Month: December 2020

Improve the indexing rate – EDAX’s optimized EBSD solution

Dr. Sophie Yan, Applications Engineer, EDAX

As an applications specialist, I have encountered various problems over the years. There is always a common goal among EBSD users—to improve the EBSD indexing rate. Even a user who mainly tests relatively easy steel samples may run into deformed samples and intergranular precipitates that are difficult to calibrate, so they still need to improve the indexing rate. Ideally, we want to get a beautiful EBSD IPF map like Figure 1; however, the reality is that we often fail to get a map with such a high indexing rate.

IPF map with a very high indexing rate.

Figure 1. IPF map with a very high indexing rate.

Recently, I received a phone call from a customer asking for help. She had tricky ceramic samples with low crystallinity and fine grains, which are hard to index. The indexing success rate was only 5.48% from the area she tried to analyze (Figure 2). She wanted to see if we could improve it.

Figure 2. Ceramic sample with an indexing success rate of 5.48%.

Of course, we can.

EDAX has a set of solutions to improve the indexing rate, as shown in Figure 3. If I had a direct detector like the Clarity™ EBSD Analysis System, I would obviously get better results. However, I only have a CMOS-based Velocity™ Super in my lab.

During the data collection process, users can optimize different parameters, such as background processing, or Hough parameters, to fit real-world samples. Combined with a unique hexagonal grid sampling and triplet indexing solution, EDAX gets a better indexing rate, which is very important for challenging samples.

If the data is not ideal, we can process the result using NPAR™. With NPAR, it averages the patterns to improve the indexing rate of challenging samples considerably. Also, in OIM Analysis™ v8.0 or higher, a module is available that can perform background processing again on the saved patterns to improve the indexing rate further.

Figure 3. EDAX’s optimized EBSD solution.

I analyzed the sample and saved the patterns. Then I used OIM Analysis to post-process the patterns, as shown in Figure 4. The original pattern is quite fuzzy, and the bands were not clear. After NPAR processing, it improves the signal-to-noise ratio of the pattern, and the bands became clearer after further background processing.

Figure 4. (a) The raw pattern, (b)NPAR, (c) NPAR+dataset background, and (d) NPAR+dataset background+dynamic background.

Of course, the processed patterns have indexing success rates. Figure 5 shows the IPF map of the data after a series of post-processing steps were taken, as described in Figure 4. The indexing success rate improved to 24.1%.

Figure 5. An IPF map with an indexing success rate of 24.1%.

For this user’s case, the indexing success rate was greatly improved and was within an acceptable range. But to achieve our goal of improving the indexing rate of challenging samples, there is much more that needs to be done.

The above indexing success rates were achieved after CI >0.1 filtering. For those points with a CI <0.1 (the black areas in the IPF map), we can further process them. EDAX recently added OIM Matrix™, which includes dictionary indexing as a supplementary solution. As we all know, the result of dictionary indexing is usually better. I would expect a higher indexing success rate on the customer sample if I could use dictionary indexing to process it further.

If we push the limit, we can use the Clarity Direct Electron System to test this sample. In fact, the super-sensitive, low-beam current requirement is ideal for testing this type of sample. Maybe we can expect a better result with Clarity?

Figure 6. Will the result improve with Clarity?

The goal of improving the indexing rate can be summed up in one sentence from a Chinese poem published in roughly 300 BC: The journey is long, but I will search up and down.

提高标定率——EDAX的EBSD解决方案

Dr. Sophie Yan, Applications Engineer, EDAX

作为应用,这些年,林林总总,我碰到各种问题。但是不能否认,对于EBSD用户来说,有一个共同的追求:提高EBSD的标定率。很少有用户没有这类困扰,即便用户主要测试的是相对容易的钢铁样品,可能也会有变形试样以及难以标定的晶间析出相;理想情况,当然是得到图1那样漂亮的EBSD图;但是受限于现实情况,我们往往不能如愿。

图1:IPF图,标定率极高

最近我接到一个客户的求助电话,她有些很难的陶瓷样品,结晶程度不高,晶粒小,标定率很低。她尝试了一个区域,标定成功率只有5.48%;如下图2。她想看看有没有办法提高。

图2:陶瓷样品,标定成功率5.48%

当然我们是有办法的。

EDAX对于提高标定率有一整套解决方案,如图3所示。从硬件选择开始,比如选择最新款直接电子检测相机Clarity,那这个样品明显会有更好的结果出现; 不过我的实验室只有CMOS相机Velocity Super……

采集过程,EDAX尽可能的开放了采集过程中的参数设置,如背底处理及Hough的参数,结合特有的六方步进处理及三条带组算法,使我们在实际操作过程中,参数更加贴近实际的样品情况,对于有挑战性的样品非常重要。

如果采集的数据结果不理想,我们可以对结果进行预处理。最有代表性的当属NPAR处理,由于对花样进行平均,能极大的提高挑战性样品的标定率。此外,在OIM8.0之后,OIM Ananlysis加入了新的模块,对保存后的花样能再次进行背底处理,进一步提升标定率。

图3:EDAX的优化EBSD解决方案

我对这个样品进行了EBSD分析,保存了花样。然后用OIMA对花样进行了预处理,见下图4.原始花样相当模糊,条带不太清楚。NPAR处理之后花样信噪比提升,条带变得较为清晰。再进一步进行背底处理,条带清晰程度进一步提升。

图4(a)原始花样 (b)NPAR (c) NPAR+dataset bkg (d) NPAR+dataset bkg+dyn bkg

使用这样处理过的花样进行标定,结果当然会有所不同。下图5是我进行图4中一系列后处理后标定的IPF图。标定成功率24.1%。

图5,IPF图,标定成功率24.1%

对这个用户,整件事情告一段落:标定成功率有了大的提升,这个结果已经在可接受的范围。但是对于我们致力于提升挑战性样品标定率的目标,其实这件事还大有可为。

以上的标定成功率都是CI>0.1过滤之后的数值。对于CI<0.1的那些点,亦即IPF图中那些黑色的区域,其实我们可以对它进一步处理:EDAX近期引入了词典算法辅助标定,作为标定算法的一部分。众所周知,词典算法标定率极高,如果能用词典算法进一步标定,相信会有更好的结果。我期待着词典算法能给这个样品更高的标定率。

如果将手段用到极致——如果有可能的话,用EDAX全新发布的探测器Clarity测试这个样品,这个超灵敏,低束流的要求其实很适合测试这一类样品——可能结果也非常值得期待?

这样看来,提升标定率这一目标,也好像适用于那一句话: 路漫漫其修远兮,吾将上下而求索?

图6.清晰度会改善结果吗?