How water during solvent drying splits itraconazole–HPMC dispersions into two phases

Background and motivation

Making a poorly soluble drug amorphous raises its apparent solubility, because the amorphous form has higher enthalpy, entropy, free energy and volume than the crystal. The advantage is lost if the drug converts back to the crystal, so the amorphous drug is mixed with a polymer to give an amorphous solid dispersion (ASD) (Section 1). Below the crystalline solubility of the drug in the polymer, crystallization cannot occur. Above the amorphous miscibility limit, the dispersion can split into a drug-rich and a polymer-rich amorphous phase, called amorphous–amorphous phase separation (AAPS) (Purohit and Taylor 2015; Zhang et al. 2011), and the drug-rich phase may then crystallize.

The paper rests on a distinction between two origins of phase separation. Thermodynamic separation follows from an unfavourable free energy of mixing and is often analysed with Flory–Huggins theory. Kinetic separation is often solvent-induced: during processing the system passes through a miscibility gap, separates, and stays trapped at the end, "even though mixing is now thermodynamically favored" (Section 1). It is usually attributed to a solubility gap between drug and excipient in the solvent (Paudel and Van den Mooter 2012; Bank et al. 1971). Note by the AI writer (not discussed in the paper): the first case belongs to the drug–polymer pair at that loading, the second to the process, so the same pair might come out single-phase if made differently.

Detection is part of the problem. In differential scanning calorimetry (DSC), one glass transition temperature ($T_g$) is read as miscible and two as separated, but this fails when the $T_g$ values are close, is complicated by surfactants, and can overestimate miscibility because heating can mix a kinetically trapped sample (Baird and Taylor 2012). Powder X-ray diffraction needs pair distribution function analysis, and solid-state NMR detects regions down to about 4 nm but is slow. Fluorescence imaging and spectroscopy, with infrared, thermal and Lorentz contact resonance measurements based on atomic force microscopy (AFM), were introduced more recently (Purohit and Taylor 2015; Li and Taylor 2016). With them, Purohit and Taylor showed that spin-coated itraconazole (ITZ) and hydroxypropylmethylcellulose (HPMC) films made from 1:1 methanol (MeOH) and dichloromethane (DCM) were phase-separated, with the probe pyrene going into the drug-rich phase. They suggested kinetic and thermodynamic origins, but the cause was not established.

The wider ITZ–HPMC literature was unsettled. Dispersions had been made by spray drying, flash evaporation, freeze-drying and solvent evaporation (Engers et al. 2010; Janssens et al. 2008), with differing conclusions on miscibility (Section 4). Purohit et al. (2017) found ITZ–HPMC spray-dried from DCM–MeOH phase-separated at all drug loadings, with no information on solvent water content, and Al-Obaidi et al. (2009) had noted that the root cause of demixing is hard to find with co-solvents. Other work, cited later in the paper, hinted that water matters: more water in the solution or higher humidity gave larger features in cast films, linked to slower evaporation (Deepak and Sundararajan 2011; Jaczewska et al. 2007), and absorbed moisture can drive AAPS in ASDs (Rumondor et al. 2011).

The paper asks whether ITZ and HPMC are "inherently immiscible/partially miscible" or whether phase separation is "induced during processing" (Section 1). The authors kept the pair fixed and changed how much water could reach the drying liquid: air humidity during spin coating, solvent dryness, HPMC grade, and water spiked into a spray-drying solvent. They looked for drug-rich domains by fluorescence imaging, AFM, and bright-field and dark-field transmission electron microscopy (TEM) with elemental analysis, and through the UV absorption and fluorescence of ITZ itself. Spin-coated films were studied first and the findings were then tested on spray-dried dispersions.

Background concepts

Spin coating. A small volume of solution is dropped onto a spinning substrate, spreads into a thin layer and dries within seconds into a film. Because drying happens in the open, the air humidity sets how much water vapour the evaporating liquid meets. The authors use spin coating as a small-scale stand-in for spray drying, which also dries a thin liquid quickly, but as droplets. A slower spin gives slower evaporation.

Drying a mixed solvent. All solutions except one control film used 50:50 v/v DCM–MeOH. ITZ is much more soluble in DCM than in MeOH and has negligible solubility in water (Section 4), and methanol absorbs water from the air. The authors make "the reasonable assumption" that DCM evaporates first, leaving a liquid richer in MeOH and water. Methanol and water do not form an azeotrope, mix incompletely at the microscopic level (Guo et al. 2003), and methanol evaporates preferentially from water–methanol drops (Sefiane et al. 2008); higher humidity also slows water evaporation. The last liquid to leave can therefore be rich in water, a very poor solvent for ITZ.

Kinetic trapping and the two routes to separation. If the drug separates while liquid is still present, later drying may bring the overall composition back into a miscible region. The solid glass, however, has too little mobility to remix, so the structure is "kinetically trapped" and records the drying path, not the dry equilibrium. Separation itself can proceed by nucleation and growth, giving droplets of a minority phase, or by spinodal decomposition, giving a disordered, interconnected two-phase structure (Bates 1991).

How the imaging methods see domains. Fluorescence optical microscopy (FLOM) here uses the fluorescence of ITZ itself (excitation 330–380 nm, emission from 420 nm), so bright domains are rich in ITZ; its resolution is about 300 nm. Contact-mode AFM traces surface height, read from line profiles, and sees only the surface. Lorentz contact resonance (LCR) is an AFM mode that maps stiffness, a stiffer material resonating at higher frequency.

How electron microscopy sees composition. In bright-field (BF) TEM, regions that scatter more electrons look dark, from composition or thickness. In high-angle annular dark-field (HAADF) imaging in scanning TEM mode, brightness rises with atomic number. Energy dispersive X-ray spectroscopy (EDX) gives elemental composition: ITZ (C₃₅H₃₈Cl₂N₈O₄) contains 15.9% nitrogen and 9% oxygen by mass and HPMC no nitrogen, so nitrogen marks the drug. Chlorine was avoided as a marker because HPMC E3 can contain up to 1% w/w sodium chloride. A fast Fourier transform (FFT) of a TEM image of one domain shows sharp spots for lattice planes and only a diffuse halo without positional order.

Liquid crystals. In a nematic phase the molecules point roughly the same way but sit at random positions; in a smectic phase such as smectic A (SmA) they also form layers. A nematic phase shows no lattice planes and "can be misinterpreted as an isotropic amorphous phase" (Section 3.3); between crossed polarizers it can show "Maltese cross" patterns. ITZ forms nematic and smectic phases in bulk and in nanoparticles, self-assembles into liquid-crystalline nanospecies when water is the antisolvent, and reorients after humid heat treatment (Mugheirbi and Tajber 2015; Mugheirbi et al. 2016). A spacing of 28.3 Å matches the length of one ITZ molecule.

Aggregates and spectra. When chromophores sit close together, their optical transitions couple and absorption and emission bands shift or split (Heyne 2016). Head-to-tail J-aggregates fluoresce strongly; face-to-face H-aggregates are strongly quenched; oblique arrangements lie in between. ITZ dispersed molecularly in a polymer cannot aggregate, so a band seen only when ITZ molecules are close together can mark self-associated drug. Emission also depends on the polarity of the surroundings (Willets et al. 2004).

Methods in detail

The work splits into four contributions: spin coating under controlled access to water, with imaging at three resolutions and a comparison of domain size with polymer water uptake; the UV absorption and fluorescence of ITZ as a probe-free marker of self-associated drug, also used on an evaporating solution; the internal order of the domains and the effect of humid storage; and a spray-drying test with known amounts of water. Contributions 1 and 2 carry the main argument, Contribution 3 characterises the domains, and Contribution 4 tests whether the result transfers to a manufacturing process. The paper contains no model or numbered equation, so all of the evidence below is experimental.

Contribution 1: Spin coating with controlled access to water

① Problem. The earlier films were made under unknown humidity, so their domains could reflect the pair or the drying conditions. Telling these apart requires changing the water available while the pair, solvent and spin programme stay fixed. It also requires detecting domains smaller than the roughly 300 nm resolution of FLOM (Section 3.2).

② What they did. ITZ and HPMC METHOCEL E3 at 30:70 w/w were dissolved at 10 mg/mL total solids in 50:50 v/v MeOH–DCM. A 100 μL aliquot was dispensed onto quartz, or onto a TEM grid taped to quartz (Figure SI.1), and spun 3 s at 500 rpm and 30 s at 2000 rpm at 21 ± 0.5 °C (Section 2.2). Humidity was 40 ± 2% RH (ambient), 20 ± 2% RH (nitrogen purge) or 7 ± 1% RH (phosphorus pentoxide around the chuck), and all films were then kept under vacuum for at least 24 h. Variants were HPMC E6 and E50 at 40% RH (higher molecular weight; supplier viscosities 5–7 and 40–60 cP against 2.4–3.6 cP for E3, Table 1); AFFINISOL HPMC HME (modified substitution, better solubility in organic solvents, 15 cP) at 20% RH from DCM–MeOH and from DCM alone; ITZ alone at 40% RH; HPMC E3 alone; E3 from solvent dried over molecular sieve at 20% RH; and a 50:50 film at 40% RH. Films were imaged by FLOM (same 0.125 s exposure for all), contact-mode AFM (several images per sample), BF and HAADF TEM at 200 keV with EDX on at least three dark and three light areas of 100 nm or more, and SEM. Water sorption of the as-received polymer powders was measured gravimetrically at 21 °C and 5, 7, 20 and 40% RH after drying at 60 °C (n = 3; Section 2.6).

③ Why it works. Each variant changes the water that can enter the drying liquid without changing the pair: drier air, dried solvent, a less hygroscopic grade, or no methanol. If domains track these changes, the cause lies in processing. The three imaging methods use different contrast (fluorescence, surface height, electron scattering), so agreement guards against artefacts such as tip contamination in contact-mode AFM, which the authors name (Section 2.5.2.4). The drug's own fluorescence replaces pyrene, which the authors call "preferable" because a dye could change the system (Section 3.1).

④ How it differs from earlier work. Purohit and Taylor (2015) used pyrene and AFM-based methods on films made under unknown humidity. Here humidity is controlled down to 7% RH, grades are varied, and HAADF with EDX adds nanometre-scale composition contrast. Note by the AI writer (not discussed in the paper): the water content of the solution or film during coating was never measured; it is inferred from humidity and from equilibrium sorption of dry powders (card note X4).

Contribution 2: The drug's own spectra as a marker of self-association

① Problem. Imaging shows where domains are, but not whether ITZ in a domain-free film is molecularly mixed. Nor does it show whether separation happens in the liquid during drying or in the solid afterwards. The authors wanted a molecular-level indicator that needs no added probe and also works on an evaporating liquid.

② What they did. UV absorption from 200 to 400 nm was recorded for films on quartz and normalised, since ITZ concentration was the same in all films (Figure 7). From these spectra, 330 nm was chosen as the excitation wavelength for fluorescence, collected at 90° (Section 2.5.1.1).Card check: the slit widths are printed as 5 and 1.5 μm; the setting could not be verified (card note X10). Emission was measured for films, for spray-dried powder brushed onto quartz in one thin layer, and for a 3 mg/mL ITZ solution in DCM–MeOH after 0, 30, 120 and 180 min of evaporation. The turbid solution was examined by dynamic light scattering (DLS; n = 3) and BF TEM. E3 films with 10, 20, 30, 40 and 50% ITZ were spun at 20% RH (Figure 10), and rotary-evaporated ITZ served as a bulk disordered reference (Section 2.4, Figure SI.7).

③ Why it works. A feature present in pure ITZ and in films with domains, but absent from films that imaging shows to be homogeneous, can be assigned to neighbouring ITZ molecules. It can then be read from bulk samples such as spray-dried powder. The evaporation experiment ties the same feature to a time in drying: if it appears in the liquid together with scattering and particles, the drug has separated before a film forms.

④ How it differs from earlier work. Tian et al. (2016) had seen an additional emission peak in immiscible indomethacin films, and earlier ITZ work used pyrene. The authors state that, "to the best of our knowledge", UV absorption shifts had not been used before for drug–polymer phase behaviour (Section 3.5). The evaporation experiment was done without polymer, so it shows how ITZ behaves in the drying liquid but not how HPMC changes that behaviour.

Contribution 3: Order inside the domains and the effect of humid storage

① Problem. A domain without crystal planes need not be isotropic amorphous, because ITZ forms liquid crystals. A second question is whether a film made dry stays single-phase in humid air. Powder X-ray diffraction was avoided because it averages over the film and needs the substrate background removed (Section 3.3).

② What they did. An FFT was taken of one domain in a BF image of the 40% RH E3 film (Figure 4a). A film spun in one step at 350 rpm for 30 s, to slow evaporation, was viewed in polarized light (Figure 4b); its humidity and composition are not stated (card note X11). The 40% RH film was stored 24 h at 40 °C/75% RH and a domain was imaged by TEM and FFT (Figure 4c, d). The 7% RH film was stored the same way, dried 24 h under vacuum and mapped at 19% RH by AFM topography and LCR at 140 kHz, with reference spectra of pure ITZ and HPMC E3 swept from 1 to 200 kHz (Figure 6).

③ Why it works. FFT tests one domain for positional order; polarized light detects orientational order that FFT cannot. LCR separates drug-rich from polymer-rich regions by stiffness without a label: the first resonance of HPMC lies at about 137 kHz and that of ITZ at about 140 kHz (read from Figure 6a), so ITZ is "slightly stiffer" (Section 3.4). ④ How it differs from earlier work. The liquid crystals of ITZ had been described in bulk and in nanoparticles (Mugheirbi and Tajber 2015; Mugheirbi et al. 2016); here the same ideas are applied to domains inside an ASD film. LCR mapping follows Li and Taylor (2016).

Contribution 4: Spray drying with known amounts of water

① Problem. Spin coating and spray drying differ in "the kinetics of solvent removal", droplet versus film geometry, solvent type and environment (Section 3.6). Spray drying, not spin coating, is how solvent-based ASDs are typically made at large scale. A water effect in films must be shown in spray-dried material before it can explain the earlier spray-drying result.

② What they did. A 10 mg/mL solution of ITZ–HPMC E3 in 50:50 DCM–MeOH was spray dried in a Büchi B-290 in closed loop with nitrogen: inlet 75 °C, outlet 46–48 °C, atomizing gas 473 normL/h, aspirator 100%, feed 10 mL/min (Section 2.3). One batch used methanol dried over molecular sieve and anhydrous DCM; three used methanol spiked with 1, 5 or 12% water and HPLC-grade DCM.Card check: whether the percentage refers to the methanol or the whole solvent, and v/v or w/w, is not stated; in the whole 50:50 solvent it would be about half (card note X5). The spray-dried drug:polymer ratio is not printed (card note X6). Only powder from the high-efficiency collector was used, kept 30 min in the collector and tray dried 1 h at 40 °C. Particles were examined by BF TEM on ultrathin grids and by fluorescence; residual solvent was measured by gas chromatography–mass spectrometry (GC–MS).

③ Why it works. Spiking makes water a controlled variable with a dose series, and the dry batch is the negative control. TEM looks inside single particles, while the fluorescence marker from Contribution 2 reads many particles at once. ④ How it differs from earlier work. Purohit et al. (2017) spray dried the same system from DCM–MeOH and found separation at all loadings, with no information on solvent water. This study makes solvent water the varied factor and includes a dry-solvent batch.

Experimental analysis

Setup in brief. The main system is ITZ:HPMC E3 at 30:70 w/w spun from 50:50 MeOH–DCM at 21 °C. The varied factors were humidity during coating (40, 20, 7% RH), HPMC grade (E3, E6, E50, AFFINISOL), solvent (as received, sieve-dried, DCM alone), polymer presence, drug loading (10–50%), water in the spray-drying methanol (none, 1, 5, 12%) and 24 h storage at 40 °C/75% RH. No dissolution or crystallization tests were done.

Result 1: humidity during spin coating decides whether domains form. In FLOM, the 30:70 E3 films spun at 40% and 20% RH showed discrete fluorescent domains, and the 7% RH film showed none at about 300 nm resolution (Figure 1a–c). AFM gave domain heights of 150–190 nm at 40% RH and 45–54 nm at 20% RH, while the 7% RH film was flat within about ±6 nm (Figure 2a, e, h). BF TEM showed the 7% RH film as "very homogeneous", "indicating a miscible system" (Figure 3c). The authors conclude "the higher the humidity, the larger the phase separated domains" (Section 3.1). Solvent dried over molecular sieve, spun at 20% RH, did not remove the domains: they merged and their fluorescence fell, "consistent with some level of mixing" (Figure SI.2). Only 7% RH gave a domain-free film (card note X21).

Table A. Water uptake of the as-received polymer powders at 21 °C (Table 1, n = 3) set against AFM domain height (text, Figure 2) and the bars of Figure 5 (read from the figure, approximate). E6 and E50 were measured only at 40% RH.

What Table A means. For E3, uptakes of 0.48, 1.517 and 3.22% w/w go with domain heights of about 0, 45–54 and 150–190 nm. At 40% RH, E6 > E50 > E3 holds for both sorption and height, and domain sizes "seem to increase" in that order (Section 3.1).Card check: for E6 the text gives 200–230 nm, the Figure 2c profile peaks reach about 230–258 nm above the axis zero, and the Figure 5 bar is about 215 nm; the measurement baseline is not stated (card note X2). AFFINISOL sorbed around 63% of E3's uptake at 40% RH, a difference that "seems to be very small" but was statistically significant; the "same trend" at 20 and 7% RH rests on E3 and AFFINISOL only (card note X3).

The authors conclude that domain size "is highly dependent on both the hygroscopicity of the polymer used and the ambient RH" (Section 4). Two points limit this. In the authors' words, "it is not clear" whether the larger E6 and E50 domains come from molecular weight or another origin (Section 3.4). And, as a note by the AI writer, the sorption values of E3, E50 and E6 differ by about 10% while heights differ by tens of nanometres, the values are equilibrium data for dry powders, and Figure 5 mixes humidity and grade changes, so the six points form a rank association rather than a measured dependence on water (card note X4). The AFFINISOL pair is also confounded: the authors say it "highlights the impact of methanol hygroscopicity", but dropping methanol also removes a poorer solvent for ITZ (card note X19).

Result 2: nanostructure of a phase-separated film. BF TEM of the 40% RH E3 film showed domains around 250 nm in diameter, each in a "donut-like" ring, within an inhomogeneous continuous phase (Figure 3a). A 50:50 film had a thinner ring, "supporting the above supposition" that the domains "seem to be encapsulated" by polymer (Figure SI.3). HAADF showed the continuous phase as a network of bright and dark regions (Figure 3b). EDX gave about 2.9% nitrogen and 2.3% oxygen in light regions against about 2.3% nitrogen and 2.85% oxygen in dark ones, so light regions are richer in drug ("suggesting").Card check: these nitrogen values lie far below the 15.9% of ITZ and the about 4.8% expected for a 30:70 blend; only the contrast is usable (card note X7). The authors propose that this network formed by spinodal decomposition ("may have occurred"). A film of HPMC alone showed features the text calls "hole-like" (Figure SI.4); in the image they look dark, like the contrast used to mark domains (card note X12).

Result 3: ITZ separates from the solvent before drying ends. A film of ITZ alone, "expected to form a smooth continuous film", showed fluorescent domains, particles 60–70 nm high by AFM, round domains by TEM and spheres by SEM (Figures 1e, 2b, 3d, SI.5). "This observation suggests that the drug underwent phase separation from the solvent" before evaporation was complete (Section 3.1). To catch this in the liquid, a 3 mg/mL ITZ solution in DCM–MeOH was left to evaporate and its emission was followed (Figure 8). The maximum sat at 375 nm at 0 min and grew without shifting by 30 min, consistent with ITZ becoming more concentrated.

At 120 min the solution was slightly turbid, the maximum had moved to 367 nm, a new peak appeared at 381 nm, scattering at the 330 nm excitation wavelength had grown ("most likely" from particles), and a second-order scattering peak appeared at 660 nm. At 180 min the emission and the 660 nm peak had grown further.Card check: the text says the 381 nm peak exceeded the 367 nm peak at 180 min; in Figure 8 both stand at about 550 a.u. (card note X9). DLS gave a hydrodynamic radius of 168 ± 5 nm (polydispersity index 0.129 ± 0.05), and TEM showed spheres about 0.1–0.2 μm across.Card check: a 168 nm radius means a diameter near 336 nm, larger than the TEM particles (card note X8). The authors conclude that ITZ nanospecies form before solidification and that the ITZ-only particles are hollow spheres formed in solution; with HPMC and atmospheric water present, similar nanospecies are taken to form the same way ("suggesting"). The vessel and temperature of the evaporation are not stated (card note X11).

Result 4: ITZ's own spectra separate films with and without domains. Table B sets the UV and emission spectra of the films against what imaging found. All films were spun from DCM–MeOH unless stated. The comparison rests on peak positions, because the normalisation of intensities is not described (card note X17).

Table B. Spectral signatures of spin-coated films (doublet from the text; other positions read from Figures 7, 9 and SI.6, approximate).

What Table B means. Films without domains share a UV doublet and a single emission band; films with domains and ITZ alone show red-shifted absorption and an extra emission peak. Because the peak occurs in pure ITZ and not when ITZ is dispersed in polymer, the authors assign it to "self-associated ITZ molecules" ("indicating"; Section 3.5), note a similar extra peak in immiscible indomethacin films (Tian et al. 2016), and conclude that "ITZ autofluorescence behavior can provide insight into ASD miscibility". For UV, the results "point toward" its use for spin-coated ASDs.Card check: the text places the extra peak "at around 385 nm", Figure 9 at about 381 nm and Figure 10b at about 386 nm; read 381–386 nm as one band (card note X1).

Result 5: the drug loading at which the marker appears. At 20% RH the authors spun E3 films with 10 to 50% ITZ and recorded both spectra (Figure 10). The question is where between 10 and 50% the marker of self-associated ITZ first appears. Table C collects the peak positions.

Table C. UV absorption and emission of ITZ:HPMC E3 films with 10–50% ITZ spun at 20% RH (read from Figure 10a, b, approximate, except the 330 nm value from the text).

What Table C means. Absorption moves steadily to longer wavelengths with loading, a shift "indicative of ITZ aggregation", and the emission feature that "appears to be characteristic of ITZ–ITZ interactions" is present at 30–50% only. "This indicates that phase separation takes place above a 20% drug loading when performing spin coating at 20% RH" (Section 3.5), consistent with the domains of the 30:70 film at 20% RH in Result 1. Note by the AI writer: no loading between 20 and 30% was made, and the 10 and 20% films emitted almost nothing overall, so the threshold lies between 20 and 30% and the UV shift is the stronger evidence (card note X14). No loading series was made at another humidity.

Result 6: order inside the domains and humid storage. The FFT of one domain showed only a diffuse halo (Figure 4a). The slowly spun film showed Maltese crosses, "characteristic of the nematic liquid crystalline phase of ITZ" (Figure 4 caption). After 24 h at 40 °C/75% RH a domain had a different internal structure, and its FFT gave a (001) spacing of 28.3 Å, the length of one ITZ molecule, which "indicates" a SmA phase and reorientation on exposure to moisture (Section 3.3). The 7% RH film also changed: it was no longer flat, showed an irregular network with submicron corrugations (height scale ±100 nm), and LCR showed interpenetrating ITZ-rich and HPMC-rich regions (Figure 6b, c). The higher polymer-rich areas "could be explained by" vertical swelling of the polymer. There was no dry 40 °C control (card note X22).

Result 7: spray drying repeats the pattern. Table D summarises the four spray-dried batches and the polymer reference. All batches used the same feed concentration and spray settings, so solvent water is the intended difference between them. Sizes of particles and domains are not given in the paper.

Table D. Spray-dried ITZ–HPMC E3 particles (peak positions from the text; intensities read from Figure 12, approximate). The dark spots are said to grow with water content; no sizes are given.

What Table D means. The dry batch matches the films made at 7% RH and every spiked batch matches the films made in humid air, with the 373 nm peak growing with water. The abstract states that "as little as 1% water in the solvent was found to induce phase separation"; the results use "Our results suggest". The authors propose that water absorbed by methanol "is the likely cause" of the separation that Purohit et al. (2017) reported. The 12% domains are compared with "the ITZ spin coated film shown in Figure 3", which could mean Figure 3a or 3d (card note X15). The pattern of emission peaks in particles and films "was not identical", and the authors consider it "likely" that ITZ arranges differently with each preparation method; rotary-evaporated ITZ, for instance, gave two similar peaks near 363 and 372 nm (Figure SI.7).

Controls and what they rule out. GC–MS found methanol around 1 ppm and DCM below detection in all samples (Section 3.7), so leftover solvent does not explain the domains. HPMC alone gave a flat emission trace, so the 373 nm peak is not from the polymer. ITZ alone is the positive control for the self-association signature, and the dry spray-drying batch the negative control for the water series. Missing are an ITZ-only film made dry (card note X20) and a dry 40 °C storage control (card note X22).

Discussion

How the authors connect the findings. Section 4 proposes one drying path for films and particles. DCM is assumed to leave first, then methanol, so the liquid becomes richer in water, from humid air or from the solvent. ITZ, "extremely poorly soluble in water", then exceeds its miscibility limit.

Further drying "may result in the system re-entering a miscible region", but low molecular mobility leaves it "kinetically trapped", so the separation "has both thermodynamic and kinetic components". The pair is therefore "thermodynamically miscible as a binary system in the final ASD (i.e., in the absence of water)" ("Our results therefore suggest"). Slower evaporation at 40% RH "is expected to promote phase separation". Inside the domains "it is probable that ITZ undergoes nematic assembly": unquenched fluorescence excludes H-aggregates, J-aggregates are excluded "based on the extent of the Stokes shift", and "we speculate" that the molecules sit oblique to each other (Section 4).

Where it applies. The evidence covers one drug with four HPMC grades, one solvent pair, spin coating and laboratory spray drying. In discussing storage, the authors stress that this system lacks favourable drug–polymer interactions (Rumondor et al. 2011). Note by the AI writer (not discussed in the paper): the mechanism, a poor solvent for the drug left behind last, should transfer best to hydrophobic drugs dried from mixtures containing methanol or another water-miscible solvent, with polymers that do not bind the drug strongly; other drugs, polymers and solvents were not tested.

What the authors recommend. Because miscibility is so sensitive to solvent water, "attention should be paid to this factor during process optimization"; water may be added deliberately, for example with a surfactant in the feed, or enter "as an unknown contaminant" through methanol. Dry storage "may also be crucial", and water during formation and storage is "a critical factor" whose "impact should be assessed for ASD formulations". Finally, "spin coating could provide a small scale approach" to study phase behaviour as a function of solvent composition and to help choose a spray-drying solvent (Sections 4 and 5).

Note by the AI writer: what this means for formulation. A phase-separated batch made from a methanol-containing solvent reflects the pair and the process together; before calling a pair immiscible, it is worth remaking it from dry solvents at low humidity. For a hydrophobic drug in a mixed solvent, what matters is which solvent leaves last, so keeping a good solvent for the drug present to the end, with dry handling of methanol, should lower the risk. Manufacture and storage are separate water risks: a film made dry still separated after 24 h at 40 °C/75% RH.

What is unresolved. The authors leave open the origin of the grade effect and call the exact ITZ structure "challenging to fully elucidate". Beyond that, no water content was measured during drying, no drug–polymer–solvent phase diagram was mapped along the drying path, and "thermodynamically miscible" rests on the absence of domains at TEM resolution, with mixing below that scale untested. Whether the domains change dissolution or crystallization, and whether dry-made material stays single-phase in long dry storage, were not studied. Note by the AI writer: useful next steps would be measuring water in the drying liquid, a loading series at 7% RH, and a spray-drying series with the water basis and drug loading stated.

Limitations

Limitations stated by the authors. FLOM has "limited spatial resolution" (about 300 nm), AFM resolution "is limited by the dimension of probe tip" and AFM "is a surface sensitive technique", and contact mode risks tip contamination, which is why several images per sample and complementary techniques were used (Sections 2.5.2.4 and 3.2). They write that "it is not clear" whether the grade effect comes from molecular weight, that the AFFINISOL sorption difference "seems to be very small", that the ITZ structure in the domains is "challenging to fully elucidate", and that the oblique arrangement is speculation. They acknowledge differences between spin coating and spray drying in "the kinetics of solvent removal", geometry, solvent and environment (Section 3.6).

The AI writer's own judgement (1): water was the variable but was never measured. The qualitative pattern is consistent: domains with humid air or added water, none with dry processing, across FLOM, AFM and TEM, across UV and fluorescence, and across two processes. But no water content of any solution or film was measured, and the size correlation pairs equilibrium sorption of dry powders with single line-profile heights, mixes humidity and grade, and does not state how many films were measured (card note X4). The controls meant to isolate water do not fully do so: sieve-dried solvent at 20% RH only reduced the domains (X21), the DCM-only film also removes a poorer solvent for ITZ (X19), and no ITZ-only film was made dry (X20). The authors' wording rises from "suggest ... may be sensitive" (Section 3.1) to "strongly support the supposition" (Section 4) and "was established" (Section 5). The data support water during processing as the most likely driver in this system, not a quantitative dependence on water content.

The AI writer's own judgement (2): weaker numbers and overall strength. The 1% threshold has no stated basis and the spray-dried loading is not printed (X5, X6); the 20% loading threshold rests on films with almost no emission (X14); EDX values are usable only as contrast (X7); the J-aggregate exclusion cites a Stokes shift that is not reported (X13); text and figures disagree on the extra peak position, the E6 heights and the 180 min peaks (X1, X2, X9); DLS and TEM sizes differ (X8); and the storage test has no dry 40 °C control (X22). None reverses the main result, but the numbers are within-paper comparisons, not transferable limits. The card rates the paper "moderate" overall: the water–domain pattern is moderate to strong, the spectral marker and solution-phase nanospecies moderate, and domain size versus hygroscopicity, dry-state miscibility with kinetic trapping, and the nematic-to-smectic assignment weak to moderate. No dissolution or stability data show whether the domains matter for performance.

Conclusion

The paper asked whether phase separation in solvent-made ITZ–HPMC dispersions belongs to the pair or to the process, and its evidence points to the process. Films spun at 7% RH and particles spray-dried from dry solvents had no domains by FLOM, AFM or TEM and showed a single ITZ emission band. Films spun at 20 or 40% RH, and particles made with 1, 5 or 12% water in the methanol, had domains and an extra emission band from self-associated ITZ, and domain height followed humidity and polymer water uptake. An evaporating ITZ solution formed particles before drying ended, supporting the picture of ITZ separating from a water-enriched liquid and being trapped by the solidifying glass.

The practical message, in the authors' words, is that solvent water "should be assessed" for ASD formulations during manufacture and storage, and that spin coating could serve as a small-scale solvent screen before spray drying. The limits are that water was varied but never measured, that some controls change more than water, and that performance was not tested. Read with those limits, the study warns that a miscibility verdict on a solvent-made ASD may describe the drying path rather than the drug–polymer pair.