Particle atomic layer deposition of alumina for flash sintering yttria-stabilized zirconia

Rebecca O’Toole – University of Colorado Boulder


Yttria-stabilized zirconia (YSZ) is a common electrolyte material for solid oxide fuel cells due to its moderate oxygen-ion conductivity. However, YSZ must be highly dense which conventionally requires sintering temperatures of ~1450°C. Alternatively, flash sintering can densify YSZ at furnace temperatures of 750°C in seconds, resulting in cost savings and better properties. During flash sintering, an electric field is applied across the sample and at a threshold furnace temperature, the sample conductivity and power dissipation increase causing densification. Typically, alumina is added to YSZ by mechanical mixing to alter grain growth and increase final density. Here, particle ALD was used to coat YSZ particles in thin films of alumina, homogeneously dispersing the alumina prior to flash sintering. Alumina addition increased the sample density, suggesting that alumina acts as a sintering aid. Additionally, ≤0.7 wt% alumina enhanced grain growth by dissolving into the grain boundaries to enhance zirconium cation diffusion.


0:00:01   Hi. My name is Rebecca Tool and today I'll be talking about particle atomic layer Deposition of
0:00:05   alumina for flash centering you tree stabilizer. Konia entering is a commonly used process for
0:00:12   technical ceramics production, where we typically start with our starting ceramic powder and press
0:00:18   that into the desired final shape of our part. We then take this part in, um, heat this toe high
0:00:24   temperatures typically around 1500 degrees C. To enable chemical diffusion and the removal of this
0:00:30   porosity. This gives us a dense structure with suitable properties for different applications. So
0:00:36   conventional centering requires high furnish temperatures around 1500 degrees C and takes hours to
0:00:42   complete S So this is a very energy intensive process. So there's interest in finding alternative
0:00:48   methods for centering ceramic particles toe high densities at lower furnace temperatures in shorter
0:00:53   times. One of these methods that was developed by our collaborators is called flash centering, and
0:00:59   this allows us to identify ceramic parts. Uh, furnace temperatures around 800 degrees C and uh on
0:01:06   lee takes seconds to complete, so it's considerably faster ah than conventional centering and
0:01:11   results in all significant amount of energy savings. Flash centering relies on the flow of current
0:01:18   through the sample during centering in order to heat this for sample temperature up well past that
0:01:24   of the surrounding furnace. So typical flash centering experiment is shown here where we have a
0:01:29   ceramic dog bone shaped sample hanging by two platinum electrodes. We place this sample in a furnace
0:01:36   and connect the electrodes to a power supply. The furnace will be heated to moderate temperatures
0:01:42   around 800 degrees C, and the power supply will set an electric field across the sample that will
0:01:48   drive current through the sample and the flow of current through the sample heats. The sample
0:01:52   temperature up well past that of the surrounding furnace, allowing you to achieve high density
0:01:57   ceramics that lower furnace temperatures. So the material that we're interested in flash centering
0:02:03   is eight y c or you tree stabilizer Konia. This is zirconia doped with eight MOL percent nutria, and
0:02:10   this material has a high oxygen ion conductivity. Making it useful is a solid oxide fuel cell
0:02:16   electrolyte. So most studies on the flash centering of eight y C are limited to the pure material,
0:02:23   and our group has shown previously that adding small amounts of aluminum oxide or alumina to eight y
0:02:28   C ceramics prior to conventional centering can enhance both the centering behavior and reduced grain
0:02:34   growth. So we'd like to see if the benefits that were observed during conventional centering extend
0:02:39   to flash centering as well. And specifically, our group has looked at adding alumina using particle
0:02:46   atomic layer deposition rld where the aluminum is added as an amorphous thin film on the surface of
0:02:52   the eight y c particles. So this is beneficial for centering as we know that the alumina is
0:02:58   uniformly distributed throughout the ceramic green body prior to centering, and it also gives us
0:03:05   information on where the alumina is located prior to identification, which will give us greater
0:03:10   insights into how the Illumina's impacting the dense fication behavior. So this was done for us by
0:03:17   our collaborators at Forge Nano, where they placed a twic particles and a fluid eyes better reactor
0:03:22   and then exposed the surface of the particles to try meth, aluminum and water with inert nitrogen
0:03:29   doses in between. So this would comprise one cycle of alumina ailed, and this process could easily
0:03:36   be repeated to grow thin films of amorphous aluminum of desired thickness So this was done for zero
0:03:44   through nine Ale de Cycles, or 0 to 9 Ale de Cycles, where I'll refer to these samples as zero
0:03:50   through nine. LD For the duration of my talk eso, in order to first characterized the amorphous
0:03:58   alumina film that was deposited by particle ale de we use. I see P o es to determine the aluminum
0:04:04   weight percent as a function of the number of ale de cycles, and we see that as we increase the
0:04:09   number of ale de cycles, we increase the wait percent of alumina approximately linearly.
0:04:16   Additionally, T e m and T E M E D s was performed to confirm that the alumina film is indeed on the
0:04:23   surface of the eight y C particles. We also saw that after nine cycles of alumina l. D. We have a
0:04:30   thin film that's approximately one nanometer and thickness.
0:04:35   So the type of flash centering experiment that we conducted for these experiments are called current
0:04:42   rate flash centering, and in this case we're taking a dog bone shaped powder sample that's hanging
0:04:48   by two platinum electrodes, and we place the sample in a furnace and heat the furnace up to 900
0:04:54   degrees C. We then ice a thermally holding this temperature for the duration of the experiment. Once
0:04:59   the sample temperature has equal abraded, we literally increase the current density through the
0:05:04   sample at a rate of 40 or 120 million amps per millimeter squared per minute until the total current
0:05:10   density reaches 120 million amps per millimeters squared. Where the current density is just defined
0:05:17   is the current through the sample divided by the sample cross sectional area. Eso We are performing
0:05:23   two different experiments for each sample type at two different current rates, one of which takes
0:05:28   three minutes to complete and another takes one minute to complete. So I'm showing the results of
0:05:34   these experiments here where the first thing that we were interested in measuring is the final
0:05:39   density of our sample after flash centering of both current rates. So I'm reporting here the
0:05:46   relative density as percent theoretical since as we approach approach 100% theoretical density, we
0:05:53   will have a fully done ceramic sample with no porosity. So in order to achieve better properties,
0:05:59   we'd like Thio get us close to 100% theoretical density as possible. So I'm showing here the
0:06:06   relative density a za function of the number of ale de cycles for both the 40 and 100 and 20 million
0:06:13   per millimeter squared permanent experiments. So for the 40 million per millimeter squared permanent
0:06:18   experiment, we see that all samples reach a similar final density of about 92 to 93% relative
0:06:25   density. And we don't see any significant differences in the final density of our sample. However,
0:06:31   for the 120 million per millimeter script permanent experiment, we see significant differences in
0:06:37   the final densities of our samples. Where samples where we've added alumina by particle l d have a
0:06:42   higher relative density than samples. Where, um, we're looking at the dense fication of just pure
0:06:49   eight y, you see, so we see that there may be some beneficial effect adding alumina by particle LD
0:06:56   on the dense fication behavior off eight y S. C. For the one 120 million per millimeters grade per
0:07:02   minute experiment where the samples not given us much time too dense. If I however, it's important
0:07:08   to note that the extent of dense fication is proportional to the temperature of the sample, which is
0:07:13   not something that we hold constant during these experiments and may vary across different sample
0:07:18   types. So in order to say definitively that the addition of alumina enhances the identification
0:07:25   behavior we need Thio look at a constant sample temperature during flash centering. So, ideally,
0:07:33   we'd like to do this by constructing a dense fication curve plot like the one shown here for
0:07:39   conventional centering. So this was conventionally centering pure eight y c and then eight y C with
0:07:45   small amounts of alumina added by particle L. D. And we can say that at a constant furnace
0:07:50   temperature or sample temperature theme. The samples with alumina added by particle L d have a
0:07:56   higher relative density, so we can say that the aluminum is enhancing the identification behavior.
0:08:02   So we'd like to construct an equivalent plot for our flash centering experiments. So this could be
0:08:08   done by estimating the relative density of our sample is a function of time. By placing a camera on
0:08:16   the sample during the flash centering experiment and measuring the shrinkage of the sample over the
0:08:20   duration of the experiment from shrinkage, we can calculate the relative density as a function of
0:08:25   time. Additionally, we can estimate the sample temperature by using a black body radiation model
0:08:31   where we just do a simple energy balance over the sample where the electrical energy dissipated in
0:08:36   the sample set equal to the energy required to increase the sample temperature by specific heat,
0:08:42   plus the energy that's lost his black body radiation.
0:08:47   So this plot was constructed for the flash centering experiments using the two methods shown
0:08:53   previously where we have relative density plotted as a function of sample temperature, and we do
0:08:58   indeed see that at a constant sample temperature are ailed. Decoded samples have a higher relative
0:09:03   density, suggesting that the alumina does as act as a centering aid and enhanced gentrification and
0:09:11   the diffusion mechanism that's controlling identification. So next we're interested in seeing how
0:09:18   the added alumina or aluminum added by particle A l D impacts grain growth and micro structural
0:09:24   evolution. So this was done using scanning electron microscopy where we image dense samples, uh, to
0:09:31   measure grain size to measure an average grain size after flash centering. So the results of this
0:09:38   analysis are shown here where we have grain size plotted as a function of the number of ale de
0:09:42   cycles for both current rate experiments. And we see uh, two interesting characteristics from this
0:09:49   plot, the first being that adding a small quantity of alumina by particle ale de actually increases
0:09:55   the final grain size. Um, indicating that alumina increases grain boundary mobility. Additionally,
0:10:03   we see that adding larger quantities of alumina begins to pan grain boundaries and we see a decrease
0:10:09   in grain the grain size of zero a zero r A decrease in the grain size of five and nine LD Relative
0:10:16   toe one LD Interestingly, we see this change in behavior occur at the soluble ity limit of alumina
0:10:24   into eight y c. So it's likely that at quantities less than one ailed that all of the added alumina
0:10:33   could dissolve into the NYC grain boundaries and lattice. And the addition of aluminum into the
0:10:39   grain boundaries increases the mobility of those grain boundaries, causing an increase in grain size.
0:10:44   However, if we add quantities of of alumina greater than the Saudi ability limit, we start to form
0:10:51   these bulk alumina inclusions which we observe in our micro grafts and their circled on day are the
0:10:57   dark gray grains shown here. These alumina inclusions can pin grain boundaries and make it more
0:11:04   difficulty for grain boundaries to migrate. UM, thereby decreasing the final grain size of five and
0:11:11   nine. LD relative toe one LD So thinking about both grand growth and density cation and the
0:11:19   diffusion processes that are controlling both um, both of these processes during grain growth, we
0:11:26   have the diffusion of atoms along the grain boundaries, air across the green boundaries, consuming
0:11:33   some grains and growing other grains and causing great growth to occur similarly for identification.
0:11:39   We have the diffusion of atoms along the grain boundaries to porosity to consume this porosity and
0:11:45   dense if I the part. So both grain growth and centering are dependent on the diffusion of atoms
0:11:50   along the grain boundaries. And we see that when we add small quantities of alumina, it's likely
0:11:55   that a luminous dissolving into the green boundaries at, uh at a certain concentration train jing
0:12:03   the structure of the grain boundaries and changing the um, diffusion rates of very species along the
0:12:12   grain boundaries specifically, in this case, the zirconium cat ion, which would control both
0:12:17   identification and grain growth. So it's likely that alumina is dissolving into these grain boundary
0:12:24   regions to change the chemical structure of the grain boundaries and enhance the diffusion of
0:12:29   zirconium cat ions, which would enhance the, um enhanced grain boundary mobility, increasing grain
0:12:36   size and enhanced centering. So, in summary, um, we studied the flash centering behavior of ale
0:12:45   decoded eight y c powders. So our collaborators at Forge Nano deposited thin films of amorphous
0:12:52   alumina on a twic particles, and the flash centering behavior of these particles was studied. And we
0:12:59   found that the dissolution of alumina into the grain boundary region likely enhances both grain
0:13:04   growth and identification by making it easier for zirconium cat ions to defuse along the green
0:13:11   boundaries. So I'd like to think my co author specifically Ebola and Rishi Raj from the mechanical
0:13:17   engineering department at sea you and then also Chris Gump at Forge Nano and my advisor Al Wymer,
0:13:25   and I'd be happy to take any questions