Mesa Delineation of Strained Layer Superlattice Infrared Detectors by means of Wet-Chemical Etch

Transcription

1 Mesa Delineation of Strained Layer Superlattice Infrared Detectors by means of Wet-Chemical Etch Nathan Henry Graduate Mentors: Brianna Klein and Nutan Guatam Faculty Advisor: Dr. Sanjay Krishna

2 Introduction A study has been performed to discover the feasibility for the use of a wet-chemical etchant to form mesa structures on Strained Layer Superlattice (SLS) devices. All etching recipes involved the use of phosphoric acid, tartaric acid and hydrogen peroxide. Etching was performed on single pixel devices 410μm x 410μm mesas. Characterization of surface morphology included RMS roughness measurements, mesa sloping angles and enhanced etch measurements. The final goal of this project is to compare noise measurements between devices formed by wet etch and dry etch procedures. Although this goal has not yet been reached, it is the next step in this research. Much was discovered about wet etch characteristics and has been reported in this paper. Most notably, refer to Figure 8 to see a mesa etch at a large temperature. Mesa walls are very smooth with minimal to no sloping. This research has been presented at Undergraduate Research and Creativity Conference (URCC), a symposium held at the University of New Mexico. Nathan received funding to attend the SPIE Defense, Security and Sensing conference in Baltimore, MD. He had the wonderful opportunity to attend many presentations on infrared detection and was able to meet people renowned in the infrared community. Background Infrared detection observes light outside of the visible spectrum, more specifically light in between 3 and 1000 micrometers (μm) in wavelength. Further classification of the infrared spectrum can be defined by six different categories: near (1-2 μm), short-wave (2-3μm), mid-wave (3-5μm), long-wave (8-12 μm), very long-wave (12-25 μm) and far (>25μm). This type of light is emitted by a wide variety of sources and the observation of this light provides many useful applications. Most think of infrared radiation as being the radiation of heat. While this is true, it does not fully define the infrared spectrum. For example, television remotes do not emit any heat when transmitting signals yet use the infrared spectrum to wirelessly transmit instructions. Many objects emit radiation in the region of mid to long wave radiation. Humans emit radiation at a wavelength of approximately 10μm. Thus, infrared technology is very useful for search and rescue missions. Many medical applications of infrared imaging exist. Research is currently being performed on detecting the emission of skin cancer. Earth s atmosphere is transparent to wavelengths within 3-5 and 8-12 μm. This is enticing to astronomers as the atmosphere diminishes much of the visible light from space. Infrared astronomy is used to measure the temperature of stars many light-years away, which can help glean a variety of information about these celestial bodies. Another major application is infrared spectroscopy. Due to molecular bonds and vibrations, the absorption of light at specific frequencies can provide a means to identifying substances as well as the calculating bond lengths and strengths.

3 Mainly, there are three ways to achieve night vision. The first is light amplification, a technology used by thermal goggles. These are usually produced at a high volume and relatively low cost, but provide a low quality of imaging. Another form of infrared technology uses slight changes in temperature to detect IR light incident on its surface. These slight changes in temperature are measured through a change in resistance, voltage, or current. An example of this type of device is the bolometer. These devices have a high volume of production with a mid-level performance quality. The last category of IR detectors has a low volume of production and a high image quality and is the subject of this report. These devices are commonly referred to as photovoltaic detectors; they utilize the properties of semiconductors to detect photons at specific wavelengths. When a photon of some particular wavelength is incident on the surface of the semiconductor, the photon s energy may be absorbed, promoting carriers to the appropriate band. Given the photon s wavelength, its energy can be calculated. If the energy is larger than the bandgap of the material then the photon will be absorbed. Electrons are promoted to the conduction and leave a hole in the valence band. This promotion of carriers causes a current that is passed to an integrated circuit and transformed into an image. In order to detect photons at specific wavelengths, the semiconductor material is grown for a specific band gap; this band gap is then tuned using an appropriate voltage bias level. Unlike the devices mentioned previously, most photovoltaic detectors need to be at a temperature of 77K in order for the camera to be operable. This is done to reduce the noise, namely dark current, in the form of thermally generated carriers. Dark current is the current present in a detector when no light is incident on its surface. When at room temperature, electron mobility is high in semiconductors, thus a current can be seen regardless of photons that are incident on the material surface. There are multiple approaches to creating these devices however this paper will focus on Strained-Layer Superlattice (SLS) IR detectors. E. Plis et al. state that fabrication of SLS structures involves the periodic layering of different semiconducting materials. Figure 1 is a scanning transmission electron microscope (STEM) image of a SLS by, J. Steinshnider et al, showing the alternating layers of the SLS material. It should be noted that the whole image is only 350 angstroms, giving an idea of exactly how thin these structures are usually grown. Electrons and holes are confined in quantum wells; one layer of the SLS acts as the electron well, having a lower energy potential, while another layer acts as a hole well, having a higher energy potential. Once the electron is confined, it possesses quantized energy levels and thus can only have a specific number of wave functions. These layers are grown so thin that the electron and hole wave functions, though confined to different layers, are able to interact. This interaction forms minibands which define a new bandgap unlike either of the constituent materials. Controlling the thickness of the alternating layers allows for tuning the bandgap. There are currently two types of SLS detectors, Type I, Type II, Type III. Materials used in Type I superlattices can include GaAs/AlAs resulting in nested band gaps. One material s conduction and valence band will be completely encompassed by the conduction and valence band of the second material. This results in the confinement of both electrons and holes in the first material. Type II superlattices are composed of materials such as InAs/GaSb, this results in the staggering of the two materials band gaps. The valence band of the first material resides at a higher potential than the conduction band of the second material. Due to this

4 configuration, the electrons are most likely to reside in the second material while the holes are most likely to reside in the first. The processing of Type II superlattices is the focus of this paper. Type III SLS devices will not be discussed as they involve II-VI materials. The first step to creating a Type II SLS device is growth, which is done through a process called Molecular Beam Epitaxy (MBE). This involves the accretion of III-V material onto a substrate, composed of GaSb. After the device has been grown, small square mesas are etched. Photoresist is applied to the sample, through a process called photolithography, to protect the sample from the etching solution. Refer to Figure 2 for a graphical illustration of photolithography and mesa delineation. Exact details as to the photolithography recipe will be given later under the Experiment section. From etching, the sample is given a three-dimensional patterning as shown in Figure 3 (a). After etching the mesas to the appropriate height, top and bottom metal contacts are deposited. These metals are evaporated with an electron beam and deposited onto a sample. Once again, photoresist is applied using a second mask. The metal is evaporated on top of the photoresist, wherever the photoresist is applied, the metal will come off when submersed in acetone causing the photoresist to become soluble there by taking the metal off the surface. This process is called lift-off. The shiny squares in Figure 3 (a) illustrate the top metal contact. For our purposes, metals involved in this deposition include platinum, titanium and gold. These metals are chosen to achieve a proper ohmic contact, rather than a chottky barrier contact. The next step, requiring a third mask is called passivation. Passivation is the application of an insulator to the sidewalls of the mesas. There are two reasons why passivation is necessary; foremost, a passivant is used to reduce leakage of current on the mesa sidewalls. Secondly, a passivant is used to make the device more robust as noise levels in the device can increase over time. After passivation, indium is deposited and reflowed into indium bumps as shown in Figure 2 (b). These bumps act as conductive glue when mounting the detector to a read-out integrated circuit (ROIC). The ROIC is used to read electrical signals from the pixels, and deliver them to a computer. The computer algorithm then translates all the electrical signals into an image. A diagram of the device is shown in Figure 4. Lastly, underfill epoxy is applied to the device. This epoxy fills in the gap between ROIC and device and helps to give it a more robust structure and longer lifetime.

5 Figure 1. STEM image of a SLS device. (J. Steinshnider et al., Phys. Rev. Lett. 85, 2953 (2000)) Figure 2. Graphical representation of Photolithography and Mesa Delineation. (a) (b)

6 Figure 3. (a) Mesas with metal deposition. (b) Indium bumps after reflow. Figure 4. Final product before substrate removal. Once mounted to the ROIC, infrared light is incident on the backside side of the sample. The incident light must first pass through the substrate before being detected. Thus it is common to either thin or completely remove the substrate. There are many other reasons why the substrate needs to be removed as well; most notably, it severely decreases the efficiency of the detector, as stated by Delaunay et al. A major issue with this is the high photon absorption possessed by GaSb, p doped is worse than n doped. Substrates also prevent Fabry-Pérot oscillation. This occurs when a photon passes through the active region and gets reflected between the walls of the device. This allows for a higher probability of absorbing the photon. The substrate prevents this from occurring because of its high absorption coefficient. Removing the substrate can also prolong the IR detector s life span. The IR detector must undergo many repetitions of thermocyling. Before use, it must be cooled down to cryogenic temperatures and returns to room temperature when not in use. The ROIC is silicon based, while the substrate is GaSb, and indium is also present between the ROIC and substrate. Due to major differences in the thermal expansion coefficients of Si, GaSb and indium, the process of thermocycling can deform and crack the SLS or even separate the ROIC from the sample. With the substrate removed, the SLS is thin enough that it can flex with the expansion and contraction of the ROIC. There are currently two methods of mesa delineation, wet and dry etching. Dry etching can be performed either by Reactive-Ion Etching (RIE) or Inductively Coupled Plasma etching (ICP). These methods use the process of ion-bombardment to remove material.

7 There are a few drawbacks with dry etching. The gasses used are not readily available and are costly. According to P. Holloway, RIE has also been shown to damage the device. Damage is induced mainly through ion implantation, a symptom of the ion bombardment process. The ion is implanted into the sample thus introducing impurities. This damage can reduce photoluminescence intensity, quantum efficiency and carrierconcentration at shallow depths. Wet etching has also shown to have a few drawbacks. Sidewalls of wet-etched mesas are typically slanted leading to a larger surface area, which in turn produces larger surface leakage. There are many inaccuracies that must be dealt with, mainly due to human error. Wet etching has however proved to show a decent mesa height uniformity, and is a more economical method. Research The research composed during the Fall 2011 and Spring 2012 semesters for the National Science Foundation sponsored Nanophotonics REU has involved mesa delineation by means of wet etching. Due to the negative effects of RIE, a study of wet etching has been performed and is the main topic of discussion in this paper. It is hypothesized that wet etching can possibly provide lower dark currents than dry etching due to less damage inflicted on the sample. Wet etching is the submersion of the device in a liquid chemical solution that reacts with the material there by etching the surface at a given rate. J.G Buglass et al. states that The mechanism of wet chemical etching consists of two distinct steps: (i) oxidation of the surface layer, and (ii) dissolution of the resulting oxidized species. Thus, there are two regimes to be considered when creating a wet etch recipe; diffusion and reaction limited. Diffusion limited solutions have etch rates limited by how fast molecules can move through the stagnant layer and react with the surface. The stagnant layer is a layer of used chemicals above the surface of the sample. Reaction limited solutions are limited by how fast the molecules react with the surface. When developing a recipe for wet etching mesas there are multiple goals. The resulting surface must be smooth. This is for a couple reasons, first being that if the surface is too rough, the metal will not properly deposit onto the sample. Second, a larger surface roughness will most likely result in higher noise levels. Another goal is to produce a chemical solution resulting in as anisotropic of an etch as possible. In other words, mesa sidewalls must be perpendicular to the surface with minimal undercutting and sloping. If the sidewall is too slanted it is possible that the next photolithography, for contact metal, would be impossible. The chemical solution is very sensitive to the type of material used. With Type Two SLSs, InAs and GaSb must both be etched in order to form the mesas. Due to this, it is desirable to obtain a recipe that holds similar characteristics for both materials in order to achieve a homogenous surface. Selectivity is one factor influencing these characteristics, it is simply a ratio of etch rates. Thus, a selectivity equal to 1 is most desirable. The solution used in the following experiments consisted of orthophosphoric acid (H 3 PO 4 ), Tartaric Acid (C 4 H 6 O 6 ), Hydrogen Peroxide (H 2 O 2 ) and Water (H 2 O). According to E Polakowska most chemical etching solutions consist of an oxidizer, etchant, complexing agent and dilution. The Hydrogen Peroxide acts as the oxidizer, this breaks the bonds of the material through the process of reduction and oxidation. The

8 etchant (orthophosphoric acid) then attacks the oxidized compounds, making them soluble and allowing them to leave the surface. The complexing agent (tartaric acid) acts to help form stable ions in the solution and prevent disassociation. Water acts as a dilutant, it decreases the solution s etch rate but can be used to select the desired etching regime. It should be noted that if the ratio of oxidizer to etchant is too large then insoluble oxides will deposit on the surface. Figure 5 shows an etched sample that has some unknown material redeposited on the surface, it is either composed of native oxides or some kind of compound created by the chemical solution. Also, strong oxidizers exhibit higher etch rates however can degrade surface morphology. Figure 5. Redeposit due to low dissolution Experiment As stated earlier, a process called photolithography is used to protect the mesas from being etched. The following is a step-by-step recipe for the photolithography process used: 1. 5 minute soak in acetone 2. Rinse with IPA to remove acetone residue 3. Rinse with Water to remove IPA residue 4. Blow dry with nitrogen 5. 5 minute dehydration bake at 150 C 6. Spin on HMDS at 3000 rpm 7. Bake for 30 seconds at 150 C 8. Spin on AZ 4330 Photoresist at 4000 rpm (resulting in a thickness of 3um) 9. Bake for 90 seconds at 90 C 10. Expose to 365nm or 405nm UV light for 6 sec 11. Develop in AZ 400K developer at a (1:4) dilution ratio All samples were mounted onto a silicon wafer, now called a silicon handle, with Teflon tape in between. The tape must be Teflon in order to prevent it from reacting with the chemical solution. This allowed for the sample to be etched face down and with as little

9 tilt as possible. Tilt has been found to reduce the uniformity of mesa height across the sample. The following is the recipe developed to apply the tape and silicon handle. 1. Obtain clean silicon handle 2. Spin photoresist on the sample twice (4000 rpm) 3. Apply teflon tape 4. Apply one drop amount of photoresist on top of the tape 5. Stick sample to photoresist 6. Bake for 5 minutes at 90 C All experiments reported here were performed on Sample A. Refer to Figure 6 for a diagram of its structure, which gives information as to the composition throughout the structure. The top GaSb layer acts as a top contact and is p-doped to insure an ohmic contact. The bottom contact, which consequently is the layer that must be reached when etching the mesa, is labeled SLS 16x7 and is n-doped to also insure an ohmic contact. There are some problems with procedure that could be improved upon. Roughness measurements were not reported in this paper because the data cannot be completely trusted. The profilometer needle is very large compared to defects on the surface. The surface roughness is also not completely uniform, thus RMS roughness measurements are very dependent on the area scanned. A better method is to use an Atomic Force Microscope (AFM); this has a much smaller needle and scans a whole area determined by the user. There is another issue that is prevalent when trying to measure the slope angle with the Scanning Electron Microscope (SEM). It is possible to get very close to a 90 angle when taking pictures of the sidewall, however it is very difficult to achieve a perfectly perpendicular view. This can end up giving inaccurate results. A better method would be to cleave the sample on a mesa and obtain a microscope image of the sidewall by sticking it to a glass slide. Figure 6. Structure of L Results and Discussion

10 A molar ratio of (1:2:1:16) of (H 3 PO 4 :C 4 H 6 O 6 :H 2 O 2 :H 2 O) was suggested at the start of the project. It became difficult to dissolve so much tartaric acid in the solution. Heat was applied to the solution to dissolve more tartaric acid however, when cooled down the acid would precipitate on the surface of the solution and the sides of the beaker. This was not acceptable, as it would prevent reproducibility. The following table gives some results when using this method. Please note that this composition is given in molar ratios. Composition (H 3 PO 4 :C 4 H 6 O 6 :H 2 O 2 :H 2 O) Rate(nm/min) Agitation Orientation Temperature ( C) 1:1:2: Yes DOWN 20 1:1:2: No DOWN 20 1:1:2:23 10 No DOWN 20 1:1:2:16 62 No DOWN 20 1:1:2:23 N/A Yes DOWN 20 1:1:2: No UP :1:2: No UP :1:2: No UP :1:2: No UP Table 1. Initial experiments showing the effect of temperature and orientation The following table gives the most recent and most optimized solutions to date. It should be noted that the tartaric acid solution was created by mixing 1 gram of tartaric acid with 1 ml of water. One problem encountered involved an enhanced etch rate close to the mesas. Refer to Figure 7 for an image that illustrates this, the etch depth on the right is greatly increased in comparison to the left. This makes it difficult to achieve the proper etch depth to expose the bottom contact. S. S. Tan et al says that this enhanced etch comes from the solution being diffusion limited. The paper states that additional flux from the x direction enhances the etch rate near the mesa. This can be reduced by thinning the stagnant layer achieved by agitating the sample during the etch process. The first three entries of the table give interesting results as to its enhanced etch profile. As the amount of the oxidant increases the etch will move more into the diffusion limited regime and thus more of an enhanced etch would be expected. Please note that the composition given in Table 2 is in volumetric ratios. Composition (H 3 PO 4 :C 4 H 6 O 6 :H 2 O 2 :H 2 O) Etch Rate (nm/min) Enhanced Etch (nm) (1:1:0.5:2) (1:1:1:2) (1:1:1.5:2) (1:1:0.5:1) (large standard deviation) (1:1:1:1) N/A N/A (1:1:0.5:1.5) Table 2. Recent experiments showing the effects of increasing the oxidant

11 Figure 7. Illustration of Enhanced Etch Results and Discussion While it may require more data points to confirm these results a few conclusions can be drawn from data acquired. Using a face-down tape method with the sample floating on surface tension resulted in a smoother and slightly slower etch. The RMS roughness was shown to decrease an order of magnitude. However, as stated previously, the profilometer may not be accurate enough to give concrete results. Temperature increased the etch rate. This is no surprise as there is more energy in the system when at a higher temperature. Surprisingly, however, an increase in temperature resulted in a much smoother and more anisotropic etch. Refer to Figures 8 (a) and (b) to compare the mesa sidewalls at both room temperature and at C. This is a very useful property and needs to be investigated further. As shown in Table 2 the etch rate is very quick, this can make it difficult to achieve the desired etch depth. (a) Room Temperature (20 C) (b) C Figure 8. A comparison of mesa sidewalls at various temperatures.

12 An important discovery was made late in the research. It was always thought that crystal direction played an important role in chemical etching however no concrete evidence was found until obtaining the SEM images shown in Figure 9 (a) and (b). These pictures are taken of the same sample, one mesa sidewall is perpendicular to the other. This shows the effect of crystal direction on sloping. In one direction (Figure 9b) it is shown there is a slight undercut with minimal sloping. Another direction (Figure 9a) illustrates a large amount of sloping. Thus the isotropic behavior of the chemical etchant must be considered. Unfortunately the miller indices, a method for determining crystal direction, of each image is not known due to the cleaving of the sample prior to this discovery. (a) (b) Figure 9. Illustration of the effect of crystal directions. Conclusions The phosphoric solution has been reported as a favorable chemical etchant by many papers. There have been favorable results achieved and interesting characteristics have been discovered. Further work is required to obtain an optimized recipe. The most recent results have proved to be promising, recently it was suggested that another brand of tartaric acid be used which will result in a smoother surface. Further work will no longer use an SEM for angle measurement and an AFM will be used for RMS roughness measurements. One of the most intriguing results of this research comes from the SEM image shown in Figure 8 (b). This image shows a very perpendicular sidewall with favorable homogeneity throughout. This is an ideal mesa and further work will be done to achieve this again. It is unknown whether this homogeneity is due to the high etch rate or temperature. The challenge with this etch recipe is its speed; with such a fast etch rate it is difficult to achieve the desired height. A procedure needs to be developed to stop the removal of material instantly and accurately. With more work it will soon be possible to perform electrical characterizations and compare with a dry etch process.

13 References Dier O. et al., (2004). Selective and non-selective wet-chemical etchants for GaSb-based materials. Semiconductor Science and Technology Chaghi R. et al., (2009). Wet etching and chemical polishing of InAs/GaSb superlattice photodiodes. Semiconductor Science and Technology Papis-Polakowska E. (2006). Surface treatments of GaSb and related materials for the processing of mid-infrared semiconductor devices. Institute of Electron Technology, 32/46, BUGLASS, J., MCLEAN, T., & PARKER, D. (1986). A CONTROLLABLE ETCHANT FOR FABRICATION OF GASB DEVICES. [Article]. Journal of the Electrochemical Society, 133(12), Delaunay, P., Nguyen, B., Hofman, D., & Razeghi, M. (2007). Substrate removal for high quantum efficiency back side illuminated type-ii InAs/GaSb photodetectors. [Article]. Applied Physics Letters, 91(23), Dier, O., Lin, C., Grau, M., & Amann, M. (2004). Selective and non-selective wet-chemical etchants for GaSb-based materials. [Article]. Semiconductor Science and Technology, 19(11), Holloway, P. H., & McGuire, G. E. (1995). Handbook of compound semiconductors: growth, processing, characterization, and devices. Park Ridge, New Jersey: Noyes Publication PAPIS-POLAKOWSKA, E. (2006). SURFACE TREATMENTS OF GaSb AND RELATED MATERIALS FOR THE PROCESSING OF MID-INFRARED SEMICONDUCTOR DEVICES (Vol. 37/38, pp. 1-34). Warszawa, Poland: Institute of Electron Technology, al. Lotników. Pils, E., Rodriguez, J. B., & Krishna, S. (2011). InAs/(In)GaSb Type II Strain Layer Superlattice Detectors: Elsevier B.V.