How strongly different sugars hydrogen bond to PVP in freeze-dried glasses, and why the sugar's Tg seems to matter

Background and motivation

Many proteins are freeze-dried (lyophilized) with sugars. When the sugar dries into an amorphous glass around the protein, the protein is more stable both during drying and in storage. Why this works was still debated when the paper was written: two explanations were in play, hydrogen bonding between sugar and protein, and the slow dynamics of the glassy sugar matrix (Pikal 1994, ref 3; refs 4–7) C2. Both ideas assume that sugar and protein are mixed at the molecular level in one amorphous phase, which raises a thermodynamic question: what makes an amorphous sugar mix with a macromolecule in the first place, and do all sugars do it equally well?

Amorphous solids lack long-range order and behave structurally like frozen liquids, so two amorphous components can form a solid solution, and the language of liquid solutions (miscibility, ideal mixing) is applied to them in modified form (Hancock and Zografi 1997, ref 1; Angell 1995, ref 2) C1. The authors' group had already studied one case in detail. Shamblin, Taylor and Zografi (1998, ref 8) colyophilized sucrose with polymers, including PVP, measured the Tg of the mixtures and used FT-Raman spectroscopy to show that sucrose hydroxyl groups hydrogen bond to the PVP carbonyl group, moving its stretching band to lower wavenumber in proportion to the sucrose content C16. From the Tg data they argued that sucrose–polymer mixing has a positive enthalpy but is carried by an entropy gain, because each sucrose–polymer hydrogen bond is formed at the expense of more extensive sucrose–sucrose bonding C32. The thermodynamic idea behind this comes from polymer blend work by Painter, Coleman and co-workers: when one component hydrogen bonds with itself and the other can only accept hydrogen bonds, hetero bonds can make mixing favourable through entropy even when the enthalpy is unfavourable (Painter 1991, ref 29) C31.

This paper extends the sucrose work to other sugars. PVP is used as a stand-in for a protein, for three reasons the authors give. It is a high molecular weight polymer with a hydrophobic backbone and a hydrophilic side group, a "very simplistic model" for a protein; it has no acidic proton, so it cannot hydrogen bond with itself, while its basic carbonyl group accepts hydrogen bonds, which makes every sugar–PVP hydrogen bond detectable without ambiguity; and both PVP and the sugars show clear glass transitions in DSC, which many proteins do not (ref 9; ref 2) C3. The question is whether disaccharides (sucrose, trehalose, maltose, lactose), a trisaccharide (raffinose) and two maltodextrins differ in how far they hydrogen bond to PVP, and whether that is linked to their thermal properties, as "a simple model system" for sugars used as excipients and protein stabilizers C3.

Background concepts

Glass transition temperature. An amorphous solid changes from a rigid glass to a rubbery or liquid state over a temperature range around its glass transition temperature, Tg. In differential scanning calorimetry (DSC) this appears as a step in heat flow. This paper reports the onset of that step, measured at a heating rate of 20 °C/min, and notes that Tg depends on scanning rate, being generally higher at higher rates (ref 1) P3·C4. For context on what the numbers mean: sucrose, maltose, lactose and trehalose all have the same molecular weight (342), yet their onset Tgs in Table 1 span 75 to 118 °C R1.

Tg of a mixture and the Gordon–Taylor equation. If two amorphous components mix ideally, the Tg of the mixture lies between the two pure-component Tgs and can be estimated by the Gordon–Taylor equation (Eq. 1 of the paper; Gordon and Taylor 1952, ref 10) E1:

In plain terms: the mixture Tg, $T_{g12}$, is a weighted average of the component Tgs, $T_{g1}$ and $T_{g2}$, where $w_1$ and $w_2$ are the mass fractions and $K$ sets how strongly component 2 pulls the average its way. The equation assumes the two components are miscible and that their free volumes simply add E1·C6. The constant $K$ is not fitted here; it is calculated by the Simha–Boyer rule (Eq. 2; Simha and Boyer 1962, ref 11) E2:

In plain terms: $K$ follows from the two Tgs and the two densities, ${\rho }_1$ and ${\rho }_2$, so the Gordon–Taylor line is a prediction with no adjustable parameter. The densities were taken from the literature: sucrose 1.43, trehalose 1.54, raffinose 1.49 and PVP 1.25 M1·M2·M5·M8. The paper prints the unit as kg/m3, but these magnitudes are g/cm3; because only the ratio enters $K$, this misprint does not change the lines X3. A measured mixture Tg below the line is called a negative deviation from ideality. The paper uses the size and sign of the deviation as a gauge of how the components interact, following the earlier sucrose study (ref 8) C6.

The PVP carbonyl band as a hydrogen-bond probe. In a Raman spectrum, the C=O stretching vibration of PVP appears near 1675 cm⁻¹ when the carbonyl is free, as in pure PVP R13.a. When the carbonyl accepts a hydrogen bond from a sugar hydroxyl group, its stretching frequency falls (ref 24) C16. The sugars have no band in this region, so in a mixture the observed band is the sum of free carbonyl and carbonyl hydrogen bonded to sugar C18. Two features of that combined band are followed. Its position moves down as more carbonyls are bonded; the authors take the position as a measure of the extent of hydrogen bonding, on the assumption that the hydrogen bonds are of similar strength for all the sugars, which they justify by the similar hydroxyl chemistry of sugars that all contain a glucose unit C20. Its width at half-height grows when several species contribute (free carbonyl and a range of bonded states) and narrows again when one species dominates C23.

First and second DSC scans. A freeze-dried solid carries a processing history. A "first scan" heats the as-dried sample directly; a "second scan" is taken after heating above Tg and cooling at a controlled rate, which erases that history and drives off residual water P3·P4. Two thermal events in a first scan could mean two phases, or one phase plus a relaxation or softening event. One test the authors use is enthalpy relaxation: a glass held (annealed) below its Tg loses enthalpy, and on reheating shows an endothermic peak at Tg; a single peak points to a single amorphous phase (Bosma et al. 1988, ref 20) C15.

Thermodynamics of mixing. Two components mix only if the Gibbs free energy of mixing is negative. The paper writes it as (Eq. 3; the number is given in the text but not printed beside the equation) E3:

In plain terms: mixing is favoured by a negative enthalpy of mixing, $\Delta H_m$, and by a positive entropy of mixing, $\Delta S_m$, which counts more at higher temperature $T$. The discussion lists three contributions (Patterson 1978, ref 28): the combinatorial entropy, always favourable but small when one component is a large polymer; a free-volume term, unfavourable when the components differ in free volume, because mixing then causes a net contraction; and intermolecular interactions, where ordinary dispersion forces usually oppose mixing and hydrogen bonds favour it C29·C30. For a polymer mixture the combinatorial entropy is too small, so specific interactions such as hydrogen bonds are needed for miscibility C30.

Methods in detail

The study has three parts: thermal analysis of the pure components and of three sugar–PVP series, compared with the Gordon–Taylor prediction; an FT-Raman comparison of the PVP carbonyl band across seven sugars; and checks that the freeze-dried mixtures are single phases and that the drying solvent did not lock in an unusual state.

Materials and sample preparation

Trehalose (as α,α-trehalose dihydrate), maltose monohydrate and raffinose pentahydrate came from Sigma; sucrose (A.C.S. quality) from J. T. Baker; PVP K90, molecular weight 1 000 000, from ISP; and two maltodextrins from V-Labs, called maltodextrin 860 and maltodextrin 1050 after the nominal molecular weights assigned by the supplier, with maltooligosaccharide molecular weights of 540–1500 and 540–2800 M1·M2·M3·M5·M6·M7·M8. Lactose appears in Table 1 and in Figures 5 and 7, but its supplier and form are not given in the materials section M4·X11. Sucrose and raffinose are nonreducing sugars; maltose, lactose and the maltodextrins are reducing sugars, which matters later because reducing sugars can react with amides on heating C13.

All materials were dried in a vacuum oven to remove water of crystallization and adsorbed water, weighed in a drybox below 10% relative humidity, and dissolved together in deionized water at 10% w/v total solids P1. The solutions were frozen on freeze-dryer shelves precooled to −40 °C and primary dried at a shelf temperature of −35 °C; once the product reached the shelf temperature, drying continued for another 24 h, after which the shelf was warmed gradually to a final secondary drying temperature of 60 °C P1. Sucrose–PVP mixtures were lyophilized by the earlier study's procedure (ref 8) P1. No cake collapsed, samples were stored over phosphorus pentoxide at room temperature, and Karl Fischer titration showed less than 0.1% w/w water in every sample R3·R4. All samples were amorphous by X-ray powder diffraction (5–40° 2θ, Cu Kα) and showed no birefringence under crossed polarizers P9·P10·R2.

Contribution 1: Tg of the mixtures against the ideal-mixing line

The problem. If a sugar and PVP mix ideally, the mixture Tg should follow the Gordon–Taylor line; a systematic departure points to non-ideal interactions. The earlier study had found negative deviations for sucrose–PVP, and the question is whether other sugars behave the same way C6.

What they did. DSC samples of 10–15 mg were sealed in crimped aluminium pans with a pinhole and run under dry nitrogen. To erase thermal history and remove residual water, each sample was heated to about 30 °C above its Tg and cooled to 100 °C below it at 20 °C/min, then heated once to 200 °C at 20 °C/min (the "second scan") P3. Tg is the onset of the transition and the mean of three independent samples; for the pure components the standard deviation was within ±1 °C P3·R1. Three further heating–cooling cycles checked for degradation P3. The scanning rate of 20 °C/min was chosen to match the earlier sucrose study C4. Tg–composition curves were measured for trehalose–PVP and raffinose–PVP from 0 to 100 wt% PVP in 10 wt% steps and plotted with the sucrose–PVP data of the earlier study (Figure 1) R6. The Gordon–Taylor line for each pair used the measured pure-component Tgs and the cited densities P6·E1·E2.

Why it works, and its limit. The Gordon–Taylor line has no fitted parameter here, so a deviation from it is a property of the data, not of a fit E2. But the line only says whether the mixture departs from additivity; it does not say which interaction causes the departure. That is why the paper pairs it with spectroscopy.

Contribution 2: the PVP carbonyl band across seven sugars

The problem. To compare how far different sugars hydrogen bond to the same partner, one needs a probe that reports only the sugar–partner bond. With PVP the carbonyl band does this, because PVP cannot hydrogen bond with itself and the sugars have no band in the carbonyl region C3·C18.

What they did. Raman spectra were collected on a Bruker RFS 100 with a 1064 nm Nd:YAG laser at typically 1 W and a liquid-nitrogen-cooled germanium detector, in 180° backscattering, averaging 500–1500 scans over 4000–50 cm⁻¹ at 4 cm⁻¹ resolution, with more scans for samples containing less polymer P7. To avoid water uptake, samples were lightly compacted into glass vials, heated under vacuum at 60 °C for 12 h and measured in the closed vials P7. The position and the width at half-height of the PVP carbonyl band were followed from 0 to 90 wt% sugar in 10 wt% steps for trehalose, maltodextrin 1050, maltose, lactose and raffinose; maltodextrin 860 was also measured but not shown; sucrose peak positions were taken from the earlier study (Figure 4) P7·R13·R14·R15·R16·R17.

Why it works, and its limit. The position of the combined band is a convenient single number. The authors' reasoning is that the shift depends on both the number and the strength of the hydrogen bonds (Moskala et al. 1984, ref 21), and that strength should be similar for all these sugars because their hydroxyl groups are chemically alike, so differences in position reflect differences in the number of bonds C20. The band was not split into free and bonded components, and no fraction of bonded carbonyl was calculated, so the measure is relative C19·X10.

Contribution 3: is each mixture one phase, and did drying from water leave a special state?

The problem. Hydrogen bonding is only meaningful as a mixing signal if the samples are single phases. Two points needed checking: first scans showed more than one thermal event, and a solid made by removing a solvent can carry a miscibility or interaction pattern set by that solvent rather than by equilibrium (Dong and Ozaki 1997, ref 23) C27.

What they did. As-lyophilized samples were heated at 20 °C/min to 200 °C without erasing their history ("first scans"), and trehalose with 0–40% PVP is shown in Figure 2 P4·R10. Trehalose containing 30% PVP, a composition where the two first-scan events are not resolved, was annealed at 110 °C for 6 h, cooled to 25 °C and reheated to 200 °C to count enthalpy relaxation peaks P5·R11. Finally, sucrose–PVP and trehalose–PVP were heated in the DSC to 30° above their Tg to let them mix to equilibrium, cooled, kept over phosphorus pentoxide, and measured again by Raman in the opened pans with about 5 min exposure to air below 20% relative humidity P8·R18.

Experimental analysis

The pure components: Tg does not simply follow molecular weight

Table A. Onset Tg of the pure components by DSC at 20 °C/min, standard deviation within ±1 °C (from Table 1; literature values from refs 13–18 as listed there) R1·M1–M8.

Tg rises broadly with molecular weight, from the disaccharides to the maltodextrins and PVP, but the four disaccharides of identical molecular weight differ strongly: sucrose is about 40° below trehalose and lactose, maltose lies in between, and raffinose, despite its higher molecular weight, is slightly below trehalose R1·C7. The disaccharide values are mostly a few degrees above the literature values, which the authors say "may be explained" by the fast scan and the vented pans that let residual water escape R1·C8. The spread of the literature values is itself large (trehalose 100–119 °C), so these Tgs are best compared with onset values at similar heating rates X13. The maltodextrin values come from first scans, because the maltodextrins degraded at the end of their Tg event and their Tg did not stay constant over repeated cycles; the authors suggest that low molecular weight components "may" cause the degradation R5·C9·X8. PVP has the highest Tg, 176 °C, and "would therefore be anticipated" to raise the Tg of the sugars when mixed with them R1.h·C10.

Mixture Tg falls below the Gordon–Taylor line

After the history was erased, each sucrose-, trehalose- and raffinose–PVP mixture showed a single glass transition at every composition, which the authors take as "confirming" single-phase systems (Figure 1) R6·C11. In all three series the measured Tg lies below the Gordon–Taylor line at every intermediate composition: "negative deviations from ideality" R6·R7·C12.

Table B. Mixture Tg and the Gordon–Taylor prediction, read from Figure 1 (approximate) R6·R7.

Note by the AI writer (not discussed in the paper): the gap is much larger for sucrose (about 26 K at 40 wt% PVP) than for trehalose and raffinose (about 9–11 K), the same order as the Raman shifts described below R6.e·R7. The paper does not compare the sizes. The sucrose curve also comes from a different study, with its own PVP end point of about 451.5 K rather than the 449 K (176 °C) of Table 1, so the comparison crosses two data sets X4. As a check on the drawn lines, the Gordon–Taylor equation with the paper's own inputs (trehalose 118 °C, PVP 176 °C, densities 1.54 and 1.25, giving $K$ = 1.073) predicts 421.2 K at 50 wt% PVP, against about 420 K read from the line in Figure 1b E1·E2.

For the reducing sugars no Tg–composition curves are reported. With maltodextrin 1050 the mixtures gave Tg values above those of both pure components and turned brown, which the authors take to reflect degradation and/or a reaction between the reducing sugars and the amide-containing PVP during heating, so they did not report these data for lactose, maltose or the maltodextrins R8·C13.

Two events in the first scan are not two phases

Before the history was erased, first scans showed two thermal events. For the pure sugars, the first lay near the Tg and the second about 20 °C higher R9. For trehalose–PVP (Figure 2), read from the curves, pure trehalose showed a step at about 112–124 °C and an endothermic dip near 145 °C; with 10 and 20% PVP the step moved to about 113–129 and 117–134 °C and the later endotherm to about 164 and 170 °C; with 30 and 40% PVP only a shoulder near 128–130 °C followed by a step at about 136–146 °C remained R10. After controlled cooling and rescanning, every sample showed one event, the reported Tg R9. Following Shalaev and Franks (1995, ref 19), who assigned the second event in lyophilized sucrose to softening of the freeze-dried material, the authors write that this "clearly explains" the pure-sugar result C14. For the mixtures they ran the annealing test on trehalose with 30% PVP: one enthalpy relaxation peak was seen, "indicating" a single phase, so the two first-scan events "are therefore not thought to be evidence of any phase separation" R11·C15.

The carbonyl band shifts down, most with sucrose

Figure 3 compares the carbonyl band of pure PVP with mixtures containing 10% PVP and 90% maltodextrin, trehalose or sucrose. Read from the figure, the band maximum moves from about 1675 cm⁻¹ for pure PVP to about 1660 cm⁻¹ with maltodextrin and about 1653–1655 cm⁻¹ with trehalose and sucrose, and the mixture bands are broader on the low-wavenumber side; free and bonded bands are not resolved R12. The downward shift is read as hydrogen bonding between polymer and sugars ("indicating") C17. Assuming similar hydrogen-bond distances and geometries in all mixtures, the free-carbonyl contribution "appears" greatest with maltodextrin 1050 and least with sucrose, so sucrose forms the most hydrogen bonds with PVP "at this composition" C19. The Figure 3 and Figure 4 captions call the maltodextrin "1060"; the text and Table 1 make clear it is maltodextrin 1050 X2.

Table C. PVP carbonyl peak position (cm⁻¹) in colyophilized mixtures, read from Figures 4 and 5 (approximate); pure PVP 1675 cm⁻¹; Tg from Table 1 R13·R15·R1.

Across composition (Figure 4), the band does not move up to 10 wt% sugar and then falls steeply between 20 and 60 wt% sugar, levelling off at high sugar content R13. At every composition from 20 wt% sugar on, the shift is largest for sucrose, smaller for trehalose and smallest for maltodextrin 1050; at 40 wt% sugar the positions are about 1663, 1667 and 1671.5 cm⁻¹ R13. The authors state this ranking with its premise: "Assuming that the peak position is an indication of the extent of hydrogen bonding", sucrose hydrogen bonds with the carbonyl more than trehalose, which bonds more than maltodextrin 1050 C21. Maltodextrin 860 produced "essentially the same" composition-dependent shift as maltodextrin 1050 (data not shown) R14·C21. Among maltose, trehalose, lactose and raffinose (Figure 5) there is "very little difference", with maltose giving a slightly larger shift R15·C22. The trehalose values agree between Figures 4 and 5 R13·R15.

The band width tells a second story

The width at half-height of the carbonyl band starts at 34 cm⁻¹ for pure PVP R16.a. With sucrose it rises to about 47 cm⁻¹ at 40 wt% sugar and falls back to about 38 cm⁻¹ at 80–90 wt% (Figure 6) R16. The earlier study attributed this to a shift from mostly free carbonyl at low sucrose content, through a mixture of free and bonded carbonyl, to mostly bonded carbonyl at high sucrose content C24. Trehalose follows a similar pattern but stays broader at high sugar content (about 41–42 cm⁻¹ at 70–90 wt%) R16·C25. With maltodextrin 1050 the band widens later (about 38.5 cm⁻¹ at 20 wt%) but then stays at about 47–48 cm⁻¹ from 50 to 90 wt%, "suggesting a mixture of several species even at high maltodextrin levels" R16·C25. Maltose, trehalose, lactose and raffinose show "a similar dependence on sugar concentration" (Figure 7) R17·C26. Read from Figure 7, lactose is the exception at high sugar content, staying at about 43.5–46 cm⁻¹ at 70–90 wt% while the other three fall to about 40.5–43.5 cm⁻¹; several lactose points at 20–50 wt% are hidden behind other markers R17·X6.

Heating above Tg changes nothing in the Raman spectra

Sucrose–PVP and trehalose–PVP heated to 30° above their Tg showed no change in carbonyl peak position or width: the spectra after lyophilization and after heating were identical R18. The authors note that, although the DSC thermograms depend on thermal history, the Raman spectra do not, and that these results "strongly support the conclusion that lyophilization of these mixtures produces single phase systems" C28.

Putting Tg and hydrogen bonding side by side

Set against Table A, the Raman data give the paper's headline. Sucrose, with the lowest Tg (75 °C), shows the largest shift; maltose (100 °C) a slightly larger shift than lactose, raffinose and trehalose (114–118 °C); and the maltodextrins (169 °C) the smallest R1·R13·R15. The abstract says the extent of hydrogen bonding "was found to correlate inversely" with the Tg of the sugar; the discussion says it "appears to vary inversely"; the conclusions say the differences "could be correlated inversely" C35. The relation is stated in words; the paper shows no plot of shift against Tg and gives no statistic C35.

Discussion

The authors' explanation: self-association against hetero bonds

The discussion builds a qualitative thermodynamic picture from the Painter–Coleman framework and the earlier sucrose work. Each step below is the authors' argument; the quantities in it (numbers of sugar–sugar bonds, enthalpy and entropy of mixing) were not measured.

In more detail: the authors start from the observation that hydrogen bonds formed with every sugar examined C33. The negative deviations of the mixture Tg from ideality are "consistent with the idea that mixing is driven by an increase in the entropy of the system", brought about by an overall loss of hydrogen bonds as sugar–PVP bonds replace sugar–sugar bonds, as proposed earlier for sucrose C34·C32. Because the extent of hydrogen bonding "appears to vary inversely" with sugar Tg, "it appears that PVP is more able to disrupt the self-association of amorphous sucrose than it is the other sugars" C35·C36. Since hydrogen bonding is one factor that sets Tg (Angell et al. 1978, ref 30), "we may speculate" that the much lower Tg of sucrose compared with the other disaccharides comes from a less extensive hydrogen-bond network in the amorphous state C37. For more extensively bonded, higher-Tg sugars "it might be predicted" that the balance between the entropy gain and the larger positive enthalpy from the net loss of hydrogen bonds yields fewer sugar–PVP bonds C38.

Two further factors are weighed. Steric effects of larger sugar molecules "may also play a role" in reducing hydrogen bonding, but cannot explain the differences among the disaccharides of equal molecular weight C39. Against this, the smaller free volume of a higher-Tg sugar reduces its free-volume mismatch with the polymer, which favours mixing C40. The authors' summary is that all the sugar–PVP systems are single phase, but "the factors contributing to the driving force for mixing clearly change" as the sugar changes C41. Extrapolating, as sugar Tg rises further, self-association should dominate and phase separation "should eventually occur"; dextran, with a Tg of around 225 °C, does phase separate completely from PVP (refs 8, 31) C42·R19.

The conclusions state the thermodynamic argument a little more firmly: "a qualitative analysis of the thermodynamics of mixing indicates" that the specific interactions contribute to miscibility, and an increase in entropy "is predicted"; and the results "appear consistent with the idea" that a sugar's Tg reflects "to some degree" its level of self-association C34·C36.

From PVP to proteins

The authors then ask what the trend might mean for sugar–protein mixtures, with explicit hedges. Proteins are charged and structured, but there are parallels: hydrogen bonds between proteins and sugars in lyophilized mixtures have been reported (Carpenter et al., ref 32), both PVP and proteins inhibit crystallization of amorphous sucrose (refs 33, 34), and both PVP and some proteins phase separate from dextran (refs 8, 35); so "it is not unreasonable to speculate that a similar balance of forces will influence mixing" in sugar–protein systems C43. Since miscibility is considered a prerequisite for protein stabilization (Pikal, ref 3), "it is therefore likely" that a specific sugar–protein hydrogen-bonding interaction is needed for a single phase, with the protein able to disrupt the excipient's self-association and vice versa C44.

They add a point about mobility. The mixture Tg is a macroscopic indicator of mobility, but not a full description of molecular dynamics: sucrose and sucrose containing 10% PVP show different enthalpy relaxation although their Tgs are the same (Shamblin thesis, ref 36) C45·R20. "It would seem reasonable" that extensive sugar–protein hydrogen bonding couples the protein to the sugar's motions, so protein dynamics would depend on that coupling and on the storage temperature relative to the mixture Tg; both hydrogen bonding and Tg "are likely to be important determinants of stability", which "may explain" why the stability of some proteins on drying does not track the Tg of the excipient (Tanaka et al., ref 37) C46. Because hydrogen bonding with PVP rose as the stabilizer Tg fell, the authors close by noting that the relative importance of hydrogen bonding versus Tg in sugar–protein systems "will need further study" C47.

Note by the AI writer (not discussed in the paper): for a formulation scientist the result points to a trade-off in excipient choice. In this model, the sugar with the highest Tg, usually preferred for a stiff, slow glass, forms the fewest hydrogen bonds with its acceptor partner, while sucrose forms the most but has the lowest Tg. The same reasoning may be worth testing for drug–polymer amorphous dispersions, where a strongly self-associated component may mix less well with an acceptor polymer even when a single Tg is seen.

Limitations

Limits the authors state

The authors flag that Tg values depend on scanning rate "to some extent" and chose 20 °C/min for consistency with the earlier study C4. They treat the peak position as a measure of the extent of hydrogen bonding only under an assumption of similar hydrogen-bond strength ("It is reasonable to assume") and similar bond geometries C20·C19. They did not report Tg–composition data for the reducing sugars because of degradation or reaction on heating C13. Their thermodynamic analysis is called "qualitative", and the extension to proteins is framed as speculation that "will need further study" C34·C43·C47.

The AI writer's own judgement

What the data support firmly. The single-phase status of the mixtures rests on three independent observations: amorphous by X-ray diffraction and microscopy, one Tg after erasing thermal history, and Raman spectra unchanged by heating above Tg R2·R6·R18. The negative deviation from the Gordon–Taylor line is clear for all three nonreducing sugars R6·R7. The ranking sucrose > trehalose > maltodextrin 1050 in carbonyl shift holds at every composition from 20 wt% sugar, by up to about 8 cm⁻¹ R13.

Where to be careful. First, the sucrose series in Figure 1 and Figure 4 is taken from the earlier study, not measured alongside the other sugars, and the Figure 1 caption even cites the wrong reference (ref 4, a 1984 book chapter, instead of ref 8); the headline comparison of sucrose with the others therefore crosses two data sets X4·X1. Second, the "inverse correlation" with Tg is a rank of three groups (sucrose, the other disaccharides with raffinose, the maltodextrins), not a relation that holds within the middle group: raffinose (Tg 114 °C) shifts slightly less than trehalose (118 °C) at 30–40 wt% sugar, and at 90 wt% sugar trehalose, lactose and raffinose all sit at about 1657 cm⁻¹, differences of 1–2 cm⁻¹ against a nominal resolution of 4 cm⁻¹ with no replicates shown X5·X10. Third, the maltodextrin Tg comes from first scans with degradation, unlike the other Tgs X8. Fourth, the annealing test that rules out phase separation in first scans was run on one composition only X9. Fifth, the self-association explanation is untested here: neither sugar–sugar hydrogen bonding (for example in the hydroxyl stretching region) nor the enthalpy and entropy of mixing were measured, so the thermodynamic story is consistent with the data rather than shown by them X7. Finally, some printed details need correcting when the paper is reused: the density unit (kg/m3 for values that are g/cm3), "maltodextrin 1060" in two captions, the fructose unit printed as αD-fructofuranoside in the sucrose and raffinose names, and the missing source of lactose; none of these changes the data X3·X2·X12·X11.

What is not covered. No protein was examined, and all samples were dry; the effect of water, which lowers Tg and competes for hydrogen bonds, is outside the study R3·C43.

Conclusion

The paper shows, for PVP K90 colyophilized with seven sugars, that every sugar hydrogen bonds to the PVP carbonyl and forms a single-phase amorphous mixture, and that the three nonreducing sugars give mixture Tgs below the ideal Gordon–Taylor line C33·C12. Its main comparative finding is the size of the carbonyl shift: largest with sucrose, intermediate and similar with maltose, lactose, raffinose and trehalose, smallest with the maltodextrins, an order the authors describe as inverse to the sugar's Tg ("appears to vary inversely") R13·R15·C35. The explanation offered, that a higher Tg marks a more self-associated sugar whose network PVP finds harder to break, is presented with hedges ("we may speculate", "appear consistent with the idea") and was not measured C36·C37·X7. Read with those limits, the study gives a clean early example of using a non-self-associating acceptor polymer and its carbonyl band to compare how different hydroxyl-rich excipients interact with a partner, and it frames the question, still relevant to lyophilized and amorphous formulations, of how hydrogen bonding and Tg each contribute to stability C47.