Phase separation that DSC misses: fluorescence probes and solid-state NMR show how drying rate sets domain size in itraconazole–HPMC dispersions

Background and motivation

Many new drug candidates dissolve poorly in water, and amorphous solid dispersions (ASDs) are a main way to raise their oral exposure (Introduction). In an ASD the drug is dispersed in a polymer matrix in a non-crystalline state, whose higher free energy gives a higher transient solubility and potentially better bioavailability (Newman et al. 2012). The same excess free energy drives crystallization unless the drug concentration in the polymer lies below its crystalline solubility in the matrix, and crystals cancel the solubility advantage (Hancock and Parks 2000; Hancock and Zografi 1997). The polymer is chosen to slow crystallization. The paper stresses the condition for this to work: drug and polymer must be mixed at the molecular level, so that polymer chains can disrupt the ordered self-assembly a crystal needs.

Miscibility can fail in two ways. The drug can crystallize, or the dispersion can split into a drug-rich and a polymer-rich phase that both stay amorphous; this is amorphous–amorphous phase separation (AAPS). AAPS matters before any crystal appears, because drug-rich domains contain less polymer and can crystallize faster (Rumondor et al. 2009), and because a hydrophobic drug-rich phase wets poorly and dissolves more slowly (Discussion). It can arise during manufacture by spray drying or hot-melt extrusion (Qian et al. 2010), from moisture taken up in storage (Rumondor et al. 2009; Purohit and Taylor 2015b), or on contact with dissolution medium (Purohit and Taylor 2015b).

The analytical difficulty is that the standard tests were built for crystals. Differential scanning calorimetry (DSC) and X-ray diffraction (XRD) detect crystalline drug well but not concentration differences within the amorphous state. The usual miscibility test, a single glass transition temperature ($T_g$) between those of drug and polymer, "often provides ambiguous results" (Introduction). By the polymer literature cited, DSC cannot resolve domains smaller than about 15–30 nm (Krause and Iskandar 1977). Qian et al. (2010) recorded a single $T_g$ in dispersions phase separated at the micron scale, and Lodge et al. (2006) recorded two $T_g$ values in blends that neutron scattering showed to be miscible. Scanning electron microscopy and atomic force microscopy (AFM) see only surfaces and cannot tell drug from polymer, and infrared and Raman spectra can be hard to interpret (refs 16–18).

Itraconazole (ITZ) with hydroxypropyl methylcellulose (HPMC) is a known difficult pair. Earlier reports (Six et al. 2003) and the authors' own work found limited miscibility after solvent evaporation. Purohit and Taylor (2015a) showed that spin-coated ITZ–HPMC films were miscible by DSC criteria but phase separated by AFM, nanoscale infrared spectroscopy, fluorescence imaging and spectroscopy, and nano-thermal analysis. Those methods worked on thin films but are hard to apply to particles, especially spray-dried particles, the form an ASD usually takes in a product. The authors name steady-state fluorescence spectroscopy as the exception, because it can be applied directly to a powder. Its best-known probe, pyrene, changes its emission with the polarity of its surroundings (Kalyanasundaram and Thomas 1977), and in a miscible ASD it reports a value between those of the pure drug and the pure polymer (Purohit and Taylor 2015b).

This paper asks whether fluorescent dyes can reveal miscibility and the size of phase-separated domains in ASDs made by four solvent-evaporation routes: film casting, spin coating, rotary evaporation and spray drying. The working idea is stated as a hypothesis ("It is hypothesized"): if AAPS occurs, non-polar probes go preferentially to the drug-rich phase and more polar probes to the polymer-rich phase. Pyrene was built into spray-dried particles, and two dyes of different polarity were used to stain the phases of films for confocal microscopy. Because fluorescence was still emerging for this purpose, solid-state nuclear magnetic resonance (ssNMR) was added as an established, orthogonal check. Holding the composition at 40:60 w/w ITZ:HPMC and changing only the route lets the paper ask how domain size depends on the rate of solvent removal.

Background concepts

Miscibility depends on the length scale of the test. Each technique sees mixing only down to some length scale, and a sample whose domains are smaller than that scale looks homogeneous to it. The paper combines methods whose windows span about three orders of magnitude: confocal microscopy resolves about 135 nm with the settings used, DSC about 15–30 nm (cited), ssNMR relaxation tens of nanometres and a few nanometres, and pyrene's excimer or exciplex band reports on neighbours within about 10–20 Å. Two methods can therefore give different miscibility verdicts on one sample without either measurement being faulty.

The single-$T_g$ criterion. A homogeneous amorphous mixture has one glass transition, while two phases large enough to behave as separate materials give two. The paper gives two reasons why a single $T_g$ can mislead. Domains smaller than about 15–30 nm give one $T_g$ (Krause and Iskandar 1977). In addition, DSC is a heating experiment: near $T_g$ mobility rises and miscibility typically increases with temperature, which "can potentially" remix a phase-separated sample into one that shows a single $T_g$ (Discussion). The second reason is an argument; remixing was not observed in this paper.

Pyrene as a polarity probe. Pyrene is a flat, non-polar aromatic molecule whose emission shows several vibronic peaks. The ratio of the first to the third peak, $I_1/I_3$, is lower in less polar and higher in more polar surroundings (Kalyanasundaram and Thomas 1977); in Figure 5 the two peaks lie at about 373 nm and about 384 nm. Here, pyrene in pure amorphous ITZ gives 1.39, in pure HPMC 1.54, and in water about 1.90. A molecularly mixed ASD should place each pyrene molecule in an averaged environment and give a value between 1.39 and 1.54 (Purohit and Taylor 2015b). A second feature is a broad, red-shifted band that appears when an excited pyrene sits next to another molecule: an excimer if the partner is pyrene, an exciplex if it is a different molecule. It requires a ground-state separation of about 10–20 Å (ref 47), so it reports on nearest neighbours.

Prodan, rhodamine 6G and confocal imaging. Prodan is hydrophobic, emits in the blue, and its emission maximum shifts with polarity. Rhodamine 6G (R6G) is cationic and relatively hydrophilic; it emits in the red but has a small Stokes shift, so its excitation and emission spectra overlap, and its molecules can self-aggregate and lose fluorescence at high concentration (ref 32). If prodan collects in the drug-rich phase and R6G in the polymer-rich phase, an image in both channels maps the phases. A confocal microscope records an optical slice of the sample in the z-plane and can excite both dyes at once, giving overlays that match exactly; its theoretical resolution here was about 135 nm.

ssNMR relaxation and spin diffusion. After a perturbation, the 1H nuclei of a solid return to equilibrium with characteristic times. The paper measures the laboratory-frame spin–lattice relaxation time $T_1$ (seconds here) and the rotating-frame relaxation time $T_{1\rho }$ (milliseconds here). Nearby protons exchange magnetization through dipolar couplings, a process called spin diffusion. If drug and polymer are intimately mixed, spin diffusion averages their relaxation times to one value; if they sit in separate domains, each keeps its own (Methods, citing refs 25–29). The distance over which magnetization spreads during one relaxation time is estimated with Eq. 2:

Here $L$ is the diffusion length, taken as the domain size; $D$ is the spin-diffusion coefficient; and $t$ is the relaxation time, $T_1$ or $T_{1\rho }$. In plain terms: the longer a relaxation time, the further magnetization travels before it decays, and the length grows with the square root of that time. Because $D$ was not measured, the authors use two literature values, 8.0 × 10⁻¹² cm²/s for rigid and 0.5 × 10⁻¹² cm²/s for mobile organic polymers (Clauss et al. 1993; Spiegel et al. 1993), which turn each time into a range; they add that "a rigorous measurement is needed" to derive a domain size. $T_1$ is a few hundred times longer than $T_{1\rho }$, so $T_1$ tests mixing over tens of nanometres and $T_{1\rho }$ over a few. Equal $T_1$ but different $T_{1\rho }$ values therefore mean mixing on the longer scale but not the shorter.Card check: the Methods state that typical $T_1$ (about 1 s) and $T_{1\rho }$ (about 10 ms) probe 20–100 nm and 1–20 nm; with Eq. 2 and the paper's D values, 1 s gives about 17–69 nm and 10 ms about 1.7–6.9 nm. The Table II ranges, computed from the measured times, do follow from the equation (story note X10). As a reference case, the paper cites Yuan et al. (2013): nifedipine–PVP dispersions at 90% drug loading gave distinct relaxation times (differences Δ of 0.4 s for $T_1$ and 13 ms for $T_{1\rho }$), while at 60% and 50% the values were almost identical, with a domain size of about 4.5 nm.

Two routes to phase separation. A solution that becomes unstable as solvent leaves can separate by nucleation and growth, in which droplets of the new phase form at discrete points and grow into round, separate domains, or by spinodal decomposition, in which the whole mixture splits into an interconnected pattern (ref 59). The final domain shape is often used to infer which mechanism operated, and the authors do this with their confocal images. "Primary drying" in the paper means the main solvent-loss step of each process.

Methods in detail

The work has four contributions. The first is the sample design: one 40:60 ITZ–HPMC solution dried at four rates, with DSC and XRD checks. The second is pyrene fluorescence as a polarity read-out that works on powders. The third is two-probe confocal imaging of films, which identifies the drug-rich phase and sizes its domains. The fourth is ssNMR relaxometry on probe-free samples, which checks the fluorescence results and bounds the domain size where imaging cannot reach.

Contribution 1: One composition, four drying rates

① Problem. To link domain size to the process, the samples must differ in how fast the solvent leaves and in little else. The earlier ITZ–HPMC study used spin-coated films, whereas the form that matters for products is spray-dried powder.

② What they did. ITZ and HPMC-606 (Shin-Etsu) were dissolved at 40:60 w/w in 1:1 v/v methanol (MeOH) and dichloromethane (DCM) at 2% w/v solids. One part of the solution received 0.2% w/w of one probe (pyrene, prodan or R6G) for fluorescence work; the other stayed probe-free for DSC, XRD and ssNMR. Film casting (slow) used quartz coverslips and Teflon sheets, 10 min in a fume hood and 24 h under vacuum, with primary drying of about 5 min; the quartz films served fluorescence and confocal work, and the Teflon films were cryomilled for ssNMR. Spin coating (fast) spun 100 μL on quartz at 300 rpm for 30 s, with primary drying of about 30 s and most solvent gone in under 10 s. Rotary evaporation (fast) used a 50 °C water bath, 24 h vacuum drying and cryomilling to 150–250 μm. Spray drying (very fast) sprayed 5 g of solids at 2% w/w, with and without 0.1% w/w pyrene, on a ProCepT Micro Spray Dryer (2 mL/min, inlet 75 ± 4 °C, outlet 50 ± 4 °C, atomization air 1.5 L/min, drying gas 0.4 m³/min), followed by overnight vacuum drying. The checks were modulated DSC at 2 °C/min to 170 °C (spray-dried powder also at 10 °C/min, data not shown), powder XRD from 5 to 35° 2θ, XRD of spray-dried powder after heating at 2 °C/min, and SEM. DSC and XRD covered only the rotary-evaporated and spray-dried powders, and replicate numbers are not printed for any method.Card check: the probe level and solids basis differ between Methods sections (0.2% w/w probe and 2% w/v solids in "Preparation of ASDs"; 0.1% w/w pyrene and 2% w/w solids in "Spray Drying"), so the pyrene level in films and rotary-evaporated samples is unresolved (story note X1). The XRD setting is printed as 40 mV, presumably 40 kV (X12).

③ Why it works. All samples come from one solution at one ratio, so the main difference between routes is the time available between reaching a miscibility limit and solidifying. The authors use exactly this when they relate domain size to primary drying time: about 5 min for casting, about 30 s for spin coating, and a few milliseconds for spray drying (Singh and Van den Mooter 2016). The DSC and XRD checks establish that the samples start amorphous, so any heterogeneity found later is amorphous–amorphous.

④ How it differs from earlier work. Qian et al. (2010) had shown that processing can induce phase separation, and the earlier ITZ–HPMC work characterised spin-coated films only. Here one composition is made by four processes, including spray drying. Note by the AI writer (not discussed in the paper): the routes also differ in substrate, sample form and probe content, and different methods were applied to different routes (Table A), so the series is a set of related samples rather than one sample measured four ways.

Contribution 2: Pyrene fluorescence as a polarity read-out for powders

① Problem. A test was needed that detects amorphous heterogeneity below the reach of DSC, works on particles, and is fast enough for screening.

② What they did. Pyrene-containing samples were measured on a Shimadzu RF-5301PC spectrofluorophotometer with a solid-state holder, with excitation at 332 nm and slit widths of 5 nm (excitation) and 1.5 nm (emission). $I_1/I_3$ was read for the rotary-evaporated, film-cast and spray-dried dispersions and for pyrene in ITZ alone and HPMC alone (Table I). Spray-dried dispersions at 10–50% drug loading were also measured, but their values are not shown. The analysis takes under 1 min. The broad excimer or exciplex band, at 470 nm in the text, was used as well (Figure 5).

③ Why it works. $I_1/I_3$ reports the polarity around the probe, and ITZ and HPMC give different values (1.39 and 1.54). Pyrene is non-polar, so in a phase-separated dispersion it "will preferentially partition" into the ITZ-rich phase, provided it has "sufficient mobility" while the phases form (Results). A molecularly mixed dispersion should therefore read between the end points and an immiscible one close to the pure drug. The excimer or exciplex band adds a proximity test: it needs a partner within about 10–20 Å, and with roughly 116 ITZ molecules per pyrene, a pyrene–pyrene partner is unlikely.

④ How it differs from earlier work. Pyrene had been used for polymerization reactions, ASD phase transformations, liquid–liquid phase separation in supersaturated solutions (Raina et al. 2015), and water-induced phase separation and miscibility in ASDs (Purohit and Taylor 2015b; Tian et al. 2016). The authors call its use on spray-dried particles "a hitherto unexplored application". The excimer or exciplex band as evidence of drug–probe contact adds to the ratio-only approach.

Contribution 3: Two-probe confocal imaging of films

① Problem. Pyrene tells whether a drug-rich environment exists, but not where it is, which phase is continuous, or how large the domains are.

② What they did. Cast and spin-coated films containing both prodan and R6G were imaged on a Nikon A1 confocal microscope with a 60× oil-immersion objective: prodan excited at 407 nm and collected at 425–475 nm, R6G excited at 561 nm and collected at 570–620 nm, with simultaneous excitation for overlays. Emission spectra of each probe in pure ITZ, pure HPMC and dispersion films were recorded at the same excitation wavelengths (Figures 6 and 7). Because the R6G spectra overlap, its emission was collected only on the red side, away from its maximum. Domains in spin-coated films were sized in ImageJ, with a watershed function to split apparent aggregates and a circularity filter to remove unresolved ones; the number of domains analysed is not printed.

③ Why it works. Prodan is hydrophobic and should partition into the ITZ-rich phase (blue); R6G is cationic and should concentrate in the HPMC-rich phase (red). The prodan spectra support the first assignment. For R6G, the text reports much lower emission in pure ITZ films and explains it by limited miscibility and self-aggregation in ITZ ("it is likely"), which "suggests" that R6G favours HPMC. The authors also draw on their earlier finding that ITZ forms the discrete phase in solvent-evaporated ITZ–HPMC.

④ How it differs from earlier work. Two-probe confocal imaging had been used for polymer-blend phase separation (Granados et al. 2001), and fluorescence imaging had been applied to ITZ–HPMC films before (Purohit and Taylor 2015a). New here are the pairing of probes of contrasting polarity to name the phases, the comparison of two drying rates, and a size distribution for the faster one.

Contribution 4: Solid-state NMR as an orthogonal check and a length-scale bracket

① Problem. Fluorescence methods were new for miscibility assessment, so their conclusions needed validation by an established technique ("it becomes essential", Discussion), and spray-dried domains lie below the confocal resolution.

② What they did. 13C cross-polarization magic angle spinning (CP-MAS) spectra of HPMC, amorphous ITZ, the spray-dried dispersion and the cryomilled film-cast dispersion, all probe-free, were recorded at 400 MHz (1H), 293 K and 12 kHz spinning (Figure 9). $T_1$ was measured by 13C-detected saturation recovery. $T_{1\rho }$ was measured by holding the 1H magnetization under a 100 kHz spin-lock field for varied times before cross-polarization to 13C, and each resolved peak's decay was fitted with Eq. 1:

Here $I(t)$ is the intensity of a resolved peak (printed as $I(t)$, although the variable is the spin-lock duration $\tau$), and $I_0$ is an amplitude from the fit. In plain terms: the longer the magnetization is held under the spin lock, the more signal is lost, and $T_{1\rho }$ is the time constant of that loss. Values from a few resolved peaks were averaged per component and converted to lengths with Eq. 2, using both diffusion coefficients.

③ Why it works. ITZ and HPMC have well-separated 13C peaks: all aromatic peaks are from ITZ, and HPMC has resolved peaks at about 103, 83 and 60 ppm (Figure 9), so each component's proton relaxation can be read through its own carbons. Because the $T_1$ and $T_{1\rho }$ lengths differ by about an order of magnitude, the pair gives an upper and a lower bound.

④ How it differs from earlier work. $T_1$ and $T_{1\rho }$ relaxometry is an established miscibility test for ASDs (Aso et al. 2007; Pham et al. 2010; Yuan et al. 2013). Its role here is new: it was run on probe-free samples from the same solutions, to check the probe-based conclusions and to supply the only size information for the spray-dried powder.

Experimental analysis

Setup in brief. Not every sample was examined by every method, and each method sees a different length scale (Table A). Pure ITZ and pure HPMC served as references for each probe and for the 13C spectra; how they were prepared is not described (story note X9).

Table A. Which sample was examined by which technique, the length scale each technique addresses, and the outcome (compiled from the Methods, Results, Table II and the Discussion).

Main result 1: by conventional tests the dispersions look amorphous and miscible. The rotary-evaporated and spray-dried powders gave only an X-ray halo, with its maximum at about 19° and about 18° 2θ (Figure 2). Modulated DSC gave one broad glass transition for each, with a midpoint of 87 °C for both (Figure 3), above the midpoint $T_g$ of pure amorphous ITZ, about 60 °C (Verreck et al. 2003). By DSC criteria both are miscible. The spray-dried trace also showed a small, sharp endotherm at about 165 °C, near the melting point of ITZ; it was absent at 10 °C/min and from the rotary-evaporated trace, and spray-dried powder heated at 2 °C/min showed ITZ crystal peaks (about 14.6, 17.6, 18.1, 20.5 and 23.6° 2θ). The authors therefore assign the endotherm to crystallization during the slow scan and take the fresh powders as amorphous. SEM showed spray-dried particles of about 10–30 μm, many near-spherical with small surface nodules (Figure 4).

Note by the AI writer (not discussed in the paper): if the drug-rich domains were nearly pure ITZ and large enough for DSC, a separate step near 60 °C would be expected. A single step at 87 °C fits either of the authors' explanations, small domains or remixing on heating, without distinguishing them. The paper also does not discuss why only the spray-dried powder crystallized during the slow scan.

Main result 2: pyrene reports a pure-ITZ environment in every dispersion.

Table B. Pyrene $I_1/I_3$ ratios (from Table I; standard deviations 0.02 or less for all samples; replicate numbers not printed; water value from the Results text, no reference given).

The three dispersions read 1.38–1.39, the value of pyrene in ITZ and clearly apart from HPMC. A molecularly mixed dispersion should read between the end points; instead every dispersion matched the pure drug, whatever its evaporation rate, and spray-dried dispersions at 10–50% drug loading gave similar ratios (values not shown). The authors write that the localised concentration of pyrene in ITZ domains "indicates" that all samples are phase separated (Results) and that the samples "demonstrated" phase separation (Conclusions).Card check: the Methods list a spin-coated pyrene sample and the Discussion places the spin-coated comparison in Table III, but Table I has no spin-coated value and Table III contains only domain sizes (story note X2).

The spray-dried spectrum (Figure 5) also showed a broad band, at 470 nm in the text and with a broad maximum at about 445–470 nm in the figure. With an ITZ:pyrene molar ratio of about 116:1, the authors find a pyrene–pyrene origin "unlikely" and read the band as evidence that pyrene sits within about 10 Å of ITZ, consistent with pyrene concentrated in ITZ-rich domains. Together, the ratio and the band are taken to show phase separation and that pyrene reports its environment at the molecular level.Card check: 116:1 fits 0.1% w/w pyrene (about 115:1 by the extractor's calculation with molar masses the paper does not print); 0.2% w/w would give about 57:1 (X1). Only the spray-dried spectrum is shown, although the text says the band appeared in all dispersions, and the pure ITZ and HPMC spectra in this region are not shown (X7).

Main result 3: the two dyes identify the phases in films. Prodan emission peaked at about 435 nm in ITZ, about 451 nm in HPMC and about 438 nm in the dispersion film, whose spectral shape matched ITZ (Figure 6, approximate). The authors read this as prodan in an ITZ-rich environment ("indicating"), in line with pyrene. The R6G evidence is weaker. The text says R6G emission was much lower in pure ITZ films than in HPMC or dispersion films, but in the 570–620 nm collection window of Figure 7 the ITZ trace is only slightly lower, from about 571 to 595 nm, and converges with the others by about 600 nm.Card check: the region near the laser line is off-scale for all three films and no integrated intensities are given, so the reduced R6G emission in ITZ is stated but not quantified (story note X6).

In the cast films (Figure 8, left), R6G stained mainly the continuous phase, with dark unstained areas, while prodan stained discrete circular domains; the overlays showed the same split. With the earlier finding that ITZ forms the discrete phase, the authors conclude that the discrete phase is drug-rich and the continuous phase polymer-rich ("Thus it is clear", Results). The spin-coated films (Figure 8, right) showed the same arrangement with much smaller domains. In the cast films, larger prodan regions were often built of smaller aggregated domains (circled in Figure 8). ImageJ analysis of spin-coated films gave a smallest domain of 300 nm, a largest of 2.8 μm, a D50 of 0.9 μm and a D90 of 1.5 μm; the smallest size agrees with the earlier AFM study. The authors call the effect of evaporation rate on domain size "very apparent", but decline to infer anything about aggregation in the spin-coated films, given the small domains and the microscope resolution.

Main result 4: ssNMR separates the cast film on both length scales and the spray-dried powder only on the shorter one. The 13C spectra had broad peaks, consistent with amorphous samples, and the spray-dried and film-cast spectra were almost identical. The authors read this as evaporation rate "presumably" not changing the molecular structure, while dynamics "could potentially" differ. Table C gives the relaxation data.

Table C. 1H relaxation times of ITZ and HPMC in the two dispersions (from Table II; averages over a few resolved peaks, standard errors in parentheses; lengths from Eq. 2 with both diffusion coefficients).

The ranges reproduce from Eq. 2. For the spray-dried $T_{1\rho }$, the mobile-system coefficient (0.5 × 10⁻¹² cm²/s, or 50 nm²/s) with 9.8 ms gives about 1.7 nm, and the rigid-system coefficient (8.0 × 10⁻¹² cm²/s, or 800 nm²/s) with 12.6 ms gives about 7.8 nm, hence 2–8 nm; the other three printed ranges reproduce the same way. In the spray-dried powder, ITZ and HPMC have essentially the same $T_1$ (3.5 and 3.4 s) but different $T_{1\rho }$ (9.8 and 12.6 ms), so they are mixed on the 30–130 nm scale but not at 2–8 nm. In the film-cast dispersion both times differ, which the authors read as phase separation above 30 nm, supporting the fluorescence and confocal results. Their wording for the spray-dried case is "suggests" (Results) and "it can be inferred" (Discussion); Table II and the Abstract state it without a hedge. The Table II caption adds that a larger Δ means more extensive phase separation (Δ of 0.1 against 0.5 s for $T_1$, 2.8 against 12 ms for $T_{1\rho }$). Note by the AI writer (not discussed in the paper): relaxation times of pure ITZ and pure HPMC are not reported, so how far each value has moved towards the other's cannot be judged from Table II.

Comparison across routes. Table D brings the size information together with the drying times and the kind of value each entry is.

Table D. Domain size by preparation route (from the Results and Discussion text and Table III, and read from Figure 10 on its log axis, approximate; the meaning of the error bars is not stated).

The trend is consistent: micrometre domains after minutes of drying, sub-micrometre domains after seconds, and nanometre-scale heterogeneity after milliseconds, a drop of about three orders of magnitude in Figure 10. The authors had expected spray-dried domains "much smaller" than the 1 μm spin-coated ones ("we can anticipate") and call the ssNMR result "entirely consistent" with that expectation. The entries, however, come from different techniques and samples: image sizes of films on quartz against a spin-diffusion bracket on powder.Card check: equal $T_1$ and distinct $T_{1\rho }$ bound the spray-dried domain dimension between about 2–8 nm and about 32–130 nm, as the Results text words it, whereas Table III, Figure 10, the Abstract and the Conclusions present 2–8 nm or below 10 nm as the size (story note X4). Table III also lists ssNMR for the film-cast row, though ssNMR gives only a lower bound there, measured on a milled Teflon-cast film rather than the imaged quartz film (X11). Read as orders of magnitude, the trend holds; read as precise sizes, the table mixes unlike quantities.

Controls and what they rule out. The pure ITZ and HPMC references define both ends of each probe scale, and the water value shows pyrene responding in the expected direction. The 10 °C/min DSC run and the XRD of heated powder show that the 165 °C endotherm arises during the slow scan, not from crystals in the fresh sample. Note by the AI writer (not discussed in the paper): because the probe-free cast film and spray-dried powder also show distinct relaxation times, the separation was not created by the 0.1–0.2% w/w of dye. What is missing is a positive control for mixing: no drug–polymer pair known to be molecularly miscible was measured with the same pyrene protocol here, so the expectation of an intermediate value rests on earlier work (Purohit and Taylor 2015b).

Discussion

How the authors explain domain formation. As methanol and dichloromethane leave, the solutes concentrate until a miscibility or solubility limit is reached, and further evaporation makes one component separate. The discrete, circular drug-rich domains are read as "suggesting" nucleation and growth rather than spinodal decomposition. Aggregation of drug-rich domains in the cast films "indicates" that they formed before the solvent had fully evaporated, while mobility was still high. Faster evaporation "is expected to" build supersaturation faster, raising the nucleation rate and giving smaller domains.

Each step after the first is inferred, not observed: domain formation was not watched during drying, and the nucleation-rate argument is an expectation that fits the end-state sizes. The mechanism is plausible and matches the size trend, but it rests on final domain shapes and three drying rates.

Where the approach applies. The evidence covers one pair, one polymer grade (HPMC-606), one solvent system and one ratio (40:60 w/w), plus unreported pyrene values at 10–50% drug loading. All routes are solvent-evaporation processes; hot-melt extrusion was not tested. Within this frame, the pyrene measurement works on films and powders and takes under a minute, which is why the authors propose it "may be useful" as a rapid screen for polymer and solvent selection (Conclusions).

What each method can tell. Pyrene answers a yes-or-no question: is there a drug-rich environment? It gave the same ratio for micrometre domains in cast films and nanometre-scale heterogeneity in spray-dried powder, so it carries no size information; the authors say as much for the spray-dried case. Confocal imaging names and sizes the phases, but only above about 135 nm. ssNMR brackets the size but needs resolved peaks and an assumed diffusion coefficient. Note by the AI writer (not discussed in the paper): because pyrene concentrates in the drug-rich phase, a small volume of drug-rich domains could dominate its signal, so an ITZ-like ratio does not show what fraction of the ITZ is in separated domains. The partitioning also depends on "sufficient mobility", a condition the authors state but did not test for the millisecond drying of the spray dryer.

What is unresolved. The authors raise one open question: evaporation rate did not change the 13C spectra, but it could affect molecular dynamics, which they call "of great interest for further exploration". Other points are left open by the design. The spray-dried domain size lies somewhere between about 2–8 nm and about 32–130 nm. The smallest domain the pyrene ratio can detect was not measured. The effect of these domains on crystallization or dissolution, the reason AAPS matters, was not measured for any dispersion. The proposed screen was tried with one polymer and one solvent system.

Implications for formulation. The authors make three recommendations. A single DSC $T_g$ should not be taken as proof of miscibility, so "it becomes important" to establish orthogonal techniques (Discussion). Fluorescence with environment-sensitive probes "show[s] promise" for spray-dried ASDs, and multi-probe confocal imaging adds a versatile tool for drug–polymer microstructure. Note by the AI writer (not discussed in the paper): three further points follow from the data. First, one composition gave micrometre or nanometre domains depending on drying rate, so a miscibility result belongs to a process and its drying rate, not only to a drug–polymer pair. Second, faster drying shrank the domains but did not remove them; even the spray-dried powder was not mixed at the few-nanometre scale. Third, the pure-component pyrene contrast was small (1.39 against 1.54), so pairs whose drug and polymer have similar polarity may be hard to resolve. Useful next steps would be to calibrate the pyrene ratio against samples of known domain size, including a known miscible pair; to make the rigorous domain-size measurement the authors call for; and to link the measured heterogeneity to crystallization and dissolution of the same powders.

Limitations

Limitations stated by the authors. For the spin-diffusion estimate, "a rigorous measurement is needed" to derive domain size, and two literature diffusion coefficients only set a range (Methods). Confocal imaging could not be used for the spray-dried particles, whose scale of phase separation lies below the theoretical resolution of about 135 nm (Discussion). For the spin-coated films, the small domains and the microscope resolution prevent any inference about aggregation. For the spray-dried powder, steady-state fluorescence detected phase separation, but information on domain size "is lacking". Because fluorescence methods were still emerging, validation against ssNMR was considered "essential". R6G emission had to be collected away from its maximum, and pyrene partitioning is stated as conditional on "sufficient mobility" (Results).

The AI writer's own judgement (1): what the sizes are. Phase separation is well supported for the cast films, where pyrene, two-probe confocal imaging and ssNMR agree, and reasonably supported for the spray-dried powder, where pyrene and $T_{1\rho }$ agree. The domain sizes are weaker. For the spray-dried powder, the data bound the size between about 2–8 nm and about 32–130 nm, yet the summary table, Figure 10, the Abstract and the Conclusions present 2–8 nm or below 10 nm as a measured size. Sizes for the same route differ between the Results, the Discussion, Table III and Figure 10, and the cross-route comparison places confocal sizes of quartz films next to ssNMR lengths of powder and of a milled Teflon-cast film. The three-orders-of-magnitude trend should be read as a ranking with order-of-magnitude values.

The AI writer's own judgement (2): coverage and reporting gaps. The statement that DSC gave a single $T_g$ for all dispersions rests on the rotary-evaporated and spray-dried powders only; the films were not tested by DSC or XRD. No spin-coated pyrene value exists, the pyrene level in films and rotary-evaporated samples is unresolved (0.1 or 0.2% w/w), probe levels in the dual-probe films are not stated, and reference-sample preparation is not described. Replicate numbers are not printed, and the ImageJ domain count is not given. The exciplex assignment rests on the spray-dried spectrum without reference spectra, and the lower R6G emission in ITZ is not borne out quantitatively by Figure 7. The formation mechanism is inferred from end-state images, as the authors' hedges ("suggesting", "is expected to") acknowledge. The general $T_1$ and $T_{1\rho }$ length scales in the Methods do not follow from Eq. 2, Figure 1b draws pyrene as a dihydropyrene, and the XRD voltage is printed as 40 mV; none of these affects the conclusions. Overall, the evidence that ITZ–HPMC is phase separated in every route is moderate to strong, the size trend is moderate, and the nanometre size for spray-dried powder is weak to moderate.

Conclusion

The paper shows that 40:60 ITZ–HPMC dispersions made by solvent evaporation are phase separated whatever the drying rate, even though the two powders tested by DSC each gave one broad $T_g$ at 87 °C. Pyrene read a pure-ITZ environment (1.38–1.39, against 1.39 for ITZ and 1.54 for HPMC) in rotary-evaporated, film-cast and spray-dried samples. Prodan and R6G showed discrete drug-rich domains in a continuous polymer-rich phase in films, and ssNMR on probe-free samples confirmed separation in the cast film on both length scales and in the spray-dried powder on the few-nanometre scale. The drying rate set the size: about 3–5 μm after minutes of casting, about 0.4–1 μm after seconds of spin coating, and nanometre-scale heterogeneity, bounded between about 2–8 nm and about 32–130 nm, after milliseconds of spray drying.

For formulation work, the paper offers a division of labour. A pyrene measurement that takes under a minute can flag a drug-rich environment in a powder, including a spray-dried one, where DSC gives a single $T_g$; confocal imaging or ssNMR is then needed to say how large the domains are. The study covers one drug, one polymer and one ratio, and it does not test whether nanometre-scale separation changes stability or dissolution. Those are the questions that would turn this analytical result into a formulation rule.