Testing an encapsulant for use in D0 smt’s HDIs
Wilhelm R. Hernández
University of Puerto Rico-Mayagüez Campus
Mayagüez, PR 00693
August 4, 1999
Dr. Cecilia Gerber
Silicon Detector Facility
This report deals with the tests performed on encapsulant 9001 version 3.1 of the Dymax Corporation for the main purpose of knowing the feasibility in its use for protecting the wire bonds in the D0 SMT HDI’s. Included are the techniques developed for removing the encapsulant, and the results obtained by the author.
As preparation for Run II of the Tevatron Collider at Fermilab, the Dzero detector is currently being upgraded in a number of ways. The Dzero Detector for Run I provided scientists around the world with a sensational detector capable of measuring the energies and tracks of the particles after particle collisions in the Tevatron ring. The Dzero detector specific’s purpose is to study p-pbar collisions. It was the instrument used to discover the top quark in 1995. For Run II of the Tevatron, luminosities in the order of 2 x 1032 cm-2s-1 will be produced and bunch-crossing times of D t m 132 ns will be encountered. In order for the Dzero Detector to be able to take advantage of the new improvements at the Tevatron it had to be upgraded. The main components of the upgraded detector are: the tracking systems, which include the Silicon Microstrip Tracker (SMT) and the Fiber Tracker, a 2 Tesla superconductiong solenoid, the Calorimeter, preshower detectors, mini-drift tubes, the Muon Detector, and the trigger control system.
A major element of the upgrade is the replacement of the inner tracking systems. The upgrade is required because of the expected radiation damage to those detectors by Run II, and also to improve the physics capabilities of the D0 detector. Included in the inner tracking systems is the SMT. Barrels and disks carrying with them silicon ladders (which will be explained later) register the track of particles that go through them.
There are 3-chip, 6-chip, and 9-chip ladders in the SMT. The chip, the SVX II, is in charge of reading the electrical pulses produced by the particles as they go through the ladders. Each chip has 128 channels for read out. Each channel is connected to one of the 128 metal strips in the silicon. The connection is made by means of wire bonds that go from one end of the chip, to one end of a strip. The pads, the place where the bonds go, are 2 mm apart, and the diameter of the wires, which are made of aluminum, is about 25 mm. The wire bonding is done with automatic bonders for the most part, that put each wire in its place with a precision of 12 mm. After all the wire bonds have been put in place, it is necessary to protect them, as they are incredibly fragile; any touching of them will destroy the bonds. One way to do this is to cover them with a substance that could hold them in place firmly. This substance, or encapsulant, has to meet certain requirements in order to be appropriate for this purpose. To evaluate if it meets the requirements, it has to be tested. The tests must include temperature cycles, and those that test the ease of taking the encapsulant off on cases where chips or wires have to be replaced. This presentation summarizes all the tests done on the encapsulant version 3.1 of the Dymax Corporation and their results. But before getting into the tests done and their results, a brief explanation is given of the silicon tracker upgrade, how the ladders work, what constitutes them, and finally, the role of the wire bonds in the detector and why they need to be encapsulated.
Description of the Silicon Microstrip Tracker (SMT)
The Silicon Microstrip Tracker (SMT) is the high-resolution section of the tracking system, and the first encountered by the particles after they collide in the Dzero Collider detector. This is part of the upgrade that will be done on the detector, which is designed to operate in the new high luminosity environment of the Main Injector at Run II of the Tevatron Collider at Fermilab. The SMT was designed to meet several goals: measure particle momentum under a solenoidal magnetic field, vertex reconstruction and radiation hardness. It is made up of six barrels, four larger diameter "H" disks and twelve small diameter "F" disks. Figure 1 shows the whole assembly of the SMT.
Figure. 1 This is how the detector will look when assembled.
Several of the Run II Collider machine parameters have an effect on the silicon design. The luminosity sets a scale for the radiation damage expected over the life of the detector, which in turn dictates the operating temperature (<10 degrees centigrade). The barrels and the disks are based on 50 m m pitch silicon microstrip detectors, 300m m thick. The way these silicon microstrip detectors work is basically simple. Each silicon sensor has a certain amount of microstrips. These microstrips run along the length of the silicon sensors. They are ac-coupled to the readout electronics by capacitors integrated directly onto each strip. A thin dielectric layer between the strip implant and the aluminum metallization forms the capacitor. When the detector is on, these micro capacitors are then charged. As a charged particle goes through the silicon sensors, it ionizes the material and electrons form. Then, these newly produced electrons will drift toward the nearest strip and produce a current pulse that is read out by one of the many chips sitting on the sensor. In this way, when a collision takes place, the microstrip detectors pick up pulses induced by a particle as it goes through them.
Figure 2. Crossection of a D0 SMT's barrel.
These pulses provide information that allows one to plot points in space with regards to the particle’s location. Then with interpolation, the trail of particles can be traced with great precision.
The job of the chip, the SVX II, is to pick up the signal and then convert it from analog to digital. Then the readout system moves the digitized data from the SVX-II chips to the Trigger Level 2 section of the Data Acquisition System located in the second floor of the fixed counting house in the D0 Assembly Building. This is how a point of the particle's track is recorded. The barrel and the F and H disks are composed of many of these sensors. Figure 2 illustrates how these sensors are arranged in a cross-section of a barrel. The beamline goes perpendicularly into the page through the center of the crossection. From here one can observe how the sensors are oriented with respect to each other. The strips in the sensors are placed at 0o, 2o and 90o angles with respect to the beam line; the alignment contributes to the 3D reconstruction of the trails.
Ladder is the name the microstrip detectors get after all the readout components have been attached. Figure 3 shows a picture of a fully assembled ladder. This is one of the three chip ladders that compose the barrels. Each barrel has 4 layers of ladders, the total number of them in all the barrels being 1068. The ladders are composed of several parts. The main parts being the silicon microstrip sensors, the High Density Interconnect (HDI), the SVX II. Also beryllium plates and carbon fiber rails are used to provide them with support. The HDI is a flex circuit that carries the SVX II chips and the passive components, and integrates a "pigtail" cable which carries signals and bias voltages to the outer radius of the detector.
Figure. 3 Top view of a fully assembled ladder.
Wire bonds on the Silicon (Si) ladders are used mostly to connect the microstrips in the Si sensors to the channels on the chips, the chips to the HDI, and the passive sensors to the active ones. For each chip we have 128 channels. Each channel connects to a single strip by means of a wire bond. The length of the wire depends on the height of loop, but it is no longer than 4 mm. After it has been bonded, it is so fragile that the slightest touch with the finger will break it. The average force that these wire bonds can sustain is about 6 grams. Therefore, it became evident for the physicist in charge of the SMT upgrade that a way of protecting them from any possible harm was needed. These harms may come from unwanted touching of the wires. To prevent this it was decided that the wire bonds should be protected using an encapsulant. Whichever encapsulant selected must comply with special requirements of texture, viscosity, thermal change resistance, etc. The remaining of this report talks about the tests to which encapsulant 9001 version 3.1 of the Dymax Corporation was subjected during the summer by the author.
In order to meet the specific requirements of protection and reliability, the encapsulant had to be chosen wisely. It must comply with certain requirements:
With the above requirements in mind, it is easy to see that not every encapsulant can fulfill the task. An older version of encapsulat was tested and selected as the encapsulant to be used when ladder production started, some time before I started to work at SiDet. Later, some problems with the encapsulant emerged. It would release toxic gases that are harmful to persons exposed to them. The ESH office prohibited the use of this encapsulant for this reason. All of this took place before my arrival to FermiLab. It was then necessary to look for another version of encapsulant. Among various alternatives, a good candidate was encapsulant version 3.1 of Dymax Corporation. The encapsulant has a viscosity of 4,500 cP. It must be refrigerated below 50o C and it has to be cured with UV Rays for at least 10 seconds after it has been applied. The encapsulant comes from the manufacturer in a special syringe. For it to be applied the syringe should be connected to a pressure pump. The person appliying the encapsulant controls the flow of it by using a pedal, which in time, regulates the air pressure. Then the UV Rays pistol is used to cure it for 10 seconds (always wear appropriate glasses when handling the UV Rays pistol). My work this summer was to test the efficiency and reliability of this encapsulant using several tests, and also to devise a way of applying it to wire bonds to be encapsulated.
Among the tests performed were temperature cycle tests, wire-pulling tests, encapsulation removal tests, and wire bond pull tests (as described in the following sections). The most important aspect of testing this encapsulant was to determine how easily a chip could be replaced, a wire bond could be pulled, and wire bonds could be replaced after an HDI has been encapsulated. The entire tests were made on four sample pieces; two of them, the gold to gold pieces, had five sets of 50 gold strips each. Each set was probed for continuity with a multimeter. Another piece was a three chip HDI with mechanical chips (mechanical, in this case and the following ones, means that their circuites are not electrically functional). The last piece was a nine chip mechanical HDI. Again, all the chips were rejects. The main purpose of these pieces was to find a way of replacing bonds already encapsulated without damaging the components, the gold pads or the chips. To actually know whether the chips get damaged after the encapsulant has been removed from the HDI pieces and wires rebonded, we would need to make tests on an electrically functioning HDI.
What follows are the distinct tests done on the encapsulant along with their results.
Tests done on encapsulant 9001 version 3.1 and their results
The first test done on the encapsulant was the thermal test. The main objective of this test was to examine how feasible the encapsulant is for wire bond protection with regard to thermal changes. The temperature cycle test provides insight on the encapsulant’s ability to resist repetitive temperature changes. As previously stated, the detector will be kept below 10oC when operational. For this test, one of the gold to gold pieces was put in a freezer, which had a maximum lower temperature of –100C. The piece went through five cycles, each one lasting 3 hours and having a temperature range from –100C to 200C . Another piece was put in another freezer. It went through three cycles, each having a temperature range from –450C to room temperature, about 300C. The detector when functional will never go below –45oC. This time each complete cycle lasted one hour. After this, all the wire bonds were tested again for continuity and they were all still good. No visual changes were found on the encapsulant. We concluded that thermal changes in the range in which the detector will be operated do not alter in any way the encapsulant’s structure and thus, the wire bonds don’t get affected.
Pulling of "hot" channels after encapsulation
The next thing we wanted to test on the encapsulant was to know how easily wire bonds could be pulled after an HDI has been encapsulated without the intention of putting them back (the wire bonds). This was done because sometimes a chip in an HDI proves to have bad channels. HDI electrical functionality tests provide the means of finding defective channels on chips, specially "hot" channels. If these channels remain "hot" all the time, they must be put out. Removing or "pulling" the wire bonds connecting these chips achives this purpose. Several techniques were carried over on all of the test pieces for this objective.
What we first did was to identify the wire bonds that were functioning on the gold to gold test pieces. This was done using a probe and a multimeter. All the test structures had a fine aluminum wire welded across the gold terminals as illustrated in figure 3. In this way it was easier to test the strips for continuity.
We started by removing one single bond using a very sharp tool like a scraper. We used one of the test structures to test this technique. The procedure was the following: we first selected one wire arbitrarily and called it T, then with the aid of a microscope and an X-acto knife we would make an incision through the encapsulant until it was possible to use tweezers to take the wire out. This was done very carefully, as we didn’t want the neighboring bonds to get damaged. We practiced the technique in another three sets. After I probe tested the sets in the test structures, I found that neighboring wire bonds got damaged indeed. On each set from three to seven wire bonds were found to be non-functioning on each side of T. A possible explanation for this could be that when the instrument was introduced on the encapsulant, the rigidity of it caused damage to surrounding bonds while the instrument was moving on the interior walls of the encapsulant. The test was repeated with other test structures but the same results were obtained. We concluded that it will not be possible to remove single wire bonds without causing damage to neighboring wires. Figure 4 shows one of the test structures after such a test.
Figure 4. Test piece after "hot" channel wire pulling test.
Removal of wire bonds after encapsulation
Sometimes it is not sufficient to have some wire bonds pulled for reasons of "hot" channels. In some instances, a full set of wires corresponding to a chip has to be removed and replaced. For this reason a method of removing only section of the encapsulant had to be devised. We started by taking all of the encapsulant off from one of the gold strip sets on a test structure. With this, our objective was to determine if the gold pads get damaged to some extent after the encapsulant is removed. This version of encapsulant peels off as a tape. Maybe this is the reason why it pulled off with it all of the wires and even two gold pads. The surface was cleaned thoroughly with ethyl alcohol, using a lint free Q-tip. After this, the pads were rebonded, except for the two gold pads that were ripped off, and tested for continuity using a multimeter and a probe. All of the rebonded wires were working properly. This test showed that the encapsulant could be strong enough to lift gold pads if we try to remove it without being extremely careful.
We also tried this on mechanical HDIs. We prepared these HDI’s as if they were good ones in the sense that chips were put on them and wire bonded onto the HDI’s. They were then encapsulated. Again, some of the gold pads were lifted up when the encapsulant was removed. In one of the HDI’s, the one corresponding to a three chip ladder, we removed the whole encapsulant as we did with the test structure. The encapsulant was covering all of the wires connecting the chips to the HDI. Three gold pads out of the 26 that connect the HDI to the SVX II were lifted up.
Another approach we tried was "sectioning", which is, taking off only a section of encapsulant. We used test structures and HDIs to try this technique. Figure 5 shows one of the test structures after being sectioned.
The technique consisted on selecting a range of wires to be replaced. For instance, in one of the test structures we selected wires 15-40 (each set on these test pieces has 50 wires) to be removed. To do this we cut very carefully the encapsulant using an X-acto knife at the boundaries of this range. Then, using scrapers and tweezers we removed the section. The problem with this approach was that even thought we try it a few times, bonds close to the boundaries of the section would always get damaged. In the case of the test structure I mentioned a total of 12 bonds were destroyed after doing this. Also, if we wanted to rebond, we needed to use the manual bonder to get to the pads close to the edge of the encapsulant because the automatic bonder wouldn’t reach to them as the encapsulant was obstructing the way.
The pads of all the test pieces were cleaned using ethyl alcohol. The cleaning process is very important as the bond pads need to be cleaned as good as possible in order for the bonds to stick to the pads. The HDI was then rebonded but only to know the rebondability of the pads - if it is possible to rebond wires onto the pads. All of the pads (only 26 are used for bonding) were bonded. Even those that were lifted from the end were bonded. This was possible because the technician that did the bonding made the length of the wires longer in order to reach the part of the gold pads that was still adhered to the HDI. Nevertheless, this technique is not recommended due to the inherent risk of damage to the pads, both on the HDI’s and on the chips.
For this reason, the manufacturing company of the encapsulant was contacted with the intention to obtain information regarding methods of getting the encapsulant off without damaging the gold pads. The information we obtained was that isopropyl alcohol (IPA) would help degrade the bond lines. This means that it helps loosing it up by getting between the encapsulant and the material it is encapsulating.
Figure 6. Encapsulated 3 chip HDI.
Testing with IPA
All the tests done previously were repeated but this time we used the IPA. We used comercial grade IPA. The IPA, as previously stated, degrades the bond lines. To illustrate what this means, imagine a piece of tape that has been put, let’s say, onto a piece of plastic. Now, if the piece of plastic with the tape on it is placed under water for a couple of minutes and then one tries to remove the tape, one notices that it comes off easier than it would have without putting it under water. This is what seems to happen to the encapsulant after it is put under IPA for a couple of hours.
To test how effective the IPA was in performing its function, we took one of the three chip mechanical HDI already encapsulated –this will never be done with an electrically working HDI- and placed it on a petri dish filled with IPA. Then we put it in the ultrasonic machine with the intention of maximizing the penetration of the IPA into the space between the encapsulant and the gold pads. The same thing was done with another test piece but this time without using the ultrasonic machine. They had the alcohol on for an hour. Then, tweezers were used to unstick the encapsulant. Using IPA did make a difference; the encapsulant was taken off with much less struggle than it would have without using the alcohol. There was not much difference in using the ultrasonic machine. The gold pads were intact.
For later tests, we used a very small amount of alcohol to avoid damping the other components in the HDI, and we would use a lint free Q-tip to apply it (it has to be lint free in order to leave no residue). In an electrically working HDI some components might get damaged if IPA gets on them, for this reason, it is imperative to avoid spreading the alcohol on the other components.
The amount of time left under the encapsulant is important. For it to be effective, the encapsulated section in the HDI should damped with IPA for at least 20 minutes.
Removal of chips after encapsulation
Another part of the tests consisted in removing chips from HDI’s that were encapsulated. The HDIs, as well as the chips, were mechanical. The first thing done was damping the encapsulant with IPA using a lint free Q-tip. Then the technique of sectioning was used to cut the encapsulant corresponding to one face of a chip in an HDI. To gets chips off a soldering iron has to be used in order to heat up the epoxy that glues the chips to the HDI. The chip came off easily. The next step was to put another chip back to know how easily was to replace a chip in an encapsulated HDI. Although the technician was able to put back another chip, she had problems in aligning it at the proper place. The reason was that some pieces of residue of the encapsulant were obstructing the way and don’t letting the chip get in the right place.
The last technique that was performed over the encapsulant had to do with heat. The idea was to apply a little bit of hot air to the encapsulant to see how that might help in removing it. What we did was take a soldering iron and use the hot air coming from the tip to heat up the top off the encapsulant but always having precaution not to heat the other components. We used the instrument at a temperature of 400oC. I suggest not using a temperature higher than this because the other components might get affected. The hot air made it possible for the X-acto knife to go through the encapsulant and take out with much less struggle than using the IPA alone. We ran five tests on one of the test structures using the method of sectioning, using the heat this time. We found out that it fewer neighboring bonds were destroyed, in fact, in the last set none of the neighboring bonds were damaged.
Pull test of rebonded wires
It was necessary to pull test the wires in all the test pieces after they were rebonded. The wire-bond pull test is the most universally accepted method used for controlling the quality of the wire bonding operation. This test is used to evaluate the strength of wire bonds. It is performed using a machine, called simply the pull-tester, which uses a very fine hook to pull the wires at the highest in point the loop. It records the amount of force it had to use to detach the wire from its place. Usually, a certain lower limit for the force is established. If the wire bonds do not comply with this requirement, they are catalogue as not apt for the job. The force limit for the wire bonds in an HDI is about 8 grams. This is for HDIs that have just been bonded for the first time but for HDIs that had removed some of the encapsulant and rebonded the limit can be lowered to 6 grams. We ran the pull test over the rebonded wires in the mechanical HDIs and found that they were barely complying with the minimum requirement. The average force found in the rebonded wires was 6.5 grams.
The tests performed over encapsulant 9001 version 3.1 of the Dymax company gave us insight about how well the encapsulant could protect the wire bonds in the HDIs. It was known also which techniques could be used to remove it in the event that chip or wire bonds have to be replaced. The technique found to work the best in doing this job was to use IPA and hot air from a soldering iron together in such method called sectioning. This encapsulant proved to be non-toxic, sufficiently protective, thermal change resistant, and flexible. It doesn’t leave excess residue either. Thus, we can say that it meets the requirements stated at the beginning of this report and can be recommended as the encapsulant for use in protecting the wire bonds at D0 SMT’s HDIs.
First of all, I would like to thank the staff of SIST, especially Dianne Engram, Prof. Davenport and Elliot McCrory for giving me this magnificent opportunity, and my mentors who assisted me this whole summer in everything I needed. This has been one of the best experiences of my college years. Also, I would like to thank the people at the Silicon Detector Facility (SiDet) for making possible this venture, the technicians (you’re the best Jill), Dr. Eric Kajfasz, my alternate supervisor, and all the people there who aided me in doing my work and those who didn’t. And finally, last but not least, my supervisor, Dr. Cecilia E. Gerber, for her support throughout my whole work at Fermi.
D0 Silicon Tracker Technical Design Report-D0 Upgrade Collaboration, 1994
The D0 Upgrade: The Detector and Its Physics-The D0 Collaboration, 1996
G. Lutz, A. S. Schwarz. Silicon Devices for Charged-Particle Track and Vertex Detection, 1995
Scott Snyder. The D0 Detector, 1995