How chain length decides whether dextran and PVP mix

Background and motivation

Whether two amorphous components form one phase or two matters in two kinds of products. In an amorphous solid dispersion, a drug that is miscible with its polymer carrier is considered more likely to stay amorphous for a long time; in a phase-separated system the drug-rich phase has less polymer available to stabilise it and is expected to crystallise more readily (Van Eerdenbrugh and Taylor, refs 2 and 3; Baird and Taylor, ref 4). C1 In freeze-dried products, polymers and small molecules are added to protect an active ingredient such as a protein during freezing, drying and storage, and to give a glass whose glass transition temperature (Tg) lies above the storage temperature (Randolph, ref 5). C2

The standard test for miscibility is differential scanning calorimetry (DSC): a single glass transition is read as a miscible blend and two transitions as phase separation. C3 The authors list the known weaknesses of this test (Baird and Taylor, ref 4). Components with similar Tgs cannot be resolved; domains below a minimum size, about 10–50 nm for polymer blends, go undetected (refs 6 and 7); broad transitions with small heat-capacity steps are hard to identify; and the scan itself heats the sample through its Tg region, which may change a miscibility that depends on temperature. C3·R18 Other methods are therefore being explored: Fourier transform mid-infrared spectroscopy (FTIR), X-ray powder diffraction with pair distribution functions, solid-state NMR, atomic force and electron microscopy, Raman mapping, local thermal analysis and nanoscale infrared spectroscopy (refs 8–24). C3 To show that any such method really detects miscibility, the authors argue, it should ideally be tested on miscible, partially miscible and immiscible systems made of the same constituents, so that chemistry does not change between samples. C8

Dextran (DEX) with poly(vinylpyrrolidone) (PVP) had been studied before, both in freeze-concentrated solutions (Izutsu and co-workers, refs 25–27) and in the solid state. C4 In the solid state the chains were long, and the blends separated: dextran of 520 kDa with PVP K 90 (1,000 kDa) was completely immiscible by DSC (Shamblin, Taylor and Zografi 1998, ref 28); dextran of 64–76 kDa with PVP K 90 was completely immiscible by X-ray powder diffraction (Newman 2008, ref 7); and Padilla 2011 (ref 12) found phase separation dominant across all compositions for both dextran of 8.5–11.5 kDa with PVP of 29 kDa and dextran of 4–6 kDa with PVP of 10 kDa. R17 One exception came from a closely related polymer: freeze-dried PVP K 90 with maltodextrins of 860 and 1,050 Da formed miscible blends (Taylor and Zografi 1998, ref 29). R17 In that study PVP hydrogen bonded less with maltodextrin than with sucrose, which was read as a difference in the extent of bonding, not in its strength. R19

The present paper builds on that contrast. Dextran and maltodextrin are both chains of D-glucose: the dextran backbone is linked 1–6 with about 5% short 1–3-linked side chains, while maltodextrin is linked 1–4 with 1,6 branches (Figure 1). M6·M7 The authors therefore treat maltodextrin as a short dextran and use "DEX" for both. C5 Both polymers are amorphous, so their blends cannot crystallise, and the only question is whether they form one amorphous phase or two. C6 Because molecular weight is a key factor in polymer miscibility (Patterson and Robard 1978, ref 30), changing only the chain lengths should give systems of varying miscibility at nearly constant chemistry. C7 The paper maps this for four DEX grades and five PVP grades over the whole composition range by DSC, measures DEX–PVP hydrogen bonding by FTIR with curve fitting, and interprets the trends with Flory–Huggins mixing thermodynamics. A1·A2·A3·A4

Background concepts

The polymers and their names. The four DEX grades are a maltodextrin of 783 Da (DEX1) and dextrans of 6,000 Da (DEX6), 40,000 Da (DEX40) and 400,000–500,000 Da (DEX450). M6·M7·M8·M9 The five PVP grades are Kollidon K 12, K 17, K 25, K 30 and K 90, called PVP12 to PVP90, with manufacturer molecular weights of 2,000–3,000, 7,000–11,000, 28,000–34,000, 44,000–54,000 and 1,000,000–1,500,000 Da (Table I). M1·M2·M3·M4·M5 Throughout, "DEX weight fraction" means the share of DEX in the total polymer mass of a blend. P1

Reading miscibility from glass transitions. A miscible blend has a single Tg that lies between those of its components. If the blend separates into two amorphous phases, each phase shows its own Tg. P3 Because each phase is a mixture, its Tg lies between the two pure-component values, so the authors assign the lower of two Tgs to the PVP-rich phase and the higher to the DEX-rich phase, judged from where the pure polymers lie. C17 Read from the PVP25 series, pure PVP25 has a Tg of about 172 °C, DEX1 about 136 °C, DEX6 about 205 °C, DEX40 about 224 °C and DEX450 about 229 °C. R1·R4·R7·R10 The maltodextrin therefore has a lower Tg than PVP25, while all three dextrans have higher ones. R1·R4·R7·R10

The PVP carbonyl band as a probe of hydrogen bonding. PVP carries a carbonyl group that can accept a hydrogen bond but has no hydrogen-bond donor; its "carbonyl" band is in fact a mixed mode of carbonyl and N–C stretching (refs 31–33). C10 In pure PVP the band peaks at 1680 cm−1. When a donor such as a DEX hydroxyl bonds to the carbonyl, the bond weakens the carbonyl vibration and a second band appears at lower wavenumber, at 1661 cm−1 in these blends. R2·C12 DEX, in contrast, hydrogen bonds with itself, both within and between chains, so a DEX–PVP bond replaces a DEX–DEX bond and the DEX hydroxyl band changes in a more complicated way; the authors therefore rely mainly on the PVP carbonyl band. C10

Mixing thermodynamics. The paper uses the Flory–Huggins lattice model. Mixing is spontaneous only when the Gibbs free energy of mixing is negative (Eq. 2):

In plain terms: the free energy of mixing $\Delta G_{mix}$ is the balance of the enthalpy of mixing $\Delta H_{mix}$, which counts the energy of the new contacts, and the entropy of mixing $\Delta S_{mix}$, which counts the gain in randomness, at absolute temperature $T$. E3 In the simplified lattice model the enthalpy of mixing per lattice site is (Eq. 3):

In plain terms: $k$ is the Boltzmann constant, ${\varphi }_1$ and ${\varphi }_2$ are the volume fractions of the two polymers, and the Flory–Huggins interaction parameter $\chi$ measures how much more or less favourable unlike contacts are than like ones; a positive $\chi$ means that mixing costs energy. E4 The entropic contribution per lattice site is (Eq. 4):

In plain terms: $N_1$ and $N_2$ are the numbers of lattice sites that one chain of each polymer occupies. Because each term is divided by its chain length, long chains gain almost no entropy by mixing; the expression is always negative, so this combinatorial entropy always favours mixing, but less and less as the chains grow. E5·C32 The authors call the Flory–Huggins equation "a good starting point" for rationalising molecular-weight effects, while acknowledging its limitations (refs 35–37). C31

Methods in detail

The work has three parts: a miscibility map built from DSC on freeze-dried blends, a measure of DEX–PVP hydrogen bonding from FTIR on spin-coated films, and a Flory–Huggins calculation of the entropic part of the free energy of mixing. The same aqueous solutions fed both measurements. P1·P2·P3·P4·P5·P6

Contribution 1: a miscibility map for 20 polymer pairs

① Problem. Earlier solid-state studies each looked at one or two dextran–PVP pairs, all of long chains, so they could not show how miscibility changes with chain length. R17·C4 A map across many grades needs one preparation and one read-out for every pair.

② How. Each DEX grade and each PVP grade was dissolved in water, and the pure solutions were mixed by volume to give 10% (w/v) total solids at DEX–PVP weight ratios from 10/90 to 90/10 in steps of 10. P1 The solutions were freeze-dried in 2 ml vials with the stoppers resting loosely on top: freezing at −2 °C for 0.5 h and −40 °C for 1 h, primary drying at −35, −20, −5 and 5 °C under 100 mTorr, and secondary drying at 20 °C under 150 mTorr, 36.5 h in total (Table II); the vials were then stored under vacuum overnight to remove residual water. P2 Samples in pin-holed aluminium pans were heated at 10 °C/min to 250 °C (200 °C for blends containing DEX1), held for 1 min, cooled to 40 °C at 10 °C/min, held again and reheated at 10 °C/min. P3 The Tg was taken as the midpoint on the second heating, except for DEX1 blends, which often degraded at higher temperature and were read on the first heating. P3 One Tg event classed a composition as miscible and two as phase separated; Table III records every cell as either an actual measurement or a value inferred from neighbouring compositions and molecular-weight trends. P3·R13

③ Why it works. All 180 blend compositions share the same chemistry, preparation and DSC protocol, so differences between them can be traced to the chain lengths. A1·A2·A3 The read-out is the simplest one available, the number of Tg events. P3 Six of the 180 cells were inferred rather than measured: the first three DEX1–PVP12 compositions, the 0.9 cells of DEX1–PVP25 and DEX1–PVP90, and DEX40–PVP12 at 0.5. R13·X9

④ Difference from earlier work. Instead of testing a single long-chain pair, the paper spans a factor of about 600 in DEX molecular weight and about 500 in PVP molecular weight (783 Da to 400–500 kDa; 2–3 kDa to 1,000–1,500 kDa), which is what turns "dextran and PVP do not mix" into a question of chain length. M1·M5·M6·M9 Note by the AI writer: the factors are ratios of the printed molecular weights, not values printed in the paper.

Contribution 2: putting a number on DEX–PVP hydrogen bonding

① Problem. Shifts of the PVP carbonyl band in the presence of a hydrogen-bond donor are well documented, but comparing spectra of 20 blends by eye is not quantitative. C10 The authors wanted a single number per sample.

② How. The solutions were spin-coated onto KRS-5 windows at 8,000 rpm for 20 s and kept overnight under dry nitrogen (below 10% relative humidity); transmission spectra from 500 to 4000 cm−1 were collected at 4 cm−1 resolution with 128 scans while the sample compartment was flushed with dry air. P4 Spectra were taken at DEX weight fractions of 0.2, 0.5 and 0.8 (plus both pure polymers) for all 20 pairs, and at every ratio for DEX1–PVP12. P4 DEX1 has a band near 1600 cm−1 that overlaps the carbonyl region, so the pure DEX1 spectrum was subtracted from DEX1 blends with a suitable factor; for the other DEX grades this band was negligible. P5·C9 Each carbonyl band was normalised to unit area and fitted, between 1590 and 1750 cm−1, as a weighted sum of two reference bands whose weights add up to 1: the pure-PVP band (free carbonyl) and the band of the 0.9 DEX1–PVP12 blend, taken as the maximally hydrogen-bonded state. P5 The quality of each fit was expressed as a percentage, the square root of the summed squared residuals divided by the summed intensities; these values are in Table S1. P5 The fitted areas of the free band ($A_F$) and of the bonded band ($A_{HB}$) were converted into a fraction with Eq. 1, taken from Ahn and Jeong (ref 33):

In plain terms: $f_{HB}$ is the share of carbonyls in the bonded state; the ratio of the molar absorption coefficients of the free and the bonded band, $a_F/a_{HB}$, was set to 1.3 from the literature, so that a bonded carbonyl, which absorbs less strongly, is counted with extra weight. E1·P5 Because the "bonded" reference is the most bonded state observed and not a fully bonded one, $f_{HB}$ counts bonded carbonyls relative to the maximum number that can bond, not relative to all carbonyls. C25 The authors expect a fraction of carbonyls to remain free even at the plateau, "likely" through steric hindrance (Garton, ref 34). C15

For DEX1–PVP12, measured at every ratio, the authors also fitted a quadratic curve and drew it in every panel of Figure 6 as a reference:

In plain terms: an empirical curve ($R^2$ = 0.993) that the authors say "has no physical significance"; it only gives a reference line for the other systems. E2 The variable $f_D$ is not defined in the paper; the equation reproduces the plotted curve only if $f_D$ is the DEX content in percent (for example 0.815 at 50%), not the weight fraction shown on the axis. E2·X4

③ Why it works. With only two reference states, every blend spectrum becomes one number, so 20 pairs at three compositions can be ranked on a common scale. P5·R14 The choice of reference fixes the top of the scale: a blend can at most reach the bonding of 0.9 DEX1–PVP12. P5·C25

④ Difference from earlier work. Earlier studies on these polymers reported the presence of bonded carbonyl bands; here the band is decomposed into a fraction that can be compared across molecular weights. C10·R14

Contribution 3: an entropy term for every pair

① Problem. If chain length is what changes, the part of the free energy that depends on chain length should explain the map. C31 The interaction parameter $\chi$, however, is "not that easily accessible experimentally". C33

② How. The authors assume ("if we assume") that $\chi$ is about the same for all grades, because the chemical composition does not change with molecular weight; differences between pairs then come only from the combinatorial term. C33 They calculated that term with Eq. 4 at 25 °C for all 20 pairs, with no enthalpic contribution, a density of 1.2 g/ml for both polymers (so that volume and weight fractions coincide), the PVP monomer as the lattice site and the mean of each molecular-weight range, and plotted it per lattice site against DEX weight fraction (Figure 7). P6·E5·R15

③ Why it works. With $\chi$ held fixed, a deeper (more negative) entropic curve means a larger drive to mix that a given unfavourable enthalpy has to overcome, so the depth of the curve can be set beside the observed miscibility. C32·C33

④ Difference from earlier work. The calculation adds no new data, but it turns "longer chains mix less" into a quantity that can be compared across the 20 pairs. R15 Note by the AI writer (not discussed in the paper): recalculating Eq. 4 with the stated inputs, and with 111.14 g/mol for the N-vinylpyrrolidone repeat unit (a value not printed in the paper), reproduces 18 of the 20 curves of Figure 7; the PVP12 curves for DEX40 and DEX450 are discussed below. E5·X1

Experimental analysis

Experimental setup. The main text shows the DSC data of one PVP grade, PVP25, with each of the four DEX grades (Figures 2a–5a), and the carbonyl spectra of DEX1–PVP12 at every ratio (Figure 2b) and of each DEX grade with all five PVP grades at one composition (Figures 3b–5b). R1·R2·R4·R5·R7·R8·R10·R11 The DSC series of the other 16 pairs (Figures S1, S3, S5, S7) and the remaining spectra (Figures S2, S4, S6, S8) are in the Supplementary Material and are described in the text only. R3·R6·R9·R12 The full DSC result is summarised as a map in Table III, and the fitted hydrogen-bonded fractions in Figure 6. R13·R14 The paper reports no replicate numbers or error bars for these measurements. X9

Table A. Glass transition temperatures (°C) of freeze-dried blends with PVP25; two values = two Tg events (phase separated). All values about, read from Figures 2a, 3a, 4a and 5a. R1·R4·R7·R10

Main result 1: the maltodextrin mixes with every PVP grade. DEX1–PVP25 shows a single Tg at every composition (Table A), and the authors report the same for DEX1 with the other four PVP grades (Figure S1); they read this as miscibility over the complete composition range ("These results suggest"). R1·R3·C11 Within that single Tg, the profile is not a smooth mixing line: it drops from about 172 °C for pure PVP25 to about 161 °C at 0.1 DEX1, stays at about 159–163 °C up to 0.6, and then falls to about 136 °C for pure DEX1. R1 The spectra agree: in DEX1–PVP12 the free carbonyl band at 1680 cm−1 gives way to the bonded band at 1661 cm−1 as DEX1 is added, and between 0.8 and 0.9 DEX1 the band no longer changes (Figure 2b), which the authors read as maximal bonding at these dilutions ("This suggests"). R2·C12·C14 A weak band at 1597 cm−1 appears at 0.9 DEX1; pure DEX1 has a band at 1601 cm−1. R2 The fitted fraction rises from about 0.28 at 0.1 DEX1 through about 0.62 at 0.3 and 0.80 at 0.5 to about 0.96 at 0.7, and is 1.00 at both 0.8 and 0.9 (the 0.9 sample being the reference itself). R14 DEX1 with the other PVP grades lies close to this curve: about 0.43–0.46 at 0.2, about 0.79–0.81 at 0.5 and about 0.91–0.98 at 0.8. R14 The authors call this "in good agreement" with the DSC finding that DEX1 is miscible with PVP independent of its molecular weight. C27

Table B. DEX weight fractions at which two phases were found, transcribed from Table III (compositions 0.1–0.9 in steps of 0.1). Six cells of Table III are inferred rather than measured; Table III marks DEX450–PVP25 at 0.1 as miscible, while Figure 5a shows two Tg events there. R13·X2·X9

Main result 2: longer dextran opens a gap that widens with both molecular weights. With DEX6 the gap appears only for the longer PVPs: DEX6–PVP25 shows two Tg events from 0.5 to 0.8 DEX, DEX6–PVP30 and DEX6–PVP90 behave similarly, the gap shrinks to the single composition 0.8 with PVP17 and disappears with PVP12 (Table B). R4·R6·R13·C16 The gap sits at high DEX fractions, which the authors note as a feature of these systems. C16 With DEX40 the gap widens to five to seven of the nine compositions, and with DEX450 every PVP grade except PVP12 retains at most two miscible compositions, PVP90 none. R13·C22 The authors summarise the map as "the miscibility gap increases with an increase in the MW of both components". C24 Read cell by cell, the map is less regular than this sentence: within DEX6 the PVP90 gap (four compositions) is narrower than the PVP30 gap (five); within DEX40, PVP90 is miscible again at 0.8 and 0.9, so its gap (six) is narrower than those of PVP25 and PVP30 (seven); within DEX450, PVP17 (eight) has a wider gap than PVP25 and PVP30 (seven each). The paper does not comment on these cells. R13·X10

Main result 3: the two phases are unequal. Where two Tgs appear, the upper (DEX-rich) Tg stays close to that of the pure dextran, for example about 205–211 °C against about 205 °C for pure DEX6, and about 216–223 °C against about 224 °C for pure DEX40, while the lower (PVP-rich) Tg moves with composition (Table A). R4·R7·C17 For the other pairs the text reports the same pattern: the PVP-rich Tg rises considerably with DEX fraction for PVP12 and PVP17 with DEX40, and with DEX450 it moves only for PVP12 and PVP17. R9·R12 The authors read this ("This suggests") as PVP being less miscible in DEX than DEX in PVP: PVP has no hydrogen bonds in the pure state and gains them on mixing, whereas DEX has to replace its own hydrogen bonds by bonds to PVP, which "probably" lowers its overall bonding (citing Shamblin 1998, ref 28); the PVP-rich phase therefore takes up more DEX than the DEX-rich phase takes up PVP. C18·C20 The compositions of the two phases were not measured. C18 In the one DEX40 series shown, the PVP-rich Tg at 0.7 and 0.8 DEX (about 204 and 203 °C) lies about 30 °C above pure PVP25, although the text describes the PVP-rich Tgs of the longer PVPs as differing little from pure PVP. R7·X5

Table C. Fraction of hydrogen-bonded PVP carbonyls, fHB, in spin-coated films. All values about, read from Figure 6, except DEX1–PVP12 at 0.8 (1.00, stated in the text); where markers overlap, the shared reading is given. R14

Main result 4: hydrogen bonding falls as molecular weight rises. Table C shows the same order as the map. With DEX6, PVP12 and PVP17 bond about as much as with DEX1 at 0.2 DEX (about 0.47), the three longer PVPs less (about 0.30–0.32); at 0.5 DEX all fall, and at 0.8 DEX the values run from about 0.87 (PVP12) to 0.62 (PVP90). R14·C28 The authors set aside DEX6–PVP25 at 0.5 DEX (about 0.30), which does not follow the other grades and had the largest fitting error of all systems ("appears to be an outlier"). R14·C28 With DEX40 only PVP12 bonds noticeably at 0.2 DEX (about 0.28), and at 0.8 DEX the values fall from about 0.68 to 0.09 between PVP12 and PVP90, "a nice trend"; the spectra show a clear bonded shoulder only for PVP12 and PVP17 at 0.8 DEX (Figure 4b). R14·R8·C29 With DEX450 only PVP12 keeps some bonding (about 0.10, 0.32 and 0.48 at 0.2, 0.5 and 0.8 DEX); for the other grades the fraction stays minimal, at about 0.2 or below even at 0.8 DEX. R14·R11·C30 The authors summarise that blends range "from completely miscible and extensively hydrogen bonding to partially miscible with an intermediate level of hydrogen bonding to virtually immiscible with little to no hydrogen bonding". C23

Table D. Minimum of the entropic free energy of mixing per lattice site, in units of 10⁻²³ J, at 25 °C. Plotted values about, read from Figure 7; the recalculated values in brackets were obtained by the AI writer from Eq. 4 with the stated inputs and are not in the paper. R15·E5·X1

Main result 5: the entropy term shrinks with chain length. In Figure 7 the DEX1 curves are deep and nearly independent of PVP grade (minima about −27 to −21.5 × 10⁻²³ J per site); with DEX6 they are much shallower and PVP12 separates clearly from the rest (about −9.1 against −4.4 to −2.8); with DEX40 and DEX450 all curves are shallower still, and for DEX450–PVP90 the entropic contribution becomes "virtually zero" (Table D). R15 The authors match these curves to the observations: the DEX1 systems, with the most favourable entropy, are fully miscible and bond extensively; the larger difference between PVP12 and the higher grades for DEX6 fits complete miscibility with PVP12 only; and in DEX40 and DEX450 PVP12 again has the deepest curve and the highest miscibility and bonding. C34·C35

Note by the AI writer (from the card's extractor notes, not discussed in the paper). Eq. 4 with the stated inputs reproduces 18 of the 20 curves, for example −27.07 for DEX1–PVP12 and −9.09 for DEX6–PVP12. E5 The two exceptions are the PVP12 curves for DEX40 and DEX450, plotted at about −3.1 and −2.8 but recalculated at −7.07 and −6.76; a PVP molecular weight of 6,000 Da, the DEX6 value, reproduces the plotted curves, so an input slip is likely, though the paper does not say. X1 With the recalculated values, DEX40–PVP12 and DEX450–PVP12 would have deeper entropy curves than DEX6–PVP17 and DEX6–PVP25 while showing wider gaps, so the depth of the curve orders the pairs within one DEX grade but not across grades. X11

A spectroscopic check on DEX–DEX bonding. To support the idea that DEX–DEX bonds are lost on mixing, the authors follow the DEX OH band in the fully miscible DEX1–PVP12 system (Figure 8). C37 Its peak maximum lies at about 3356 cm−1 at 0.9 DEX1 and about 3358 cm−1 in pure DEX1, about 3364 cm−1 at 0.8, about 3370 cm−1 from 0.4 to 0.7, about 3373 cm−1 at 0.1 and 0.3, and about 3383 cm−1 at 0.2. R16 The authors read the overall upward (blue) shift with more PVP as "indicating" fewer DEX–DEX hydrogen bonds as DEX–PVP bonds form. C37 The shift is not monotonic, four compositions share one value at 4 cm−1 resolution, and only one system was measured. X6

Discussion

The authors' explanation. The argument runs in four steps. The combinatorial entropy of mixing is always positive and therefore always favours mixing, so wherever a blend shows a miscibility gap there must be an opposing, unfavourable enthalpy ("it is anticipated"). C32 For DEX and PVP the enthalpy comes from hydrogen bonds: forming DEX–PVP contacts costs DEX–DEX (and PVP–PVP) contacts, and since many blends separate, the enthalpic part "has to be positive"; DEX–DEX interactions are energetically more favourable than DEX–PVP ones, which "may stem from" DEX–DEX hydrogen bonds being stronger and/or more numerous. C36 Holding $\chi$ constant, only the entropy term changes between pairs, and it shrinks quickly as the chains get longer. C33·R15 Hence short chains mix, because the entropy gain outweighs the positive enthalpy, and long chains separate, because it no longer does. C35·C39

The authors frame this carefully in their conclusions: it "was surmised" that the enthalpy of mixing is positive and that the entropy of mixing "rapidly becomes less favorable with increasing molecular weights and therefore cannot offset the unfavorable enthalpy of mixing". C39 The abstract is firmer, stating that the combinatorial mixing entropy "was observed to decrease"; it was in fact calculated, not measured, and neither the mixing enthalpy nor $\chi$ was determined. C39·X7

Scope. The results describe dry binary polymer blends: freeze-dried powders for DSC and films dried under nitrogen for FTIR, with no drug and no water beyond what drying left behind. P2·P3·P4 The two measurements were made on different sample forms: the DSC samples were heated to 200 or 250 °C before the Tg was read on reheating (DEX1 on the first heating), while the FTIR films were never heated. The authors themselves note, in the introduction, that heating through the Tg region may change a temperature-dependent miscibility, so the agreement between the two methods is an agreement between differently treated samples. C3·X8 The hydrogen-bonded fraction is relative to one reference blend (DEX1–PVP12 at 0.9), and the paper does not say whether the free-PVP reference was the matching grade or PVP12 for every pair. P5·X12

What the paper offers. The authors' own conclusion is practical for method development: because chemically similar DEX–PVP systems with tunable miscibility can be made simply by choosing grades, and have now been characterised, they "can be considered as excellent model systems for the evaluation of analytical techniques aiming at characterizing miscibility behavior". C40 This answers the requirement they set in the introduction, that a new method be tested on miscible, partially miscible and immiscible systems of the same constituents. C8

Note by the AI writer (not discussed in the paper): implications for formulation. In a freeze-dried product that contains a polysaccharide and PVP, the grade choice can decide whether the matrix is one phase or two: in this series, every dextran of 6 kDa or more separated over part of the composition range with PVP K 25 or longer, while the 783 Da maltodextrin mixed with all PVP grades. R13 More generally, the data show that two polymers that hydrogen bond with each other can still separate when both chains are long and one of them associates strongly with itself; chain length is then a lever on miscibility alongside chemistry. R13·R14 The paper tested no drug, no humidity and no storage, so how far these trends carry over to drug–polymer dispersions remains open.

Limitations

Limitations stated by the authors. The authors state the limits of DSC itself in the introduction: close Tgs, small domains, broad transitions and the effect of heating during the scan. C3 They call the Flory–Huggins equation only "a good starting point", with "a number of limitations", and note that $\chi$ "is not that easily accessible experimentally", so their comparison rests on assuming that $\chi$ does not depend on molecular weight. C31·C33 They describe the maximal hydrogen-bonded state as not necessarily involving all carbonyls, since a fraction is "likely" unable to bond, and they flag DEX6–PVP25 at 0.5 DEX as an apparent outlier with the largest fitting error. C15·C28 They also restrict the quadratic reference curve to a fitting aid with "no physical significance". E2

The AI writer's own judgement (1): the thermodynamic explanation. The molecular-weight trend in miscibility and in hydrogen bonding is direct and consistent across DSC and FTIR. R13·R14 The explanation is weaker. It rests on two inferences that were not measured, a positive mixing enthalpy and a $\chi$ independent of molecular weight, and on a calculated entropy term. X7·C33 Two of the 20 plotted curves (PVP12 with DEX40 and with DEX450) do not follow from the stated inputs; recalculated, they are more than twice as deep, and with those values the depth of the entropy curve no longer orders the pairs across DEX grades by their observed miscibility, although it still does within each grade. X1·X11 The spectroscopic support for the enthalpy argument, the OH band shift of one system, is non-monotonic and resolution-limited. X6

The AI writer's own judgement (2): the data behind the map. Most of the data are in Supplementary Material that was not available here: Tg series for 16 of the 20 pairs and most spectra. X9 No replicate numbers or error bars are given, and 6 of the 180 map cells are inferred. X9 Table III and Figure 5a disagree for DEX450–PVP25 at 0.1 DEX, and Table III marks as inferred a DEX1–PVP25 point that Figure 2a plots. X2·X3 The statement that miscibility falls with the molecular weight of both components holds for the size of the gap, not for every cell. X10 One statement about DEX40 PVP-rich Tgs does not match the only DEX40 series shown, and the quadratic reference curve uses an undefined variable. X5·X4

Conclusion

The paper makes three points. First, a DSC map of 20 DEX–PVP pairs at nine compositions shows miscibility falling as the molecular weight of either polymer rises, from a maltodextrin of 783 Da that mixes with every PVP grade to a 400–500 kDa dextran that separates from PVP K 90 at every composition. R13·C38 Second, FTIR curve fitting shows DEX–PVP hydrogen bonding falling in step, from the maximum observed in DEX1–PVP12 to about 0.09 at 0.8 DEX for DEX40 and DEX450 with PVP K 90. R14·C23 Third, the authors relate both trends to a Flory–Huggins combinatorial entropy that shrinks with chain length and, they surmise, can no longer offset a positive mixing enthalpy that arises from DEX–DEX hydrogen bonds being replaced. C39·C36

The first two points are measurements; the third is an interpretation built on an assumed constant $\chi$ and an inferred enthalpy, with two of its plotted curves not following from the stated inputs. X7·X1 The lasting use the authors propose is methodological: a set of blends of near-constant chemistry and known, tunable miscibility on which new methods for detecting phase separation can be tested. C40