“It Happens”

The lessons we learn in this life come with a cost—be it the time we spend reading books, or the consequences from the mistakes we make. Yesterday’s lesson has to do with polishing the capsules for my experiments. The cost: I need to re-do two weeks of work, and will have less data than I’d hoped to put into the conference abstract that is coming due soon.

What happened? Well, with my experiments the first step is to seal powder with a composition similar to that of natural rocks into tiny gold capsules (2 mm outside diameter, ~5-6 mm long). These capsules get placed into a nest of MgO (which looks just like normal chalk-board chalk, but contains Mg instead of Ca), which goes into a graphite tube; the tube goes into a cylinder made of salt, and that cylinder and its contents goes into a large steel container, which is hooked up to a piston-cylinder apparatus with thermocouple in such a way that it is possible to inflict quite high pressures and temperatures upon those little gold capsules. We run these experiments for two to four weeks, and once that time has elapsed the second step is to remove the capsules from their nest, clean off any of the MgO-graphite-salt that has become stuck to the outside of the capsule, mount them into epoxy and then carefully polish the resultant disk until the capsule is at the surface, with the inside of the capsule exposed.

During the experiments the powder within the capsule undergoes chemical reactions and minerals grow (just as happens in real rocks when they are metamorphosed by enduring such pressures and temperatures in the real world). Because we start with powder the minerals within the capsule are often not as well interlocked with one another as is the case with real rocks. Therefore our normal procedure is to polish just until the uppermost surface of the gold has been removed, and then add additional epoxy, which soaks into the powder and holds it all together, letting that dry before doing the final polish in preparation for analysis with the electron microprobe. Yesterday that isn’t quite how it worked.

Before I left for AGU we “uploaded” my fourth experiment into the piston cylinder machine. That “run” completed its two weeks of mineral growth at high pressure and temperature (650° C, 22 kbars) during my absence, and my boss “downloaded” it, and left the package of salt-graphite-MgO-gold capsules-contents in my experimental drawer for me to deal with upon my return. Last week, after my return from my post-AGU holidays, I removed most of the salt-graphite-MgO layers from the now somewhat deformed gold capsules (it would be odd if they didn’t deform under that much pressure), but a small amount of it remained suck to the outside of the capsules—small crystallized bits with one edge wedged between wrinkles in the capsules surface, or merely adhered to the outside. This is normal, and our standard policy is to just ignore such small contaminants on the outside of the capsule, since our analytical methods are able to focus upon individual mineral grains within the capsule, so long as they are 1 μm or larger (remember that there are 1000 μm in every millimeter).

Therefore I dropped them off to be mounted into epoxy on Friday, and picked them up on Monday morning. I did a bit of polishing on Monday, just enough to remove the surface epoxy, revealing the tiniest bit of gold at the surface of the disks (one per capsule). Then I decided that I was really too sleepy to be engaging in such work (not only had I not yet completely adjusted back to this time zone after my flight from the US the week before, I’d also, foolishly, stayed up quite late both nights of the weekend) and I set the project aside for the day, and returned to it Tuesday evening.

In retrospect, I now believe that my brain was still not functioning at full capacity on Tuesday (no doubt do to once again, foolishly staying up late on Monday night). Why do I think that my brain wasn’t working properly? Because when I looked at the disks to be polished I could clearly see a bit of the “stuff” adhering to the outside of the capsules, so when I started polishing I told myself that I needed to polish past that “stuff” to get to the capsule itself and its contents. Therefore, each time I paused in my work to look at the capsules under the microscope I was convinced that the non-gold “stuff” I was looking at was naught more than the external contamination from the salt-graphite-MgO casing that had been around the capsules during the experiment. It never looked like the dark, powdery stuff that I’d found within the capsules of my earlier experiments, and, so, thinking I’d not yet polished deep enough to expose that powder, I kept polishing. And polishing. And polishing. Eventually, I suddenly had a gap in the gold between the top and the bottom ends of the capsule for one of them, and looking at it from the side view, saw that there wasn’t much left to the capsules at all. Confused, but undaunted (a *huge* clue that my brain wasn’t working properly), I then picked up the other capsule (I’d been switching between them as I worked, to keep them at more or less the same point in the polishing process) and proceeded to polish it, too, down past the middle of the capsule, with only the very ends remaining in the epoxy disk. Then I began to panic as I realized that if I hadn’t yet found the top of the contents, then, perhaps, what I had been thinking of as “outside contamination” was really the contents I’d been seeking. Oops. Major oops.

How was it possible to thus mistake it? Well, part of it is probably the difference in quality between this run and my earlier runs. The first couple of times I attempted to weld shut the capsules my seal was not, quite, perfect for all of them, which resulted in the water (which we put inside of the capsules along with the powder) to boil out of the capsules. Without that bit of fluid present the chemical reactions are inhibited and the minerals don’t grow as large as they do when the fluid is present. Therefore the resultant product is darker and more powdery than one obtains when the capsules are properly sealed and the fluid remains inside to participate in the chemical reaction. The hard “white stuff” I had been so carefully removing because it wasn’t the black powder I was expecting was, in fact, exactly what I wanted to see. Some combination of lack of sleep and three weeks of holiday (including the week at AGU) since last I’d done anything with these experiments conspired to make me fail to realize this on time to save any of it. Sigh.

My boss tells me that “it happens”. He did it once; our PhD student did it once. He also says that it is a mistake that doesn’t tend to get made a second time. I bet! I know that I shall be really, really paranoid about the polishing in the future to be certain that I never do this again. So now I need to go back to step 1 again and seal some fresh powder (plus water plus graphite) into new capsules, and give them a couple of weeks in the piston cylinder to grow new minerals. Then I can polish them, correctly, without throwing away the important part, and analyze the results in the microprobe. Then I can add that data to our growing database, and, perhaps, someday, one of you might compare the compositions of the minerals present in your real rocks with the data from my experiments and use that to determine at what temperature and pressure your rocks probably formed. When you do, and the list pressure-temperature combinations we tried isn’t quite as extensive as you were hoping it would be, remember just how much effort goes into obtaining each data set, and how easily things go awry making it necessary to start the process over from the beginning.

water is important for growth, even for minerals

One of the joys about the learning to run experiments process is that one gets to learn the results of both intended and unintended phenomena. In my experiments the intention is to seal powder of known composition into gold capsules along with a sufficient H₂O and graphite to ensure that the chemical reactions which take place when we elevate the pressure and temperature (to simulate what happens to rocks buried at great depth) take place in “water-saturated” conditions (which is to say there is enough water available for the growth of minerals which require water as part of their chemical formula, such as the micas). However, learning to weld the capsules is a difficult process (I’ve got a draft post on that topic just waiting for me to get photos that actually display the features I want them to show).

As a result of my welding trials and tribulations I’ve had mixed success in the “sealing” part of the above paragraph. Despite the issues with my first attempts at sealing, we ran my first experiment nonetheless, giving two samples a week and a half at elevated pressure/temperature (in this case 650 C and 25 kbars). Once they were “cooked” we had the gold capsules mounted into small disks of epoxy, then carefully polished the disks until the insides of the capsules were exposed. During the polishing stage we received our first confirmation that they had not achieved the same level of “sealed”. Apparently when properly sealed the presence of water inside the capsules ensures that the pore space in between the grains of powder are occupied, and as a result even the high pressures to which we subject them aren’t enough for the new minerals to properly interlock when they grow. As a result, while there are new crystals present, the texture isn’t very rock-like, and when polishing it is easy to accidentally remove clumps of the sample itself. This is the texture we were anticipating, and, for one of the samples run in the first experiment, this is exactly what happened. In these cases we polish only enough to just expose the inside of the capsule, then add more epoxy, letting it soak down into those pore spaces and let it dry before completing the polishing process without so much risk in losing what we are trying to polish.

However, in the other of the two samples run in the first experiment I must not have done the final welding properly, because the contents of the capsule were much harder, and held together better, meaning that the pore space was not held open with fluid when the minerals were growing. This was obvious during the polishing process, so I was able to go quite a bit deeper into the capsule (remember these are only 2 mm in diameter and about 5 mm long so “deeper” is only a relative term) before needing to add the additional epoxy.

Today we got to look at these samples in the microprobe, and as expected from the difference in their textures noted while polishing them, they are rather different from one another. The one wherein I had issues with the welding did contain some water; we know this because there are very small grains of mica present. However, neither was it water-saturated, so it lost some due to the poor seal of the capsule. It contains many, many very tiny grains of garnet (~1 micron diameter; remember that there are 1000 microns in every millimeter) which nucleated on their own, and very thin rims of garnet on the “seeds” which had been included in the powder to encourage garnet growth. The rest of the sample is even finer grained “matrix” minerals, which are going to be difficult to analyze. The other, water saturated, sample contains fewer, larger, grains of garnet, and the rims of new garnet growth on the “seeds” are much thicker than in the first sample. While it, too, is generally fine-grained, it will be easier to find single crystals large enough to get a good analysis of their compositions (which we need if we are going to accomplish our goals).

Having had this first look at the samples we’ve set the probe to create “element maps”, pretty full-colour pictures showing which areas are high (warm colours) and which areas are low (cool colours) in specific elements. Once we have these maps, we will use them to select the grains for the detailed compositional analysis. But even before we do that, I now have a better understanding of the difference between water-saturated and water-under saturated environments in terms of the ease at which minerals grow.