Boost Light Element Accuracy in WBG Semiconductors with TEM/EDS Self-Calibration

TEM/EDS analyses can be divided into two main types: qualitative and quantitative. Works of EDS qualitative analyses are to characterize elements in micro-nano regions and approximate proportions of elements involved. EDS quantitative analyses then do more calculation to make processed data approach a certain accuracy, thus materials analyzed can be identified to be what kind of compounds or alloys. In traditional textbooks and papers, adequate standard specimens were required to be analyzed at same operation parameters. Results of standard specimens were then used to calibrate data of unknown specimens.

Due to employing thin foil specimens, EDS analyses in TEM neglects effects of absorption and fluorescence, takes atomic number effect into consideration only. In recent decades, famous EDS manufacturers had developed sets of EDS software which can perform standardless quantitative analysis by using the K-factor model proposed Cliff and Lorimer ^[2] after collecting amounts of EDS data. Advantages of standardless quantitative analysis using K-factor includes dislodging purchase of standard specimens and obtaining accurate results. Now this technique has been widely used in the field of TEM/EDS analysis, and data processed by this technique have well accepted by famous journals in the field of science and engineer of materials ^[3].

In nature, geese fly in V shape by automatic alignment, as shown in Figure 1(a). The process of making metal silicide at sources, drains, and gates of MOS structures, as shown in Figure 1(b), can be completed by coating a layer of metal and adequate thermal treatment without any mask. Nano particles are arranged in order when they are put in
appropriate amount of liquid, as shown in Figure 1(c). All these examples indicate that behaviors of self-alignment exist in nature and physics. They are waiting for being explored. In TEM/EDS quantitative analyses, behavior of self-calibration was also found in some material systems from years of TEM works.

K factors of different manufacturer are slightly different. K factors of elements with atomic number larger than 11 are almost same because energies of X-rays emitting from these elements are higher than 1.0 KeV. For elements of C, N, and O, corresponding energies are lower than 0.6 KeV. K factors of these three elements vary from manufacturer to manufacturer due to absorption effect. The difference in K factors will result in large deviation in EDS quantitative analysis. Unfortunately, oxide, nitride, and carbide are all important materials for semiconductors of first, third, and fourth generations.

Materials discussed in this article are divided into two groups according to their morphology. When regions to be analyzed are random distributed, they are the type I material system (material system A), while the type II material system (material system B) are thin-film materials.

(1) Material System A

Supposing x grams of element A and y grams of element B are melted into a single-phase alloy, the weight percent of A and B in this alloy is C_A(= [x/(x+y)] x 100%) and C_B (= [y/(x+y)] x 100%) , and C_A+ C_B = 1. When a designed heat treatment is applied to this alloy, the alloy decomposes into (eutectoid) two phases, α and β. Both α and β grains are consisted of elements A and B, but have different compositions with each and the matrix. Figure 2 displays a schematic diagram of a simple model showing the microstructure of this kind of material. Under same (S)TEM operation conditions and same collection time. I_A^o and I_B^o stand for intensities of elements A and B of a region including dozens of α grains and β grains. I_A^α and I_B^α stand for intensities of elements A and B of the α grain. I_A^β and I_B^β stand for intensities of elements A and B of the β grain.

In this case, the composition of this alloy sample is known, and can be taken to be a standard. Compositions of α phase and β phase can be calculated from some simple equations list below:
C_A^α = [I_A^α / I_A^o] x C_A
C_B^α = [I_B^α / I_B^o] x C_B
C_A^β = [I_A^β / I_A^o] x C_A
C_B^β = [I_B^β / I_B^o] x C_B

Author had used this method to analyze composition of α phase and β a ternary super alloy for advisor in 1992. The calculated result has only 2 at% difference with the data reported by a paper, which is the only one at that time, of this material system. Both eutectoid materials and high-entropy alloys should be good candidates for this kind of quantitative EDS micro-analysis technique.

(2) Material system B

Figure 3 illustrates a schematic diagram of a typical multi-thin-film material. Many current semiconductor devices have similar configuration. If some layers in this thin film material have one or two elements in common, and one or two layers of thin films are stoichiometric composition, then they can be used to be standards to calibrate compositions of other layers. Firstly, a set of spectrum image is acquired by using STEM/EDS^[4], after processed by the built-in EDS program, a line profile is extracted from the processed data. If compositions of thin films are not correct, the proportion of the light element is modified by timing a constant larger than one to meet the stochiometric composition of the compound. Compositions of thin films containing this light element will be changed. An example will be given to prove the feasibility of this method.

The example given below was published in IPFA 2022^[5] and discussed briefly in iST Classroom (Read more: How to Identify & Analyze Ga2O3?). A further discussion about how to further process results obtained by TEM built-in EDS software. A STEM BF image of a cross-section sample is shown in Figure 4(a). This sample is a GaN device undergoing heat treatment in oxygen rich environment. A layer of SiO₂ was then coated upon it by magnetron of sputtering after cooling. Figure 4(a) indicates that two layers of reactants, IA and IB, were formed on the surface of GaN. Figure 4(b) ~ 4(e) are EDS maps of Ga, N, O, and Si respectively. Figure 5(a) is the corresponding STEM HAAD image of Figure 4(a). The yellow dash arrow in the STEM HAAD image marks the trace of EDS line profiles.

Figure 5(b) is the EDS line profiles extracted from the EDS spectrum image along the trace shown in Figure 5(a). EDS analysis shows that reactants IA and IB contains Ga and N only and their composition are very close. The atomic percentage ration of oxygen and gallium of the reactant calculated from a flat region in the EDS line profiles, as marked by the red dotted rectangle in Figure 5(b), is 1.23 to 1. The chemical formula of the reactant, according to this ratio, is Ga₅O₆ which is not able to be found in web search.

An article in early iST’s Classroom mentioned that X-rays of N and O were partially absorbed during escaping off the specimen (Read more: TEM EDS Analysis Inaccuracy Caused by Light Element Absorption). Intensities of X-rays of these light elements detected by the EDS detector are usually lower than those generated in the specimen. Checking data at both sides in Figure 5(b) carefully, we found that the atomic percentage of oxygen is not twice as much as silicon at the left side, and the atomic percentage of nitrogen is significantly less than that of gallium at the right side. The built-in EDS software in TEM can process signals detected, but can’t deduce the extent of absorption. So, if we accept these kinds of data without any question, there will be a big mistake in this result.

Now, let us increase the atomic percent of oxygen to be twice of silicon and the atomic percent of nitrogen to be same to that of gallium, then recalculate the atomic percent. The modified results are shown in Figure 5(c). The composition of the reactant deduced from the same region in the line profile discussed early changes to be Ga₂O₃ which is the most stable phase of gallium oxide ^[7]. Ga₂O₃ is claimed to be the semiconductor material of the fourth generation.

In this case, the thin film of unknown composition happens to be located between two layers of stoichiometric compounds, GaN and SiO2. The composition of this unknown thin film can be calculated out more accurately by correcting these two layers of known compositions to be right compositions. This kind of calibration by using neighbor phase of known composition is the key point of the TEM/EDS self-calibration quantitative analysis.