Dopant‐Assisted APCI for Enhanced Sugar Analysis: Overcoming Ionization Bottlenecks

ABSTRACT

Rationale

The investigation of lignocellulosic biomass by atmospheric‐pressure chemical ionization (APCI) mass spectrometry (MS) has revealed significant issues for carbohydrate ionization, particularly during biomass pyrolysis analyses. Unequal ionization of carbohydrates and significant fragmentation have been reported. The introduction of suitable dopants into the ion source shows promise for limiting fragmentation.

Methods

To investigate these ionization disturbances, both direct infusion (DI) and modified direct insertion probe (DIP) APCI coupled with a Fourier‐transform ion cyclotron resonance mass spectrometry (FT‐ICR MS) were employed. Solvents with increasing proton affinity (PA) were used to assess solvent PA influence in DI‐APCI ionization processes. The DIP system was modified to enable the introduction of an inert gas stream passing through a dopant‐rich atmosphere prior to entering the ion source.

Results

A significant PA difference between the solvent and the analyte promotes carbohydrate dissociation after their ionization. Experiments without solvent and with deuterated compounds showed that carbohydrate ionization and dissociation may be self‐induced by water formed during carbohydrate dehydration. The introduction of ammonia in the ion source promoted the [M + NH₄]⁺ adduct formation, reducing ionization‐induced dissociation and increasing the intensity of the carbohydrate signal by at least one order of magnitude in DI‐APCI. The method was successfully applied to both standards and cellulose pyrolysis samples.

Conclusions

The dopant‐assisted APCI (dAPCI) approach appears highly promising for the rapid assessment of biomass fast pyrolysis, particularly when highly oxygenated compounds are key markers. Careful selection of dopants and solvents is essential in carbohydrate analysis to preserve analyte structural integrity during ionization.

1. Introduction

The analysis of carbohydrates and their derivatives is of major importance in many scientific fields, including life sciences and renewable energy research. In biological studies, carbohydrates are central to metabolomic and glycoproteomic investigations aimed at understanding cellular processes and biomolecular interactions [1, 2]. In the energy sector, carbohydrates also constitute the main structural components of lignocellulosic biomass, which is largely composed of cellulose and hemicellulose and represents one of the most abundant renewable carbon resources [3, 4, 5, 6, 7]. Indeed, the thermochemical conversion of lignocellulosic biomass is a promising alternative to petroleum, offering a sustainable solution to mitigate global warming [8].

Fast pyrolysis is an attractive route for the valorization of lignocellulosic biomass, mainly producing a complex liquid product commonly referred to as bio‐oil. This mixture contains thousands of oxygenated compounds derived from carbohydrate depolymerization, including several platform molecules of industrial interest such as levoglucosan, one of the major products of cellulose pyrolysis [9, 10]. Because of the extreme molecular complexity of these mixtures, advanced analytical approaches are required for their characterization [11].

Ultra‐high‐resolution mass spectrometry has therefore emerged as a powerful tool for the molecular‐level analysis of bio‐oils [12, 13, 14, 15, 16]. In this context, rapid direct analysis methods such as direct insertion probe (DIP) [17, 18] or atmospheric solid analysis probe (ASAP) [19] coupled with atmospheric pressure chemical or photo‐ionization sources (APCI or APPI) have recently attracted significant interest. These techniques enable fast introduction, without tedious sample preparation, thermal desorption and degradation of the investigated samples, allowing rapid screening of complex pyrolysis mixtures. Voellinger et al. recently described a modification of the DIP setup enabling real‐time analysis of fast pyrolysis products of glucose and cellulose. In this study, the neutral species produced by pyrolysis were ionized by APCI(+) [18]. However, unlike electrospray ionization, this ionization method does not allow controlled formation of sodium or potassium adducts. Ionization therefore mainly relies on gas‐phase proton transfer reactions, leading predominantly to protonated species. The detection of protonated carbohydrates is particularly challenging because these species are relatively unstable in the gas phase [17, 20, 21, 22, 23], and extensive fragmentation may occur during APCI ionization [20, 24]. These analytical issues are particularly critical in the study of biomass conversion products.

Several strategies based on dopant‐assisted ionization have been proposed to overcome these limitations. In APCI, the introduction of gaseous dopants can modify ion–molecule reaction pathways and enhance ionization efficiency for specific classes of analytes. For example, ammonium hydroxide or ammonium acetate can promote the formation of ammonium adducts in APCI(+) [20, 25]. Indeed, ammonia is one of the most commonly used dopants in APCI [26] because it significantly modifies the reagent‐ion chemistry generated in the corona discharge. In conventional APCI, ionization is mainly initiated by hydronium ions and protonated solvent clusters [26, 27], whereas the addition of ammonia promotes the formation of NH₄ ⁺ and proton‐bound ammonia clusters [27], which become the dominant reagent ions under appropriate operating conditions. These ions may subsequently ionize analytes either by proton transfer or by ammonium‐adduct formation [27, 28], depending primarily on the gas‐phase basicity and proton affinity (PA) of the analyte. The relative abundance of these reagent ions is also influenced by the solvent composition [29] and the competitive ion–molecule reactions occurring within the APCI source. Therefore, the use of ammonia can improve both ionization efficiency and selectivity while altering the dominant ionization pathways compared with conventional APCI. These mechanistic aspects have been extensively investigated and provide the framework for interpreting the ionization behavior discussed in the present work [26, 27, 29]. Chlorinated compounds promote chloride attachment [30, 31] and the formation of stable bidentate or tridentate chelated [M + Cl]⁻ ions with saccharides, providing structural information that complements protonated and ammonium‐adduct species. Previous studies on cellulose pyrolysis have shown that chloride adducts and ammonium adducts provide complementary information on carbohydrate products [28], while the relative abundance of ammonium‐bound oligomers (notably tetramers and pentamers) evolves with the pyrolysis residence time, reflecting changes in the product distribution. Comparative APCI and ESI studies have further demonstrated that the nature of the dopant strongly influences the ionization pathways and the distribution of carbohydrate ions [32]. Finally, it was also clearly demonstrated that such dopants provide promising routes to improve the detection of compounds which are poorly ionized by proton transfer such as carbohydrates.

The objective of the present work is to develop a rapid and efficient analytical approach for characterizing carbohydrate pyrolysis products by integrating open‐DIP APCI with a dopant‐assisted ionization strategy. In order to rationalize the ionization processes involved, the fragmentation behavior of carbohydrates and their primary dehydration products was first investigated under direct infusion APCI (hereafter referred to simply as APCI) conditions using solvents with different PA. These experiments provide mechanistic insights that enable the optimization of dopant‐assisted open‐DIP APCI for the rapid analysis of biomass pyrolysis products.

2. Materials and Methods

2.1. Chemicals

Methanol (99.9%, Carlo Erba), absolute anhydrous ethanol (99.9%, Carlo Erba), 2‐propanol (99.9%, VWR), ammonia solution (30% in water, Carlo Erba), methanol‐d4 (≥ 99.8 atom% D, Sigma‐Aldrich), deuterium oxide (99.9 atom% D, Sigma‐Aldrich), and LC/MS‐grade water (Carlo Erba) were used as received. Avicel PH‐101 cellulose (> 99%, Sigma Aldrich), levoglucosenone (> 95%, Sigma‐Aldrich), levoglucosan (1,6‐anhydro‐β‐D‐glucose, > 99%, Sigma‐Aldrich), glucose (> 98%, Prolabo), and KIMBLE KIMAX open‐ended glass capillaries (Sigma‐Aldrich) were used as received.

Glucose was solubilized in water, methanol, ethanol, and 2‐propanol to obtain solutions of final concentration of 10⁻⁴ mol·L⁻¹. An equimolar solution (10⁻⁴ mol·L⁻¹ each) of glucose, levoglucosan, and levoglucosenone was prepared in methanol.

To synthesize glucose‐d5, 100 mg of glucose were solubilized in 1 mL of D₂O. In these conditions, labile hydrogen (hydrogen of the hydroxyl groups) underwent H/D exchange. After standing overnight at room temperature, the resulting solution was frozen and lyophilized for 24 h. The glucose‐d5 obtained was kept under argon atmosphere to prevent back D/H exchange. The ESI(+) spectrum of a 10⁻³ M solution of glucose‐d5 in methanol‐d4 is shown in Figure S1. Based on the isotopic pattern of the [C₆H_xO₆D_y + Na]⁺ ion (x + y = 12, y = 0–5), it was defined that a 95 atom% D exchange was obtained (Supporting Information 1).

2.2. Direct Infusion APCI and Open‐DIP APCI

An APCI II ion source (Bruker Daltonics) was employed. The corona needle voltage was set at 4500 V, with a discharge current of 4 μA. The drying gas temperature was set at 220°C, with a 2.5 L·min⁻¹ flow rate of N₂. The nebulizer gas pressure was set at 2 bar.

For APCI, the APCI heater temperature was set to 250°C, and the sample was introduced in the ion source at a flow rate of 3 μL·min⁻¹.

For open‐DIP‐APCI experiments, a DIP mounted on the APCI II ion source (Bruker Daltonics) was used. KIMBLE KIMAX open‐ended glass melting capillaries (100 mm long, 1.5 mm internal diameter) were employed. These capillaries allow a controlled flow of an inert gas stream (N₂, 99.995%, Air Liquide), enabling the rapid transport of neutral species generated either by thermal desorption or by thermal degradation of the sample toward the APCI ion source. For liquid samples, 1 μL of solution was loaded at one end of the capillary using a micropipette. After solvent evaporation at room temperature, the capillary was mounted onto the DIP probe head using its opposite end. For solid samples, a precisely weighed 100 μg portion of the compound was introduced into the end of the capillary opposite to a glass wool filter (a few millimeters in length). Glass wool served to prevent solid particles from being carried into the ionization source by the gas flow.

Possible contaminants of the capillary and glass wool arising from handling were removed by inserting the capillary (with or without the glass wool, depending on the investigated sample) into the heated region of the ion source for 1 min at 530°C. The capillary was then carefully removed from the DIP probe head using tweezers before loading the liquid or solid sample.

For open‐DIP‐dAPCI, ammonia‐doped nitrogen was introduced into the capillary at a flow rate of 3 mL·min⁻¹. When introduced through the exhaust into the ion source, the flow rate was increased to 60 mL·min⁻¹ to overcome the counter‐current gas flow. To dope the nitrogen stream with ammonia, the nitrogen was first passed through the headspace of a flask containing a 1% aqueous ammonia solution. The experimental setup is schematically illustrated in Figure S2.

2.3. Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FT‐ICR MS)

Ions formed by APCI(+) or dAPCI(+) were analyzed using FT‐ICR mass spectrometry (7 T FT‐ICR MS, solariX 2xR, Bruker Daltonics, Bremen, Germany). Both ion source and instrument parameters were optimized using FTMS Control software version 2.3.0 (Bruker Daltonics). Prior to analysis, the mass spectrometer was externally calibrated using permanent well‐known contaminants easily ionized by APCI(+), such as palmitic and stearic acids, and the ICR cell was shimmed and gated.

All mass spectra were acquired over the m/z 86–520 range using a 4‐megaword time domain. A mass resolution of 500 000 was achieved at m/z 200, with a mass measurement accuracy better than 0.5 ppm. The ion accumulation time was set to 50 ms. In some experiments, the accumulation time was decreased to 20 ms to avoid overloading the ICR cell. Reported mass spectra correspond to the average of 50 scans in APCI and to the average of all scans acquired over a 30–60 s thermogram, representing the total observed ion current as a function of the time, in DIP‐APCI experiments. The DIP‐APCI experiments were performed in triplicate.

MS/MS spectra were acquired by collision‐induced dissociation (CID) in the quadrupole cell of the FT–ICR instrument, using collision energy in the 0–20 eV range. High purity argon (Air Liquide, Paris, France) was used as the collision gas. The isolation window width was set to 1 Da, and each spectrum was obtained by summing 50 individual scans. In MS/MS experiments, ion accumulation time was increased to 200 ms.

2.4. Data Treatment

DataAnalysis 5.2 software (Bruker Daltonics, Bremen, Germany) was used to internally calibrate the FT‐ICR mass spectra using a list of known compounds, yielding a standard deviation of the mass error ≤ 100 ppb. Peak lists from the different mass spectra were exported with a signal‐to‐noise ratio ≥ 4. Molecular formula assignment and graphical representations were obtained using the in‐house‐developed Punc'data software [33]. Only cations were considered, as no radical cations were observed. Two assignment passes were defined with the following constraints: pass 1, C_2–40H_4–60O_0–25; and pass 2, C_2–40H_4–80O_0–30N_1–2. The second pass allowed, in particular, to identify NH₄ ⁺ adduct ions. Double bond equivalent (DBE) values were restricted to the 0–40 range, and the Golden Rules [34] were applied to filter the assigned molecular formulas. The maximum mass error tolerance was set to 0.5 ppm. Only the molecular formula assignments detected in at least two of the three replicates were considered in DIP‐APCI experiments.

The assigned datasets were represented using H/C vs. O/C van Krevelen diagrams [35] allowing classification of compounds into specific biochemical families such as carbohydrate compounds that typically exhibit 1.4 < H/C < 2.0 and 0.7 < O/C < 1.0 [36]. In the present work, van Krevelen diagrams proved useful for recognizing and describing dehydration phenomena in carbohydrates, as depicted in Figure S3.

3. Results and Discussion

3.1. Investigation of Glucose by Direct Infusion APCI(+)

The mass spectrum obtained from an aqueous solution of D‐glucose by APCI(+) is shown in Figure 1a. No intense ion was observed at m/z 181.07067, corresponding to the theoretical m/z value of the protonated glucose. The absence of protonated glucose is attributed to its facile dehydration, whereby protonation of hydroxyl groups facilitates water elimination and the formation of a stable secondary carbocation, consistent with the well‐established fragmentation behavior of protonated alcohols in the gas phase [37]. Thus, ions at lower m/z ratios were detected, corresponding to chemical formulas consistent with sequential water losses. These observations agree with the results reported by Chen et al. [20].

FIGURE 1.

APCI(+) FT‐ICR mass spectra of glucose (10⁻⁴ mol·L⁻¹) in (a) H₂O, (b) MeOH, (c) EtOH, and (d) isopropanol, at a 250°C vaporizer temperature. The m/z ranges were selected to highlight the main ions associated with glucose degradation products. When water is used as a solvent (a), the base peak is relative to the protonated palmitic acid (m/z 257.24806). The [Palmitic acid —H₂O + H]⁺ ion was also observed at m/z 239.23749.

Since the ions observed in the mass spectrum are not radical species, they correspond to protonated molecules or other cations formed during APCI processes. In APCI(+), ions of the type [M + H]⁺ are frequently formed through gas‐phase proton transfer between a protonated solvent molecule or solvent cluster and the analyte molecule M, as described in Equation (1). This reaction is favored toward the formation of the [M + H]⁺ species when the gas‐phase PA of molecule M exceeds that of the solvent (S) or one of its clusters (S_n).

$$M_{\left(g\right)}+{\left[S\text{or}{S}_n+H\right]}_{\left(g\right)}^{+}\to {\left[M+H\right]}_{\left(g\right)}^{+}+S_{\left(g\right)}\text{or}n{S}_{\left(g\right)}$$ (1)

Although Equation (1) illustrates proton transfer from protonated solvent molecules, protonation in APCI involves a more complex network of ion–molecule reactions. Hydronium ions and their hydrated clusters are the dominant reagent ions under atmospheric‐pressure conditions and may protonate analytes directly, while simultaneous competition with solvent molecules influences the overall protonation efficiency. The relative contribution of these pathways depends on the solvent composition and the gas‐phase ion chemistry of the APCI source [29, 38].

A comparison of the gas‐phase proton affinities of water and D‐glucose indicates that this reaction is strongly favored and exothermic by approximately 140 kJ·mol⁻¹. Ab initio calculations performed by Chen et al. [20] have shown that the activation barriers associated with proton migration between hydroxyl groups are sufficiently low to allow the coexistence of several protomers. Moreover, the calculated activation barrier for the dehydration of protonated glucose was found to be on the order of ~140 kJ·mol⁻¹. This barrier can therefore be readily overcome due to the elevated temperature, frequent ion–neutral collisions at atmospheric pressure, and the exothermic proton transfer from protonated water, which together promote rapid water elimination, in agreement with the high rate reaction constant reported by Chen et al. [20].

To verify whether one of the possible pathways leading to the observed species involves dissociation of a protonated glucose molecule, tandem mass spectrometry experiments were performed. As mentioned earlier, under the experimental conditions used, the [M + H]⁺ ion was not stable; therefore, the MS/MS study was restricted to the ions at m/z 163.06010 ([Glucose − H₂O + H]⁺), 145.04954 ([Glucose − 2H₂O + H]⁺), and 127.03897 ([Glucose − 3H₂O + H]⁺). The corresponding mass spectra are shown in Figure S4. This investigation was further complemented by examining the dependence of the fragment ion distribution on the activation energy (Figure S5). MS/MS spectra and energy‐resolved breakdown curves derived from these spectra clearly show the relationship among the ions, corresponding mainly to the sequential loss of one to four water molecules from the protonated glucose ion. The collision energy required for the formation of the ions at m/z 145.04954, 127.03897, 11 705 462 ([Glucose − 2H₂O − CO + H]⁺), 109.02841 ([Glucose − 4H₂O + H]⁺), 99.04406 ([Glucose − 3H₂O − CO + H]⁺), and 97.02841 ([Glucose − 3H₂O − H₂CO + H]⁺) is relatively low, of the same order of magnitude, or even lower, than the energy released during protonation of glucose by protonated water molecules (Figure S5a).

The detection of the [Glucose − (H₂O)_1–4 + H]⁺ ions at m/z 163.06010, 145.04954, 127.03897, and 109.02841 may also have an alternative origin. Indeed, glucose is poorly stable in the gas phase at temperatures above 80°C. It has been shown that after isomerization to fructose in the gas phase, three water molecules are successively eliminated, leading to the formation of 5‐hydroxymethylfurfural (5‐HMF), whose protonated form corresponds to m/z 127.03897. The gas‐phase proton affinities of the products resulting from the successive dehydration of glucose are higher than that of a water molecule (Table 1). Consequently, these species can be protonated under the APCI(+) conditions used, and the exothermicity of these reactions appears sufficient to also permit post‐protonation dissociation processes, as observed for protonated glucose.

TABLE 1.

Proton affinities (PA) of the solvents used, ammonia, and selected compounds of interest.

Formula	Name	PA (kJ/mol)	References
H2O	Water	691.0 682.0	Hunter & Lias [39] Chen et al. [20]
CH3OH	Methanol	754.3 753.1	Hunter & Lias [39] Chen et al. [20]
C2H5OH	Ethanol	776.4	Hunter & Lias [39]
CH3CH(OH)CH3	Isopropanol	793.0	Hunter & Lias [39]
C6H12O6	β‐D‐glucose	845.2	Chen et al. [20]
C6H10O5	/	828.4	Chen et al. [20]
C6H8O4	/	811.7	Chen et al. [20]
C6H6O3	/	761.5	Chen et al. [20]
NH3	Ammonia	853.6	Hunter & Lias [39]
C6H6O3	5‐Hydroxymethylfurfural	867.3	Garzoli et al. [40]
C5H4O2	Furfural	847.9	Ricci et al. [41]

To assess the significance of post‐protonation dissociation processes on the distribution of ions observed, glucose solutions in different solvents were investigated. The gas‐phase PA of the solvents considered gradually increases from methanol to isopropanol (Table 1), allowing control over the exothermicity of the protonation reaction. The APCI(+) conditions used were identical to those employed in the study of the aqueous glucose solution, and the results are shown in Figure 1. The results are also presented in the form of van Krevelen diagrams in Figure 2.

FIGURE 2.

van Krevelen diagrams obtained for glucose (10⁻⁴ mol·L⁻¹) in (a) H₂O, (b) MeOH, (c) EtOH, and (d) isopropanol using APCI(+). Dot size is proportional to the ion relative abundance. Only species with a relative abundance > 1% are shown after the background signal has been removed. H/C calculations for [M + adduct]⁺ ions were calculated without considering the adduct contribution in order to superimpose the green dots of the [M + NH₄]⁺ onto the corresponding red dots of the [M + H]⁺ ions.

In the different experimental conditions employed, several product ions originating from consecutive dehydration of glucose ions were observed. Indeed, the [Glucose − (H₂O)_1–3 + H]⁺ ions were predominantly observed. However, their relative distributions strongly depend on the difference in PA between glucose and the considered solvent (Figure 3). The smaller this difference is, the greater the contribution of the less dehydrated species. In solvents (ethanol and isopropanol) with PA closest to that of glucose, even the protonated glucose ion can be observed. Using water as a solvent resulted in a very low intense ion (< 1%) associated with the intact protonated glucose molecule, which was also observed as [M + NH₄]⁺ adduct ion, even though no ammonia was added in source prior to the analysis. It is most likely that ammonia remnants in‐source were enough to allow formation of this adduct. Since the temperature, pressure, and nebulizing gas flow conditions were identical to those used in the study of the aqueous glucose solution, it is reasonable to conclude that the observation of the [Glucose − (H₂O)_1–3 + H]⁺ ions was primarily associated with post‐protonation dissociation processes rather than protonation of thermally degraded glucose products.

FIGURE 3.

Relative abundance distribution of [C₅H₄O₂ + H]⁺, [C₆H₆O₃ + H]⁺, [C₆H₈O₄ + H]⁺, [C₆H₁₀O₅ + H]⁺, and [C₆H₁₂O₆ + H]⁺ ions as a function of the solvent used: (blue) water, (green) methanol, (yellow) ethanol, and (red) 2‐propanol. m/z values are given under the formula.

In addition to the [Glucose − (H₂O)_1–3 − (CH₂O)_0–1 + H]⁺ ions discussed in the previous paragraph, ions corresponding to the attachment of a solvent molecule were also observed. These species can be written in the general form [Glucose − (H₂O)_1–3 − (CH₂O)_0–1 + Solvent + H]⁺. Such species have previously been reported in studies of methanolic glucose solution [20]. Interestingly, the intensity and the number of these ions increase with the PA of the solvent used, suggesting that ion–molecule association processes occurring in the ion source contribute to their formation and stabilization. Indeed, the formation and relative abundance of the observed protonated solvent clusters can be rationalized by their gas‐phase ion–molecule chemistry. Protonated alcohol clusters are stabilized through hydrogen‐bond interactions between the protonated solvent molecule and neutral solvent molecules. As the PA and hydrogen‐bond acceptor strength increase from water to methanol, ethanol, and propanol (Table 1), the corresponding proton‐bound clusters become increasingly stable, leading to higher abundances in the mass spectra. Conversely, hydrated clusters are less strongly bound and may partially dissociate during ion transfer from the APCI source to the mass analyzer, which may explain their low abundance or absence in the recorded spectra. In addition, hydrated reagent ions such as (H₂O)_nH₃O⁺ and mixed solvent clusters may be formed under atmospheric‐pressure conditions depending on the residual water content and source conditions.

These species are clearly highlighted in the van Krevelen diagrams (blue dots) shown in Figure 2. Most of the ions involving the attachment of a solvent molecule align along a diagonal parallel to that defined by the dehydration products of protonated glucose, consistent with the progressive loss of H₂O from the glucose backbone followed by association with a solvent molecule. Indeed, the interactions responsible for the attachment of solvent molecules are relatively weak. A slight increase in the drying gas flow (up to 3.5 L·min⁻¹) or the application of a low collision energy led to a major disappearance of the [Glucose + Solvent + H]⁺ ions and to the near absence of protonated glucose, to the benefit of the dehydration products. In fact, the mere selection of these ions by the quadrupole in the ion transfer line was sufficient to induce their activation (see Figure S6). These results further confirm the very low stability of protonated glucose in the gas phase and its high propensity to undergo dehydration. Nevertheless, its association with a molecule of solvent ensures its stabilization, and the stabilization of the various anhydro‐glucose isomers.

3.2. Investigation of Glucose by Open‐DIP‐APCI(+)

A 1‐μL aliquot of a glucose solution (10⁻³ mol·L⁻¹ in MeOH/H₂O, 1:1 v/v) was introduced into a capillary, which was inserted in the heater of the APCI II source after solvent evaporation. The heater temperature was set to 200°C to ensure sample volatilization. The average mass spectrum is shown in Figure 4.

FIGURE 4.

Open‐DIP APCI(+) average mass spectrum of glucose at the 200°C set temperature.

As observed in APCI(+), protonated glucose was not significantly observed unlike its first‐, second‐, and third‐step dehydration products observed at m/z 163.06012, 145.04952, and 127.03898, respectively. Ions with high m/z values were also detected, especially at m/z 325.11294, corresponding to the C₁₂H₂₁O₁₀ ⁺ ion. The literature reports that glucose heated to temperatures around 200°C undergoes dehydration to form levoglucosan (C₆H₁₀O₅) [42], which can then recombine to produce cellobiosan (C₁₂H₂₀O₁₀) [43]. At high temperature, cellobiosan itself undergoes dehydration processes. Although these transformations mainly yield carbohydrate derivatives, one important by‐product is the formation of significant amounts of water molecules. The processes occurring in the APCI ionization source can lead to the formation of protonated water molecules, which are then able to transfer a proton to compounds with higher PA than H₂O. As shown previously, if this reaction is sufficiently exothermic, proton transfer can be followed by dissociation processes that, in the case of these carbohydrate derivatives, result in the elimination of water molecules, as observed in Figure 4. Interestingly, when extending the m/z range up to 1000, protonated tetramers and pentamers of anhydro‐oligosaccharides were also observed. However, in contrast to smaller anhydro‐oligosaccharides, these higher oligomers showed limited dehydration following protonation. This behavior may result from a smaller difference in PA between higher oligomers and water, combined with the ability of larger oligomers to dissipate the excess energy generated during exothermic protonation through a larger number of degrees of freedom.

It therefore appears that, even when present only in small amounts in the APCI source, water molecules may act as the main driving force for ion formation in the open‐DIP‐APCI analysis of glucose and more generally of carbohydrates. To study the ability of water molecules formed during the thermal degradation of glucose under the DIP‐APCI(+) conditions used to induce the formation of protonated species, the behavior of glucose‐d5 was compared with that of non‐deuterated glucose. Both experiments were performed on the same day, within a few minutes of each other. Non‐deuterated glucose was analyzed first, followed by glucose‐d5. A similar thermogram was obtained for both experiments (Figure S7). The TIC remained approximately constant initially and then increased sharply once the temperature applied to the sample became sufficiently high to induce the thermal degradation of glucose. Prior to this sharp increase, the TIC was mainly associated with the desorption of surface contaminants from the capillary, notably fatty acids such as palmitic and stearic acids. The mass spectrum obtained for non‐deuterated glucose was identical to that reported in Figure 4, whereas that obtained for glucose‐d5 showed a broad distribution of deuterium incorporation in the ions formed during thermal degradation. As previously mentioned, the experimental conditions used did not allow observation of protonated (or deuterated) glucose. Nevertheless, the behavior of ions associated with palmitic and stearic acid contaminants is of particular interest, especially those observed at nominal m/z 258 and 286, respectively. At these m/z values, both the ¹³C isotopic contribution of the [M + H]⁺ ions and the [M + D]⁺ ions may be observed. At temperatures below those required to induce glucose degradation, only the ¹³C isotopic contribution of the [M + H]⁺ ions was observed, regardless of the sample (non‐deuterated glucose or glucose‐d5). This was also the case at higher temperatures when non‐deuterated glucose was analyzed. In contrast, at the same temperature, additional contributions were observed at m/z 258.25378 and 286.28508, corresponding to the [M + D]⁺ ions of palmitic and stearic acids, respectively.

These results clearly indicate that water or deuterated water molecules generated during the thermal degradation of glucose or glucose‐d5 can act as proton or deuteron donors in APCI(+) to enable ionization of the evolved compounds.

The open‐DIP‐APCI(+) mass spectrum of glucose, Figure 4, also reveals the presence of ions that can be attributed to ammonium adducts. This is particularly the case for the ions observed at m/z 180.0867 and 198.0972, which may correspond to the ammonium adducts of levoglucosan or its isomers (mono‐dehydrated glucose) and glucose, respectively. Interestingly, no ammonium adducts were observed for the di‐ and tri‐dehydrated derivatives of glucose. NH₃ may be used in chemical ionization (CI) and promotes the formation of ammonium adduct ions [44, 45, 46]. As previously mentioned, ammonia has been employed to assist ionization in dAPCI experiments [24, 31, 47]. It is important to note that neither ammonium hydroxide nor ammonium salts had been introduced into the APCI ionization source during the hours preceding the open‐DIP‐APCI analysis of glucose. This observation suggests that the formation of ammonium adducts with carbohydrate species can occur readily, even in the presence of trace amounts of ammonia, and may suppress the dissociation of the resulting ions due to the high PA of NH₃, which reduces the exothermicity of the ionization process (ammonium adduct formation vs. protonation).

3.3. Direct Infusion and Open‐DIP dAPCI(+) of Carbohydrates

The use of dopants in APCI(+) is well established, and NH₃ as a dopant has yielded interesting results through the formation of stable [M + NH₄]⁺ adducts rather than promoting proton transfers [24, 31, 45, 46, 47]. Indeed, compounds with a PA lower than that of ammonia preferentially form the adduct, since the H⁺ transfer from ammonia to the analyte is not thermodynamically favored [45]. Optimal adduct yields are obtained when the PA of the analyte closely matches that of ammonia, resulting in a stable [M··H⁺··NH₃] intermediate [44]. The PA of the main carbohydrates ranges from 750 to 860 kJ·mol⁻¹ [20, 21, 40, 48], which is close to that of NH₃ (853.6 kJ·mol⁻¹). Consequently, ammonia is well suited for conducting dAPCI(+) measurement of carbohydrates [39].

However, ammonia is also known to exhibit selective reactivity with different functional groups. It forms imines when associated with carbonyl compounds in both the liquid and gas phases [49, 50], and reacts with carboxylic acids to form amides. Nucleophilic substitution (S_N) of alcohols has also been demonstrated under ammonia atmosphere in CI sources [51]. This reactivity complicates data interpretation, as ammonium adducts cannot be easily distinguished from imines, amides, or S_N products without performing MS/MS experiments. To evaluate the ability of ammonia to react with the compounds produced by biomass pyrolysis, particularly its potential to form imines in the ion source with carbonyl‐containing species, as previously demonstrated in ESI(+) ionization [47], the behavior of vanillin was investigated. Vanillin is a well‐known compound produced during the pyrolysis of lignocellulosic biomass through the degradation of lignin [52, 53]. Vanillin was introduced into the DIP capillary under pyrolysis conditions. The dopant was delivered to the source via the DIP capillary (Figure S2), providing ammonia in the heated environment of the DIP source at 450°C. Under these conditions, gas phase formation of imines was found to be negligible. Indeed, the abundance of the imine‐linked peak was less than 1% of the vanillin peak, despite the higher PA of the imine (Figure S8).

To evaluate the ability of dAPCI(+) to limit post‐ionization dissociation processes of carbohydrates and their derivatives, an equimolar mixture of glucose, levoglucosan, and levoglucosenone was investigated using APCI(+) and open‐DIP‐APCI(+), both with and without ammonia as a dopant. The individual mass spectra of glucose, levoglucosan, and levoglucosenone acquired in the presence of ammonia are presented in Figure S9. The corresponding mass spectra acquired from the mixture under the various analytical conditions investigated are shown in Figure S10. The APCI ionization conditions employed were strictly identical, particularly with respect to the dry gas flow rate and temperature (2.5 L·min⁻¹, 220°C), as well as the APCI heater, which was set to 250°C. Direct infusion of mixed samples (10⁻⁴ mol·L⁻¹ each) was performed with methanol as a solvent. For open‐DIP experiments, 1 μL of a concentrated mixed (10⁻³ mol·L⁻¹ each) solution was introduced into the capillary, which was then inserted into the DIP‐APCI source after complete evaporation of methanol. The dAPCI experiments were carried out using the experimental setup described in Figure S2, with the dopant flow being introduced by the source gas exhaust to further prevent cross‐reactions. The results are presented as van Krevelen diagrams in Figure 5.

FIGURE 5.

van Krevelen diagrams of an equimolar (10⁻⁴ mol·L⁻¹ each) glucose (pink arrow and circle), levoglucosan (blue circle), and levoglucosenone (orange circle) methanolic solution obtained by (a) APCI(+), (b) dAPCI(+) with NH₃, (c) open‐DIP APCI(+), and (d) open‐DIP dAPCI(+) with NH₃. C₅H₄O₂ is highlighted with red circles. Dot size is proportional to relative abundance. Purple dots indicate neutrals associated with [M + H]⁺ ions, while green dots indicate neutrals associated with [M + NH₄]⁺ ions.

As previously described, the use of a low PA solvent results in an excess of internal energy in protonated analytes, which ultimately leads to dissociation of these ions. In this context, several unforeseen compounds were observed, the most intense being attributed to C₅H₄O₂ and putatively assigned to furfural or its isomers.

The relative intensity of peaks corresponding to protonated or ammoniated standards as well as those of peaks related to C₅H₄O₂ has been reported in Figure 6 for each method used.

FIGURE 6.

Relative intensities of the signals corresponding to the three standard compounds and to C₅H₄O₂, tentatively assigned to furfural, normalized to the sum of the total intensity of all attributed signals. Bar colors indicate the analytical technique used: APCI (red), dAPCI (blue), DIP‐APCI (green), and DIP‐dAPCI (yellow). Both [M + H]⁺ and [M + NH₄]⁺ were included in the intensity calculations.

Upon introduction of NH₃ introduced into the ion source near the corona needle, the targeted ions exhibited different relative intensities. Overall, ion intensities slightly decreased (Figure 5b,d), likely as a result of the introducing a compound with high PA and modifying the gas flows within the source. It is also worth noting that the relative abundance of the protonated molecule compared with that of the ammonium adduct strongly depends on the difference in PA between ammonia and the analyte (Table 1). The smaller the difference in PA, the higher the relative abundance of the [M + NH₄]⁺ ion. Accordingly, in APCI(+), the ammonium adduct is readily observed for glucose and levoglucosan but is almost absent for levoglucosenone, which has a substantially lower PA than ammonia. In the case of glucose, whose PA is the closest to that of ammonia, this trend is partially offset by its limited thermal stability, particularly under DIP‐dAPCI(+) conditions (Figure S10). However, this also offers significant advantages for carbohydrates ionization. The formation of [M + NH₄]⁺ ammonium adducts prevents complete glucose degradation during dAPCI ionization, while the intensity of levoglucosan increases despite the lower degree of glucose degradation (Figure 6). In contrast, the signal relative to C₅H₄O₂ compound decreased significantly. Similar trends were observed for dopant effects in open DIP‐APCI, with a marked decrease in the relative intensity of ions associated with furfural and its isomers, as well as levoglucosenone, while levoglucosan intensity increased. However, no ions corresponding to protonated or ammonium‐adducted glucose were observed in DIP‐dAPCI, suggesting that stabilization of ionized glucose species requires interaction with a solvent molecule.

Interestingly, in DIP‐APCI, cellobiosan isomers (C₁₂H₂₀O₁₀) were formed. These reactions appear to be limited under direct infusion conditions, likely due to the high abundance of solvent molecules in the gas phase, which both lowers the effective temperature and decreases the likelihood of reactions between species derived from the thermal degradation of glucose. In DIP‐APCI, several high‐mass ions were also formed, with oxygen counts from 6 to 10. These species likely result from the combination of carbohydrates or from the degradation of the levoglucosan dimer itself. Such proportions of O_6–10 compounds were not observed under DIP‐dAPCI ionization. The introduction of a dopant into the source appears to limit oligomer formation from glucose and its derivatives, possibly due to the high stability of [levoglucosan+NH₄]⁺ adduct ion.

Despite the inability to stabilize glucose‐linked ions under DIP‐APCI ionization, dopant addition reduced furfural formation. This is further evidence that water molecules, produced during dehydration of the standards and present in the ion source, participate in the ionization process and contribute to dissociation effects.

3.4. Application of Open‐DIP‐dAPCI(+) to the Study of Cellulose Pyrolysis Products

Cellulose pyrolysis generates a complex mixture of oxygenated products, including anhydrosugars, furans, and small aldehydes and ketones [54]. Understanding the distribution and reactivity of these compounds is crucial for both fundamental studies of biomass thermochemistry and the development of valorization strategies. More specifically, levoglucosan is a well‐known marker of cellulose pyrolysis. Its quantification by GC–MS often leads to overestimation, as higher‐order anhydro‐oligosaccharides can degrade in the injector or chromatographic column [55]. In this context, open‐DIP‐dAPCI(+) offers a powerful approach to directly probe cellulose pyrolysis products, as it combines soft ionization with the ability to limit post‐ionization dissociation, allowing for the detection of intact carbohydrates and their dehydration derivatives. In the following, this method is applied to investigate the composition of cellulose pyrolysis products under controlled conditions.

Cellulose was introduced into the DIP capillary, which was positioned within the APCI II source heater maintained at 380°C, where the sample underwent pyrolysis. Experiments were performed with and without ammonia dopant, and the results are presented as van Krevelen diagrams in Figure 7.

FIGURE 7.

Van Krevelen diagrams obtained in the study of cellulose fast pyrolysis by open‐DIP‐APCI(+) at 380°C, without (left) and with (right) NH₃ dopant. Dot size is proportional to relative abundance. Purple dots indicate neutrals associated with [M + H]⁺ ions, while green dots indicate neutrals associated with [M + NH₄]⁺ ions. The corresponding spectra are shown in Figure S11.

Open‐DIP‐APCI(+) of cellulose mainly led to the observation of [M + H]⁺ ions, as previously observed by Voellinger et al. while ammonium [M + NH₄]⁺ adducts were nearly absent [18]. In contrast, in open‐DIP‐dAPCI(+), ammonium adducts accounted for approximately 80% of the intensity of identified ions. Significant differences were observed when the distribution of detected ions was considered in terms of their O/C and H/C ratios. By open‐DIP‐APCI, ions generally displayed lower O/C and H/C ratios compared with open‐DIP‐dAPCI(+), particularly for compounds aligning along the line with a slope of 2 associated with dehydration products of carbohydrates and anhydro‐oligosaccharides.

It is reasonable to consider, at first approximation, that the main neutral cellulose pyrolysis products are largely identical under both operating conditions, as trace amounts of ammonia in the nitrogen flow are expected to minimally affect primary cellulose pyrolysis processes. As previously demonstrated in glucose studies, the formation of ammonium adducts at the expense of protonated species reduces the exothermicity of the ionization process and thereby limits, or even suppresses, post‐ionization dissociation processes, particularly water‐loss reactions. Consequently, open‐DIP‐dAPCI provides a more faithful representation of the actual distribution of primary sugar derivatives produced by cellulose pyrolysis compared with open‐DIP‐APCI.

Interestingly, small carbohydrates are predominantly observed as ammonium adducts. For example, the intensity of the [M + NH₄]⁺ ion for cellobiosan is four times higher than that of the [M + H]⁺ ion in DIP‐dAPCI(+). Larger carbohydrates, however, exhibit a more balanced distribution between protonated and ammonium‐adduct forms [20]. This can be explained by the lower propensity of larger oligosaccharides to undergo dissociation after protonation, as the relative excess internal energy from the exothermic proton‐transfer reaction is distributed over a larger number of vibrational modes.

Another noteworthy observation concerns larger oligomers, whose relative intensities are lower in dAPCI. This cannot be attributed to their dissociation, as the ionization conditions are softer than in APCI. It is well established that, during cellulose pyrolysis, larger anhydro‐oligosaccharides (beyond trimers) are formed via secondary pyrolysis processes involving primary pyrolysis products, particularly levoglucosan [56, 57]. These secondary reactions occur over timescales of several seconds following primary pyrolysis event. In the open‐DIP setup, the residence time in the DIP capillary is well below 1 s, and ammonium ion attachment is very rapid. Therefore, NH₄⁺ attachment effectively suppresses or at least limits the formation of secondary pyrolysis products, yielding a more specific representation of the distribution of primary pyrolysis products, as previously demonstrated for glucose oligomers formation in DIP‐dAPCI.

4. Conclusion

Glucose ionization processes in APCI(+) were investigated under conditions with or without solvents and dopants. The impact of PA differences between the solvent and glucose was demonstrated, as increasing this difference enhances the exothermicity of proton transfers, leading to the dissociation of poorly stable ions via dehydration.

In the absence of solvent, as in DIP‐APCI(+), carbohydrates remain exposed to high temperatures and ambient gases within the ion source, leading to extensive dehydration. Water generated during glucose dehydration may further promote post‐ionization dissociation via proton attachment, as suggested by DIP‐APCI(+) experiments using glucose‐d5.

The addition of ammonia as a dopant in the ion source, combined with the study of an equimolar mixture of glucose‐derived compounds, demonstrated its ability to mitigate various degradation pathways in both DI‐APCI and open‐DIP‐APCI. Signals attributed to furfural, a degradation product that can constitute the base peak in absence of dopant, and its isomers showed a substantial decrease in relative intensity in presence of ammonia, despite the similar PA values of ammonia and furfural. However, this approach was insufficient to prevent glucose dehydration under DIP‐dAPCI(+) conditions.

In cellulose pyrolysis experiments, DIP‐dAPCI(+) results are consistent with the predominant formation of cellobiosan and levoglucosan, with limited degradation. Larger anhydro‐oligosaccharides were not observed in significant amounts; likely because their formation is suppressed by the rapid NH₄ ⁺ attachment, which inhibits secondary pyrolysis reactions.

These findings are expected to be of significant interest in carbohydrates analysis, particularly in the context of biomass pyrolysis, and provide valuable insight to APCI(+) users regarding the limitations associated with analysis with or without solvents and dopants.

Author Contributions

Nathan Traullé: conceptualization, investigation, writing – original draft, data curation, validation, methodology. Théo Voellinger: investigation, software, validation, writing – review and editing, data curation, methodology. Jasmine Hertzog: writing – review and editing, validation, methodology. Sébastien Schramm: methodology, validation, writing – review and editing. Vincent Carré: conceptualization, methodology, validation, writing – review and editing, supervision. Frédéric Aubriet: conceptualization, funding acquisition, methodology, validation, writing – review and editing, project administration, supervision, resources.

Funding

This work was supported by Agence Nationale de la Recherche, REALYTIC project (22‐CE29‐0011).

Supporting information

Figure S1: ESI(+) spectra of a 10⁻³ M solution of prepared deuterated glucose in CD₃OD.

Data S1: Determination of glucose deuterium enrichment rate.

Figure S2: Experimental setup used for the introduction of ammonia as a dopant into the nitrogen capillary gas flow and into the APCI ion source. The nitrogen flow rate was set to 3 mL·min−1 in open‐DIP dAPCI and to 60 mL·min−1 in dAPCI.

Figure S3: Van Krevelen representation of the most common thermal degradation products of glucose. The grey line represents the so‐called “sugar line”, characteristic of successive H2O losses.

Figure S4: FT‐ICR MS/MS spectra of the ions at m/z (a) 163.06010, (b) 145.04952, and (c) 127.03897 observed during APCI(+) analysis of a glucose solution in EtOH. Argon was used as the collision gas, and the center‐of‐mass collision energy (ECOM) was set to 0.9 eV.

Figure S5: Energy‐resolved breakdown curves derived from the FT‐ICR MS/MS spectra of the ions at m/z (a) 163.06010, (b) 145.04953, and (c) 127.03897 formed during APCI(+) analysis of a glucose solution in EtOH. Symbol shapes and colors correspond to the following m/z signals: (blue dots) 163.06010, (green squares) 145.04953, (orange diamonds) 127.03897, (black circles) 109.02841, (purple crosses) 99.04406, (pink squares) 117.05462, and (red triangles) 97.02841. The collision energy is expressed as the center‐of‐mass collision energy (ECOM), which corresponds to the fraction of the precursor ion kinetic energy that is transferred into internal energy during the collision with the collision gas. This internal energy is available to induce fragmentation of the precursor ion. Lines are shown as guides for the eye.

Figure S6: APCI(+) FT‐ICR mass spectrum obtained after the mere isolation (without additional activation) of the [Glucose − H2O + EtOH + H]+ ion (m/z 209.10158). Red asterisks indicate ions resulting from the dissociation of a co‐isolated ion at m/z 209.12025. Blue “1” corresponds to m/z 173.08080 ([Glucose − 3H2O + EtOH + H]+), and blue “2” corresponds to m/z 191.09138 ([Glucose − H2O + EtOH + H]+).

Figure S7: Thermogram (blue line) obtained during the desorption/pyrolysis of non‐deuterated glucose and glucose‐d5. The selected m/z ranges correspond to the 13C isotopic contribution of the [M + H]+ and [M + D]+ ions of palmitic and stearic acids (nominal m/z 258 and 286, respectively), observed at different time points along the thermogram during open‐DIP analysis of non‐deuterated glucose (red) and glucose‐d5 (green). Arrows indicate the time points selected for comparison.

Figure S8: NH3 open‐DIP dAPCI(+) FT‐ICR mass spectrum of vanillin at 450°C.

Figure S9: APCI(+) FT‐ICR MS spectra acquired from 10−4 M methanolic solutions of (a) glucose, (b) levoglucosan, and (c) levoglucosenone in the presence of ammonia. Background ions were removed for clarity. Blue peaks correspond to protonated species, green peaks to ammonium adducts, and red peaks to protonated methanol adducts.

Figure S10: Mass spectra obtained in the study of an equimolar (10−4 mol·L−1 each) glucose, levoglucosan, and levoglucosenone methanolic solution obtained by (top left) APCI(+), (top right) dAPCI(+) with NH3, (bottom left) open‐DIP APCI(+), and (bottom right) open‐DIP dAPCI(+) with NH3. Ammonium adducts are in green color. The peak marked with a green asterisk corresponds to [Levoglucosenone + NH4]+.

Figure S11: Open‐DIP‐APCI(+) FT‐ICR mass spectra acquired during cellulose pyrolysis at 380°C: (left) in the absence of ammonia and (right) in the presence of ammonia.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.