Introduction

Over the years, the demand for high-quality food has increased substantially [1]. Consumers require products with minimal risk for disposal to negative health effects. There is a global need to improve food safety and quality [2] since in many countries economic profit is emphasized ahead of nutritious products that meet strict quality and safety standards [3].

The food industry is of utmost significance as it is one of the most important economic axes in the world [2]. To meet the challenges, scientists and specialists of the food sector are designing, develo**, and improving technologies to enable trustworthy, fast, and economically feasible analysis of the quality of food. Most important traditional detection techniques used in food analysis have included high performance liquid chromatography (HPLC), mass spectroscopy (MS), lateral flow strip (LFS), Southern Blot (SB) and enzyme-linked immunosorbent assay (ELISA) [4, 5]. Although these techniques often give accurate results, they are complex, costly, require chemical reagents, are time-consuming, and, above all, destructive (i.e., they alter the physical, chemical and/or nutritional characteristics of the food), which makes accurate measurement of labile food components like vitamin C difficult. Due to continuous technological advancement especially in the area of computational and data processing, other techniques have been developed that do not require prior sample preparation and are more efficient and non-destructive [6, 7]. These include computer vision (CV) and vibrational spectroscopies, encompassing infrared (IR), near (NIR), mid (MIR) and far infrared (FIR) spectroscopy, Raman spectroscopy, ultraviolet–visible (UV–Vis) spectroscopy, and hyperspectral imaging (HSI) [90] obtained a validation coefficient of 0.79–0.91 in predicting the vitamin C content of fresh tomatoes. Beghi et al. [91] used a portable vis–NIR spectrometer to monitor the vitamin C content of Golden Delicious and Stark Red Delicious apples. In this case, the validation had a low R2 value of 0.50. Similarly, the correlation coefficient between the evaluation with a portable NIR spectrometer and the reference values for vitamin C content in Kakadu plums and plum puree was 0.55 and 0.86, respectively [92]. Even smartphone-based spectrometers have been developed as demonstrated in the studies of Kong et al. [93] and Aguirre et al. [94], whose equipment was successfully employed to assess the vitamin C content in a lemon beverage and orange juice, respectively.

The development of this miniaturized equipment is a great achievement in the field of spectroscopy, but there are some limitations that influence the results obtained. The size of the portable equipment limits the evaluation of high volume samples. When performing the evaluation in the field (in situ), there is a high risk of contamination and interference from environmental factors [74]. They also have somewhat higher detection limit and lower sensitivity compared with conventional equipment. In a study predicting the quality of mushrooms of the genus Tuber, the accuracy of the benchtop NIR spectrometer was from 93.33–100%, and the values of three portable NIR spectrometers were 83.33–100%. Similar results were shown when evaluating bioactive compounds from green tea [79].

Another disadvantage is that the construction of calibration models for each portable equipment requires considerable cost and time [95]. One option is for models to be built on benchtop spectrometers (master or primary unit) and transferred to portable spectrometers (slave or secondary unit) [96] to avoid the whole calibration process. It has been reported that large spectral databases of various samples have been transferred from benchtop to portable spectrometers [74]. The transfer can also be made between equipment of different makes/models [97, 98]. There are several chemometric or standardization techniques to achieve this and maintain model accuracy [99]. For example, Igne et al. [100] evaluated the quality of soybean with four benchtop NIR spectrometers of two different brands by employing eight calibration transfer techniques based on the removal of orthogonal signal. Zeng et al. [96] evaluated pigments in tea leaves with a laboratory Raman spectrometer and the calibration models were transferred to a portable Raman spectrometer with the direct standardization (DS) technique. For the evaluation of fresh cow milk, Melenteva et al. [101] transferred calibration models from a diode-array spectrometer to a shortwave visible-NIR spectrometer with model transfer by slope and bias correction (SBC). To transfer the calibration model between three NIR spectrometers of different brands to predict apple TSS, Hayes et al. [95] employed orthogonal projection (TOP), piecewise direct standardization (PDS), difference spectrum adjustment (DAS), and model updating (MU). Salguero-Chaparro et al. [102] used PDS, TOP, and SBC to transfer calibration models on olive quality assessment from a benchtop NIR monochromator to a portable NIR spectrometer. The disadvantage of the transfer is that, after a certain time, the model obtained loses accuracy. To avoid this, the models should be recalibrated gradually [102].

UV spectroscopy has degradative effects on bioactive compounds such as vitamin C. The use of IR spectroscopy is not recommended for foods with water content higher than 80% [103] since water is a strong absorber whose hydrogen bonds interfere significantly in the IR region [104]. This is a serious limitation for many foods of plant origin that are significant sources of vitamin C, but also have a high proportion of water in their composition. This problem can be solved with a suitable preparation of the sample, but this demands time and is costly. Raman spectroscopy is a potential option in these cases because the weak Raman scattering has less interference with water, and it can be used in solid, semi-solid, and liquid samples quickly and economically; in addition, by omitting the sample preparation, the analysis would be non-destructive [8]. This has been demonstrated in several studies such as in the evaluation of milk; El-Abassy et al. [105] were able to successfully (R2: 0.92–0.99) determine milk fat content using Raman spectroscopy.

Despite the potential of Raman scattering, it is weak, resulting in low analytical sensitivity [106] compared to other spectroscopy techniques such as NIR and MIR spectroscopy. To counteract this, enhancement of Raman spectroscopy can be utilized in form of SERS. Another disadvantage of the Raman spectrum is that it can only collect a small amount of information from the sample and, therefore, the results obtained are not complete [106]; the same is also true for IR spectroscopy [107]. The low sensitivity of Raman spectra was demonstrated by He et al. [108], who were unable to discriminate the Raman spectra of three prohibited food additives dissolved in ethanol. On the other hand, by employing gold-plated silicon as SERS substrate, the weak signal was improved, enabling successful distinguishment of the samples.

Considering the above, before performing any analysis in food, it is necessary to define which vibrational spectroscopy technique is going to be used, taking into account its advantages and disadvantages in relation to the nature and composition of the sample. For example, many compounds cannot be detected based on IR spectrum; specifically, minerals are not active in the NIR region, but they are active in the Raman region [107]. It should be noted that, even if the most optimal vibrational spectroscopy technique is chosen, it is challenging to analyze multiple compounds simultaneously [24]. This can, however, be solved with correct data processing of the spectrum using chemometrics. Yang and Irudayaraj [57] concluded that infrared spectroscopy (NIR and IR-photoacoustic) is better than Raman spectroscopy for the prediction of vitamin C content in some foods and pharmaceuticals. The authors also used FT-Raman spectroscopy and highlighted the importance of such a chemometric method for improving the sensitivity of the technique and reducing data acquisition time. In the same study, it was established that the O–H groups have a weak absorbance in the Raman spectrum, but a strong absorbance in the infrared region, with the opposite occurring with the non-polar C–C groups.

Finally, to enhance the robustness of the models generated with respect to vitamin C assessment, it is necessary to have a comprehensive database. This can be achieved by analyzing different foods or, if the focus is on a specific food, samples of different species and origins should be used [7, 28]. Relatively more robust models were obtained when evaluating vitamin C in three bell pepper cultivars at different growth stages by using Vis–NIR spectroscopy (average R2: 0.74 and RPD: 2.2) compared to shortwave NIR spectroscopy (average R2: 0.72 and RPD: 2.13) [109]. In addition, excellent results were obtained in the evaluation of vitamin C in two green-fleshed kiwifruit cultivars employing FT–NIR spectroscopy [110]; in eleven yellow-fleshed peach cultivars, in three white-fleshed peach cultivars, in four yellow-fleshed nectarine cultivars and in one white-fleshed nectarine cultivar using UV spectroscopy [111]; in four plum cultivars, using NIR, MIR and Raman spectroscopy [112]; and in apricot pastes of eight cultivars using FT-IR spectroscopy with attenuated total reflectance [113].

Vibrational spectroscopy techniques and chemometric methods will continue to evolve in parallel with other technological advances to address the above limitations. On the other hand, researchers should be informed and trained on the importance of using these techniques not only in the evaluation of vitamin C, but also in the evaluation of other compounds in different food matrices. If the use of spectroscopy expands further, this will increase the volume of the production of equipment and, consequently, lead to a significant cost reduction.

Conclusions

Vibrational spectroscopy and its recent advancements are attracting the attention of researchers around the world due to its possibilities in food quality assessment, the benefits including rapid, accurate and inexpensive non-destructive analysis. Specifically, the usage of UV–Vis, IR and Rahman spectroscopies has increased in the quantification of vitamin C as a quality indicator of various food products; the techniques are used together with chemometric preprocessing, discrimination, and modeling tools to, in addition to improving the efficiency of obtaining spectral data, help researchers in the subsequent data analysis and interpretation of the results. To avoid limitations of vibrational spectroscopy techniques, it is necessary to define the appropriate technique according to its advantages and disadvantages with respect to its compatibility with the nature of the food sample.