Optimization of ultrasound-assisted extraction using a choline chloride-lactic acid green solvent for gingerbread sample preparation before elemental analysis by inductively coupled plasma mass spectrometry Authors: Agnieszka Kowaluk, Dobrochna Rabiej-Kozioł, Leszek Pazderski, Aleksandra Szydłowska-Czerniak Published: 2026 ABSTRACT AbstractNatural deep eutectic solvents (NADESs), composed of choline chloride (ChCl) and lactic acid (LacA), have been developed as sustainable and green extractants for the determination of chemical elements in gingerbread cookies by inductively coupled plasma mass spectrometry (ICP–MS). Three ChCl:LacA solvents with molar ratios of 1:2, 1:1, and 3:2 were prepared and characterized using Fourier‑transform infrared spectroscopy (FTIR), proton nuclear magnetic resonance spectroscopy (1H NMR), and thermal analysis (thermogravimetry, TGA, and derivative thermogravimetry, DTG). The response surface methodology (RSM) combined with a Box-Behnken design (BBD) was used to optimize green ultrasound-assisted extraction (UAE) conditions, including the ChCl to LacA molar ratio, water content (WC) in NADESs, and sonication time. Evidently, there were significant differences in the amounts of macronutrients and manganese between the samples prepared using the proposed NADES extraction and those obtained after microwave‑assisted acid digestion (MW‑AD) and quantified by ICP‑MS. Therefore, under the predicted optimum extraction parameters with ChCl:LacA (1:1), a WC of 34 %, and a sonication time of 40 min, the concentrations of three microelements (Zn, Cu, and Fe) were consistent with the predicted values from the response surface fitting equations. In addition, the decision tree learning model helped determine the optimal conditions for green extraction of these inorganic analytes from gingerbread and validated that RSM predictions are consistent and accurate. The results indicate that the UAE using ChCl:LacA is a promising analytical procedure for the simultaneous extraction of zinc, copper, and iron from gingerbread cookies.Keywords: Confectionery products, Green solvents, Ultrasonication, Spectroscopic and thermal methods, Elemental analysis, Multi-response optimization and decision tree Introduction Green solvents derived from sustainable crops or other eco‑friendly processes, due to their low toxicity, ease of recycling, and rapid biodegradation, represent a promising option for industrial applications, including cleaning, separation and extraction processes, as well as pharmaceutical, food, and cosmetic processing, and material fabrication. Among green solvents, which can help solvent-based separation processes and various industries become greener and leaner, deep eutectic solvents (DESs) play a key role, aligned with the principles of green chemistry [1]. Recently, natural deep eutectic solvents (NADESs), prepared by mixing appropriate molar ratios of hydrogen bond acceptors (HBAs) and hydrogen bond donors (HBDs) derived from natural components such as organic acids, organic bases, amino acids, sugars, amines, and choline derivatives, have gained significant attention for their environmentally friendly and efficient extraction capabilities, attributed to their low volatility, high solubility, and biodegradability. The co-occurrence map in Fig. 1 visually depicts the two keywords “deep eutectic solvents” and “extraction,” grouping them into distinct clusters that reflect the primary research directions in this area. The co-occurrence map of deep eutectic solvent and extraction keywords (an enlarged map fragment on the right side). The 1182 keywords, of which 203 met the threshold of appearing at least 5 times, were analyzed using VOSviewer (version 1.6.20). These keywords were clustered into 21 groups, each representing a unique aspect of research based on the DES application in separation processes. The term “deep eutectic solvent” centralizes the blue cluster (20 items), involving green extraction processes that include various matrices (biomaterials, nanomaterials, residues, biomass), and their characterization and purification. It can be noted that the other clusters are linked to terms, including components utilized in the DES preparation, sample preparation, extraction techniques such as sonication, bioactive compounds isolated by DESs, analytical methods applied for the determination of isolated antioxidants, and chemometric and statistical analyses (Box-Behnken design, central composition design). An enlarged map fragment shows that DESs have been used recently as effective extractants for most organic and inorganic analytes, including heavy metals (mercury) and other elements, from various matrices (Fig. 1). Since the mid‑2010 s, there has been a notable increase in publications on the use of DESs as green extractants for organic analytes, and a similar trend has been observed in studies focused on the extraction of inorganic elements from various food matrices. Some hydrophilic and hydrophobic DESs (choline chloride (ChCl):oxalic acid (OxA), lactic acid (LacA), mandelic acid (ManA), tartaric acid (TarA), glycerol (Gly), urea (U), ethylene glycol, riboflavin, cystein, phenol (Ph); xylitol (Xyl):citric acid (CitA), malic acid (MalA); glycolic acid (GlyA):menthol; menthol:formic acid (FA), lauric acid (LA); CitA:MalA, sucrose (Scr); trihexyltetradecylphosphonium chloride:pivalic acid) were applied for the extraction of macro- (Na, K, Ca, Mg, P), micronutrients (Cu, Fe, Zn, Ni, Mn, Sn, V, Se) and toxic elements (Hg, As, Cd, Pb, Ba, Co) from edible oils and fats, vegetables and fruits, mushrooms, fishes, plants, beverages, honey, eggs, dairy, meat and flour products [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22]. The selection of suitable techniques, including conventional mechanical extraction [4], [9], [7], [12], [17], [20] and unconventional extraction processes: microwave-assisted extraction [8], [14], ultrasound-assisted extraction and ultrasound liquid–liquid microextraction [2], [3], [11], [13], [14], [21], [22], dispersive liquid–liquid microextraction [15], [16], [19], and liquid phase microextraction [6], [10], [18] for the separation of inorganic elements from food samples using proposed hydrophilic and hydrophobic DESs, has depended on practicality, simplicity, and economic factors. However, the more satisfactory efficiency of these various extraction processes, in combination with the use of green solvents, can be achieved by optimizing extraction parameters, including HBA to HBD ratio, sample mass to solvent ratio, water content in DESs, pH, time, temperature, and others. Procedures for optimizing extraction factors using multivariate techniques appear to be among the most appropriate, as they are efficient, faster, more economical, and allow for exploration of the relationships between the dependent and independent variables involved in the extraction process. Response surface methodology (RSM) and central composite, Box–Behnken, or Doehlert designs are effective statistical tools that enable the optimization of extraction conditions by simultaneously evaluating the influence of multiple independent variables. RSM and Box–Behnken and/or central composite designs were applied to predict the optimal extraction conditions for chemical elements, including cadmium from milk and cheese [6] vanadium from vegetables, fruits, meats, fish, mushrooms, eggs, cheese, honey, black tea, and coffee [21] lead and cadmium from vegetables, dry tea, and mint [2] as well as iron, copper, and lead from vegetable oils [19]. Generally, after extraction and preconcentration procedures, atomic absorption spectrometry (AAS) is a widely used analytical method for determining elements in food samples due to its ease of use, low cost, and selectivity [3], [4], [6], [7], [9], [10], [11], [12], [16], [17], [18], [20], [21]. Only some researchers employed plasma techniques, such as inductively coupled plasma optical emission spectrometry (ICP-OES) [8], [13], [15], inductively coupled plasma mass spectrometry (ICP-MS) [13], [14], and microwave-induced plasma optical emission spectrometry (MIP-OES) [19], which offer multi-elemental analysis and better validation parameters (high sensitivity, low detection limits, wide dynamic ranges, good accuracy and precision), particularly in trace and ultra-trace analysis. Furthermore, square-wave anodic stripping voltammetry (SWASV) at a mercury-modified carbon working electrode was used to determine lead and cadmium in vegetable oils after a reversed-phase dispersive liquid–liquid microextraction procedure based on the decomposition of the DES (menthol:FA) [15]. The applicability and accuracy of the proposed analytical methods for the determination of different elements in food samples have often been evaluated using certified reference materials (CRMs) with declared amounts of these elements. Therefore, the keyword “certified reference materials” appears on the co-occurrence map (Fig. 1). However, there are limited certified reference materials for the determination of metals and metalloids in this type of sample. Although, to the best of our knowledge, the most effective DESs for the simultaneous extraction of several inorganic analytes from gingerbread have not yet been identified for the future development of a quantitative reference material containing fixed amounts of chemical elements. Only in our previous study, we proposed three NADESs (Xyl:MalA, ChCl:MalA, and ChCl:LacA) for the preparation of uniced and iced gingerbread samples prior to the determination of macroelements (K, Na, Mg, and Ca), microelements (Mn, Zn, Fe, and Cu), and the toxic Cd element using the ICP-MS technique [23]. These results confirmed that the extraction efficiency of chemical elements from gingerbread depended on the components (HBA and HBD) of NADESs, which likely form complexes with target metal ions through interactions between metal ions and monocarboxylic acids (LacA), as well as through complexation with polycarboxylic acids (MalA). Moreover, the solvation of metal ions in choline-based solutions enabled the extraction of inorganic analytes from solid samples. Interestingly, the highest extraction efficiency of microelements from gingerbread was achieved using a NADES containing ChCl, LacA, and water [23]. On the other hand, ChCl and LacA, as components of NADESs, exhibited low toxicity, high biodegradability, affordability, and wide availability on an industrial scale. For this reason, NADES solutions composed of ChCl and LacA represent a promising green solvent for the extraction of inorganic analytes from confectionery matrices. Additionally, in recent years, there has been an increasing tendency to employ machine learning (ML) methods to predict and optimize extraction conditions for various analytes from food samples, including antioxidants and phytochemicals from pomegranate peels [24], Moringa oleifera oil from kernels [25], cinnamon essential oil from dried cinnamon raw materials [26], and cellulose from hemp bast fiber [27]. Among these methods, decision tree learning is particularly attractive due to its transparency and interpretability. Decision trees are also well-suited for relatively small datasets, as they do not require large sample sizes for model training. Moreover, the literature indicates that no studies have yet evaluated the optimization of DES–based extraction of chemical elements from food matrices using both RSM and ML algorithms. Therefore, the aim of this study was to develop an optimized ultrasonic treatment procedure for gingerbread samples using ChCl:LacA-based mixtures as green solvents prior to multielement analysis by ICP-MS. A gingerbread sample was selected as a prospective candidate for a reference material with a well-defined elemental composition to assess the accuracy of analytical methods for the determination of inorganic analytes in confectionery products. The conditions of the ultrasound-assisted extraction (UAE) process, including the ChCl:LacA molar ratio, the water content (WC) in these NADESs, and the sonication time, were optimized using RSM based on a Box–Behnken design (BBD). The proposed ChCl:LacA-based ICP-MS method was validated and successfully applied for the effective extraction and simultaneous quantification of Zn, Cu, and Fe in gingerbread samples. Furthermore, a decision tree, a machine learning technique, was employed to accurately evaluate the optimization results based on the parameters identified through RSM and to reduce potential deviations within the system. In addition, more sophisticated techniques, such as Fourier transform infrared spectroscopy (FTIR), 1H nuclear magnetic resonance (1H NMR), and thermal analysis, were employed to validate the formation of the selected NADESs and to better understand their chemical interactions and thermal stability. Materials and methods Chemicals, reagents, standards, and samples All chemicals were of analytical grade or HPLC grade and purchased from MERCK Sp. z o.o. (Warszawa, Poland). Choline chloride (ChCl; 99 % purity) and lactic acid (LacA; 88 % purity) were used for the preparation of NADESs. Calibration solutions for ICP‑MS elemental analysis were prepared by diluting single‑element mass concentration standards obtained from the Central Office of Measures. Deionized distilled water (DDW) was sourced from an Arium® Water Purification System (Sartorius, Goettingen, Germany). Certified reference material, Polish Virginia Tobacco Leaves (INCT-PVTL-6), was used as a reference material for the ICP-MS method recovery study. The kindly provided gingerbread cookies, without filling or coating, from a local confectionery company, were ground to a powder using a laboratory mill (FW100 model, Chemland, Stargard, Poland) and sieved through a nylon sieve with a 500 μm mesh. Natural deep eutectic solvents (NADESs) Preparation Three NADESs, consisting of choline chloride (ChCl) as the HBA mixed with lactic acid (LacA) as the HBD at different molar ratios (1:2, 1:1, and 3:2), were prepared using the heating method described in our previous article [23]. Appropriate amounts of ChCl and LacA, which constituted each NADES, were transferred into three glass beakers and then heated at 80 °C on a magnetic stirrer–hot plate (400 rpm; 06–MSH–PRO–T, Chemland, Stargard, Poland) until clear solutions were obtained. After preparing three NADESs, 20, 30, and 40 wt% of water were added to reduce the overall viscosity and enhance element extraction efficiency. Characterization by Fourier transformed infrared (FTIR) spectroscopy The infrared spectra of the prepared NADESs containing ChCl and LacA in different molar ratios, as well as the initial pure compounds, were obtained by attenuated total reflectance − Fourier transform infrared spectroscopy (ATR-FTIR) on a Bruker VERTEX 70v FTIR spectrometer (Bruker Optics, Ettlingen, Germany) equipped with a diamond ATR mode, with a spectral resolution of 4 cm−1 and an average accumulation of 64 scans. The samples were directly placed on the ATR crystal surface of the ATR probe and analyzed from 4000 to 400 cm−1. Characterization by nuclear magnetic resonance (NMR) spectroscopy The 1H NMR measurements were performed by a Bruker Avance III 400 MHz NMR spectrometer (Bruker Corporation, Billerica, MA, USA), at 295–300 K. The 1H NMR spectra of the prepared NADESs containing ChCl and LacA in different molar ratios were recorded by two alternative ways: (a) in DMSO‑d6, CD3CN or CD3OD; (b) for neat mixtures placed in coaxial capillary inserts into the 5 mm NMR tubes filled with neat CDCl3 as an external field-frequency lock and a chemical shift reference. In contrast to FTIR measurements, no water was added to the NADESs. 1H NMR spectra of pure ChCl or LacA were simply recorded in DMSO‑d6. The 1H chemical shifts were referenced to TMS, with residual 1H signals of the NMR solvents (DMSO‑d6: 2.50 ppm; CD3CN: 1.94 ppm; CD3OD: 3.31 ppm) or neat CDCl3 (7.26 ppm) as primary references [28], [29]. Characterization by thermal analysis The thermal decomposition of the prepared NADESs and the initial reagents was studied using thermogravimetric analysis (TGA) on a Discovery SDT 650 simultaneous DSC–TGA instrument (TA Instruments, New Castle, Delaware, USA). TGA measurements were performed under a nitrogen flow (20 mL/min) over the temperature range of 25–550 °C, with a heating rate of 10 °C/min. Optimization of ultrasound-assisted extraction of elements from gingerbread cookies Single-factor analysis Based on literature data [30], [31], [32] and our previous results reported in [23], ChCl:LacA was selected from three prepared NADESs, such as Xyl:MalA, ChCl:MalA, and ChCl:LacA. A single‑factor experimental design was employed to investigate the effects of individual variables, including the ChCl:LacA molar ratio, water content (WC) in the eutectic mixture, liquid-to-solid ratio, sonication time, and extraction temperature on the concentrations of macroelements (Na, K, Ca, and Mg) and microelements (Mn, Zn, Cu, and Fe) in gingerbread extracts, while keeping the remaining variables constant (supplementary material, Table S1). Design of response surface methodology From the single-factor experiment results, three independent variables were chosen: ChCl:LacA molar ratio, water content (WC) added to ChCl:LacA, and extraction time in the ranges 0.5 − 1.5 mol/mol, 20 − 40 %, and 20 − 40 min, respectively. The RSM approach based on a three‑level, three‑factor BBD was applied to optimize and statistically evaluate the effects of these three selected factors on the extraction efficiency of chemical elements from gingerbread. The minimum, intermediate, and maximum levels (coded − 1, 0, +1, successively) for each factor were: ChCl:LacA (1:2, 1:1, and 3:2), WC (20, 30, and 40 %), and extraction time (20, 30, and 40 min). The UAE of elements from pulverised gingerbread was performed according to the procedure described in our previous article [23] with some modifications. Briefly, 5 g of the sample was weighed, and 30 mL of the ChCl:LacA solvent at the respective molar ratio was added. It was then subjected to an ultrasonic bath (Labo Plus, Warszawa, Poland) with an ultrasound power of 540 W and an ultrasound frequency of 37 kHz for 20, 30 and 40 min at 50 °C. A total of 15 experiments were performed, corresponding to 12 factorial points and 3 replicates of the central point. In RSM, the BBD was used to predict the value of the dependent variables (concentrations of chemical elements) studied according to equation (1): where Yn is one of the eight predicted responses: cNa, cK, cCa, cMg, cMn, cZn, cCu, and cFe, ChCl:LacA, WC, and t represent the independent variables; β0 is the constant; β1, β2, and β3 are the linear term coefficients; β11, β22, and β33 are the quadratic term coefficients; and β12, β13, β23, β112, and β113 are the cross-term coefficients. The quality of the developed partial cubic model (PCM) was evaluated by calculating the coefficient of determination (R2), while analysis of variance (ANOVA) was used to assess the statistical significance of the proposed model based on the regression and the mean square of the residual error. The graphical representation of equation (1), referred to as the response surface, was applied to present the interactions between the independent variables (extraction conditions) and their effects on each response (studied element content). Optimal conditions for NADES-based extraction of three microelements effectively separated from the gingerbread sample were obtained by solving the regression equations. Moreover, the microwave-assisted acid digestion (MW-AD) method served as the reference in this study, carried out in a closed Anton Paar Multiwave Pro microwave system in an 8NXF100 rotor (Anton Paar GmbH, Graz, Austria) according to the procedure described previously [23]. Machine learning method − decision tree modelling A decision tree machine learning algorithm was chosen to test the accuracy of the RSM results. The incorporation of decision trees into optimization research enhances the conventional RSM technique, providing a data-driven approach that can yield more accurate and dependable results. A decision tree is a hierarchical, tree-like structure in which each internal node represents a decision based on an attribute, each branch denotes the outcome of that decision, and each leaf node signifies a final decision or a predicted value. A decision tree model was developed using variables including the ChCl:LacA molar ratio, WC in NADES, and sonication time. Then, decision tree analysis was conducted, and its performance in determining and optimising element concentrations under UAE-NADES conditions was evaluated. The model was used for data classification and analysis. The minimal gain, minimal leaf size, and minimal size for split were 0.01, 1, and 4, respectively. However, a maximal tree depth of 8 was used in the study to obtain a decision tree with the best accuracy. The optimal parameter combination is determined by trying all possibilities. Accuracy was calculated as a percentage of correct predictions. A pruning technique was applied to prevent overfitting, and the number of pre-pruning alternatives was selected as 3. Analysis of chemical elements by inductively coupled plasma mass spectroscopy (ICP-MS) method An inductively coupled plasma mass spectrometer (ICP-MS Agilent 7800, Agilent Technologies, Inc., Tokyo, Japan) equipped with a crossflow nebulizer (gas flow = 1.01 L/min) for sample introduction into the plasma, spray chamber (temperature = 2 °C), operating at RF power (1550 W) in plasma generation (15.0 L/min), and auxiliary gas (0.90 L/min) was used for Na, K, Ca, Mg, Mn, Zn, Cu, and Fe determinations. The instrumental details and operational conditions are detailed in supplementary material (Table S2). Data analysis All experiments were conducted in triplicate and presented as: mean value of three replicates ± standard deviation (SD). The percentage recovery of each element in the prepared green extract and in wet‑digested Polish Virginia Tobacco Leaves (INCT‑PVTL‑6) was calculated and compared with the certified element concentrations of this reference material. Differences between the measured and certified values were assessed using Student’s t-test at a 95 % confidence level. The Duncan test was performed to analyze the significant differences (p < 0.05) between the obtained results of element concentrations. The statistically significant effects of the independent variables (UAE conditions) and their interactions were evaluated using analysis of variance (ANOVA) at a 95 % confidence level (p < 0.05). All data were statistically analyzed and optimized using RSM in Statistica 8.0 software (StatSoft Inc., Tulsa, OK, USA). However, the decision tree models were created with RapidMiner AI Studio, version 2026.1.1 (Altair Engineering Inc., Troy, MI, USA). Results and discussion Characterization of hydrophilic choline chloride:lactic acid solvents FTIR analysis ATR-FTIR spectra were used to identify the structures of the pure components (ChCl and LacA), to confirm the formation of the NADESs, and to analyze the differences among the proposed precursor molar ratios and the amounts of added water. Characteristic bands corresponding to the ChCl and LacA functional groups were illustrated in Fig. 2. ATR‑FTIR spectra of the pure components (choline chloride − ChCl and lactic acid − LacA) and the prepared natural deep eutectic solvents (NADESs) composed of ChCl and LacA at different molar ratios of: a) 1:2, b) 1:1, and c) 3:2 with water contents ranging from 20 to 40 %. As described in our previous study [23], ChCl presented a hydroxyl band at 3219 cm−1, along with bands in the 1200–880 cm−1 region associated with vibrational bands of the quaternary ammonium group, including N–H stretching and symmetric C–N+ stretching vibrations. The FTIR spectrum of LacA displayed a broad band in the 3600–3100 cm−1 region and a band at 2986 cm−1, which were assigned to − OH and − CH stretching vibrations, respectively. In addition, a strong absorption band at 1719 cm−1 corresponded to C=O stretching vibrations, while the bands at 1453 cm−1 and 1375 cm−1 were attributed to − CH bending vibrations. For all ChCl–based NADESs, the bands observed at 1477 cm−1 and 1418 cm−1 were ascribed to CH2 bending vibrations, while the vibrational bands at 1349 cm−1 and 1142 cm−1 disappeared. These spectra were consistent with previously published studies by other authors [33], [34]. Following NADES formation, the − OH stretching vibrations of LacA (3600–3100 cm−1) shifted toward lower wavenumbers, appearing as a broad band in the range of 3361–3336 cm−1, depending on the NADES composition. In addition, the − CH vibrational band at 2986 cm−1 observed in the LacA spectrum shifted to 3034 cm−1 in the NADES spectrum. The C=O stretching band also shifted from 1719 cm−1 in the LacA spectrum to 1727–1721 cm−1, depending on the obtained NADES. Similar observations for the formation of ChCl:LacA NADESs were reported by Karimarji et al. [34]. Further confirmation of NADES formation was provided by the appearance of an absorption band at 1643–1638 cm−1, which was attributed to the angular deformation of − OH groups from water molecules (Fig. 2). Moreover, in agreement with the study by Malebrán et al. [35], the recorded spectra showed a decrease in the sharp signal associated with carbonyl stretching vibrations at 1727–1721 cm−1 with increasing water content from 20 to 40 %. Meanwhile, characteristic absorptions of the C–N single bond in ChCl were observed at 3026 cm−1 and 1481 cm−1, with the intensity of the latter also decreasing as water content increased. Additionally, regardless of the amount of added water, the sharp signal corresponding to the carbonyl stretching vibration at 1727–1721 cm−1 exhibited the highest intensity for the ChCl to LacA ratio of 1:2. The same trend was found for the bands at 1131–1127 cm−1, assigned to C–O stretching vibrations, and at ∼ 1270 cm−1, attributed to C–O stretching vibrations of LacA. Thus, the ATR-FTIR spectra confirmed the formation of hydrogen bonds in the NADESs and demonstrated differences arising from variations in reagent ratios and the amount of water added. 1H NMR analysis The 1H NMR chemical shifts of ChCl, LacA and the NADESs listed in Table 1 (the assignments were based on signals integration and multiplicity), while the 1H NMR spectra are reproduced in the supplementary material (Figs. S1-S7). Table 11H NMR chemical shifts of choline chloride, lactic acid and the studied NADESs, together with signals multiplicity (s – singlet, d – doublet, t – triplet, q – quartet, m – multiplet).SolventCholine chloride (cholinium cation)Lactic acid3xCH3N-CH2O-CH2OHCH3CHOHCOOHChCl(DMSO‑d6)3.13 s3.42 m3.79 m5.63 tLacA(DMSO‑d6)1.26 d *1.27 d **1.37 d **4.02 q *4.17 q **4.91 q **5.41 dca. 12.4 saChCl:LacA (1:1)(DMSO‑d6)3.10 s3.39 m3.81 mca. 5.4 a,b1.21 d *1.26 d **1.38 d **4.02 q *4.17 q **4.91 q **ca. 5.4 a,bca. 12.4 saChCl:LacA (1:1)(CD3CN)3.15 s3.44 m3.95 mca. 2.7 a,b,c1.33 d *1.36 d **1.46 d **4.21 q *4.27 q **5.05 q **ca. 2.7 a,b,cn.d.ChCl:LacA (1:1)(CD3OD)3.24 s3.53 m4.01 mca. 4.9 a,b,d1.38 d *1.42 d **1.50 d **4.23 q *4.33 q **5.06 q **ca. 4.9 a,b,dn.d.ChCl:LacA (1:1)neat e3.25 s3.57f3.96fca. 6.2 a,b1.30 d *1.43 d **1.55 d **4.27 q *4.35 q **5.00 q **ca. 6.2 a,bn.d.ChCl:LacA (1:2)neat e3.25 s3.58f4.01fca. 6.4 a,b1.35 d *1.38 d **1.46 d **4.32 q *4.39 q **5.02 q **ca. 6.4 a,bn.d.* major form, ** minor form(s), a Broad signal, b OH groups of choline and lactic acid yield one shared, broad signal with unobservable multiplicity, c Additional sharing with residual water in acetonitrile‑d3 solvent (2.13 ppm for pure CD3CN), d Additional sharing with residual water in methanol‑d4 solvent (4.87 ppm for pure CD3OD), e 1H NMR spectrum of the neat NADES, recorded in a coaxial capillary insert into the 5 mm NMR tube filled with CDCl3 as an external field-frequency lock and a chemical shift reference,f Broad signal with unobservable multiplicity. 1H NMR chemical shifts of choline chloride, lactic acid and the studied NADESs, together with signals multiplicity (s – singlet, d – doublet, t – triplet, q – quartet, m – multiplet). * major form, ** minor form(s), a Broad signal, b OH groups of choline and lactic acid yield one shared, broad signal with unobservable multiplicity, c Additional sharing with residual water in acetonitrile‑d3 solvent (2.13 ppm for pure CD3CN), d Additional sharing with residual water in methanol‑d4 solvent (4.87 ppm for pure CD3OD), e 1H NMR spectrum of the neat NADES, recorded in a coaxial capillary insert into the 5 mm NMR tube filled with CDCl3 as an external field-frequency lock and a chemical shift reference,f Broad signal with unobservable multiplicity. The 1H NMR signals of C-bonded protons in the cholinium cation and the LacA molecule, detected for the ChCl:LacA (1:1) solutions in various NMR solvents, are similar to those for pure components – which is best exemplified upon the same DMSO‑d6 solvent. However, there is a noticeable difference concerning the hydroxyl groups, present in both species. In all NMR solvents the two OH resonances of the ChCl:LacA (1:1) system overlap into one broad signal (at ca. 5.4 ppm for DMSO‑d6, at ca. 2.7 ppm for CD3CN, at ca. 4.9 ppm for CD3OD) with unobservable multiplicity – although for pure ChCl and pure LacA (each in DMSO‑d6) these were well-resolved multiplets (triplet and doublet, respectively). So large changes of 1H NMR chemical shifts indicate some interactions between OH groups in both species forming this NADES (probably O–H…O intermolecular hydrogen bonds) and/or fast exchange of mobile hydroxyl protons between Ch+ cation and LacA molecule. A similar phenomenon was already described for the same or similar ChCl:LacA (1:1, 1:2, 1:3) systems in CD3OD but the authors focused mainly on 1H chemical shifts averaging, neglecting signals broadening and the multiplicity loss [36]. In contrast, some other authors who studied 1H NMR spectrum of the ChCl:LacA (1:1) system in DMSO‑d6 did not report the OH signals at all [37]. Even more interesting are the 1H NMR spectra measured for the presently studied neat NADESs. The reason for their recording was the obvious but often omitted fact that each ChCl:LacA mixture, when dissolved in any NMR solvent, principally becomes a different physicochemical system than before dissolving; moreover, it is questionable whether such a solution remains eutectic at all. Despite this manifest circumstance, the majority of 1H NMR spectra of the previously studied NADESs were recorded just in NMR solvents. In particular, these were the NADESs containing ChCl and various carboxylic acids, measured in the following NMR solvents: CDCl3 (ChCl + malonic acid [38]); DMSO‑d6 (ChCl + acetic acid [39], ChCl + OxA [40], [41], ChCl + LacA [37]); CD3OD (ChCl + LacA [36]); D2O (ChCl + acrylic or methacrylic acid [42], ChCl + OxA [40]). The problem is that the authors of the above papers usually assumed tacitly they measured by 1H NMR the same eutectic systems like those prepared for some other purposes (e.g. for extraction of organics from food etc.) – but it was not true. In fact, the mixtures obtained for 1H NMR measurements, consisting of both components forming a given NADES and an excess of the NMR solvent (thus, containing three different chemicals), most likely have no more eutectic properties – which were usually a crucial feature from the viewpoint of the other scientific analysis. It seems not too probable, for instance, that the mixture of 1 % ChCl, 1 % of LacA, and 98 % of DMSO‑d6, CD3CN or CDOD (an example from the present work) is still eutectic; at least, such a default assumption cannot be made a priori. In contrast, the already reported NADESs were 1H NMR-studied as neat liquids only exceptionally, e.g. ChCl + OxA [43]. Hence, we have decided to perform 1H NMR measurements of the presently studied NADESs also as neat liquids, i.e. upon maintaining their eutectic nature. However, our attempts to detect the 1H NMR spectra of neat ChCl:LacA (1:1), (1:2) and (3:2) mixtures in a classic way, i.e. in standard 5 mm NMR tubes, have generally failed due to the bad signals quality. It is a well-known phenomenon, caused by high viscosity of such NADESs and very large concentrations of both components in the liquid phase. In order to solve this problem, we have placed each studied NADES in a coaxial capillary insert into a 5 mm NMR tube filled with neat CDCl3 (which played the role of an external field-frequency lock and a chemical shift reference). The 1H NMR measurements for these systems yielded much better spectra, although some signals, especially those deriving from ChCl, were significantly broadened and lost their multiplicities. Another issue was that the 1H NMR spectrum for ChCl:LacA (3:2) mixture did not correspond well to the real composition of the liquid phase, as we observed, during the ongoing measurement, slow precipitation of an excess of ChCl and the gradual decline of current integrations of the choline signals; this is why we have finally not included the respective spectral data into Table 1. The 1H NMR chemical shift differences between neat ChCl:LacA (1:1) and (1:2) mixtures seem to be rather negligible. It is interesting, however, that while these parameters for the C-bonded protons in the cholinium cation and the LacA molecule, detected for neat liquids, are quite similar to the ones recorded in the three earlier used NMR solvents – those for OH protons (yielding one shared 1H signal derived from both species) have become by ca. 1–3 ppm larger. Such a deshielding effect, observed for the neat NADESs, evidently reflects the formation of O–H…O hydrogen bonds, in accordance with general principles of 1H NMR spectroscopy. Thermal analysis The thermal properties of NADESs are crucial for their future applications. Among these properties, the onset decomposition temperature is a key parameter, as it defines the maximum temperature at which NADESs can remain in the liquid state without undergoing thermal degradation. Consequently, this temperature determines the practical application range of NADESs when used as solvents [44]. The TGA curves of the nine prepared NADES and their individual components, showing their mass-loss behavior upon heating to elevated temperatures, are shown in Fig. 3a-c. TGA (a, b, c) and DTG (d, e, f) curves of NADESs composed of choline chloride (ChCl) and lactic acid (LacA) at different molar ratios (1:2, 1:1, 3:2) with water contents ranging from 20 to 40 %. An alternative and complementary approach for thermal stability assessment is the determination of the peak temperature (T_peak), corresponding to the maximum of the derivative thermogravimetric (DTG) curve (Fig. 3d-f). The use of DTG analysis is particularly important to avoid ambiguities associated with conventional TGA curves, especially in systems exhibiting overlapping or gradual mass-loss processes [44], [45]. The nature and strength of hydrogen bonding interactions in NADESs had a significant impact on their thermal stability. All investigated NADESs contained water in the range of 20–40 wt%. In all cases, the first degradation step corresponds to dehydration, typically occurring at temperatures close to 100 °C. The initial mass losses exceeded the nominal water content in several systems, including ChCl:LacA (1:2) + 20 % H2O (27.3 %), ChCl:LacA (1:2) + 30 % H2O (41.6 %), ChCl:LacA (1:1) + 20 % H2O (26.6 %), ChCl:LacA (1:1) + 30 % H2O (33.4 %), ChCl:LacA (1:1) + 40 % H2O (41.2 %), ChCl:LacA (3:2) + 20 % H2O (21.9 %), ChCl:LacA (3:2) + 30 % H2O (31.2 %), and ChCl:LacA (3:2) + 40 % H2O (50.0 %). This behavior can be attributed to the hygroscopic nature of both the starting materials and the NADESs themselves, as well as to moisture uptake during preparation and storage [45]. Another explanation for mass losses exceeding the amount of added water is the partial overlap of dehydration and decomposition processes, resulting from the relatively low thermal stability of LacA. Similar behavior was reported by Jurić et al. [45] for ChCl–LacA systems, where water evaporation was completed below 100 °C, while decomposition commenced at comparatively low temperatures. The initial decomposition temperature made it possible to determine the highest temperature at which the NADESs can remain in the liquid state and therefore be used without undergoing decomposition. The temperature range was from 80.7 °C for ChCl:LacA (1:2) + 40 % H2O to 96.8 °C for ChCl:LacA (1:1) + 20 % H2O, with an average value of 89.3 °C. These results indicate that the proposed NADESs would be useful for the extraction processes, where increasing the temperature is important for releasing extractable components [44]. Numerous studies have demonstrated that water strongly influences the intrinsic structure of NADESs. Theoretical investigations suggest that NADESs can accommodate hydration levels of up to approximately 30 wt% while largely preserving the characteristic features of the eutectic mixture. Further water addition (up to 50 wt%) leads to a transition from water-in-NADES systems to NADES-in-water systems [35]. Interestingly, in the present study, increasing water amounts did not result in a systematic shift of T_peak toward lower temperatures. This observation suggests that the added water content did not significantly disrupt the hydrogen-bonding network between ChCl and LacA, which may be related to the specific nature of the hydrogen bond donor and the resulting interaction strength within the NADES. Nevertheless, NADESs containing 20 % water exhibited steeper TGA curve slopes than those containing 30 or 40 % water (Fig. 3a-c). The subsequent decomposition step was primarily governed by the thermal stability of LacA, which acts as the HBD, whereas ChCl (the HBA) exhibited relatively high thermal stability (T_peak = 313.9 °C). Consistent with the findings of Komesu et al. [46], the DTG profile of pure LacA showed three decomposition stages: the first at approximately 93 °C, corresponding to the evaporation of water and methanol impurities (13.1 % mass loss), the second at around 202 °C (80.7 % mass loss), associated with LacA evaporation, and the third at approximately 352 °C (6.0 % mass loss), attributed to lactide decomposition. In NADES systems, the decomposition of LacA generally occurred in two stages. The first evaporation stage was observed between 226 and 250 °C, followed by a second stage exhibiting T_peak values in the range of 254–269.5 °C. The DTG curves indicate that hydrogen bonding interactions contributed to the enhanced thermal stability of LacA. Notably, some NADESs exhibited additional decomposition or evaporation stages (e.g., ChCl:LacA (1:2) + 30 % H2O at 151.5 °C; ChCl:LacA (1:2) + 40 % H2O at 148, 191.5, and 251 °C; ChCl:LacA (1:1) + 20 % H2O at 218 and 269.5 °C; ChCl:LacA (3:2) + 40 % H2O at 201.2 °C). These additional features may result from complex electrostatic effects, variations in water content, and diverse intermolecular interactions. The corresponding processes were also evident in the TGA curves as subtle changes in slope, manifested as slight inflections. Similar observations, confirming enhanced LacA stability through hydrogen bonding and suppression of lactide formation, were reported by Škulcová and Jablonský [47] for NADESs based on betaine, glycine, and alanine. The final decomposition step of ChCl:LacA occurred at lower temperatures than that of pure ChCl, typically in the range of 289.0–303.2 °C (average 295 °C). An exception was observed for ChCl:LacA (1:1) + 40 % H2O, where the final decomposition stage appeared at an even lower temperature (279.3 °C). NADESs with a 1:1 M ratio of ChCl to LacA displayed a steeper TGA slope than the 1:2 and 3:2 systems, with the exception of the 3:2 system containing 20 wt% water, which likewise exhibited a pronounced slope (Fig. 3a-c). Overall, the NADESs prepared with different component ratios showed similar thermograms, with no unequivocal shifts in T_peak values in the DTG curves, whereas variations in water content were mainly reflected in changes in the slope of the TGA curves (Fig. 3). Analytical features of the developed ICP-MS method based on green ultrasound-assisted extraction The certified reference material (Polish Virginia Tobacco Leaves, INCT‑PVTL‑6) was subjected to ultrasound-assisted extraction (UAE) using a ChCl:LacA (1:1, w/w) mixture containing 30 % water for 30 min. The extracted macroelements (Na, K, Ca, and Mg) and microelements (Mn, Zn, Cu, and Fe) were analyzed by ICP‑MS. The certified and experimentally obtained concentrations of the chemical elements are listed in Table 2. Table 2The concentrations of macro- and microelements in certified reference material (Polish Virginia Tobacco Leaves (INCT-PVTL-6)) prepared using UAE-ChCl:LacA procedure (ChCl:LacA (1:1), WC = 30 % and sonication time = 30 min) and microwave-assisted acid digestion (MW-AD).ElementCertified value ± SD(mg/kg)UAE-ChCl:LacAMW-ADObtained value* ± SD (mg/kg)Recovery (%)tcalc*Obtained value* ± SD(mg/kg)Recovery (%)tcalc*Na62.4 ± 2.178.1 ± 2.412511.31758.1 ± 4.1931.814K26400 ± 90026793 ± 1771013.84124452 ± 1686931.999Ca22970 ± 78017151 ± 8767511.49222224 ± 818971.578Mg2410 ± 901409 ± 1205814.4312320 ± 129961.207Mn136 ± 5144 ± 101061.384133 ± 2973.460Zn43.6 ± 1.444.8 ± 21031.03839.9 ± 2.8922.286Cu5.12 ± 0.205.80 ± 0.41132.9414.33 ± 0.4843.547Fe258 ± 9261 ± 1.51013.460256 ± 3991.153*n = 3; SD − standard deviation; *Student’s t-test: tcalc − calculated absolute value, tcritic = 4.303 (p = 0.05). The concentrations of macro- and microelements in certified reference material (Polish Virginia Tobacco Leaves (INCT-PVTL-6)) prepared using UAE-ChCl:LacA procedure (ChCl:LacA (1:1), WC = 30 % and sonication time = 30 min) and microwave-assisted acid digestion (MW-AD). *n = 3; SD − standard deviation; *Student’s t-test: tcalc − calculated absolute value, tcritic = 4.303 (p = 0.05). It can be noted that the calculated t values (tcalc = 1.038 – 3.841) for potassium and all studied microelements were lower than the critical t value (tcrit = 4.303) at the 95 % confidence level (n = 3). These results confirm that the UAE–ChCl:LacA procedure is free from systematic errors. Accordingly, no statistically significant differences were observed between the certified and experimental values for the potassium, manganese, zinc, copper, and iron. In contrast, the obtained t-values of 11.317, 11.492, and 14.431 for sodium, calcium, and magnesium, respectively, were significantly higher than the critical t-value. Thus, the null hypothesis is rejected due to significant differences between the certified values and mean concentrations of these three analytes determined in the ChCl:LacA-based extract. Moreover, the prepared NADES was evaluated for recoveries of four macro- and four microelements after UAE-ChCl:LacA and ICP-MS determination (Table 2). Application of the NADES yielded recoveries within the acceptable range (101 – 113 %) for potassium, manganese, zinc, copper, and iron, indicating high method accuracy. However, recovery values (58 – 125 %) for sodium, calcium, and magnesium were outside the acceptable range. Meanwhile, preparation of certified material using the conventional MW-AD method resulted in recovery values ranging from 84 % to 99 % for all studied chemical elements, whereas the calculated t values (tcalc = 1.153 – 3.547) were lower than the critical t value (tcrit = 4.303) (Table 2). Comparison of the mean values of element concentrations using Student’s t-test (95 % confidence interval) and recovery values, indicating that the proposed ChCl:LacA procedure combined with ICP-MS can potentially be applied to the analysis of selected micronutrients. In addition, statistically significant differences between the recovery values for sodium, calcium and magnesium obtained after green extraction and reference MW-AD suggest the existence of different types of chemical interactions between the solvent and these analytes in the matrix. On the other hand, calibration curves at various concentrations were prepared using element standard solutions, and the validation parameters (coefficient of determination (R2), slope b ± Sb, intercept a ± Sa, limits of detection (DL), and quantification (QL)) are listed in Table 3. Table 3Analytical characteristics of the proposed ICP-MS procedure.ParameterChemical elementsNaKCaMgMnZnCuFeConcentration range (mg/L)0 – 20 – 1000 – 1000 – 1000 – 20 – 20 – 0.10 – 2b ± Sb*29.40 ± 1.0231.71 ± 0.082.317 ± 0.0070.167 ± 0.00114.458 ± 0.0758.661 ± 0.04996.46 ± 0.96298.3 ± 1.2a ± Sa*2.58 ± 0.834.68 ± 0.720.757 ± 0.064−0.006 ± 0.0120.016 ± 0.0140.024 ± 0.0090.0342 ± 0.0090.37 ± 0.24Sy/x*1.2034.9395.48 × 10-21.87 × 10-32.63 × 10-31.14 × 10-31.45 × 10-37.23 × 10-1R20.99951.00001.00000.99980.99990.99990.99960.9999RSDslope (%)3.50.20.30.80.50.61.00.4DL (mg/L)1.35 × 10-15.14 × 10-17.80 × 10-23.70 × 10-26.00 × 10-44.35 × 10-44.97 × 10-58.00 × 10-3QL (mg/L)4.09 × 10-11.5582.36 × 10-11.12 × 10-11.82 × 10-31.32 × 10-31.51 × 10-42.42 × 10-2* n = 5; b − slope; a − intercept; Sb, Sa − standard deviations of slope and intercept; R2 – determination coefficient; Sy/x – standard deviation of y-residuals; limit of detection (DL) = (3.3 × Sy/x)/b; limit of quantification (QL) = (10 × Sy/x)/b. Analytical characteristics of the proposed ICP-MS procedure. * n = 5; b − slope; a − intercept; Sb, Sa − standard deviations of slope and intercept; R2 – determination coefficient; Sy/x – standard deviation of y-residuals; limit of detection (DL) = (3.3 × Sy/x)/b; limit of quantification (QL) = (10 × Sy/x)/b. The calibration curves showed good linearity (R2 = 0.9995 – 1.0000) over the standard concentration ranges of 0 – 100 mg/L for K, Ca, and Mg; 0 – 2 mg/L for Na, Mn, Zn, and Fe; and 0 – 0.1 mg/L for Cu. Furthermore, the DL and QL values calculated from the calibration curves and presented in Table 3 demonstrate the high sensitivity of the proposed method. The relative standard deviation of the slopes (RSDslope) ranged from 0.2 to 3.5 %, indicating good within-day precision (repeatability) of the developed methodology. Optimization of extraction procedure of elements via Box–Behnken response surface methodology The experimental and predicted concentrations of macro- (Na, K, Ca, and Mg) and microelements (Mn, Zn, Cu, and Fe) in gingerbread extracts prepared using NADES-based UAE are listed in Table 4, Table 5. Table 4Three-level Box-Behnken design with experimental and predicted values for responses: macronutrient concentrations in gingerbread extracts.Exp.Independent variablesDependent variablesChCl:LacA (mol/mol)WC (%)Time(min)cNa ± SD (mg/kg)cK ± SD (mg/kg)cCa ± SD (mg/kg)cMg ± SD (mg/kg)Exp.Pred.Exp.Pred.Exp.Pred.Exp.Pred.10.5 (−1)20 (−1)30 (0)50.93 ± 0.05d52.72151.39 ± 6.54c,d154.861.97 ± 0.10f2.1580.45 ± 1.05f,g78.4121.5 (1)20 (−1)30 (0)51.16 ± 0.35d49.37142.69 ± 10.03b,c139.222.34 ± 0.10h2.1682.61 ± 1.64g,h84.6530.5 (−1)40 (1)30 (0)64.02 ± 1.44i65.81184.27 ± 10.43e187.742.02 ± 0.07f2.2063.35 ± 0.85b61.3141.5 (1)40 (1)30 (0)52.03 ± 0.10d,e50.24139.18 ± 0.41b,c135.712.21 ± 0.03g2.0351.29 ± 0.14a53.3350.5 (−1)30 (0)20 (−1)83.35 ± 0.72j81.56220.26 ± 1.27f216.792.56 ± 0.03j2.3877.15 ± 2.95e,f79.1961.5 (1)30 (0)20 (−1)54.86 ± 1.28f56.65143.51 ± 3.04b,c146.981.98 ± 0.01f2.1684.75 ± 0.67h82.7170.5 (−1)30 (0)40 (1)45.76 ± 0.00a,b43.97121.03 ± 1.85a,b117.561.71 ± 0.06d,e1.5376.75 ± 0.89e78.7981.5 (1)30 (0)40 (1)48.14 ± 0.92c49.93116.23 ± 5.09a119.701.41 ± 0.06a1.5975.58 ± 0.49d,e73.5491 (0)20 (−1)20 (−1)53.16 ± 0.83e,f53.16130.22 ± 0.11b,c130.221.78 ± 0.02e1.7870.38 ± 0.45c70.38101 (0)40 (1)20 (−1)47.05 ± 0.17b,c47.05124.25 ± 1.67a,b124.251.42 ± 0.02a1.4283.21 ± 0.04g,h83.21111 (0)20 (−1)40 (1)44.22 ± 0.49a44.22113.02 ± 4.39a113.021.55 ± 0.03b,c1.5571.54 ± 2.81c71.54121 (0)40 (1)40 (1)54.53 ± 0.76f54.53142.77 ± 0.20c,d142.771.75 ± 0.05e1.7584.14 ± 3.12h84.14131 (0)30 (0)30 (0)57.70 ± 0.45g58.29151.83 ± 3.76c157.361.62 ± 0.02c,d1.6172.60 ± 1.50c,d74.47141 (0)30 (0)30 (0)57.27 ± 0.75g58.29158.85 ± 11.72d157.361.54 ± 0.01b,c1.6173.12 ± 0.15c,d74.47151 (0)30 (0)30 (0)59.91 ± 0.19h58.29161.41 ± 5.92d157.361.66 ± 0.01c,d1.6177.70 ± 0.91e,f74.47After MW-AD digestion590.5 ± 111384.7 ± 135254.6 ± 10330.3 ± 1.0*n = 3 SD – standard deviation; ChCl:LacA – choline chloride to lactic acid molar ratio; WC – water content in ChCl:LacA solvents; Time - sonication time; Exp. – experimental results; Pred. – predicted values; probability, p = 0.05; coded levels are given in parentheses; different letters within the same column indicate significant differences between concentrations of macronutrients (one-way ANOVA and Duncan test, p < 0.05); MW‑AD – microwave‑assisted acid digestion. Three-level Box-Behnken design with experimental and predicted values for responses: macronutrient concentrations in gingerbread extracts. *n = 3 SD – standard deviation; ChCl:LacA – choline chloride to lactic acid molar ratio; WC – water content in ChCl:LacA solvents; Time - sonication time; Exp. – experimental results; Pred. – predicted values; probability, p = 0.05; coded levels are given in parentheses; different letters within the same column indicate significant differences between concentrations of macronutrients (one-way ANOVA and Duncan test, p < 0.05); MW‑AD – microwave‑assisted acid digestion. Table 5Three-level Box-Behnken design with experimental and predicted values for responses: micronutrient concentrations in gingerbread extracts.Exp.Independent variablesDependent variablesChCl:LacA (mol/mol)WC (%)Time(min)cMn ± SD (mg/kg)cZn ± SD (mg/kg)cCu ± SD (mg/kg)cFe ± SD (mg/kg)Exp.Pred.Exp.Pred.Exp.Pred.Exp.Pred.10.5 (−1)20 (−1)30 (0)0.50 ± 0.03f0.524.68 ± 0.08a5.110.78 ± 0.01b,c0.8312.28 ± 0.07g11.6721.5 (1)20 (−1)30 (0)0.56 ± 0.00h0.5420.38 ± 0.52g19.960.47 ± 0.02a0.437.04 ± 0.26b7.6530.5 (−1)40 (1)30 (0)0.53 ± 0.01f,g0.554.30 ± 0.02a4.730.74 ± 0.01b0.7911.22 ± 0.95f10.6141.5 (1)40 (1)30 (0)0.37 ± 0.01e0.359.59 ± 0.38b9.170.92 ± 0.01d0.884.14 ± 0.15a4.7550.5 (−1)30 (0)20 (−1)1.15 ± 0.01l1.1332.59 ± 0.49j32.171.32 ± 0.05f1.2813.50 ± 0.15h14.1161.5 (1)30 (0)20 (−1)0.96 ± 0.02k0.9843.10 ± 0.33l43.530.68 ± 0.01b0.737.47 ± 0.11b6.8670.5 (−1)30 (0)40 (1)0.39 ± 0.01e0.3727.88 ± 0.56h27.460.87 ± 0.01c,d0.8310.88 ± 0.04e,f11.4981.5 (1)30 (0)40 (1)0.31 ± 0.01d0.3334.96 ± 0.75k35.391.02 ± 0.01e1.079.45 ± 0.25c,d8.8491 (0)20 (−1)20 (−1)0.19 ± 0.01a0.1930.73 ± 0.47i30.730.76 ± 0.02b0.7612.55 ± 0.39g12.55101 (0)40 (1)20 (−1)0.70 ± 0.01i0.7014.33 ± 0.42d14.330.68 ± 0.00b0.6810.03 ± 0.19d,e10.03111 (0)20 (−1)40 (1)0.83 ± 0.00j0.8316.99 ± 0.26e16.990.44 ± 0.03a0.4410.27 ± 0.36e10.27121 (0)40 (1)40 (1)0.54 ± 0.02g0.5417.95 ± 0.81f17.951.96 ± 0.02h1.9610.12 ± 0.57d,e10.12131 (0)30 (0)30 (0)0.27 ± 0.01c0.2513.91 ± 0.22d12.971.45 ± 0.05g1.439.05 ± 0.02c9.32141 (0)30 (0)30 (0)0.23 ± 0.00b0.2512.80 ± 0.05c12.971.35 ± 0.05f1.439.12 ± 0.00c9.32151 (0)30 (0)30 (0)0.25 ± 0.01b,c0.2512.19 ± 0.04c12.971.49 ± 0.02g1.439.80 ± 0.11c,d9.32After MW-AD digestion7.19 ± 0.15.63 ± 0.200.58 ± 0.0112.32 ± 0.01*n = 3 SD – standard deviation; ChCl:LacA – choline chloride to lactic acid molar ratio; WC – water content in ChCl:LacA solvents; Time - sonication time; Exp. – experimental results; Pred. – predicted values; probability, p = 0.05; coded levels are given in parentheses; different letters within the same column indicate significant differences between concentrations of micronutrients (one-way ANOVA and Duncan test, p < 0.05); MW‑AD – microwave‑assisted acid digestion. Three-level Box-Behnken design with experimental and predicted values for responses: micronutrient concentrations in gingerbread extracts. *n = 3 SD – standard deviation; ChCl:LacA – choline chloride to lactic acid molar ratio; WC – water content in ChCl:LacA solvents; Time - sonication time; Exp. – experimental results; Pred. – predicted values; probability, p = 0.05; coded levels are given in parentheses; different letters within the same column indicate significant differences between concentrations of micronutrients (one-way ANOVA and Duncan test, p < 0.05); MW‑AD – microwave‑assisted acid digestion. It can be noted that the efficiency of the proposed green extraction of chemical elements (except Mg and Zn) was most effective (cNa = 83.35 mg/kg, cK = 220.26 mg/kg, cCa = 2.56 mg/kg, cMn = 1.15 mg/kg, cCu = 1.32 mg/kg, and cFe = 13.50 mg/kg) after treatment of ChCl:LacA (1:2) with the intermediate amount of water (WC = 30 %) at the shortest sonication time (20 min). Simply changing the molar ratio of the NADES components from 1:2 to 3:2, while adding the same amount of water (30 %) at the same ultrasound treatment time (20 min), resulted in the extraction of the highest amounts of magnesium (84.75 mg/kg) and and zinc (43.10 mg/kg). However, the lowest extraction efficiencies for Na, K, Zn, and Cu were achieved with the lowest amount of water added to NADESs (20 %) and a longer process time (30–40 min). This can be explained by the fact that high density and viscosity of NADESs with a low amount of added water reduce the solubility of metal ions and their extraction efficiency due to low diffusivity [48]. Although the presence of the carboxyl substituent of one component of the used NADESs influenced both density and viscosity of ChCl:LacA, as the carboxyl group is a substantial hydrogen‑bond donor with a high capacity to form hydrogen bonds. Furthermore, the use of NADESs with high viscosity and surface tension in the ICP‑MS technique reduces the transport efficiency of the sample into the plasma, as these properties affect spray formation in the nebulizer and droplet separation in the nebulization chamber, ultimately decreasing the concentrations of the inorganic analytes detected [14]. Unexpectedly, the contents of Mg (51.29 mg/kg) and Fe (4.14 mg/kg) were the lowest for the gingerbread sample treated with NADES with the highest molar ratio of ChCl to LacA (3:2), which contained the highest amount of water (40 %) for a shorter sonication time (30 min) (Table 4, Table 5). Probably, a lower molar proportion of LacA in the mixture with ChCl (3:2) led to a trend similar to that observed for density and viscosity, resulting in a decrease in the extraction efficiency of these two elements. This behavior is due to the simple chemical structure of LacA and its intermolecular interactions, which are reflected in its relatively low melting point of around 17 °C and its liquid state at room temperature [36]. In addition, the Duncan test indicated that changing the molar ratio of ChCl to LacA from 1:2 to 3:2, while keeping the lowest water content (WC = 20 %) and the extraction time (30 min) constant, did not significantly affect the Na, K, and Mg contents in the prepared gingerbread extracts (Table 4). On the other hand, the concentrations of the three micronutrients (Mn, Zn, and Cu) in the extracts obtained after 30 min of ultrasound treatment with the NADES composed of ChCl and LacA at the lowest molar ratio (1:2), and with added water (20 % or 40 %), also did not differ significantly (Table 5). This suggests that the lowest molar ratio of HBA to HBD had a greater effect on the viscosity and solubility of micronutrients than the reduction in viscosity from adding more water to NADES. However, increasing the water content of the extractant ChCl:LacA (1:2) from 30 % (exp. 7) to 40 % (exp. 3) did not significantly increase iron concentrations in the gingerbread extracts after a shorter ultrasound treatment time (Duncan test, Table 5). Moreover, the Duncan test indicated that increasing the WC from 20 % to 40 % in ChCl:LacA (1:1) and prolonging the extraction time (exps. 10, 11, and 12) did not significantly affect the Fe contents in the prepared samples (10.03 – 10.27 mg/kg). Importantly, similar values for each analyzed macro- and microelement can be observed in the gingerbread extracts prepared under the same UAE–NADES conditions (experiments 13, 14, and 15). However, significant differences in the concentrations of all macroelements (Na, K, Ca, and Mg) and manganese were observed between the gingerbread samples prepared using the proposed green UAE–NADES procedure (showing approximately 5–10 times lower values) and those obtained using the conventional MW-AD method (Table 4, Table 5). The less efficient extraction of sodium, potassium, calcium, magnesium, and manganese into the proposed NADESs can be explained by the formation of more stable complexes within the gingerbread matrix itself than those formed with LacA in the ChCl:LacA phase. Therefore, gingerbread samples pose an analytical challenge due to their high content of sugars, fats, and other compounds strongly bound to these elements, which hinders their release into green solvents and interferes with element separation. Evidently, the nature and composition of ChCl:LacA have a significant influence on the dissolution and extraction recovery of the three micronutrients (Zn, Cu, and Fe) from the complex gingerbread matrix. LacA, as a precursor of the prepared NADESs, likely acts as a chelating agent, promoting the effective separation and transfer of Zn, Cu, and Fe from the gingerbread matrix into green solvents. Importantly, the higher concentrations of zinc (4.30 – 43.10 mg/kg) and copper (0.44 – 1.96 mg/kg) in ChCl:LacA-based extracts compared to those determined in gingerbread samples after the conventional MW-AD procedure (cZn = 5.63 mg/kg and cCu = 0.58 mg/kg) may be attributed to the formation of more soluble chelate complexes with Zn and Cu ions in the NADES systems, which are more readily extracted from the matrix. This confirms the differing binding affinities of chemical elements within gingerbread matrices, as well as the distinct affinity of LacA, as an NADES precursor, for inorganic analytes, which can be exploited for their selective separation. It is likely that metal ions are directly coordinated to the oxygen atoms of the hydroxyl and carboxyl groups of LacA. Furthermore, replacing the aggressive nitric acid and hydrogen peroxide used in standard wet digestion with ChCl:LacA-based green extraction prevents inorganic analyte losses due to carbonisation and volatilisation. Similarly, in our previous studies [23], ChCl-based DESs containing MalA or LacA as HBD were more effective green solvents for the extraction of Zn (c = 2.41 – 28.66 mg/kg), Cu (c = 0.56 – 3.45 mg/kg) and Fe (2.96 – 8.96 mg/kg) from uniced and iced gingerbread samples. This suggests the formation of stable chelating structures between these metal ions and the carboxyl and hydroxyl groups of carboxylic acids, which are subsequently transferred to a DES-rich phase. In this case, the carboxyl groups of these two DESs proved highly effective for the extraction of three micronutrients from the gingerbread sample. Thus, DESs acted as both extraction solvents and complexing reagents. Apart from our study, according to the literature review, NADESs composed of ChCl and LacA have not been used so far for the preparation of gingerbread samples prior to determining their elemental profiles. However, ChCl:LacA (1:1) combined with UAE (ultrasonication time = 45 min) has been proposed as an effective extraction medium for iron from bovine liver (194–331 mg/kg), sheep liver (111–308 mg/kg), and chicken liver (253 mg/kg) prior to determination by flame atomic absorption spectrometry [30]. The same research group [31] selected NADESs composed of ChCl and LacA for the digestion and ultrasound-assisted liquid-phase microextraction of copper in liver samples. The obtained copper concentrations ranged from 27 to 159 mg/kg. Furthermore, Shishov et al. [32] demonstrated that ChCl and LacA (1:1) containing 30 % water was the most effective green solvent used in a simple analytical procedure for the determination of 20 chemical elements (Ag, Al, B, Ba, Ca, Co, Cr, Cu, Fe, K, Mg, Mn, Mo, Na, Ni, Pb, Sn, Ti, V, Zn) in oil samples by the ICP-OES technique. Fitting the RSM models for predicting chemical element concentrations In this study, a three-level, three-variable BBD (Table 4, Table 5) based on RSM was applied to statistically specify the effect of the UAE conditions, namely, the ChCl and LacA molar ratios, amount of water (WC) in NADESs and sonication time, affecting concentrations of macro- and micronutrients in 15 gingerbread extracts. The experimental profiles of chemical elements were fitted to the PCM (equation (1), and the least‑squares method was used to calculate the regression coefficients for the linear, quadratic, and interaction terms, which are given in supplementary material Table S3. The significance of UAE conditions and their interactions was verified by using the analysis of variance (ANOVA) results of the fitted models for all estimated responses (amounts of four macroelements: Na, K, Ca, and Mg and four microelements: Mn, Zn, Cu and Fe in prepared extracts). In ANOVA, the F-test, lack of fit, determination coefficient (R2), and adjusted coefficient (R2adj) were calculated to estimate the goodness of fit of the polynomial models (Table 6). Table 6ANOVA results for the responses: concentrations of macro- and microelements in the gingerbread extracts.Model parametersdfSSMSF valueSSMSF valuecNacKRegression111397.49127.0463.32*11817.791074.3543.67*Residual329.759.92145.5348.51Lack-of-fit125.7425.7412.8396.3396.333.92Pure error24.012.0149.2024.60Total141427.2411963.32R2, Adjusted R20.9767, 0.89130.9866, 0.9373cCacMgRegression111.500.1436.46*1221.82111.0714.10Residual30.270.08949.0916.36Lack-of-fit10.260.2669.43*33.3333.334.23Pure error20.00750.003715.757.88Total141.761270.91R2, Adjusted R20.8353, 0.23120.9549, 0.7897cMncZnRegression111.3290.121302.09*1785.69162.34213.48*Residual30.00440.00152.970.989Lack-of-fit10.00360.00369.031.451.451.90Pure error20.000800.000401.520.76Total141.3341788.66R2, Adjusted R20.9961, 0.98180.9984, 0.9926cCucFeRegression112.4210.22042.32*73.006.6438.66*Residual30.02660.00893.291.10Lack-of-fit10.01620.01623.122.952.9517.20Pure error20.01040.00520.340.17Total142.44776.29R2, Adjusted R20.9897. 0.95200.9573. 0.8008* p < 0.05; SS – sum of squares; MS – mean square. ANOVA results for the responses: concentrations of macro- and microelements in the gingerbread extracts. * p < 0.05; SS – sum of squares; MS – mean square. As shown, F-values ranging from 36.46 to 302.09 and low p-values (p < 0.05) indicated that the empirical models for all chemical elements (except Mg content) were significant. Therefore, these mathematical models are valid and convenient for predicting determined elements in gingerbread extracts prepared under any combination of variables (UAE conditions using ChCl:LacA as a green solvent). Moreover, the lack-of-fit values were insignificant (p > 0.05) for the models predicting concentrations of Na (p = 0.070), K (p = 0.186), Mg (p = 0.176), Mn (p = 0.095), Zn (p = 0.302), Cu (p = 0.220), and Fe (p = 0.054), suggesting that these models were well fitted with respect to pure error, which was not significant. High coefficients of determination (R2 = 0.9549 – 0.9984) and R2adj values (0.7897 – 0.9926) further confirm the goodness-of-fit of the proposed models for the studied inorganic analytes (except Ca). Moreover, the small differences between R2 and R2adj (8.54 %, 4.93 %, 16.52 %, 1.43 %, 0.58 %, 3.77 %, and 15.65 % for models predicting Na, K, Mg, Mn, Zn, Cu, and Fe concentrations in ChCl:LacA-based extracts, respectively) indicate strong correlations, high reliability, and a close agreement between predicted and experimental values. Consequently, the generated models can be effectively used as reliable tools for analyzing and predicting green extraction processes for these inorganic analytes. On the contrary, the proposed PCM mathematical model for predicting Ca content did not adequately fit the data, as the lack-of-fit test was significant (F = 69.43, p = 0.014), with the p-value well below 0.05 (Table 6). Moreover, the R2 and the R2adj were 0.8353 and 0.2312, respectively. Although the model explained 83.53 % of the variability in the Ca response, the low R2adj (23.12 %) indicated a poor relationship between the response and the fitted model. The large discrepancy between R2 and R2adj (60.41 %) suggests that the model contains substantial unexplained variation and may be overfitted. Therefore, only the quadratic of ChCl:LacA, the linear of extraction time, and the two interactions ((ChCl:LacA)2 × t and WC × t) parameters of the empirical model were statistically significant (F = 21.00 – 165.58, p = 0.0060 – 0.044) for the Ca amount in the gingerbread extracts. Unexpectedly, two effects: linear term of WC and interaction (ChCl:LacA)2 × WC significantly (F values = 32.37 and 86.56, p < 0.05) influenced the Mg content. However, the variable with the most significant negative impact on concentrations of two macronutrients (Na and K) was the linear term for sonication time (F = 202.21 and 128.76, p = 0.0049 and 0.0077), while the interaction (ChCl:LacA)2 × t had a strong positive effect on their amounts (F = 114.39 and 83.03, p = 0.0086 and 0.012). Moreover, the same interaction parameter showed the most significant positive effect on Mn content in ChCl-based extracts (F = 1116.28, p = 0.00090). In contrast, only three terms of interaction ((ChCl:LacA)2 × t, (ChCl:LacA)2 × WC, and ChCl:LacA × t) (F = 1.23 – 3.87, p = 0.19 – 0.38) produced insignificant effects on Zn concentration in the prepared extracts, whereas the variable with the strongest negative effect on the extraction efficiency of this element was the quadratic term of sonication time (F = 1237.46, p = 0.00081). Meanwhile, the variables with the most significant negative impact on Fe content were the linear term of the ChCl:LacA molar ratio, the quadratic term of extraction time, and the linear term of the amount of water added to green solvents (F = 50.83 – 284.94, p = 0.0035 – 0.019). However, the interaction between the molar ratio of NADES components and the time of ultrasonic treatment of gingerbread samples positively and significantly affected the iron content (F = 30.82, p = 0.031). Therefore, the analyzed gingerbread extracts exhibited the highest Fe levels when treated with NADES containing ChCl and LacA at a molar ratio of 1:2, with a lower WC and shorter ultrasound treatment times. In addition, all interactions (except ChCl:LacA × WC), quadratic terms of WC2 and (ChCl:LacA)2, and linear WC factor (F = 25.50 – 123.08, p = 0.0080 – 0.037) were statistically positively significant for Cu concentration in extracts. This suggests that a higher water amount in NADESs with a high molar ratio between ChCl and LacA reduced their viscosity and enhanced the efficiency of Cu extraction from gingerbread cookies. Despite the good fit of the polynomial models for the studied chemical elements (except for Ca), unsatisfactory recovery values for Ca and Mg (below 90 %) and for Na (above 120 %) indicated that the ChCl:LacA system had insufficient extraction efficiency. Consequently, these analytes could not be determined with adequate accuracy using the ICP-MS method. Moreover, significantly lower amounts of Na, K, Ca, Mg, and Mn were found in the green extracts compared to wet-digested gingerbread samples. This effect, resulting from matrix interferences, namely the presence of these metal ions bound to the solid matrix and the limited ability of the proposed NADESs to transfer these inorganic analytes into the eutectic phase, restricted the applicability of the ChCl:LacA-based procedure for gingerbread sample preparation prior to the quantitative analysis of macroelements and manganese. However, good accuracy and precision (expressed as RSD < 5 %), low DL and QL values, as well as comparable concentrations of the three micronutrients in samples prepared using green NADESs and the conventional MW-AD method indicate that the ICP-MS method based on ChCl:LacA extraction, without the use of harmful acids or strong oxidizing agents, constitutes an environmentally friendly and reliable approach for the simultaneous determination of zinc, copper, and iron in gingerbread cookies. Analysis of response surfaces The response surface profiles were estimated only for the three micronutrients (zinc, copper, and iron) whose amounts in the NADES‑based gingerbread extracts were comparable to those in the mineralizates obtained after MW‑AD digestion. The three‑dimensional response surface plots illustrating the effects of interactions between pairs of independent variables (NADES‑based extraction conditions), while keeping the third variable constant at its central level, on the predicted concentrations of three micronutrients (Zn, Cu, and Fe) determined by ICP‑MS are shown in Fig. 4. Response surface plots depicting: (a) the concentration of zinc with the combined effects of the water content in the ChCl:LacA solvents and sonication time; (b) the concentration of copper with the combined effects of the choline chloride to lactic acid molar ratio and water content in the ChCl:LacA solvents; (c) the concentration of iron with the combined effects of the choline chloride to lactic acid molar ratio and sonication time. Evidently, the response‑surface profiles for Zn, Cu, and Fe obtained using ChCl:LacA at various molar ratios, water contents, and UAE durations were distinct from one another. The parabolic shapes of these response surfaces indicate that the quadratic terms of the studied independent variables were significant. Fig. 4a confirms that the positive quadratic effect of water added to the NADES, together with the negative quadratic and linear effects of extraction time, increased the Zn concentration in the gingerbread extracts, particularly when higher water amounts were combined with shorter extraction times. Although the significant positive interaction between the corresponding variables (WC × t) caused an increase in Zn content in the sample ultrasonically treated with the longest time (40 min) by NADESs containing higher WC. Moreover, the effect of the ChCl:LacA molar ratio, water in the extractant, and their interaction on the Cu concentration in gingerbread extracts was visualized in Fig. 4b. The significant positive linear and quadratic effects of WC and the quadratic term of the ChCl:LacA molar ratio on Cu content resulted in enhanced Cu content with increasing ratio of HBA to HBD and water level, thereby decreasing NADES viscosity and improving Cu extraction from the gingerbread matrix. However, the response surface and contour plot for Cu content in the ChCl-based extract showed a maximum at an intermediate ChCl:LacA ratio (1:1) and WC of 30 – 40 %. Additionally, Fig. 4c shows that the highest Fe content in the investigated extracts was obtained with the lowest molar ratio of ChCl to LacA (1:2) and the shortest sonication time (20 min) for the eco-friendly preparation of gingerbread cookies before the determination of micronutrients. The elliptical contour plot and the shape of the Fe content surface indicate that the negative linear effect of the ChCl:LacA molar ratio, the quadratic effect of extraction time, and the positive interaction between these two variables were significant. The chosen mathematical models enable the calculation of optimal operating parameters for UAE–NADES of three micronutrients (Zn, Cu, and Fe) from gingerbread cookies. The optimum conditions for the proposed green extraction were as follows: ChCl:LacA = 1:1, WC = 34 %, and t = 40 min, under which the predicted concentrations were cZn = 25.25 mg/kg, cCu = 1.81 mg/kg, and cFe = 10.23 mg/kg. The calculated optimum parameters were used in three additional experiments to validate the adequacy of the models obtained from the RSM. The mean experimental concentrations of Zn, Cu, and Fe were 24.06 ± 0.27 mg/kg, 1.71 ± 0.01 mg/kg, and 11.43 ± 0.75 mg/kg, respectively. The insignificant differences (p > 0.05) between the predicted and experimental values confirmed that the proposed models were accurate and adequate for optimizing the ChCl–based extraction of the three micronutrients from gingerbread cookies prior to their determination using the ICP–MS method. Development of machine learning algorithm − decision tree A constructed dataset representing the effectiveness of UAE–ChCl:LacA for the extraction of Zn, Cu, and Fe from gingerbread prior to ICP‑MS determination was analyzed using a decision tree approach. Three decision trees were constructed using RapidMiner, and the structures of the regression tree models are illustrated in Fig. 5. Decision tree diagrams with independent variables (ChCl:LacA molar ratio, water content in the ChCl:LacA solvents, and sonication time) affecting concentrations of a) zinc, b) copper, and c) iron in gingerbread extracts. Each internal node represents a binary partition based on a threshold value of one of the input parameters, namely, the ChCl:LacA molar ratio, water content in the deep eutectic mixture, and sonication time, selected to minimize prediction error. These decision rules divide the input space into regions with similar residual behavior, while the terminal nodes (leaves) represent the predicted response values. Furthermore, the presence of a parameter at the first split of the decision tree indicates that this variable plays a key role in explaining the system and determining the outcome. In the decision tree presented in Fig. 5a, sonication time was identified as the most important independent factor influencing Zn concentration in the prepared extracts. The highest Zn content (35.473 mg/kg) was achieved after ultrasonic treatment of the gingerbread sample with ChCl:LacA containing less than 35 % water for a duration of less than 25 min. In contrast, the water content in the ChCl:LacA green solvent was the predominant factor affecting Cu concentration in the extracts (Fig. 5b). The Cu content increased with higher water content (> 25 %) and higher ChCl:LacA molar ratios (> 0.750 mol/mol). Moreover, analysis of the decision tree for Fe concentration in gingerbread extracts (Fig. 5c) revealed that the ChCl:LacA molar ratio was the dominant factor influencing iron content. Extracts prepared using ChCl:LacA at a molar ratio below 0.750, with lower water content (< 25 %) and shorter sonication times (< 25 min), exhibited higher Fe concentrations. The RSM plots (Fig. 4) also support these findings. Sonication time and the ChCl:LacA molar ratio exerted the most significant negative effects on Zn and Fe concentrations, respectively. In contrast, the water content in the green solvent showed a strong positive effect on Cu concentration. Although the RSM analyses did not clearly elucidate the combined effects of the variables on element extraction efficiency, the decision tree model clarified this distinction. Conclusions This study investigated the potential of prepared NADESs to extract inorganic analytes from gingerbread cookies. Spectroscopic analyses, including FTIR and 1H NMR, confirmed the eutectic formation of the NADESs and elucidated the hydrogen‑bond interactions between ChCl and LacA. However, thermal analysis assessed the extent of hydrogen‑bond weakening and bond dissociation between ChCl and LacA in the prepared NADESs, as well as the thermal decomposition of these systems, which began with the loss of water. Moreover, BBD and RSM were effectively applied to optimize the UAE conditions for extracting chemical elements from gingerbread cookies with ChCl:LacA prior to their determination by ICP‑MS. All analyzed macronutrients and manganese were less effectively extracted into the proposed NADESs, which may be due to the formation of more stable complexes within the gingerbread matrix. Therefore, the optimum conditions for the effective extraction of three microelements (zinc, copper, and iron) from gingerbread samples were determined to be a molar ratio of ChCl to LacA = 1:1, a WC of 34 %, and an extraction time of 40 min. Under these optimum conditions, the concentrations of Zn, Cu, and Fe in the extracts were 24.06 mg/kg, 1.71 mg/kg, and 11.43 mg/kg, respectively. These findings demonstrate the potential of decision trees as a complementary approach to RSM, especially for visualizing variable interactions and enabling interpretable decisions in the optimization of zinc, copper, and iron extraction from gingerbread cookies. This advanced, sustainable UAE technique, employing nontoxic, renewable, and efficient green solvents for the simultaneous extraction of zinc, copper, and iron from gingerbread, may serve as a viable alternative to microwave digestion for elemental analysis prior to ICP-MS determination. CRediT authorship contribution statement Agnieszka Kowaluk: Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Dobrochna Rabiej-Kozioł: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Leszek Pazderski: Writing – original draft, Visualization, Methodology, Investigation, Data curation, Conceptualization. Aleksandra Szydłowska-Czerniak: Writing – review and editing, Supervision, Software, Resources, Project administration, Funding acquisition, Data curation, Conceptualization. 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. Acknowledgments The authors would like to warmly thank Jean Guédon for his invaluable assistance and fundamental support in laboratory analyses. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.ultsonch.2026.107907. Supplementary data The following are the Supplementary data to this article: Supplementary Data 1 Supplementary material: Tables S1–S3, including the single-factor experimental design, operating conditions of the ICP-MS spectrometer, and regression coefficients of the partial cubic model (PCM), as well as 1H NMR spectra of choline chloride (ChCl), lactic acid (LacA), and natural deep eutectic solvents (NADESs) composed of these two components. References Associated Data Supplementary Materials Supplementary Data 1 Supplementary material: Tables S1–S3, including the single-factor experimental design, operating conditions of the ICP-MS spectrometer, and regression coefficients of the partial cubic model (PCM), as well as 1H NMR spectra of choline chloride (ChCl), lactic acid (LacA), and natural deep eutectic solvents (NADESs) composed of these two components.