ALD of Hydrogen Diffusion Barriers
Sarah Bull – University of Colorado Boulder
Tungsten nitride and boron nitride were analyzed experimentally and computationally as barriers for hydrogen diffusion using ALD and density functional theory.
0:00:02 Hello, everyone, and thank you for attending my talk on Atomic Layer deposition or a L. D of boron 0:00:07 nitride as a hydrogen environmental barrier coating. My name is terrible and I'm coming to you from 0:00:12 Alan Wimmer's group at the University of Colorado, Boulder. I want to focus on one application, in 0:00:18 particular of these ale, the environmental barrier coatings on that's for use in nuclear thermal 0:00:22 propulsion or NTP engines. These engines are of interest because they can achieve twice the specific 0:00:28 impulse of our best chemical combustion engines, thereby having trends at times and making further 0:00:33 destinations such as a journey to Mars. More feasible possibility. These engines work by pumping 0:00:39 liquid hydrogen into a nuclear reactor core where energy released from vision is used to heat that 0:00:45 hydrogen to 2700 Calvin before pumping it out of a supersonic nozzle to generate thrust. Inside the 0:00:53 nuclear reactor core house thes sir met fuel elements. Sermet refers the fact that they're both a 0:00:57 ceramic uranium oxide and a metallic the tungsten matrix. These Dermot fuel elements are meter in 0:01:04 length and contain 123 millimeter hydrogen flow channels throughout the length. As you would imagine, 0:01:10 when you subject these materials to hot hydrogen flow issues arise as you can see from the image on 0:01:15 the left after just three hours of testing in hot hydrogen flow. But 2000 Kelvin the outside coding 0:01:21 His craft on the fuel, um in itself is fractured, and this is due to multiple causes, including 0:01:26 chemical reaction of uranium and hydrogen to form uranium hydroids thes hydride. They're less dense 0:01:32 than the surrounding material, so they will expand volumetric Lee and exert pressure on the 0:01:36 surrounding matrix that's releases plastic deformation and fracture. In addition, vacancies in the 0:01:43 tungsten lattice contract up to 12 hydrogen atoms each. These hydrogen atoms will interact and 0:01:48 recombined for bubbles, which, much like the hydroids, will expand and exert pressure on the 0:01:54 surrounding material. Other defects, such as grain boundaries and dislocation cores, will also lower 0:01:59 the ideal fracture energy and make thes substrates more susceptible to hydrogen embroilment. In 0:02:05 addition, high concentrations of hydrogen grain boundaries can stabilize surfaces and lead to extra 0:02:11 thermic cleavage by lowering the inter facial grain boundary strength. 0:02:18 One possible solution is to prevent the interaction of hydrogen with the sermon do. Actually, this 0:02:24 idea is not new. However, it has previously been investigated the use of chemical vapor deposited, 0:02:28 or CBD, taxed in cote coated particles. However, as you can see here, the CBD tungsten coding is 0:02:37 nonconforming and contains low energy diffusion pathways for hydrogen still to defuse into an attack 0:02:43 the substrate. In addition, CVD is not self limited, so it cannot be used on the outside of the Sir 0:02:49 Matt fuel elements, um, as it would clog the entrances to these hydrogen flow channels without 0:02:55 coding the length. However, we hypothesize that switching away from CBD and to L. D will allow us to 0:03:02 deposit can formal and pinhole free environmental barrier cove ing on alleviate hydrogen in brittle 0:03:07 mint in addition to coating around the field particles before they're in bed in the Matrix. We can 0:03:13 also use a lady to coat the outside of the sermon elements within the hydrogen flow channels. And 0:03:18 this is due to the self limited nature of ale de where we can see here from the image on the right, 0:03:23 uh, FNL decoding within that high aspect ratio poor shown by Becker at all. 0:03:32 We wanted to first investigate this hypothesis through use of depositing theory legacy material of 0:03:37 tungsten. And to do this, we used by turbulent, you know, by dimethyl amino, tungsten and ammonia. 0:03:43 This deposits tungsten nitride. However, because tungsten night dried is not a thermodynamic lee 0:03:48 stable material, we can easily and yield the coding in order to remove nitrogen on leave behind a 0:03:54 crystalline tungsten coding. To do this, we use the food eyes bed. It will be set up shown here 0:03:59 where we have our tungsten containing precursor in a bubble ammonia in a lecture bottle on. We dosed 0:04:06 these alternating into the fluid eyes bed where we, um where we contain our Petraeus stabilizer 0:04:13 konia substrate 300 degrees Celsius and use a mass spectrometer in order to analyze the species in 0:04:19 the outlet stream. Using the set up, we coated the Y C microspheres the 65 and 100 l d cycles. After 0:04:29 coding we used f i b milling of the thickest coding of 100 l d cycles followed by E. D. S, in order 0:04:36 confirmed the presence of tungsten within that coating. After confirming the deposition of a 0:04:42 tungsten code containing coding, we use differential thermal analysis in order to investigate the 0:04:49 effect of hydrogen on our sample. So shown here is the DT a result with exo thermic peaks shown in 0:04:55 the downward direction, and I've highlighted the three exo thermic peaks that are present within 0:04:59 this temperature range from 0 to 1600 Celsius we tested are three samples uncoated 65 100 ale 0:05:07 decoded. You dress stabilizer Kony and 6% hydrogen and argon, as well as our thickest coding of 100 0:05:14 cycles and 100% and are are gone environment in order to determine which keeps air due to hydrogen 0:05:20 versus increased thermal energy. The peaks, highlighted in pink and blue, are due thio um, 0:05:28 decomposition and crystallization of our coding and these values of 550 around 800 degrees Celsius 0:05:35 correlate well with literature where we see an initial decomposition of the tungsten nitride in 0:05:40 orderto leave behind a 2 to 1 Texans nitrogen strike geometry around 800 degrees Celsius. We have a 0:05:46 final decomposition in order to leave behind a pure crystalline tungsten coding beyond 800 degrees 0:05:53 Celsius. Our environmental barrier coating is no longer tungsten nitride but pure tungsten. 0:05:59 Highlighted in green is our hydrogen reaction with the sample we see here that as we increase our 0:06:06 cycle coding are the code the cycle count for our coding from 0 to 6500 l D cycles. We also increase 0:06:14 that peak reaction temperature from 11 99 to 12. 14 to 12 37 degrees Celsius. So we have confirmed 0:06:21 our hypothesis that we can deposit a coding with L. D in order to reduce some of the hydrogen 0:06:26 embroilment. However, because the NDP temperature is closer to 2500 degrees Celsius, we're still far 0:06:33 from that operation temperature, so two possible solutions exist. One is to increase film thickness 0:06:39 on the others. To change ABC material toe one with a lower hydrogen, diffuse it Vitti and thereby 0:06:45 still stay within those low cycle accounts. Um, in order to take advantage of the higher activation 0:06:50 energy for hydrogen diffusion, 0:06:54 we chose to look into the second possible. The second solution and that was because this ale de 0:07:00 chemistry deposit, tungsten nitride, is extremely slow. It's about two hours per cycle as well as a 0:07:06 slow growth rate per cycle. A swell. 0:07:11 Therefore, in order to positive thicker film, we would have to switch our ale de chemistry, so we 0:07:16 decided instead to look into changing ABC material. To do this, we use density functional theory in 0:07:23 order. Thio predicts the activation energy for hydrogen diffusion into the various or factory 0:07:29 materials shown here. Activation energy can easily be calculated with density. Functional theory is 0:07:35 the difference in energy between the transition state and the relaxed initial state along the 0:07:39 hydrogen diffusion pathway. As you can see here, many of the night rides that we looked at our 0:07:47 predicted to perform better than the pure metallics, tungsten and molybdenum in particular. Boron 0:07:53 nitride is predicted to perform best, and we chose to investigate further hexagonal boron nitride 0:07:59 using density functional theory. Although cubic boron nitride is predicted to perform better, um, it 0:08:05 requires high pressures and temperatures in order for crystallization, whereas with atomic layer 0:08:10 deposition, it's more likely that we'll see that hexagonal crystal structure because hexagonal boron 0:08:17 nitride is asymmetric. I wanted to look further into the various diffusion pathways for hydrogen in 0:08:22 this material, keeping in mind that the lowest energy surface is that 001 surface and therefore the 0:08:29 most likely to be present on the surface of our coated sample. we can see here that as we compared 0:08:35 to fusion through a sheet versus between these hexagonal boron nitride sheets that we see a much 0:08:42 lower diffusion activation energy down to one e V for diffusion for surface diffusion on the 0:08:50 underside of the top sheet. 1.11 for surface diffusion on the second sheet and all the way down 2.5 0:08:57 eveyone diffusing from a boron atom to a boron atom between sheets. So this means that micro 0:09:03 structure is extremely important and thes boron nitride coatings must be deposited with sheets 0:09:11 parallel to the substrate surface. In order to take advantage of that three and a quarter evey 0:09:16 activation energy and depositing codings perpendicular to the surface will allow for essentially 0:09:23 channels to be formed of for low energy diffusion pathways. Done. 2.5 TV When looking into what 0:09:32 causes the high activation energy for diffusion through a sheet versus between these streets, my 0:09:38 first conclusion was to look into Starik hindrance where we see that when a hydrogen has to diffuse 0:09:44 through the surface, it has to push out the born in the nitrogen lattice ions from its diffusion 0:09:49 pathway. However, with hydrogen diffusion between sheets, we can see that there's very little 0:09:54 movement of the brawn and the nitrogen atoms in order to compensate for the fusing hydrogen. In 0:10:01 order to look into this derrick hindrance, I calculated reconfiguration energies for these various 0:10:07 diffusion pathways at the initial. The transition in the final state on this reconfiguration energy 0:10:12 was calculated as the difference of energy between the born and the nitrogen atoms at those states 0:10:18 versus the relaxed initial structure before being in contact with hydrogen so we can see those 0:10:25 reconfiguration. Energy is shown in orange across each bar. Thes bars correlate with activation 0:10:31 energy, the same as was reported on the previous slide. So we can see here that the highest 0:10:37 reconfiguration energy correlates with the highest activation energy. Um, and from this, this 0:10:43 suggests that there's a strong correlation between reconfiguration, energy and activation energy on 0:10:49 a strong energy penalty for this increased reconfiguration energy for after looking into boron 0:10:57 nitride, computational e, I looked into it experimentally through deposition using atomic layer 0:11:03 deposition with boron, try chloride and ammonia. Shown here is my mass spec trace where we can see 0:11:08 the self limited nature of this chemistry, where we see hydrogen chloride is a by product for both 0:11:13 the boron and the ammonia dose. Um, and we can see that as that hydrogen chloride signal falls, we 0:11:20 see the increase in the precursor signal, indicating a self limited chemistry and saturation of the 0:11:26 substrate. This chemistry had a much faster deposition rate than with Thompson. I tried off about 0:11:32 three nanometers per cycle after characterization on zirconia. We used this chemistry to deposit 115 0:11:40 165 and 310 foreign nitride LD cycles on the issue a stabilizer konia the same substrate as was 0:11:48 previously used for the tungsten nitride chemistry. We again used F i b milling followed by BDs in 0:11:54 order to confirm the presence of Born and I, John and Nitrogen. Within our film shown here is the E. 0:12:02 D s results, where we can see the peak due to boron as well as a slight impurity of iron within our 0:12:07 film. In addition, we used XPS in order to analyze the elemental composition of the surface of these 0:12:14 samples. And we see here again the presence of born and nitrogen, as well as a slight chlorine 0:12:20 impurity on these iron and chlorine contaminants within the film are likely dio to that boron. Try 0:12:26 chloride precursor. Possible corroding inside of that precursor bottle or potentially picking up 0:12:33 iron from the steel fluid eyes bed during deposition. After confirming the presence of boron and 0:12:40 nitrogen within the film, we again use differential thermal analysis in order to extract the 0:12:45 behavior of these samples in a 6% hydrogen and argon environment, or 100% are gone environment. So 0:12:54 we tested all three film thicknesses 101 115 1, 65 310 as compared to that uncoated control again 0:13:02 highlighted in blue. We see here born nitride crystallization, we tried to confirm with X rd. 0:13:09 However, these films are too thin compared to the y si substrate. However, the temperatures for 0:13:16 these peaks correlate well with previous reports and literature of chemical vapor deposit, boron 0:13:22 nitride films crystallizing below 1000 degrees Celsius and chemical vapor infiltration or on nitride 0:13:28 films crystallizing between 194 and 644 degrees Celsius. Finally, the peak highlighted in green eyes 0:13:37 due to hydrogen reaction with the substrate, and it's actually only present within our uncoated y si 0:13:43 substrate on this indicates that we have quenched that peak with our with our boron nitride film on. 0:13:50 This confirms our hypothesis that we can use boron nitride l d. In order to decrease hydrogen 0:13:57 diffusion and its associate ID and brittle mint of our ailed of our NTP substrates eso we've been 0:14:04 able to push that hydrogen peak reaction temperature above the temperature range shown here from 0 0:14:10 to 1400 degrees Celsius. So this warrants further testing at higher temperatures closer to that of 0:14:16 an NDP engineer 2500 degrees Celsius. We've also confirmed our hypothesis from density functional 0:14:23 theory that boron nitrite is predicted to perform much better than tungsten nitride where I've show 0:14:29 here the a sample from that previous slide from the DT analysis of tungsten nine child where we 0:14:35 still see those peaks from the tux tonight. Drug coated samples. However, we do not see those peaks 0:14:41 in the boron nitride coated sample. As a summary, we first deposited the legacy material of tungsten 0:14:48 using L. D of tungsten. I tried and decomposition in order to form a pure tungsten coding. We then 0:14:56 use Dita and saw that the hydrogen peak reaction temperature increases with increased cycle account. 0:15:01 However, it's still far below the NTP right operating temperature. We use density functional theory 0:15:07 to predict better barrier materials, and we saw the boron nitride has shown experiment is shown to 0:15:13 perform best using dft on experimentally. Also confirmed to perform better than for than tungsten. I 0:15:20 tried. We also showed using density functional theory that diffuse city of boron nitride is 0:15:27 extremely dependent on micro structure. On. In order, Thio realize that high activation energy boron 0:15:33 nitride sheets must be peril to the substrate surface. We also saw that the large activation energy 0:15:40 of hydrogen diffusion through a born I judge she likely rises from Starik. Hindrance, a za part of 0:15:46 feature work we look into. We want to look into testing boron nitride above 1400 degrees Celsius on 0:15:53 closer to that 2500 degrees Celsius operating temperature and compare these boron nitride codings to 0:15:59 thicker tungsten codings, um used deposited from a different body chemistry than that used for 0:16:06 tungsten nightshade. With that, I would like to acknowledge both of my advisers, Alan Weimer and 0:16:12 Charles Musgrave, as well as my lab mates within those groups and funding sources, and I look 0:16:17 forward to hearing your questions