IMAGING GREEN MICROSTRUCTURES

Transcription

1 Graduate Student: Ian Maher Advisor: Dr. Richard Haber IMAGING GREEN MICROSTRUCTURES The Center for Ceramics Research at Rutgers University (CCOMC) October INTRODUCTION Different factors in the processing of compacted green body ceramics lead to microstructural variations that are difficult to account for. There is a study dedicated to determining a methodology accounting for these variations. 2. BACKGROUND Dry pressing of ceramic powders is a widely used process due to its low cost, high production rates, and shape forming abilities. Fine ceramic particles typically achieve high-density parts at lower temperatures and times due to enhanced sintering kinetics. However, fine particles do not flow due to the interparticle attractive forces, which are greater than the force of gravity. This reason is what hinders die filling uniformity and the green (non-sintered) bulk consistency of the compacted ceramic. To improve the flowability of fine ceramic particles, a granulation technique is desirable. In dry pressing, agglomerated powders are created in the form of flowable granules. A key requirement in this process is a uniform density distribution of primary ceramic granules. A desired uniform green microstructure shows no remnants of the agglomerated granules. If remnants remain, intergranular pores will persist within the microstructure after densification, thus depleting the reliability of the ceramic s properties. Large pores that persist throughout the sintered microstructure can become stress concentrators and cause failures during production. A major challenge of dry pressing ceramics is processing a flaw free green microstructure that will return a highly reliable sintered ceramic. To form reliable dry pressed ceramics, characterization and evaluation of the ceramic green microstructure is necessary. However, the major issue is an absence of sufficient analytical techniques needed to characterize the microstructure of the green compact. Uematsu et al. [1] and Saito et al. [2, 3] have been successful in developing methods using optical microscopy and confocal laser scanning microscopy (CLSM) respectively to analyze the internal structures of granules and compacted green bodies. A limitation for these techniques is the inability to characterize specimens greater than 0.5 mm [1-25]. Hondo et al. [26, 27] has developed a method to use micro-x-ray computed tomography (micro-ct) to visualize density gradients and large pore evaluation within compacted bodies. Unfortunately, this method showed it is unable to characterize microstructural variations less than one micron. It has been stated that it is impossible to prepare samples for SEM to achieve high-resolution microstructure analysis of the compacted green ceramics [3, 9, 23]. Visualization of spray-dried granules and compacted green bodies obtained by a high resolution SEM will be beneficial in understanding what process parameters influence microstructural uniformity during the compaction process. Having the capabilities to visualize green microstructural variations in three-dimensions will aid in improving the processing of ceramics. This can only benefit research areas such as modeling of binder removal through porous mediums by understanding pore channel networks and residual defect formations as a

2 function of processing parameters. Coupling the improved visualization techniques of these phenomena with experimental results, processing and formation of materials will be improved. Alumina (Al2O3) is an important material in the technical ceramics industry due to its favorable thermal, mechanical, and electrically insulating properties. Spray-drying is the most widely used granulation technique for large production of dry pressing ceramics. This candidate system and granulation process will solely be investigated for this thesis. The effects that processing parameters have during the spray-drying process on the morphology of spray-dried granules will aid in the understanding of how granule characteristics effect the compaction behavior of the granules and the rearrangement of the particles that comprise them. This will provide an understanding of what granule characteristics promote or inhibit microstructural uniformity. The processing roles to be varied will include the viscosity, the type and percentage of organic binder, and the specific gravity of the slurry. In addition, the atomization and feed rate effects during the spray-drying process will be observed. Alumina granules will be compacted at various uniaxial pressures and characterized using a field emission scanning electron microscope (FESEM) (Zeiss Sigma). Microstructural imaging will be used as a tool to visualize the compaction behavior of the spray dried powder as well as performing mechanical analysis of the compaction behavior. 3. GOAL The goal of this project centers on visualizing the particle and pore arrangements in dry pressed spray-dried alumina as a function of processing parameters to form a fundamental relationship of experimental and visualization results to improve the characterization capabilities of compacted green microstructures. The processing parameters to be controlled include the slurry properties, operational parameters of the spray-drier, and granule characteristics prior to compaction. 4. OBJECTIVES The first objective is to process spray-dried granules. A16SG Alumina (Almatis Inc.) will be the commercial powder investigated. This is the powder due to its fine and narrow particle size distribution. PVA is the chosen binder system for due to its wide use in spray-drying alumina. PEG is the chosen plasticizer for the slurry and ammonium polyacrylate (Darvan 821A) will be the chosen dispersant. This candidate system will be used to investigate and visualize the binder segregation during the spray-drying process in attempt to understand how it effects the compaction behavior and characteristics of the generated granules. The next objective will involve examining compaction behavior of the granules. Compaction behavior will be investigated as a tool to understand the crushing and adhesion characteristics of alumina granules. Understanding what governs the granule characteristics as a function of spray-drying process parameters will aid in the understanding of how the granules will behave during compaction. To visualize the alteration of processing parameters on the formation of spray-dried alumina and its impact on the compaction characteristics and microstructure of the system, spraydried alumina granules and compacted alumina will be characterized by the use of a field emission scanning electron microscope (FESEM) (ZEISS). Due to the porous microstructure of the green alumina granules and compacted samples, an infiltration epoxy that infiltrates sub-micron pores

3 will be needed. Spurr Low Viscosity Kit (Electron Microscope Sciences) epoxy resin will be used to embed the green body. This epoxy resin allows infiltration within the interparticle pores without dispersion or breakage. Once embedded in the epoxy, samples will be polished and prepared for microstructural characterization. Since our alumina samples are porous, techniques need to be implemented to mitigate charging with use of an SEM. These techniques include the use of conductive materials such as silver paste, carbon tape, and a nanolayer of sputtered gold. To fully understand microstructural variations and pore arrangements within green ceramics, visualizing three-dimensional green microstructures is necessary. Visualizing density gradients and pore channel orientations will aid the understanding of what processing effects may cause microstructural variations and uniformity. This objective will incorporate quantifying the visualization results in terms of the processing parameters performed. Understanding the occurrence of various processing phenomena three-dimensionally will aid in the understanding of better ceramic processing. For an example, exploring the interconnectivity of catalyst supports to develop a relationship between performance and microstructure would be applicable. For this research, it will involve understanding what phenomena during the slurry preparation, spraydrying, and dry-pressing of alumina granules correlate to microstructural uniformity and improved reliability of alumina ceramics. Three-dimensional visualization will aid in understanding what parameters dictate microstructural uniformity or the formation of density gradients within compacted ceramics. Large intergranular voids, pore networks, and the formation of residual defects will be explored to understand what parameters dictate different microstructural characteristics. 5. EXPERIMENTAL APPROACH 5.1. COMPACTION PLASTICITY OF SPRAY-DRIED ALUMINA GRANULES Alumina slurries were processed and milled for 24 hours [28, 29]. The alumina powder used in the slurry was A16 alumina from Almatis Incorporated. The binders used in the slurry were polyethylene-glycol (PEG) and polyvinyl-alcohol (PVA). The PEG used was PEG 300 from Acros Organics and the PVA used was a 20% aqueous solution prepared by SELVOL (SELVOL E 205 PVA) and distributed by Sekisui and St. Gobain. The dispersant used in the slurry was sodium polyacrylate (ACUMER 9400). Different percentages of the PVA binder were added at 0.75% and 1.5% on the total slurry weight and 1.35% and 2.7% based on the solids weight within the slurry [28, 29]. The PEG 300 was kept constant in all slurries at 0.15% based on the solids weight and acted solely as a plasticizer for the slurry. All the raw materials were milled for 23 hours and then more dispersant was added in the final milling hour to drop the viscosity prior to spray drying [28, 29]. The initial percentage of dispersant added for the first 23 hours was 0.3% based on the solids weight within the slurry. The additional percentage of dispersant added within the last hour of milling varied and depended on the percentage of PVA in the system and on what was the desired viscosity of the slurry before spray drying. Dispersant was added based on volumetric amounts of a 50/50 solution of ACUMER 9400 and deionized water. Different viscosities were examined prior to spray drying to determine if there was a difference in the granules compaction behavior since the characteristics of the granules depend on the process parameters and slurry characteristics during spray drying [30]. The spray dryer used for this study was a Niro Atomizer Minor Plant with a fountain nozzle. The slurries were pumped into the nozzle at a constant speed and atomized into the drying

4 chamber at a pressure of 30 psi. The inlet temperature of the spray dryer was set to 150 C with the outlet ranging from C [8, 9]. The powder was then screened through varied sized sieves to evaluate particle size analysis on the spray dried granules. The moisture of the spray dried granules varied from % moisture based on weight and no further heat treatment needed to be conducted to ensure the mechanical properties of the organic binder were governed. Compaction analysis of the powder was also conducted to determine compaction behavior of the powder. An Instron 5982 mechanical tester was used to perform the compaction curve analysis up to a force of 20 kn with a 13 mm die. The compaction rate used on the Instron was 0.05 mm per minute. The Instron returns the displacement change and the respective force measurement. From this data, density and pressure can be calculated and plotted as shown below in Figure 1 to determine the three stages of compaction. Stage one is where granules flow and rearrange, stage two is where granules begin to deform, and stage three is where granules begin to densify and join [31]. A schematic of the compaction die used is shown below in Figure 2. Figure 1. (Left) Example of a compaction curve showing the different stages during compaction. Figure 2. (Right) A schematic of the compaction die used with the Instron compressive tester. After analyzing the stages of compaction, powder was pressed at various samples to analyze the microstructure of the alumina green body at different stages of compaction. Samples were pressed using a 13 millimeter die. Alumina granules were also analyzed for visualization. Pressed alumina samples and alumina granules were then heat treated prior to microstructural analysis. Two different heat treatments were analyzed during this study. The first heat treatment ramped up to 150 C at a rate of 10 C per minute and dwelled at 150 C for two hours until it ramped back down to room temperature at the same rate of 10 C per minute. The second heat treatment ramped up to 500 C at the same rate in the previous heat treatment and dwelled for 2 hours and ramped back down to room temperature at the same rate of 10 C per minute [12, 31]. The pressed samples were then infiltrated in a low viscosity Spurr epoxy kit prepared by Electron Microscopy Sciences (EMS) under vacuum using a Buehler Cast N Vac The epoxy was heated in an oven at 70 C for fifteen minutes to lower the viscosity prior to infiltrating. The epoxy was left under vacuum for fifteen minutes to release any air within the liquids, causing the air bubbles to rise [31]. The samples were left under vacuum for thirty minutes to ensure proper infiltration into the alumina samples [31]. The samples were then cured in an oven overnight (roughly sixteen

5 hours) at 70 C. The epoxy was then polished using a Buehler mechanical polisher down to 1,200 grit pad and then continued to be polished on cloth pads down to a diamond suspension of 0.05 microns [31]. The polished samples were then coated in silver paste around every part of the epoxy except for the polished surface of the sample and sputtered with fifteen nanometers of gold using an EMS Model 150T ES sputter coater. A Field Emission Scanning Electron Microscope (FESEM) from Zeiss was used during the microstructural analysis and imaging section of the study. An InLens detector was used at an EHT of 5kV HOLLOW CORING AND PROCESSING OF ALUMINA SPRAY DRIED GRANULES Alumina slurries were processed and milled for 24 hours in 500 ml Nalgene bottles and 22 hours in 1 Liter ceramic jars [28, 29]. The main concern with changing the time was to ensure all slurries had the same milled particle size. The alumina powder used in the slurry was A16 alumina from Almatis Incorporated (Almatis Inc. Leetsdale, PA.). The binders used in the slurry were polyethylene-glycol (PEG) and polyvinyl-alcohol (PVA). The PEG used was PEG 300 from Acros Organics (Acros Organics ThermoFisher Scientific, Waltham, MA.) and the PVA used was a 20% aqueous solution prepared by SELVOL (SELVOL E 205 PVA) and distributed by Sekisui (Sekisui Secaucus, N.J.). The dispersant used in the slurry was ammonium polyacrylate (Darvan 821A). PVA binder was added at 1.50% based on the solids weight within the slurry [28, 29]. The PEG 300 was kept constant in all slurries at 0.15% based on the solids weight and acted solely as a plasticizer for the slurry. The slurries were milled for 23 hours (21 hours for ceramic jars) followed by adding more dispersant in the final milling hour to drop the viscosity prior to spray drying [28]. The initial percentage of dispersant added was 0.1% based on the solids weight within the slurry. The viscosities varied from 100 cp, 250 cp, and 400 cp (Brookfield, 20 RPM, and Spindle RV03). The specific gravity of the slurries varied from 1.48 (50% solids), 1.70 (55% solids), and 1.80 (60% solids). The additional percentage of dispersant added within the last hour of milling varied depending on the specific gravity of the slurry and the desired viscosity range. The total dispersant amounts varied from 0.25% to 0.46% based on solids weight. Two processing roles are being examined to see if there is a relationship between slurry characteristics and the characteristics of the spray dried granules, primarily, what drives the formation of hollow coring in spray dried granules. The viscosity and specific gravity of the alumina slurries varied prior to spray drying in attempt to correlate a connection to the compaction behavior of the spray dried granules. The ranges of viscosity and specific gravity mentioned above were the values analyzed for this study. The spray drying process is the same as described in section I of the alumina study, except the air pressure in the nozzle was decreased to 20 psi. After spray drying the granules were then screened through varied sized sieves to evaluate particle size analysis on the spray dried granules. The moisture of the spray dried granules varied from % moisture based on weight and no further heat treatment needed to be conducted to ensure the mechanical properties of the organic binder were governed. Granules were then heat treated at 150 C with a ramp heat of 10 C per minute. Granules were dwelled for two hours at 150 C and then cooled to room temperature, 25 C, at a rate of 10 C per minute. Granules were infiltrated using the Buehler Cast N Vac 1000 and Spurr Low Viscosity epoxy from EMS, the same process described above in section I. The same polishing process and imaging process was followed as described in section I GRAIN GROWTH OF SINTERED ALUMINA SAMPLES

6 For this study, granules that were spray dried from the same process as section II above were evaluated. Granules that were spray dried at 55% solids (specific gravity of 1.7) and a viscosity of 250 cp (Brookfield, 20 RPM, and Spindle RV03) were only used for this study. Alumina granules will be compacted at various pressures along the compaction curve to map out different stages of compaction. These samples will then be sintered in a box furnace as shown in the following diagram. The sintering profile is also described below. Figure 3. Pressure less sintering of compacted alumina samples. Ramp up to 500 C at a rate of 3 C per minute, dwell for 30 minutes. Ramp up to 1,600 C at a rate of 10 C per minute, dwell for 30 minutes. Ramp down to 25 C at a rate of 10 C per minute. Samples will be sintered on pre-sintered Almatis A16 spray dried powder discs with a second top disc placed on top of the samples The top alumina disc is to ensure that the low compacted samples do not exhibit exaggerated surface grain growth in the upwards direction. This is to hopefully keep most of the diffusion during sintering in the bulk of the sample, to correlate a relationship between compaction pressure and grain growth in the bulk and not the surface. Density values will be calculated using the Archimedes principal. Moisture used as a plasticizer before compaction will also be analyzed as a function of sintered density. The sintered alumina samples will be cut in half for infiltration and polished for microscopy imaging. This study will be conducted using granules that were spray dried at a constant viscosity of 250 cp (Brookfield, 20 RPM, Spindle RV03) and specific gravity D GREEN COMPACTED ALUMINA MICROSTRUCTURAL VISUALIZATION This study was designated towards the visualization of particle and pore arrangements in compacted alumina. A three-dimensional layering technique to obtain microstructural images using the FESEM will be developed. This will be coupled by using destructive polishing methods to calculate the depth of the polishing layer lost (z-displacement between layers). The use of the geometry of Vickers and Berkovich indents will be used. To do this, a registration material will be needed to accurately place indents around the diameter of the sample. The hope is to obtain an accuracy of one micron in the z-displacement loss to be coupled with a three-dimensional reconstruction software. In between each polishing layer, microstructure images using the FESEM will be taken. The registration material and indents will be used to find the same area of the sample with ease. The FESEM s stage navigation system will be implemented to know the x and y displacement values in between each image taken. The sample will be prepared by using conductive materials to mitigate charging. The compacted samples infiltrated in epoxy will be sputtered with nm of gold along with addition of carbon tape on the surface of the sample to mitigate charging. The surface of the carbon tape will be painted with silver to aid in charge mitigation during characterization. After obtaining two-dimensional FESEM images with accurate z-displacements, a threedimensional reconstruction software (FEI Avizo) will be used to obtain a three-dimensional image

7 of a compacted alumina sample. Images will be stitched and layered on top of one another to form the framework of high-resolution three-dimensional FESEM images of a compacted green alumina microstructure. Identifying a precise layered thickness to improve the understanding of the compacted microstructure will be conducted iteratively with software reconstruction to determine optimal layered thickness based on average particle size, granule size, and pore size diameter. This is to ensure maximum visualization of submicron particle and pore interconnectivity. h = d 2 2 tan( ) (1) Figure 4. A schematic diagram of a Vickers indent. Shown in the schematic are the diagonal values and the depth (h) calculated based on the geometry of the indent [32]. 6. RESULTS AND DISCUSSION 6.1. COMPACTION PLASTICITY OF SPRAY-DRIED ALUMINA GRANULES The compaction curves for various alumina slurries are shown below. Powder was pressed at different particle size distribution to determine the effect particle size during dry pressing has on the compaction of the spray dried powder and the microstructure within the green body. Alumina slurries were also spray dried at different viscosities to determine what roles during processing have an effect on the compaction plasticity and microstructure of the green body. Figure 5a shows the compaction behavior of alumina powder using the low binder percentage (0.75%) and high binder percentage (1.5%) of PVA superimposed on one graph. The low binder slurry was spray dried at a viscosity measurement averaging 60 centipoise using a Brookfield viscometer with an RV02 spindle at a speed of 50 RPM (Brookfield Engineering Ametek, Middleboro, MA.). The high binder percentage (1.5%) of PVA spray dried at a viscosity measurement averaging 850 centipoise using an RV05 spindle at a speed of 20 RPM. These slurries contained the same amount of dispersant at 0.3% based on solids. The compaction curves in Figure 4a were completed with a particle size distribution of the spray dried granules ranging from 75 to 150 microns (+200 and -100 U.S. Mesh size). Figure 5b shows the compaction behavior for alumina spray dried powder with the low and high PVA amount superimposed on the same compaction graph. The low binder slurry was spray dried at a viscosity measurement averaging 235 centipoise using an RV05 spindle at 50 RPM. The compaction behavior of the high PVA amount spray dried at a viscosity range averaging 100 centipoise using an RV05 spindle at a speed of 50 RPM. Compaction curves shown in Figure 2b used a certain percentage of fine and coarse powder in the particle size distribution. The particle size ranges and their respective percentages

8 are as follows: coarse powder ( microns) at 12% of the distribution, medium powder ( microns) at 74%, and fine powder (45 75 microns) at 14%. Theoretical Density (%) 60% 50% 40% 30% Low Binder High Binder µm Theoretical Density (%) 60% 50% 40% 30% Low Binder High Binder Coarse/Medium/Fine 20% (a) 20% (b) Pressure (MPa) Pressure (MPa) Figure 5. Plot shows both curves for low and high binder granules compacted using different particle size distributions superimposed on the same plot for comparison. The density for the higher binder granules is greater for both cases shown in Figures 5a and 5b. Considering Figure 5a, the first stage of compaction is almost horizontal (the vertical section of the curve before this is just noise from the Instron moving down until it hit the powder). Once the second stage of compaction starts (roughly 0.4 MPa for the low binder granules and 0.7 MPa for the high binder powder), the slope of the high binder curve becomes steeper than the low binder curve. This shows densification occurring quicker within the high binder than the low binder granules during the second stage of compaction. The slope of the low binder powder changes into the third stage of compaction around 8 MPa of pressure. The high binder granules remain constant at higher pressures, showing densification occurring at a greater rate than the low binder granules. Comparing the low and high binder granules shown in Figure 5b, during the first stage of compaction the slopes of the curves seem to be similar, with the higher binder granules showing a higher density than the low binder powder. The slopes for each curve change roughly around 0.04 to 0.5 MPa for the high binder and 0.5 to 0.6 MPa for the low binder granules. During stage two, when deformation occurs, the slope of both curves look to be very similar until the low binder curve changes the slope during the third stage of compaction (roughly around 14 MPa). The high binder curve shows a slight change in its slope but seems to continue on slope similar to stage two, showing that the higher binder granules again have a greater densification rate than the low binder granules. Below are SEM images of 2-D polished microstructure images of alumina pressed samples. Samples were pressed at 5 MPa (Figure 6), 7 MPa (Figure 7), and 100 MPa (Figure 8) using the different percentages of coarse, medium, and fine powder. Figures 6 and 7 were pressures taken from the second stage shown on the compaction curve whereas Figure 8 was taken during stage three of compaction. There was a difference in the microstructure from the two different heat treatments. The 150 C heat treatment did not seem to alter the particle arrangement in the microstructure too much within the pressed body but as you can see below, the 500 C heat treatment seemed to deform the granule deformation representation

9 a little bit greater than the 150 C heat treatment. The high binder granules showed greater adhesion and deformation as shown in Figure 7, showing the densification of the sample was greater in the case of the high binder granules. In Figure 8, the one noticeable thing of the low binder granules when compared to the high binder granules was the greater amount of pores in the microstructure. This could be due to greater adhesion of the granules during compaction exhibited by the high binder granules. (a) (b) Figure 6. SEM Heat Treatment difference shown between (a) 150 C heat treatment held for 2 hours and (b) 500 C heat treatment held for 2 hours. Both samples were low binder PVA powder pressed at a pressure of 5 MPa. (a) (b) Figure 7. SEM microstructure images of (a) low PVA binder and (b) high PVA binder pressed alumina samples at a pressure of 7 MPa, roughly 40 45% dense using the 150 C heat treatment.

10 (a) (b) Figure 8. SEM microstructure images of (a) low PVA binder and (b) high PVA binder pressed alumina samples at a pressure of 100 MPa, roughly 60 65% dense using the 150 C heat treatment. Figure 9. Images represent a 2-dimensional polished section of alumina spray dried granules. All granules shown are high binder (PVA) alumina granules.

11 Figure 9 shows 4 SEM images of alumina granules infiltrated for visualization. These images show images of mechanically polished alumina granules with spray dried using the high binder formulation heat treated at 500 C for 5 hours [12] HOLLOW CORING AND PROCESSING OF ALUMINA SPRAY DRIED GRANULES Five varying spray dried granules were analyzed in terms of morphology from the spray dryer. Their tap densities are 0.92 g/cm 3, 1.00 g/cm 3, 1.00 g/cm 3, 0.94 g/cm 3, and 0.95 g/cm 3 for 50% Solids (400 cp), 55% Solids (400 cp), 60% Solids (400 cp), 100 cp (55% solids), and 250 cp (55% solids) respectively. These values are an average of three calculations for three varying granule sizes. Tap density values show specific gravity and viscosity of the aqueous slurry promote denser granules and less hollow coring. SEM image analysis shows the variations in hollow coring for the spray dried granules and shows proof of the past statement. Hollow coring was extensive in the 100 cp (55% solids) granules compared to the rest and showed much greater hollow cores in size. 400 cp and 60% solids showed some irregular shapes but less frequent hollow coring that were smaller in size compared to the 100 cp and 50% solid granules. A few micrographs of the granules are shown below in Figures Compaction analysis was conducted in attempt to understand the effect hollow coring and granule morphology has on compaction behavior of ceramic compacts. Compaction curves of each varying granule are shown below in Figure 16. The slopes of these curves give us insight as to what is occuring. The granule yield strengths were 0.7 MPa, 0.7 MPa, 0.7 MPa, 0.7 MPa, and 1 MPa for 50% Solids (400 cp), 55% Solids (400 cp), 60% Solids (400 cp), 100 cp (55% solids), and 250 cp (55% solids) respectively. This shows that hollow coring does not affect the yield strength of the granule. The segregated organic binder layer on the surface of the granule dictates the yield strength of the spray dried granule. The greater the hollow coring of the granules, the faster the deformation process of compaction (stage II) occurs. The densification process behaves very similarly for all granules except for 250 cp. Further analysis will be conducted. SEM image analysis shown in Figures 17 and 18 below show the microstructure of a compacted alumina sample in the parallel and perpendicular directions of the compaction pressure respectively. This is another tool developed to visualize microstructure variations of ceramic compacted green bodies. Figure 10. As-received Almatis A16 Alumina.

12 Figure 11. FESEM microgrpahs of 100 cp (55% solid) granules at a sive size of 55 µm to 75 µm. Hollow core size is very large (almost at granule wall) when compared to others. Figure 12. FESEM microgrpahs of 250 cp (55% solid) granules at a sive size of 55 µm to 75 µm. Hollow core size is very large (almost at granule wall) when compared to others.

13 Figure 13. FESEM micrograph of 400 cp (55% solid) granules at a sive size of 55 µm to 75 µm. Irregular shaped granules were more abundent but smaller hollow coring was shown. Figure 14. FESEM micrograph of 400 cp (50% solid) granules at a sive size of 55 µm to 75 µm. Irregular shaped granules were more abundent but smaller hollow coring was shown.

14 Figure 15. FESEM micrograph of 400 cp (60% solid) granules at a sive size of 55 µm to 75 µm. Denser granules were more abundent but smaller hollow coring was shown. Figure 16. (Left) Compaction behavior of spray dried granule with varying slurry viscosities prior to spray drying. (Right) Compaction behavior of spray dried granules with varying specific gravity, or percent solids, prior to spray drying.

15 Figure 17. Alumina compact pressed at 100 MPa with 1.5 wt. % PVA (based on solids) spray dried at 250 cp of granule sieve size of 75 µm to 150 µm. Parallel direction to compaction pressure shown. Figure 18. Alumina compact pressed at 10 MPa with 1.5 wt. % PVA (based on solids) spray dried at 250 cp of granule sieve size of 75 µm to 150 µm. Parallel direction to compaction pressure shown. In attempt to understand processing effects on microstructural uniformity, moisture was added to specific spray dried granules to understand what role moisture plays alongside PVA binder during compaction. It is known that moisture softens the spray dried granules and lowers the Tg of the organic binder. The yield strength decreases but the hope is to visualize the microstructure to account for voids and pores. Spray dried granules in the size range of 55 µm to 75µm of the 250 cp (55% Solids) were placed in a humidity cabinet at 95% relative humidity to add moisture to the granules. Granules were compacted to 100 MPa at moisture percentages of 0.1% (Dry), 0.5%, 1%, 3%, and 6%. The mechanical properties of the PVA binder are governed under 0.5% moisture. SEM images of the varying moisture percentages are shown below in Figures

16 Figure 19. FESEM microstructure of a 100MPa uniaxial compact at dry or 0.1% moisture content. Figure 20. FESEM microstructure of a 100MPa uniaxial compact at 0.5% moisture content.

17 Figure 21. FESEM microstructure of a 100MPa uniaxial compact at 1% moisture content. Figure 22. FESEM microstructure of a 100MPa uniaxial compact at 3% moisture content.

18 Figure 23. FESEM microstructure of a 100MPa uniaxial compact at 6% moisture content. At 0.5%, the orientation of the granules can still be seen. At 1%, the boundaries of the granules are mostly gone and cannot be seen anymore. This is a visualization of the softening of the segregated binder layer of the granules. During spray drying, the PVA binder migrates to the surface of the ceramic particles during the evaporation of the water. This creates a segregated layer of organic binder that dictates the knitting and adhesion of the granules during compaction. Figure 24. FESEM microstructure of a 100MPa uniaxial compact at 0.5% moisture content before (Left) and after (right) binder burnout. The sample was heated up to 500 C at a rate of 3 C per minute and dwelled at 500 C for 2 hours before being ramped down to room temperature at a rate of 10 C per minute. The green densities of the alumina compacts for 6%, 3%, 1%, 0.5%, and dry (0.1%) are 56.9%, 56.5%, 56.1%, 56.7%, and 56.0% respectively. These densities are geometric densities calculated by the volume of the compact and weight after the compact was heat treated to remove the binder. 6% and 0.5% moisture correlate to the highest green densities, almost 57% in respect to theoretical density of alumina (3.95 g/cm 3 ). There is a drop-in density after 0.5% moisture and then slight increase until 6% moisture as shown. This shows that 0.5% moisture may be optimal

19 in terms of governing mechanical properties of the PVA organic binder but softening the binder enough to reduce voids and hollow cores remaining after compaction. SEM images of a few of these compacts before and after binder burnout can be seen in Figure GRAIN GROWTH OF SINTERED ALUMINA SAMPLES Samples have been compacted at 3.5 MPa, 10 MPa, 35 MPa, and 100 MPa. The respective percent theoretical densities calculated using Archimedes principal are 85%, 87%, 94%, and 96% to the respective pressure. Three different sintering temperatures of 1600 C, 1625 C, and 1650 C were analyzed to determine densification effect on the samples at various compacted pressures. There was no significant difference in the densities so sintering process was no longer investigated. The samples were ramped up to 500 C at a rate of 3 C per minute to remove moisture and binder. Then the samples were ramped up to 1600 C at a rate of 10 C per minute and dwelled for an hour before being ramped back down to room temperature at the same rate of 10 C per minute. The alumina samples were cut in half and polished down to 0.25 µm. The alumina was thermally etched up to 1500 C at a rate of 3 C per minute up to 500 C and then 10 C per minute up to 1500 C and finally ramped down to room temperature at the same rate. Microstructure images of the sintered samples are shown below in Figures Moisture as a plasticizer was also investigated to visualize grain growth and density as a function of moisture after uniaxial compacted alumina. FESEM comparison images are shown below in Figures 28 and 29. A table comparing density values as a function of uniaxial pressure and moisture is shown below in Table I. Density values were calculated using Archimedes method. Density increased as a function of uniaxial pressure. Adding moisture to the spray-dried alumina prior to compaction allowed the theoretical density to reach up to 98.0% at 100 MPa, an increase of roughly 1.5%, and 98.7% theoretical density at a uniaxial pressure of 150 MPa. Grain growth was investigated at various regions within the compact, but unfortunately no significant difference was noted, only a slight increase in exaggerated grain growth as uniaxial pressure is increased, as shown in Figure 30 below. Table I. Density values as a function of uniaxial compaction pressure and moisture content. Values shown are for an average of 3 samples. Moisture Content 1% 3.5 MPa 10 MPa 35 MPa 100 MPa 150 MPa 3.39 g/cm % 4% NA 3.52 g/cm % 3.71 g/cm % 3.74 g/cm % 3.76 g/cm % 3.82 g/cm % 3.82 g/cm % NA 3.85 g/cm %

20 Figure 25. Microstructure of a compact uniaxial pressed at 10 MPa and sintered at 1600 C. Figure 26. Microstructure of a compact uniaxial pressed at 35 MPa and sintered at 1600 C. Figure 27. Microstructure of a compact uniaxial pressed at 100 MPa and sintered at 1600 C.

21 Figure 28. FESEM microstructure comparison of a compact uniaxial pressed at 100 MPa with spray-dried alumina granules at 0.1% moisture (left) and 4.0% moisture (right). Figure 29. FESEM microstructure comparison of a compact uniaxial pressed at 100 MPa (left) and 150 MPa (right) with spray-dried alumina granules at 4.0% moisture.

22 Figure MPa (left) and 150 MPa (right) grain growth shown. Largest grains for 100 MPa average 10 µm. The largest grains for 150 MPa are µm and are more abundant D GREEN COMPACTED ALUMINA MICROSTRUCTURAL VISUALIZATION Initially three epoxies were tested (Buehler s TransOptic, EpoMet, and EMS Spurr Low Viscosity epoxy). The only epoxy that represented the geometry at multiple indent loads was the TransOptic epoxy. An indent using the TransOptic epoxy is shown below in Figure 31. Infiltrating the green compacted ceramic alumina had to be re-developed to account for the TransOptic Epoxy encompassing the alumina sample completely all around. In order to do this, a hole in the center of an empty TransOptic disc needed to be machined to fit the compacted sample with a tight tolerance around the compacted disc. The sample is 8.5 mm in diameter and round hole roughly 9.1 mm was machined in order to encompass the sample all around. Then the low viscosity, vacuum infiltrated epoxy was used to properly infiltrate the green compact within the machined center to infiltrate the porous microstructure to ensure a proper polish. The sample was then polished down to 0.05 µm. Vickers indents were placed as 1 and 2 kilogram loads around the sample. Image analysis was conducted on the sample and considered layer 1. The indents were used as registration marks to note where on the sample the images were being taken. Figure 31. Vickers indent on a Buehler TransOptic epoxy polished down to 0.05 µm. The challenge at this stage was not losing the indents around the sample. The indents were lost and the infiltrated sample had to be re polished to start the process over and a new layer 1 was imaged. The sample was polished starting with a 9 µm diamond pad down to a 0.05 µm diamond pad. The depth lost between layer 1 and layer 2 averaged 4.3 µm with a standard deviation of 1.1. The minimum and maximum depths calculated were 2.7 µm to 6.1 µm. This is close to the onemicron accuracy we are hoping to achieve but the attempt is to achieve it on a larger scale of hundreds of imaged layers for reconstruction purposes. Layer 1 is shown below in Figure 32 and layer 2 is shown in Figure 33.

23 Figure 32. FESEM microstructure of a compact uniaxial pressed at 175 MPa Layer 1. Figure 33. FESEM microstructure of a compact uniaxial pressed at 175 MPa Layer 2. Upon further investigation, using the TransOptic epoxy as the registration materials proved difficult. At certain stages of re-polishing the sample to determine the depth lost, the geometry of multiple Vickers indents around the sample were not consistent. This is shown in Figure 34. This phenomenon is due to the relaxation of the material after the load during indentation is removed. Equation 1 is related to the depth of the indent under load and not once the load is replaced. The actual depth is less than the projected depth due to the relaxation of the material. This is described in more detail below, showing the stiffness of the material providing the unloading curve on a load vs. displacement curve. Multiple correction factors have been addressed for both Vickers and Berkovich indents. Understanding the optimal correction factor and method to calculate depth loss will be investigated and developed over the next few months.

24 Figure 34. Polished section of Vickers indent geometry on Buehler s TransOptic epoxy. As one can see, the dvalues are not consistent, altering the depth lost calculation due to the relaxation of the material. Figure 35. Profile of indentation under load and once the load is replaced. The geometry of the indent changes based on the properties of the indented material. A new registration material had to be investigated as well. Soft aluminum and polycarbonate plastics were examined as possible materials that could be used. Polycarbonate showed the same representation as Buehler s TransOptic, excellent geometry representation but terrible consistency throughout the indentation. Multiple soft aluminums were tested that included 1100, 3003, 7075, and 2024 grade aluminum and 2024 grades were rods of various wall thicknesses that were 9.5 mm inner diameter. 1100, 3003, and 7075 were sheets of aluminum 3.2mm thick that were cut and drilled to fit the 8.5 mm compacted body within. The softer two of the group, 1100 and 3003 showed similar hardness as the infiltrated green body and therefore, a similar polishing procedure. However, the pile-up behavior after indentation was difficult to assess and overcome during the polishing investigation. The same polishing procedure was repeated twelve times to determine repeatability of depth lost. Both the 1100 and 3003 showed no correlation or repeatability in depth lost. Determining a correction factor for the shape of the indent after the load is removed was difficult. The inconsistent depth loss calculations are shown below in Figure 37. The average depth lost is shown as an average of 12 indents per sample. The standard deviation is shown in the error bars. Both metals resulted in a negative depth lost at some points,

25 Depth Loss (um) as the standard deviation range was so high. The inconsistencies of depth lost from sample to sample were difficult to assess. As the same polishing procedure was used for all twelve calculations, it is assumed to depth lost inconsistencies are a result from the relaxation of the registration aluminum after the load was displaced. As the variation in depth loss varies from 5 µm to 0.05 µm, the statistical variation will not allow us to gain the one micron accuracy needed between each polishing layer. Figure 36. Pile-up can be seen in the surface of the 3003 aluminum metal (left) and 1100 metal (right) using a Keyence VHX-5000 optical microscope Grade Al Bar

26 Depth Loss (um) Depth Loss (um) 3003 Grade Al Bar Grade Al Tube Figure 37. Depth loss calculations based on Vickers indent geometries for three aluminum registration samples grade aluminum (top), 3003 grade aluminum (middle), and 3003 grade aluminum tube (bottom). It was determined a higher hardness aluminum might be needed to assess the depth lost inconsistencies and 2024 grade aluminums were harder than the green body but still soft enough that polishing was not an issue. Both metals polished fine with no issues on the resultant green body. Aluminum 2024 showed less pile-up than the 7075 aluminum and was investigated further aluminum also took longer to polish due to its higher hardness, which can be seen in Figure 38 below comparing Vickers indents on both metals.

27 Nanoindentation was performed using Micro Materials Vantage NanoTest instrument to obtain indents of depths down to 0.5 µm. Coupling Berkovich indents with Vickers indents at depths of 5, 10, and 20 µm will be valuable in assessing an accurate depth lost while polishing. Twenty four Berkovich indents of 6 various depths were placed in a ring around the outer diameter of the sample to achieve polished layers 0.5 µm apart. These twenty four indents consisted of four indents at depths of 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 µm. Ten Vickers indents at various loads were placed around the outer diameter of the sample as well to be used as registration marks. Indents up to 1.5 µm were polished away and images of four layers were obtained using a FESEM. These four layers are assumed to be 0.5 µm apart based on the Berkovich indents. The nanoindentation software gives the user the load versus depth during the indentation and calculates the unloading curve once the load is displaced. This relaxed depth was averaged over the four indents at each load. Comparing the equipment s results with the corrected depth using the area of the indent was investigated and results are shown below in Table II. Using the corrected depths as a function of the depth polished off, the lost layering depth was assumed. Microstructure images were taken using the FESEM at each of the four layers. These images can be seen below in Figures The respective depth lost between layer 1 and 2 was 0.67 µm, the layer lost between layer 2 and 3 was 0.47 µm, and the layer lost between layer 3 and 4 was 0.43 µm. Figure 38. Table II. Optical microscope images of 7075 grade aluminum (left) and 2024 grade aluminum (right) 500 gram Vickers indents. Depth loss calculations from the Nanoindenter and calculated by the true area of the indent. Layer Depth Loss Nanoindenter Corrected Area Average FESEM Area Unloaded Depth Depth Depth Loss Layer µm 0.59 µm 0.73 µm 0.67 µm Layer µm 0.42 µm 0.50 µm 0.47 µm Layer µm 0.44 µm 0.44 µm 0.43 µm

28 Total Area: A t = 1 3 a2 sin (65.3 ) 4 Projected Area: A p = 3 4 a2 Figure 39. Total area correction for Berkovich indenter [32]. Figure 40. FESEM Layer 1

29 Figure 41. FESEM Layer 2 that is 0.67 µm from layer 1 Figure 42. FESEM Layer 3 that is 0.47 µm from layer 2 Figure 43. FESEM Layer 4 that is 0.43 µm from layer 3

30 In attempt to increase the layering of the 3-D method to reduce the time limiting step of smaller layers (0.5 µm) and to visualize density gradients throughout the entirety of the sample, larger layering depths were attempted. Vickers micro-indentation was the primary indentation method used for this study, without nano-indentation this time. Different loads of Vickers indents were chosen based on their respective loading and unloading depths calculated. Below in Table III are the indent loads chosen and their respective depths before and after stiffness correction was applied. The maximum depth value is based on the geometry of the indent but this calculation is respective of the indent under load, not when the load is replaced. Therefore the stiffness correction, or the slope of the dp/dh line of an unloading curve, was calculated for the unloaded depth [32, 33]. Stiffness: S = γ 2 E π (1 v 2 ) Area (Eq. 1) Table III. Vickers load indentation values used and their respective indent depth calculations of the maximum depth (hmax) and the unloaded depth (hfinal). The unloaded depth is the depth calculated with the stiffness correction factor explained above. Values shown are for an average of 4 indents each load and standard deviation values are shown in parenthesis. Indent Maximum Depth (hmax) Unloaded Depth (hfinal) 0.1 kg 4.96 µm (+/- 0.74%) 4.53 µm (+/- 0.88%) 0.5 kg µm (+/- 0.87%) µm (+/- 1.02%) 1.0 kg µm (+/- 0.79%) µm (+/- 0.93%) 2.0 kg µm (+/- 0.76%) µm (+/- 0.90%) 3.0 kg µm (+/- 0.78%) µm (+/- 0.92%) 10.0 kg µm (+/- 0.39%) µm (+/- 0.46%) Micro-indentation was performed using Buehler Wilson VH3300 to obtain Vickers indents. A 2024 T-3 aluminum tube was used again as the primary registration material. Twenty four Vickers indents of six various depths were placed in a ring around the outer diameter of the sample to achieve polished layers 5-10 µm apart. These twenty four indents consisted of a total of four indents at depths of 5.0, 10.0, 15.0, 20.0, 25.0 and 50.0 µm. These Vickers indents were also used as registration marks to navigate different regions of the sample. Indents were polished away and microstructure images of ten layers were obtained using a FESEM. Using the corrected depths as a function of the depth polished off, the lost layering depth was assumed. These images can be seen below in Figures

31 Figure 44. FESEM Layer 1 Figure 45. FESEM Layer 2 that is 5.55 µm from layer 1 Figure 46. FESEM Layer 3 that is 6.07 µm from layer 2

32 Figure 47. FESEM Layer 4 that is 8.57 µm from layer 3 Figure 48. FESEM Layer 5 that is 4.11 µm from layer 4 Figure 49. FESEM Layer 6 that is 4.57 µm from layer 5

33 Figure 50. FESEM Layer 7 that is 5.54 µm from layer 6 Figure 51. FESEM Layer 8 that is 2.33 µm from layer 7 Figure 52. FESEM Layer 9 that is 5.93 µm from layer 8

34 Figure 53. FESEM Layer 10 that is 2.58 µm from layer 9 Figure 54. FESEM Layer 11 that is 1.88 µm from layer 10 Using the Avizo software to stitch and align 2-D FESEM images together, a 3-D volume rendering image was achieved and can be seen below in Figure 55. Thresholding the 2-D images using pixelated greyscales and different filters to achieve similar different thresholding attributes, 3-D volume rendering images of a thresholded compacted sample was achieved and can be seen below in Figure 56.

35 Figure µm 3-D Image of an Alumina Uniaxial Dry-Pressed Compact at 175 MPa.

36 Figure µm Thresholded 3-D Image of an Alumina Uniaxial Dry-Pressed Compact at 175 MPa. The issue that arises with using larger depths is the relaxation of the material after the indentation load is replaced, and how to calculate the depth of the indent at certain points throughout the mechanical polishing process. Below in Table IV is the depth loss accuracy at different layering points and their respective standard deviations of every indent (maximum average of 24 indents). One can calculate the unloaded depth of the indent with the stiffness correction equation, however, this correction cannot work when half of the indent has been polished away. Using the geometry of the indent will represent the depth of the indent under load. The inaccuracies of depth calculation of these hardness indents are the same issues that arise with hardness indentation values and calculations, the relaxation of the material. The relaxation of the indents of the 2024 T3 aluminum registration material is shown below in Figure 57. The hope is to develop a method that can improve this inaccuracy and standard deviation of the depth loss calculated.

37 Figure 57. Relaxation profile show of the aluminum indents. Notice the curvature of the indent near the surface of the polished metal, showing the relaxation profile of the indentation.

38 Figure 58. Relaxation profile show of the aluminum indents. The top image is from a 1.0 kgf Vickers indent and the bottom indent is from a 10.0 kgf Vickers indent. The curvature and the relaxation behavior of the indents varies, showing exceedingly more curvature the larger the force. This becomes an issue when attempting to measure the depth of the indent accurately. Table IV. Depth Loss Accuracy Calculated for Each Layer Layer Depth Loss (µm) Standard Deviation µm % µm % µm % µm % µm % µm % µm % µm % µm % µm % 7. RESEARCH PLAN (Next 6 Months) More accurate compaction behavior analysis will be conducted and determined to appropriately calculate the density with respect to pressure within the compacted samples. Mercury porosimetry analysis can be used as a relationship between pressed density and pore analysis within the compacted sample. Porosity measurements on compacted samples and spray dried granules will be attempted using a mercury porosimeter. Porosity measurements will be a priority moving forward in attempt to understand what role granule porosity correlates to the compacted