Copolymer analysis method, copolymer analysis program, and copolymer analysis device

Abstract

To provide an analysis method that can obtain predetermined sequence information on a copolymer from a chromatogram.SOLUTION: There is provided a copolymer analysis method for estimating a copolymerization index which is an index of the ratio of bonds between specific repeating units in a copolymer including two or more repeating units. The copolymer analysis method includes the step for estimating the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed using a learned model that estimates the copolymerization index from the chromatogram.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to a method for analyzing a copolymer, a program for analyzing a copolymer, and an apparatus for analyzing a copolymer. [Background technology]

[0002] JP 2022-13303 A discloses a data analysis device and an analysis method for predicting the structure and composition of a polymer raw material from the measured one-dimensional spectral information of the polymer raw material. The publication lists the ratio of cis, trans, vinyl, and styrene in diene rubber, as well as dyad and triad chains, as examples of structural information to be predicted. The publication also lists the amount of butadiene and the amount of styrene, the mass average molecular weight, the viscosity, the degree of branching, and the number of branches as examples of composition information to be predicted.

[0003] Japanese Patent Application Laid-Open No. 2022-13310 discloses a data analysis device and an analysis method for predicting physical property information of a polymer raw material from structural information of the polymer raw material. The publication lists elastic modulus, Poisson's ratio, viscosity, glass transition temperature, and crystallization temperature as examples of physical property information to be predicted.

[0004] JP 2004-233095 A discloses a method for analyzing organic oligomers. In this method, an organic oligomer with known analytical values ​​is first analyzed by liquid chromatogram and / or supercritical fluid chromatogram, and multiple regression analysis is performed between each peak of the chromatogram and the known analytical values, and a regression coefficient is obtained based on the peak with a high correlation with the known analytical values, and a regression equation is determined. Based on this result, an organic oligomer with unknown analytical values ​​is analyzed by liquid chromatogram and / or supercritical fluid chromatogram, and the analytical values ​​are calculated from the regression equation. In this publication, the epoxy equivalent, viscosity, softening point, hydrolyzable chlorine content, 1,2-glycol content, and total chlorine content of the epoxy resin are given as examples of analytical items.

[0005] JP 2021-107813 A discloses a polymer property estimation device. This polymer property estimation device estimates physical property values ​​from NMR spectrum data of a polymer sample using a learning model generated by machine learning. The publication lists melting point, boiling point, sublimation temperature, glass transition point, crystallization temperature, etc. as examples of physical property values ​​to be estimated.

[0006] JP 2021-149449 A discloses a material searching device that searches for a material with predetermined properties. This device identifies a predicted range of spectral properties exhibited by a material with more suitable properties based on actual measurement information including the synthesis conditions of the material, the evaluation value of the properties of the material, and spectral data, and searches for a material having this predicted range of spectral properties using an optimization method.

[0007] Japanese Patent Application Laid-Open No. 2021-76602 discloses an information processing device. The information processing device has an information acquisition means for acquiring quantitative information estimated by inputting spectral information of a sample containing a test substance into a learning model, and a contribution degree acquisition means for acquiring a contribution degree related to the acquired quantitative information. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Patent Publication No. 2022-13303 [Patent Document 2] Patent Publication No. 2022-13310 [Patent Document 3] JP 2004-233095 A [Patent Document 4] Patent Publication No. 2021-107813 [Patent Document 5] Patent Publication No. 2021-149449 [Patent Document 6] Patent Publication No. 2021-76602 Summary of the Invention [Problem to be solved by the invention]

[0009] The present inventors have been focusing on the arrangement (lineup) of repeating units contained in a copolymer to develop copolymers with novel properties. For example, in a copolymer containing repeating units A and B, even if the ratio of repeating units A to B is constant, the properties may change by changing the arrangement. Specifically, the properties may differ between a copolymer having a block of consecutive repeating units A and a block of consecutive repeating units B (block copolymer) and a copolymer in which repeating units A and B are irregularly arranged (random copolymer).

[0010] Nuclear magnetic resonance (NMR) is commonly used to analyze the sequences of such copolymers. Although analysis by high performance liquid chromatography (HPLC) such as liquid chromatography / mass spectrometry (LC / MS) is theoretically possible, it is not practical for several reasons, as explained below.

[0011] First, in order to quantitatively analyze a copolymer by HPLC, it is necessary to prepare a standard substance (reference) of each repeating unit, either a pure product or a copolymer with a known structure. This is because the sensitivity of the HPLC detector varies greatly depending on the substance. In many cases, it is not easy to obtain such a reference. In contrast, in NMR, the sensitivity is constant regardless of the substance if the relaxation time, etc. is set appropriately, so quantitative analysis is possible even without a reference. In addition, in HPLC, when the analysis target is a polymer, molecules with different chain lengths each have a separate peak, so many peaks overlap overall, making it difficult to determine the peak area. In contrast, in NMR, the signal of the same partial structure becomes one regardless of the chain length of the polymer, making it relatively easy to quantify.

[0012] On the other hand, NMR measurement equipment is more expensive than HPLC measurement equipment, and if you do not own an NMR measurement equipment, you will have to outsource the analysis to an external institution, so you cannot use it frequently.

[0013] An object of the present invention is to provide an analytical method, an analytical program, and an analytical device that can obtain predetermined sequence information of a copolymer from a chromatogram. [Means for solving the problem]

[0014] A copolymer analysis method according to one embodiment of the present invention is a copolymer analysis method for estimating a copolymerization index, which is an index of the ratio of bonds between specific repeating units in a copolymer containing two or more types of repeating units, and includes a step of estimating the copolymerization index of a target copolymer from a chromatogram of a sample containing the target copolymer, using a trained model that estimates the copolymerization index from a chromatogram.

[0015] A copolymer analysis program according to one embodiment of the present invention is a copolymer analysis program that estimates a copolymerization index, which is an index of the bond ratio between specific repeating units in a copolymer containing two or more types of repeating units, and causes a computer to execute a step of estimating the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram.

[0016] A copolymer analysis apparatus according to one embodiment of the present invention is an apparatus for estimating a copolymerization index, which is an index of the ratio of bonds between specific repeating units in a copolymer containing two or more types of repeating units, and is equipped with an estimation device that estimates the copolymerization index of a target copolymer from a chromatogram of a sample containing the target copolymer, using a trained model that estimates the copolymerization index from a chromatogram. Effect of the Invention

[0017] According to the present invention, it is possible to obtain specific sequence information of the copolymer from the chromatogram. [Brief description of the drawings]

[0018] [Figure 1]FIG. 1 shows a portion of the structure of a copolymer containing lactic acid units LA and caprolactone units CL. [Figure 2A] FIG. 2A is a 1H-NMR spectrum of a copolymer having a relatively small ratio of bonds between lactic acid units LA and caprolactone units CL. [Figure 2B] FIG. 2B is a 1H-NMR spectrum of a copolymer having a relatively high ratio of bonds between lactic acid units LA and caprolactone units CL. [Diagram 3] FIG. 3 is a TIC chromatogram of the sample in FIG. 2A. [Figure 4] Figure 4 is a scatter plot (S-Plot) in which the covariance obtained by PLS-DA is plotted on the horizontal axis and the correlation coefficient is plotted on the vertical axis. [Diagram 5] FIG. 5 is a scatter plot in which the horizontal axis represents Fα=Iα* / (Iα+Iα*) and the vertical axis represents the peak area of ​​the peak at RT: 6.58 minutes and m / z: 599.268. [Figure 6] FIG. 6 is a diagram showing a schematic diagram of a nonlinear mapping. [Figure 7] FIG. 7 is a TIC chromatogram measured in the example. [Figure 8] FIG. 8 is a scatter plot in which the predicted values ​​of Fα are plotted on the vertical axis and the true values ​​of Fα (values ​​calculated from NMR spectra) are plotted on the horizontal axis. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] The present inventors have investigated a method for obtaining specific sequence information of a copolymer from data (chromatogram) obtained by measuring the copolymer by chromatography. More specifically, they have investigated a method for estimating the bond ratio between specific repeating units in a copolymer containing two or more kinds of repeating units from the chromatogram.

[0020] Nuclear magnetic resonance (NMR) is generally used to obtain information about bonds in copolymers. FIG. 1 is a diagram showing a part of the structure of a copolymer containing lactic acid units LA and caprolactone units CL. FIG. 2A shows a copolymer (hereinafter referred to as "sample 1") in which the ratio of bonds between lactic acid units LA and caprolactone units CL is relatively small. 1 FIG. 2B is a H-NMR spectrum of a copolymer having a relatively large ratio of bonds between lactic acid units LA and caprolactone units CL (hereinafter referred to as "sample 2"). 1 This is the H-NMR spectrum. 1 The H-NMR spectrum was measured using a Bruker AVANCE III HD 400MHz nuclear magnetic resonance spectrometer (NMR) by dissolving the copolymer in deuterated chloroform at 10 mg / ml. The chemical shift standard was TMS (tetramethylsilane).

[0021] As shown in Figure 1, the caprolactone unit CL has five methylene groups. These methylene groups are referred to as the α-, β-, γ-, δ-, and ε-position methylene groups, starting from those directly bonded to the carbonyl carbon of the ester bond.

[0022] The methylene group at the α position is the methylene group bonded to the carbonyl carbon of the ester linkage. The methylene group at the ε position is the methylene group bonded to the oxygen of the ester linkage.

[0023] The signal derived from the proton of the methylene group at the α-position appears at a different chemical shift depending on whether the methylene group is bonded to the caprolactone unit CL via an ester bond or to the lactic acid unit LA. Here, the methylene group bonded to the carbonyl carbon of the ester bond and bonded to the lactic acid unit LA via the ester bond is referred to as α. * The methylene group bonded to the caprolactone unit CL via the ester bond is represented as the methylene group at the α-position as described above. * The signals of are α-position and α *Specifically, the signal of α appears at about 2.30 to 2.35 ppm, while the signal of α * The signal appears at approximately 2.35 to 2.50 ppm.

[0024] Similarly, the signal derived from the proton of the methylene group at the ε position appears at a different chemical shift depending on whether the methylene group is bonded to the caprolactone unit CL via an ester bond or to the lactic acid unit LA. Although not shown in Figure 1, the methylene group bonded to the oxygen of the ester bond and bonded to the lactic acid unit LA via the ester bond is called the ε * The methylene group bonded to the caprolactone unit CL via the ester bond is represented as the methylene group at the ε position, as described above. * The signals are at ε and ε positions, respectively. * Specifically, the ε signal appears at about 4.00 to 4.10 ppm, whereas the ε signal appears at about 4.00 to 4.10 ppm. * The signal appears at approximately 4.10 to 4.20 ppm.

[0025] Here, Iα, Iα * , Iε and Iε * are respectively α and α * , ε and ε * The integrated intensity of the signal is taken as the intensity. 1 The integrated intensity of the H-NMR signal is proportional to the number of protons. * , Iε and Iε * are α-position and α * ε position, ε position and ε * The amount is proportional to the number of protons in the methylene group at the

[0026] Therefore, Iα is a quantity proportional to the number of bonds between caprolactone units CL and CL (hereinafter referred to as "CL-CL bonds"), and Iα *is proportional to the number of bonds between caprolactone units CL and lactic acid units LA (hereinafter referred to as "CL-LA bonds"). Therefore, for example, Fα = Iα * / (Iα+Iα * ) can be used as an indicator of the ratio of bonds between specific repeat units (CL-LA bonds in this example). * / (Iα+Iα * ) is the ratio of the number of CL-LA bonds to the sum of the number of CL-CL bonds and the number of CL-LA bonds. Instead of Fα, Fε = Iε * / (Iε+Iε * ) may be taken as an indication of the ratio of bonds between particular repeat units.

[0027] Hereinafter, this "index of the ratio of bonds between specific repeating units" will be referred to as the "copolymerizability index." As shown in Figures 2A and 2B, Sample 2 (Figure 2B) has a higher α * It can be seen that the signal of is larger and the ratio of CL-LA bonds is higher in sample 2 than in sample 1. In this way, the copolymerization index can be quantitatively analyzed from the NMR data.

[0028] Next, we investigated whether a similar analysis could be performed from chromatograms, that is, whether the copolymerization index could be calculated from the chromatograms.

[0029] The same samples (sample 1 and sample 2) as those in Figures 2A and 2B were measured by liquid chromatography mass spectrometry (LC / MS). The measurement conditions are shown in Table 1. The samples were dissolved by adding 100 times the weight of methanol, filtered through a 0.2 μm filter, and further diluted 100 times with methanol before use.

[0030] [Table 1]

[0031] Figure 3 shows the total ion current (TIC) chromatogram of sample 1. In the chromatogram, molecules with different chain lengths each appear as a separate peak. For example, there are four different combinations of two lactic acid units LA and two caprolactone units CL (LA-LA-CL-CL, LA-CL-AL-CL, LA-CL-CL-LA, and CL-LA-LA-CL), which appear as separate peaks, respectively. * The ratio of Fα to Iα is different. As in the case of NMR, Fα = Iα * / (Iα+Iα * When trying to calculate the intensity of all peaks, this calculation must be performed for all peaks. In principle, it is possible to identify all peaks and add up their intensities, but in practice this is difficult.

[0032] Thus, it is difficult to calculate the copolymerization index directly from the chromatogram.

[0033] Next, the peak area of ​​each peak was calculated from the TIC chromatograms of each of Sample 1 and Sample 2, and partial least squares discriminant analysis (PLS-DA) was performed to search for peaks that characterize Sample 1 and Sample 2. Figure 4 is a scatter plot (S-Plot) in which the covariance obtained by PLS-DA is plotted on the horizontal axis and the correlation coefficient is plotted on the vertical axis. The peak group in region A1 of Figure 4 is the peaks that are highly correlated with Sample 1 (a copolymer with fewer CL-LA bonds), and the peak group in region A2 is the peaks that are highly correlated with Sample 2 (a copolymer with more CL-LA bonds). Table 2 shows the peak groups contained in regions A1 and A2.

[0034] [Table 2]

[0035] From Table 2, it can be seen that in region A1, there are many peaks derived from oligomers consisting of a single repeating unit, and from oligomers consisting of a single repeating unit to which one or two other repeating units are bonded. On the other hand, in region A2, there are many peaks derived from oligomers in which the same number of lactic acid units LA and caprolactone units CL are bonded. This suggests that the components with a difference in the number of repeating units of 2 or less may characterize the copolymerization index.

[0036] Next, we investigated several copolymers produced under different conditions. 1 Fα was determined by measuring H-NMR spectrum. TIC chromatograms were measured for these copolymers, and the peak areas of the peaks at RT: 6.58 min and m / z: 599.268 were calculated. Figure 5 is a scatter plot in which Fα is plotted on the horizontal axis and the peak area of ​​the peak on the vertical axis. As shown in Figure 5, although there was some variation, a certain degree of correlation was observed between the peak area of ​​the peak and Fα.

[0037] In this way, the chromatogram contains information that is statistically correlated with the copolymerization index. From this fact, the inventors conceived the idea that the copolymerization index can be estimated from the chromatogram by relating the chromatogram and the copolymerization index using a machine learning technique.

[0038] The present invention has been completed based on the above findings. Hereinafter, the embodiment of the present invention will be described in detail. Note that the above-mentioned relationship between the peak area of ​​the peak at RT: 6.58 minutes and m / z: 599.268 and Fα is shown as an example in which information correlated with the copolymerization index is contained in a chromatogram. The present invention does not attempt to estimate the copolymerization index based only on the peak area of ​​a specific peak in a chromatogram.

[0039] [Analysis method of copolymer] A copolymer analysis method according to one embodiment of the present invention is a method for estimating a copolymerization index, which is an index of the ratio of bonds between specific repeating units in a copolymer containing two or more types of repeating units. The analysis method according to this embodiment includes a step of estimating the copolymerization index of a target copolymer from a chromatogram of a sample containing the target copolymer, using a trained model for estimating the copolymerization index from a chromatogram.

[0040] The copolymer to be analyzed in this embodiment is an organic polymer or oligomer containing two or more kinds of repeating units. In FIG. 1, a copolymer containing lactic acid units LA and caprolactone units CL is illustrated, but the copolymer to be analyzed in this embodiment is not limited to this. The copolymer to be analyzed in this embodiment may be a random copolymer, an alternating copolymer, a block copolymer, or a graft copolymer.

[0041] The copolymer to be analyzed in this embodiment preferably has a number average molecular weight Mn of 50,000 or less. If the number average molecular weight Mn is too large, the chromatogram peaks may overlap too much, which may reduce the accuracy of the estimation. The number average molecular weight Mn is more preferably 30,000 or less, even more preferably 10,000 or less, even more preferably 5,000 or less, and even more preferably 1,000 or less.

[0042] The analysis method according to the present embodiment is a method for estimating a copolymerization index, which is an index of the ratio of bonds between specific repeating units. For example, in a copolymer containing repeating units A and B, there are three types of bonds: a bond between repeating units A and A (AA), a bond between repeating units A and B (AB), and a bond between repeating units B and B (BB). Any one of these bonds can be determined as the "bond between specific repeating units."

[0043] Here, if the bond (AB) between repeating unit A and repeating unit B is defined as a "bond between specific repeating units," a high copolymerization index means that the proportion of AB bonds in the copolymer is high. In other words, the higher the copolymerization index, the more frequently the repeating unit A and the repeating unit B are switched, and the smaller the size of the block in which the repeating unit A or the repeating unit B is continuous.

[0044] The copolymer to be analyzed may contain three or more types of repeating units. For example, a copolymer containing three types of repeating units may have six types of bonds. In this case, any one of the six types can be determined as the "bond between specific repeating units."

[0045] The copolymerization index is an index of the ratio of the "bonds between specific repeating units" selected as above. The copolymerization index does not need to be a "ratio" in the strict sense, but may be an "index" normalized to allow relative comparison. For example, when the copolymer contains a first repeating unit and a second repeating unit, the copolymerization index may be expressed by the following formula: Copolymerizability index=(number of bonds between the first repeating unit and the second repeating unit) / {(number of bonds between the first repeating unit and the second repeating unit)+(number of bonds between the first repeating unit and the first repeating unit)}

[0046] The above is just one example, and for example, the denominator in the above formula may be (the number of all bonds) or (the number of bonds between the first repeating units).

[0047] In the analysis method according to the present embodiment, as described below, a trained model generated by machine learning is used to estimate the copolymerization index of the copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed. At the stage of generating the trained model, it is necessary to input the true value of the copolymerization index as training data. The copolymerization index can be calculated, for example, from an NMR spectrum.

[0048] In FIG. 2A and FIG. 2B, for a copolymer containing lactic acid units LA and caprolactone units CL, Fα=Iα * / (Iα+Iα * 2A and 2B, the copolymerization index can be calculated by the same method for other copolymers. For example, the copolymerization index can be calculated as (integral intensity of signal due to bond between specific repeating units) / {(integral intensity of signal due to bond between other repeating units)+(integral intensity of signal due to bond between specific repeating units)}. In FIG. 2A and FIG. 2B, 1 We have explained the case where the copolymerization index is calculated from the H-NMR spectrum. 1 In addition to H-NMR, 13 C-NMR, 19 F-NMR or 31 P-NMR may also be used.

[0049] The chromatogram can be measured, for example, by liquid chromatography or gas chromatography.

[0050] The liquid chromatograph preferably includes, as a detector, one or more detectors selected from the group consisting of a UV / VIS detector, a mass analyzer, a fluorescence detector, a differential refractive index detector, an electrical conductivity detector, an electrochemical detector, an evaporative light scattering detector, and a corona charged particle detector. The detector is more preferably a mass analyzer, a differential refractive index detector, or a corona charged particle detector, and particularly preferably a mass analyzer.

[0051] The gas chromatograph preferably includes a pyrolysis device as a front stage. The gas chromatograph preferably includes, as a detector, one or more detectors selected from the group consisting of a mass spectrometer, a hydrogen flame ionization detector, a thermal conductivity detector, a barrier discharge ionization detector, an electron capture detector, a thermal ionization detector, a flame photometric detector, and a sulfur chemiluminescence detector. The detector is particularly preferably a mass spectrometer.

[0052] A chromatogram is one-dimensional spectral data. The more data points included in a chromatogram, i.e., the higher the time resolution, the higher the estimation accuracy may be. On the other hand, since too many data points consume a large amount of calculation resources, the number of data points may be reduced by thinning the data or taking a moving average before using the data for learning and estimation. In addition, normalizing a chromatogram by a certain time may improve the estimation accuracy.

[0053] [Trained model] A trained model is an estimation model that estimates output data (objective variables) from input data (explanatory variables). A trained model is obtained by machine learning using training data created in advance for multiple copolymers (hereinafter referred to as "training samples").

[0054] First, for each of a plurality of learning samples, a chromatogram is obtained and a copolymerization index is calculated, and training data is created in which the chromatogram is the input data (explanatory variable) and the copolymerization index is the output data (objective variable). As described above, the copolymerization index can be calculated from the NMR spectrum.

[0055] The training data is preferably prepared for a plurality of copolymers that contain the same repeating unit as the copolymer to be analyzed, but have different molecular weights and copolymerizability indices. The more training samples there are (i.e., the more training data there are for machine learning), the more accurate the trained model that can be obtained.

[0056] The trained model is generated by machine learning using the created training data. The specific machine learning method is not particularly limited, and any method can be used.

[0057] Machine learning has two components: the evaluation method and the kernel function. Evaluation methods include models that use the least squares method (the squared distance from the correct answer) as the evaluation criterion and include corrections to prevent overfitting (least squares method, LASSO, Ridge, Elastic net, Huber, PLS, etc.), models that use absolute values ​​rather than squared values ​​as the evaluation criterion (SVR, etc.), and models that use multiple decision trees (Random Forest, etc.). From these, the one that is most suitable for the inherent variability and experimental error of the data used is selected, and the one with the highest fitting accuracy is selected. Essentially, the general procedure is to evaluate the accuracy of all models and select the best one.

[0058] As described above, the machine learning method is not particularly limited, but for the application of this embodiment, partial least squares regression (PLS regression) is relatively suitable. PLS regression is a method that enables linear regression by setting new spatial coordinate axes by reconfiguring variables and projecting them onto the axes. PLS regression is particularly suitable for cases where there is multicollinearity between input data (chromatograms in this embodiment) (where parameters are not independent and there is a strong correlation between parameters).

[0059] For all models other than RandomForest, the kernel function (the shape of the baseline when fitting) can be specified as follows, for example. From these, select the one that is appropriate for the copolymer to be analyzed. ·liner (linear model, straight line) ·polynomial (polynomial model) ·RBF (nonlinear model)

[0060] However, scikit-learn, which is often used to perform machine learning, is designed so that only linear models can be selected other than Ridge and SVR. Even in this case, nonlinear data can be used. Also, linear regression becomes possible by performing nonlinear mapping. For example, as shown in Figure 6, even if the data can only be fitted to the boundary with a nonlinear line in a 2D image, by increasing the dimension to 3D, it becomes possible to perform "straight line / linear" fitting.

[0061] The correlation between the chromatogram and the copolymerization index does not necessarily fall into only one of the above categories, and the correlation (here, this means the shape of the correlation function) differs depending on the copolymer being analyzed. Although it is possible to apply a nonlinear model to data that is linearly correlated, there is a risk of overlearning due to errors in the measured data (for example, fitting with a wavy curve that matches the errors in the measured data instead of fitting with a straight line, returning an incorrect value when predicting). Therefore, when it is known that the correlation is linear, it is preferable to use a linear model. However, when the correlation is unknown, fitting is possible by selecting a nonlinear model.

[0062] Once the trained model is obtained, the copolymerization index of the analyzed copolymer can be estimated from a chromatogram of a sample containing the analyzed copolymer using the trained model. In other words, the copolymerization index of the analyzed copolymer can be estimated from a chromatogram without acquiring an NMR spectrum.

[0063] [Effects of this embodiment] The above describes a copolymer analysis method according to one embodiment of the present invention. According to this embodiment, it is possible to obtain specific sequence information of a copolymer from a chromatogram. More specifically, it is possible to estimate a copolymerization index, which is an index of the ratio of bonds between specific repeating units, in a copolymer containing two or more types of repeating units.

[0064] According to this embodiment, the copolymerization index of the copolymer to be analyzed can be estimated from the chromatogram without acquiring an NMR spectrum. Furthermore, according to this embodiment, there is no need to prepare a standard substance (reference).

[0065] Even when NMR is used, the signals of impurities may overlap, making it difficult to obtain the correct integrated intensity. When the sample is an industrial product, it often contains undisclosed components in a non-negligible amount in addition to unintended impurities. In addition, when the number of components is large, peaks may overlap, making analysis difficult. According to this embodiment, it may be possible to estimate the copolymerization index of such a sample that is difficult to analyze by NMR.

[0066] [Copolymer analysis programs, etc.] The above-mentioned copolymer analysis method can also be realized as a computer program. The copolymer analysis program according to one embodiment of the present invention is a copolymer analysis program that estimates a copolymerization index, which is an index of the bond ratio between specific repeating units in a copolymer containing two or more types of repeating units, and causes a computer to execute a step of estimating the copolymerization index of the copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram. This computer program may be recorded on a computer-readable recording medium.

[0067] The above-mentioned copolymer analysis method can also be realized as a computer system. A copolymer analysis system according to one embodiment of the present invention is a copolymer analysis system that estimates a copolymerization index, which is an index of the ratio of bonds between specific repeating units in a copolymer containing two or more types of repeating units, and includes a memory and a processor, and the processor executes a step of estimating the copolymerization index of the copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram, according to a program in the memory.

[0068] The copolymer analysis apparatus according to one embodiment of the present invention is an apparatus for estimating a copolymerization index, which is an index of the bond ratio between specific repeating units, in a copolymer containing two or more types of repeating units, and includes an estimation device that estimates the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model for estimating the copolymerization index from a chromatogram. The copolymer analysis apparatus according to this embodiment may further include a model generation device that generates a trained model by machine learning using teacher data having a chromatogram of the copolymer as an explanatory variable and a copolymerization index calculated from a nuclear magnetic resonance spectrum as a target variable.

[0069] This invention can contribute to achieving Goal 9 (Industry, Innovation and Infrastructure) and Goal 12 (Responsible Consumption and Production) of the 17 Sustainable Development Goals (SDGs) established by the United Nations. EXAMPLES

[0070] The present invention will be described in more detail below with reference to examples, but the present invention is not limited to these examples.

[0071] A copolymer of polylactic acid and polycaprolactone was produced by the method described below.

[0072] When multiple types of polyester resins are heated and stirred in an environment where moisture is present, decomposition and generation of ester bonds occur simultaneously, resulting in a polyester copolymer. At this time, the copolymerization index can be adjusted by adjusting the reaction conditions (moisture content, temperature, reaction time, stirring speed, etc.). The present inventors have also filed patent applications for this method and the invention of the product produced by this method (PCT / JP2023 / 002986, Patent Application No. 2023-125568, Patent Application No. 2023-125569).

[0073] Copolymers were produced by varying the reaction conditions by adding polylactic acid (PLA) and polycaprolactone (PCL) in the ratio (mass ratio) shown in Table 3. The molecular weight of the produced copolymer was measured by GPC.1 The formation of a copolymer was confirmed from the H-NMR spectrum, and Fα, which was explained in Figures 2A and 2B, was calculated as the copolymerization index. The results are shown in Table 3.

[0074] [Table 3]

[0075] The TIC chromatograms of these copolymers were measured by LC / MS. The measured TIC chromatograms are shown in Figure 7. Note that the TIC chromatograms were measured twice for each sample.

[0076] An estimation model (trained model) with chromatograms as input data (explanatory variables) and Fα as output data (objective variable) was generated by machine learning. Machine learning was performed using the open source library scikit-learn. PLS regression was used as the learning algorithm. In order to reduce the load, the chromatogram data was thinned so that the number of data points per dataset was about 300 to 400. In addition, general preprocessing such as setting the RT range, removing outliers, and normalization was performed before use. #1, #2, #4, #5, and #6 were used as learning data (training data), and #3 and #7 were used as validation data (validation data).

[0077] Figure 8 is a scatter plot with the predicted values ​​of Fα plotted on the vertical axis and the true values ​​of Fα (values ​​calculated from the NMR spectrum) plotted on the horizontal axis. The solid marks represent training data, and the open marks represent validation data. As shown in Figure 8, we confirmed that predictions could be made with a certain degree of accuracy even for the validation data. It is expected that the accuracy of predictions will be further improved by increasing the amount of training data.

[0078] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-mentioned embodiment, and various modifications are possible within the scope of the invention.

Claims

1. A method for analyzing a copolymer, comprising the steps of: estimating a copolymerization index, which is an index of a ratio of bonds between specific repeating units in a copolymer containing two or more kinds of repeating units; A method for analyzing a copolymer, comprising a step of estimating the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram.

2. 2. The copolymer analysis method according to claim 1, wherein the trained model is generated by machine learning using training data for a plurality of copolymers, the training data having copolymer chromatographs as explanatory variables and the copolymerizability index calculated from nuclear magnetic resonance spectra as a response variable.

3. The nuclear magnetic resonance １ H-NMR, １３ C-NMR, １９ F-NMR or ３１ The method for analyzing the copolymer according to claim 2, wherein the analysis is by P-NMR.

4. The copolymer includes a first repeat unit and a second repeat unit, The method for analyzing a copolymer according to claim 1 , wherein the copolymerization index is represented by the following formula: Copolymerizability index=(number of bonds between the first repeat unit and the second repeat unit) / {(number of bonds between the first repeat unit and the second repeat unit)+(number of bonds between the first repeat unit and the first repeat unit)}

5. The method for analyzing a copolymer according to claim 1 , wherein the copolymer is a random copolymer, an alternating copolymer, a block copolymer or a graft copolymer.

6. The chromatogram is measured by liquid chromatography; The method for analyzing a copolymer according to claim 1, wherein the liquid chromatograph includes, as a detector, one or more detectors selected from the group consisting of a UV / VIS detector, a mass spectrometer, a fluorescence detector, a differential refractive index detector, an electrical conductivity detector, an electrochemical detector, an evaporative light scattering detector, and a corona charged particle detector.

7. The chromatogram was measured by a gas chromatograph; 2. The method for analyzing a copolymer according to claim 1, wherein the gas chromatograph includes, as a detector, one or more detectors selected from the group consisting of a mass spectrometer, a hydrogen flame ionization detector, a thermal conductivity detector, a barrier discharge ionization detector, an electron capture detector, a thermal ionization detector, a flame photometric detector, and a sulfur chemiluminescence detector.

8. A copolymer analysis program for estimating a copolymerizability index, which is an index of a ratio of bonds between specific repeating units in a copolymer containing two or more kinds of repeating units, comprising: A copolymer analysis program that causes a computer to execute a step of estimating the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram.

9. An apparatus for analyzing a copolymer, which estimates a copolymerization index, which is an index of a ratio of bonds between specific repeating units in a copolymer containing two or more kinds of repeating units, comprising: An apparatus for analyzing a copolymer, comprising an estimation device that estimates the copolymerization index of a copolymer to be analyzed from a chromatogram of a sample containing the copolymer to be analyzed, using a trained model that estimates the copolymerization index from a chromatogram.

10. The copolymer analysis apparatus according to claim 9 , further comprising a model generation device configured to generate the trained model by machine learning using training data having a chromatogram of the copolymer as an explanatory variable and the copolymerizability index calculated from a nuclear magnetic resonance spectrum as a response variable.

Images