JP2002214163A - X-ray powder diffraction pattern screening method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a screening method by measuring the X-ray powder diffraction patterns of a set of samples, sorting them in a few individual component or a factor with a standard technique, judging scores corresponding to the factor for individual X-ray powder diffraction patterns, plotting the scores, inspecting the graph of the scores, and checking a cluster, a trend, or an outliner having a possibility of indicating a new substance or defect data. SOLUTION: This screening method can be used in combination with scores for expressing the patterns in the sample set, and it comprises a step for determining a fixed number of factors by a major component analysis, a step for determining the scores of the factors on the X-ray powder diffraction patterns of the set, a step for plotting the scores in a two or more dimensional space, and a step for inspecting the plots of the scores.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to the rapid screening of multiple X-ray powder diffraction patterns, such as those made through combinatorial chemistry.

[0002]

BACKGROUND OF THE INVENTION Combinatorial chemistry is becoming increasingly used, for example, in the formation of new compounds. A large number of different compounds can be formed simultaneously, which may take days or weeks, in minutes or hours. However, with the rapid synthesis of new compounds, the task of identifying a large number of newly synthesized compounds has occurred. Chemists have long favored X-ray powder diffraction analysis techniques to identify the structure of new compounds. However, the overall identification process can be very time consuming to compare each X-ray powder diffraction pattern with a large number of known patterns in the library. Material Data In
c. While a pattern recognition or "search and match" computer program, such as Jade 5.0, available from Tektronix, helps to compare unknown X-ray powder diffraction patterns more efficiently than those in a library of known patterns, Large amount of X generated in one combination chemistry application
Line powder diffraction patterns can be unwieldy with conventional standard procedures.

[0003]

SUMMARY OF THE INVENTION The present invention focuses on using a statistical method of principal component analysis to more efficiently manage large amounts of X-ray powder diffraction patterns. The use of principal component analysis allows each X-ray powder diffraction pattern to be reduced to a set of scores that can be plotted in two or more dimensions. Inspection of the resulting plot reveals a great deal of information to chemists who are familiar with X-ray powder diffraction analysis. For example, X-ray powder diffraction patterns that are likely to correspond to the same compound or structure can be identified by the proximity of their scores in one cluster, thereby allowing, for example, search and match type The overall number of X-ray powder diffraction patterns that must be interpreted can be reduced by comparing with a library of known X-ray powder diffraction patterns using a software program. A change in the score plot may also indicate an outlier corresponding to an X-ray powder diffraction pattern that exhibits unique properties compared to the entire sample set. The chemist then pays attention to the X-ray powder diffraction pattern that is most likely to be the desired new compound without wasting time on the resources of the sample as indicated by clusters that are likely to be multiple samples of the same structure. You can concentrate. Thus, the plot reveals only a few of the multiple X-ray powder diffraction patterns that need to be further investigated.

[0004] Principal component analysis includes, for example, near infrared spectroscopy for process control equipment (US Patent Application No. 5,862,
060)). Principal component analysis has also been used to determine the concentration of controlled substances, such as heroin and cocaine, when present in a mixture with other known compounds. Minami Y, Miya
zawa K, Hida H: Advances in X-ray analysis, 27 (1996) 107-
115; see Okuyama Y, Analytical Science, 7 (1991) 941-945. Haryu M, Minkkinen P, Volkanenn J, Chemo
metrics and intelligent Laboratory Systems, 23
(1994) 341-350 discloses and predicts the ammonium nitrate solid phase transition path between IV, III, and II based on X-ray powder diffraction patterns and microcalorimetry by applying partial least squares regression and principal component analysis. I have. However, the present invention uses principal component analysis in combination with multiple X-ray powder diffraction patterns to obtain a large amount of information on samples that may vary widely. That is, the present invention contemplates a discovery method applied to a very large number of samples where a large number of known or unknown substances may be present in the sample set. Therefore, it differs from the prior art, which is limited when all substances in the sample are known apriori, or when the number of substances that may be present is very limited. ing.

It is an object of the present invention to provide a method for rapidly screening multiple X-ray powder diffraction patterns. This represents a very large number, possibly over a thousand, of the number of angular intensity data pairs present in each X-ray powder diffraction pattern, and is representative of that pattern, and is easily represented for screening purposes. The plot is achieved by reducing the number to a decimal number called a visualizable score, usually 2-5.

[0006]

SUMMARY OF THE INVENTION The present invention comprises the step of first obtaining an X-ray powder diffraction pattern of a corner member in a set of samples. Principal component analysis (PCA) is then used to extract a certain number of factors representing this data set. In connection with these factors, the PCA simultaneously corresponds to one of the above extracted factors, and at the same time a set of scores assigned to each sample along with each score representing each pattern in the sample set. Occurs. The scores for each factor are determined with respect to the X-ray powder diffraction pattern of the sample set, and the scores are plotted in two or more dimensions. The resulting plot is analyzed visually or statistically to identify clusters, trends, or outliers that may be indicative of new material, or data that may be defect data.

In a more specific embodiment of the invention, one sample subset and corresponding X-ray powder diffraction pattern can be selected. Preferably, the sample subset forms one cluster in the first plot described above. Such clustering can be identified visually or by defining those clusters using various statistical techniques. Again, a principal component analysis determines a certain number of factors that can be used in combination with the scores of those X-ray powder diffraction patterns in the sample subset. Scores for each factor are also determined for each X-ray powder diffraction pattern of those sample subsets, and those scores are plotted in two or more dimensions. As before, the resulting plot is visually inspected or statistically analyzed to identify clusters, trends, or outliers. Using the selected number of previous subsets at each iteration, using progressively smaller subsets, the resulting plots show random scattering or those to stop the occurrence of sub-clusters. The entire method can be repeated any number of times until another reason is obtained, such as chemical knowledge of the sample.

[0008]

The present invention can be applied to any combination of compounds whose structure can be analyzed by X-ray powder diffraction pattern. The nature of the chemical reaction used to create the compound to be analyzed is not important. However, the present invention is most effective when a large number of compounds are analyzed and analysis is required, such as in a combination chemistry instrument. For example, 48, 96, and in some cases, as many as 384 compounds may be produced simultaneously in one combined chemistry device. A small number of combinatorial chemistry experiments require the analysis of 1000 or more samples. A preferred analytical method commonly used to identify the structure of such samples is X-ray powder diffraction. However, X-ray powder diffraction patterns are generally complex and require considerable time and skill to interpret. Producing 1000 or more samples for individual analysis on a daily or weekly basis is a daunting task for most laboratories.

[0009] A feature of the present invention is that compounds can be analyzed using statistical analysis without the need to painstakingly interpret each individual X-ray powder diffraction pattern to determine the structural characteristics (identities) of each sample continuously. Can generate a considerable amount of information. That is, for example, if one set of samples is created in a combinatorial manner,
In the conventional procedure, each X-ray powder diffraction pattern is individually viewed, a display of the sample is made, and the next sample is moved. In a combinatorial fashion, or whenever a large number of samples are made, it is possible that many of those samples are the same or very similar or at least a mixture of a few pure compounds. As a result, a lot of time is spent interpreting the X-ray powder diffraction pattern, which shows the sample is actually the same. If you can look at hundreds of X-ray powder diffraction patterns at the same time and absorb the details, you can put together the same or very similar patterns or use them to identify the structures they represent. You will want to analyze only one of the representative patterns. However, as noted above, an X-ray powder diffraction pattern consists of hundreds of data points, which is too complex to capture tens or hundreds of patterns simultaneously.

[0010] PCA groups such samples into a much simpler task of interpreting hundreds of patterns, thereby identifying the structure of much fewer clusters identified. In addition, the identified clusters can be further analyzed to create much simpler sub-clusters of the material present in these initial clusters. The process of the present invention begins with taking an X-ray powder diffraction pattern of a set of samples. X
Line powder diffraction pattern techniques are known in the prior art and will not be described in detail here. Whinston, C, X-Ray Method
s; Richard, FEEd; Analytical Chemistry by Open Le
arning; John Wiley &Sons; New York, 1987 and X-Ray
Spectrometry, Herglotz HK, Birks, LS Ed .; Prac
ticak Spectroscopy Series, Vol. 2; Marcel Dekker; N
See ew York, 1978 for details. The x-ray technique and instrumentation used are not critical to the success of the present invention, but for a given set of samples, for each sample in that set (the intent of the analysis is It is not significant (unless it identifies instrumentational differences that contribute to the XRD pattern for the sample). The X-ray powder diffraction patterns of any of the samples are generally specific _2θ
vs. 2θ as a two-dimensional representation of the intensity of diffracted or scattered X-ray radiation. That is, 1
One axis indicates intensity and the other indicates 2θ diffraction angle. Whatever the details, each X-ray powder diffraction pattern can be viewed as one vector.

These patterns are then processed by well-known statistical techniques of principal component analysis to produce a small number of principal components or factors that reflect the dominant deviation in the expected sample X-ray powder diffraction pattern. Be sent out. In other words, major components indicating significant contributions to pattern changes are obtained from those samples by principal component analysis (or relevant statistical methods such as SIMCA or partial least squares). Then, any new samples of this set, or any new samples scanned later, can be assigned various degrees of involvement of these key components, which will most likely reproduce the pattern. The amount of each major component required is called its score, and it is these scores that are plotted for visual inspection.

Using mathematical expressions, the matrix X
Applying principal component analysis to a set of patterns, denoted by, for example, the principal component “load” P (indicating the spectral components involved) as the eigenvector of the equation (XX) P = PT, Xス コ ア Create a score matrix that is TP. For the purposes of the process envisioned in this example, only two to five key components are usually taken into account to capture data on a very wide range of compounds resulting from various chemical reactions. Can be put in. A factor of 2-5 may not be enough to reflect all of the compounds shown in the data, but the number of score dimensions can be easily visualized, touched above, and Satisfactory results have been obtained in identifying clusters and outliers for the cluster / subcluster method described in detail. Then, the X-ray powder diffraction pattern of the sample is indicated by the scores of the main components used. It is these scores that are plotted and visually inspected as described below. Thus, there is no need to perform pattern matching between a sample's X-ray powder diffraction pattern and a known library of chemical and structural patterns to obtain a significant amount of information. In fact, it is not necessary to know the nature of the sample itself.
What is important is to measure the X-ray powder diffraction pattern of the sample, identify a set of key components for that sample set, and determine and plot the score for each X-ray powder diffraction pattern. . Once the scores for each X-ray powder diffraction pattern have been determined, those scores are plotted on a graph. The number of major components used also indicates the type of plot of those scores. For example, if two principal components or factors are used, their scores are plotted on a two-dimensional graph; if three principal components are used, their scores are plotted on a three-dimensional graph; To show the dimensions of a multigraph or Visible Decis
A multidimensional visualization program such as SeeIT provided by ion can be used. PCA scores can be created and plotted using either the Pirouette provided by Informetrix or a multivariate statistical package such as UnScramber provided by Camo. Once these scores have been created,
SeeiT, Spotf from Spotfire
ile or any of a number of multivariate visualization programs such as Advanced Visual Systems.

Although the graph of scores is usually visually inspected by analysts, other algorithms can be used to analyze the clustering and patterns of the scores. A surprising amount of information can be gleaned from the graph. The closer the two scores are on the graph, the more similar the two samples are, for example, especially when using the first two or three principal components or factors. Conversely, if a unique substance is present, that is, a new compound may be present, its score is usually some distance from the main cluster, that is, the "outlier" Thus, such new substances can be detected much more easily in large samples. Or, using another expression, the needle would move away from the haystack. The scores for similar substances are summarized on the graph, and if the identity of one of the samples in the uniform cluster is known, the identity of the remaining samples in the cluster is also known.
The time and effort savings in cluster inspection can be enormous. For example, if 100 samples were subjected to the process according to the invention as described above, and the resulting score was split into three distinct clusters (further analysis would not indicate the presence of subclusters), Only three X-ray powder diffraction patterns are required for further processing using a search and match program in combination with a library of known X-ray powder diffraction patterns. First search
The results of the and match routine are extrapolated to each of the X-ray powder diffraction patterns in the cluster from which the first representative X-ray powder diffraction pattern was taken. The result of the second search and match routine is extrapolated to each X-ray powder diffraction pattern in the cluster where the second X-ray powder diffraction pattern was taken, and so on. With the time and effort required to individually identify the three X-ray powder diffraction patterns, the identity of all 100 samples can be estimated with a certain degree of certainty. Similarly, new structures can be detected by comparing the X-ray powder diffraction pattern of the sample with a known X-ray powder diffraction pattern if no correspondence can be found. New structures detected for one sample in one cluster can be extrapolated to each of the X-ray powder diffraction patterns in the cluster from which the representative X-ray powder diffraction pattern was taken.

As noted above, score plots can be examined visually or using a statistical pattern recognition algorithm to determine characteristics such as clusters and outliers. An appropriate statistical technique is density mapping to examine K recent neighbors, Mahalanobis distances, and their properties. Still other clustering methods that can be used in examining score plots include Jardin and Sibson node analysis, Forsy centroid, MASLOC center type method, fuzzy clustering, minimum spanning tree method, and McQueen's K-means method And so on. Chemometric: A Textbook, DL et a
See l., Elsevier, 1988pp, 371-383, 392-399, 431.

Known samples can be included as part of a sample set to help identify the structural properties of the samples in a cluster. The known sample is analyzed with the remaining samples as described above.

The location of the score on the score graph will help identify the composition represented by the cluster or outlier. For example, if a known sample is in one well-defined cluster, it is that the remaining samples in that cluster also have structural features very similar to the structure of the known sample. Is a good indicator of This is because the X-ray powder diffraction patterns from the known material are very consistent with the X-ray powder diffraction patterns for the other samples in the cluster. As a result, if the scores for the known samples are calculated for the cluster samples, they will be very similar. Equally important is that the analyst can focus on those samples whose plotted scores do not fit into any of the clusters. These data points are called "outliers." Outliers are typically new materials or defect data.
In any case, these small outliers can be examined in more detail and most of their samples can be assigned to known categories. Again, the time and effort savings for the analyst can be substantial. Identifying and concentrating on the X-ray powder diffraction patterns that are most likely to indicate a new material avoids wasting resources on X-ray powder diffraction patterns associated with less promising materials.

A plot of these scores may indicate a transition from one phase to another within one sample, or the presence of more than one phase. Thus, depending on the degree to which the X-ray powder diffraction pattern is additive to the mixture of pure powders, the score of the mixture will also show a linear average of the scores of those pure substances weighted by their respective proportions. . For example, the score for a mixture of 50% A and 50% B will be approximately halfway between the scores indicating pure A and pure B. Conversely, the score for a sample composed of 20% A and 80% B will be between the two pure substance scores, but only 5 times closer to A than B . The score of the mixture of the three pure phases will lie on the plane formed by the scores of the three pure phases.

Additional information can be obtained by repeating the analysis portion of the above method one or more times on successively smaller subsets of samples. For example, in one specific embodiment of the invention, clusters can be shown on a graph of scores. The samples that make up the cluster are selected as one subset, and principal component analysis can be applied to only that subset. It is not necessary to generate an additional X-ray powder diffraction pattern, but only to apply principal component analysis to the first generated pattern. The major components that are largely responsible for pattern changes within this subset are principal component analysis (or
(SIMCA or partial least squares). The PCA score for each sample in the subset is determined and plotted as described above. It then examines the score graph for clustering (which may indicate transitions or multiple phases), outliers, or trends. Using principal component analysis and plotting the scores of a smaller subset of the sample may reveal previously undetected differences within that subset. In other words, for a complete sample set, the clusters may have appeared to be very relevant, but a closer inspection of the samples that make up the cluster would show that subclusters and other Useful information may be revealed. More information can be gained by repeating such operations on progressively smaller subsets of the sample being analyzed, and such operations can be continued until no useful additional information is available.

The point at which the work is stopped may be when the analyst no longer finds a difference between the overlapping patterns in the cluster, or because of the uniform density or randomness of the scores when plotted. You may decide. To aid in determining whether a cluster should be decomposed into multiple sub-clusters, the overlap of all X-ray powder diffraction patterns corresponding to the samples making up that cluster can be examined.
If the overlap of all patterns indicates a difference, sub-clustering only this cluster, applying principal component analysis, and plotting the scores as described above may provide more detailed information.

Without limiting the scope of the invention, examples of the invention are set forth below, with specific conditions applied to specific embodiments of the invention, for illustrative purposes only. These embodiments clearly illustrate the method of the method described herein and its advantages.

Example 1 Three different zeolites were synthesized and a mixture of the zeolites was prepared. This set of samples included samples having zeolite weight ratios as shown in Table 1. Zeolite marking follows the standards set by the Structural Committee of the International Zeolite Association. Meier, W.
M .; Olson, DH; Baerlocher Atlas of Zeolite Struct
_{ure Type, 4 th Revised Ed .;} Rees, V, C., von Ballmoo
s, R. Eds; published for the Structure Committee of the International Zeolite Association, Elsevier; New York, 1996, pp 104-105 and pp1.
See 30-133. FAU is Faujalight, LTL is Linde
Type L and LTA indicate Linde type A.

[0022]

[Table 1]

Scintag Theta-Theta with copper radiation source
Was used to obtain an X-ray powder diffraction pattern for each sample in the sample set. 1A and 1B show the X-ray powder diffraction patterns of these samples, and it is difficult to find any correlation when viewing the X-ray powder diffraction patterns individually. The X-ray powder diffraction pattern was analyzed using principal component analysis to determine three factors. From these factors, a score for each sample in the sample set was calculated and plotted on a three-dimensional graph. FIG. 2 is a phase diagram marked with the sample composition used. FIG. 3 shows a plot of the score calculated for the corresponding sample. A comparison of FIGS. 2 and 3 clearly shows that the plots of the scores are very similar to the state diagram. In FIG. 3, the plotted score pattern is a triangle, and the pure phase is at the corner of the triangle. The scores of the remaining samples also settled within the triangle, indicating the various mixtures and the approximate amount of each zeolite, thus indicating that principal component analysis can be successfully used to analyze the X-ray powder diffraction pattern. I have. In other words, PCA can not only analyze samples composed of pure phases, but also samples containing various mixtures of these pure phases. In addition, the scores of these mixtures are seen to be on the line connecting the scores of the pure phases.

Example 2 Blank and zeolite FAU and LTA standard samples were analyzed with X-ray powder diffraction patterns to obtain X-ray powder diffraction patterns using standard X-ray powder diffraction techniques. The diffractometer used was Br
It was ucker AXS D8 Advance. The radiation source is high intensity X
The wire tube was operated at 40 kV and 40 ma. Diffraction patterns from copper K-alpha radiation were obtained by using appropriate computer techniques. 3.6 The sample a force is applied to the horizontal direction ^{0 (2θ)} / min to 5 ⁰ -40 ⁰ (2
θ). Principal component analysis was applied to all X-ray powder diffraction patterns to determine four factors. Scores for each of the analyzed samples and blanks were calculated and plotted in a two-dimensional graph. A plot of the scores of the first two factors is shown in FIG. By visually inspecting FIG. 4, the presence of three score clusters can be easily identified. One cluster was identified as FAU, another as LTA, and the third as blank.

The blank clusters appeared visually elongated. Therefore, the principal component analysis is performed for X corresponding to the blank in FIG.
Performed on a line powder diffraction pattern, four factors were determined. Factor scores for each of the analyzed blanks were calculated and plotted on a two-dimensional graph. A plot of the scores obtained for the first two factors is shown in FIG. By visual inspection of FIG. 5, three clusters of scores in the blank sample could be easily identified. The overlaid X-ray powder diffraction patterns for each of the three clusters of FIG. 5 are shown in FIG. By examining the cluster of FIG. 5, the analyst was able to determine that the deviations in the background material and plate alignment used to load the sample resulted in a variety of different X-ray powder diffraction patterns for blanks. Could be determined. Thus, PCA not only can distinguish between different sample compositions, but also can identify incorrect data and isolate the source of such errors.

Example 3 Hydrothermal properties of a Zn—Cu—V—O system were examined using a combination method. Selected formulations were examined under various reaction conditions. One specific synthesis condition is at 150 ° C.
Digestion for another day, another reaction condition was digestion at 200 ° C. for 8 days. All of the obtained samples were analyzed by X-ray powder diffraction to obtain an X-ray powder diffraction pattern. As in Example 2, an X-ray powder diffraction pattern was obtained using a standard X-ray powder diffraction pattern technique. The diffractometer used was Br
In the ucker AXS D8 Advance, a high intensity X-ray tube source was operated at 40 kV and 40 ma. Diffraction patterns from copper K-alpha radiation were obtained by appropriate computer techniques. Were scanned in the range of 3.6 samples force is applied horizontally ⁰ (2 [Theta]) / min 5 ⁰ -40 ^{0 (2θ).} Principal component analysis was applied to all X-ray powder diffraction patterns to yield four factors. Factor scores were calculated for each of the analyzed samples and blanks, and the first three scores were plotted on a three-dimensional graph. The resulting score plot is shown in FIG. Visual inspection of FIG. 7 showed two different score clusters. The cluster labeled E was identified as a substance synthesized at 150 ° C, and the other clusters labeled F were identified as substances synthesized at 200 ° C.

The sample cluster synthesized at 200 ° C. contained some unexpected patterns. Therefore, 20
Principal component analysis was applied to the X-ray powder diffraction patterns corresponding to a subset of the samples synthesized at 0 ° C. and four factors were determined. The score of the factor and the first three scores for each of the samples are shown in FIG.
A plot was made on the three-dimensional graph shown in A. The visual review of FIG. 8 shows the trend of scores within that subset. This trend was identified as indicating a sample that was quenched while undergoing interphase trait changes due to high temperatures. The scoring trend showed an association with the change in some peaks of the sample when the phase change occurred.

The visual inspection of FIG. 8 showed one cluster with the two side chains of the outlier extending radially along the Factor 1 and Factor 2 axes. These samples differed significantly from the samples in the main or vertex clusters, indicating the first three factors regarding their importance. Five outliers along the factor 1 axis were identified as new by X-ray powder diffraction analysis and the new material was named CuZnVO # 15. The X-ray powder diffraction patterns for these five samples are shown in FIG. Within these five outlier groups, the most important difference is the peak intensity. Similarly, factor 2
The side chains of the eight outliers along with were identified as new material and were named ZnVO # 13. The X-ray powder diffraction pattern corresponding to the latter set of samples is shown in FIG. As in the previous case, the main difference between these patterns was the peak intensity.

With these two groups removed from the initial set of patterns to be analyzed, principal component analysis was performed on the remaining samples to examine differences between the components of the major cluster. A plot of the resulting score
See Figure 11. Visual inspection of this plot shows that A, B,
Four major clusters labeled C and D are visible.
The component of cluster A was identified as a new material CuZNVO # 12. Figure 1 shows the overlapping pattern of the components of cluster A
See Figure 2. Cluster B was predominantly occupied by a material designated CuZnVO # 6, which contained only zinc cations, ie, was found in some of the above-mentioned materials without copper. Some of the samples in Cluster B contained an impurity designated ZnVO # 1. FIG. 13 shows an overlapping pattern of the components of the cluster B. C
In the cluster labeled, the presence of a substance named CuVO # 1 in each pattern is predominant. The overlapping pattern associated with cluster C is shown in FIG. Within this cluster, three subclusters are visible, which are C ₁ , C ₂ ,
And it is shown as C _3. The X-ray powder diffraction pattern for the sample of Cluster C was analyzed by principal component analysis and the resulting score plot is shown in FIG. 15, revealing three subclusters. Figure 16 shows CuVO
# Shows the X-ray powder diffraction pattern that is associated with the cluster C ₁ corresponding to the peak in the first structural + 3.02a. FIG. 17 shows cluster C
₂ shows the X-ray powder diffraction pattern associated with No. ₂ , indicating that the only component of the C2 cluster is CuVO # 1. Figure 18 shows that shows the X-ray powder diffraction pattern in the cluster C _3, the cluster C ₃ contains several peaks from minor component and CuVO # 1. The fourth cluster, labeled D, contained a sample that contained less crystalline material, as indicated by the noise levels in the pattern as shown in FIG. The peak at the higher angle corresponds to CuO. The remaining scores on the plot of FIG. 11 are not sufficiently clustered, which also indicates that they are mutually unique and should be analyzed separately. Among the remaining samples, the furthest component from the beginning contained the new CuZnO # 10.

[Brief description of the drawings]

FIGS. 1A and 1B show zeolite FAU, LTL, LT
FIG. 3 shows a series of X-ray powder diffraction patterns corresponding to a set of samples containing A, and mixtures thereof.

FIG. 2 is a three phase diagram of FAU, LTL, LTA and mixtures thereof used to generate an example test set.

FIG. 3 applies principal component analysis to the X-ray powder diffraction patterns of FIGS. 1A and 1B corresponding to the sample of FIG. 2 to determine a fixed number of factors, and each X-ray powder of the sample set. It is a three-dimensional graph of the score obtained by calculating the score of each factor with respect to a diffraction pattern.

FIG. 4 illustrates a method for determining a certain number of factors.
Obtained by applying principal component analysis to the X-ray powder diffraction patterns of a random set of samples composed of FAU, LTA, or blanks, and calculating the score for each factor for each X-ray powder diffraction pattern It is a two-dimensional graph of a score.

FIG. 5 applies principal component analysis to the X-ray powder diffraction pattern of the blank sub-cluster of FIG. 4 and for each X-ray powder diffraction pattern in the blank cluster of FIG. It is a two-dimensional graph of the score obtained by calculating the score of each factor.

FIG. 6 includes a superimposed X-ray powder diffraction pattern for each of the clusters of FIG. 5, where three clusters of blank samples correspond to the three types of artifacts associated with that blank sample cell. FIG.

FIG. 7 illustrates a method for determining a certain number of factors;
Two sample subsets made in the Cu-Zn-VO system, one synthesized at 150 ° C and the other at 200 ° C
7 is a three-dimensional graph of scores obtained by applying a principal component analysis to the X-ray powder diffraction patterns of the subset synthesized in step (a) and calculating the scores for each factor for each X-ray powder diffraction pattern.

FIG. 8 shows a sample made with a Cu—Zn—V—O system synthesized at 200 ° C. to determine a fixed number of factors.
3 is a three-dimensional graph of scores obtained by applying principal component analysis to the X-ray powder diffraction patterns of one sample subset and calculating a score for each factor for each X-ray powder diffraction pattern.

FIG. 9 shows overlapping X-ray powder diffraction patterns corresponding to the five outliers along the factor 1 axis of FIG.

FIG. 10 is an illustration of 8 along the factor 2 axis of FIG.
FIG. 4 shows overlapping X-ray powder diffraction patterns corresponding to two outliers.

FIG. 11 shows Cu-Zn- after excluding two groups of outliers to determine a fixed number of factors.
A score of 3 obtained by applying principal component analysis to the X-ray powder diffraction patterns of the remaining samples made in the VO system and calculating the score for each factor for each X-ray powder diffraction pattern. It is a dimensional graph.

FIG. 12 is a diagram showing overlapping X-ray powder diffraction patterns corresponding to cluster A in FIG. 11;

FIG. 13 is a diagram showing an overlapping X-ray powder diffraction pattern corresponding to cluster B in FIG. 11;

FIG. 14 is a diagram showing overlapping X-ray powder diffraction patterns corresponding to cluster C in FIG. 11;

FIG. 15 shows Cu-Zn- after excluding two groups of outliers to determine a fixed number of factors.
A score of 3 obtained by applying principal component analysis to the X-ray powder diffraction patterns of the remaining samples made in the VO system and calculating the score for each factor for each X-ray powder diffraction pattern. It is a dimensional graph.

FIG. 16 is a diagram showing overlapping X-ray powder diffraction patterns corresponding to cluster C ₁ in FIG.

FIG. 17 is a diagram showing overlapping X-ray powder diffraction patterns corresponding to the cluster C ₂ in FIG. 15;

Figure 18 is a diagram showing an overlapping X-ray powder diffraction pattern corresponding to the cluster C ₃ in Figure 15.

FIG. 19 is a diagram showing overlapping X-ray powder diffraction patterns corresponding to cluster C ₄ in FIG. 15.

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Sheryl M. Brach United States Illinois, Death Plains, East Argon Quinn Load 25 UOP LLC. (72) Inventor Gregory Jay. Luis United States Illinois, Death Plains, East Argon Quinn Load 25 UOP ELC F-term (reference) 2G001 AA01 BA18 CA01 GA01 GA13 HA01 HA05 HA20 JA12 MA04

Claims (12)

[Claims] 1. A method for screening a plurality of X-ray powder diffraction patterns corresponding to a set of samples, comprising: a) a constant number that can be used in combination with a score to represent each pattern in the sample set. Determining said factors by principal component analysis; b) determining a score for each factor for each X-ray powder diffraction pattern of said set; c) plotting said scores in a space of two or more dimensions; d. X) examining the score plot. 2. The test of the plot of scores comprises outliers,
The method of claim 1, wherein the method identifies a feature selected from a group consisting of clusters and trends. Determining a structure of at least one sample in at least one cluster from a corresponding X-ray powder diffraction pattern by comparing with a known X-ray powder diffraction pattern; Assigning the determined structure for one sample to all the samples in the cluster. 4. Determining one outlier or clustered sample to show erroneous data from a corresponding X-ray powder diffraction pattern, and determining a false error determined for said one sample in one cluster. 3. A method according to claim 2 including the step of assigning data to all samples in the cluster. 5. The X-ray of claim 2 including the step of determining the structure of at least one outlier by comparing the X-ray powder diffraction pattern of the outlier with a known X-ray powder diffraction pattern. Powder diffraction pattern screening method. 6. The method of claim 1, wherein the structure in one cluster is identified as new by comparing an X-ray powder diffraction pattern of the one sample, and wherein the structure determined for the one sample in one cluster is new. Assigning a new structure to all samples in the cluster. 7. The method according to claim 2, wherein at least one of said samples has a known structure, and the unknown is identified by informing said clustering function of said known sample pattern. X-ray powder diffraction pattern screening method. 8. A) comparing the X-ray powder diffraction patterns corresponding to the samples in one cluster and examining the differences; and b) X-ray powder corresponding to the samples in the clusters whose patterns show differences. Principal component analysis applied only to the diffraction pattern provides a number of subset factors that can be used to represent each X-ray powder diffraction pattern within the sample subset in combination with the subset score of the subset factor. Determining) c) determining a subset score for each subset factor for each X-ray powder diffraction pattern of the sample subset; and d) plotting the subset score obtained in two or more dimensions. E) examining the plot of the subset scores. 3. The X-ray powder diffraction pattern screening method according to claim 2, comprising a step. 9. The method further comprising: a) selecting a sample and its corresponding X-ray powder diffraction pattern and a subset; and b) principal component analysis in combination with a subset score of a subset factor. Determining a number of subset factors that can be used to represent each X-ray powder diffraction pattern within the subset; c) a subset of each subset factor for each X-ray powder diffraction pattern of the sample subset. 3. The method of claim 2, further comprising the steps of: determining a score; d) plotting subset scores obtained in two or more dimensions; and e) examining said subset score plot. X-ray powder diffraction pattern screening method. 10. The X-ray powder diffraction pattern screening method according to claim 9, wherein the sample subset belongs to the cluster according to claim 2. 11. The method of claim 9, further comprising the step of repeating the above operations a) to e) at least once, each iteration being performed using a selected number of sample subsets. X-ray powder diffraction pattern screening method. 12. The method of claim 9, wherein examining the plot of subset scores leads to identification of features selected from a group consisting of outliers, clusters, and trends.