, but 1364_19 was a perfectly suitable for our purposes. The initial individual channel IV data for this detector is shown in figure 9.

Figure 9: Initial individual channel IV data for detector 1364_19.

This series of glue tests was identical to test series two. All gluing took place on a fan bench to keep a constant flow of air over the curing glue to sweep away the outgassing products and keep them from settling on the detector in the familiar "bathtub ring" structure. Also the Rohacell piece for this test was not taken directly off of a spec. C-cage, but was a close approximation. Again the initial gluing came out perfectly; no glue slopped onto the detector edges, and there was no sign of a "bathtub ring." Subsequent retesting showed minor increases in the leakage current at 40 and 60 Volts; see figure 10.

Figure 10: Individual channel IV data for detector 1364_19 after initial gluing.

Next, glue was deliberately added to the entire top edge of the detector, and some down both sides. The top surface of the detector was left clean. Again, the detector was left to cure under moving air to prevent the deposition of outgassing products. The results of the subsequent individual channel IV test, shown in figure 11, show slight increases in the leakage currents of all channels at 40 and 60 Volts.

Figure 11: Individual channel IV data for detector 1364_19 after second gluing.

For the final glue test, Araldite was deliberately added to the top surface of the detector. A finger of glue was added that reached around the edge of the detector from the backplane, across both guard rings, and into the active area of the detector crossing many microstrips. A separate isolated dot of glue was also added to the central active area of the detector. All spy and bond pads were left clean, so the glue had physical contact only with the silox coating on the top surface of the detector, the aluminum on the back plane, and the raw silicon edge of the detector. Again, curing took place under moving air. Figure 12 shows that again, there were slight increases in the leakage currents at 40 and 60 Volts.

Figure 12: Individual channel IV data for detector 1364_19 after third and final gluing.

For the final test, the top surface of 1364_19 was cleaned off with isopropyl alcohol. While there were no visible depositions of outgassing products, it was thought that perhaps the slight changes in the detector performance were due to depositions too faint to show up under our microscope. Unfortunately, retesting after cleaning resulted in only negligible differences from the data taken after the last gluing; see figure 13.

Figure 13: Individual channel IV data taken after cleaning off the top surface of the detector with isopropyl alcohol.

V. Conclusions

Araldite is an acceptable choice for a glue to affix our silicon microstrip detectors to the Rohacell C-cages. The polyimide coated detectors show no change in performance at all except when allowed to cure in static air, in which case the "bathtub ring" structure appears to short out all strips. Cleaning off this structure will however restore the detector to its original condition. The Silox coated detectors do show marginal increases in leakage current in all channels at 40 and 60 Volts bias. These increases are small enough to be negligable.

These tests also tell us that certain precautions should be taken during glueing. First and foremost, curing must take place under moving air. Should outgassing products manage to condense onto the detector and adversely affect the detector's performance, cleaning the detector with isopropyl alcohol should fully restore the detector to operable condition. Second, care should be taken not to get glue on the raw edges of the detectors. While these tests indicate that this is not a catastrophic problem should it occur, it is apparent that the uncoated edge of the detector is somewhat sensitive to glue contamination.

References:

[1] Araldite Epoxy, AW106, HV953, Ciba Formulated Systems Group, Ciba-Geigy Corporation, 4917 Dawn Avenue, East Lansing, MI 48823-5691 USA

[2] Micron Semiconductor Ltd., 1 Royal Buildings, Marlborough Road, Churchill Industrial Estate, Lancing, Sussex BN158UN U.K.

Please send comments and suggestions to: