How sorbed water binds to PVP, PVP/VA and PVAc, and why the glass transition changes the picture

Background and motivation

Many pharmaceutical excipients are hygroscopic. In an amorphous solid, water vapour from the air is not only adsorbed on the surface but absorbed into the bulk, and hydrophilic amorphous polymers can take up large amounts of it (Zografi 1988; Dole 1950; Oksanen 1990). How much they take up depends on temperature and humidity, but also on the chemistry of the polymer and on how the water changes the structure of the solid (refs 3–5). The absorbed water matters for products: it can change chemical stability, powder flow and compactability, and so "potentially" product performance (Ahlneck 1990). C1·C2

The best-known effect of sorbed water is plasticization. Water increases the free volume of the polymer and lowers its glass transition temperature, $T_g$. From a plot of $T_g$ against water content one can read the water content, commonly called $W_g$, at which $T_g$ falls to the temperature of the surroundings; above it the polymer is no longer a hard glass but a much less viscous rubber (Oksanen 1990). Earlier work by Zografi and co-workers on poly(vinylpyrrolidone), PVP, linked this change to the state of the water: in the glass both polymer and water are "considered" to have restricted mobility and the water is often called "tightly bound", while in the rubber both are far more mobile and the water is "thought to be" more "solvent-like" (Oksanen 1993). C3·C4

What had not been done, the authors write, was a detailed mechanistic study of the chemical interaction of water with PVP: how the water molecules actually attach to the polymer. Vibrational spectroscopy is a natural tool for this, because infrared and Raman spectra respond to hydrogen bonds; in particular, the carbonyl band of PVP is known to move to lower wavenumber when it hydrogen-bonds to small molecules or other polymers (Taylor 1998; Moskala 1985). C5·C6·R21

The paper therefore sets out to study, by FT-Raman spectroscopy, the interactions between solid polymers and the water sorbed in them, to learn something about the structure of that water, and to relate the spectral changes to the chemistry of each polymer and to the plasticizing effect of water. The four polymers were chosen to separate these factors. PVP K90 and PVP K12 have the same monomer but very different molecular weight and dry $T_g$. PVP K12 and the copolymer PVP/VA (poly(vinyl pyrrolidone-co-vinyl acetate), about 60:40 by weight) have about the same dry $T_g$ but different chemistry and water uptake. And PVP/VA can be compared with the two homopolymers of its monomers, PVP and poly(vinyl acetate), PVAc. C5·C7·A1·M1–M4

Background concepts

Hydrogen-bond acceptors without donors. All three polymer types carry a carbonyl group on the side chain: a tertiary amide inside the pyrrolidone ring of PVP, and an ester in the acetate group of PVAc (Figure 1). Both are hydrogen-bond acceptors, but none of the polymers has an acidic proton, so the dry polymers cannot hydrogen-bond to themselves. A carefully dried sample therefore shows the carbonyl in its non-bonded state, and any water that bonds to it shows up as a change of that band. C9·M1–M4

The carbonyl shift. When a water molecule donates a hydrogen bond to a carbonyl oxygen, the C=O stretching vibration is weakened and its Raman band moves to lower wavenumber. The paper measures this as the shift of the band maximum from its position in the dry polymer, in cm⁻¹. The size of the shift depends on how much water is bound and on how strong the bond is; the authors note that bond strength in turn depends on steric effects, on the acidity of the donor proton and on the basicity of the carbonyl (Pimentel 1960; Nyquist 1990). P6·C28

Counting water per monomer unit. To compare polymers on an equal footing the authors express water content as a mole fraction, counting the polymer not by its chains but by its monomer units: molecular weight 111 for the vinylpyrrolidone unit, 86 for the vinyl acetate unit and 101 for the PVP/VA "unit". For PVP K12 and PVP K90 they also use the equivalent count of water molecules per monomer unit (Figure 12). On this basis the two PVP grades, which have the same number of monomer units per gram, can be compared directly. E1·E3·C35

Glassy and rubbery at the measurement temperature. All spectra were recorded at room temperature, 22 °C. A sample whose $T_g$ (here the onset of the DSC step) lies above 22 °C is glassy during the measurement; one whose $T_g$ lies below is rubbery. Because water lowers $T_g$, each polymer passes from glassy to rubbery at some water content, and the paper reads every spectral change against that point. P1·P4·R11·C23

Principal components analysis. Besides the carbonyl band, water causes many small changes in the polymer bands between 1550 and 600 cm⁻¹: slight changes in band heights and small shifts, too subtle to follow one by one. Principal components analysis (PCA) reduces each spectrum in a set to a few numbers. The first principal component (PC1) is the single direction in which the spectra of a set vary most; each spectrum's score on it says where that spectrum lies along this direction, and the loadings say which wavenumbers contribute. The sign of a score carries no meaning. Water itself has no Raman bands in this region, so the analysis sees only the polymer. P5·R13·C24

Methods in detail

The study has one backbone: a single humidity series, applied to all four polymers, on which three kinds of measurement were made in parallel. Water uptake and $T_g$ say how much water each sample holds and whether it is glassy or rubbery at 22 °C; the Raman spectra then say what that water does to the polymer. A1·A2·P1

Materials and humidity control

PVP K90 and PVP K12 came from BASF; the manufacturer gives weight-average molecular weights of 1,000,000–2,000,000 and 2000–3000 Da. PVP/VA, also from BASF, is a random copolymer of vinyl pyrrolidone and vinyl acetate at a weight ratio of about 60:40, with a reported weight-average molecular weight of 45,000–70,000. PVAc came from Aldrich with a stated weight-average molecular weight of 12,800; it arrived as large chunks and was milled to a fine powder, while the other polymers were supplied as fine powders. M1·M2·M3·M4

Humidity was set by saturated salt solutions in desiccators: lithium chloride (11% RH), potassium acetate (22%), magnesium chloride (33%), potassium carbonate (44%), magnesium nitrate (54%), sodium nitrite (65%), sodium chloride (75%), potassium chloride (85%) and potassium nitrite (94%), plus dry samples. Samples kept at low humidity were handled and capped in a glove box below 5% RH. The potassium chloride condition appears as 85% RH in the methods and as 84% in the captions of Figures 5 and 7. P1·X7

Water uptake and glass transition

For the isotherms, 0.5 g of dried polymer was stored at each humidity for 6 weeks and weighed periodically until the weight was constant; the equilibrated weight was recorded, the sample was dried under vacuum at 70 °C for 24 h, and the dry weight was recorded. The isotherms in Figure 3 were measured at 22 °C. P3·R3

For the glass transition, about 5 mg of polymer was equilibrated in an open aluminium pan, sealed hermetically and scanned in a Mettler Toledo 820 DSC at 20 K/min. The onset of the step is reported, taken from a second scan after the sample had been heated 20 K above $T_g$ to erase its thermal history. P4·R11

Raman spectra and the carbonyl shift

Samples were stored in 5 mm NMR tubes over the salt solutions for 4–6 weeks, a time shown earlier to be enough for equilibration, and then capped tightly so that their water content could not change during the measurement. Spectra were recorded on a Perkin-Elmer System 2000 FT-Raman spectrometer with a 1064 nm Nd:YAG laser (typically 800 mW at the sample), an InGaAs detector and 180° back-scatter, averaging 64–512 scans at 4 cm⁻¹ resolution, with more scans for samples containing less polymer; the tube rotated during the measurement and indene served as a wavenumber standard. P2

The authors first tried to describe each humidified carbonyl band as a mixture of two reference spectra: the dry polymer for the non-bonded state and a polymer solution for the bonded state. The calculated spectra did not fit the measured bands well, which they read as "suggesting" that such a two-state picture is not enough (the fits are not shown). They therefore used the simpler shift of the band maximum, which allows the polymers to be compared "at least qualitatively". R5·C15

Using the peak position as a measure of hydrogen bonding assumes that water acts on the carbonyl mainly through hydrogen bonds and not through the extra mobility it gives the polymer. To test this, the spectrum of dry PVP K12 was recorded at room temperature, where it is glassy, and at 130 °C, where it is rubbery. The carbonyl position differed by less than 1 cm⁻¹, against shifts of up to about 22 cm⁻¹ caused by water, and the authors take the assumption as valid. P7·R7·C16·R8

The paper does not say that the water content of each Raman sample was measured directly. The mole fractions plotted in Figures 6 and 10 agree with those calculated from the gravimetric isotherm of Figure 3 (for example, about 0.81 for PVP K12 at 94% RH), so they appear to be taken from the separate sorption samples. E1·X6

Principal components analysis

For each polymer, 30 spectra (10 conditions including dry, with three spectra from independent samples per condition) were first corrected for intensity differences by multiplicative scatter correction over 1800–600 cm⁻¹, then mean-centred and analysed by PCA over 1550–600 cm⁻¹ with random cross-validation. In every polymer the first principal component described more than 85% of the variation, and only this component is reported. Each point in Figure 10 is the mean score of the three spectra at one humidity. P5·R14

Experimental analysis

The dry spectra

Table 1 and Figure 2 give the starting point. PVP K12 and K90 have almost identical spectra, as expected for the same monomer. The amide carbonyl of PVP lies at 1675 cm⁻¹ and the ester carbonyl of PVAc at 1733 cm⁻¹; in PVP/VA the two appear as separate, well-resolved bands at 1678 and 1735 cm⁻¹, so the pyrrolidone and acetate parts of the copolymer can be followed independently. Two further bands are diagnostic: the ring-breathing band at 933 cm⁻¹ occurs only with the pyrrolidone ring, and the O–C=O deformation at 631 cm⁻¹ (634 cm⁻¹ in PVP/VA) only with the acetate. The C–C backbone band lies at 1023 cm⁻¹ in both homopolymers. R1·R2·C8·C10

How much water, and when each polymer turns rubbery

The four isotherms (Figure 3) differ widely. At 94% RH the PVP grades hold more than 40% water (about 0.43 and 0.41 weight fraction for K90 and K12), PVP/VA about 0.33 and PVAc only about 0.03. PVP K90 sorbs somewhat more than K12 at low humidity (about 0.105 against 0.069 at 33% RH) and the curves merge at high humidity, as reported earlier; the authors call this molecular-weight effect "slight". PVP/VA sorbs less than either PVP at low humidity, and the difference narrows as humidity rises. R3·R16·R17·C11·C12

Water lowers $T_g$ in all four (Figure 8). Dry PVP K90 has by far the highest $T_g$: 175 °C in the Discussion, about 169 °C in Figure 8. PVP K12 and PVP/VA both start near 100 °C and fall similarly with water, which the authors read as "approximately the same plasticizing ability"; PVAc starts near 30 °C and loses only about 20 °C over its small range of water contents. Over the humidities used, the drier samples of each polymer are glassy at 22 °C and the wetter ones rubbery. M1–M4·R11·C22·C23·X1

Table A. Water uptake (weight fraction, 22 °C) and the water content at which the DSC onset $T_g$ passes the measurement temperature, read from Figures 3 and 8 (approximate). The last column matches these water contents to the isotherm; it is the AI writer's mapping, not a column of the paper. R3·R11·M1–M4·X1·X5

The ordering in the last two columns frames everything that follows. PVP K12 is rubbery at 22 °C from 44% RH upward, whereas PVP K90, though it sorbs more water, stays glassy up to 65% RH. The K90 transition is also broad: at a water mole fraction of 0.69 (the 75% RH sample) its $T_g$ spans 15–23 °C from onset to offset, so at 22 °C that sample sits inside the transition rather than clearly above it. R11·R12·C32·C26

Water binds the pyrrolidone carbonyl much more than the acetate

As water content rises, every carbonyl band moves to lower wavenumber. Figure 5 shows this for PVP K90: the band maximum moves from about 1674 cm⁻¹ in the dry sample, step by step, to about 1653 cm⁻¹ after storage at 94% RH, and the band broadens on its low-wavenumber side. The authors read the downshift as "indicating" hydrogen bonding between water and PVP, as reported before for PVP with small molecules and polymers. R6·C14·R21

Figure 6 puts all five carbonyls on one axis of water mole fraction. At low water the three pyrrolidone carbonyls (K90, K12 and the VP part of PVP/VA) shift by similar amounts, and all reach about 22–22.5 cm⁻¹ at the highest water contents. The acetate carbonyl shifts far less at the same water: in PVP/VA about 2.8 cm⁻¹ against 10.9 cm⁻¹ for its pyrrolidone at a mole fraction of 0.44, and 12.1 against 22.0 cm⁻¹ at 0.73. PVAc reaches only about 1 cm⁻¹, and its data end at a mole fraction of about 0.135 because it sorbs so little water. R8·C19·C20·C27

Table B. Carbonyl shift (cm⁻¹) at the stated mole fraction of water, read from Figure 6 (approximate); dry positions from Table 1. R1·R8

The authors explain the difference by basicity. The strength of a hydrogen bond depends on the basicity of the acceptor, and the pyrrolidone group is "considerably more basic" than the acetate group, so it forms the stronger bond with water and shifts further at a given water content; the largest difference between the polymers "can therefore be explained" by their chemistry. Basicity itself was not measured in the paper. C28·C29

Two points qualify the numbers. The shift is a qualitative measure, adopted because the two-state band fit failed, so equal shifts in two polymers need not mean equal numbers of bonded carbonyls. And the mobility control (less than 1 cm⁻¹ for dry PVP K12 between room temperature and 130 °C) shows that softening the polymer alone does not move the band, which is what lets the authors read the shift as hydrogen bonding. R5·C15·R7·C16

The shift steps up where the polymer turns rubbery

On closer inspection each pyrrolidone curve in Figure 6 is roughly linear up to a point, then jumps, then continues roughly linearly. The jump sits at a different water content in each polymer (between mole fractions of about 0.31 and 0.41 for K12, 0.44 and 0.53 for PVP/VA, and 0.69 and 0.73 for K90), and in PVP/VA the acetate carbonyl jumps at the same point as the pyrrolidone. The authors describe this in words ("it is apparent"); they fit no lines and give no statistics. R8·C18·X5

Matched to the humidity series, these jumps fall in the same humidity step in which $T_g$ passes 22 °C for K12 (33 → 44% RH) and PVP/VA (65 → 75% RH); for K90 the jump comes one step later (75 → 85% RH), after its broad transition. Each jump is located only to one humidity step. R8·R11·R12·X5

Figure 12 makes the comparison between the two PVP grades explicit by replotting the shift against the number of water molecules per monomer unit, with arrows where each grade turns rubbery: near 0.48 water molecules per monomer for K12 and near 2.2 for K90. Below the K12 arrow the two curves lie close. Between the two arrows K12 shifts clearly more than K90 at the same water per monomer. Beyond the K90 arrow the K90 curve jumps, and both end near 22.5 cm⁻¹ at 4.3–4.6 water molecules per monomer. R9·C33

Table C. Carbonyl shift (cm⁻¹) of the two PVP grades at similar numbers of water molecules per monomer unit, read from Figure 12 (approximate); states from the arrows in Figure 12. R9·X8

The authors conclude that these results "support the conclusion" that the physical state of the polymer, as well as its chemistry, influences how much water hydrogen-bonds to it, and that the K12–K90 difference "strongly suggests" that factors other than chemistry affect water uptake. They describe the two curves as "virtually identical" while both grades are glassy; in Figure 12 the K12 points already lie about 1–2 cm⁻¹ above the K90 curve there, against about 4 cm⁻¹ once K12 is rubbery. C34·C30·C33·X8

The first one or two water molecules, and the state of the water

Figure 12 also shows where most of the shift happens. For K12 the shift reaches about 14.8 cm⁻¹ at one water molecule per monomer and about 18.0 cm⁻¹ at two, then only about 19.8 cm⁻¹ at three and 22.6 cm⁻¹ at 4.3. The authors summarize this as the largest shift between 0 and 2 water molecules per monomer ("~15 cm⁻¹ for K12"), a further shift up to 3 and a levelling off by 5; the plotted value at two molecules is closer to 18 cm⁻¹, and the Results section had described the same data, on the mole-fraction axis, as showing "no sign of leveling off". From this pattern they "imply" that more than one water molecule is associated with each carbonyl, in line with an earlier infrared study (Kobyakov 1981). R9·C37·C38·R19·C17·X4

The water itself was followed through its OH-stretching band between 3500 and 3100 cm⁻¹ (Figure 7), a weak Raman band. Pure water shows a broad double band, with a high-frequency and a low-frequency component and a shoulder on the low-wavenumber side. Water sorbed in PVP K12 at low humidity looks different: the band is weak and flat with its maximum near 3400 cm⁻¹; according to the text the high-frequency component sits above its position in pure water, the low-frequency component is weaker and the shoulder is weak. As humidity rises, a second maximum near 3320 cm⁻¹ and a shoulder near 3230 cm⁻¹ grow, and at 94% RH the band closely resembles that of pure water without being identical. PVP K90 and PVP/VA behave similarly (spectra not shown); in PVAc there is too little water to measure. R10·C21

The authors join these observations. The water is least like bulk water exactly in the range (0–2 water molecules per monomer) where the carbonyl shift grows fastest, which is "consistent with a model" in which the first one or two water molecules hydrogen-bond directly to the side-chain group of the polymer, while further water molecules bond to other water molecules and look more like, but not entirely like, bulk water. The two populations are inferred from the shapes of the two bands; the paper does not separate them spectroscopically. C39

PVP/VA: the acetate joins in once the copolymer is rubbery

The copolymer shows the effect of the physical state most clearly. Its acetate carbonyl barely moves while PVP/VA is glassy (about 2.8 cm⁻¹ at a mole fraction of 0.44), then jumps to 5.6 cm⁻¹ at 0.53 and reaches 12.1 cm⁻¹ at 0.73; the jump falls in the humidity step in which the copolymer passes $T_g$. Above that step the pyrrolidone of PVP/VA also shifts slightly more than PVP at similar water (about 17.4 cm⁻¹ at 0.53 against 14.7 cm⁻¹ for K12 at 0.50). R8.c–d·R8.b·R11.c·X5·C19·C40

The water uptake of PVP/VA tells the same story. Treating the copolymer as 60% PVP K12 and 40% PVAc and adding their isotherms in that ratio predicts its isotherm well up to 65% RH ("excellent agreement"), but above that the measured water content is "considerably higher" (Figure 4). E2·R4·C13

Table D. PVP/VA water uptake (weight fraction, 22 °C), observed and calculated from the PVP K12 and PVAc isotherms in a 60:40 ratio, read from Figure 4 (approximate); below 44% RH the two agree within the reading. R4·E2

The authors "speculate" that in the glassy copolymer water goes mainly to the pyrrolidone groups, which bind it more favourably, so the acetate carbonyl hardly shifts, while the pyrrolidone shifts as in PVP. Once PVP/VA is rubbery the acetate groups also hydrogen-bond with water, not at the expense of the pyrrolidone, whose interaction is even slightly enhanced relative to PVP. They see "further evidence" for this enhanced bonding in the excess water uptake, which appears "only when the polymer is in the rubbery state". As a cause they "speculate" that, with both groups on the same chain, the glass-to-rubber change gives a chain conformation unlike that of the homopolymers, which makes the side groups more available to water, strengthens their bonds or, "less likely", changes their basicity indirectly. "It thus appears" that for this polymer the physical state is very important, particularly for the acetate group. C41·C42·C43·C44

Note by the AI writer (not discussed in the paper): the additivity baseline is built from PVP K12, which is itself rubbery at 22 °C from 44% RH, and from PVAc, which passes its $T_g$ between about 54 and 65% RH. Above 44% RH the "calculated" curve therefore already contains the sorption of rubbery homopolymers. Figure 4 shows that the copolymer sorbs more than its parts once it is rubbery, but the comparison does not by itself separate a glassy-versus-rubbery effect from a copolymer effect. X9·R11

A principal-component score that breaks at the glass transition

Besides the carbonyl, water causes small changes across the 1550–600 cm⁻¹ region of PVP K90 (Figure 9): changes in band height ratios and small band shifts, marked by arrows near 1458, 1315, 1118, 1020 and 940 cm⁻¹. The PC1 loadings of PVP K90 (Figure 11) are largest at the ring-breathing band near 940 cm⁻¹ and at the CH₂ bands between about 1420 and 1490 cm⁻¹, so these bands carry most of the change. R13·R15

The PC1 scores (Figure 10) behave like the carbonyl shift. For PVP K12 the score rises from about −28.6 to −17.3 over the first four conditions, jumps to about +3.6 at the next, then rises again in steps of about 3–6. PVP K90 and PVP/VA follow the same pattern with the jump at other water contents. R14·C24

Table E. First principal-component scores, read from Figure 10 (approximate); the last column compares the jump with the $T_g$ crossings of Figure 8. R14·R11·X2·X3·X5

For K12 and PVP/VA the authors state that the break corresponds to the water content range in which the polymer changes from glassy to rubbery. For K90 the break comes at a slightly higher water content, which they say "can probably be explained" by the breadth of the transition: at a mole fraction of 0.69 the K90 $T_g$ spans 15–23 °C, so that sample is still inside its transition at 22 °C, and allowing for the width, K90 follows the same trend. For PVAc they report "no clear trends". C25·C26·R12·C24

Note by the AI writer (not discussed in the paper): the PVAc panel of Figure 10 deserves a second look. Its x axis, labelled mole fraction, runs only to about 0.031, which matches the weight fraction of water in PVAc at 94% RH (about 0.029); with the monomer molecular weight of 86 the mole fraction would be about 0.125, and Figure 6 does plot PVAc out to about 0.13. Read as weight fraction, the PVAc scores fall from about 5.9 to −8.7 over the water range in which PVAc passes 22 °C (about 0.010–0.013), and then level off. That is a step, not an absence of trend, and it sits where the other polymers would lead one to expect it. X2·X3·R14.d·R11.d

Discussion

How the authors connect the findings

The paper's argument can be laid out as one chain with two branches, one for the glassy and one for the rubbery polymer. Neither branch is "good" or "bad" for a formulation in itself; they are two regimes.

Two parts of this chain are measured directly: the carbonyl shift with its dependence on chemistry, and the coincidence of the spectral steps with the $T_g$ crossings. The links between them are the authors' explanations. Basicity is invoked but not measured. The rise in site accessibility is introduced with "One can hypothesize" and the K12–K90 difference is said to "most likely" reflect it, with reference to the Vrentas free-volume model, which predicts how plasticization proceeds from the dry $T_g$ of each grade (100 °C for K12 and 175 °C for K90 in the text). In the abstract the same difference is "attributed" to differences in the "structural relaxation changes accompanying plasticization" of the two grades. The two populations of water are "consistent with a model", and the conformational reading of the PC1 break rests on "we can hypothesize". C29·C35·C36·E4·C39·C47

The argument for that last step is worth spelling out. Water uptake shows no sudden change at $T_g$ (the isotherms in Figure 3 are smooth), yet the polymer bands change abruptly there. If the bands simply tracked the amount of water, they should change smoothly too; the authors therefore attribute the discontinuity to the polymer itself, to the change in its conformational freedom and molecular mobility on passing $T_g$. They stress that Raman spectroscopy "cannot be used to directly measure mobility and dynamics" in amorphous systems; what it shows are the consequences of mobility for the vibrations. C47·C49·R3

The authors present the spectroscopic discontinuity at $T_g$ as new: such discontinuities are well known from many techniques, but "to the authors' knowledge" had not been reported for vibrational spectra. Their overall conclusion is that sorbed water changes the polymer Raman spectrum "not only because of hydrogen bonding" but also through its plasticizing effect on polymer mobility; chemistry "dominates" the polymer–water interaction, and the structural state of the polymer has a "notable effect". C52·C51·C50

What it suggests for formulation

The paper itself makes few formulation statements. Its motivation is that sorbed moisture affects chemical stability, flow and compaction, and its practical message is methodological: multivariate analysis of Raman spectra can tell glassy from rubbery samples. C2·C48

Note by the AI writer (not discussed in the paper): three points follow from the data for anyone working with PVP or PVP/VA (copovidone) as a carrier. First, molecular weight matters at ordinary humidity: at 22 °C, PVP K12 is already rubbery between 33 and 44% RH, while K90 and PVP/VA stay glassy up to 65% RH. A low-molecular-weight grade may lose its glassy state on an ordinary laboratory shelf. Second, the pyrrolidone carbonyl that binds water here is the same acceptor through which many drugs hydrogen-bond to PVP and PVP/VA in amorphous solid dispersions. The large shift produced by the first one or two water molecules per monomer means that sorbed water is a strong competitor for that site from low humidity onward, and more so once the polymer is rubbery; this paper contains no drug, so the competition itself is untested. Third, in PVP/VA the acetate group becomes a second water-binding site only above the glass transition, and the copolymer then sorbs more than the sum of its homopolymers, so uptake predicted from homopolymer data would be too low at high humidity. R11·R3·R9·R8·R4

Limitations

Limits stated by the authors. The carbonyl shift is used only as a qualitative measure, because a two-state description of the band did not fit. PVAc absorbs so little water that its acetate cannot be compared with the acetate of PVP/VA over the same range, and its water band cannot be measured at all. The mechanism of the PC1 break is offered as a hypothesis, with the explicit warning that Raman cannot measure mobility directly. The explanations for the copolymer are labelled as speculation. C15·C27·R10·C47·C49·C41·C43

The AI writer's own judgement (1): what is measured and what is inferred. The strongest result is the difference between pyrrolidone and acetate: it is large (about 22 against 12 and 1 cm⁻¹), it appears in three polymers, the two carbonyls of PVP/VA are measured in the same sample, and the mobility control rules out a pure temperature or softening effect. The state effects are weaker evidence. Each step in the shift and in the PC1 score lies between two adjacent humidities, located by eye, with no fit, no statistics and undefined error bars, and the $T_g$ crossings are known to the same one-step resolution. The coincidence is real in three polymers, but finer humidity steps would be needed to show that the spectral step and the glass transition coincide rather than merely fall in the same interval. The water contents of the Raman samples appear to come from separate gravimetric samples, and the drying of the dry Raman and DSC samples is not described. R8·R7·X5·X6

The AI writer's own judgement (2): numbers that differ inside the paper. Several stated values do not match the figures: the dry $T_g$ of PVP K90 (175 °C in the text, about 169 °C in Figure 8); the K12 shift at two water molecules per monomer ("~15 cm⁻¹" in the text, about 18 cm⁻¹ in Figure 12); the levelling of the shift (absent in the Results, present by five water molecules in the Discussion); the "virtually identical" glassy-range curves of K12 and K90, which differ by about 1–2 cm⁻¹; the PVAc axis of Figure 10, which appears to be weight rather than mole fraction; and the 84 against 85% RH label for the potassium chloride condition. None of these changes the main conclusions, but they matter for anyone reusing the numbers. X1·X4·X8·X2·X7

The AI writer's own judgement (3): scope. All measurements are at one temperature, on neat polymers equilibrated from the vapour. Free volume, bonding-site accessibility, chain conformation and the two proposed populations of water were inferred, not measured, and the additivity baseline for PVP/VA already includes rubbery homopolymers above 44% RH. The paper establishes a spectroscopic picture of water in these polymers; it does not test how that picture changes when a drug is present. C35·C39·C43·X9

Conclusion

The paper makes three contributions. First, with FT-Raman spectra of four polymers across a humidity series, it shows that sorbed water hydrogen-bonds to the side-chain carbonyls and that the pyrrolidone carbonyl binds it far more than the acetate, with a mobility control that excludes softening alone as the cause. Second, by placing every spectrum against the water content at which $T_g$ falls to 22 °C, it shows that the physical state of the polymer modulates the interaction: PVP K12 binds more water per monomer than K90 only while K12 is rubbery and K90 still glassy, and the acetate of PVP/VA joins in, with extra water uptake, only once the copolymer is rubbery. Third, it shows that a single principal-component score of the polymer bands breaks at the glass transition, a spectroscopic marker of $T_g$ that the authors had not seen reported for vibrational spectra. The authors' own summary is that chemistry "dominates" the polymer–water interaction while the structural state of the polymer "also had a notable effect". C50·C34·C33·C41·C42·C25·C48·C52·R7

The mechanistic explanations, carbonyl basicity, plasticization-driven access to bonding sites, one or two directly bound water molecules per carbonyl, and a changed chain conformation in the copolymer, remain hypotheses that the data are consistent with but do not test. For formulation work the most direct lessons are the ranking of the two carbonyls as water binders, the strong dependence of the glass transition humidity on PVP grade, and the possibility of reading the glassy or rubbery state of a PVP-type matrix from its Raman spectrum. C29·C35·C39·C43·R11·C48