Atomic Layer Deposition for Materials-Based H2 Storage:

Mg(BH4)2 as a case study

Noemi Leick –  National Renewable Energy Laboratory


To meet the requirements for vehicular solid-state hydrogen (H2) storage, novel materials such as metal borohydrides have increasingly been investigated, in particular, magnesium borohydride (Mg(BH4)2). While these materials have a high H2 capacity (> 14 wt%), poor hydrogenation-dehydrogenation cyclability and material degradation, (e.g., loss of boron), need to be overcome. Prior research has indicated that nano-encapsulation and chemical additives can address these challenges. Therefore, we pursued these two strategies simultaneously with atomic layer deposition (ALD) on Mg(BH4)2. We investigated the use of metal-oxides (e.g., Al2O3, TiO2, CeO2), Pt- group metals (e.g., Pd, Ru) as well as pulsing only one precursor molecule (e.g., Al(CH3)3, BBr3, TiCl4), and assessed these modified Mg(BH4)2 in terms of their H2 storage properties. This presentation will also present the benefits and limitations of using vapor-phase techniques to modify the properties of Mg(BH4)2.


0:00:01   Hello. My name is Noemi  Leick and I work at the National renewable energy laboratory today. I will
0:00:09   be talking to you about A. L. D. For material based hydrogen storage and we'll use magnesium boron
0:00:16   hydride as a case study.
0:00:20   This cartoon depicts the hydrogen economy and how it ties in with energy generation and demand. And
0:00:28   you can see from main applications such as transportation, agriculture industry and buildings that
0:00:37   energy storage will play a critical role in deploying hydrogen from intermittent renewable energy
0:00:43   sources. The hydrogen material advanced Research consortium, also called High Mark is focusing on
0:00:54   hydrogen storage for light and heavy duty vehicles but recently also for another class of materials
0:01:02   called carriers. And an example of that is for example to Halloween in methyl cyclo hacks ain. And
0:01:11   the idea of those is to be able to transport excess renewable energy created for example in a place
0:01:20   where there's a lot of solar to um another place where there is not that much renewable energy
0:01:28   sources um that is possible. So transporting at this energy in the form of hydrogen is very
0:01:36   appealing and those forms can also be used in long term storage applications. As we saw earlier in
0:01:43   the cartoon. When we talk about hydrogen storage technologies, we focus mainly On two types physical
0:01:53   based and material based. The physical hydrogen storage technology that is currently already applied
0:02:02   in um light duty vehicles is The 700 bar compressed gas technologies, Which has about 40 g per liter
0:02:12   of volumetric capacity. And from the Spider chart, you can see that this technology reaches almost
0:02:21   all of the D. O. E. Targets. However, we see that for the low volumetric density and the high system
0:02:29   costs, there is a true um potential for material.
0:02:36   While there are all kinds of different material used to store hydrogen, we're going to divide those
0:02:43   into two classes. The first have weak interactions with hydrogen. Here's an example of moth five and
0:02:53   high mark um published on a nickel based moth that has the world record Room temperature capacity of
0:03:03   11 g per lir. In this particular talk, we will focus on the second class of material which is
0:03:12   chemise or hydrogen to the material. So it has a strong interaction and one of this example are
0:03:20   complex hydride with sodium aluminium hydride, but we will focus on magnesium boron hydride powders.
0:03:29   For this talk that is porous and nano structure to increase the hydrogen diffusion in and out of the
0:03:36   material as well as increase the reaction rates.
0:03:42   Magnesium world hydrated has been a little bit the rock and roll star of the last five to 10 years
0:03:50   and that's because of its high hydrogen capacity that completely exceeds the GOP targets, The
0:03:56   volumetric capacity is 82 g per liter. In gravimetric is about almost 15 weight percent. Um to
0:04:05   remind you the deal we target is 40 g per liter. The idea on how this would work is um to cycle
0:04:15   magnesium borough hydrate from its hydrogenated form to magnesium bow ride and then re hydrogenated
0:04:23   this back to magnesium boron hydride by inserting hydrogen to the system. It's d Hydrogenation
0:04:31   process occurs around 300°C,, which is too high for Dont targets. The other thing is along its d
0:04:40   hydrogenation pathway and not only releases hydrogen but also some um die Brain B two H 6 which is
0:04:50   not only toxic to people but also to the fuel cell. The other thing that is detrimental is the
0:04:57   formation of B 12 H 12 that is so stable that it's viewed as a thermodynamic energy Well, um when
0:05:09   you look at the absorption side of things, 950, bar 400°C and 108 hours or energetically absolutely
0:05:20   unfeasible for applications, so there is a lot of room for improvement of the system. The approach
0:05:27   we are taking to address these challenges in this project is to use atomic layer deposition in order
0:05:35   to retain the nano structure for cycle ability um while not decreasing the gravimetric capacity of
0:05:42   magnesium boron hydride too much. But we also hope that the encapsulation will help us manipulate
0:05:49   the thermodynamic pathway of the hydrogen release, as well as mitigate material loss. Like we just
0:05:55   talked about in the form of boron depending on the material that we choose to deposit. Um If it's a
0:06:03   catalyst, we might also be able to enhance the reaction rates. In this particular talk, we're going
0:06:09   to focus in on aluminum oxide, L. D. And the process that we used was room temperature L. D. And we
0:06:18   increase the precursor exposure times to eat seconds in order to um account for the um porosity and
0:06:27   high surface area of magnesium boron hydride.
0:06:33   We're gonna talk about three things. The first is how aluminum oxide lt. So the use of both try
0:06:43   metal, aluminum and water alter the performance of magnesium boron hydride. And then we'll focus in
0:06:52   on what water alone does and what tm a alone does. We've actually screened a lot more materials than
0:07:01   just aluminum oxide here. You can see the hydrogen signal released during the heating of the
0:07:06   material. After the LT process
0:07:12   we tried syria um titania mixtures there of, we incorporated also catalyst metals like platinum and
0:07:21   palladium and you can see that aluminum oxide still has the highest low temperature hydrogen release
0:07:32   on this side. You can see the signals for just palladium and routinely um compared to the blue curve,
0:07:39   which is the uncoated and except for aluminum oxide, they don't perform much better than that.
0:07:50   One thing that will be important to highlight is aluminum oxide was the only L. D. Process that we
0:07:56   performed at room temperature. All the other ones necessitated higher temperatures than that. And
0:08:01   we'll come back to that in a little bit.
0:08:06   I'm going to start by showing you a thickness series that we did on aluminum oxide. This is the
0:08:13   temperature program description. So the hydrogen signal as a function of heating of the sample after
0:08:19   A. L. D. The encoding material um is shown in blue and we increase the number of cycles up to the
0:08:30   red signal, which is 100 cycles A. L. D. When we quantify the hydrogen signal from that, we could
0:08:38   see a big improvement from 0.6 week percent to 1.6% in the um 200 degrees Celsius range. So that's
0:08:48   very promising. Another very encouraging step was to see the dye brian signal. So this time it's di
0:08:56   brain from the uncoated material. We could see that it went down to pretty much zero when it was
0:09:03   coated with 100 cycles aluminum oxide. So all of those things were extremely encouraging. And we
0:09:10   couldn't wait to do even more cycles with that hydrogen cycles. So we did Pressure composition
0:09:19   temperature measurement with a 100 cycles L. D. Coded material. And here I am showing you deception
0:09:27   kinetics for the uncoated versus the coating materials. Where we could increase the kinetics by a
0:09:35   factor of around threat. And down here is the absorption kinetic plot where um u unfortunately you
0:09:44   can't see it that well because I'm on this. But the green plot is again the coded material and the
0:09:53   kinetics are not that much better in this case of the hydrogen absorption step. So now there are
0:10:01   more cycles. And this plot is getting a little bit messy. But on Top here, you can see the uncoated
0:10:08   materials and you can see three um D hydrogenation re hydrogenation cycles for this material and the
0:10:17   same for the coating material in green at the bottom here. And except for the first day
0:10:24   hydrogenation um cycle, the encoded encoding material performed very similarly. So that is a little
0:10:36   disappointed.
0:10:39   So the first description steps performs better with aluminum oxide. Not only from a point of view of
0:10:46   quantity of hydrogen diz orbed, but also with the kinetics. However, absorption of hydrogen still
0:10:54   remains challenging with or without aluminum oxide. LD
0:11:01   moving away a little bit from the performance of the sample. We started looking into what is
0:11:08   actually occurring with the sample um during the L. D. Process itself here you can see a porous
0:11:16   cemetery plot of the thickness series performed and um the very narrowly defined poor of eight
0:11:25   Angstrom and the need material kind of goes away as you increase the number of cycles. We can't
0:11:31   really speak of airport size distribution anymore as we increase the number of cycles. And as for
0:11:37   the specific surface area, it also decreases as a function of cycles. So that still doesn't really
0:11:44   tell us whether or not we have a coding or if it's mainly infiltration. So we did some small angle X
0:11:53   ray spectroscopy on the coding materials and focusing in on this region here, Which is the sensitive
0:12:01   one to the around 10 inches from poor region. We can see two things. Not only that this shoulder is
0:12:10   increasing in intensity but there's also a shift occurring toward lower cure numbers and all of that
0:12:18   is suggesting that um T. M. A. And water are infiltrating into the pores. And in a way that even at
0:12:28   100 cycles A. L. D. Um the pores are accommodating TME and water coming in and there is some kind of
0:12:38   swelling mechanism that is occurring. So this picture is trying to depict this in an exaggerated way.
0:12:47   So if this is magnesium bore hydrate that is encoded whether it's poor structure inside. As we look
0:12:55   at 100 cycles aluminum oxide particle, we would have some kind of coating but also some pores that
0:13:03   are completely um trapped. And we've had some swelling because even at 100 cycles um team and water
0:13:12   are still able to get into the material. So that was very interesting to us and has some um
0:13:21   consequences as to how to design um L. D. Processes. Another finding came from differential scanning
0:13:33   calorie symmetry of the neat material that exhibits to very characteristic and a thermic phase
0:13:41   transitions with comparing this with the aluminum coded sample. Which didn't really clearly show any
0:13:50   of those to enter therms and had a little bit of an extra thermic um bump over here and coupled that
0:13:58   with X ray diffraction that showed it's a loss of crystalline itty as a function of temperature. We
0:14:06   started wondering if maybe there is some in situ hydrologists reaction that is occurring as we heat
0:14:12   the samples the coded samples. So I'm now going to start um diving into the hydraulic sis reactions
0:14:22   that could be occurring. We are suggesting that there are two different steps of hydraulics
0:14:28   reactions, one occurring in A. L. D. And um another one that would release water and hydrogen during
0:14:38   the heating at the samples. So we're kind of done right now talking about the dual process using T.
0:14:47   M. A. And water. And we're really going to focus in on hydra Alice is now this data is a little bit
0:14:54   um overwhelming. So I'm going to try to walk you through as slowly as I can. This is um still mass
0:15:04   normalized mass spectrometer signals as we heat the samples after it's been exposed to Ditto and T.
0:15:15   M. A. N. D. 20 And we did different numbers of cycles. So here you see three cycles versus 100
0:15:22   cycles. And then here 10 cycles of T M. A. N. D. 20 versus 100 cycles TME and decorated water. So
0:15:33   look at the intensity as well. Um they're not all at the same scale. One thing that we see is that
0:15:42   the signal for HD is much higher after 100 cycles of D. Two exposure compared to the three cycles
0:15:51   due to exposure that is lowered when we intercollegiate T. M. A. In between the water pulses. So in
0:16:01   a more traditional aluminum oxide process, you can see that we still have um measurable amounts of
0:16:09   HD. The hydrogen coming from the sample and the deuterium coming from the deuterium exposure. Either
0:16:16   in the water pulse or only detail. So that already shows us already right here that the water paul's
0:16:27   during L. D. Is not self limiting. When we use magnesium boron hydride as it seems. When we now take
0:16:35   a look at the signals from water and die brain released during the heating of the same sample, we
0:16:43   can see that um a lot of die Broin is released With low numbers of cycles, three cycles of T. two
0:16:54   and 10 cycles of TM. And deuterium here. And when we increase those numbers of cycles, we can
0:17:00   measure a lot of water also in the Form of D. 2 0. itself. Which clearly means that hydroxyl roots
0:17:08   are left behind on the surface after L. D. And also we can see that that during these high numbers
0:17:17   of cycles almost no diaper rain is formed when we heat up the sample. So taking all of those
0:17:27   findings together and looking at hydrologists reactions that that can occur. We are suggesting that
0:17:36   during A. L. D. We already have one of the hydraulic Asus reactions that occurs which um comes along
0:17:44   with the formation of hydroxyl groups. Die Brain meaning die brain is already released during the
0:17:51   field. Which means that with the increasing number of cycles of L. D. We lose die brian. Which is
0:18:00   probably why when we heat them after, we barely see any dye brian left. And we also lose some
0:18:08   hydrogen during this L. D. Process. So this is during a L. D. At room temperature when we now go
0:18:17   over after L. D. So when we do temperature program description or any kind of other heating which is
0:18:23   necessary to D hydrogen eight our material. Um we think that this um water which comes from the
0:18:32   combination of these hydroxyl groups, is taking place this reaction we are forming boring species
0:18:40   which kind of kills are usable species. So we have material laws occur. We are releasing water which
0:18:50   is consistent with the data and we're releasing also way more hydrogen and in this reaction. So
0:19:00   unfortunately the discouraging results that we were seeing as we were cycling hydrogen in and out of
0:19:08   the material probably comes from These two reactions here.
0:19:16   And so that also explains the fact that when we did the other processes with other metal oxides that
0:19:24   required higher temperatures than room temperature. And the use of water, we actually, it looks like
0:19:32   there is less hydrogen because we lost probably way more hydrogen um during the L. D. Cycles.
0:19:39   Because in these cases we have both hydrologists reactions that are occurring during the L. D.
0:19:48   So the fact that water was involved and that we were doing L. D. At um higher temperatures is
0:19:57   explaining why we didn't see as good of a hydrogen gee hydrogenation performance in the other
0:20:05   materials. We looked at.
0:20:11   Now we're gonna be looking at T. M. A. Pulsing. Only if water is doing something maybe T. M. A. Is
0:20:20   also already reacting with magnesium for hydrated.
0:20:25   Looking at 100 cycles of T. M. A. Or I should say 100 times pulsing T. M. A. N. Magnesium boron
0:20:33   hydride. We um did TPD and DSC which you can see here at identical rates.
0:20:43   And this event a That occurs around 90°C actually corresponds to the structural phase
0:20:50   transformation of the gamma phase to the epsilon phase. But this release is very rapid, which is
0:21:01   interesting. So, already event A is telling us that T. M. A. Is reacting with magnesium boron
0:21:08   hydride. And a second indication of that is of N. C. Which corresponds to Magnesium hydride to
0:21:17   magnesium transition, which usually occurs 120 five degrees higher than what we observed here.
0:21:26   While the overall hydrogen capacity was low about 2.5 hydrogen weight percent. The rapid release was
0:21:34   really intriguing to us and we decided to test it with less TME exposure and see if we could retain
0:21:42   that.
0:21:47   So here you see uh the T. P. D. Of uncoated magnesium borough hide right here in black. Then the 10
0:21:56   cycles T. M. A. In red. And then the data I just previously showed you is now in blue. So this is
0:22:03   the um rapid release Occurring around 100°C.. And um this blue curve really has almost no
0:22:15   similarity with the uncoated material in the black curve. And that was also the case with the
0:22:23   structural information we got from X. R. D. However, the 10 cycle Tm a simple shows a lot of
0:22:32   similarities. The shape is the same and it actually decreased the onset Description temperature by
0:22:39   around 100%, 100 degrees Celsius as well as retaining its um Full hydrogen capacity. If you look up
0:22:51   to 500°C,, it still was able to deliver around 14 weight percent. And this difference here is
0:22:58   inherent to the air bars we get from TPD. So that is very encouraging. Very low numbers of pulsing
0:23:08   of T. M. A. Was able to trigger that kind of structural changes. However, it retained its um native
0:23:17   capacity and most of its structural um characteristics. This information, along with NMR and
0:23:29   infrared spectroscopy, helped us see that there is no reaction with boron and almost no
0:23:35   incorporation of aluminum containing species. Which means that um The reaction is mainly a ligand
0:23:44   exchange reaction between the b. h44 and the method group of T. M. A, which is in line with their
0:23:51   very similar venables, radi I. So overall the complete reaction that we are suggesting which is a
0:23:59   ligand exchange reaction is that the method groups go over to the magnesium and the bh we'll go over
0:24:06   to the aluminum, a traditional ligand exchange reaction. And um so basically we usually see that in
0:24:15   a traditional L. D. Process between the two precursors itself, it looks like in this particular case
0:24:24   when we use a very reactive substrate, the substrate is taking on the role as of the second
0:24:32   precursor. So now we see like an exchange reaction Just between one precursor and the substrate. And
0:24:41   we think that um as we increase these numbers of cycles, well we are close to completion of this um
0:24:49   reaction by the time we reached 100 cycles or 100 pulses. However using only 10 of those cycles, we
0:25:01   are probably only partially reacting, meaning that the magnesium still has one Bh four at least and
0:25:09   S. C. H. Three. And if we think about our infiltration model um the which could explain why there Is
0:25:17   such a drastic change in temperature between the encoded and then only 10 pulses of T. M. A. is that
0:25:25   the team they can infiltrate the entire um structure of magnesium bora hydrates. So it's a very
0:25:32   surface modification And most of the magnesium for hydra it still retains its to be H4 functional
0:25:43   groups. So all of this to say that the team a half cycle is also not self limiting,
0:25:54   wrapping up all of the data presentation. We conclude that none of the precursor pulses were self
0:26:02   limiting in the case of tri metal, aluminum and water on magnesium boron hydride. So if we transfer
0:26:12   that to more reactive substrates we can say that L. G. Does not retain its unique attributes and
0:26:20   actually becomes just in a more general term a pulsed fiber face technique which still offers a lot
0:26:29   of um promising properties.
0:26:34   So transferring what we just showed to more complex hydrates in general we recommend to use a short
0:26:42   exposures to pre closures, either in short exposure times or low number of cycles. Um one exposure
0:26:52   to one precursor can be sufficient and triggering what kind of reaction you want. You can take into
0:26:59   account. The van evolves radi i of the functional groups of your substrate and of your precursor.
0:27:07   And by matching the pore sizes of your material with the size of your leggings, you can control if
0:27:15   you want a more infiltrated or coding um dominated growth mechanism of your phone and last. We need
0:27:27   more chemistries that at the same time they're room temperature and avoid the use of water and
0:27:32   oxygen and other oxygen precursors. So I'll definitely be on the lookout for some of those more
0:27:38   exquisite chemistries. With that. I would like to thank the hydrogen storage group at enroll the
0:27:47   Department of Energy for funding high Mark and among others this research. And finally, I'd like to
0:27:53   thank you for your attention