JEOL Resources

AccuTOF™ GC-Alpha & LIFDI

Fri, 01 May 2026 06:42:31 GMT

The best analytical tool for organometallics and air-sensitive samples

AccuTOF™ GC-Alpha is a superior time-of-flight mass spectrometer (TOF MS) system that simultaneously accomplishes high-resolution analysis, high mass accuracy and high-speed data acquisition. Equipped with the unique triple ion source EI/FI/FD that supports LIFDI, it offers the ultimate flexibility for GC-MS and direct-insertion MS applications.

msFineAnalysis AI Novel Qualitative Analysis Software for JMS-T2000GC with AI Structural Analysis

Fri, 10 Oct 2025 10:02:55 GMT

JEOL NEWS Vol.58 No.1
Ayumi Kubo, MS Business Unit, JEOL Ltd.

JEOL developed msFineAnalysis as qualitative analysis software for our gas chromatograph time of flight mass spectrometer (GC-TOFMS). We implemented deconvolution detection, variance component analysis, and other features in the software through updates. We have recently developed a new version of the series called msFineAnalysis AI. msFineAnalysis AI is equipped with a structural analysis method using artificial intelligence (AI), called “AI structural analysis.” AI structural analysis enables the identiﬁcation of molecular formulas as well as structural formulas of compounds that are not registered in the NIST 20 library (unknown compounds). The workﬂow of AI structural analysis is as stated below.

First, msFineAnalysis’s integrated analysis function identiﬁes the molecular formula of an unknown compound. Next, based on the identiﬁed molecular formula, structural formula candidates are extracted from PubChem, the database containing over 100 million compounds. The AI predicts electron ionization (EI) mass spectra from the extracted structural formula candidates. Then, the structural formula candidates are ranked by comparing the predicted mass spectra with the measured mass spectrum. Finally, a candidate that ranks ﬁrst is adopted as the analysis result.

Using the NIST 20 library, we trained the AI to predict mass spectra from structural formulas and evaluated its accuracy. From the results of accuracy evaluation, we conﬁrmed that AI structural analysis is useful in the structural analysis of unknown compounds. In this report, we will introduce features of msFineAnalysis AI and provide our evaluation results.

Structural Analysis of Polyethylene Terephthalate Film by using a Py-GC-HRTOFMS and msFineAnalysis AI

Fri, 10 Oct 2025 09:58:09 GMT

MSTips No. 391

Introduction

Electron ionization (EI) is one of the most popular ionization methods used in gas chromatography-mass spectrometry (GC-MS). Consequently, compounds are typically identified by a mass spectral database search using EI mass spectra. Because molecular ions are often weak or absent in 70 eV EI mass spectra, identification of unknowns can be difficult by EI alone. In these cases, soft ionization (SI) can be very helpful for producing and identifying molecular ions. Recently, JEOL began developing an integrated qualitative analysis workflow that automatically combines and interprets the information from EI and SI data. And then in 2018, we introduced our integrated qualitative analysis software “msFineAnalysis” which uses both EI and SI data to improve compound identification for GC-MS applications.

Despite the fact that msFineAnalysis was automatically able to determine the molecular formula and partial structure information from EI fragment ion formulas, the actual structural formulas still required manual analysis using chemical compositions. To address this, we then developed an automated structure analysis software package entitled "msFineAnalysis AI" which uses artificial intelligence (AI) to predict EI mass spectra from chemical structures. We have used our newly-developed AI model to create a database of predicted EI mass spectra for around 100 million compounds. In this work, we introduce a polymer materials application that uses msFineAnalysis AI for structural analysis.

AI Structural Analysis

The AI structural analysis workflow is shown in Figure 1. In this method, we used deep learning to construct an AI model that can predict the EI mass spectrum from a structural formula. We then submitted approximately 100 million compound structure formulas to our AI model in order to generate predicted EI mass spectra. The structural formula and the predicted EI mass spectra associated with each compound are included with the software as an “AI library” database that also includes database search function based on the mass spectral pattern. Additionally, msFineAnalysis AI uses the molecular formulas uniquely determined during automatic integrated qualitative analysis in order to narrow down the possible candidate structural formulas.

The predicted EI mass spectrum narrowed down by molecular formula and the actual EI mass spectrum are used to then calculate a score from the similarity of their spectral pattern, and the candidate structural formulas are then listed in order of high similarity to low similarity.

Figure 1: Workflow for structural analysis of unknowns using msFineAnalysis

Experimental

A commercially-available polyethylene terephthalate film was used as a test sample in this study. We performed Py-GC-HRTOFMS measurements using both EI and field ionization (FI) modes with a combination EI/FI ion source. The qualitative data processing was performed with msFineAnalysis AI (JEOL). Measurement conditions are shown in Table 1.

Pyrolysis Conditions
Pyrolyzer	EGA/PY-2020D (Frontier Lab)
Pyrolysis Temperature	600°C
GC Conditions
Gas Chromatograph	7890 GC
Gas Chromatograph	(Agilent Technologies)
Column	ZB-5MSi (Phenomenex)
Column	30 m × 0.25 mm, 0.25 μm
Oven Temperature	40°C (2 min) - 20°C/min
Oven Temperature	-320°C (30 min)
Injection Mode	Split mode (100:1)
Carrier flow	He: 1.0 mL/min

MS Conditions
Spectrometer	JMS-T200GC (JEOL Ltd.)
Ion Source	EI/FI combination ion source
Ionization	EI+: 70 eV, 300 μA
Ionization	FI+: -10 kV
Mass Range	m/z 29 - 600
Data Processing Conditions
Software	msFineAnalysis AI (JEOL Ltd.)
Library database	NIST20, AI Library (JEOL Ltd.)

Table 1: Measurement and analysis conditions

Results and Discussion

Comparison of AI structural analysis results and references

Among the observed polyethylene terephthalate (PET) pyrolysis products, AI structural analysis was performed for four components not registered in the NIST library database and for which structural formulas were proposed in reference [1]. Figure 2 shows the TIC chromatograms obtained from the Py-GC-EI and FI measurements. The peaks with IDs [058], [059], [115], and [159] in Figure 2 are the four components analyzed in this study. Figure 3 shows the measured EI mass spectra for these four components (upper, black), the structural formula proposed in the reference literature (right side of the spectrum), and its predicted EI mass spectrum (lower, red).

Figure 2: Py-GC-EI and FI TIC chromatograms for Poly (ethylene terephthalate)

Figure 3: Measured EI mass spectra and predicted EI mass spectra of the proposed structural formula in reference [1]
for ID [058], [059], [115], [159] in Figure 2

AI structural analysis results are shown in Table 2. In the table, the "AI Score" is a score (up to 999) calculated by msFineAnalysis AI that represents the cosine similarity between the measured and predicted EI mass spectra. "Rank" indicates the score rank of the structural formulas listed in Figure 3, and "Total" indicates the number of candidate structural formulas. Three of the four components analyzed in this study (ID [058], [059], [115]) obtained over 800 score which means a high degree of similarity, and the fragment ions observed in the measured mass spectra and the predicted mass spectra were in good agreement. The number of candidate structural formulas ware about 970-1400, and the structural formulas proposed in the reference were obtained as within the top two.

For ID [159], the AI Score was calculated at 510 score, and although this score was lower than the other components, the structural formula proposed in the reference literature was obtained at #3 rank in 311 candidates.

As a result, it was found that the structural formulas proposed in the reference literature were obtained in the top three positions for all the components evaluated in this study.

Reference [1] data
Notation	Assignment of Main Peaks
C	CH2=CHOCOC6H4COOCH=CH2
D	CH2=CHOCOC6H4COOH
F	C6H5COOCH2CH2OCOC6H4COOCH=CH2
H	C6H5COOCH2CH2OCOC6H4COOCH2CH2OCOC6H5

msFineAnalysis AI result
ID	RT (min)	IUPAC name	PubChem CID	AI Score	Rank	Total
058	10.49	bis(ethenyl) benzene-1,4-dicarboxylate	15374889	820	2	1471
059	10.81	4-ethenoxycarbonylbenzoic acid	22223159	914	2	1013
115	15.52	1-O-(2-benzoyloxyethyl) 4-O-ethenyl benzene-1,4-dicarboxylate	101115782	877	1	975
159	22.66	bis(2-benzoyloxyethyl) benzene-1,4-dicarboxylate	53951693	510	3	311

Table 2: AI structural analysis results

Conclusion

Structural analysis using AI was performed on four components not registered in the NIST library database, and results were compared with structural formulae proposed in the reference literature.

In spectral pattern comparisons, Three of the four components analyzed in this study (ID [058], [059], [115]) cosine similarity scores were over 800, indicating that AI-predicted mass spectra showed a high degree of similarity to measured mass spectra. The number of candidate structural formulas ware about 970-1400, and the structural formulas proposed in the reference were obtained as within the top two. For ID [159], the AI Score was calculated at 510 score, and although this score was lower than the other components, the structural formula proposed in the reference literature was obtained at #3 rank in 311 candidates. The prediction by AI showed high accuracy, indicating that the method is effective for structural analysis of pyrolysis products.

Qualitative analysis of GC-MS data can be greatly assisted by using EI and SI data together with msFineAnalysis AI, especially when trying to identify unknown compounds in complex samples.

Reference

[1] Shin Tsuge, Hajime Ohtani, Chuichi Watanabe (2011), Pyrolysis - GC/MS Data Book of Synthetic Polymers, Elsevier

Structural Analysis of Additives and Related Compounds in Vinyl Acetate by using a Py-GC-HRTOFMS and msFineAnalysis AI

Fri, 10 Oct 2025 09:44:50 GMT

MSTips No. 390

Introduction

Electron ionization (EI) is one of the most popular ionization methods used in gas chromatography-mass spectrometry (GC-MS). Consequently, compounds are typically identified by a mass spectral database search using EI mass spectra. Because molecular ions are often weak or absent in 70 eV EI mass spectra, identification of unknowns can be difficult by EI alone. In these cases, soft ionization (SI) can be very helpful for producing and identifying molecular ions. Recently, JEOL began developing an integrated qualitative analysis workflow that automatically combines and interprets the information from EI and SI data. And then in 2018, we introduced our integrated qualitative analysis software "msFineAnalysis" which uses both EI and SI data to improve compound identification for GC-MS applications.

AI Structural Analysis

The AI structural analysis workflow is shown in Figure 1. In this method, we used deep learning to construct an AI model that can predict the EI mass spectrum from a structural formula. We then submitted approximately 100 million compound structure formulas to our AI model in order to generate predicted EI mass spectra. The structural formula and the predicted EI mass spectra associated with each compound are included with the software as an "AI library" database that also includes database search function based on the mass spectral pattern. Additionally, msFineAnalysis AI uses the molecular formulas uniquely determined during automatic integrated qualitative analysis in order to narrow down the possible candidate structural formulas.

Figure 1: Workflow for structural analysis of unknowns using msFineAnalysis

Experimental

A commercially-available vinyl acetate resin was used as a test sample in this study. We performed Py-GC-HRTOFMS measurements using both EI and field ionization (FI) modes with a combination EI/FI ion source. The qualitative data processing was performed with msFineAnalysis AI (JEOL). Measurement conditions are shown in Table 1.

Pyrolysis Conditions
Pyrolyzer	EGA/PY-2020D (Frontier Lab)
Pyrolysis Temperature	600°C
GC Conditions
Gas Chromatograph	7890 GC
Gas Chromatograph	(Agilent Technologies)
Column	DB-5msUI (Agilent)
Column	15 m × 0.25 mm, 0.25 μm
Oven Temperature	50°C (1 min) - 30°C/min
Oven Temperature	-330°C (1.7 min)
Injection Mode	Split mode (100:1)
Carrier flow	He: 1.5 mL/min

MS Conditions
Spectrometer	JMS-T200GC (JEOL Ltd.)
Ion Source	EI/FI combination ion source
Ionization	EI+: 70 eV, 300 μA
Ionization	FI+: -10 kV, 6m A/10 msec (Carbotec)
Mass Range	m/z 35 - 800
Data Processing Conditions
Software	msFineAnalysis AI (JEOL Ltd.)
Library database	NIST20, AI Library (JEOL Ltd.)

Table 1: Measurement and analysis conditions

Results and Discussion

Figure 2 shows the TIC chromatograms obtained from the Py-GC-EI and FI measurements. Diethylene Glycol Dibenzoate used as a plasticizer was detected at RT 8.45 min (ID [134]). In addition, compounds presumed to be pyrolysis products of Diethylene Glycol Dibenzoate were detected in the RT 5.02 min (ID [085]), RT 5.98 min (ID [108]) and RT 6.98 min (ID [127]). All of these compounds were not registered in the NIST library database. Therefore, we performed AI structural analysis on these three components and compared them with the structural formulas of pyrolysis products expected to be generated from Diethylene Glycol Dibenzoate.

The measured EI mass spectra of these three components (top, black) and the predicted structural formula (right side of the spectrum) and its predicted EI mass spectrum (bottom, red) are shown in Figure 3. Regarding Diethylene Glycol Dibenzoate, the EI mass spectrum registered in the NIST library database is displayed in blue at the bottom, not the predicted EI mass spectrum.

Figure 2: Py-GC-EI and FI TIC chromatograms for Poly(vinyl acetate)

Figure 3: Measured EI mass spectra and predicted EI mass spectra of the proposed structural formula for ID[085], [108], [127], [134] in Figure 2

AI structural analysis results are shown in Table 2. In the table, the "AI Score" is a score (up to 999) calculated by msFineAnalysis AI that represents the cosine similarity between the measured and predicted EI mass spectra. "Rank" indicates the score rank of the structural formulas listed in Figure 3, and "Total" indicates the number of candidate structural formulas. All three of the components analyzed in this study obtained a score of 850 or higher, indicating a high degree of similarity, and the fragment ions observed in the measured mass spectra and the predicted mass spectra were in good agreement. The number of candidate structural formula all exceeded 2,000 for each of the components, the presumed structural formula was obtained as the first or second candidate for all three components.

msFineAnalysis AI result
ID	RT (min)	IUPAC name	PubChem CID	AI Score	Rank	Total
085	5.02	2-Ethenoxyethyl benzoate	21930739	858	1	4600
108	5.98	2-(2-Hydroxyethoxy)ethyl benzoate	88603	857	2	4544
127	6.98	2-[2-(2-Hydroxyethoxy)ethoxy]ethyl benzoate	89963	872	1	2433

Table 2: AI structural analysis result

Conclusion

In this MSTips, we introduced our newly-developed software msFineAnalysis AI, which contains AI structural analysis functionality to enhance qualitative analysis workflow. Additionally, the structural analysis application of additives and related compounds in vinyl acetate using msFineAnalysis AI was also presented.

Structural analysis using AI was performed on three components not registered in the NIST library database, and results were compared with the structures of pyrolysis products expected from Diethylene Glycol Dibenzoate. In spectral pattern comparisons, all cosine similarity scores were over 850, indicating that AI-predicted mass spectra showed a high degree of similarity to measured mass spectra. Even though the number of candidate structural formula exceeded 2,000 for each of the components, the presumed structural formula was obtained as the first or second candidate for all three components. The prediction by AI showed high accuracy, indicating that the method is effective for structural analysis of pyrolysis products and additives.

Qualitative analysis of GC-MS data can be greatly assisted by using EI and SI data together with msFineAnalysis AI, especially when trying to identify unknown compounds in complex samples.

Structural Analysis of Acrylic Resin Oligomers by using a Py-GC-HRTOFMS and msFineAnalysis AI

Fri, 10 Oct 2025 09:28:53 GMT

MSTips No. 389

Introduction

Electron ionization (EI) is one of the most popular ionization methods used in gas chromatography-mass spectrometry (GC-MS). Consequently, compounds are typically identified by a mass spectral database search using EI mass spectra. Because molecular ions are often weak or absent in 70 eV EI mass spectra, identification of unknowns can be difficult by EI alone. In these cases, soft ionization (SI) can be very helpful for producing and identifying molecular ions. Recently, JEOL began developing an integrated qualitative analysis workflow that automatically combines and interprets the information from EI and SI data. And then in 2018, we introduced our integrated qualitative analysis software "msFineAnalysis" which uses both EI and SI data to improve compound identification for GC-MS applications.

AI Structural Analysis

The AI structural analysis workflow is shown in Figure 1. In this method, we used deep learning to construct an AI model that can predict the EI mass spectrum from a structural formula. We then submitted approximately 100 million compound structure formulas to our AI model in order to generate predicted EI mass spectra. The structural formula and the predicted EI mass spectra associated with each compound are included with the software as an "AI library" database that also includes database search function based on the mass spectral pattern. Additionally, msFineAnalysis AI uses the molecular formulas uniquely determined during automatic integrated qualitative analysis in order to narrow down the possible candidate structural formulas.

Figure 1: Workflow for structural analysis of unknowns using msFineAnalysis

Experimental

A commercially-available acrylic resin was used as a test sample in this study. We performed Py-GC-HRTOFMS measurements using both EI and field ionization (FI) modes with a combination EI/FI ion source. The qualitative data processing was performed with msFineAnalysis AI (JEOL). Measurement conditions are shown in Table 1.

Pyrolysis Conditions
Pyrolyzer	EGA/PY-3030D (Frontier Lab)
Pyrolysis Temperature	600°C
GC Conditions
Gas Chromatograph	8890 GC
Gas Chromatograph	(Agilent Technologies)
Column	ZB-5MSi (Phenomenex)
Column	30 m × 0.25 mm, 0.25 μm
Oven Temperature	40°C (2 min) - 10°C/min
Oven Temperature	-320°C (15 min)
Injection Mode	Split mode (100:1)
Carrier flow	He: 1.0 mL/min

MS Conditions
Spectrometer	JMS-T2000GC (JEOL Ltd.)
Ion Source	EI/FI combination ion source
Ionization	EI+: 70 eV, 300 μA
Ionization	FI+: -10 kV, 40 mA/30 msec
Mass Range	m/z 35 - 800
Data Processing Conditions
Software	msFineAnalysis AI (JEOL Ltd.)
Library database	NIST20, AI Library (JEOL Ltd.)

Table 1: Measurement and analysis conditions

Results and Discussion

Comparison of AI structural analysis results and references

Among the observed acrylic resin pyrolysis products, AI structural analysis was performed for four components not registered in the NIST library database and for which structural formulas were proposed in reference [1]. Figure 2 shows the TIC chromatograms obtained from the Py-GC-EI and FI measurements. The peaks with IDs [038], [040], [055], and [063] in Figure 2 are the four components analyzed in this study. Figure 3 shows the measured EI mass spectra for these four components (upper, black), the structural formula proposed in the reference literature (right side of the spectrum), and its predicted EI mass spectrum (lower, red).

Figure 2: Py-GC-EI and FI TIC chromatograms for an Methyl methacrylate-methyl acrylate copolymer

Figure 3: Measured EI mass spectra (upper, black) and predicted EI mass spectra (lower, red) of the proposed structural formula in reference [1] for ID[038], [040], [055], [063] in Figure 2

AI structural analysis results are shown in Table 2. In the table, the "AI Score" is a score (up to 999) calculated by msFineAnalysis AI that represents the cosine similarity between the measured and predicted EI mass spectra. "Rank" indicates the score rank of the structural formulas listed in Figure 3, and "Total" indicates the number of candidate structural formulas. All four of the components analyzed in this study obtained a score of 750 or higher, indicating a high degree of similarity, and the fragment ions observed in the measured mass spectra and the predicted mass spectra were in good agreement. The number of candidate structural formulae all exceeded 3,000, but in three of the four components, structural formulas proposed in the reference literature were obtained within the top 1% of the candidates.

Reference [1] data
Notation	Assignment of Main Peaks
d2	C=C(C)-C-C(C)(COOC)-C ?
d4	C=C(C)-C=C(COOC)-C ?
A2'	C=C(COOC)-C-C(COOC)-C ?
D1	C=C(COOC)-C-C(C)(COOC)-C ?

msFineAnalysis AI result
ID	RT (min)	IUPAC name	PubChem CID	AI Score	Rank	Total
038	7.41	Methyl 2,2,4-trimethylpent-4-enoate	12512240	872	2	5548
040	8.32	Methyl 2,4-dimethylpenta-2,4-dienoate	71327190	865	18	3769
055	11.04	Dimethyl 2-methyl-4-methylidenepentanedioate	12037869	753	37	3109
063	11.69	Dimethyl 2,2-dimethyl-4-methylidenepentanedioate	10035672	825	9	3732

Table 2: AI structural analysis results

Conclusion

Structural analysis using AI was performed on four components not registered in the NIST library database, and results were compared with structural formulae proposed in the reference literature. In spectral pattern comparisons, all cosine similarity scores were over 750, indicating that AI-predicted mass spectra showed a high degree of similarity to measured mass spectra. Even though the number of candidate structural formulae exceeded 3,000 for each of the components, the structural formulae proposed in the reference literature for three of the four components was in the top 1% of candidates. The prediction by AI showed high accuracy, indicating that the method is effective for structural analysis of pyrolysis products.

Qualitative analysis of GC-MS data can be greatly assisted by using EI and SI data together with msFineAnalysis AI, especially when trying to identify unknown compounds in complex samples.

Reference

[1] Shin Tsuge, Hajime Ohtani, Chuichi Watanabe (2011), Pyrolysis - GC/MS Data Book of Synthetic Polymers, Elsevier

Introduction of AI Structure Analysis Function in Automatic Structure Analysis Software msFineAnalysis AI

Fri, 10 Oct 2025 09:09:17 GMT

MSTips No. 388

Introduction

Electron ionization (EI) is one of the most popular ionization methods used in gas chromatography-mass spectrometry (GC-MS). Consequently, compounds are typically identified by a mass spectral database search using EI mass spectra. Because molecular ions are often weak or absent in 70 eV EI mass spectra, identification of unknowns can be difficult by EI alone. In these cases, soft ionization (SI) can be very helpful for producing and identifying molecular ions. Recently, JEOL began developing an integrated qualitative analysis workflow that automatically combines and interprets the information from EI and SI data. And then in 2018, we introduced our integrated qualitative analysis software "msFineAnalysis" which uses both EI and SI data to improve compound identification for GC-MS applications.

Despite the fact that msFineAnalysis was automatically able to determine the molecular formula and partial structure information from EI fragment ion formulas, the actual structural formulas still required manual analysis using chemical compositions. To address this, we then developed an automated structure analysis software package entitled "msFineAnalysis AI" which uses artificial intelligence (AI) to predict EI mass spectra from chemical structures. We have used our newly-developed AI model to create a database of predicted EI mass spectra for around 100 million compounds. In this work, we introduce AI structure analysis function in automatic structure analysis software msFineAnalysis AI.

Figure 1 Image of analysis result in msFineAnalysis AI

About AI Structure Analysis Function

AI structure analysis function performs automatic structure analysis for unknown compounds using two AIs (main AI, support AI) that complementarily combine machine learning and deep learning.

Figure 2 shows the workflow of AI structural analysis by the main AI. In the main AI, a model for EI mass spectra prediction from structural formulas was constructed using deep learning, and predicted EI mass spectra of 100 million compounds were included in the software as an "AI library" database. The database search function using the "AI library" is implemented similarly to traditional library searches using the commercially available EI mass spectra database. Structural formula candidates are narrowed down by molecular formulas uniquely determined by integrated qualitative analysis, so more correct structural formulas can be obtained quickly. The predicted EI mass spectra were compared with measured EI mass spectra, then the scores were calculated from the spectral patterns, and candidate structural formulas were arranged in order of highest score. Finally, the correct structural formula is selected by combining the obtained structural formula candidates with the sample information and the knowledge and know-how obtained from the previous analysis.

Figure 3 shows the workflow of partial structure prediction by the support AI. The support AI assists interpreting analysis results by predicting the partial structure from the measured EI mass spectrum. It is possible to analyze the composition formula of fragment ions and neutral losses obtained from accurate mass analysis and assist in the interpretation of structural information proposed by the main AI.

Figure 2 Main AI workflow

Figure 3 Support AI workflow

GUI of AI Structure Analysis Result

Figure 4 shows the AI structure analysis result of Acrylic Resin Oligomers by msFineAnalysis AI. The target of analysis is a dimer component that is not registered in the NIST library database. The left side of the analysis result screen shows the structure candidates by the main AI, and the right side shows the analysis results by the support AI. Detailed structural information can be obtained even for unknown compounds that have not been registered in the database.

On the main AI analysis result screen, a list of predicted structural formulas is shown at the bottom of the screen, and it is possible to check the AI structural analysis results all at once. The AI score indicates the similarity between the AI library and the measured mass spectrum, and it is shown at the bottom of each structural formula. Furthermore, information on the selected structural formula is posted at the top of the screen. We can see where the selected structural formula is in the histogram. It also includes a filtering function by partial structure and monomer, which enables structural analysis results to reflect the presence or absence of substructures predicted by the support AI described below.

On the support AI analysis result screen, predicted partial structure information is shown at the bottom of the screen. On the list, the left side is the partial structure predicted to be present, and the right side is the partial structure predicted not to be present. The partial structure with blue background matches the structural formula selected in the main AI, while the partial structure with red background does not match. Measured mass spectrum and the predicted composition formula of each fragment ion/neutral loss is posted at the top of the screen. It is also possible to confirm and edit comments for each estimated composition formula.

Figure 4 GUI of msFineAnalysis AI

Conclusion

In this MSTips, we introduced our newly-developed software msFineAnalysis AI, which contains AI structural analysis functionality to enhance qualitative analysis workflow. This software performs automatic structure analysis for unknown compounds using two AIs (main AI, support AI) that complementarily combine machine learning and deep learning. No knowledge of mass spectrometry and AI are required as the software automatically interprets complex mass spectra.

Qualitative analysis of GC-MS data can be greatly assisted by using EI and SI data together with msFineAnalysis AI, especially when trying to identify unknown compounds in complex samples.

Qualitative Analysis of Chemical Components in an Herbal Medicine

Fri, 10 Oct 2025 08:59:21 GMT

MSTips No.433

Introduction

Herbal medicines contain multiple effective ingredients, making it difficult to study them systematically compared to Western drugs. Initial steps in examining effective ingredients of an herbal medicine include qualitative analysis of chemical components contained in its raw material. For this application, gas chromatography - mass spectrometer (GC-MS) is widely used. However, typical database search based on GC-MS often fails to fully identify such chemical components.

JEOL’s GC-high resolution mass spectrometer (GC-HRMS), JMS-T2000GC, is capable of: 1) high precision mass analysis; 2) detection of molecular and fragment ions through electron ionization (EI) and soft ionization (SI); and 3) auto analysis of acquired data by msFineAnalysis AI to determine chemical formulas and predict chemical structures. In this work, we identified the chemical components contained in an herbal medicine using a JMS-T2000GC and msFineAnalysis AI.

Measurement

Dry wild chrysanthemum flowers (Flos Chrysanthemi Indici), a commercial herbal medicine, was used as a sample. Following the steps shown in Figure 1, 1.1 g of the sample was immersed in 10 mL of ethanol (≥99.5%) and subjected to ultrasonic extraction at room temperature for 30 minutes.

The supernatant of the extracted liquid was collected, added anhydrous sodium sulfate, and let stand for 30 minutes.

The resulting solution was injected to the GC-HRTOFMS. EI and FI (Field Ionization) were used as ionization methods for GC-HRTOFMS measurement, and msFineAnalysis AI was subsequently used for analysis of the acquired data. Table 1 shows the measurement conditions.

GC-HRMS
Gas Chromatograph	8890 GC (Agilent Technologies, Inc.)
Mode	Splitless
Inlet temperature	280 °C
Column	DB-5MS, 30 m x 0.25 mm, 0.25 μm (Agilent Technologies, Inc.)
Oven	50 °C (1 min) → 10 °C/min → 320°C (10 min)
Carrier gas	He, 1.0 mL/min
Injection volume	1 μL
TOFMS	JMS-T2000GC (JEOL Ltd.)
Ionization	EI+: 70 eV, 300 μA
Ionization	FI+: -10 kV, 40 mA
Monitor ion range	m/z 10-800
Analysis software	msFineAnalysis AI (JEOL Ltd.)

Table 1. Measurement conditions

Figure 1. Experimental process

Results and Discussion

Figure 2 shows the total ion current chromatograms (TICCs) of EI and FI. A total of 141 components were detected, including eucalyptol, camphor, borneol, and bornyl acetate marked by ● in the figure. These 4 components are reported to be active compounds⁽¹⁾.

For 81 components, their chemical formulas predicted from the molecular ions did not agree with the results of NIST database search.

However, msFineAnalysis AI contains an AI database of EI mass spectra estimated from the chemical structures in addition to the NIST database. The 81 components, when examined against the AI database, turned out to be the compounds, which were in good agreement with EI fragment ions patterns and had the chemical formulas predicted by molecular ions. Figure 4 shows the results. These peaks were identified primarily as terpenes, terpenoids, fatty acids, alkanes, and sterols. The figure shows their groups and detection ranges.

Figure 2. EI and FI TICCs of Flos Chrysanthemi Indici ethanol extract.

Figures 3 and 4 respectively show the mass spectra and predicted chemical structure of the peak marked by ▼ in Figure 2, which was detected at a retention time of 21.96 minutes.

Its chemical formula was estimated to be C₁₇H₁₈O₅ from the exact mass verified in the FI mass spectrum in Figure 3. NIST database search for the EI mass spectrum did not result in any compounds showing a similarity of 700 or more.

Meanwhile, Figure 4 shows the results of msFineAnalysis AI structural analysis of (2R)-4,7-dimethyl-1-propan-2-yl-1,2,3,5,6,8a-hexahydronaphthalen-2-ol. The AI-predicted mass spectrum and measured mass spectrum of the compound have high levels of AI score and AI probability, providing effective structural information.

Figure 3. EI and FI mass spectra of the component at a retention time of 21.96 minutes in FlosChrysanthemi Indici ethanol extract.

Figure 4. Structural Analysis results of the component at a retention time of 21.96 minutes by msFineAnlaysis AI.

Summary

The results of this work demonstrates the effectiveness of msFineAnalysis AI. It is a powerful and efficient tool for qualitative analysis, saving time and labor in chemical formula determination and chemical structure prediction. Accurate qualitative analysis of herbal components is useful for discovery and quality control in the field of herbal medicine. The GC-HRMS system with msFineAnalysis AI is expected to be used for wide-ranging qualitative analysis of chemical components in natural resources including herbal medicines.

References

[1] Jiesheng Ye,. Chemical Papers, 2009, 63, 5, 506-511. DOI: 10.2478:-009 to -0056:-0

Qualitative Analysis of Impurities in a Solvent for Semiconductor Fabrication using msFineAnalysis AI

Fri, 10 Oct 2025 08:42:48 GMT

MSTips No.434

Introduction

Reliability of semiconductor devices is determined by the level of cleanliness of silicon wafers. Therefore, it is critical to prevent contamination from the impurities in cleaning solutions. To process any cleaning solution to an ultra-pure level, it is important to identify the chemical components of impurities to be removed. For this purpose, gas chromatography-mass spectrometer (GC-MS) is widely used.

Recently, techniques used for qualitative analysis are increasingly sophisticated, including exact mass analysis using GC-high resolution mass spectrometer (GC-HRMS), NIST database of electron ionization (EI) mass spectra, and molecular ion analysis using soft ionization (SI). Meanwhile, volumes of data acquired from exact mass analysis are enormous, requiring expertise in MS data analysis and vast amounts of time for interpretation.

JEOL’s latest analytical software, msFineAnalysis AI, is designed to provide speedy analysis of GC-HRMS data acquired by both EI and SI, determine chemical formulas, and predict chemical structures. In this work, we used msFineAnalysis AI to identify impurities in 2-methoxy-1-methylethyl acetate (PGMEA), a cleaning solution for wafer surfaces.

Measurement

A commercial PGMEA (≥99.5%) was used as a sample. As a GC column, Rtx-BAC PLUS1 was used because impurities were expected to be highly polar, similarly to PGMEA. EI and FI (Field Ionization) were used as ionization methods. msFineAnalysis AI was subsequently used for analysis of the acquired data. Table 1 shows the measurement conditions.

Table 1. Measurement conditions

Results and Discussion

Figure 1 shows the total ion current chromatograms (TICCs) of EI and FI. Other than air and PGMEA, a total of 12 components, considered to be impurities, were detected. These components other than air and water were verified in both EI and FI TICCs. For some components, their formulas predicted from the molecular ions did not agree with the results of NIST database search.

However, msFineAnalysis AI contains an AI database of EI mass spectra estimated from the chemical structures in addition to the NIST database. These components, when examined against the AI database, turned out to be the compounds, which were in good agreement with fragment ions patterns and had the chemical formulas predicted by molecular ions.

Water molecules, which can be easily mixed with PGMEA, as well as 1-methoxypropan-2-ol[1], which is used as a synthetic raw material for PGMEA, were identified as impurities.

2-methoxypropyl acetate, a structural isomer of PGMEA, was also identified as an impurity.

Figure 1. EI and FI TICCs of PGMEA

Figure 2. EI and FI mass spectra of the component at a retention time of 6.61 minutes
Upper: EI, Lower: FI

Figures 2 and 3 respectively show the mass spectra and predicted chemical structures of the peak marked by ▼ in Figure 1, which was detected at a retention time of 6.61 minutes.

No molecular ion was detected in the EI mass spectrum in Figure 2. The molecular ion of this component was detected in the FI mass spectrum, demonstrating the importance of using soft ionization as well. The exact mass of the molecular ion detected suggested C8H18O3 as its chemical formula. NIST database search for the EI mass spectrum did not result in any compounds showing a similarity of 750 or more.

However, the AI structural analysis in Figure 3 resulted in 3-(3-hydroxybutan-2-yloxy)butan-2-ol, which had a chemical formula of C8H18O3 and also showed good agreement between the measured and AI predicted EI fragment patterns.

Figure 3. EI and FI mass spectra of the component at a retention time of 6.61 minutes

Summary

The results of this work demonstrate the effectiveness of msFineAnalysis AI. It is a powerful and efficient tool for qualitative analysis, saving time and labor in chemical formula determination and chemical structure prediction. In addition to qualitative analysis of impurities in organic solvents including PGMEA, the GC-HRMS system with msFineAnalysis is expected to be effective for qualitative analysis of organic contaminants on semiconductor surfaces by examining ROSE test solutions before and after cleaning semiconductor devices.

Reference

[1] Arif Hussain, Yus Donald Chaniago, Amjad Riaz, Moonyong Lee,. Ind. Eng. Chem. Res. 2019, 58, 6, 2246−2257.
DOI: 10.1021/acs.iecr.8b04052

Quantitative NMR Analysis Using JASON SMILEQ: Novel Methods for Improving Accuracy, Part 3. Elucidating Factors through Simulation Analysis

Thu, 02 Oct 2025 12:46:52 GMT

How Does the Uncertainty of Standard Samples Impact Quantitative Analysis Results?

Uncertainty Report and ANOVA Report Results

From the previous analysis, it has been confirmed that repeated errors across the entire measurement system are very small, demonstrating the stability of the measurement process. On the other hand, it is suggested that the uncertainty associated with standard samples may propagate throughout the measurement results.

Analyzing the Impact of Standard Sample Uncertainty

Standard samples are often difficult to substitute, and there are challenges in directly testing their uncertainty impact on measurement results through experimentation. To address these challenges, detailed analyses using simulations prove effective. In this study, computational methods were employed to clarify the tendencies of uncertainty caused by standard samples affecting measurement results.

Analysis Method

The influence of standard sample uncertainty on the integral values obtained through experiments and the resulting quantitative values was investigated. This analysis employed the following methods to examine the impact in detail. Calculations were performed using Python^Ⓡ3 and the report data.

Analysis of Coefficients of Variation: The standard deviation (SD) and coefficient of variation (CV) of integral and quantitative values in experimental data were compared to assess the impact of standard sample uncertainty on the data.
Analysis Using Sensitivity Coefficients: Simulations based on sensitivity coefficients were conducted to analyze the effects of standard sample uncertainty.
Analysis Using Monte Carlo Method: The Monte Carlo method was employed to simulate the influence of uncertainty on the entire measurement and to analyze the detailed characteristics of data distribution.

1. Analysis of Coefficients of Variation

Coefficient of Variation (CV)

The coefficient of variation is a metric that expresses the variability of data relative to the mean. It is calculated by dividing the standard deviation by the mean and is typically displayed as a percentage. This metric is effective in assessing the variability of measurements and serves as an important tool for determining the stability of measurement systems.

Coefficient of Variation After Quantitative Calculations

Figure 1 compares the standard deviation and coefficient of variation for integral values (Non-Normalized) and quantitative values (Normalized) before quantitative calculations. Both sets of values were normalized using their average values to analyze variations more effectively. This approach focused on examining the amplitude of fluctuations and ensured a consistent basis for comparison. After the calculations, the coefficient of variation was found to be as small as 0.21%, confirming that these corrections successfully reduced data variability. Additionally, it was suggested that the final quantitative values might be highly dependent on the characteristics of the standard samples.

Figure 1. Comparison of Standard Deviation and Coefficient of Variation

2. Analysis Using Sensitivity Coefficients

Sensitivity Coefficient

The sensitivity coefficient is a quantitative metric that indicates the extent to which each factor impacts measurement or calculation results. Specifically, it is used to assess how small variations in individual factors contribute to the outcomes. By utilizing this metric, critical factors within a system or analytical model can be identified. Below, the significance and details of simulations based on sensitivity coefficients are explained.

Significance of Simulations Utilizing Sensitivity Coefficients

By applying sensitivity coefficients in simulations, it becomes possible to effectively quantify how specific factors influence the results. In this study, the contribution of standard sample characteristics to overall measurement results was assessed, and their tendencies were clarified.

Simulation Results Using Sensitivity Coefficients

Simulation Range

Simulations were conducted to evaluate the impact on quantitative values when the uncertainty of standard samples (0.25%) fluctuated within its surrounding range.

Simulation Results

From the simulations, the average values and standard deviations of the quantitative values were calculated (Figure 2 (a)). Figure 2 (b) presents a plot comparing the standard deviations of the quantitative calculation results and simulation results against the integral values of the standard samples in the experimental data. The trends in standard deviation were consistent with the experimental results, confirming the accuracy of the simulation's approach.

Characteristics of Experimental Results

The experimental results showed slightly higher values compared to the computational model results, suggesting the possible influence of other factors. Below is an explanation of the impacts of the uncertainties under consideration on the simulation:

Characteristics of Small Uncertainty Ranges
A narrow uncertainty range implies that the variability captured by the model is minimal, making the contribution of specific factors more prominent. In scenarios dominated by the uncertainty of standard samples, external factors are more likely to have a significant impact within this limited range of variability.
Impact on Model Responsiveness
In modeling with small uncertainty ranges, the sensitivity coefficient may not fully reflect the variability. Specifically, the extent to which the model can account for external factors becomes a critical point. Simulations capable of appropriately reproducing the cumulative effects of small fluctuations are more likely to exhibit realistic behavior, even within a narrow range of uncertainty.

Conclusion and Model Applicability

In cases where the uncertainty range is small, the sensitivity and corrections of the model play a crucial role. However, when accounting for complex external factors and interactions, comprehensive methods like the Monte Carlo approach may prove more effective. Especially when experimental data exhibits higher variability or deviation than the model, using such comprehensive methods enables modeling that aligns more closely with the experimental environment.

3. Analysis Using Monte Carlo Method

Overview of the Monte Carlo Method

The Monte Carlo method is a technique that combines random sampling with statistical methods to analyze complex problems. By repeatedly performing numerous simulations based on the distribution of input variables, it identifies the distribution and tendencies of the output results. This method is particularly useful in the following ways:

Reproducing the Behavior of Complex Systems: Capable of analyzing overall behavior even in scenarios involving numerous factors.
Evaluating Uncertainty: Examines in detail how uncertainty impacts results.
Analyzing Entire Distributions: Allows for a visual understanding of not only mean values but also variability and ranges of output results.

Below, we explain the significance and details of simulations utilizing the Monte Carlo method.

Significance of Simulations Using the Monte Carlo Method

The Monte Carlo method was employed to analyze the impact of standard sample uncertainty on the variability and distribution of measurement results. This approach clarified not only the mean values and standard deviations but also the range and shape of variability in measurement results, enabling a comprehensive understanding of the overall data distribution. Additionally, the contributions of standard samples to the stability of the measurement system were quantitatively evaluated, providing direction for improving reliability.

Simulation Results Using the Monte Carlo Method

Simulation Range

Simulations were conducted to evaluate the impact on quantitative values when the uncertainty of standard samples (0.25%) fluctuated around the baseline. The ranges of variation included ±0.05% (i.e., 0.20%-0.30%) and ±0.10% (i.e., 0.15%-0.35%).

Modeling and Statistical Evaluation of Uncertainty

Uncertainty was analyzed using probabilistic models to assess the trends in data distribution. The variations of standard samples were statistically reproduced using uniform distributions and Gaussian distributions (normal distributions), and these results were compared with experimental data. Figure 3 shows a plot comparing the calculated results with experimental results: (a) Distribution of quantitative values. (b) Comparison of normalized mean values, standard deviations, and coefficients of variation.

Evaluation Methods for Simulation Results

To evaluate the simulation results, the most realistic model was considered by calculating the following score. This score was computed as a metric to quantify the variability of measurement results and is based on the formula:
(Score) = (Mean Difference (%) + SD Difference (%) + CV Difference (%)) /3

Mean Difference (%) Evaluates the difference between the average values of experimental results and simulation results, indicating the overall alignment.
SD Difference (%) Assesses the difference in standard deviations between experimental data and simulation data, measuring the degree of alignment in variability.
CV Difference (%) Evaluates the difference in coefficients of variation, focusing on relative variability in the data.

By integrating these three differences and calculating the average, the score was determined. This score is effective for comprehensively evaluating data alignment and the impact of uncertainty.

Figure 3. Simulations Using the Monte Carlo Method: (a) Simulation results (b) Comparison of mean values, standard deviations, and coefficients of variation

Most Consistent Model

A uniform distribution within the uncertainty range of 0.35% (0.9965 to 1.0035) achieved a score of 2.05%, confirming it as the best fit to the experimental data. Figure 4 shows a plot comparing the simulation results and experimental results at an uncertainty of 0.35%. The results are interpreted as follows:

Uncertainty Range of Standard Samples
Although an uncertainty of 0.25% was set as the theoretical standard, the simulation results demonstrated that a range of 0.35% was the most consistent with the experimental data. This difference suggests that the uncertainty of the standard samples may propagate as an error factor throughout the experiment.
Analysis of Distribution Shape
The better fit of the uniform distribution compared to the Gaussian distribution implies that the experimental environment lacks significant variations. In scenarios where a uniform distribution is a better fit, it is likely that the environment and processes are relatively stable, with variations confined to a consistent range.
Ripple Effects and Error Factors
The simplicity and uniformity of fluctuations in the overall experimental system suggest that the uncertainty of standard samples directly impacts the variability of measurement data, reflecting it as an overall error factor.

Figure 4. Comparison of Data Distribution and Experimental Results within a 0.35% Uncertainty Range

Summary of Simulation Results Using Sensitivity Coefficients and the Monte Carlo Method

Differences in Approaches

The sensitivity coefficient was utilized as a method to locally quantify the contribution of standard samples and to analyze in detail the impact of specific factors on measurement results. This approach focuses on particular factors and evaluates their influence in depth. On the other hand, the Monte Carlo method comprehensively analyzes the random behavior of multiple factors, aiming to reproduce the overall behavior of measurement results. It emphasizes visualizing overall distributions and ripple effects.

Differences in Objectives

The objective of sensitivity coefficient analysis was to locally evaluate the influence of specific factors on measurement results. This analysis clarified the extent to which standard sample uncertainty contributes to the results. Conversely, the Monte Carlo analysis aimed to reproduce the overall distribution of measurement data and to examine the ripple effects of uncertainty, as well as the potential influence of external factors and interactions beyond the standard samples.

Overall Conclusions

The simulation results using sensitivity coefficients indicated that the uncertainty of standard samples significantly impacts the overall trend of uncertainty in measurement results. Additionally, the results suggested that external factors and interactions, beyond standard samples, might slightly influence the measurement outcomes. The Monte Carlo results showed that a uniform distribution with an uncertainty of 0.35% (0.9965–1.0035) best fit the experimental data. This finding revealed that the uncertainty of standard samples (0.25%) is a primary source of variability, while its effects are slightly amplified by other factors. Moreover, the suitability of a uniform distribution suggests that the overall experimental environment is relatively simple and exhibits stable fluctuations within a defined range.

SMILEQ Report: Summary of Uncertainty Factor Analysis

Based on the results of the uncertainty report and ANOVA report, the primary factors affecting quantitative analysis results were analyzed. Additionally, simulations were utilized to conduct a detailed analysis of uncertainty factors. In particular, the SMILEQ report's indications regarding the impact of standard sample uncertainty were reproduced, allowing for a more quantitative evaluation of its contribution to overall measurement results. Although evaluations of results influenced by extended uncertainty have been conducted previously, there are few cases where additional factor analysis has clarified the underlying causes. From the findings of this report, it is demonstrated that comprehensive analysis using the SMILEQ report provides specific methods and directions for improving the accuracy and reproducibility of quantitative analysis.

[1] JEOL Analytical Software Network
[2] Spectral Management Interface Launching Engine for Q-NMR
[3] Python is a registered trademark of the Python Software Foundation.

Quantitative NMR Analysis Using JASON SMILEQ: Novel Methods for Improving Accuracy, Part 2. Analysis of Factors through Variance Analysis

Thu, 02 Oct 2025 12:46:18 GMT

From Uncertainty Report to Variance Analysis

Deviation in the Uncertainty Report:
The Uncertainty Report calculates Expanded Uncertainty by analyzing interactions between individual factors and overall data variations. This serves as a critical foundation for integrated analysis of uncertainty across the entire measurement process.

Role of Variance Analysis:
Variance analysis (ANOVA: Analysis of Variance) is a statistical method for determining how multiple factors in data affect the results. It isolates the pure deviations of each factor and clarifies their contribution levels. By separating interactions between factors, ANOVA establishes a solid basis for detailed analysis of uncertainty sources.

Comparison Through the ANOVA Report:
The results of the ANOVA Report allow for thorough analysis of uncertainty sources and the contribution rates of individual factors. This facilitates the development of clear guidelines for improving the measurement process. Such analysis plays an essential role in enhancing data reliability and accuracy.

Variance Analysis: Two-way ANOVA

Two-way ANOVA (Two-factor Analysis of Variance) is a method used to evaluate the effects of two different factors (Factor A and Factor B) on data, as well as their interaction. Below are the primary elements involved in the calculation:

Factor A Evaluates the impact of Factor A on the data.
Factor B Evaluates the impact of Factor B on the data.
Interaction Assesses the effects of the interaction between Factor A and Factor B on the data (Factor A × Factor B).
Error Represents random variability not attributable to the above factors or their interaction.

The Total Variance is determined by combining the contributions of all these elements (Factor A, Factor B, Interaction, and Error). Figure 1 illustrates the mechanism of Two-way ANOVA through a schematic diagram.

Figure 1. Schematic Diagram of Two-way ANOVA

JASON ANOVA: 2 way ANOVA

JASON Variance Analysis provides reports based on two distinct analytical models. The JASON 2 way ANOVA focuses on the effects of independent factors, offering a simple method for individually evaluating elements that influence the data. Its main features include:

Factor A (Sample) Evaluates the independent contribution of Sample without considering direct relationships with Signal, treating it as a background element.
Factor B (Signal) Assesses the direct impact of Signal on the data and evaluates its independent contribution.
Interaction Does not include independent interaction terms.

This approach is particularly suitable when differences between samples are minimal or when the effects of Signal itself are the main focus. Compared to JASON’s other model, the 2-way nested ANOVA, it offers a simpler structure that is effective for analyzing individual factors. Figure 2 presents a schematic diagram illustrating the mechanism of the JASON 2 way ANOVA.

Figure 2. Schematic Diagram of JASON 2 way ANOVA

JASON ANOVA: 2-way nested ANOVA

JASON 2-way nested ANOVA is a comprehensive method for evaluating the relationships and interactions between multiple factors. Its main features include:

Factor A (Sample) Assesses elements influencing data across multiple levels. Similar to the 2 way ANOVA, it is treated as a background factor.
Factor B (Signal) Analyzes the direct impact on data, evaluating it in relation to the levels of Sample.
Interaction Independent interaction terms are not included.

This model is particularly suitable for scenarios where Sample's contribution is critical or where Signal's effect needs to be analyzed in conjunction with Sample interactions. It supports complex data analysis and yields more accurate results. Figure 3 illustrates a schematic diagram of the mechanism of JASON 2-way nested ANOVA. Additionally, Table 1 compares the differences between the two analytical models.

Figure 3. Schematic Diagram of JASON 2-way nested ANOVA

Table 1. Differences Between 2 way ANOVA and 2-way nested ANOVA

Details of the ANOVA Report

In the ANOVA report, variations attributed to each factor are calculated as Mean Square Variance, a key metric for interpreting variance analysis and quantitatively assessing the contribution of each factor. From the Mean Square Variance, deviations for individual factors can be calculated. These deviations, determined independently of other factors or interactions, are referred to as Pure Deviation. By using Pure Deviation, it is possible to extract uncertainty that captures the impact of each factor individually. This process quantitatively evaluates the precision and reliability of the data, laying the groundwork for deeper insights into analytical results. Figure 4 provides an overview of the metrics included in the 2 way ANOVA report, along with detailed explanations.

Figure 4. ANOVA Report: (a) 2 way ANOVA report, (b) Metrics and Descriptions

Comparison Between 2 way ANOVA and Uncertainty Report Results

Figure 5 compares the results of the 2 way ANOVA and the uncertainty report (Application Note NM250002) using a radar chart. The purple color represents the results of the 2 way ANOVA, while the blue color indicates the results of the uncertainty report. The following trends are observed:

Sample Shows lower values compared to the uncertainty report, suggesting that only the pure effects have been extracted.
Signal Displays smaller values, clearly indicating that the influence of other factors and interactions has been reduced.
Repetition Shows very small values, reflecting a high level of stability in the measurement process itself.

Differences between the Results of the Uncertainty Report The differences between the results of the two reports can be explained by the characteristics outlined in Table 2. The uncertainty report provides an evaluation that considers the "worst-case scenario" and tends to assess repeated errors as relatively large. On the other hand, variance analysis offers an approach that reflects the "overall characteristics of the data," extracting uncertainties isolated from other factors. Additionally, variance analysis enables detailed analysis of individual factors.

Figure 5. Comparison Between 2 way ANOVA and Uncertainty Report Results

Table 2. Key Characteristics of the Uncertainty and ANOVA Reports

Analysis of Differences from the Uncertainty Report Results

To analyze the differences from the uncertainty report results, the following steps are undertaken to examine the variance analysis outcomes. First, the validity of the variance analysis results is verified. Then, an analysis of factors is conducted using the results of the two variance analysis models.

Verification of Data Statistical Properties
The statistical properties of the underlying data are evaluated to confirm the validity of the uncertainty and variance analysis results. The report data was analyzed using Python for this evaluation. The following methods were employed:
- IQR Test
- Shapiro-Wilk Test
- QQ Plot
- KDE Plot
Factor Assessment in Variance Analysis
The contribution of data variability and uncertainty to each factor is assessed. This process involves comparing the results of the 2 way ANOVA and the 2-way nested ANOVA to organize the relative impact of each factor.

1. Verification of Data Statistical Properties: Outlier Examination (IQR Test)

The IQR Test (Interquartile Range Method) is one of the techniques used to detect outliers in data. The IQR is defined as the difference between the third quartile (Q3) and the first quartile (Q1), representing the range of the central 50% of the data. In the box plot shown in Figure 6, the following elements are illustrated:

Box (IQR): Represents the range from Q1 to Q3.
Line inside the Box: Indicates the median of the data (50th percentile).
Whiskers: Represent the overall range of the data.

As a result of this analysis, no outliers were detected, and it was confirmed that the variability within the data is small.

Figure 6. IQR Test Results

1. Verification of Data Statistical Properties: Normality Check

Shapiro-Wilk: Test The Shapiro-Wilk test is a statistical method for testing the normality of data. In this test, the null hypothesis assumes that "the data follows a normal distribution." The calculated p-value of 0.6092 is significantly greater than 0.05, confirming that the data follows a normal distribution.

QQ Plot (Quantile-Quantile Plot): The QQ plot is used to compare the distribution of data against a theoretical normal distribution. The closer the points of the actual data align with the straight line representing the theoretical values, the higher the normality. Figure 7 (a) illustrates the QQ plot created using the calculated results. This plot statistically confirms the normality of the data.

KDE Plot (Kernel Density Estimate): The KDE plot is a method for smoothing the distribution of data, providing a more accurate display of data density compared to histograms. Figure 7 (b) shows the KDE plot created using the calculated results. This plot visually confirms the normality of the data.

Assessment of Variance Analysis Validity: Based on the verification results from the above methods, the validity of the variance analysis has been confirmed. Normality has been evaluated both statistically and visually, providing a reliable foundation for trustworthy analytical results.

Figure 7. Verification of Data Statistical Properties: (a) QQ Plot, (b) KDE Plot

2. Factor Assessment in Variance Analysis

Figure 8 compares the results of the 2 way ANOVA and 2-way nested ANOVA using a radar chart. The purple color represents the results of the 2 way ANOVA, while the darker purple indicates the results of the 2-way nested ANOVA. The following trends were observed:

Sample Both models show similar values, confirming the stability of variability. However, the 2 way ANOVA results suggest the potential influence of interactions with other factors.
Signal The 2-way nested ANOVA shows lower values, while the 2 way ANOVA produces relatively higher values. These findings indicate that the Signal might be sensitive to repeated errors, Sample, or interactions with standard samples.
Repetition The values for repeated errors are nearly identical across both models (0.02), confirming a high degree of stability. Additionally, it was noted that the overall variability of the measurement data is kept to a minimum.

Figure 8. Results of JASON ANOVA

Summary of Variance Analysis Results

Based on comparisons with the uncertainty report, the following points have been confirmed through variance analysis:

Validity of Variance Analysis The uncertainty associated with each factor has been statistically validated, and its influence on quantitative analysis results has been clarified.
Repeated Errors Repeated errors show extremely small values, supporting the stability of the measurement process and reliability of the data.
Impact of Standard Samples The uncertainty from standard samples has been found to propagate throughout the measurement, establishing them as key influential factors.
Impact of Interactions Interactions between Signal and Sample may contribute to the uncertainty within the measurement data.

These findings provide critical insights for enhancing the reliability of quantitative analysis.

How Does the Uncertainty of Standard Samples Impact Quantitative Analysis Results?

From the previous analysis, it has been confirmed that repeated errors across the entire measurement system are very small, demonstrating the stability of the measurement process. However, further investigation is needed to explore how the uncertainty associated with standard samples affects the measurement results. To address this issue, detailed analyses using methods such as simulations would be effective. Particularly, evaluating the impact of standard sample characteristics on the overall data can provide crucial insights to improve measurement accuracy and reliability. The detailed exploration of this analysis will be covered in "Part 3. Elucidating Factors through Simulation Analysis."

[1] JEOL Analytical Software Network
[2] Spectral Management Interface Launching Engine for Q-NMR

JEOL Resources

AccuTOF™ GC-Alpha & LIFDI

The best analytical tool for organometallics and air-sensitive samples

msFineAnalysis AI Novel Qualitative Analysis Software for JMS-T2000GC with AI Structural Analysis

Structural Analysis of Polyethylene Terephthalate Film by using a Py-GC-HRTOFMS and msFineAnalysis AI

Introduction

AI Structural Analysis

Experimental

Results and Discussion

Comparison of AI structural analysis results and references

Conclusion

Reference

Structural Analysis of Additives and Related Compounds in Vinyl Acetate by using a Py-GC-HRTOFMS and msFineAnalysis AI

Introduction

AI Structural Analysis

Experimental

Results and Discussion

Conclusion

Structural Analysis of Acrylic Resin Oligomers by using a Py-GC-HRTOFMS and msFineAnalysis AI

Introduction

AI Structural Analysis

Experimental

Results and Discussion

Comparison of AI structural analysis results and references

Conclusion

Reference

Introduction of AI Structure Analysis Function in Automatic Structure Analysis Software msFineAnalysis AI

Introduction

About AI Structure Analysis Function

GUI of AI Structure Analysis Result

Conclusion

Qualitative Analysis of Chemical Components in an Herbal Medicine

Introduction

Measurement

Results and Discussion

Summary

References

Qualitative Analysis of Impurities in a Solvent for Semiconductor Fabrication using msFineAnalysis AI

Introduction

Measurement

Results and Discussion

Summary

Reference

Quantitative NMR Analysis Using JASON SMILEQ: Novel Methods for Improving Accuracy, Part 3. Elucidating Factors through Simulation Analysis

How Does the Uncertainty of Standard Samples Impact Quantitative Analysis Results?

Uncertainty Report and ANOVA Report Results

Analyzing the Impact of Standard Sample Uncertainty

Analysis Method

1. Analysis of Coefficients of Variation

Coefficient of Variation (CV)

Coefficient of Variation After Quantitative Calculations

2. Analysis Using Sensitivity Coefficients

Sensitivity Coefficient

Significance of Simulations Utilizing Sensitivity Coefficients

Simulation Results Using Sensitivity Coefficients

Simulation Range

Simulation Results

Characteristics of Experimental Results

Conclusion and Model Applicability

3. Analysis Using Monte Carlo Method

Overview of the Monte Carlo Method

Significance of Simulations Using the Monte Carlo Method

Simulation Results Using the Monte Carlo Method

Simulation Range

Modeling and Statistical Evaluation of Uncertainty

Evaluation Methods for Simulation Results

Most Consistent Model

Ripple Effects and Error Factors The simplicity and uniformity of fluctuations in the overall experimental system suggest that the uncertainty of standard samples directly impacts the variability of measurement data, reflecting it as an overall error factor.

Summary of Simulation Results Using Sensitivity Coefficients and the Monte Carlo Method

Differences in Approaches

Differences in Objectives

Overall Conclusions

SMILEQ Report: Summary of Uncertainty Factor Analysis

Quantitative NMR Analysis Using JASON SMILEQ: Novel Methods for Improving Accuracy, Part 2. Analysis of Factors through Variance Analysis

From Uncertainty Report to Variance Analysis

Variance Analysis: Two-way ANOVA

JASON ANOVA: 2 way ANOVA

JASON ANOVA: 2-way nested ANOVA

Details of the ANOVA Report

Comparison Between 2 way ANOVA and Uncertainty Report Results

Ripple Effects and Error Factors
The simplicity and uniformity of fluctuations in the overall experimental system suggest that the uncertainty of standard samples directly impacts the variability of measurement data, reflecting it as an overall error factor.