Lab 2: PWELL Lithography and Diffusion

Transcription

1 Brandon Baxter, Robert Buckley Lab Instructor: Liang Zhang Course Instructor: Gary Tuttle EE Due: February 23, 2017 Lab 2: PWELL Lithography and Diffusion 1. Overview The purpose of this lab was to create the p-well region of our CyMOS device. This p-well is a lightly doped region of the device in the n-doped silicon wafer which can act as the channel for the NMOS part of the device. To create this p-well, we first performed photolithography to etch a region of the oxide layer where we wanted the dopant to diffuse to the silicon. Photolithography is basically the process of cutting with light. In this case, we cut holes in the oxide layer. First we covered it with photoresist, which is a material that is sensitive to ultraviolet light exposure. Since the photoresist was a positive photoresist, the material that was exposed to the UV light would be removed during the subsequent etching process, thus exposing the underlying oxide. Additionally, the non-exposed photoresist material would remain on the wafer, protecting the oxide layer beneath it. To control which parts of the photoresist material on our silicon were exposed to the UV light, we used a chrome-coated mask. This allowed us to expose only the parts of the photoresist we wanted to remove. So after stripping the unwanted photoresist, the exposed oxide was etched off completely. The mask we were using was specifically designed for this p-well formation process, so we ultimately etched the oxide in the regions where the p-wells would be created. After exposing the silicon for the p-well, we had to deposit some boron, which is a type of p-dopant. For this deposition process, we used boron nitride BN-975 source wafers made by Saint-Gobain Ceramics. The deposition created a thin, concentrated p-doped region in the very top surface layer of our silicon wafer. After depositing these boron atoms, we had to increase the depth of this p-doped region to create the p-well. This would also decrease the concentration of the boron atoms as they spread out further into the silicon wafer, which is important for a lightly doped p-well. To accomplish this, first we had to remove the boron skin by oxidizing it. Then we had to allow these atoms to diffuse further into the silicon using high temperatures. At the beginning of this step, we can pump in some extra oxygen in the form of water vapor, which will reform the oxide layer above this p-well. Using this method, we can complete two steps in one high-temperature processing stage. 2. Photolithography For the photolithography process, we did not have to start with a standard clean procedure beforehand. In fact, we were specifically asked not to start with this step because it can make it harder for the photoresist to stick to the wafer. So the first step was to warm up the mask aligner machine by turning on the main console of the aligner and the power supply used by the ultraviolet lamp. The ultraviolet bulb is used to shine a constant intensity of UV radiation to the photoresist layer on our wafers. The light from this bulb shines through the apertures of the mask, which we placed in the mask holder of the machine. We made sure the mask covered the vacuum ring so that it would be properly secured in place when the vacuum pump was turned on. The UV bulb needs to draw about 1

2 275 watts of power, so it takes about 15 minutes to fully warm up. While we were waiting, we started putting the photoresist on our wafers. To spread the photoresist layer on our wafers, we used a spinner. The spinner had a rotating shaft, connected to a vacuum pump. On the shaft, we placed a 2-¾ inch diameter chuck, which provided a flat surface to place our wafer on top of. The chuck had holes in it that allow the vacuum to provide suction and hold our wafer in place. The spinner was set to rotate with the following settings: Rotation Speed: Rotation Duration: 4000 rpm 25 s The purpose of using the spinner machine is to spread a small amount of liquid across a surface, in order to create a layer that is relatively even in thickness. For each wafer, we applied 4 drops of HMDS (hexamethyldisilazane) to the center of our wafer, then we ran the spinner by pressing the pedal. After that, we applied half a pipette of photoresist to the center of the wafer. Before running the spinner, we removed any photoresist bubbles we saw on the wafer surface by sucking the bubble back into the pipette. This process was repeated for each of our device wafers. After doing this procedure, our wafers had the following cross-section profile: When handling the wafers to place them on the spinner chuck or remove them from the spinner, we used metal tweezers. This was acceptable because we did not perform the standard clean, and we were not doing high temperature processing of our wafers. As a result, we were not too concerned with these metal contaminants coming into contact with our wafers. After repeating this process and getting the photoresist spun onto all of our wafers, we had to pre-bake this layer. So we set our oven to operate with the following settings: Oven Temperature: 85 C Time Pre-Baking: 31 min 2

3 Once the wafers were in the oven, we started cleaning the spinner bowl by rinsing it with acetone to dissolve the extra photoresist. We let the acetone-photoresist liquid drip down into the acetone-photoresist disposal bottle, since we cannot let these chemicals go down the regular drain. Finally, we wiped down the spinner chuck with a few drops of acetone on a cleanroom wiper. After the wafers finished pre-baking, we waited for them to cool down. Then, we had to expose the photoresist with UV light while covering it with the mask pattern for the p-well process. This was a necessary step to remove the photoresist material above the future p-wells. First, we turned on the microscope light and selected the Soft Contact mode on the machine. Using only one objective on the machine, we focused on the features of the mask. Loading a wafer onto the wafer tray of the machine, we paid close attention to how we placed the wafer, for easier alignment in future lithography processes. For our wafers, we placed the flat edge of the wafer toward us and parallel to the edge of the wafer tray. Closing the tray, we brought our wafer close to the mask by rotating the lever on the left side of the machine toward us. Even though the light that says Contact would turn on, we saw that our wafer still was not in contact with the mask. In fact, this light just indicates that the lever is in its fully-rotated position. So we had to decrease the z-axis separation of the wafer, after rotating the lever away from us to bring the wafer back down again. We turned the knob counter-clockwise in front of the machine to decrease the separation distance. Then, we tried again, by turning the lever toward us and checking for contact between the mask and wafer. Once the mask was truly in contact with our wafer, we normally would have had to align the wafer to the features of the mask to ensure that the proper areas were exposed to UV light. However, since this was the first time exposing our wafers, there was not an existing pattern to align to. So we did not have to make any fine adjustments to the position of our wafer relative to the mask. Now that the mask was in contact with the wafer, our wafer cross-section profile looked like this: Next, we had to expose our wafer with UV light through the mask. So we set the exposure time to be for 120 seconds, made sure everyone had their goggles on, then hit the exposure button. Then, the upper part of the mask aligner machine usually would not move forward like it was supposed to, so we simply pushed it back and released, and it would slowly slide forward and click into position. For extra protection from the UV light, we also turned away from the machine to look away from it, even though we had our goggles on. We did not know if this step made a difference, but we decided it could not hurt us, so we continued to repeat it for each wafer we exposed. 3

4 During this step, we exposed our photoresist to UV radiation through the openings in the mask. This process is shown in this diagram below of our wafer profile: After exposing all of our wafers, we had to develop the photoresist. The UV radiation chemically changed parts of the photoresist by breaking through the bonds of its long polymer chains. This makes it easier for the developing agent to remove those parts of the photoresist. For development, we used the fluid in the bottle labeled MIF-300, pouring about 1 centimeter of this liquid into a glass dish. We also filled the cascade rinse tub and got out an empty wafer carrier to be ready for the rinsing step. Placing one wafer at a time into the developing liquid, we let the photoresist develop while gently moving the glass dish side-to-side to keep the fluid flowing around the wafer surface. After about 70 seconds or so, we removed the wafer with tweezers and placed it in the wafer carrier. Then we submerged the carrier in the cascade rinse tub, with the bubbles turned on. We repeated this process for the rest of our device wafers, giving each wafer a total developing time of 60 to 90 seconds. We left the wafer carrier in the cascade rinse tub the entire time, except for brief moments when we wanted to load another wafer that was done developing. After the last wafer was loaded, we let the wafer carrier remain in the cascade rinse tub for another 3 minutes, so that all of the wafers would be rinsed for at least that amount of time. To dispose the used developing fluid, we poured it into the developer waste bottle. Then, we thoroughly rinsed out the glass container with deionized water and set it to dry. Now that the exposed photoresist was completely removed, our wafers had a cross-section profile that looked like this: 4

5 After rinsing off the developing fluid, we dried our wafers using a nitrogen sprayer. Using pressurized nitrogen gas, we had a quick way of drying without damaging the surface of our wafers. Then, we looked at our wafers with the yellow filter on the microscope to inspect that the photoresist was exposed and developed properly. We checked that there were sharp edges and corners in the features of the photoresist layer. Though the instructions specified that we did not need to inspect every wafer, we wanted to practice using the microscope by focusing with it and navigating across our wafer surface, so we inspected each of our device wafers. We also took some photos with the digital camera on our microscope. See the Results section toward the end of the document to view these photos. Once we verified that our photolithography was done properly, we needed to do a post-bake on our wafers. So we set our oven and timer to the following settings: Oven Temperature: Time Pre-Baking: 120 C 25 min After removing our wafers from the oven and letting them cool, we had to do the etching procedure in the furnace lab. But before we left the lithography lab, we put away the mask. Then, we turned off the microscope light, the video monitor, the UV lamp power supply, and the power to the mask aligner machine. Then we closed the air and vacuum switches and turned off the vacuum pump. For the etching process, we first placed one of our test wafers in the buffered oxide etch (BOE) tub and measured how long it took to etch and remove the exposed oxide layer. Because our device wafers had the same oxide thickness, we knew they would take the same amount of time to etch. After several minutes of etching for our test wafer, we would pull the carrier out from the solution to check if there was still wetting on the surface of our wafer. Even though it was very simple, this check is a pretty good indicator that the oxide layer had been completely removed by the etching solution. This is because silicon dioxide is hydrophilic, which means that water tends to stick to it. However crystalline silicon is hydrophobic, so water drops cleanly slide off of its surface. So if we still had any oxide on the surface, when we pull out our wafer from the tub, we should see this wetting behavior. Timing the point when it transitions to non-wetting behavior, we found that it took about 5 minutes and 40 seconds in total to etch the oxide. So once we knew how long the etching process would take using our test wafer, we etched our device wafers for the same amount of time in the buffered oxide etch. Then, we rinsed the wafers in the cascade rinse tub for 2 minutes. At this point, we had removed the oxide that was not covered by photoresist, but we still had the unexposed photoresist on our wafer surface. So the cross-section profile of our wafers looked like this: 5

6 After the etching fluid was rinsed off, we had to remove the extra photoresist from the surface of our wafers. To do this, we used acetone, which is the same chemical as nail polish remover. We let it soak in the first acetone tub for 3 minutes and then in a second acetone tub for 1 minute to remove the photoresist. Finally, we placed the wafer carrier in the methanol tub for another minute. Afterwards, we rinsed off all of these chemicals in the cascade rinse tub for another 2 minutes and dried our wafers in the spin rinser/dryer machine. Now that the extra photoresist was removed, the cross-section profile of our wafers looked like this: 3. Boron Deposition and Drive (a) Standard Clean Because boron deposition is a high-temperature process, we first had to perform the standard clean procedure to remove impurities from the surface of our wafers. This procedure is necessary to prevent unwanted materials from diffusing and being baked into the wafer under high temperatures and ruining our devices. For the standard clean process, we prepared the ammonium hydroxide and hydroperoxide solution in the SC-1 tub. We also prepared a hydrochloric acid solution in the SC-2 tub. We then raised the temperature of both solutions to about 75 degrees Celsius. Then, we placed the wafers for fifteen minutes in the mixture of ammonium hydroxide and hydroperoxide, which is a basic solution. The step removes any organic contaminants, like oils, dead skin cells, and any leftover photoresist from the previous lithography step. After rinsing our wafers in the cascade rinse tub, we placed the carrier in the dilute hydrofluoric acid solution for fifteen seconds. Then, we submerged our wafers for another fifteen minutes in SC-2, which is an acidic solution. This step removes any metallic contaminants, like from contact with the metal tweezers, the spinner chuck, or the metal surfaces of the microscope. Also, this is the reason why we only use plastic tweezers to handle our wafers after doing the standard clean. Once we finish rinsing them again, we put our wafer carrier in the spin rinser/dryer machine for a full cycle. This machine quickly dries our wafers without damaging their surface, by spinning them and using pressurized nitrogen. (b) Boron Deposition Before starting the deposition process, our lab instructor helped us calculate the amount of time we would need to soak our wafers, based on the boron dose we wanted. Our lab instructor told us that the desired dose was the following value: 6

7 13 2 Q desired = 7.5 * 10 cm Additionally, we tried to determine our solubility concentration limit by looking at the following chart from the lecture notes: Knowing that we would be doing our processing at temperatures of 850 degrees Celsius, we were not exactly certain of the solubility limit value for boron in silicon from this chart. But we knew it would be somewhere between 7 * cm 3 and 1.2 * cm 3. Using a logarithmic chart that our lab instructor showed us, he showed us that the solubility limit value at 850 degrees Celsius would be about 9.5 * cm 3 for N s. We also had to determine our diffusion coefficient, knowing that it can be approximated as an Arrhenius value dependent on temperature. So we referenced the following table from the lecture slides to determine the Arrhenius constants for boron: 7

8 As a result, at 850 degrees Celsius (or 1123 degrees Kelvin) we got the following diffusion coefficient for boron: 3.5 ev k * 1123 K D = 1.0 * exp( ) 16 D = 1.96 * 10 cm 2 /s Then we used the expression that gives us the dose for a constant-source diffusion, and solved for the diffusion time, t: Q = 1 π * 2N s Dt Q π = 2N s D * t Q π * 1 2 N s D = t t = 2 Q π 4 N s 2 D Then we plugged in the values determined above for the dose, diffusion coefficient, and the solid solubility limit. This gave us the following value for t, our diffusion time: t = * ( 7.5 * 10 cm ) π 19 4 ( cm 3 * * ) 2 16 ( cm 2 * * /s ) t = seconds So the diffusion time for our desired dose is about 41 minutes and 38 seconds. After performing the preliminary standard clean on our wafers, we started the deposition process. From the previous photolithography step, we removed precise sections of oxide from our wafers, exposing the silicon underneath. These regions correspond to the p-well region of our devices, which need to be doped with a p-type material through this deposition process. For this lab, we use boron for the deposition. Boron has 3 valence electrons whereas silicon has 4. So because boron has one less valence electron than silicon, it is a p-type dopant. This is indicated in the periodic table chart below: 8

9 Our boron dopant comes from a solid source wafer product of grade BN-975, made by Saint-Gobain Ceramics. According to the datasheet [1], the composition of our wafers is mostly boron nitride as well as some boron trioxide. As discussed in lecture, using this product provides a major safety advantage, since boron nitride is not very toxic to humans. Additionally, using a solid source makes it rather difficult to accidentally ingest this substance, unlike gaseous sources for boron. Below is a picture of the boron source wafers of grade BN-975, which is what we used for this lab: [1] The first task for the diffusion process was to prepare the source boat, which was located at the center of our furnaces. Because the idle furnace temperature is still 400 degrees Celsius, we needed to put on thermally insulating gloves. To retrieve the source boat, we first removed the end cap of the second tube in the furnace, which was the boron deposition tube. Then we used a long push rod to pull out the source boat at a relatively slow 9

10 rate of 2 inches every 12 seconds. This rate of pulling is twice the rate we normally use because the temperature gradient was not as large, since the furnace is at its lower, idle temperature. Once the source boat was close to the edge of the tube, we used the short push rod to pull it onto a white plate. After removing the source wafers, we ramped up the furnace temperature to our desired diffusion temperature. According to the datasheet, the temperature ranges vary depending on the grade of the planar diffusion source product [1]: Grade BN-975 BN-1100 BN-1250 BN-HT Temperature Range C C C C So for BN-975 source wafers, we needed to perform the p-well diffusion at a temperature between 800 and 975 degrees Celsius. We followed this recommendation by setting our furnace to ramp up to 850 degrees Celsius. After that, we set the nitrogen flow rate of the furnace to 2 standard liters per minute (slpm). Next we loaded our wafers onto the source boat, using plastic tweezers to prevent getting metal contaminants on them. We placed our wafers in the empty slots directly adjacent to the source wafers. When doing this, we made sure that the shiny side faced the source wafer since this was the device side that needed to be doped. After loading all of our wafers onto the source boat, there were still some source wafers with one or both sides uncovered. So to prevent uneven diffusion, we used guard wafers to cover both sides of all source wafers, without losing track of where our own wafers were located. Once the furnace reached 850 degrees Celsius, we put on the thermally insulating gloves and pushed the quartz boat into the furnace tube very slowly. First using the short push rod, then using the long push rod, we slowly nudged the quartz boat deeper into the furnace tube at a rate of 1 inch every 12 seconds. After getting the boat to the center of the furnace, we then replaced the end cap of the tube. To begin the recovery process, we first opened the oxygen cylinder valves, including the main valve on the oxygen tank and the small valve along the gas line after the regulator. We then set the oxygen flow to be 1 slpm and decreased the nitrogen flow to be at the same rate, thus keeping our total gas flow at 2 slpm. We continued providing this flow for 20 minutes. For the recovery process, this percent of ambient gas flow matches the datasheet specifications. The datasheet [2] recommends using an ambient of 50% nitrogen and 50% oxygen under temperatures between 750 and 850 degrees Celsius. Not only were we doing our recovery process at 850 degrees Celsius, but we were also providing equal flow rates for the oxygen and nitrogen gas. According to the datasheet, this nitrogen and oxygen ambient grows a thin layer of silicon dioxide in the parts of the wafer where the silicon is exposed. This 10

11 layer masks the boron trioxide diffusion, which helps reduce gradients in the sheet resistivity of the wafers. This difference in sheet resistivity can occur since the first wafer inserted into the furnace is also the last wafer leaving the tube. So the first wafer effectively experiences a longer diffusion and gets a slightly greater dose of boron. This can slightly decrease the sheet resistance. We also inject some hydrogen gas, by opening the hydrogen cylinder and regulator valves. We turned on the flow to a rate of 40 standard cubic centimeters per minute (sccm) and maintained that rate for 2 minutes. This follows the two-step diffusion procedure discussed by the BN-975 datasheet [2], which recommends the hydrogen injection to last for 1 to 4 minutes. The datasheet claims that using hydrogen injection at low temperatures (less than 975 degrees Celsius) can be advantageous for shallow structures, including CMOS p-well applications. It also describes that the hydrogen injection can help improve the uniformity of the boron diffusion across the surface of the silicon wafers. It can also control defects in the silicon like oxidation induced stacking faults and point defects simultaneous to the deposition process. The datasheet [2] explains that the purpose of this source process is to transfer dopant to the surface of our wafers. This covers our wafers with a glass layer of metaboric acid, which is a substance with the chemical formula HBO 2. The datasheet also describes that to safely inject hydrogen, we must be sure that an excess amount of oxygen is included in the tube. This helps ensure that all of the unbound hydrogen atoms are consumed to form water vapor. Once the hydrogen flow was shut off, we had to perform the soak process. This is where we let the boron glass layer diffuse into our wafers. We started this process by shutting off the oxygen flow and increasing the nitrogen flow back up to 2 slpm to maintain the same total gas pressure. Describing the soak process in a bit more detail, the datasheet [2] explains that the dopant glass undergoes a reduction reaction when the oxygen and hydrogen gas are discontinued and the nitrogen gas flow rate is increased. This results in a thin layer of silicon-boride at the surface of the silicon. The datasheet claims that this thin layer helps control damage to the crystal at the silicon and silicon-boride interface. We left our wafers in the furnace for the desired soak time, which we calculated to be 41 minutes and 38 second. We also had to consider the time that it takes to slowly pull out the wafers, which takes about 5 minutes. So to ensure that the wafers were not experiencing the soak process for too long, we made sure to begin pulling the wafers out after about 36 minutes and 38 seconds. Using thermally insulating gloves, we pulled the wafers out at the typical rate of 1 inch every 12 seconds. We first used the long push rod and then switched to the short push rod once the wafers were close to the edge. Then we brought the wafers onto a white plate and replaced the end cap of the tube. Afterwards, we set the oven back to its idle temperature of 400 degrees Celsius and its idle nitrogen flow rate of 0.3 slpm. At this point, our wafers still had the glass dopant layer on the surface, which needed to be etched off. After letting the wafers cool down a bit, we used plastic tweezers to remove our wafers and place them into a teflon wafer carrier. Then, we filled the rest of the source wafer 11

12 boat with guard wafers before we put it back in the furnace. We pushed the source wafer at a rate of 1 inch every 12 seconds, first using the short push rod then switching to the long push rod. Finally, we had to deglaze our wafers by placing them in the buffered oxide etch tub for 30 seconds. Afterwards, we rinsed them in the cascade rinse tub and dried them by placing them in the spin rinser/dryer machine for one cycle. (c) Boron Drive At the end of the boron deposition lab, our lab TA told us he would be the one doing the boron drive process for our wafers. This boron drive procedure accomplishes two main tasks. First, it creates our p-well junction. Second, it performs an oxidation process that covers the exposed silicon with a layer of oxide. The first main steps of the boron drive include a low-temperature oxidation which removes the boron skin layer that forms during the previous deposition process. Then a wet-oxidation process is done to form the layer of oxide over the exposed areas of silicon. During this time the drive diffusion process begins that allows the boron dopant to diffuse deeper into the wafer and form the p-well. The first step includes a standard cleaning procedure. Afterwards, the oxidation tube of the furnace needs to be brought to its standby state at 800 degrees Celsius with a nitrogen flow rate of 1 slpm. After checking that the bubbler has plenty of deionized water, we can set the nitrogen gas flow rate through the bubbler to 200 sccm. We can also set the temperature of bubbler to 98 degrees Celsius. After this setup of the furnace is done, the quartz boat for the third tube needs to be removed from the furnace onto a ceramic plate using a short push rod. Then after waiting a few minutes to let it cool, our wafers were loaded into the boat with plastic tweezers. The entire boat was returned to the mouth of the furnace tube with the device side facing the inside of the furnace. Then, the wafers were slowly pushed to the center of the furnace at a rate of 1 inch every 12 seconds. The pushing first used the short push rod then switched to the long push rod to get the wafers to the center of the furnace. Finally, the end cap was replaced onto the end of the tube. Once the wafers were back in the furnace and the bubbler temperature reached about 98 degrees Celsius, the dry nitrogen flow was switched off. This began the low-temperature oxidation process, which lasts about 30 minutes. According to the datasheet [3], the purpose of this low-temperature oxidation process is to oxidize the silicon-boron layer and a thin layer of silicon that lies beneath this boron skin. The boron skin forms when excess boron atoms react with the silicon surface, and it is usually only 20 nanometers in thickness. Additionally, by oxidizing these layers, the crystal defects will be immobilized in the oxide. This prevents these defects from propagating into the silicon wafer. A diagram of this boron skin from the cross-section profile is shown below: 12

13 4. Results After the time for this low-temperature oxidation process is complete, the wafers were slowly removed from the furnace at a rate of 1 inch every 12 seconds. Once the wafers cool down, they are placed into a teflon wafer carrier and etched in the buffered oxide etch tub for 30 seconds. Finally, they are rinsed in the cascade rinse tub for 3 minutes and dried in the spin rinser/dryer for 1 cycle. Next comes the oxidation drive process, which requires us to place the wafers back into the oxidation boat and reload this carrier back into the furnace. The wafers were reloaded at the same rate of 1 inch every 12 seconds, then the temperature of the oven was ramped up to the desired oxidation temperature of 1135 degrees Celsius. Once the bubbler settled to being within 2 degrees of the 98 degrees Celsius temperature setting, the nitrogen gas was switched off and redirected to flow through the bubbler. This action starts the wet oxidation procedure to create the oxide layer above the exposed silicon. Once the oxidation time of about 13.5 minutes was complete, the bubbler was switched off. Then the nitrogen gas was again flowed into the oxidation tube at 1 slpm to continue on to the boron drive process. For the drive process, the wafers were left in the high temperature oven for another 15 hours and 46.5 minutes. This makes the total time an even 16 hours for our p-well drive process, since we need to consider the diffusion that occurred during the previous oxidation step as well. Afterwards, the furnace temperature was allowed to ramp down. The boat could be pulled out after the temperature reached less than 800 degrees Celsius. The wafers were removed at our typical pull-out rate of 1 inch every 12 seconds. After letting the wafers cool for a few minutes, they were removed from the boat with plastic tweezers. The boat was returned to the mouth of the oxidation tube, and the furnace was set back to 600 degrees Celsius with a nitrogen flow rate of 0.3 slpm. During the photolithography process, we checked the features of our pattern from the PWELL mask using a microscope to make sure the photoresist developed correctly. We made sure that the features were all distinct, with sharp edges and corners. We could see our wafers on the television screen hooked up to the microscope. We tried to take a picture of our observations on the television screen with a camera phone: 13

14 However, these images were a little bit difficult to see clearly unless you were in the lab at the time. So with our instructor s help, we were able to take some pictures of our wafers from the digital camera that was hooked up to the microscope. These pictures give a much better representation of the devices on our wafers: 14

15 As seen in the photos above, the mask pattern came out rather nicely. The features had sharp lines, even for finer details of the mask. So we determined that the p-well photolithography was quite successful, and our lab TA agreed. We were not able to actually check if our boron profile was successful, however the lithography process seemed to show satisfactory results during inspection. When doing the boron deposition step, we did not experience any major issues. However during the boron drive, which was performed by our instructor, there seemed to be a small issue with too much oxide being formed for our group s wafers. Dr. Tuttle measured the oxide layer for several groups and determined that the final oxide thickness was about 820 nanometers. Since we tried to grow only about 310 nanometers of oxide, this was significantly more than we were expecting. We checked this actual thickness at the beginning of the following lab ourselves, and found that the average oxide thickness for our device wafers was about 860 nanometers. At first, Dr. Tuttle did not know the exact cause of this issue. Initially, he thought that possibly one of the valves got stuck and there was still some gas containing oxygen flowing into the furnace tube. Another possibility that Dr. Tuttle was considering was that the nitrogen gas ran out during the night, and that the tube back-filled with some air. However after further investigation, Dr. Tuttle determined that the issue was due to a puddle being formed in the tube connecting the bubbler to the furnace. This caused a problem because after the wet oxidation process was stopped, there would still be water in the tube that would provide a source of oxygen. This extra source of oxygen would be available to react with the silicon to form more silicon dioxide, which would increase the thickness of our oxide layer. However, despite this unexpected extra growth, Dr. Tuttle and our lab instructor both believe that this extra amount of oxide should not be too much of a problem for our devices. The only difference now is that it will take longer to etch for the next steps. This increased etch time in the 15

16 next lithography process can lead to more undercutting, as shown in the following sequence of diagrams: So for the upcoming PMOS source and drain lithography process, the exposed silicon may be a bit wider than we were hoping due to this undercutting effect. The amount of undercutting increases with thicker oxide layers, so our thicker layer would create a more noticeable effect. But overall, the unexpected increase in our oxide layer should not become a major issues and it will not harm the functionality of our devices. Indeed, unexpected things happen. And even if our devices are not perfect, hopefully they will work one way or another. Regardless of the outcome, the overall process was a good learning experience, and in the end that is what ultimately matters. 16

17 5. References [1] Saint-Gobain Ceramics, PDS Products Boron Nitride P-Type Source Wafers, Revision A, BN-975 datasheet (PDF) [2] Saint-Gobain Ceramics, Increased Yield Using PDS Products Grade BN-975 with Hydrogen Injection, Revision A, BN-975 hydrogen injection specification (PDF) [3] Saint-Gobain Ceramics, Silicon Boron Layer (Si-B) & Low-Temperature Oxidation, Revision A, BN-975 low-temperature oxidation specification (PDF) Appendix We had to determine the total diffusion time necessary for the constant-source diffusion process. This calculation and our reasoning is explained in detail in section 3b. The same calculations are repeated here, in a condensed form: Known or given information: 13 2 Q desired = 7.5 * 10 cm T = 1123 K N s = 9.5 * cm 3 (at 1123 Kelvin) D o = 1 cm 2 /s (for boron dopant) E A = 3.5 ev (for boron dopant) Calculations performed: D = D o * E kt A exp ( ) 3.5 ev k * 1123 K D = 1 cm 2/s * exp ( ) 16 D = 1.96 * 10 cm 2 /s Q = 1 π * 2N s Dt Q π * 1 2 * N s * D = t 17

18 t = Q 2 π 4 N s 2 D t = * ( 7.5 * 10 cm ) π 19 4 ( cm 3 * * ) 2 16 ( cm 2 * * /s ) t = seconds ( or 41 min. and 38 sec.) 18

19 Here is a scanned copy of our process traveler sheet for the p-well photolithography process: 19

20 Here is a scanned copy of our process traveler sheet for the boron deposition process: 20