Seeing nanoscale phase separation in telaprevir–polymer films: infrared, thermal and mechanical probes on an AFM tip

Background and motivation

Many new drugs dissolve too poorly in water to be absorbed well. One established remedy is the amorphous solid dispersion (ASD): the drug is made amorphous and blended with a polymer, which raises its dissolution rate and the solution concentration it can reach. Whether the drug and polymer actually mix, their miscibility, matters for shelf life. In a miscible blend the high-molecular-weight polymer slows drug crystallization (Yoshioka 1995, ref 8); in a phase-separated blend the amorphous drug "may crystalize quickly" and lose the solubility and bioavailability advantage (Rumondor 2009, ref 9). C1

Measuring miscibility is the hard part. The standard test is differential scanning calorimetry (DSC): one glass transition temperature (Tg) is read as a miscible blend, two as a phase-separated one (Baird 2012, ref 10). The introduction stresses that a single Tg "may not be a reliable indicator" (Qian 2010, ref 11), and that DSC struggles in three situations: drug and polymer with similar Tgs, separated phases of similar composition, and very small domains (refs 12, 13). C2 Electron microscopy and atomic force microscopy (AFM) can resolve nanoscale domains but cannot say what they are made of; solid-state NMR can probe miscibility and domain size, but each measurement takes hours (Lauer 2011, ref 14; Marsac 2010, ref 15; Pham 2010, ref 16). C3 What was missing was a way to tie the microstructure seen in an image to the chemical composition of each phase.

Telaprevir (TPV) is a demanding test case. It is a poorly water-soluble hepatitis C protease inhibitor whose crystalline form has low oral bioavailability, so it was developed and marketed as an ASD with hydroxypropyl methylcellulose acetate succinate (HPMCAS) (Kwong 2011, ref 4; Newman 2015, ref 17). C4 Its DSC Tg, 102.7 ± 0.4 °C, lies close to those of the three polymers studied: hydroxypropyl methylcellulose (HPMC, Methocel E5) at 140 ± 5 °C, HPMCAS (HF grade) at 122 ± 1 °C and the vinylpyrrolidone–vinyl acetate copolymer PVPVA (Kollidon VA 64) at 108 ± 1 °C (Table 1). M1·M2·M3·M4 With PVPVA the two Tgs are about 5 °C apart, so DSC cannot be "readily employed" to detect phase separation in that pair. C26

The group had earlier used AFM-based infrared spectroscopy (AFM-IR, also called nanoIR) to image phase separation in a drug–polymer blend (Van Eerdenbrugh 2012, ref 29) and to study drug nanocrystals, where the same instrument also records a contact-resonance frequency map (Harrison 2013, ref 18). Local thermal analysis with a heated tip had been applied to solid dispersion films (Qi 2013, ref 20; Qi 2011, ref 48), and a fluorescent probe that prefers one phase had been used to reveal phase separation in itraconazole–HPMC blends (Purohit 2015, ref 41). C43·C39 The stated objective of this paper is to develop new approaches to characterize ASD miscibility with nanoscale methods, by combining infrared spectroscopy, thermal analysis and Lorentz contact resonance (LCR) mechanical measurements with standard AFM imaging, and to apply them to TPV films with HPMC, HPMCAS and PVPVA as a function of polymer type and drug:polymer ratio.

Background concepts

Miscibility gap, dispersed and continuous phase. When a drug and polymer are not fully miscible, there is a range of compositions, the miscibility gap, in which the blend splits into two phases. Inside the gap the composition of each phase is fixed by thermodynamics; raising the overall drug loading changes only how much of each phase forms. The paper uses this principle when it interprets spectra taken at different loadings. C20 In a film, the minority phase usually appears as discrete domains (the dispersed phase) inside a continuous matrix. In all films studied here the discrete domains stand above the matrix in the AFM height image, so "domain" and "raised region" refer to the same thing. R8·C13

Spin-coated films. All samples are thin films made by spinning a drop of drug–polymer solution in 1:1 methanol:dichloromethane onto a 1 cm zinc sulfide (ZnS) flat and drying it overnight under vacuum. P1 Such films are convenient for surface probes but differ from a spray-dried powder, which the authors note would need embedding and microtoming before these methods could be applied. C53

AFM-IR. A pulsed, tunable mid-infrared laser (pulse length about 10 ns) is focused on the spot where the AFM tip touches the sample. When the sample absorbs at the laser wavelength it heats and expands within nanoseconds and kicks the cantilever, which rings down at its resonance frequencies; the amplitude of this "ringdown" is proportional to the absorption integrated through the sample thickness. P4·E1 Sweeping the laser wavelength gives a local IR spectrum; scanning the sample at a fixed wavelength gives an IR image. Three consequences shape the whole paper. First, signal intensity depends on how thick the absorbing material is, so the authors compare band ratios rather than intensities. C12 Second, the signal comes from the whole thickness under the tip, so a spectrum taken on a thin domain also contains signal from the material beneath or beside it, an effect called spectral mixing. C14·C50 Third, the lateral resolution is limited: the best reported is around 20 nm and a typical value is around 100 nm, depending on film thickness and thermomechanical properties. R24 The frequency of the ringdown also shifts with the stiffness of the material under the tip, so a contact-resonance frequency map can be recorded at the same time as the IR image. P5·E2

nanoTA. In local thermal analysis a microfabricated cantilever with a heater in its tip is pressed onto the surface and heated at 5 °C/s. The tip first rises as the material under it expands; when the material softens, the tip sinks in and the deflection turns over. The temperature at which the derivative of deflection against temperature crosses zero is taken as the softening temperature, the local analogue of Tg. P7 Because the tip penetrates once the material softens, the sampled region is typically about 100 nm across and about half that deep. C50

Lorentz contact resonance (LCR). Here a U-shaped probe carries an alternating current in a magnetic field, so a Lorentz force drives the cantilever while the tip stays in contact. Sweeping the drive frequency from 1 to 1000 kHz gives a mechanical spectrum whose resonance peaks shift to higher frequency on stiffer material; driving at one chosen frequency while scanning gives an amplitude image in which regions whose contact resonance matches that frequency light up. P8·C24 The authors list three advantages over piezo-driven contact resonance: no parasitic piezo peaks, contrast based on stiffness rather than on phase (which friction and adhesion can disturb), and one cantilever for a wide stiffness range. P8 The contact area, and with it the resolution and the depth probed, is typically in the 10 nm range, so LCR is strictly surface-sensitive. C50

Pyrene fluorescence. Pyrene is a small hydrophobic fluorescent molecule. Added at 0.01% (w/v) to the casting solution, it "tends to accumulate" in the more hydrophobic phase of a phase-separated film, here the drug-rich phase, which therefore fluoresces more brightly (Purohit 2015, ref 41). P9·C39 This is an optical method, so it cannot resolve domains smaller than the diffraction limit of light. R22

Table A. The probes compared in the paper, with the sampling sizes the authors give in the Discussion and the situations in which each one gave little or no contrast. C50·C46·C48·C49·R11·R15·R16·R22

Methods in detail

The study has one simple design: one drug, three polymers, a series of drug:polymer ratios, and for one high-loading film per polymer every probe is placed on the same two phases. The flow below shows the order of the work. A1·A2·A3·A4

Films and reference measurements

TPV (Attix) and each polymer were dissolved in 1:1 (v:v) methanol:dichloromethane at 5% (w/v) total solids. Between 30 and 50 µL were spun on a ZnS flat, 6 s at 50 rpm and then 50 s at 3100 rpm, and the films were dried overnight in a vacuum oven at room temperature. Drug:polymer ratios of 1:9 to 5:5 (w/w) were made with all three polymers, plus 6:4 with HPMC and HPMCAS and 7:3 and 8:2 with HPMCAS. P1·M1·M2·M3·M4 The paper does not report film thickness, storage time before measurement, or the humidity during storage and measurement. X9

Height images were taken on the nanoIR2 in contact mode (0.5 Hz, 256 × 256 points). For TPV–PVPVA at 30–50% drug the surface was so rough that tapping mode (0.3 Hz) was used to avoid damage. Domain sizes were measured with the software ruler as the average of 5 readings. P2 Reference spectra of the pure components came from films spun on KRS-5 windows and measured in transmission (128 scans), and from AFM-IR on the pure films (9 spectra each). P3·P4·R2 The DSC Tgs of the pure components (Table 1) are midpoints of the second heating at 20 °C/min, in triplicate; HPMC was run with one thermal cycle only. P6·X12

Chemical contrast: nano-IR spectra and images

Before each session the IR background was recorded from 1200 to 1800 cm⁻¹, and the laser spot was centred on the tip at 1450, 1525 and 1670 cm⁻¹ in four steps from an 800 µm to a 50 µm search area (Figure S1). Local spectra were recorded from 1200 to 1800 cm⁻¹ every 4 cm⁻¹ with 256 co-averages (2.5 min per spectrum), using the second cantilever mode (centred at 200 kHz). Five spots per phase were averaged and normalized, except for 60:40 TPV–HPMC, where the Figure 4A caption gives n = 4. P4·X11

The key choice is the marker band. In both bulk and AFM-IR spectra, a band at 1525 cm⁻¹ appears in telaprevir and in none of the polymers (Figure 2B, C). R2 A theoretical IR spectrum of telaprevir could not be computed; from its position the authors suggest the band "may be" a pyrazine ring stretch. R3·C11 Each polymer then contributes a band of its own: 1450 cm⁻¹ for HPMC, 1236 and 1740 cm⁻¹ for HPMCAS, 1244 and 1676 cm⁻¹ for PVPVA. R2 Because intensity follows thickness, the comparison is always the ratio of a polymer band to the 1525 cm⁻¹ band. C12

Spectra of the domains and the matrix were taken on 60:40 TPV–HPMC, 80:20 TPV–HPMCAS and 50:50 TPV–PVPVA, and on 50:50 TPV–HPMCAS (Figure S3). Spectra of the domains alone were taken across the loading series (Figure 5); the matrix was too narrow there to place the tip. P4·R12 IR images were recorded at 1525 cm⁻¹ and at a polymer band (1457, 1740 or 1676 cm⁻¹) over the same 4 × 4 µm area, each image taking about 43 minutes, and divided to give a ratio image; a frequency map was recorded at the same time. Imaging used 60:40 TPV–HPMC, 80:20 TPV–HPMCAS and 30:70 TPV–PVPVA, chosen because their domains were large enough to give IR contrast and to image the matrix. P5

Thermal contrast: nanoTA

The heated probes were calibrated with three polymer standards (polycaprolactone 55 °C, polyethylene 116 °C, polyethylene terephthalate 235 °C) through a quadratic fit of deflection against heating voltage. After a height image located the domains, the tip was ramped at 5 °C/s on chosen spots of each phase: five per phase on 60:40 TPV–HPMC and 30:70 TPV–PVPVA, six on 80:20 TPV–HPMCAS (Figure 7). Ramps stopped automatically when the deflection or the probe resistance dropped by 0.15 V within 20 ms, which protects the probe. Pure films of each component were measured the same way. P7·M12

Mechanical contrast: LCR

Mechanical spectra of the pure films were swept from 1 to 1000 kHz (drive strength 20%, 100 kHz/s). P8 The first and largest peak, the first flexural mode, is normally used for imaging because it is the most sensitive. R18 From these spectra the authors chose two imaging frequencies per film, one that favours the drug and one that favours the polymer, and then tuned them on each film for best contrast, since the probe drags across the surface during imaging and since each phase may contain some of the other component. C34 The final pairs were 128 and 123 kHz (60:40 TPV–HPMC), 128 and 124 kHz (80:20 TPV–HPMCAS), and 710 and 129 kHz (50:50 TPV–PVPVA), each image taking about 14 minutes. P8

An optical cross-check: pyrene fluorescence

To confirm phase separation with a method that does not use an AFM tip, films containing 0.01% (w/v) pyrene were spun from 100 µL on quartz slides and imaged with 330–380 nm excitation and emission collected above 420 nm. P9 Only high-loading films were imaged: pure HPMCAS as a control, 60:40 TPV–HPMC, 80:20 TPV–HPMCAS and 50:50 TPV–PVPVA, because the domains of lower-loading films are below the optical diffraction limit. R22

Experimental analysis

Where domains appear: the height images

Films of the pure components were smooth, except pure HPMCAS, whose height variations the authors read as uneven film formation (Figure 2A). R1·C5 With drug added, the three polymers behave differently (Figure 3). R4·R5·R6

Table B. What the height images show for each polymer, from the text and Figure 3 (sizes are the authors' ruler readings, given as "about"). R4·R5·R6·R7

For HPMC and PVPVA, discrete domains are present already at the lowest loading, 10% drug, and they grow as the loading rises. The authors read them as "most likely" (HPMC) and "probably" (PVPVA) telaprevir-rich, since their size follows the drug content. C6·C8 For HPMCAS, the surface shows only the irregular texture of the pure polymer film up to 30% drug; small circular domains appear at 40% and become clearer at 50%, from which the authors conclude that phase separation "may have occurred" above 40%. R5·C7 The gaps between these HPMCAS domains are only 21 nm at 40% and 25 nm at 50% drug. R7 The authors are explicit that height alone can neither prove phase separation (the 50% HPMCAS film is the example) nor tell what each domain contains. C9

What the domains contain: local nano-IR spectra

The spectra of domains and matrix answer the composition question for one film per polymer (Figure 4). R8·R9·R10

Table C. Polymer bands against the telaprevir band at 1525 cm⁻¹ in the local spectra, as described in the text; the paper gives no numerical ratios. R8·R9·R10·R11

In all three films the domains carry relatively more of the telaprevir band than the matrix. The authors conclude that the dispersed phase is telaprevir-rich and the continuous phase polymer-rich, with graded confidence: "indicating" for HPMC, "most likely" and "appear to be" for HPMCAS, and "preliminary analyses ... suggest" for PVPVA. C13·C16·C19 HPMCAS had to be measured at 80% drug: at 40–50% the domains are crowded with gaps of 21–25 nm, well below the typical resolution of about 100 nm, so the tip cannot record either phase cleanly. C15·R7·R24 At 50:50 HPMCAS the two phases gave the same spectrum, and an attempt to image the film by IR failed. R11

Both phases show bands of both components in every film. The authors give two causes that they do not separate: partial miscibility, so that each phase really contains some of the other component, and spectral mixing through the film thickness or from the neighbouring phase below the resolution of the method, "or some combination". C14 For HPMCAS they add that its higher miscibility means each phase holds a lot of telaprevir, which makes the phases harder to tell apart chemically. C17

The loading series. Spectra of the domains across loadings (Figure 5) show the telaprevir band growing and the polymer bands shrinking as the loading rises: for HPMC the 1450 cm⁻¹ band is about equal to the 1525 cm⁻¹ band at 2:8 and clearly lower at 5:5, and for PVPVA the band near 1675 cm⁻¹ stands above the 1525 cm⁻¹ band at 1:9 and 2:8 and below it at 4:6 and 5:5 (read from Figure 5, approximate). R12 Read naively, the domains get richer in drug as more drug is added. The authors point out that this contradicts thermodynamics, since inside a miscibility gap only the amounts of the phases should change. C20 One possibility is a non-equilibrium, kinetically trapped film; "a more likely explanation", in their view, is that spectral mixing falls as the domains grow wider and taller, so less matrix signal leaks into the domain spectrum. C21 Neither domain thickness nor phase composition was measured, so the two readings remain open. X10

Images: ratio maps and frequency maps

In all three imaged films (60:40 HPMC, 80:20 HPMCAS, 30:70 PVPVA) the IR image at 1525 cm⁻¹ matches the height image, with the strongest telaprevir signal on the raised domains (Figure 6). R13·R14·R15 The images at the polymer bands (1740 and 1676 cm⁻¹, not shown) are also strongest on the domains. The authors explain this with the physics of the method: AFM-IR signal grows linearly with the thickness of the absorbing material and also depends on its thermal expansion, effects that are roughly the same at every wavelength. R14·R15·E1 Dividing a polymer image by the drug image removes them. The ratio images match the height images, with a higher polymer/drug ratio in the matrix, "confirming" phase separation and a polymer-rich continuous phase; the authors recommend that ratio images "should be considered" for miscibility work. C23·C22

The frequency maps recorded with the IR images show the drug-rich domains of the HPMC and HPMCAS films at higher contact-resonance frequency than the surrounding polymer, which the authors read as stiffer drug-rich domains. R13·R14·C25 For PVPVA the text describes the matrix as mostly uniform and lower in frequency than the domains, and concludes that the polymer-rich phase is "slightly stiffer". R15·C25

Note by the AI writer (not discussed in the paper): a lower frequency should mean a softer phase under the rule the paper itself states (contact resonance rises with stiffness), and in Figure 6C the matrix appears as noisy mixed colours rather than uniformly low. The PVPVA stiffness ranking is better taken from the LCR spectra below. X6·E2

Softening temperatures: nanoTA

Table D. Glass transition by DSC (Table 1, mean ± SD of three) and softening temperatures by nanoTA (Table 2, mean ± SD; 5 or 6 ramps per phase in the films). M1·M2·M3·M4·R16

For pure telaprevir, nanoTA and DSC agree (102 against 102.7 °C). For the pure polymers nanoTA reads higher; the text gives the offset as about 20–60 °C, while the tables give 69 °C for HPMC, 29 °C for HPMCAS and 24 °C for PVPVA. R16.a–d·C27·X5 NanoTA values are therefore compared only with other nanoTA values below.

In 60:40 TPV–HPMC the domains soften at 130 ± 2 °C, a little above pure telaprevir, which the authors attribute "possibly" to some HPMC in the drug-rich phase; the matrix softens at 156 ± 5 °C, well below pure HPMC, "most likely" because amorphous telaprevir plasticizes it. Together these "confirmed" partial miscibility in both phases. R16.e–f·C28·C30 A softening point near 130 °C, far below the 240 °C melting point of telaprevir measured by DSC, also shows that the drug in the domains is amorphous, so the phase separation is not crystallization. C29·M1 In 80:20 TPV–HPMCAS, Table 2 gives 124 ± 14 °C for the domains and 143 ± 3 °C for the matrix, consistent with a telaprevir-rich and an HPMCAS-rich phase. R16.g–h·C31

Note by the AI writer (not discussed in the paper): the text on manuscript page 20 assigns "around 124 °C" to the continuous phase and "around 143 °C" to the dispersed phase and then calls them telaprevir- and HPMCAS-rich "respectively", which would put the drug in the matrix. Table 2, the curves of Figure 7B (domains turn over first) and every other result put the drug in the domains, so the text has most likely swapped the phases; this article uses Table 2. X1

In 30:70 TPV–PVPVA the domains and matrix soften at 131 ± 5 and 146 ± 4 °C. The authors describe this as "a similar temperature range", "readily explainable" by the similar Tgs of the two components, and suggest that the matrix softening above pure PVPVA "may be caused" by drug–polymer intermolecular interactions (ref 39). R16.i–j·C32 Their overall verdict is that nanoTA shows the physical state of each domain but is limited when the Tgs are similar, and was most useful here with the higher-Tg cellulose polymers. C48

Stiffness: LCR spectra and images

The pure-film spectra (Figure 8) place the first resonance of telaprevir at about 130 kHz, of HPMC at about 127.5 kHz and of HPMCAS at about 128 kHz (read from Figure 8B, approximate): telaprevir is slightly stiffer than both cellulose polymers, and HPMC slightly softer than HPMCAS. PVPVA and telaprevir peak at virtually the same first-mode frequency, though the PVPVA peak is much taller; at the second-largest peak, near 710 kHz, PVPVA lies slightly higher (about 711.5 against 710 kHz), so PVPVA is slightly stiffer than telaprevir. R18·C33

Table E. LCR imaging frequencies and the phase that gives the stronger signal at each (Figure 9). P8·R19·R20·R21

The LCR images "clearly indicate" two phases with different mechanical properties, and for HPMC and HPMCAS the polymer-rich regions are the softer ones. C35 For PVPVA the 129 kHz contrast comes mainly from the taller PVPVA resonance peak, while 710 kHz picks out the telaprevir-rich regions. C37 In the HPMC film some small discrete regions gave the polymer-type response and also showed weak absorbance at 1525 cm⁻¹ in the IR image, so not every small domain is drug-rich. R19·C36 Because LCR resolution is limited by the tip contact rather than by film thickness, the authors report "greatly enhanced" resolution: the smallest feature discriminated was about 98 nm across, and the 50:50 TPV–HPMCAS film, where nano-IR failed, gave good LCR contrast (data not shown). C38·C49·R19·R20

Note by the AI writer (not discussed in the paper): in Figures 6A and 9A the red and blue circles that mark these small regions and the 98 nm feature are printed on opaque white boxes that hide the image underneath, so the reader cannot check them in this manuscript version. For HPMCAS, the pure HPMCAS peak sits at about 128 kHz, the frequency used to light up the drug-rich domains; the assignment relies on the frequency shifts the authors expect in mixtures. X7·X14·C34

An optical check: pyrene

The pure HPMCAS film fluoresced evenly. In 60:40 TPV–HPMC, 80:20 TPV–HPMCAS and 50:50 TPV–PVPVA films the dispersed phase was brighter than the matrix (Figure 10), which the authors take as "further confirmation", from a method other than AFM, that the dispersed phase is drug-rich. R22·C40 These films were cast from a larger volume on quartz rather than on ZnS, so their microstructure need not match the AFM films in detail. P9·X13

Discussion

Ranking the three polymers

Putting the probes together, the authors conclude that telaprevir and HPMCAS are phase separated above 30% drug when spin-coated under ambient conditions, and that TPV–HPMCAS "appears to be the most miscible" of the three pairs, with PVPVA and HPMC showing phase separation at lower loadings. C52 The abstract states the same ranking more firmly: phase separation "was observed" with HPMC and PVPVA above 10% drug, while HPMCAS showed "a smaller miscibility gap" with separation above about 30–40%. C52·C17 They note that the marketed telaprevir product (Incivek) is a spray-dried HPMCAS dispersion, and call for work on how excipients and processing affect TPV–HPMCAS microstructure. C52·C53

Note by the AI writer (not discussed in the paper): the HPMCAS onset is stated four ways in the paper (above 40% from the height images, above 30% from the IR section, about 30–40% in the abstract, 30–40% and above 30% in the Discussion). Chemical contrast between two HPMCAS phases was shown only at 80:20; at 50:50 the spectra were identical, and no spectrum or image tests 30 or 40%. The onset therefore rests on the height images. Likewise, for HPMC and PVPVA domains are already present at 10% drug, so "above 10%" in the abstract reads best as "at or below 10%", and at 10–20% the evidence is again topography plus spectra of the domains alone. X2·X3·X4

Why each probe works or fails

The Discussion turns the results into a guide to the methods. Two properties of a sample decide which probe can separate its phases: how different the two phases are in composition and properties, and how large the domains are compared with the volume each probe samples. C46·C50

How the authors explain the success and failure of the probes; green and red mark where nano-IR works and where it fails, not better or worse formulations. C46·C47·C49·C50

For nano-IR, the authors name two possible reasons for the failure at 50:50 HPMCAS: the small domains (122 nm wide, 13 nm high) and the low spectral contrast that follows from the relatively high TPV–HPMCAS miscibility, or both; IR radiation may also penetrate the thin domain so that the material beneath contributes. C46 NanoTA can confirm phase separation and the amorphous state but is limited when the components have similar Tgs. C48 LCR is independent of IR absorptivity and thickness and is therefore "especially beneficial" for very small domains and weak IR absorbers. C49 The authors close the Discussion by calling nano-IR, nanoTA and LCR "clearly powerful tools in miscibility evaluation", particularly in concert with AFM imaging, because together they give both domain size and phase composition; the abstract is more measured ("appear to be promising"). C51

What it suggests for formulation

The authors explain why composition and domain size matter beyond the yes-or-no of phase separation. The lower the polymer content of the drug-rich phase, the greater the risk of failure by crystallization during storage, since polymer in a separate phase cannot inhibit crystallization as well as in a miscible blend; and domain size "is likely to" affect how the ASD performs on dissolution. C41·C42 They also note that TPV–HPMCAS microstructure "seemed to be somewhat dependent" on environmental moisture, with day-to-day changes in films around 20–50% drug, an observation they consider worth further study because amorphous products always contain some water. C18·R23

Note by the AI writer (not discussed in the paper): for a formulator, the practical reading is that with telaprevir the choice of polymer moves the onset of phase separation in films several-fold, from at most 10% drug with HPMC and PVPVA to about 30–40% with HPMCAS, the polymer of the marketed product. With PVPVA a DSC scan would be expected to show one Tg even for a separated film, since the two Tgs are about 5 °C apart. These onsets come from thin films cast from methanol:dichloromethane with uncontrolled humidity; spray-dried or extruded material, other solvents, or stored product could separate at other loadings. R4·R5·R6·M1·M4·X9

Limitations

Limits the authors state. They write that it is "hard to conclusively determine" phase separation or composition from height images alone. C9 They describe AFM-IR resolution as a limit for very small domains ("analytically challenging"), give two unresolved causes for both components appearing in both phases, and call the composition trend across loadings more likely an artefact of the method than a real effect. C15·C14·C21 They note that humidity could not be controlled during measurement although water is known to promote phase separation. C18 They state that nanoTA has limited use when the Tgs are similar. C48 They accept that the LCR imaging frequency may differ from the pure-component peaks because of probe friction and partial miscibility. C34 And they note that spray-dried particles would need embedding and microtoming before these methods could be applied. C53

The AI writer's own judgement.

• The evidence that the domains are drug-rich is strong for one high-loading film per polymer, where four independent signals agree. The loading onsets, which carry the ranking of the polymers, rest mainly on height images. R8·R9·R10·R16·R22·X3·X4
• No phase composition is quantified; band ratios rank the phases but give no weight fractions. C12·X3
• Parts of the text disagree with the paper's own data: the HPMCAS nanoTA phases are swapped relative to Table 2, the nanoTA–DSC offset is given as 20–60 °C where the tables give 24–69 °C, and the PVPVA frequency map is read as "stiffer" at lower frequency. X1·X5·X6
• The two results that favour LCR (the 98 nm feature and contrast at 50:50 HPMCAS) cannot be checked: the first is hidden under an annotation box and the second is not shown. X7·R20
• The claim that similar Tgs rule out DSC was not tested: no blend was run by DSC, and the HPMC and HPMCAS Tgs differ from telaprevir by 37 and 19 °C. X8·C44
• Replication is thin (4–6 points or ramps per phase, one film per condition), film thickness and humidity are not reported, and different techniques used different films (30:70 or 50:50 PVPVA; quartz substrates for fluorescence). X11·X9·X13

Conclusion

The paper shows that an AFM fitted with an infrared, a thermal and a mechanical probe can find and characterize nanoscale phase separation in drug–polymer films where DSC would struggle. In spin-coated telaprevir films the drug-rich phase forms discrete domains, from below 50 nm to a few hundred nanometres across, in a polymer-rich matrix, and each phase still contains some of the other component. C54·C14 Telaprevir stays mixed with HPMCAS to about 30–40% drug, the highest of the three polymers, while HPMC and PVPVA separate from 10% drug or below. C52·C17·R4·R6 No single probe covers every case: nano-IR needs domains large and different enough, nanoTA needs Tgs that differ, and LCR, which reached about 98 nm, carries the smallest domains. C46·C48·C49 What the paper leaves open is the actual composition of the phases, the role of moisture, and whether the same picture holds in spray-dried material. X3·C18·C53