How PVP breaks up the acid dimers of amorphous indomethacin: an infrared and Raman study of solid dispersions

Background and motivation

An amorphous solid dispersion of a poorly soluble drug in a polymer is meant to dissolve faster than the crystal, because the amorphous drug is more soluble and the hydrophilic carrier improves wetting; binary amorphous mixtures of this kind, sometimes called glass solutions, are considered homogeneous systems in which the drug is dissolved in a glassy solvent (Chiou and Riegelman 1971; Ford 1986). C1 Poly(vinylpyrrolidone) (PVP) is one of the most common carriers: over 60 drugs had been dispersed in it by 1986 (Ford 1986). The catch is that the amorphous phase is metastable relative to the crystal, so it "may" transform during storage and on dissolution. C2

PVP was known to inhibit crystallisation both in solution (Sekikawa et al. 1978) and in amorphous solid dispersions (Yoshioka et al. 1995). In solution PVP interacts with many organic molecules, and the inhibition "has been suggested" to depend on how strongly drug and polymer interact. C3 A second explanation is mobility: crystallisation from the amorphous state is related to molecular mobility, and PVP, with a glass transition temperature (Tg) of 177 °C, should raise the Tg of the mixture and slow the molecules at a given temperature. C4·M4 Yoshioka, Hancock and Zografi had tested this for indomethacin: PVP inhibited crystallisation significantly even at relatively low levels, and a close analysis showed that this could not be explained solely by the antiplasticising effect of the polymer. C5 Something more specific seemed to be at work.

The obvious candidate was a molecular interaction between drug and polymer, and vibrational spectroscopy is the natural tool to look for one. Earlier spectroscopic studies of solid dispersions (El-Hinnawi and Najib 1987; Sekizaki et al. 1995; Tantishaiyakul et al. 1996) had, however, compared the spectrum of the crystalline drug with that of the dispersion, in which the drug is usually amorphous, and had inferred interactions from the differences. Because crystallinity by itself changes vibrational spectra (Bugay and Williams 1995), such comparisons mix two effects. The authors therefore insist that the amorphous drug must be the reference state. C7 They also frame the dispersion in physical terms: below its Tg a binary amorphous mixture can be regarded as a highly viscous, non-equilibrium supercooled liquid, but little was known about how like and unlike molecules associate in such glasses. C6

Indomethacin is a good test case. It has two monotropic polymorphs, the stable γ form and the metastable α form, several solvates and an amorphous form (Yamamoto 1968; Borka 1974). M1 The γ crystal structure is known: two molecules in the unit cell, with the carboxylic acid groups paired as cyclic dimers (Kistenmacher and Marsh 1972); the hydrogen bonding in the α and amorphous forms was unknown. M1·M2 The paper therefore has two aims: to establish how indomethacin molecules are hydrogen-bonded in each solid form, using IR and FT-Raman spectra, and then to identify any specific interaction with PVP in solid dispersions, judged against amorphous indomethacin and physical mixes. A1·A3·C24

Background concepts

The carbonyl stretch as a probe of hydrogen bonding. The C=O stretching vibration absorbs between about 1750 and 1600 cm⁻¹. When the oxygen of a carbonyl accepts a hydrogen bond, the band moves to lower wavenumber, and a carbonyl that is not hydrogen-bonded (a "free" carbonyl) absorbs at higher wavenumber. R7·C28 Indomethacin has two carbonyls, the benzoyl (amide) C=O on the indole nitrogen and the acid C=O, but only one proton donor, the acid OH (Figure 3). PVP and its small analogue N-methyl-2-pyrrolidone (called methylpyrrolidone in the paper) have only acceptors, the lactam carbonyl oxygen and the ring nitrogen. M1·M4·M5 So any new hydrogen bond between drug and polymer must show up in two places: in the acid C=O band of indomethacin, and in the C=O (or C–N) band of PVP. C25

Cyclic acid dimers and the mutual exclusion rule. Carboxylic acids hydrogen-bond in three ways: as cyclic symmetric dimers, as open chains (the paper calls them "catamers"), or with other functional groups (Leiserowitz 1976). A cyclic dimer has a centre of symmetry, so the mutual exclusion rule applies: the in-phase (symmetric) stretch of the two C=O groups changes no dipole moment and is seen only in the Raman spectrum, while the out-of-phase (asymmetric) stretch changes the dipole moment and is seen only in the IR, at a noticeably higher wavenumber. Comparing IR and Raman spectra therefore tests for cyclic dimers. C14 Liquid acetic acid shows the rule cleanly: its asymmetric C=O stretch is at 1716 cm⁻¹ in the IR and its symmetric stretch at 1669 cm⁻¹ in the Raman spectrum (Figure 4c). R5.j A band that appears at the same position in both spectra cannot come from a centrosymmetric dimer and points to another bonding pattern. C16

Chains and their free ends. In a hydrogen-bonded chain, each acid donates its OH to the next molecule's C=O, so the molecule at the end of a finite chain is left with a free carbonyl. Liquid acetic acid, described in the literature as mainly cyclic dimers with other hydrogen-bonded species and monomers, has an IR shoulder (1759 cm⁻¹ in Table 1, 1758 cm⁻¹ in the sentence that cites it) assigned to the free carbonyl of a chain end group (Haurie and Novak 1965). R8·R5.k·X1 This is the analogy the authors later use for amorphous indomethacin.

The reference state and the control. Two reference spectra are used for every dispersion. Amorphous indomethacin, made by quench-cooling the melt, removes the effect of crystallinity. A physical mix of amorphous indomethacin and PVP at the same composition removes simple superposition: if the dispersion spectrum differs from the physical-mix spectrum, the difference must come from molecular mixing. C24·P1·P4

Stoichiometry in weight terms. The paper expresses compositions as weight percent of PVP. To ask whether there is enough PVP to bind every indomethacin molecule, the authors convert to mole ratios with 111 as the molecular weight of one PVP repeat unit. E1·M4 They find that a 1:1 ratio of PVP monomer to indomethacin is reached at 24% weight fraction PVP. R18 They give no equation; the relation is stated in words, and this article does not add one.

Methods in detail

The study has three experimental parts and one calculation. All solids were measured by FT-Raman (Bruker RFS 100, Nd:YAG laser at 1064 nm, typically 500 mW, 180° back-scattering, 200 scans at 4 cm⁻¹ resolution, 4000–50 cm⁻¹) and by transmission FTIR (Mattson Galaxy 5020 with DTGS detector, 500 scans at 4 cm⁻¹, 4000–400 cm⁻¹). P7·P8 Solids for IR were diluted to 0.8–1% in dry KBr, ground in a Wig-L-Bug mixer and pressed into pellets; liquids were held between NaCl windows. P8 The sample temperature is not stated. X7

Part 1: how each solid form of indomethacin is hydrogen-bonded

Problem. Before a drug–polymer interaction can be read from a dispersion spectrum, one needs to know how the drug is bonded on its own in the amorphous state, which is the form it takes in the dispersion. Only the γ structure was known. M1·M2

What was done. γ-indomethacin was used as supplied by Sigma. The α form was made by dissolving indomethacin in methanol and precipitating it with water at room temperature, following Kaneniwa et al. (1985). Amorphous indomethacin was made by melting at 165 °C for 5 minutes and quench-cooling in liquid nitrogen, then warming to room temperature under vacuum (to avoid condensing moisture) and storing over phosphorous pentoxide at −20 °C. M1·P1·P2 IR and Raman spectra of the three forms were recorded over the full range (Figures 1 and 2) and in the carbonyl region (Figure 4), and the carbonyl bands are tabulated in Table 1. R2·R3·R4·R5

Why it works. The γ form gives a calibration: its acid groups are known to form cyclic dimers, so its spectra show what the dimer looks like in IR and Raman. Liquid acetic acid gives a second calibration for a simple acid. The α and amorphous forms are then read against these two. C14·C15·R5.j

Part 2: solid dispersions against physical mixes

Problem. Does PVP change the bonding of indomethacin when the two are mixed at the molecular level, and through which groups? A3

What was done. PVP-K90 (GAF Corp.) was dried at 105 °C for 12 hours under vacuum. For each dispersion, 5 g of indomethacin and PVP in the required ratio were dissolved in 100 ml of anhydrous methanol at 65 °C; the solvent was removed under vacuum at 50 °C on a rotary evaporator and the residue dried under vacuum at room temperature for 24 hours. All compositions were amorphous by X-ray powder diffraction, and the dispersions were stored at −20 °C over phosphorous pentoxide. M4·P3·R1 Physical mixes were made from amorphous indomethacin and PVP at the same compositions; how they were mixed is not described. P4·X7 IR spectra of the carbonyl region were recorded at 0, 5, 10, 20, 30, 40, 50, 60, 70 and 100 wt% PVP for both series (Figure 5); Raman spectra of the dispersions were recorded but are described only ("data not shown"), as is the PVP C–N stretch region. R10·R11·R12·R13·R14

Why it works. Because the physical mix has the same composition, any band present in the dispersion but not in the physical mix needs molecular contact. The authors expect the acid C=O band and the PVP carbonyl band to respond, since the PVP carbonyl is the most favourable acceptor site, the nitrogen being sterically hindered (Molyneux 1983). C25·C26

Part 3: liquid models as an assignment key

Problem. In the dispersion spectra the PVP carbonyl absorbs in the same range as the indomethacin benzoyl group and partly under the acid band, so the changes are hard to assign directly. C27

What was done. Two liquid systems were measured by IR between NaCl windows: acetic acid mixed with methylpyrrolidone from 10 to 90 wt% methylpyrrolidone in 10% steps (Figure 6a), and indomethacin dissolved in methylpyrrolidone at 30 to 70 wt% methylpyrrolidone (Figure 6b). Acetic acid models the carboxylic acid group of indomethacin; methylpyrrolidone models the PVP repeat unit (Figure 3). P5·M5·M6

Why it works. In a liquid the only thing that can change the bands is hydrogen bonding between the components, so the liquid spectra show what acid–amide bonding looks like without the complications of a glass. Because the dispersion changes turned out to be "qualitatively similar" to both models, the model assignments are used for the dispersions. C32

Part 4: is there enough PVP to bind every drug molecule?

Using 111 for a PVP monomer unit, a 1:1 mole ratio of PVP monomer to indomethacin is reached at 24% weight fraction PVP. If each PVP carbonyl could accept two hydrogen bonds without steric hindrance, the authors state that 12% would be enough to bind every indomethacin molecule. E1·R18 These thresholds are then compared with the compositions over which the spectra still change. C37

Note by the AI writer (not discussed in the paper). The 12% figure is half of 24%, that is, the weight fraction was halved instead of the mole ratio. From the paper's own 24%, half a monomer per indomethacin molecule corresponds to about 13.6 wt% PVP. The difference does not change the argument, because the spectra keep changing well above both values. X2·E1

Experimental analysis

The three solid forms

Table A collects the carbonyl bands of Table 1. The γ form has an IR acid band at 1717 cm⁻¹ and no Raman band at that position; its Raman benzoyl band at 1698 cm⁻¹ is highly symmetrical in shape, so the authors consider it "unlikely" that the symmetric acid stretch hides beneath it, and conclude that the symmetric stretch "appears" not to be Raman active, as reported before for other dimer-forming acids. The 1717 cm⁻¹ band "must" therefore be the asymmetric stretch of the known cyclic dimer. R5.a·R5.b·R6·R7·C15

Table A. Carbonyl band positions and the authors' assignments, transcribed from Table 1 (p.1694); s strong, m medium, w weak, sh shoulder; a dash means no band listed. R5

Amorphous indomethacin. The amorphous form behaves like γ in the acid region: an IR band at 1710 cm⁻¹ with no Raman counterpart, which "suggests" that the acid groups are hydrogen-bonded as cyclic dimers, as in the γ polymorph. R5.h·R6·C19 It differs from γ by a shoulder at 1735 cm⁻¹. The authors note that the IR band shape of liquid acetic acid is similar, with its free chain-end shoulder, and find it "reasonable, therefore, to suppose" that amorphous indomethacin, a non-equilibrium supercooled liquid below its Tg, likewise consists mainly of cyclic dimers with a small proportion of molecules in hydrogen-bonded chains, the 1735 cm⁻¹ shoulder being the free carbonyl of chain-end molecules. A further argument is that α-indomethacin has a band at exactly the same position, also assigned to chain ends. R5.g·R5.c·R8·C20

α-indomethacin. The α spectrum is "considerably more complex" in this region. Besides a weak-to-medium IR band at 1735 cm⁻¹ and the benzoyl band, it has acid bands at 1681 (IR shoulder, Raman 1680) and 1649 cm⁻¹ that appear in both IR and Raman spectra. That coincidence "indicate[s]" a bonding pattern other than cyclic dimers. R5.c–f·R6·C16 The "most likely" alternative is chains of molecules linked through the acid groups; this "would be consistent" with α unit-cell parameters calculated from a synchrotron powder diffraction pattern (Dinnebier et al., unpublished results), which require a trimeric repeat unit stackable in one direction and 6 molecules in the unit cell. Chain ends would carry free carbonyls, which "could account" for the 1735 cm⁻¹ band. C17·M2 The authors add that interpretation is "hampered" by the lack of single-crystal data and that α bonding "appear[s]" more complex than γ. C18

Outside the acid group. The other spectral regions tell a different story, summarised in Table B. In the Raman CH stretching region, amorphous and α indomethacin share one aryl CH peak at 3070 cm⁻¹, while γ has an additional peak at 3022 cm⁻¹; the authors say that the amorphous CH pattern "generally resembles" that of α more than that of γ. R3·C10 In the Raman lattice region (150–50 cm⁻¹), γ shows sharp, well-resolved peaks, read as crystalline order, whereas α and amorphous indomethacin show broad peaks, read as more disordered structures. R4·C21 The benzoyl C=O lies at 1692 cm⁻¹ (IR) and 1698 cm⁻¹ (Raman) in γ, high for an amide group (usually 1695–1630 cm⁻¹). One explanation offered is that mesomerism is reduced when the amide nitrogen is part of a ring with its own resonance energy, as in indole ketones; another is that the single-crystal structure shows an unusually long C=N bond, attributed to the carbonyl being twisted out of the indole plane. R5.b·R7·R9·C11 In α (1688/1692 cm⁻¹) and amorphous indomethacin (1684/1681 cm⁻¹) the benzoyl band is lower, "suggesting" fewer conformational restrictions. R5.d·R5.i·C12 Finally, the acid OH stretch is a very broad IR band from 3400 to 2500 cm⁻¹ in all three forms, typical of a hydrogen-bonded OH, with differences in shape that "suggest" variations in hydrogen bonding. R2·C9

Table B. The three solid forms compared region by region (Table 1; text pp.1692–1695; Figures 1, 2, 4). R2·R3·R4·R5

The authors find it "interesting" that the amorphous form resembles α in the CH and lattice regions but resembles the more ordered γ crystal in how its acid groups are bonded. C10 They draw a general lesson from these differences: spectra change with crystal form and structural order, which "emphasise[s] the need to characterise the structural form of the compound in the system of interest". C13

Solid dispersions against physical mixes

No dispersion showed X-ray diffraction peaks over the whole composition range, so indomethacin was amorphous throughout. R1·C23 The IR carbonyl region (Figure 5) then shows a clear contrast. In the physical mixes, the doublet of amorphous indomethacin (bands near 1710 and 1684 cm⁻¹) persists at every PVP content, with the PVP band at 1679 cm⁻¹ added on top and growing with PVP content. R10·R5.l In the dispersions the spectrum changes progressively, and the dispersions are "clearly" different from the physical mixes of the same composition. C26 No change was seen in the PVP C–N stretch in either IR or Raman. R13

Table C. IR carbonyl changes in indomethacin–PVP solid dispersions (text p.1696, Figure 5). The paper gives no band intensities; "falls", "grows" and "present" are the authors' descriptions. R10·R11·R12·C33

Between 5 and 20% PVP the cyclic dimer band loses intensity while the non-hydrogen-bonded acid carbonyl band gains. At 20% PVP and above a single broad band at 1726 cm⁻¹ is present, which becomes a shoulder as the indomethacin content falls further and the PVP carbonyl dominates the spectrum. The authors assign it to the free acid C=O of an indomethacin molecule that is hydrogen-bonded to PVP through its hydroxyl. R11·C33 A band at 1636 cm⁻¹, absent in the physical mixes, appears at 5–50% PVP and is assigned to hydrogen-bonded PVP carbonyl; at higher PVP content it merges with the free PVP carbonyl band and remains only as a shoulder (60–70%), leaving a much broader band than in the physical mix. R12·C33 In the Raman spectra (not shown), the acid carbonyl is not seen and the PVP carbonyl coincides with the benzoyl band; with increasing PVP this combined band shifts progressively to lower frequency and broadens relative to the physical mixes, which the authors call "consistent with" the IR data. R14·C34

Note by the AI writer (not discussed in the paper). Two different bands are called "non-hydrogen-bonded acid C=O" in this paper: the 1735 cm⁻¹ shoulder of neat amorphous indomethacin (chain ends) and the 1726 cm⁻¹ band of the dispersions (drug bonded to PVP through its OH). Where the free band sits between 5 and 20% PVP, and how the 1735 cm⁻¹ shoulder turns into the 1726 cm⁻¹ band, is not reported. X4

The liquid models

Acetic acid with methylpyrrolidone (Figure 6a). The pure liquids show the acetic acid dimer C=O at 1716 cm⁻¹ and the methylpyrrolidone amide C=O at 1688 cm⁻¹. As methylpyrrolidone is added, the free-acid shoulder at 1759 cm⁻¹ grows relative to the dimer band, and the methylpyrrolidone carbonyl first appears at 1652 cm⁻¹, 36 cm⁻¹ below the pure liquid. R15 The authors account for this shift by hydrogen bonds between acetic acid and methylpyrrolidone, which cannot form in pure methylpyrrolidone because it has no donor: "It appears" that methylpyrrolidone disrupts the acid dimers by bonding to the acid hydroxyl. C28 With more methylpyrrolidone, a second band appears at the pure-liquid position of 1688 cm⁻¹. The text places the point where the hydrogen bonding of methylpyrrolidone is saturated and free methylpyrrolidone appears between 30 and 40% methylpyrrolidone; in the figure a separate band between the 1716 and 1652 cm⁻¹ peaks is visible from about 50%. Over this range three species coexist: methylpyrrolidone bonded to acetic acid, acetic acid dimers and free methylpyrrolidone. R15·C29

Indomethacin in methylpyrrolidone (Figure 6b). When indomethacin is dissolved in methylpyrrolidone, its 1710 cm⁻¹ dimer band disappears and the 1735 cm⁻¹ shoulder grows into a single band at 1723 cm⁻¹. At low methylpyrrolidone content a bonded methylpyrrolidone carbonyl band appears at 1649 cm⁻¹. The free methylpyrrolidone band coincides with the indomethacin benzoyl band, but it grows relative to the 1649 cm⁻¹ band as methylpyrrolidone is added even though the benzoyl contribution is being diluted, "suggesting" free methylpyrrolidone at higher concentrations; the saturation resembles the acetic acid case, and analogous species "can be postulated". R16·C30 "Interestingly", the authors add, methylpyrrolidone has an effect of similar magnitude on the indomethacin spectrum as PVP. C31

Table D. The same pattern in the two liquid models and the solid dispersion (Table 1; text pp.1695–1696). R5.l·R12·R15·R16·R11

Table D shows why the authors use the models to interpret the glass: in all three systems the amide carbonyl moves down by a few tens of wavenumbers when it accepts the acid OH, and the acid C=O that leaves the dimer appears as a separate higher-frequency band. C32·C33 Note by the AI writer (not discussed in the paper). The statement that the effect of methylpyrrolidone is "similar" in magnitude to that of PVP, and the abstract's "comparable ... at the same weight percent", are not backed by a stated measure or by named compositions; Figure 6b has no composition below 30 wt% methylpyrrolidone, whereas the dispersion changes are described from 5 wt% PVP. X6

Incomplete binding

With 24 wt% PVP for a 1:1 ratio and 12 wt% (as printed) if each PVP carbonyl took two hydrogen bonds, all indomethacin could in principle be bonded at modest PVP contents. "This is clearly not the case", the authors write, because the spectra keep changing at PVP contents above both values. "This suggests" that the equilibrium constant for forming PVP–indomethacin species is low and/or that the stoichiometry is less than 1:1. R18·C37 The picture they draw (Figure 7) has three species: indomethacin bonded to PVP, free PVP and indomethacin dimers. At low PVP, indomethacin is present as dimers and PVP-bonded molecules, with some free PVP; as PVP increases, free PVP increases and dimers "may still be present depending on the indomethacin-PVP equilibrium constant". C38

Discussion

What the hydrogen bond does

The authors' reading of the data is that indomethacin forms hydrogen bonds with PVP "at the expense of dimer formation": the spectral data "support the fact" of this interaction, which in the abstract is described as hydrogen bonding between the drug hydroxyl and the polymer carbonyl. C35 Why PVP can take the OH away from a dimer partner is explained by basicity: the PVP amide group is more basic than the carboxylic acid group (Stewart 1985), which "would explain" its ability to disrupt the dimers. C36 The chain of reasoning is summarised below.

One phase, behaving like a solution

A mixture with several hydrogen-bonded species could in principle be several phases. The authors argue that it is not: the presence of several species "in no way indicates that more than one phase is present"; rather, drug and polymer interact at the molecular level and the species are distributed in a single phase. C39 Two pieces of evidence are offered. First, PVP–indomethacin dispersions were reported earlier to have a single Tg over the whole composition range (Yoshioka et al. 1995), which is generally taken to indicate one phase (Painter et al. 1991). R17·C39 Second, the solid dispersion and the indomethacin–methylpyrrolidone solution are very similar in the size of the interaction, its nature and the species present. C31·C39 The authors add that this "does not preclude" non-ideal mixing, or concentration fluctuations of the solution being frozen in when the glass forms. C40 Their description of the glass as a non-equilibrium supercooled liquid of "solvent" and "solute" molecules that form both homo- and hetero-associations "seems to be reasonable", and fits the older idea that PVP can molecularly disperse many compounds, more readily those that hydrogen-bond with it (Shefter and Cheng 1980). C41 In the summary this becomes the "solution" nature of the binary amorphous system, "at least in terms of molecular interactions if not kinetics". C42

Crystallisation inhibition

The interaction has two consequences: a dimer is broken, and the proton donor that a dimer needs is engaged with another molecule. Because the γ crystal consists of dimers, the authors argue that forming a critical nucleus of γ-indomethacin "would require the breaking of the PVP-indomethacin hydrogen bond" so that molecules can dimerise and take up the right orientation. C43 Because α also appears to use the acid group, in chains rather than dimers, hydrogen bonding with PVP "would thus also be anticipated" to inhibit crystallisation to α. C44 The abstract puts it cautiously: PVP "may influence" the crystallisation kinetics by preventing the self-association of indomethacin; the summary says that the hydrogen bond "offers an explanation" of how PVP inhibits crystallisation from the amorphous phase at levels where the antiplasticising effect is minimal. C45 No crystallisation, nucleation or stability measurement is reported in this paper, so this part of the argument rests on the earlier observation of Yoshioka et al. and on reasoning. X8·C5

What it suggests for formulation (note by the AI writer, not stated in the paper)

For a carboxylic acid drug in PVP, the paper gives a practical recipe for detecting a drug–polymer hydrogen bond with a standard IR spectrometer: record the amorphous drug and a physical mix of the same composition, then look for the loss of the IR-only dimer band, the appearance of a free acid C=O band and a down-shifted polymer carbonyl band in the dispersion. Because the physical mix showed none of these changes, the same spectra also tell whether a preparation has achieved molecular mixing. The incomplete binding suggests that adding polymer beyond a stoichiometric amount does not guarantee that every drug molecule is bound. Whether the bond is the reason PVP slows crystallisation would need stability data on the same dispersions, which this paper does not provide. X5·X8

Limitations

Limits stated by the authors. Interpretation of α-indomethacin is "hampered by the lack of single crystal data". C18 The PVP carbonyl absorbs in the same range as the indomethacin benzoyl group and influences the acid region "to a lesser degree", so the dispersion bands overlap. C27 The single-phase picture does not exclude non-ideal mixing or frozen-in fluctuations. C40 The "solution" analogy is claimed for molecular interactions, explicitly "if not kinetics". C42

The AI writer's own judgement (1): qualitative reading of overlapping bands. The paper reports band positions and describes intensity changes in words; no band intensities, band areas or curve fits are given. The conclusion that binding is incomplete above 24 wt% PVP rests on continued spectral change in a region where the growing PVP carbonyl band itself changes the composite band, so it is a qualitative inference rather than a measured bound fraction. X5·C37 The two-bond threshold of 12 wt% is miscalculated (about 13.6 wt% from the paper's own 24%), without effect on the argument. X2

The AI writer's own judgement (2): assignments by analogy. The chain-end assignment of the 1735 cm⁻¹ shoulder in amorphous indomethacin rests on the analogy with liquid acetic acid and on the α band at the same position; the α chain model in turn rests on band coincidence and unpublished unit-cell data. C20·C17·X10 Two different bands, 1735 and 1726 cm⁻¹, are both called non-hydrogen-bonded acid carbonyl, and how one becomes the other between 5 and 20% PVP is not shown. X4 The comparison with methylpyrrolidone, which carries much of the "solution" argument, gives no compositions or measure of "magnitude". X6

The AI writer's own judgement (3): what is cited rather than measured. Phase homogeneity rests on a single Tg reported in earlier work; the X-ray patterns of this paper exclude crystals but cannot distinguish one amorphous phase from two. X9·R17·R1 The crystallisation role of the hydrogen bond is not tested here. X8 The preparation of the physical mixes, the sample temperature and water content at measurement are not stated, and the IR samples were ground with KBr, whose effect on the amorphous samples is not reported. X7 Smaller text inconsistencies do not affect the conclusions: the acetic acid shoulder is given as 1759 cm⁻¹ in Table 1 and 1758 cm⁻¹ in one sentence, and the text refers to the "pyrrole ring" of PVP where Figure 3 shows a pyrrolidone ring. X1·X3

Conclusion

Read against amorphous rather than crystalline indomethacin, the IR and Raman spectra show that amorphous indomethacin keeps mainly the cyclic acid dimers of the γ crystal, with a small proportion of chain ends, while α-indomethacin is bonded differently, probably in chains. C19·C20·C17 In solvent-evaporated dispersions with PVP, but not in physical mixes, the dimer band falls, a free acid C=O band appears at 1726 cm⁻¹ and a hydrogen-bonded PVP carbonyl band at 1636 cm⁻¹, a pattern reproduced by acetic acid and indomethacin in methylpyrrolidone; the authors conclude that the acid OH of indomethacin hydrogen-bonds to the PVP carbonyl at the expense of dimers. R10·R11·R12·R15·R16·C35 Binding is incomplete even with PVP in excess, the system is regarded as one solution-like phase, and the bond is offered as an explanation for crystallisation inhibition that this paper does not test. C37·C42·C45·X8