Introduction
This is the result of several experiments on the adhesion failure of photoresist during the spray wet chemical etching process of InGaP/GaAs NPN HBT. Several factors that may influence adhesion were identified, and the design of experiments (DOE) method was used to study the effects and interactions of selected factors. The most significant adhesion improvement identified was the incorporation of native oxide etching immediately prior to photoresist coating. In addition to improving adhesion, this precoat treatment also changes GaAs such that the reaction-limited etch is more isotropic compared to the unsurfaced wafer ; the profiles both have a positive taper direction, but the taper angles are not the same . The altered profile allowed us to use the evaporated metal of 5200 without planarization to produce a fully detectable HBT with an 5×5 m emitter.
Keywords: photoresist adhesion, gallium arsenide , wet etching, surface treatment, profiling
Introduction
photoresist adhesion results in wet etching and subsequent yield of electrical and optical devices plays a key role in. There are many factors that can cause photoresist to adhere to a semiconductor substrate. However, there is very little information on gallium arsenide in the public literature, and methods commonly used for silicon, such as hexamethyldisilazane (HMDS) pretreatment may not be effective for gallium arsenide.
Our historical etch process made two major process changes, first, we switched from Clariant AZ4330 photoresist to Shipley SPR220-3. We have found that the latter resist has better rotational uniformity and resolution, but its adhesion to GaAs is slightly worse than AZ4330. Second, we moved wet etching from a manual immersion-based process to the SSEC 3300 spray etch system. While it is possible to produce better etch uniformity and repeatability, spray etch systems can be a harsh test of photoresist adhesion and can lead to process failure in the wrong circumstances, as shown in Figure 1 , Figure 1 shows a typical device from one of the first batches encountered in the spray etching system.
The initial photolithography process served as a control for the studies described later. Lithography consists of HMDS vapor primer at 1°C, 5 kRPM coating for 2.2 m film and 115html 90 second soft bake at 1°C on a Microtec ACS200 coated / developing track . The samples were then exposed with an ASML PAS5000/55 i line projection stepper ( dose =370mJ/cm2), post-exposure baked at 115°C for 1 second, and Co. NMD-W (2.38% TMAH) Spin develop 1html in developer for 1 minute. They were then post-development baked at 120°C for two minutes and subjected to oxygen plasma deslag in a Tepla 300 barrel asher for one minute. The slag is 200 watt, 750 millitorr plasma , the flow rates of nitrogen and oxygen are 500 sccm and 10 sccm respectively. Typically, during the slag removal process, the sample is placed horizontally on a metal plate.
and then clean the sample 10 second with 20 H2O:1 NH4OH (clean before etching), and 1 h3po 4:4 H2O 2:45 H2O Etched in solution. After photoresist stripping, bond quality was determined by visual inspection with an light microscope . Each wafer in the study was given a subjective bonding rating, with 10 being the best and 0 being the worst.Serpentines are particularly helpful in judging bond quality because they provide ample opportunity for bond failure within a relatively small viewing area. Figure 2 shows an example of such a snake structure, illustrating a resolved failure mode.
Initial segmentation
To study resist adhesion issues, manual, immersion-based etching using n type (100) GaAs machinery was performed before resist adhesion DOE (the etch depth was approximately 2 m) Process separation. The split shows that increased dehydration bakes (120°C10 minutes) and higher post-development bakes (130°C2 minutes) had little effect on improving adhesion compared to the control. A more aggressive, five-minute, 600 W, 225 mTorr, O2 flow rate 600 sccm deslagging before photoresist coating resulted in reduced adhesion quality. Letting the samples sit for three days instead of etching immediately reduced the adhesion of all samples except the 20 H2O:1 NH4OH pre-coated sample. 20 H2O: 1 The pre-coating treatment of NH4OH 10html produces excellent adhesion in 1 second. Additionally, a 10 second 10 H2O:1 HCl pre-etch clean improved adhesion over an ammonium hydroxide-based clean.
A surprising result of these initial splits is the profile of the (100) N type GaAs wafer, which is consistent in both crystallographic directions when the 20 H2O:1 NH4OH pre-coat treatment is applied as shown in Figure 3 Tapered. The slope of the precoated profile is approximately 50 °, when viewing a typically retrograde cross-section, the slope is approximately 40°.
anti-adhesion agent
To further characterize and optimize the etching process, resist adhesion DOE was performed. The etch depth was approximately 0.8 μm and the samples were etched in a spray etching tool the day after photolithography.
An important process detail is the oxygen plasma residue of the sample, which is placed vertically on the quartz boat instead of horizontally on the metal grid. When comparing control samples to other experiments, the use of quartz boats may significantly reduce adhesion under certain conditions because it thermally insulates the wafers, allowing them to heat up during the deslagging process. This results in a poor surface for the resist to adhere to. The second detail is that the NH4OH-based pre-etch cleaning is performed in a spray etch tool, while the HCl-based pre-etch cleaning is performed manually by soaking in an acid bench. The increased jet pressure may contribute to the poor adhesion observed when performing NH4OH based pre-etch cleans without any pre-coat treatment.
A surprising result of these initial splits is the profile of the (100) N type GaAs wafer, which is consistent in both crystallographic directions when the 20 H2O: 1 NH4OH pre-coat treatment is applied as shown in Figure 3 It's tapered. The slope of the precoated profile is approximately 50 °, when viewing a typically retrograde cross-section, the slope is approximately 40°.
Anti-adhesion agent
To further characterize and optimize the etching process, a resist adhesion DOE was performed. The etching depth was approximately 0.8 μm, and the samples were etched in a spray etching tool the day after photolithography.
An important process detail is the oxygen plasma residue of the sample, which is placed vertically on a quartz tank rather than horizontally on a metal grid. When comparing control samples to other experiments, the use of a quartz bath may significantly reduce adhesion under certain conditions because it thermally insulates the wafers, allowing them to heat up during the deslagging process. This results in a poor surface for the resist to adhere to. The second detail is that the NH4OH based pre-etch cleaning is done in the spray etch tool, while the HCl based pre-etch cleaning is done manually by soaking in the acid bench. The increased spray pressure may contribute to the poor adhesion observed when performing NH4OH-based pre-etch cleans without any pre-coat treatment. The three factors considered in the 3x3x2 DOE are pre-coat treatment (20 DI:1 NH4OH, 10 DI:1 HCl or none), post-development bake (120°C for 2 minutes or none) and pre-etch clean (20 DI: 1 NH4OH, 10 DI:1 HCl or None). Table 1 details the levels of each factor and the final bond grade.
As shown in Figure 4, the etched profile of the sample incorporating the pre-coated treatment from DOE has a slight lip on the top of the table. This lip could be explained by partial adhesion loss using the spray tool, or the inability of the current spray etch process to remove material at the photoresist/GaAs interface. We are working on ways to eliminate the observed lip.
