Effect of porphyrin structure on aroma sensing response based on DFT: A case study of black tea fermentation

Abstract

Porphyrin-based colorimetric sensing arrays (CSA) are frequently used to evaluate the black tea quality, but the mechanisms behind the interaction between porphyrin structures and volatile organic compounds (VOCs) remain unclear. Six VOCs that can be used to distinguish the degree of fermentation in black tea were identified and analysed. Response surface optimisation revealed the optimal conditions for CSA: reaction temperature of 65 °C, reaction time of 8 min and dye volume of 5 μL. The predictive coefficients and relative predictive deviations of the simplified quantitative model ranged from 0.82 to 0.94 and 1.72 to 2.92, respectively. Finally, density functional theory (DFT) calculations were used to elucidate the intrinsic relationship between the structural features of porphyrin molecules and their responsiveness to sensing by analysing variations in binding energy, dipole moment, charge and bond length. This provides a theoretical basis for constructing targeted CSA to monitor aroma.

Highlights

• •A total of 6 key VOCs of black tea fermentation were screened based on SPME-GC/MS.
• •DFT was used to study porphyrin-VOC structure-function relationship.
• •The CSA reaction conditions were optimised using response surface analysis.
• •The accuracy of porphyrin modelling with p < 0.05 was 94.12 %.
• •The theoretical basis for constructing a targeted aroma monitoring CSA is provided.

1. Introduction

Black tea is popular with consumers for its strong aroma, clean and mellow taste, and unique quality (Fang et al., 2023). The fermentation of black tea involves substantial changes in volatile organic compounds (VOCs). By monitoring changes in the levels of key VOCs during fermentation, the degree of fermentation can be determined (Chen et al., 2022).

Traditional sensory review methods are heavily influenced by the experience and subjectivity of the assessor (Li et al., 2022). By contrast, gas chromatography is accurate and objective but is also expensive and technically challenging to implement (Wang et al., 2020). Methods such as the electronic nose are affected by the humidity of the environment in which black tea is fermented, leading to unstable test results (Banerjee et al., 2019). Given the practical obstacles associated with these instrumental methods in terms of cost, technical requirements, and environmental adaptability, relying on experienced operators to conduct sensory evaluations remains the most feasible and widely used method for assessing fermentation levels for most manufacturers.

The colorimetric sensor array (CSA) is a sensing technology that mimics the mammalian olfactory system (Suslick et al., 2007). CSA technology has been successfully applied to evaluations of tea aroma. Wang designed a cellulose CSA to monitor the withering dynamics of black tea (Wang, Cai, et al., 2025). Li proposed a simple CSA combined with a hyperspectral system to enable the rapid, quantitative analysis of black tea aroma (Li et al., 2022). An combined hyperspectral imaging technology with a CSA to evaluate the aroma quality of fermented black tea (An et al., 2022). These studies have enabled the rapid detection of black tea aromas. However, simple CSAs are unable to detect subtle differences in flavour, and to improve responsivity, researchers have employed various substrate materials, such as silica gel plate substrates and paper (Li, Askim, & Suslick, 2019). The strength of the response exhibited by porphyrins employed in CSAs can be improved by using nanoscale porphyrins, which exert a small size effect, a quantum tunnelling effect, and a surface effect (Chen et al., 2021). Metal-organic frameworks (MOFs) have become a promising surface-modified functional material for colorimetric sensing systems as a functionalized modification material that can significantly improve detection sensitivity and selectivity (Mantovani et al., 2007).

Although researchers have improved the response effect in CSAs by changing the materials used in the CSA and optimising the experimental conditions, the mechanism through which dyes bind to volatile organic compounds remains unclear, and the construction of CSAs is currently based on empirical results. The detection ability of a CSA is poor if the array cannot be constructed in a targeted manner. To further investigate the binding mechanism between porphyrin and VOCs, Waluk studied the levels of individual porphyrin molecules by using optical and scanning probe microscopy (Waluk, 2017). Saleh described the principle of porphyrin-based MOF materials interacting with volatile compounds and outlined their future prospects (Saleh et al., 2025). Yang used a microemulsion-assisted self-assembly method to prepare nano-sized indium porphyrin and explained its working principle (Yang et al., 2024). Previous research has employed DFT calculations to investigate CSA interactions with tea aroma. Dong used DFT to explore the colorimetric sensing mechanism of metal porphyrins in relation to roasting degree (Dong et al., 2025). Kang analysed the response mechanism of CSA to fermented VOCs in kombucha extracts using DFT calculations (Kang et al., 2023). In studies investigating the binding mechanism between porphyrin structures and VOCs, Gu employed DFT calculations to analyse the molecular interactions between silver porphyrin and VOCs (Gu et al., 2016). Similarly, Wang utilized DFT to explore the response mechanism of metal porphyrins and ZIF-8 to VOCs (Wang, Shoaib, et al., 2025). Gu investigated the binding mechanisms of eight different central-atom porphyrins with trimethylamine (Gu et al., 2020). The above study only involved a single porphyrin structure and was unable to fully reveal the relationship between structure and sensing ability. In the present study, the binding energies, bond lengths, dipole moments, and charges of porphyrins with different structures reacting with VOCs were investigated through DFT calculations. The results refined the theory of the ‘structure-response’ mechanism, revealed the mechanism of action of porphyrin structures on sensing capabilities, and provided theoretical guidance for the subsequent targeted construction of applicable CSAs.

This study employed DFT to investigate the influence of the structure of porphyrins and the MOF materials used in CSAs on a CSA's response properties. This study did so in a case study where a CSA was applied to the quantitative analysis of characteristic aromas in black tea fermentation. The following experiments were performed. (1) This study used solid-phase microextraction (SPME)–gas chromatography (GC)/mass spectrometry (MS) to screen key the VOCs generated during black tea fermentation. (2) Response surface methodology was used to determine the reaction duration, temperature, and dye volume that maximise CSA response. (3) The binding energies, dipole moments, charges, and bond lengths of the reaction between porphyrin and VOCs were calculated using DFT. The results were employed to compare the effects of the central metal atom, axial coordination, and peripheral substituents on the porphyrins' responsiveness. (4) Finally, correlation analyses combined with support vector regression (SVR) and least squares support vector regression (LS-SVR) were used to quantify the aromas that are characteristic of specific stages of black tea fermentation. Specific information is displayed in Illustration of the experiment flow-chart.

2. Experimental

2.1. Chemicals and reagents

Benzaldehyde, (E)-2-hexenal, decanal, (E)-linalool oxide (furanoid), methyl salicylate and geraniol were purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). All odorants used for GC analysis had purities exceeding 95 %. Specifically, ethyl decanoate (≥99 % purity) was obtained from Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). The n-alkane standards (C₅-C₄₀) were acquired from Sigma-Aldrich (St. Louis, MO, USA). Additional reagents including absolute ethanol (≥99 %) and dimethylacetamide were procured from Yongda Chemical Reagent Co. (Tianjin, China). A C2 reverse silicone plate was obtained from Merck KGaA (Darmstadt, Germany). Porphyrins and MOF were purchased from Sigma Aldrich Chemical Co., Inc. (Shanghai, China). Specific material information is displayed in Table S1.

2.2. Sample materials

The tea samples used in this study were collected from Youfangdian Jinzhai Xiangyuan Company, Lu'an City, Anhui Province, China. The tea variety is Shuchazao, and the fermentation process is carried out in accordance with the standard manufacturing process for black tea. A chamber with humidity of 90 % ± 5 % and temperature of 30 °C ± 3 °C was used to withering and ferment the fresh leaves. Fermentation was conducted continuously over a period of 5 h, Fermentation humidity is 90 %, temperature is 30 °C. Samples weighing 500 g were taken at half-hourly intervals during this process and stored in a refrigerator at 4 °C, finally, transfer to a laboratory freezer set at −80 °C for storage.

Table S2 presents results of a sensory review of the characteristics of black tea corresponding to specific fermentation durations, namely, the tea's appearance, infusion, flavour, taste, and infused leaf. These five characteristics were defined to contribute 25 %, 10 %, 25 %, 30 %, and 10 %, respectively, to the overall determination of fermentation stage. The evaluation was performed with reference to the evaluation factor scoring coefficient for Congou black tea (three replicates) in the GB/T 23,776–2018 standard. In this study, a fermentation duration of 2.5–3.0 h was considered to result in moderate fermentation, and samples fermented for 0.0–2.0 and 3.5–5.0 h were considered to have been insufficiently and excessively fermented, respectively. Further information is presented in Table S2.

The black tea fermentation sensory review involved seven volunteers aged 22–45 years who authorised sensory review data. Specific information on the sensory review is presented in the Human Sensory Ethical Inspection.

2.3. Detection of volatile compounds

Headspace SPME was used to extract the VOCs generated during the fermentation of black tea. First, 0.6 g of freeze-dried black tea at one of three fermentation levels was placed in a 20 mL headspace sample bottle. Then, 5 mL of boiling water was added, after which 40 μL of 5 ppm ethyl decanoate for internal standard quantification, and 1.5 g of sodium chloride powder to stimulate the aroma of the tea leaves (Wang et al., 2024). Finally, the bottle was sealed with polytetrafluoroethylene and equilibrated for 5 min at 60 °C. An activated SPME fibre tip was inserted into the headspace vial, and adsorption was allowed to proceed for 50 min, these steps ensure that the aroma of the tea is fully preserved. Finally, GC/MS detection was performed per a method developed by the research team (Huang et al., 2022).

Quantification was performed through both internal and external standard methods. The concentrations of aroma-active compounds were calculated using ethyl decanoate as an internal standard and a standard curve generated from standard compounds of known concentration. OAVs were calculated as the ratio of the concentration of each aroma-active compound in water to its respective threshold value. Aroma-active compounds with OAV > 1 are substances for potential aromatic characterisation. Variable importance in projection (VIP) values were used in multivariate statistical analyses. Metabolomics data were subjected to principal component analysis and hierarchical cluster analysis, conducted using SIMCA-P 14.1 software (Conrado et al., 2021). Model quality was assessed using R² and Q². Major volatiles were indicated by both OAV > 1 (Ma et al., 2022) and VIP > 1 (Wold et al., 2001).

2.4. CSA system and information acquisition

2.4.1. Preparation of CSA

Through single factor experiments, this study determined that dye volume, reaction duration, and reaction temperature were the three factors that should be considered in the response surface optimisation. Five temperatures were considered: 50 °C, 55 °C, 60 °C, 65 °C, and 70 °C. Five dye volumes were employed: 4.0, 4.5, 5.0, 5.5, and 6.0 μL. Finally, five durations were used: 7.0, 7.5, 8.0, 8.5, and 9.0 min. The optimal parameters and thus the optimal reaction performance of the CSA were determined by combining the stabilisation time of the reaction substrate silica gel plate and the three information acquisition modes. To simulate the real-world environment in which black tea is fermented, the environmental humidity was set to 70 %, 80 %, and 90 % and the environmental temperature was set to 20 °C, 30 °C, and 40 °C; this ensured the feasibility of the CSA's application.

In this study, nine porphyrin materials were reacted with key VOCs, and five porphyrin materials (Table S1) were selected on the basis of the reaction strength. Porphyrins of three different structures from these five materials were then proportionally mixed with an MOF material to sensitise them and to form a 2 × 4 CSA. The optimised parameters were employed to apply the CSA to the key VOCs generated in black tea fermentation. Finally, the strength of the reaction with the VOCs was assessed using the CSA Euclidean distance, differences in prereaction and postreaction images, and quantitative modelling.

The experimental method was as follows. Three porphyrin materials and an MOF material were weighed and dissolved in 2 mL of N,N-dimethylacetamide (Li, Xie, et al., 2019). The resultant solutions were sonicated for 30 min and stored away from light. The porphyrin and MOF materials were then mixed in three ratios—1:1, 1:3, and 3:1—and the mixtures were sonicated for 15 min. A C₂ reversed-phase silica gel plate (Merck KGAA, Frankfurt, Germany) was used as a substrate for CSA; 5 μL of each dye was absorbed by a pipette gun, applied to the silica gel plate, and stabilised under a fume hood for 12 min. Experiments were performed after this stabilisation period. The VOCs (E)-2-hexenal, benzaldehyde, methyl salicylate, geraniol, (E)-linalool oxide (furanoid), and decanal were diluted with anhydrous ethanol at concentrations of 0.01, 0.1, 0.5, 1, 5, and 10 ppm. This was set based on the actual concentration detected in the black tea samples.

2.4.2. CSA response and information collection

Before the stabilised CSA was used in an experiment, it was passed through a scanner for images of its initial state to be obtained. Then, the key VOCs were placed in a Petri dish of 75 mm diameter, and the colour-sensitive plate was fixed on the lid of the Petri dish (with the dyes facing inward); 10 mL of standard solution was transferred to the Petri dish with a pipette, and the dish was immediately placed in an oven at 65 °C for 8 min, during which reactions occurred. Immediately after the reactions, the colour-sensitive plate was removed and placed in a scanner for images to be obtained，using an canon lide 400 (Canon, Japan). All collected images had a resolution of 6016 × 4016 pixels and were saved in JPEG format. Twenty parallel experiments were conducted simultaneously. Difference images obtained from the initial and final images were created using Adobe Photoshop and MATLAB, and Euclidean distance processing was performed to determine the CSA response intensity. In addition, Pearson's correlation and a correlation heat map were used for analysis. The colour difference was calculated as follows:

$$∆R=\left|R_a-R_b\right|$$ (1)

$$\begin{array}{c}∆G=\left|G_a-G_b\right|\end{array}$$ (2)

$$\begin{array}{c}∆B=\left|B_a-B_b\right|\end{array}$$ (3)

Ra, Ga, and Ba are the initial red, green, and blue gray values before colour sensitivity point response, while Rb, Gb, and Bb correspond to the red, green, and blue gray values after colour sensitivity point response. These variables provide characteristic information for subsequent analysis. The Euclidean distance is used as a response intensity indicator, and the formula is as follows:

$$\begin{array}{c}\mathrm{ED}=\sqrt{\left(∆R^2+∆G^2+B^2\right)}\end{array}$$ (4)

2.5. DFT calculations

DFT computations were performed using Gaussian 16 software (Revision B01). The popular hybrid density functional B3LYP [2–4], with the D3 version of Grimme's dispersion with the Becke–Johnson damping function (Grimme et al., 2011), was used for these computations. The Turbomole series double zeta basis set, Def2-SVP, was used for modelling the main group atoms and transition metal atoms. Atomic charge was analysed through natural population analysis. The independent gradient model based on the Hirshfeld partition (IGMH) was implemented using the Multiwfn program (Lu & Chen, 2012), and the VMD program (version 1.9.3) was used to analyse the interaction between the tetraphenylporphyrin (TPP) molecules and VOCs (Lu & Chen, 2022). Binding energies, bond lengths, dipole moments, charges, and electron clouds were calculated. The binding energies of interactions were calculated as follows:

$$\begin{array}{c}{E}_{\mathit{binding}}=E_{\mathit{AB}}-\left(E_A+E_B\right)\end{array}$$ (5)

where E_AB is the electronic energy of the optimised molecule AB after binding and E_A and E_B are the electronic energies of the optimised isolated molecules A and B.

The mechanism of the influence of the metal centre, axial ligand and peripheral substituents on the binding mode and response strength is revealed by the Independent gradient model based on Hirshfeld partition (IGMH), the IGMH model was used to visualise the interactions arising from the response of porphyrins to VOCs, the interaction regions are usually visualised using equivalent surfaces of σginter and σgintra, and the interactions are represented by different colours (Lu & Chen, 2021).

Data visualisations were generated using Origin 2021 (OriginLab, MA, USA) and ChiPlot (ChiPlot), and variable selection and model construction were performed using MATLAB 2014a (Mathworks, Natick, USA).

2.6. Data processing and quantitative predictive modelling

In this study, SVR and LS-SVR algorithms were used to construct a quantitative model for predicting the key VOCs generated during the fermentation of black tea. First. A total of 120 data samples were randomly divided at a ratio of 2:1 to ensure sample uniformity. Each sample includes colour information from the RGB, HSV and LAB colour spaces of the sensor. The calibration set comprises 80 samples and is primarily used to train the model parameters. The prediction set comprises the remaining 40 samples and is solely used to evaluate the predictive performance of the trained model on the data (Jiang et al., 2022). For both SVR and LSSVR models, hyperparameter optimisation is conducted via cross-validation, with the optimal parameters then used to train the final model (Viscarra Rossel et al., 2006). The square root of the calibration coefficient of determination (Rc), root mean square calibration error, square root of the prediction coefficient of determination (Rp), root mean square prediction error, and relative prediction deviation (RPD) were used as evaluation metrics. The evaluation indices for RPDs were as follows: very good: (RPD ≥ 2.0), excellent: (1.8 ≤ RPD < 2.0), and fair: (1.4 ≤ RPD < 1.8) (Munnaf et al., 2021).

3. Results and discussion

3.1. Screening for key VOCs in black tea fermentation

The results of orthogonal partial least squares discriminant analysis (OPLS-DA) of 44 VOCs in black tea samples fermented for different durations, presented in Fig. 1A(a), indicated that the samples could be well distinguished, hierarchical Cluster Analysis (HCA) also verified this result in Fig. 1A(b). The reliability of the established OPLS-DA model was verified through cross validation, presented in Fig. 1A(c). Fig. S1A shows that 20 substances—including benzaldehyde, (E)-2-hexenal, (E)-2-octenal, (E)-linalool oxide (furanoid), (E,E)-2,6-nonadienal, methyl salicylate, and geraniol—had VIP > 1. This indicated that these VOCs had a strong influence in the fermentation of black tea. (See Scheme 1.)

Fig. 1.

Analysis of aroma substances in black tea fermentation process. A: a) Orthogonal partial least squares discriminant analysis (OPLS-DA) results; and b) HCA cluster analysis and c) Results of 120 permutation of OPLS-DA; B: Changes in the content of key VOCs during fermentation. C: Heat map of fermented aroma of black tea.

Scheme 1.

Illustration of the experiment flow-chart.

The results obtained using the aroma activity value method are presented in Fig. S1B. Of the 20 substances with VIP > 1, six were screened with OAV > 1 and quantified using external standard curves, which revealed content changes during the fermentation process. These changes are illustrated in Fig. 1B. Of the key VOCs, benzaldehyde and (E)-linalool oxide (furanoid) were found to increase with the degree of fermentation, whereas decanal and methyl salicylate decreased. Notably, the (E)-2-hexenal content increased and then decreased. The changes in the type of aroma could be summarised as a decrease in grassiness and an increase in pleasant scents, such as floral, fruity, and sweet notes, as the degree of fermentation was increased.

The total volatile aroma content is illustrated in Fig. 1C. The contents of alcohols, aldehydes, and other aromatic compounds gradually increased with fermentation duration. The total amount of each aroma reached its maximum value at a fermentation duration of 3.5 h.

3.2. Construction and optimisation of CSA

The results of porphyrin screening are shown in Fig. S2C. The porphyrins were reacted with 10 ppm of geraniol, and five porphyrins were selected for subsequent experiments on the basis of their different structures and response strengths. The selected porphyrins were TPP, CoTPP, MnTPP, MnClTPP, and MnClTPP [p-PhSO₂]. MOF material was added to MnTPP, MnClTPP, and MnClTPP [p-PhSO₂] at 3:1 for sensitisation studies.

The optimal response conditions are displayed in Fig. 2B. A temperature of 65 °C, reaction duration of 8 min, and dye volume of 5 μL were the parameters that yielded the optimal CSA response. The CSA response was found to be strongly influenced by temperature and duration. The optimal silica gel plate stabilisation time was 12 min (Fig. 2A(b)), and the use of the scanner was the best image acquisition method (Fig. 2A(c)).

Fig. 2.

CSA response condition optimisation. A: a) CSA stability verification results. b) Optimization of silicone plate stability time; c) Optimization of information acquisition methods. B: Optimization of CSA reaction conditions response surface methodology.

Black tea is fermented in a highly humid environment, and the stability of the constructed CSA was determined by simulating such an environment. The results are shown in Fig. 2A(a). Overall, CSA response gradually increased with reaction temperature. The Euclidean distance was found to increase by 18 % when the reaction temperature was increased from 30 °C to 40 °C and to increase by 11 % when the reaction temperature was increased from 40 °C to 50 °C. At 90 % humidity, the effective data remained within a reasonable range. The CSA was discovered to not be greatly affected by humidity.

3.3. Analysis of the response effect of the CSA

The difference images are presented in Fig. 3A. CSA response increased with concentration; this is represented in the graph by an increase in the brightness of the dye image. With regard to the sensitisation effect, the addition of MOF material increased the brightness of the difference images. The MOF-sensitised peripherally substituted porphyrin MnClTP P[p-PhSO₂] had the highest brightness, and this finding agreed with the Euclidean distance results.

Fig. 3.

Response results of CSA with key aroma of black tea fermentation. A: CSA differential images of black tea fermentation aroma response. B: CSA versus black tea fermentation aroma Euclidean distance line plot. C: Porphyrin and CSA correlation.

The quantitative response of the CSA to the key VOCs is illustrated in Fig. 3B. Addition of MOF material increased the intensity of the response for all dyes, with the increase most pronounced for MnClTPP [p-PhSO₂]. The introduction of peripheral substituents enhanced the Euclidean distance to a greater degree than the introduction of axial coordination groups. In addition, the porphyrin containing sulfonated phenyl in the periphery of MOF-MnClTPP [p-PhSO₂] reacted best and led to the highest Euclidean distance reaction values. The responsiveness of the porphyrin materials increased as the VOC concentration was increased.

In order to facilitate the observation of differences in sensing performance between porphyrin structures and subsequent refined modelling of porphyrin arrays. The correlations between the porphyrin reaction effects and aroma substances were analysed by constructing a Mantel test correlation heat map (Fig. 3C). The map revealed strong correlations between the same porphyrin structures. TPP, MnTPP, MnClTPP, MnClTPP[p-PhSO2] show great differences between the different structures. Mantel's test showed a significant correlation between the porphyrin structure and the interaction with VOCs. MOF-MnClTPP[p-PhSO₂] showed strong positive correlation with geraniol (r = 0.90, p = 0.002), (E)-linalool oxide (furanoid) (r = 0.82, p = 0.001) and (E)-2-hexenal (r = 0.81, p = 0.002), whereas MOF- MnClTPP was weakly correlated with substances such as benzaldehyde (r = −0.25, p = 0.865). CoTPP was moderately correlated with decanal (r = 0.69, p = 0.01), while TPP was moderately correlated with methyl salicylate (r = 0.73, p = 0.001) and (E)-2-hexenal (r = 0.68, p = 0.001). This result confirms that metal center type and peripheral substituent modification modifications are key factors in modulating the binding selectivity of volatile organic compounds.

3.4. DFT analyses

3.4.1. Selection of base state molecules

Before DFT calculations could be performed, the models of individual molecules had to be constructed, and whether the constructed molecular models reflected the true structure of the molecules had to be determined. The geometry was then optimised using appropriate generalisations and basis sets. The results are presented in Table S3. Fig. 4D illustrates the weak interaction between TPP and VOCs, with strong hydrogen bonding very strong hydrogen bonding and van der Waals forces between the central atom (Mn²⁺ and CO²⁺) and the VOCs, but the presence of axial ligands (Cl⁻ and OAc⁻) reduces the interaction force between the two. In terms of peripheral substituents (p-PhSO₂ and p-CH₃), although the PhSO₂ region shows a higher red/yellow colour (repulsive force). However, the charge modulation of p-PhSO₂ and the extended range of action may have improved the binding ability to VOCs.

Fig. 4.

Gaussian calculations of charge, dipole moment variation, and IGMH diagrams for the reaction of CSA with VOCs. A and B: Changes in the dipole moment of porphyrins reacting with volatile organic compounds. C: Dipole moments of porphyrin reacting with VOCs. D: IGMH plot of the optimised porphyrin molecule versus ligand.

3.4.2. Binding energies

The binding energy is the energy released when particles combine from a free state to form a composite particle (Zhang et al., 2019); a higher binding energy corresponds to a more stable molecular structure. The calculated average binding energies for porphyrin materials reacting with VOCs are shown in Fig. 5A. The absolute values of these energies were, in descending order, as follows: MnTPP (−28.15 kcal/mol) > CoTPP (−26.86 kcal/mol) > MnClTPP [p-PhSO₂] (−25.95 kcal/mol) > MnClTPP [P-CH₃] (−25.51 kcal/mol) > MnClTPP (−24.67 kcal/mol) > MnOAcTPP (−24.66 kcal/mol) > TPP (−22.04 kcal/mol). These results indicate a clear relationship between molecular structure and binding energy; the more similar the structures, the closer the binding energies. The binding energy for a Mn centre atom was larger than that for a Co centre atom, whereas the binding energies for axially coordinated Cl⁻ and axially coordinated OAc⁻ were relatively close. The binding energy of the peripheral substituent p-PhSO₂ was greater than that of the peripheral substituent benzyl group. The average binding energies of porphyrins with different structures and key VOCs had the following order: central atom > peripheral substituent > axially coordinated > original porphyrin. The energy of TPP binding to VOCs was significantly lower than that of other porphyrins binding to VOCs, perhaps due to the effect of the central atom and peripheral substituents.

Fig. 5.

Gaussian calculations of binding energies, bond lengths and dipole moments for the reaction of CSA with VOCs. A: Binding energy of porphyrins reacting with VOCs. B: Bond lengths of porphyrins reacting with VOCs. C: Dipole moments of different porphyrin structures.

3.4.3. Bond lengths

A bond length indicates the tightness of intermolecular bonding; a shorter bond length corresponds to a tighter bond (Kang et al., 2023). The length of the bond between a dye molecule and a VOC affects the response strength of the dye. The results of the bond length calculations are displayed in Fig. 5B. The lengths of the bonds between the central metal porphyrin and the VOCs were found to be shorter than the other bond lengths. Porphyrins with axial coordination bonds and peripheral substituents formed longer bonds with VOCs. Among them, MnTPP (2.37 Å), CoTPP (2.75 Å) had shorter bonds.

3.4.4. Dipole moments and charges

The dipole moment is a quantitative measure of molecular polarity; molecules with higher dipole moments are generally more polar, and more-polar molecules tend to exhibit higher reactivity (Dong, 2011). The dipole moment of MnClTPP with a peripheral substituent of sulphonatophenyl was discovered to be 8.9 debye (Fig. 5C); this value was the highest achieved in the investigated CSA and could lead to strong polarisation and high reactivity with VOCs. The dipole moments of porphyrins bound to VOCs are represented in Fig. 4C, with the largest dipole moments found for MnTPP and MnClTPP [p-PhSO₂] (Fig. 4A and B).

The basis of all chemical reactions is the exchange of electrons between reactants. When porphyrins bind to VOCs (Wu et al., 2024), their electron distribution changes, and this change can be quantitatively represented by the change in the charge distribution. In general, the variation in charge for the different VOCs were approximately the same. MnTPP and MnClTPP [p-PhSO₂] had high charge, whereas MnOAcTPP and MnClTPP [P-CH₃] had low charge. The large differences in charge between MnTPP and MnClTPP [p-PhSO₂] before and after their response reaction with VOCs are shown in Table S4.

3.5. Quantitative prediction of key aroma substances in black tea fermentation

Two chemometric methods were used for quantitative prediction. In LS-SVR, the benzaldehyde prediction model had the highest accuracy when applied to the prediction set (accuracy = 0.85) but fell short of the requirement for accurate prediction. The porphyrin materials that were significantly associated with the aroma of black tea at different stages of fermentation (p < 0.05) were identified through correlation analysis; the selection of porphyrin materials is detailed in Table 1. In all modelling results, Rp > 0.80 and RPD > 1.40. Notably, the Rp of (E)-linalool oxide (furanoid) increased by 0.10, benzaldehyde by 0.07, geraniol by 0.12, decanal by 0.15, (E)-2-hexenal by 0.02, and methyl salicylate by 0.06. Among these, (E)-linalool oxide (furanoid) yielded the most favorable modelling results, with an Rp of 0.94 and an RPD of 2.92.

Table 1.

Results of correlation modelling analysis.

VOCs	Methods	Porphyrins	Parameters	Calibration set		Prediction set		RPD
				RMSEC	Rc	RMSEP	Rp
(E)-Linalool oxide (furanoid)	LS-SVR	8/8	Sig = 850.7	11.0900	0.9609	2.028	0.8419	1.8280
			Gam = 24.65
		7/8	Sig = 340.7	0.3818	0.9946	1.2697	0.9412	2.9200
			Gam = 28.93
	SVR	8/8	c = 36.75	0.0001	0.9999	3.9232	0.7369	0.9431
			g = 0.2500
Benzaldehyde	LS-SVR	8/8	Sig = 638.5	1.0230	0.9606	2.0410	0.8398	1.8170
			Gam = 11.21
		3/8	Sig = 170.2	0.6160	0.9859	1.5433	0.9118	2.4020
			Gam = 12.50
	SVR	8/8	c = 36.75	0.0690	0.9948	4.3373	0.7057	0.8530
			g = 0.009
Decanal	LS-SVR	8/8	Sig = 204.5	0.3985	0.9941	2.4410	0.7605	1.5190
			Gam = 18.49
		6/8	Sig = 213.8	0.7379	0.9797	1.8130	0.8760	2.0450
			Gam = 14.08
	SVR	8/8	c = 12.12	1.3888	0.9066	5.0920	0.6402	0.7266
			g = 0.001
(E)-2-Hexenal	LS-SVR	8/8	Sig = 504.2	1.3130	0.9342	1.9710	0.8514	1.8810
			Gam = 3.808
		5/8	Sig = 164.0	0.3775	0.9947	1.8730	0.8670	1.9800
			Gam = 23.99
	SVR	8/8	c = 4.0000	3.4991	0.7646	4.5005	0.6922	0.8221
			g = 0.0009
Methyl salicylate	LS-SVR	8/8	Sig = 344.2	0.8874	0.8978	2.3861	0.7727	1.5540
			Gam = 14.90
		7/8	Sig = 628.4	1.2810	0.9375	2.1520	0.8199	1.7230
			Gam = 17.65
	SVR	8/8	c = 21.11	2.0797	0.8643	6.6362	0.5356	0.5575
			g = 0.0009
Geraniol	LS-SVR	8/8	Sig = 344.6	0.3289	0.9812	0.9373	0.8380	3.9560
			Gam = 11.70
		7/8	Sig = 428.1	0.5856	0.9870	1.7560	0.8841	2.1120
			Gam = 19.81
	SVR	8/8	c = 21.13	0.1830	0.9871	2.9547	0.8123	1.2522
			g = 0.005

Note: Porphyrin materials with p > 0.05 were screened for modelling by correlation analysis.

Benzaldehyde: TPP, CoTPP, Mof-MnClTPP [p-PhSO₂].

(E)-Linalool oxide (furanoid): TPP, MnTPP, MnClTPP, MnClTPP [p-PhSO₂], CoTPP, Mof-MnClTPP [p-PhSO₂].

Decanal: TPP, MnTPP, MnClTPP, MnClTPP [p-PhSO2], CoTPP, Mof-MnClTPP [p-PhSO₂].

(E)-2-Hexenal: TPP, MnTPP, MnClTPP, MnClTPP [p-PhSO₂], Mof-MnClTPP [p-PhSO₂].

Methyl salicylate: TPP, MnTPP, MnClTPP, MnClTPP [p-PhSO₂], CoTPP, Mof-MnTPP, Mof-MnClTPP [p-PhSO₂].

Geraniol: TPP, MnTPP, MnClTPP, MnClTPP [p-PhSO₂], CoTPP, Mof-MnTPP, Mof-MnClTPP [p-PhSO₂].

3.6. Discussion

DFT calculation is a computational method to deal with multi-electronic systems at the microscopic level, through which inter-atomic interactions can be clarified.DFT calculations can be used to analyse the binding mechanism of porphyrins with different structural modifications to aroma substances at the molecular and electronic scales. Compared with the traditional empirical and semiempirical methods, DFT has higher accuracy and universality, and it can provide key theoretical guidance for the screening of porphyrins that are more sensitive to aroma substances.

Modifications to the central atom (Mn²⁺ and Co²⁺) and peripheral substituents (p-CH₃,p-PhSO₂) were found to alter the overall morphology of the molecule and had a greater influence on responsiveness than did the introduction of axial ligands (Cl⁻,OAc⁻). The ability of MnTPP, a fundamental porphyrin structure, to bind with VOCs is predominantly constrained by the porphyrin's electronic characteristics and the spatial arrangement of the metal centre (Fang et al., 2017). In this study, the colorimetric sensing performance of MnTPP was found to be superior to that of CoTPP. The strong reactivity, low ionisation energy, and high reducibility of manganese atoms mean that MnTPP is more highly sensitive to VOCs. DFT calculations revealed that incorporation of axial ligands decreases the binding energy, resulting in increased bond lengths and reduced colorimetric sensing capability between porphyrin and VOCs. This phenomenon likely stems from spatial hindrance caused by the axial ligand at the manganese sites, where this hindrance reduces the porphyrin's affinity for and strength of binding with VOCs.

In this study, MnClTPP[p-PhSO₂] and MnClTPP[p-CH₃] in the presence of VOCs exhibited significant differences in response. The results of the DFT calculations indicated that the variations in binding energy and bond length were minimal and that the main differences were in the dipole moments. This suggests that the nature of the peripheral substituents has a strong influence on the colorimetric sensing performance of porphyrins, especially with regard to the dipole moment and charge of porphyrins. Phenylsulfonyl groups exert strong electron-withdrawing effects on the porphyrin ring, considerably reducing the π-electron density (Armentrout, 2018). This modification enhances the oxidation of the Mn centre through conjugated mechanisms, resulting in high colorimetric sensing performance of MnClTPP [p-PhSO₂]. By contrast, phenylmethyl groups, which have weak electron-withdrawing capability and low polarity, exert a negligible influence on the overall properties of porphyrins.

Nondestructive methods are often used to monitor the degree of black tea fermentation during tea processing. Li achieved rapid monitoring of black tea's fermentation quality and a 100 % recognition rate by using a solution-phase sensor array combined with ultraviolet–visible spectroscopy (Li et al., 2022). However, this method is complex and cannot be used for testing aroma. Hyperspectral imaging and CSA were used by An to assess the fermentation aroma of black tea, and an aroma discrimination rate of 94.29 % was achieved (An et al., 2022). However, the mechanism through which CSA dyes bound to VOCs was not revealed. DFT calculations have been applied to VOCs analysis. This covers the molecular mechanisms of VOCs interactions with polymers in red wine (Koch et al., 2023), peptide-VOC interactions in baijiu (Jia et al., 2022) and VOC-sensing mechanisms in rice (Zhang et al., 2019). Additionally, Anju used DFT to elucidate the interaction mechanism between porphyrin-based materials and VOCs (Anju et al., 2024). Kang investigated the binding of substituents, axial ligands, and simple volatiles in CSA sensors (Kang et al., 2021), but their CSA was not applied to the study of VOCs in tea and the associated binding mechanisms were not fully described.

In the present study, three porphyrins with different structures were employed to achieve rapid quantitative discrimination of key VOCs generated during black tea fermentation. In addition, binding energies, bond lengths, dipole moments, charges, and electron clouds were used to determine that the central atom and peripheral substituents of a porphyrin strongly influence a CSA's response properties. DFT calculations provide a critical theoretical basis for the design of CSA. The results of this research can be used to predict and screen porphyrin structures that exhibit high selectivity and sensitivity towards specific VOCs. Future research will explore the applicability of this method in various fields. Examples include designing targeted CSA for quality monitoring during tea processing, providing theoretical guidance for the design of compact, portable colorimetric devices and selecting cost-effective yet efficient porphyrin materials to minimise detection costs.

Although we have used DFT to study the binding mechanism between porphyrin and VOCs, this study mainly focuses on the unique aroma of black tea fermentation, which has a relatively simple aroma type. We plan to further investigate the impact of different porphyrin structures on responsiveness and apply the research findings to a broader range of VOCs. By leveraging the unique structural characteristics of various porphyrins, we aim to design simplified, low-cost, and practical colorimetric sensor arrays. This on-demand sensor unit design strategy can rapidly adapt to numerous fields requiring complex VOCs fingerprint analysis, such as food, agriculture, and environmental monitoring, demonstrating immense market potential and socio-economic value.

4. Conclusion

In this study, a CSA was constructed by exploring different-structure porphyrins and combining them with MOF material. Six key VOCs—benzaldehyde, (E)-2-hexenal, decanal, (E)-linaoxide (furanoid), methyl salicylate, and geraniol—were identified through SPME-GC/MS. Subsequently, the most responsive dyes were screened through correlation analysis and used for quantitative discrimination of the key VOCs generated during black tea fermentation. The final results indicated that the Rp values for the prediction set were all >0.80. Porphyrin structure was found to have a strong influence on binding energy, dipole moment, bond length and charge. In particular, the difference between the central atom and peripheral substituents of a porphyrin has a strong effect on response ability. This paper provides a theoretical basis for the design of CSAs.

CRediT authorship contribution statement

Yifan Zuo: Writing – review & editing, Writing – original draft, Formal analysis. Shuai Dong: Data curation, Conceptualization. Jingfei Shen: Validation, Supervision. Yurong Chen: Resources, Project administration. Qianfeng Yang: Visualization, Validation. Yongning Wei: Methodology, Investigation. Jingming Ning: Funding acquisition, Formal analysis. Luqing Li: Project administration, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material: CSA and VOCs Gaussian calculation research

Data availability

Data will be made available on request.