October 12th 3D HEDM Data

High Energy X-Ray Diffraction Microscopy (I believe I've mentioned it previously before) is an amazing technique which allows non-destructive 3D characterization of a microstructure. Chris Hefferan recently defended his thesis on studying the annealing of pure nickel. This is very similar to my project, although at the same time a more general understanding of whether this technique can be used to observe grain growth changes as well as any change in the grain boundary character. Of course, the most readily apparent observation is the saturation of Sigma3 boundaries in the microstructure. This was further investigated to determine if it was just the increase volume of Sigma3 boundaries in the microstructure, or if more boundaries were becoming closer in alignment to the Sigma3 nature. It turned out that the first had very little contribution, while the second played a much more major role.

I recently obtained a copy of the data of the sample in its fourth annealing state. The goal is to compare the twin morphologies found in this 3D data set to both what Yuan and I were trying to achieve, and any other data set we can find out there, and determine a scheme for classifying twins in 3D.

The .vtk file of the data is 2.5 GBs.

...

ParaView has crashed two or three times dealing with the size of this data, but the few times I have managed to get it to fully work, this thing is an absolute monstrosity. The number of grains in the sample are impressive as well as their well-defined shape prior to any smoothing/meshing. There are well over 10,000 grains in the system, which each time I attempt to threshold the sample, I watch in fear of the rainbow cursor spinning indefinitely. However, here are some preliminary attempts, in which we can clearly see we conserve the corner twin on the left and the complete twin regardless of how we approach the slice:

What's immediately interesting already is the lack of well defined, straight coherent twin boundaries. Instead, depending on both how and where we slice it, we can have different degrees of steps/facets in the 2D image. Back to more investigation!

October 8th MS&T 2012

Is being held here in Pittsburgh!

It actually started yesterday, although I was pre-occupied with sending both my father and Yuan to the airport, and the rain discouraged me from coming out again later in the evening.

Attending in-town conferences always present a difficulty in balancing between classes (which I decided to ditch), research work (some last minute SEM work), and sessions. And with all that, there's the other difficulty of picking which sessions to attend due to overlaps.

That being said, I did attend some very good and novel talks on grain boundary quantification, specifically looking at the topology of grain boundary networks as well as how to further/properly classify triple junctions from stereology. It'll be hard to find someone doing the exact same work as you (after all that is research), but there will always be similarities. However, in most cases, I find that these talks simply either reinforce what I already know, or challenge me to think a little bit harder about what my data is showing. Intrinsically, this might be a little bit of a research competition kick to think what can I show or prove better.

The talks that I come away with the most are normally those that aren't related to my research directly though. They're indirectly related in that their objective is looking for another method of characterization (typically through heavy mathematics). Although the reasoning behind this is most likely due to the fact that annealing twins have more or less been very well documented now (despite the lack of explanation still for the last hundred years). Therefore it isn't that we need more experimental work necessarily, but more ways to consider how we approach the problem and what we're looking for. I consider how these ideas can be applied to my own research then, and very generally what information it may provide that is different from what we haven't seen already.

Unfortunately I will not be doing the same courtesy as I just mentioned above, but I will still be presenting a poster tomorrow.

October 7th Reflections

I dropped off Yuan at the airport today. In another month I'll be joining him there at CEMEF to perform a series of in-situ experiments. I assume that I'll be as busy there just as he was here, and probably towards the end of the exchange just as homesick. During the last few days, Yuan and I did less experimental work, but significantly more discussion between our data, our grain encounter program, and our ideas for twin formation mechanisms. We talked about each proposed mechanism, and I mentioned that it seemed impossible during our timespan to find the answer.

He quickly interjected, "That is not our responsibility. We cannot expect ourselves to change the world. Our responsibility first and foremost as PhD students is obtain our degree through doing good research."

At this point I contemplated what good research was. Isn't good research something that answers a fundamental question? Which in our case would be determining the origin of annealing twins.

Yuan told me something that Nathalie told him:

Research is the equivalent of a forest. We are completely lost, there are no paved paths that lead us, and instead we have to look for an exit ourselves. We don't know how large the forest is, we don't know what direction is right, and we don't who (or if anyone) has came through here before. All we can do is search for our way out.

That search is the one we find in research.

There is no right or wrong answer in our journey. As long as continue to trodden into uncharted territory, then we will always be finding something new.



Continue moving forward and continue exploring.

Ciao!

September 24th Hacking It

Made an EBSD scan of another one of our 304L Stainless Steel samples which came out much better than the first time, despite going over relatively the same polishing steps. My suspect that Yuan has a significantly better cleaning technique then I do (nobody has actually really gone over what the final cleaning steps should be). That being said, there is always something to be learned.

The third 304L Stainless Steel was also prepared today. This will most likely be the specimen we use to build a 3D-dataset from. The other two were initially used to confirm whether we had a working polishing procedure or not. The last few days which we have spent trying to figure all of this out is a reminder that every lab may have very different procedures. Even something as simply as polishing may differ, or something like furnace ramp up times. That being said, I should definitely try and prepare my samples way advance in time to circumvent these issues, such that they can be loaded on the EBSD immediately and be tested. This should be a relatively easy procedure, with the exception of thinning down the sample to just 200um. Seeing one of my classmates grind down her samples to 100um, I realize that this may be a much harder task than I originally anticipated.

The actual title of this blog post again has to deal with some of issues we've been having. I've mentioned the problems with coming up with an appropriate sample holder (which was resolved by created a custom one). Over the weekend while attempting to make our scans, we noticed another issue in that for our longer samples, we didn't have a secure method to clamp the sample in the SEM. The sample holder that was available at the same time was too small to do anything. So our approach was to use a sawblade and wafer down the sides to increase the space in the middle between the clamping sidewalls. While Yuan and I tried our best, this unfortunately didn't work out as well as we wished, and will most likely be taking it the machine shop.

At some point Yuan had mentioned that this is "relatively unimportant". I thought about it and realized that I agreed completely with him. This is the work that will never be mentioned in a paper, an interview, and perhaps not even in a meeting. We should always be more concerned with the final outcome then the intermediate steps. As long as the intermediate steps are sufficient to produce the results we need, then it is okay to proceed. Recognize your problems, solve then in the easiest manner, and continue onwards.

Perfectionism will hinder research.

September 19th Last Meeting

Ended up staying in the office significantly longer than I anticipated. Although for the most part that was my fault for going on a mountain bike ride at 5:00 pm. But for the most part I view these activities as relatively constructive to my overall productivity. They enable me to refocus on the task at hand after looking at papers over and over again.

Today was also Nathalie's last day in town before leaving back for CEMEF. We had another meeting which more or less went over everything that we've discussed the last few days. This included again the current data that we have, the data that we plan on obtaining over the next two and half weeks, what we are trying to answer during this time, and how we plan to analyze all this. Overall this has been extremely motivating for me in realizing that I am doing something relatively much larger. That we are tackling a very interesting problem behind annealing twin formation, and we potentially do have solutions. At the same time, it's also provided a strong insight into how I should consider structuring my Research Performance Evaluation (RPE), which is more or less the equivalent to qualification exams.

The last couple of months, I felt extremely obsessed with attempting to come up with a mechanism for how twins form in the microstructure as my advisors had asked me to do. However, over the last few days, I've had the opportunity to look at the bigger picture again and understand that perhaps there isn't one defining mechanism. There is a possibility that multiple mechanisms may play a role to the formation, as is the case to many things we find in the field of science. This realization has allowed me to approach my topic with a more general hypothesis, which is more easily testable.

A good hypothesis should of course investigate something that we don't know, but can be tested and verified. Amusingly, the most previous form of my hypothesis has already been proven wrong by my recent results, so time for a better one.

On another note, November:

September 18th Probability Calculations

A few days ago, I posted a picture of abnormal grain growth (AGG) in the electrodeposited nanocrystalline nickel I've been working with. This picture, is not only extremely pretty, but features two unique features in the field of AGG:

• The abnormal grains are extremely faceted with {100} type planes
• The abnormal grains exhibit a high fraction of Sigma3 boundaries between them

The second point introduced an interesting mathematical calculation for us to work on. For these bi-crystals of abnormal grains that exhibit a Sigma3 boundary between them, do these arise based on random probability? Or do they occur because of some other hidden mechanism behind them...

Optical Image of Nano-Ni

To perform this calculation, we must first discretize the number of possible misorientations. Based on this value, we then determine the number of possible twin-related misorientations (note this includes both coherent and incoherent). Calculations have been done by previous researchers, which show that the probability of finding a twin-related grain is approximately 1/50, or 0.2%

Our next approach then is for a given volume of space (although I've started this calculation with just area first), determine the number of starting grains. We pick one grain, and allow it to grow abnormally. While in most cases, we assume that the grain growth occurs by consuming one grain at a time, I've taken it a "step" further, and assume that the grain grows n^2 to maintain its faceted nature.

For each step that the grain grows, it meets an increased number of neighbors. At the same time, as the grain grows, the number of grains encountered in the starting system decreases. Therefore the probability of meeting a twin-related grain (originally 0.2%) increases overtime with each growth step. When a grain reaches a certain size then, the probability that it should have met another related grain with a twin orientation more or less exceeds one.

We can approach this calculation in two ways, one is to determine the number of steps necessary before this occurs, and the second is to determine the volume necessary for a given abnormal grain size. The second has a value that can actually be determined, whereas the first shows a trend of probability over time. The preliminary calculation does seem to "agree" with some of the current experimental results.

An issue that becomes immediate obvious based on the picture to the left, is that the abnormal grain growth is NOT homogenous. In fact the grains are coarser towards the bottom and nucleate extensively around the crack.

This would seem to imply that the nucleation event and encounter with a twin-related grain is not purely statistical based on the size gradient in the sample. Unfortunately, no answer is offered at the moment, but does raise some interesting questions on how we may manipulate the "degree" of abnormal grain growth in our samples. 

September 17th Twinning Meeting

In reality, the twinning meeting was more or less a continuation of the grain boundary engineering meeting from last Friday. Yuan and I met up with Nathalie first to go over the questions that we want answered. This focused on two major aspects in the field of annealing twins:

• What do twins look like in 3-D? What morphological categories can we associate them with?
• Do twins occur with recrystallization or with grain growth? And what influence does GB velocity as well as curvature have in that case?

These questions lack the answers primarily due to lacking data, but also conflicting data. In most cases of the first, only the recent tools have come about in allowing us to find the answers we desire. These involve both high energy x-ray diffraction microscopy (HEDM) as well as the serial sectioning combined with EBSD/FIB. The case of the latter arises simply due to the fact that the original experiment premise wasn't designed for such an investigation of twinning. A primary issue that comes from grain sizes varying far too much as a result of the annealing procedures, where it has been clearly shown that grain size does play a role to twin density.

Tony arrived a little later, at which point we were going over the data we presently have all together. This includes previous work on Stainless Steel 304L, Inconel 718 (both provided by CEMEF), as well as the work on high purity nickel and nanocrystalline nickel (done at CMU). This is important in determining the next steps that we take, that is, what is it that we're trying to find out, primarily in the three weeks that Yuan is here in Pittsburgh.

We've currently concluded with two goals:

1. Develop a 3-D data set via serial sectioning (by vibromet) and EBSD to confirm the morphology of twins (at a high resolution) in either 304L or Ni
2. Streamline and unify our data from both CMU and CEMEF such that we can actually compare out values to one another, and produce a stronger overall report on what we know

I'm slowly finding out that while such research meetings feel very time-consuming and exhaustive, they are definitely very helpful in focusing on the topics of interest. Whereas some of my prior work seems to have been all over the place without a clear goal necessarily. The next few weeks will definitely be busy, but also very rewarding.

On another note, after three days of work, my code to obtain all boundaries of recrystallized grains (those including the border between recrystallized and deformed grains) now works. The next step is cleaning up the code to something... more bearable for the next user

September 15th Coding Work

In my previous posts, I mentioned the need to capture the boundaries between the recrystallized and deformed grains. Partitioning by the OIM Data Analysis software by TSL/EDAX is limiting because the only boundaries that are exported are those between the recrystallized grains. Therefore there is a significantly loss of information on all the edges bordering the deformed regions. This cause the Sigma3 content in this region to become extremely saturated due to the exclusion of the high angle boundaries with the deformed regions, as well as limit the ability to analyze misorientations at the border between deformed and recrystallized grains.

I've written a simple C++ code in the past that basically picks the reconstructed boundaries of interest simply based on GrainIDs. This was easy to achieve in the EBSD scan I was working with since I was just looking at boundaries around two abnormal grains (see the picture below). The program went through and selected all reconstructed boundaries that contained one of the abnormal grains.

Abnormal grain growth observed in electrodeposited nanocrystalline nickel annealed for 2 hours at 500C.

We can use the grain orientation spread (GOS) criterion to select for the grains of interest. The issue is that the reconstructed boundaries text file only contains GrainIDs (and this cannot be changed). We can however export a Grain File (.txt) out that contains the GrainIDs and their average GOS. The proposal (and currently being worked on) is to first analyze the Grain File for all grains of interest based on their GOS, and store the GrainIDs of all grains of interest.

Using this list, we can select for all reconstructed boundaries containing on of these grains. This works much better than before because we still get all boundaries between recrystallized grains, but also the boundaries between recrystallized and deformed grains, and thirdly, none of the boundaries between neighboring deformed grains (which would contribute to a larger fraction of random boundaries).

Off to more work now...

September 14th GBE Meeting

On Wednesday evening I picked up Yuan Jin from the airport, the PhD student working on the GBE/Formatting project at CEMEF Nice, France. The following day I finally had the opportunity to meet my future adviser Nathalie Bozzolo as well as future group mate Andrea. I suppose the word "future" is no longer appropriate seeing as we've all now met in person and had the opportunity to discuss research ideas.

Today consisted of meeting between Andrea, Yuan, Nathalie, Greg, Tony, and I, going over a variety of topics. We started on Andrea's presentation was on inhomogenous grain growth in Ni-based superalloys. The word "inhomogenous" is an interesting choice because the more research that is being done, the less "abnormal" abnormal grain growth seems to appear. In his particular work, Andrea was studying how deformation levels as well as the sub-solvus annealing temperature influences the inhomogenous grain growth. The other point of interest in these inhomogenous grains are the large number of twins present (which relate back to Yuan and I's work). This is also a very particularly interesting work because these inhomogenous grains occur with the presence of the Delta-phase, making us question the role the solute drag. Or the major question in addition to why do they occur, is also why do they stabilize and stop?

Yuan gave a presentation afterwards on some data analysis on samples prepared by Thomas, a Master's student working at CEMEF. In this study, the annealing twin density based on the average grain size distribution was studied. Grain size was manipulated by both the initial deformation level, as well as the recrystallization temperature. These samples were dynamically recrystallized and followed the trend behaviors as predicted by Pande. The purpose was to possibly look into how grain boundary velocity may be playing a role to the twinning mechanism. Tony mentioned Carl Necker's inadvertent anneal, which raises the question again of whether it is grain boundary velocity or temperature that is the contributing factor.

The rest of the afternoon consisted over going both of our paper submissions for the 5th International Recrystallization and Grain Growth Conference. Needless to say, my initial draft was picked apart by my advisers, however, that is always quite the learning experience.

Somethings that were mentioned and need to be done:

• How can we related the moving boundary plane to the the (111) twinning plane?
• How can we partition to look at the boundary of deformed and recrystallized grains from GOS?
• To follow up, what sort of behaviors do we expect to observe at the interface
• Use of DSC furnace of rapid thermal anneals of sample to cut down on time
• Probability of Sigma3 boundary occurring based on every 50 grains in a volume of space
• Observe the number fraction and length fraction during both recrystallization and grain growth and how these two regimes differ from each other
• Determine a way to measure the connectivity of twin length fractions (as discussed yesterday with Professor Rohrer)

On another note: First conference paper submitted...

Can't wait until May...

September 13th Weekly Meeting

Fraction recrystallized vs. time and the corresponding EBSD maps for 5, 10, 20, and 30 minutes

Things that were discussed with Professor Rohrer:

• Does Sigma 3 content actually increase based on fractions or just propagate/elongate with grain growth based on the length/number ratio?
• To better answer this, we should group the clusters of Sigma3 boundaries: This is achieved by taking the Sigma3 out list text file, and starting with the first line, find what endpoints it connects to (at most two more). But more importantly is to see how many more connections are formed since then. These connections should be stored as to eliminate any redundant "chains"
• Obtained CSL_filter programs to test for the behavior of Sigma 9 and Sigma 27 in more detail. Future proposed work (based on above) should be incorporate the "chain" behavior
• Other idea will be determine how this relates to the percolation theory, and secondly, how well this measurement of connectivity corresponds to Betti number / CHomP analysis
• Triple junction analysis data should be grouped into more general categories, but promising sign that the 3,3,9 fractions are increasing, and especially the coh3,incoh3,9 combination the most
• Take 10 coh3,coh3,9 boundaries, and zoom to a higher resolution and confirm that there is no such "bending" at the triple junction. Basically it is very unlikely for two fully coherent Sigma 3s to come together (but not impossible) to form the last Sigma 9 junction.

Things that were discussed with Professor Rollett:

• Re-plot graphs based on fraction recrystallized from the Hardness test measurements and not the grain orientation spread measurements
• Also plot the existing graph against fraction recrystallized rather than time to determine whether new boundaries are arising in the recrystallized areas or not
• Calculate the stored energy in the microstructure based on the GOS data and how that corresponds to grain boundary migration velocity
• Use the Avrami equation to determine both the number of recrystallizing nuclei (I) and also the rate of growth (G). How do these values compare to that which Yuan has produced?
• Is the fraction of Sigma 3 boundaries static in the deformed microstructure as expected? Also analyze how triple junction character changes in deformed area