How the chain length of PVP decides when an amorphous solid dispersion stops releasing congruently

Background and motivation

Amorphous solid dispersions (ASDs) are molecular mixtures of a poorly soluble drug and a polymer. The usual explanation of their benefit is the higher solubility and faster dissolution of the amorphous drug, but the authors point to evidence that ASDs also release drug by mechanisms of their own, which "may" improve bioavailability further; if so, two physically stable ASDs of the same drug could still perform very differently in vivo. C4

Two release behaviours, seen mainly at low drug loading (DL), are behind the benefit. First, drug and polymer can dissolve congruently, that is together and in the same proportion as in the ASD; the release rate is then set by the polymer, can be several orders of magnitude above that of the neat amorphous drug, and is comparable to that of the neat polymer (Indulkar 2019; Que 2019; Saboo 2019). Second, release is fast and complete even under non-sink conditions: once the dissolved drug exceeds its amorphous solubility, the solution forms drug-rich nanoparticles by liquid–liquid phase separation (Ilevbare and Taylor 2013). C5 Such nanoparticles raised bioavailability in an earlier in vivo study, "possibly" through a reservoir effect and particle drift, and congruent release is therefore described as possibly "the optimal release outcome" for an ASD. C6 Drug-rich nanoparticles whose glass transition temperature (Tg) lies below the experimental temperature are called nanodroplets. C11

The benefit disappears above a drug loading called the limit of congruency (LoC). The transition is sharp: release drops toward that of the neat amorphous drug ("falling off the cliff"), as seen for ASDs of PVP, PVPVA and HPMC and, to a lesser extent, HPMCAS. Across drug–polymer systems the LoC lies between 5 and 40% DL. C7·R13 Two factors were known to move it: drug–polymer interactions (Que 2021; Saboo 2020) and the drug's Tg (Hiew and Taylor 2022). C2 In water an ASD forms a viscous surface gel layer, as a dissolving polymer does, and the drug–water incompatibility inside this layer complicates its erosion. C8 Imaging work from the same group showed that in copovidone ASDs of low-Tg drugs the drug-rich phase forms stable nanodroplets in the gel layer below the LoC, but a continuous drug-rich matrix above it, which sits on the surface and blocks release (Deac 2023; Deac 2024; Yang 2023). This paper names that mechanism Failure Mode II. C9

The authors set out to widen the list of factors and failure modes behind the LoC. They chose polymer molecular weight because it strongly shapes the gel layer and so "could have implications" for the LoC. PVP was used because it is sold in many molecular-weight grades, and clotrimazole because it does not interact strongly with PVP (Que 2020), which removes drug–polymer interaction as a confounding factor; clotrimazole also crystallizes slowly. C10 The release of drug and polymer from compacts was measured for six grades at several loadings, and the phase behaviour at the ASD–solution interface was imaged by confocal fluorescence microscopy. A1·A2·P5·P8

Background concepts

Congruent release, the LoC and the gel layer. When water reaches a glassy ASD it turns the surface into a rubbery, viscous gel. The gel layer is bounded on the inside by the glass–gel boundary, where the dry glass starts to swell, and on the outside by the gel–solution boundary, where the polymer is dilute enough for its chains to leave into the medium. In the paper's view, anything held in the gel (dye molecules, drug-rich droplets) leaves at the gel–solution boundary together with the polymer chains; this is why release from a congruent ASD looks like the release of the neat polymer. C31·C34 Water entering the gel can also open a miscibility gap: the gel splits into a drug-rich hydrophobic phase and a water- and polymer-rich hydrophilic phase. If the polymer stays mainly in the water-rich phase, the separation is called segregative; if it follows the drug, associative. Segregative separation is typical of weakly interacting drug–polymer pairs. C28·C29

Two morphologies of the drug-rich phase. While the drug-rich phase is the minority it forms droplets dispersed in the continuous water-rich phase; this is the morphology of congruent release. As the drug loading rises, the volume of the drug-rich phase rises too, and above a certain volume the morphology inverts: the drug-rich phase becomes continuous and the water-rich phase is dispersed in it. This phase inversion is Failure Mode II. A continuous drug-rich phase cannot be carried out of the viscous gel and, on the surface, it acts as a barrier to further water penetration. C39

Polymer chain entanglement. The new element of this paper is polymer physics. Long, flexible chains that overlap enough form entanglements, a transient network that raises viscosity sharply. Short or stiff chains cannot entangle. The critical molecular weight for entanglement can be estimated from the chain's flexibility and its monomer mass (Wool 1993): C14

In plain terms: a chain must be roughly 30 times longer than its "stiffness length" before it can entangle. $M_c$ is the critical molecular weight for entanglement, $C_{\infty }$ the characteristic ratio (a measure of chain stiffness; 14 for PVP, taken from the literature) and $M_0$ the monomer molar mass (111 g/mol for PVP). With these values $M_c$ = 46,620 g/mol, and the equation reproduces this number exactly. E1·M2–M7 Entanglement also needs enough polymer: when the chains are diluted by water or drug below a critical concentration $c_e$, they disentangle (Graessley 1980): C14

In plain terms: the longer the chain relative to $M_c$, the lower the polymer concentration at which it still entangles. $\rho$ is the dry polymer density (not printed) and $M_w$ the weight-average molecular weight of the grade. Table 1 gives $c_e$ as a volume fraction, which equals $M_c/M_w$ (for PVP K29/32, 46,620/58,000 = 0.80). E2·M5

The size of the mesh formed by the entangled chains is the reptation tube diameter $d_T$, the distance between entanglement points: C14

In plain terms: the mesh grows with the square root of the number of monomers between entanglements. $b$ is the statistical segment length, about 6 Å for most linear polymers, which gives $d_T$ = 8.7 nm for undiluted PVP. The results section prints the segment length as "~6 nm"; only 6 Å reproduces the 8.7 nm, so "nm" there is a misprint. E3·X1 On dilution the mesh opens, up to a maximum at $c_e$: C14

In plain terms: the more the gel is swollen by water, the wider the mesh, until the chains disentangle altogether at $c_e$. The widest meshes are 10, 36 and 80 nm for K29/32, K60 and K90. E4·M5·M6·M7 Finally, the coil size of an unperturbed chain is given by its radius of gyration $R_g$: E5

In plain terms: the coil size grows with the square root of the chain length; $R_g$ runs from 1 nm for K12 to 27 nm for K90. E5·M2·M7 The physical point of these numbers is diffusion: particles much smaller than $d_T$ (water, single drug molecules) move through an entangled polymer as if the entanglements were not there, up to several orders of magnitude faster than the Stokes–Einstein relation predicts, whereas particles much larger than $d_T$ are slowed by the network as Stokes–Einstein predicts. C14

Table A. The six PVP grades (from Table 1 of the paper; MW is the vendor's nominal weight average). Only the three grades above M_c = 46,620 g/mol can entangle. The text says R_g was estimated for K29/32, K60 and K90, but Table 1 lists it for all six grades. M2–M7·C16·X4

Methods in detail

The study has a simple architecture: measure where congruent release fails for each PVP grade, then look at the ASD–water interface of one short grade (K12) and one long grade (K29/32) to see why. The entanglement calculation runs alongside as the interpretive frame.

Materials and dispersions

Clotrimazole (Figure 1a) melts at 141.2 ± 0.3 °C, has a Tg of 28.0 ± 0.3 °C and an amorphous solubility of 10.3 ± 0.6 µg/mL in pH 8.0, 50 mM phosphate buffer; its calculated pKa is 6.26 and logP 5.84, so the buffer keeps it largely unionized. M1·C12 PVP K12 came from BASF and K17, K25, K29/32, K60 and K90 from Ashland. M2–M7 ASDs were made by dissolving drug and polymer in ethanol and removing the solvent by rotary evaporation (70 °C bath, then 80 °C under vacuum for about 5 min, drying overnight). PVP K60 is sold as a 45% aqueous solution, so its ASDs were made via a 1-propanol azeotrope, an 80 °C bath and a 120 °C oven overnight. P1·P2 For imaging, two dyes were added: prodan (0.1 wt% of solids), which goes to the drug-rich phase and is shown blue, and Alexa Fluor 488 (0.003 wt%), which goes to the water-rich phase and is shown red. M8·M9·P1

Release tests and the definition of the LoC

Powder was pressed at 1500 psi for 1 min into a die so that only a flat 0.5 cm² face was exposed (Woods intrinsic dissolution apparatus). The die was rotated at 100 rpm in 100 mL of pH 8.0, 50 mM phosphate buffer at 37 °C. ASD compacts weighed 100 mg and neat PVP compacts 200 mg (250 mg for K12). P5 The paper describes the tests as non-sink; by simple arithmetic (AI writer's note), even a 5% DL compact holds 5 mg of drug, about five times what 100 mL of buffer holds at the amorphous solubility of 10.3 µg/mL. P5·M1·R4 Samples were diluted with ethanol to dissolve the drug-rich droplets; clotrimazole was measured by reversed-phase HPLC and PVP by size-exclusion HPLC, except K12, which is too small to separate from buffer components and was measured by an iodine colorimetric assay (absorbance at 500 nm, linear from 5 to 40 µg/mL). P6·P7·C13 Each test was run in triplicate. R1

The grades and loadings tested were K12 and K17 at 5%, K25 at 20 and 23%, K29/32 at 30, 33 and 35%, K60 at 35 and 38%, and K90 at 38, 40 and 45% DL; the first four grades were followed for 90 min, K60 and K90 for 120 min. A2 Release was called congruent when more than 80% of both drug and polymer dissolved. Because scatter made incongruence hard to detect in some runs, the LoC was defined operationally as the highest drug loading at which drug release exceeded 80%. C36·R2 The paper does not say at which time point this criterion applies. X6

Imaging the interface

Films at high humidity. About 5 µL of an ethanol solution (50 mg/mL solids, with both dyes) was dried on a glass slide, then stored above water in a sealed jar at 37 °C (100% RH). Films were taken out at intervals and imaged by confocal microscopy in a chamber at 37 °C and 100% RH (60× oil objective, 1024 × 1024 pixels). P10 Water vapour hydrates a film slowly, so phase separation and the later evolution of the domains can be followed over hours.

Compacts in buffer. Dyed 10% DL compacts of K12 and K29/32 were fixed to the glass bottom of a dish, covered with a cover slip and imaged at the bottom edge every 30 s after buffer (37 °C, pH 8.0) was added. Each image is 15 stitched frames, giving a strip across the dry ASD, the gel layer and the solution. The K12 compact was fused to the glass by heating to about 170 °C for 30 s, the K29/32 compact by wetting the glass with methanol and drying it off under vacuum. P8·X12

Phase compositions and droplet stability

To mimic the composition inside a gel layer, about 200 mg of a 10% DL ASD was rotated with 300 µL of buffer (ASD:water 40:60) at about 90 °C for 15 min, giving an opaque droplet suspension, then cooled and centrifuged at 37 °C. Both phases were dissolved in deuterated DMSO and quantified by 1H NMR (Bruker 800 MHz, clotrimazole singlet at 6.806 ppm, water at 3.321 ppm, PVP band 1.10–2.39 ppm). P9 The methods state triplicate samples, the figure caption n = 4. X3

For droplet stability, 100 µL of a 10 mg/mL ethanol solution of clotrimazole was injected into 10 mL of medium at 37 °C, without PVP or with 1 mg/mL of PVP K12 or K29/32 (a polymer-to-drug ratio of 10:1), and the droplet size was followed by dynamic light scattering (DLS). P11

Experimental analysis

The LoC rises steeply with chain length

Figure 2 is the core data set: twelve release profiles, each with drug (triangles) and PVP (circles), against the 80% line. In every test the apparent clotrimazole concentration exceeded the amorphous solubility and the medium became turbid, which the authors read as drug-rich nanodroplets forming. R4·C18 Table B condenses the profiles.

Table B. Release of clotrimazole and PVP from 100 mg compacts in pH 8.0 buffer at 37 °C, read from Figure 2 (approximate; mean of n = 3). "Verdict" follows the paper's text and its 80% drug criterion. R1·R2·R12

The short grades fail in a telling way. With K12 at 5% DL the polymer dissolved completely (about 107%) but drug release stopped at about 35%; with K17 the polymer again dissolved fully while the drug stalled at about 74–77%. The authors describe both as incongruent "even at a 5% drug loading". R1·R2 From K25 upward, low loadings released fully and congruently and higher loadings did not. R2 In the paper's summary, incongruent release appeared at ≥5% DL for K12 and K17, ≥23% for K25, ≥35% for K29/32 and ≥45% for K90; at 35% K29/32, 38% K60 and 40% K90 the scatter made incongruence hard to see, which is why the 80% drug criterion was adopted. R2·C35·C36

Figure 7 plots the resulting LoCs against molecular weight: below 5% for K12 and K17, 20% for K25, and about 33, 35 and 38% for K29/32, K60 and K90 (the bars read about 32.5, 35 and 37.5%). R12 The authors call the effect "surprising": the LoC rose rapidly, across a narrow molecular-weight threshold, to over 30%, and the threshold lies close to the computed $M_c$ of 46,620 g/mol; above it the LoC hardly changed. C37·E1 Two qualifications follow from the data themselves. Each LoC is only bracketed by two tested loadings (20–23% for K25, 33–35% for K29/32, 35–38% for K60, 38–40% for K90, below 5% for K12 and K17), and K25, at 34,000 g/mol below $M_c$, already reaches 20%, so the rise spans K17 to K29/32 with K25 in between. X5·X9 Congruently releasing ASDs of the longer grades also released more slowly, following the trend of the neat polymers (Figure S5), which supports the picture of polymer-controlled release; the ASD comparisons, however, are at different loadings. C19·R3·R1

Failure Mode I: droplets clump together with short chains

The imaging compares K12 and K29/32 at 10% DL. Before exposure the dyed films were uniformly magenta, so drug and polymer were mixed. R6·C21 After 3 h at 100% RH both films had separated into blue drug-rich droplets in a continuous red water-rich phase (Figure 3a, c). After 11 h the droplets were larger in both, but in the K12 film groups of touching droplets formed strands, outlined in yellow in Figure 3b, whereas the K29/32 droplets stayed separate (Figure 3d). R7 The authors write that this agglomeration "is likely the mechanism behind the poor release of this system" and name it Failure Mode I. C22

The compacts in buffer show the same contrast at the dissolution front (Figure 5). Water entered the dry ASD quickly and formed a gel layer in which the dyes separated; the glass–gel boundary appears as a sharp colour change and the gel–solution boundary where the red signal falls to background. R10·C25·C32 In the K12 compact, after 2 min, the droplets were dispersed in the inner part of the gel but formed a strand-like, interconnected network near the gel–solution boundary, with denser agglomerates nearest the solution; they gathered at the interface instead of leaving, and this continued over time (Figure S2). R10 The authors' reading: as the water-rich phase around the network was diluted and the polymer diffused out, the network collapsed; it was anchored in the gel layer and "had enough strength to withstand the hydrodynamic forces", so the droplets stayed behind. C26 In the K29/32 compact, after more than 7 min, no agglomeration was seen either in the gel layer or among droplets already released into solution (Figure 5b, Figure S3). R10

Failure Mode II: phase inversion between 30 and 40% DL

Films at higher loadings were imaged after 1 h at 100% RH (Figure 4). For K12 the drug-rich phase was dispersed at 20 and 30% DL and continuous at 40% DL, with the water-rich phase now forming droplets inside it; K29/32 behaved the same at 30 and 40% DL. R8 The authors identify this as phase inversion, Failure Mode II, and conclude that it occurs between 30 and 40% DL for both grades. C23·C40 For K12 this means that two failure modes exist in one system: Failure Mode I already limits release below 5% DL, Failure Mode II appears only between 30 and 40% DL, so, were agglomeration prevented, the LoC "would have been limited by Failure Mode II". C24 One detail of Figure 4 deserves a look: at 30% DL the K12 domains are several micrometres across and many touch, much coarser than the small K29/32 droplets at the same loading, although the text groups both as droplets. R8·X15

Molecular weight does not change the phase compositions

If the two grades produced drug-rich phases of different composition, phase inversion could happen at different loadings. Figure 6 rules this out for K12 and K29/32 at ASD:water 40:60 (Table C). C40·R11

Table C. Composition (wt%, mean ± SD) of the two phases formed from ASD:water 40:60, read from Figure 6 (approximate). The x-axis of Figure 6 labels the polymer "PVPVA"; it is PVP. R11·X2

The droplets are about 80 wt% drug and the water-rich phase about 40 wt% PVP and 60 wt% water in both systems. The authors call the compositions "statistically" similar, which "suggests" that molecular weight had no significant effect on the thermodynamics of the system. PVP stays out of the drug-rich phase, a segregative separation typical of weakly interacting pairs; strongly interacting pairs separate associatively, with the polymer following the drug. C27·C28·C29 No statistical test is named, and the mixtures were heated to about 90 °C before analysis, so they describe this protocol rather than the gel layer during release. X14·X16

PVP does not stabilize the droplets in dilute suspension

The most obvious way a polymer could keep droplets apart is steric stabilization: chains adsorbed on the droplet surface repel each other, and longer chains repel more strongly. DLS tested this. In buffer without PVP the droplets started at about 540 nm and grew to about 1 µm; with 1 mg/mL K12 they started at about 340 nm and grew to about 610 nm; with K29/32, about 290 to 630 nm. PVP made the starting droplets smaller, but the starting size did not depend on the grade, and the normalized diameter grew at the same rate in all three media (Figure S6). R5 The authors conclude that clotrimazole nanodroplets are unstable and that PVP is "not a good stabilizer" in dilute suspension, even at a polymer-to-drug weight ratio of 10:1. C20

Discussion

Two independent ways to lose congruent release

The authors assemble the results into one picture (Figure 7). Water turns the ASD surface into a viscous gel that splits into drug-rich droplets and a PVP-rich water phase; the droplets are held in the gel until the polymer is dilute enough to leave at the gel–solution boundary, and then leave with it. Release is then limited by water penetration and the release of polymer chains, which is why ASDs of longer grades release more slowly. C31·C33·C34 From there the outcome depends on chain length and loading:

In the authors' words, "it appears that there are two independent mechanisms" behind the limit of congruency. Below $M_c$ (K12, K17 and K25) the lower LoC "appears to be caused by nanodroplet agglomeration in the gel layer", a mechanism they describe as newly discovered and distinct from phase inversion. Above $M_c$ agglomeration is inhibited and the LoC is set by the loading at which phase inversion occurs: for K29/32 and "possibly" for K60 and K90, whose LoCs of about 33–38% fall inside the 30–40% inversion window. C42·C38·C41·R12 Because inversion requires a drug-rich phase that is the majority, and an inverted system no longer makes droplets at all, Failure Mode I comes at lower loadings than Failure Mode II, which the authors call "conceptually expected". C48

Note by the AI writer (not discussed in the paper): the Failure Mode I assignment for K17 and K25 and the Failure Mode II assignment for K60 and K90 are inferences. Agglomeration was imaged only for K12, and phase inversion only for K12 and K29/32 films; the 30–40% band drawn across all six grades in Figure 7 is, as its caption says, inferred from those two grades. X7·X8

Why long chains keep the droplets apart: the "fishnet" hypothesis

The authors state that the exact mechanism "has not been fully elucidated" and offer hypotheses. C43 Steric stabilization is set aside: since neither K12 nor K29/32 slowed droplet growth in dilute suspension, "we expect that stabilization through steric repulsion is not the likely mechanism" for the stable droplets in the K29/32 gel layer. C44·R5

The alternative is entanglement. The coincidence of the LoC jump with $M_c$ "gives support to the hypothesis that polymer chain entanglement may be contributing" to droplet stability. C45·R12 Entanglement raises viscosity sharply, but only for objects larger than the mesh: water and single drug molecules, much smaller than $d_T$, move as if there were no entanglements, while droplets much larger than $d_T$ are held back (Figure S8). The authors picture the entangled chains as a "fishnet" in which small objects fall through and large ones are caught. The numbers support the geometry: the droplets are a few hundred nanometres across (Figure 5, Figure S6), while the widest mesh, for K90 at its entanglement concentration, is 80 nm. C45·E4·R5·M7 A second effect may add to this: entangled polymers swell more and form a thicker gel layer, which "is expected to increase the distance between the nanodroplets", so that when the diluted chains finally disentangle the droplets are far enough apart not to agglomerate at once. C46 K29/32 chains are expected to disentangle at a polymer volume fraction of 0.8 or below, which in the authors' view explains why K29/32 gave no stabilization in dilute solution; K12 chains are too short to form a fishnet at all. C47·M5·M2 The conclusion section restates the idea: theoretical considerations "suggest" that the entangled high-MW chains stabilize the droplets by hindering their translational diffusion and increasing their spacing through swelling. C45·C46

Note by the AI writer (not discussed in the paper): for K29/32 the fishnet works only above 0.80 v/v polymer. The water-rich phase of the 40:60 mixtures holds about 39–41 wt% PVP, and the paper reports no polymer fraction for the part of the gel layer where the K29/32 droplets stayed dispersed, so whether those chains were entangled there is not shown. The swelling argument is the authors' answer to this, but it is not measured. X10·R11

What the authors draw for formulation

The authors present the two failure modes as separate design problems. A holistic approach "should consider both mechanisms": if another polymer or additive raised the loading at which phase inversion occurs, the system might become limited by Failure Mode I instead. C50 To avoid Failure Mode I, formulators should avoid conditions that promote agglomeration, even where this runs against accepted practice: smaller droplets, usually sought for absorption, "can also make the nanodroplets more unstable", and a low-MW polymer gives faster polymer-controlled release but can trigger Failure Mode I. C51 Surfactants lower Tg and physical stability but "may prevent" Failure Mode I by stabilizing the droplets, and partly ionized droplet surfaces are more stable, so the drug pKa and the medium pH matter. C52 Failure Mode II "will always occur in unionized conditions as drug loading is increased", is driven by drug–polymer–water thermodynamics and can only be pushed to higher loadings by choosing polymer and additive chemistry; strong drug–polymer interactions, which help storage stability, trigger it at lower loadings, and in highly interacting systems it has been seen at only 5–10% DL. C53·C49·R14

Note by the AI writer (not discussed in the paper): for a formulator, the practical message is that the polymer grade is a formulation variable in its own right. In this system the LoC moved from below 5% to about 38% with grade alone, so an LoC measured with one PVP grade cannot be transferred to another. Longer chains buy loading at the cost of release rate, and the LoCs of the entangling grades already sit in the inversion window, so further gains here would need a change in thermodynamics rather than a longer chain.

Limitations

Limits the authors state. The mechanism by which high molecular weight prevents agglomeration "has not been fully elucidated"; entanglement and swelling are offered as hypotheses. C43·C45·C46 The phase-inversion range for K17, K25, K60 and K90 "was inferred based on data from K12 and K29/32" (Figure 7 caption), and the LoC of K60 and K90 is "possibly" limited by phase inversion. R12·C41 The LoC was defined by an 80% drug-release criterion because scatter made incongruent release hard to detect in some runs. C36

The AI writer's own judgement.

• The release data are direct and replicated (n = 3), and the jump from below 5% to over 30% DL between the short and long grades is far larger than the scatter. This is the solid core of the paper. R1·R12
• Each LoC is bracketed by only two tested loadings, Figure 7 shows bar heights rather than printed numbers, and the 80% criterion was applied without a stated time point to tests of 90 or 120 min in which some profiles were still rising (K90 at 38, 40 and 45% DL). X5·X6
• Failure Mode I rests on images of one grade (K12) at 10% DL, while K12 release was measured only at 5% DL; one compact per grade was imaged, and the two compacts were fused to the glass differently and photographed at different times. X7·X11·X12
• The paper does not report the outcome of its polarized-light check for crystals on compacts that did not release fully, so surface crystallization during release is not formally excluded. X13
• The "narrow threshold" near $M_c$ includes K25 at an intermediate 20%; the step spans 10,000 to 58,000 g/mol. X9
• Minor source inconsistencies: the segment length printed as "~6 nm" on p.5 (6 Å reproduces the calculation), Figure 6 labelling the polymer "PVPVA", triplicate samples in the methods but n = 4 in Figure 6, and R_g listed for all six grades although the text names three. None changes the conclusions. X1·X2·X3·X4
• One drug, one pH (8.0, unionized drug) and one polymer family were tested. Whether the fishnet picture carries over to other polymers or to drugs whose droplets have a higher Tg is open.

Conclusion

The paper shows that the molecular weight of PVP strongly changes the limit of congruency of clotrimazole ASDs: below 5% DL for K12 and K17, 20% for K25 and about 33–38% for K29/32, K60 and K90, with the jump close to the computed entanglement weight of 46,620 g/mol. R12·E1·C37 Confocal imaging identified a second way congruent release fails besides phase inversion: with short chains the drug-rich nanodroplets agglomerate in the gel layer and stay behind, at loadings far below the 30–40% at which the drug-rich phase turns continuous. R7·R10·R8·C54 Because the phase compositions are the same for short and long chains, molecular weight "does not appear" to move phase inversion; instead, the authors propose that entangled long chains hold the droplets apart, a hypothesis they hedge with "may be contributing" and "suggest", and did not measure. C27·C40·C45·C43 For formulation the paper adds a variable, polymer chain length, and a warning that solving one failure mode can expose the other. C50·C55