Keeping vitamin C amorphous in freeze-dried polymer dispersions: hydrogen bonding, Tg and water uptake compared

Background and motivation

Vitamin C is a widely used nutrient in fortified foods and supplements, and one of the least stable vitamins; its stability often limits shelf life (Hung et al. 2007). It normally exists as a crystal. Earlier work from the Mauer group (West and Mauer 2013) had noticed that VitC dried together with anthocyanin extracts gave X-ray amorphous powders and hypothesized that the pectin in those extracts was responsible C2. In pharmaceutics and in work on phytochemicals such as curcumin and resveratrol, amorphous forms are made on purpose to raise solubility and bioavailability (Leuner and Dressman 2000; Li et al. 2013; Wegiel et al. 2013, 2014). Amorphous solids, however, are thermodynamically less stable than crystals and tend to crystallize during storage and processing, so polymers are added to keep them amorphous C1.

Which property of a polymer makes it a good crystallization inhibitor is an open question, and the introduction lists three candidates. The first is the glass transition temperature: polymers are usually chosen with a Tg above that of the amorphous compound, and the resulting rise in dispersion Tg (antiplasticization) has been correlated with crystallization inhibition (Craig et al. 1999; Khougaz and Clas 2000), although not at low polymer contents in some systems (Kestur et al. 2011; Matsumoto and Zografi 1999) C3. The second is hydrogen bonding between polymer and compound, which "has also been suggested" as the primary inhibition mechanism (Matsumoto and Zografi 1999; Taylor and Zografi 1997; Wegiel et al. 2013) C4. The third is hygroscopicity: water is a plasticizer that lowers Tg (Roos and Karel 1991b) and also disrupts compound–polymer hydrogen bonds (Rumondor et al. 2010) C5.

VitC is a demanding test case. It crystallizes so readily that freeze-drying it alone gives crystals, and spray drying without additives has also failed to make it amorphous (Islam et al. 2010) R1·R27. The paper therefore sets two goals: first, to make amorphous VitC dispersions with polymers by freeze-drying (lyophilization) and follow their physical stability at controlled temperature and RH; second, to compare three polymer properties, the ability to hydrogen bond with VitC, Tg and hygroscopicity, and find which one most influences crystallization inhibition (Section 1) A1·A2·A3·A4·A5·A6.

Six polymers were chosen to differ in these properties. Four were food-grade pectins with different degrees of esterification (DE): pectin from citrus peel (CT, DE about 61%), pectin from apple (AP, DE 70–75%), esterified pectin from citrus (EST, DE at least 85%) and an esterified potassium-salt pectin from citrus (K, DE 55–70%, 4.1–6.4% potassium) M2·M3·M4·M5. Two were synthetic polymers common in pharmaceutical amorphous solid dispersions: polyvinylpyrrolidone (PVP, 40,000 MW) and polyacrylic acid (PAA, 450,000 MW), chosen for their different Tg and hydrogen-bonding groups M6·M7.

Background concepts

Amorphous versus crystalline, and how it is detected. A crystal gives sharp X-ray powder diffraction (XRPD) peaks; an amorphous solid gives only broad halos. In this paper a sample is called crystalline when its XRPD pattern shows "sharp, distinguishable peaks above the noise level", and amorphous otherwise (Section 2.4) P6. "XRPD amorphous" therefore means that no crystal is detected, not that none is present; the amount of crystal was not quantified anywhere in the paper X15.

Hydrogen bonds seen by infrared spectroscopy. In a hydrogen bond a donor group (here an OH or a COOH) shares its hydrogen with an acceptor (a C=O, an ether oxygen or another OH). Both partners change their vibrations: the O–H stretching band and the C=O stretching band move, usually to lower wavenumber, and change shape. The paper reads a shift to lower wavenumber as "stronger and/or more extensive intermolecular interactions" (Section 3.2.1, citing Taylor and Zografi 1997 and Wegiel et al. 2013) C15. Crystalline VitC has four sharp O–H bands at 3215, 3315, 3410 and 3526 cm⁻¹, which the paper assigns to steric hindrance by the ring in the crystal structure (Nath et al. 2005) R8·C17. When these four bands disappear and a single broad band takes their place, the VitC–VitC contacts of the crystal are gone C18.

The shifts in Table 3 are defined in its footnotes: for the hydroxyl region the shift is the dispersion's peak position minus that of the freeze-dried (FD) polymer; for the two carbonyl regions it is the dispersion's position minus that of freeze-dried VitC E5. A negative value therefore means a band that moved to lower wavenumber.

Which pairs can hydrogen bond. The paper ranks donors and acceptors in Table 2. Acceptor strength comes from the pK_BHX hydrogen-bond basicity scale (Laurence et al. 2009), with classes weak below 0.5, medium 0.5–1.8, strong 1.8–3.0 and very strong above 3.0; no value existed for the carboxylic acid acceptor, so it was taken as equal to an ester carbonyl M1·M2·M6·M7. VitC has a strong hydroxyl donor and medium hydroxyl and carbonyl acceptors; its ring ether is not involved. Pectin has strong hydroxyl and carboxylic acid donors and medium acceptors (hydroxyl, ester, carboxylic acid, ether). PVP has no donor at all but a strong amide carbonyl acceptor. PAA has a strong carboxylic acid donor and a medium carboxylic acid acceptor M1·M2·M6·M7. The working hypothesis is that "polymers with stronger hydrogen bond donors (HBD) and weaker hydrogen bond acceptors (HBA) than the target compound, or vice versa" form the best dispersions, because the strongest donor pairs with the strongest acceptor C14. VitC is a weak acid with pK_a1 4.15; all solutions before freeze-drying had a pH of 2.68–3.17, so VitC was taken to be neutral, a strong donor and a medium acceptor R7·C16. (The text prints five pH values for six solutions, so one value is missing X2.)

Degree of esterification. Pectin is mainly a chain of galacturonic acid units. The DE is the share of these acid groups that carry a methyl ester. A lower DE leaves more free carboxylic acid groups, which are strong donors; this is the property the authors use to explain the difference between CT and EST C24.

Glass transition and the Gordon–Taylor equation. Below its Tg an amorphous solid is a glass with low molecular mobility; above Tg it becomes a rubber, mobility rises and crystallization is expected to speed up C11·C32. The paper measures the onset Tg by differential scanning calorimetry (DSC), and estimates the Tg of a mixture with the Gordon–Taylor equation (Eq. 2):

Here $w_1$ and $w_2$ are the weight fractions of components 1 and 2, $T_{g1}$ and $T_{g2}$ their glass transition temperatures in kelvin, and $k$ a constant, described as the ratio of the free volumes of the two components E2. In plain terms: the Tg of the blend is a weighted average of the two component Tgs, and $k$ sets how much weight component 2 carries. The constant can be fitted to data or estimated from the densities ${\rho }_1$, ${\rho }_2$ with the Simha–Boyer rule (Eq. 3):

In plain terms: if the densities are known, the constant (printed here as $K$) follows from the densities and Tgs of the two components. The paper does not use this route; it fits $k$ instead, and no densities are printed E3. The Tg of water is taken as 136 K (Haque and Roos 2008) M12. A third, empirical relation appears in Section 3.4: the Tg of a compound is "typically 2/3 of the Tm measured in Kelvin", $T_g\approx \frac{2}{3}{T}_m$, where $T_m$ is the melting temperature of the crystal E4.

Moisture sorption and what a weight loss means. A dynamic vapor sorption (DVS) instrument raises RH step by step and records the equilibrium weight gain. An amorphous solid usually sorbs more water than its crystal. If it crystallizes during the run, the crystal expels water and the profile shows a weight loss at that RH (West and Mauer 2013) C27. To see whether making a dispersion changes water uptake, the paper compares each 50:50 dispersion with the uptake predicted if VitC and polymer sorbed independently (Eq. 1):

In plain terms: for a 50:50 blend the expected water uptake (in % weight change) is the average of the uptakes of the freeze-dried polymer and freeze-dried VitC at the same RH E1. Uptake above this line is read as amorphous VitC taking up extra water C28.

Methods in detail

The study combines four kinds of measurement on the same freeze-dried materials: XRPD to tell amorphous from crystalline, attenuated total reflectance FTIR to look for hydrogen bonds, DVS and desiccator weighing for water uptake, and DSC for Tg. A Gordon–Taylor fit adds estimated Tgs for amorphous VitC and the pectins. The flow below shows how the parts fit together.

Contribution 1: making amorphous VitC by freeze-drying and following it in storage

The problem. Spin coating and rotary evaporation need organic solvents in which VitC dissolves poorly, and aqueous methods that use heat may degrade VitC. Freeze-drying needs no heat, VitC is very soluble in water, and many food polymers are water soluble, so it was the method of choice (Section 3.1) C7.

What they did. For the 50:50 dispersions, 0.25 g VitC and 0.25 g polymer were dissolved in 45 mL water in a 50 mL tube; for the ratio series, 0.1 g of solids was dissolved in 9 mL water, from 0 to 100% polymer in 10% steps; VitC or polymer alone was dissolved at 0.5 g in 45 mL. The solutions were mixed until clear, frozen at −20 °C for at least 24 h and freeze-dried for at least 90 h, then sealed in 20 mL vials P1·P2·P3. The methods do not list which polymers went into the ratio series; Table 3 reports CT and EST at every ratio and PVP from 40%, and the text reports 30% CT, EST and PVP in storage P2. Storage used desiccators over Drierite (0% RH) and saturated potassium acetate, magnesium sulfate and sodium chloride (23, 54 and 75% RH, values given for 25 °C) at 25 °C and in incubators at 40 °C, for 130 days; XRPD (Cu Kα, 5–35° 2θ, 4°/min, 0.04° step) was run at day 0 and weekly P5·P6·M8·M9·M10·M11. The RH that the salts gave at 40 °C is not stated X14.

Why it works. The design crosses six polymers with four RHs and two temperatures, so the effect of each factor on whether VitC stays amorphous can be read from a simple yes/no table. How it differs from earlier work. The same polymer-screening logic had been applied to resveratrol and curcumin (Wegiel et al. 2013, 2014); here it is applied to a water-soluble, fast-crystallizing vitamin, with food pectins as the main carriers and freeze-drying instead of solvent evaporation.

Contribution 2: reading VitC–polymer hydrogen bonds by FTIR

The problem. Water also shifts the bands that report hydrogen bonding, so interactions must be measured on dry samples. What they did. ATR-FTIR spectra (4000–400 cm⁻¹, 128 scans) were recorded on samples stored at 0% RH, under a nitrogen purge, and the O–H (3800–3000 cm⁻¹) and C=O (1900–1500 cm⁻¹) regions were followed P7. PVP, the most hygroscopic polymer, served as an internal moisture check: its carbonyl band appeared at 1678 cm⁻¹, close to 1680 cm⁻¹ for dry PVP and far from 1652 cm⁻¹ for PVP at 94% RH (Taylor, Langkilde and Zografi 2001) R14. The authors write that they "cannot ensure that no moisture was sorbed", but that it "seems" little was, and attribute the shifts to VitC–polymer interactions C6. To see what water does, the 50:50 VitC:EST dispersion was also measured after 7 days at 54 and 75% RH and 25 °C P7·R10.

Why it works. Crystalline VitC has a distinctive set of four sharp O–H bands, so their disappearance is an easy marker that VitC is no longer in the crystal, and band shifts compare the strength of the new contacts across polymers C18·C15. Difference from earlier work. The approach follows Taylor and Zografi (1997); new here is the pairing with Table 2's donor and acceptor strengths to predict which polymer should interact most.

Contribution 3: water uptake and Tg as rival explanations

Water uptake. DVS used samples of 900–1100 mg in aluminum pans, held at 5% RH for 3 h and then stepped from 5 to 90% RH at 25 °C; a step ended when the weight changed by less than 0.01% in 15 min (minimum 50 min, maximum 10 h) P8. Every 50:50 dispersion was compared with a 50:50 physical mixture of crystalline VitC and the polymer, and with the additive prediction of Eq. 1 P4·E1. The paper describes 5% RH increments, but the plotted profiles have 16 points spaced about 5.6% RH apart (about 5, 11, 16, 22 … 89.5% RH), so RH thresholds read from the figures fall on that grid X16. Separately, samples in the storage desiccators were weighed weekly until the weight changed by less than 0.01 g, which took 4 to 7 weeks; these values are not printed P9.

Tg. After storage, 2.5–4 mg of each sample was sealed in a hermetic pan, heated to 120 °C, cooled to −80 °C and reheated to 100 °C at 10 °C/min; the onset Tg was taken from the second heating P10. Pectin Tgs could not be detected by DSC or modulated DSC, which the authors attribute to heterogeneous DE, and no amorphous VitC could be made without polymer, so its Tg could not be measured either M2·M1·C33. To estimate it, the authors fitted $k$ in Eq. 2 to Tgs of 50:50 VitC–PVP dispersions at different water contents, obtaining $k$ = 4.287, and then used this $k$ with the dry 50:50 VitC–PVP Tg to back-calculate the Tg of VitC P11·R21·R22. The data for this fit are cited as "Table 5", which is not printed in the article X4. The pectin Tgs were then calculated from the 50:50 VitC–pectin Tgs, assuming VitC at −7.09 °C R24.

Experimental analysis

Day 0: pectins and PVP make VitC amorphous, PAA does not

Freeze-dried VitC without polymer was substantially crystalline, with an XRPD pattern identical to the starting material, so it crystallized back to the same form (Figure 1) R1·C9. The authors therefore class VitC as a Class I compound, one that crystallizes rapidly, using the scheme of Van Eerdenbrugh, Baird and Taylor (2010) C8. All six polymers were XRPD amorphous before and after freeze-drying R2. At day 0 the 50:50 dispersions with CT, AP, EST, K and PVP showed no peaks, while the PAA dispersion showed peaks of crystalline VitC; the authors conclude that it "was not possible" to make an amorphous VitC dispersion with PAA under these conditions R3·C10.

FTIR agrees. In the pectin and PVP dispersions the four crystalline O–H bands of VitC were absent; in the PAA dispersion they were present (Figure 2a) R9·C18. Note by the AI writer (not discussed in the paper): in Figure 2a two traces carry the label "50:50 VitC:K" and none carries "PAA"; the upper of the two shows the crystalline bands that the text assigns to PAA, so it is most likely the PAA spectrum with a misprinted label X1.

Storage: humidity decides, not temperature or polymer

Every amorphous 50:50 dispersion stayed XRPD amorphous for the full 130 days at 0 and 23% RH, at both 25 and 40 °C. At 54 and 75% RH every one of them showed at least some crystallization within one week, again at both temperatures (Table 1) R4·R5. At 50% polymer the four pectins did not differ, and PVP behaved like them R5·C13. The authors conclude that "the rate of crystallization seemed to be influenced more by storage RH than storage temperature within the experimental space", as reported for resveratrol–PVP and felodipine–PVP dispersions C12·R26. The FTIR spectra of the EST dispersion after a week support the XRPD result: at 0% RH the crystalline O–H bands were still absent, at 54 and 75% RH they had started to reappear (Figure 2b) R10.

Note by the AI writer (not discussed in the paper): Table 1's caption says "after 7 days at 25 °C", but the table lists rows for 25 and 40 °C; the 7-day table and the 130-day statement in the text agree in pattern X3. Because no RH between 23 and 54% was tested, the stability limit can only be placed somewhere in that interval X14.

FTIR: the strongest interaction with PVP, the weakest with the potassium pectin

Table A. FTIR band positions and shifts (cm⁻¹) of the 50:50 dispersions at day 0, from Table 3. O–H shifts are relative to the freeze-dried polymer; C=O shifts relative to freeze-dried VitC (1678 and 1754 cm⁻¹) R11.f·R11.m·R11.t·R11.aa·R11.ae·R9.

At 50:50 the O–H band moved to lower wavenumber in the order PVP (−232 cm⁻¹) > EST (−96) > AP (−67) > CT (−36) > K (−19) (Table A) R11·C19. PVP also moved the VitC carbonyl band in region I by −23 cm⁻¹, which no pectin did R11.ae. The authors call the large PVP shifts "predictable": PVP has no donor of its own, so VitC's donors are free to bind its strong amide carbonyl acceptor (Kestur et al. 2011) C20. In the potassium pectin K the O–H shift was smaller than in CT, which has a similar DE, and a new peak appeared in the carbonyl region of the VitC:K spectrum only; the authors propose that the positively charged potassium "could interact with the VitC HBA and the pectin", competing with VitC–pectin contacts R13·C21. PAA showed no change in band shape or position from crystalline VitC, so the authors find "limited interactions" C25. They generalise that a stronger donor in VitC and a stronger acceptor in the polymer, or the reverse, "was essential" for the hydrogen bond C26.

Note by the AI writer (not discussed in the paper): the text lists crystalline VitC carbonyl bands at 1647 and 1751 cm⁻¹, while Table 3 uses 1678 and 1754 cm⁻¹ for freeze-dried VitC, which the text says is the same crystal form; the shifts are computed from the Table 3 values, and every printed shift reproduces from the printed positions X8·E5.

Pectin DE: CT needs less polymer, EST bonds more strongly once amorphous

Table B. The ratio series at the first polymer content where Table 3 lists a single O–H band (cm⁻¹) R11.e·R11.t·R11.ad.

In the ratio series the four crystalline O–H bands weakened as the pectin content rose and were replaced by a single broad band at a lower wavenumber than pure pectin, which the authors read as VitC–pectin hydrogen bonding (Figure 3) R12·C22. This happened from 40% CT but only from 50% EST: "More EST (50%) was needed than CT (40%) to disrupt VitC crystallization" (Table B) R11.e·R11.t·C22. Once the single band had formed, however, EST gave the larger shift (−96 against −42 cm⁻¹), "a stronger hydrogen bond interaction" C22.

The carbonyl region tells the other half. Between 20 and 40% CT both VitC carbonyl bands moved down (region I from 1678 to 1668, 1667 and 1670 cm⁻¹ at 20, 30 and 40%; region II from 1754 to 1753, 1751 and 1748 cm⁻¹), while EST at the same contents moved region I by only −3 to +7 cm⁻¹ R11.a·R11.c–e·R11.q–s. The authors attribute this to VitC's carbonyl accepting hydrogen bonds from CT's donors, which "further confirmed" that VitC–VitC contacts were being replaced by VitC–CT contacts as the crystalline O–H bands faded C23. Their explanation for the two regions is a hypothesis: CT, with a lower DE, carries more free carboxylic acid, whose OH is strongly polarized by the adjacent carbonyl and is a strong donor; these donors bind VitC's carbonyl, so less CT is needed to break up the crystal, but the same groups leave CT's other hydroxyl donors less available, giving the smaller O–H shift ("It is hypothesized") C24.

Storage of the ratio series is reported through one example: after 50 days at 25 °C and 23% RH, the 30% CT and 30% PVP dispersions were XRPD amorphous, while the 30% EST dispersion had partially crystallized R6·C13. From this and the FTIR data the authors conclude that lower-DE pectin is the better inhibitor at low polymer content C13·C39.

Note by the AI writer (not discussed in the paper): the summary sentence that "the pectin with lower DE had stronger interaction with VitC" rests on the carbonyl region; the O–H shifts point the other way (EST −96 and AP −67 against CT −36 cm⁻¹ at 50:50), and the CT carbonyl shifts at 20 and 30% come from dispersions that still contain crystalline VitC X17. Also, the lowest region I position between 20 and 40% CT is at 30% (1667 cm⁻¹), not at 40% as the text says; the 40% minimum holds for region II X10.

Water uptake: amorphous VitC sorbs extra water; K and PAA dispersions crystallize in the instrument

Among the pure materials at 25 °C, crystalline VitC took up almost no water below its deliquescence point of 98% RH; the pectins sorbed similar amounts to each other; PVP sorbed the most (about 54% weight gain at about 90% RH); PAA sorbed less than the pectins below about 60% RH and more above it (about 42% at 90% RH) (Figure 4a) R15·M1.

Table C. Weight gain (%) of 50:50 samples at the last point, about 89% RH, 25 °C, read from Figure 4b (approximate) R16.

Every pectin and PVP dispersion sorbed more water than its physical mixture, by about 6 to 13 percentage points at about 89% RH (Table C) R16·R17. The authors attribute this to amorphous VitC being more hygroscopic than crystalline VitC ("can be attributed") C31. For PAA, whose dispersion was already crystalline, dispersion and mixture differed little, which the authors note coincides with PAA's failure to prevent crystallization R16.f·C31. The pectin dispersions also sorbed more than the additive prediction at all RHs except where they lost weight; for PVP the difference was smaller, "most likely because PVP was more hygroscopic" R17·C28.

Note by the AI writer (not discussed in the paper): the plotted pectin predictions (about 17–18% at 89% RH) are not half of the pectin curves in Figure 4a (about 26–29%, which would give about 13–15%), while for PVP and PAA they are. The prediction uses freeze-dried polymers whose isotherms are not shown, and Figure 4a may show the polymers as received, so for the pectins the comparison with the physical mixture is the safer one X6.

Weight losses mark crystallization. The VitC:K dispersion lost weight above about 60% RH; XRPD on samples from the run showed no peaks at 50% RH and crystalline peaks at 72% RH, and Figure 5b is labelled "Rapid crystallization above 61%RH" R17·R18. The VitC:CT dispersion only levelled off between about 55 and 67% RH, and XRPD at 50 and 72% RH showed no crystals (Figure 5a) R18. The other pectin dispersions also levelled off between 50 and 70% RH without losing weight R17·C29. The authors link K's early crystallization to its weakest FTIR interaction ("consistent with") C29. The PAA dispersion lost weight above 40% RH; since it was already partly crystalline, the authors read this as "the likelihood of at least some initial amorphous VitC" recrystallizing R17·C30.

Tg: no ranking of stability

Table D. Onset Tg (°C) of the 50:50 dispersions after 35 days of storage at the stated temperature and RH, from Table 4; n.d. = not determined R19.

The pure polymers had onset Tgs of 157 ± 2 °C (PVP) and 130 ± 2 °C (PAA), in line with literature values M6·M7. In the dispersions, comparing Tg with the storage temperature shows both states among the samples that stayed amorphous (Table D): at 0% RH the pectin dispersions were glassy, except VitC–CT at 40 °C (Tg 31 °C); at 23% RH they were above Tg, except VitC–AP (28 °C) and VitC–EST (26 °C) at 25 °C; and the PVP dispersions were above Tg at both RHs and both temperatures R20. All of them stayed amorphous for 130 days, and at 23% RH the larger distance above Tg at 40 °C than at 25 °C made no difference. The authors conclude that "the Tg of the dispersion had limited effect on the physical stability of VitC" C34. (The text names the 23% RH exceptions as "CT–AP and CT–EST"; Table 4 shows they are the VitC–AP and VitC–EST dispersions X18.)

Two further comparisons point the same way. PAA has a Tg below that of PVP but above the estimated Tgs of CT and EST, yet its dispersion was the least stable; so "the Tg of polymers was found to have limited correlation with the stability", especially without intermolecular interactions, as found earlier for resveratrol (Wegiel et al. 2013) C40·R24·R26. And the VitC–CT dispersion had a lower Tg than VitC–K (29 against 42 °C dry at 25 °C), yet VitC–K crystallized during sorption where VitC–CT did not; the dispersion Tgs "were not correlated to their crystallization tendencies" R19.a·R19.d·R18·C41.

Table E. Estimates of the Tg of amorphous VitC (text p.874) R22·R23·E4.

The Gordon–Taylor route gave −7.09 °C for amorphous VitC, which the authors place between the literature DSC value of −17 ± 2 °C, obtained after heating VitC to its melt where "chemical degradation" is likely, and 10.95 °C from the two-thirds rule (Table E) R22·R23·C35. Note by the AI writer (not discussed in the paper): the two-thirds rule with Tm 190 °C gives 463.15 K × 2/3 = 308.8 K, or 35.6 °C, not 10.95 °C; the conclusion that −7.09 °C lies between the two estimates still holds X7. The text also describes the literature protocol as heating VitC "to slightly above its melting point, 100 °C", which does not match a Tm of 190 °C X7. Recomputing the Gordon–Taylor step reproduces −7.02 °C only when $k$ multiplies the VitC term and the dry Tg after storage at 25 °C (24 °C) is used; the paper does not say which component is which, and the same $k$ was fitted to water-plasticization data E2·X5. With VitC at −7.09 °C the authors estimate pectin Tgs of 83.5 (CT), 130.8 (AP), 79.5 (EST) and 207.3 °C (K), which they call similar to the 93.5–96.4 °C reported by Lai et al. (2000); another study reported about 37 °C for four pectins (Iijima et al. 2000) R24·R25·C36. The $k$ used for the pectins is not stated, and no single $k$ reproduces all four values from Table 4 X5.

Discussion

How the authors put the findings together. In Section 3.5 the authors weigh the three candidate properties against the outcomes. Hydrogen bonding wins: "a good agreement was found between the FTIR spectra and XRPD patterns", because every dispersion that showed VitC–polymer interactions was amorphous and stayed so at low RH, while the PAA dispersion, with "minimal or absent" interactions, was crystalline C38. Tg loses: dispersions above and below their Tg behaved alike, and the high-Tg PAA was the worst carrier C34·C40·C41. Hygroscopicity is placed in between: "not a key element", since PAA sorbed the least water below 50% RH and still failed, but "still found to influence" stability; PVP's much higher water uptake "was thought to be the explanation" for why its stronger hydrogen bonding gave no advantage over pectin at 54 and 75% RH C42·C43. The chain they propose can be drawn as follows.

The conclusion drawn is a recommendation: to inhibit the "self-assembly" of compounds with a high crystallization tendency such as VitC, "selection of a polymer based on its ability to form intermolecular interactions with the target compound is an important aspect" (Section 4) C48. The abstract adds a ranking of crystallization inhibitor properties, PVP > pectin with lower DE > pectin with higher DE >> PAA C50, and the conclusions close by noting that the evidence for amorphous VitC "opens the door to further investigations of the properties of amorphous forms of this vitamin" C49.

Where the approach applies. The results concern 50:50 (and, for three polymers, 30–100%) freeze-dried VitC dispersions stored at 25 and 40 °C between 0 and 75% RH for up to 130 days A1·A2·A3·A4·A5. Within that space the picture is internally consistent across XRPD, FTIR and sorption. The comparison with drug systems is direct: as for resveratrol–PVP and felodipine–PVP, RH dominated over temperature, and as for resveratrol, polymer Tg did not predict which polymer inhibited best R26.

What it suggests for formulation. Note by the AI writer (not discussed in the paper): two practical points follow. First, for a fast crystallizer the polymer has to bind the compound; a high Tg without a complementary hydrogen-bonding group (PAA) did not prevent crystallization even during drying. Second, binding sets whether the amorphous form can be made, but water sets how long it lasts: no polymer here protected VitC for a week at 54% RH, so an amorphous VitC ingredient would need dry storage or moisture-protective packaging, and a polymer that bonds to VitC while sorbing less water than PVP would be the logical next candidate. Chemical stability of the vitamin, which is what limits shelf life in practice, was not measured.

Limitations

Limits the authors state. The authors acknowledge that they "cannot ensure that no moisture was sorbed" during FTIR handling C6; that heating VitC to measure its Tg "will likely cause chemical degradation" C35; that the PAA dispersion was at least partially crystalline at the start of its sorption run C30; and, in the introduction to Section 3.5, that "it is unclear which of these factors plays the biggest role" and that the literature conflicts C37. The pectin Tgs could not be measured and are estimates C33·R24.

The AI writer's own judgement (1): what the hydrogen-bonding conclusion rests on. The contrast that carries the main conclusion is between PAA and the other five polymers. PAA is the only polymer without hydrogen-bond evidence, but it also differs in Tg, molecular weight (450,000 against 40,000 for PVP) and water uptake, and its FTIR spectrum was recorded on a dispersion that was already crystalline, so a missing interaction could be a consequence of crystallization as well as a cause X13. Among the five polymers that did bond, stability at 50:50 was identical, so the strength of bonding (from −19 to −232 cm⁻¹) did not translate into any measured difference in storage stability R4·R5·R11. The ranking PVP > lower-DE pectin is therefore supported by FTIR shifts only: at 30% polymer PVP and CT both stayed amorphous, and K, which crystallized first in sorption, is not placed in the ranking X11. The claim that lower-DE pectin interacts more strongly rests on the carbonyl region, while the O–H shifts are larger for EST and AP than for CT X17, and the DE effect on storage stability is shown by one printed example X12.

The AI writer's own judgement (2): what the data can and cannot resolve. Crystallinity was recorded as present or absent, without quantification, and no replicate numbers or uncertainties are given for storage, FTIR or the dispersion Tgs; the 0% RH, 25 °C PVP dispersion sits only 1 °C below its storage temperature X15. The RH at 40 °C is not stated and no RH between 23 and 54% was tested, so the humidity limit is bracketed but not located X14. The Tg estimates rest on an unprinted "Table 5", an unstated assignment of components in the Gordon–Taylor equation and a $k$ fitted to water-plasticization data; the comparison value from the two-thirds rule is misprinted (10.95 °C, recomputed 35.6 °C) X4·X5·X7. Smaller slips do not change the conclusions: the PAA spectrum in Figure 2a appears under a "K" label, the pH list is one value short, the Table 1 caption omits 40 °C, the VitC carbonyl positions differ between text and Table 3, Table 3 lists no region-I band for the neat pectins although Figure 3 labels one at 1616–1624 cm⁻¹, and the plotted RH steps are about 5.6% rather than 5% X1·X2·X3·X8·X9·X16. On balance, the formation and storage results are clear, and the attribution to hydrogen bonding is plausible but rests on one non-bonding polymer.

Conclusion

Freeze-drying alone cannot make vitamin C amorphous, but freeze-drying it 50:50 with pectins or PVP can, and these dispersions stay amorphous for at least 130 days if kept at 23% RH or below, at 25 or 40 °C R1·R3·R5·C44. At 54% RH and above they crystallize within a week whatever the polymer R5. FTIR shows that the polymers that make VitC amorphous hydrogen bond with it, most strongly PVP, and that PAA, whose acid groups mirror VitC's rather than complement them, does not R11·C25. Neither the Tg of the polymer nor that of the dispersion orders the outcome, and water uptake modulates it only at high RH C34·C40·C43. Among the pectins, the lower-DE CT removes VitC crystallinity at a lower content than the higher-DE EST R11.e·R11.t·R6. The authors' recommendation follows: choose the polymer by its ability to form intermolecular interactions with the compound C48.