How glassy vemurafenib crystallizes: four polymorphs, their growth rates, and why the fastest grower wins

Background and motivation

Most new drug candidates dissolve poorly in water; by the biopharmaceutics classification system, 90% of development compounds and 40% of marketed drugs fall into this group (Loftsson 2010, ref 1). Making the drug amorphous is one of the main ways around this. An amorphous solid lacks a crystal lattice, so less energy is needed to dissolve it, and both solubility and dissolution rate rise. The price is physical instability: an amorphous drug is thermodynamically metastable or unstable and tends to recrystallize during storage or during dissolution, and with it goes the solubility advantage. Inhibiting recrystallization is therefore central to amorphous formulations, and that in turn needs a basic understanding of how, and how fast, amorphous drugs crystallize C1.

The authors distinguish two routes from the amorphous state to the crystal. Melt crystallization happens while a melt is cooled. Cold crystallization happens when a glass, or the supercooled liquid made by heating it, is warmed. The glass transition temperature ($T_g$) separates the two amorphous states: below $T_g$ the material is a glass, and between $T_g$ and the melting point ($T_m$) it is a supercooled liquid. Cold crystallization has been studied widely in polymers, where amorphous chain segments rearrange into a crystal (Nagarajan 2015, ref 2; Zhou 2016, ref 3; Wellen 2005, ref 4); in a small-molecule drug, single molecules rearrange into the lattice C2.

Pharmaceutical work on amorphous materials had concentrated on molecular mobility (refs 5–7), on polymeric additives (Tian 2015, ref 8) and on water (Purohit 2015, ref 9). Which crystal phases appear when an amorphous drug crystallizes, and in what order, had been followed in comparatively few studies (refs 10–16), several of them by Yu and co-workers on concomitant polymorphs and cross-nucleation C3.

Vemurafenib is a demanding case. Its aqueous solubility is below 0.1 µg/ml (Shah 2013, ref 17) M1. Preliminary work from the same group found a strong crystallization tendency and a maximum crystal growth velocity of 106 ± 4 µm/s at 250 °C, fast compared with values reported for other compounds (Trasi 2014, ref 18) R25. On that basis the amorphous drug "might be anticipated" to be physically unstable, yet the commercial formulation contains vemurafenib in amorphous form, as an amorphous solid dispersion (ASD) R26·C4. The aim of the study was therefore twofold: to measure the crystallization kinetics of amorphous vemurafenib as a function of temperature and to map its polymorphic landscape A1.

Background concepts

The supercooled liquid as a crystallization medium. Between $T_g$ (108 °C for vemurafenib at 10 °C/min) and $T_m$ (271 °C for the stable crystal) the amorphous drug is a liquid that is mobile enough to crystallize M1·R19. Holding it at a fixed temperature in this range (here called the cold-crystallization temperature, $T_{cc}$) and watching crystals appear under a polarizing microscope is the core experiment of the paper P2.

Polymorphs, morphologies and concomitant polymorphism. Polymorphs are different crystal structures of the same molecule. A single polymorph can grow in several shapes (morphologies), such as needles or spherulites, depending on conditions; a spherulite is a round aggregate of crystal fibres radiating from a centre. Concomitant polymorphism means that two or more polymorphs crystallize at the same time from the same liquid. The paper relies on Raman microscopy to tell polymorphs from morphologies: crystals with one structure give one spectrum, whatever their shape P4·C58.

Monotropy and enantiotropy, and the heat-of-fusion rule. Two polymorphs are monotropic if one is more stable than the other at every temperature below melting, and enantiotropic if the order switches at a transition temperature below the melting points. Burger and Ramberger's heat-of-fusion rule (ref 28) is a quick test that the paper names but does not print: if the higher-melting form also has the larger enthalpy of fusion, the pair is monotropic E3.

The Hoffman equation. To estimate how much free energy is gained when a liquid crystallizes at a temperature $T$ below the melting point, the authors use the Hoffman approximation (Eq. 1, cited to Andronis and Zografi, ref 29) E1:

In plain terms: the driving force for crystallization grows from zero at the melting point as the liquid is cooled further below $T_m$, and it scales with the enthalpy of fusion $\Delta H_m$. Here $\Delta G_v$ is the free-energy difference between liquid and crystal, $T$ the crystallization temperature and $T_m$ the melting temperature of that crystal form; temperatures must be absolute (kelvin) for the published curves to be reproduced E1·R22. Applied to two polymorphs, the equation gives two free-energy curves; where they would cross is the transition temperature, and if that crossing lies above both melting points it is only a "virtual" transition temperature, a sign of monotropy C51.

The bell-shaped growth curve. Crystal growth from a supercooled liquid is slow just below $T_m$, because the driving force is small, and slow close to $T_g$, because the molecules hardly move; in between, growth passes through a maximum (Okui, ref 20) C17. Naito (ref 21) noted that this maximum, the "maximum crystal velocity", typically lies at about 0.94 of the melting temperature E2:

In plain terms: the fastest growth is expected a few percent below the melting point on the absolute scale. For vemurafenib α ($T_m$ = 271 °C = 544.15 K) this gives about 511.5 K (238 °C); the authors give 240 °C R24·E2.

Ostwald's law of stages, nucleation and growth zones. Ostwald's law of stages says that the least stable polymorph crystallizes first and then converts step by step to more stable ones (ref 22). Concomitant polymorphism has been explained by similar nucleation rates of two forms (refs 23–25), but Yu (ref 15) pointed out that a fast-nucleating polymorph may grow slowly, so that a slow-nucleating but fast-growing polymorph can be the one that is seen or even dominates; Yu also described cross-nucleation, the heterogeneous nucleation of one polymorph on another (refs 13, 16) C37. The paper builds on the related idea, taken from polymer science (Woo, ref 27), that each polymorph has its own temperature "zone" for nucleation and another for growth C42.

Methods in detail

The study uses one compound and no excipient. Everything starts from one glass, made by melting the as-received crystal and quenching it, and the glass is then crystallized at many temperatures or heated along defined paths. Four contributions can be separated: a map of which crystal forms grow at which temperature, the growth rate of each form, the paths by which the forms convert, and a thermodynamic ranking of the two main forms.

Making the glass

The as-received vemurafenib (more than 99% pure, α-form) was placed between two coverslips, held at 285 °C for 1 min to melt it completely and quenched to room temperature on a metal block M1·P1. The sample did not change colour, HPLC showed no detectable degradation, and the polarizing microscope showed no crystals R2. Slow cooling does not work for this drug: in the DSC, a melt cooled at 10 °C/min crystallized as α already at 248.9 °C, only about 25 °C below its melting point, which the authors read as "suggesting that vemurafenib is a relatively poor glass former" R1·C5. Only faster cooling avoided crystallization and gave the glass R2.

Contribution 1: a temperature map of the crystal forms

① Problem. The DSC trace of the heated glass (Figure 3A) shows two exotherms, a small endotherm–exotherm pair and a final melting peak, which already hints that a polymorph other than α crystallizes first R3·C6. The question is which forms appear, and where.

② What they did. Glass films on a Linkam THMS 600 hot stage were held at fixed temperatures from 120 °C, just above $T_g$, to 270 °C, just below $T_m$, and photographed at regular intervals under a polarizing microscope P2. Each crystal type was then identified on the spot by confocal Raman microscopy (532 nm laser, 10× objective, spot diameter about 2.1 µm, resolution about 2.0 cm⁻¹), which the authors consider fine enough to record a spectrum from each spherulite P4·C58. Samples crystallized at 120, 130, 150, 190, 230 and 260 °C were also measured by PXRD (Cu Kα, 4.5–30° 2θ) P5. ATR-FTIR was possible only for the amorphous form, α and β, the forms that could be made in sufficient quantity P6.

③ Why it works. Microscopy tells morphologies apart, but only a structural fingerprint tells polymorphs apart. Raman microscopy supplies that fingerprint locally, so a mixture of three forms in one film can be resolved crystal by crystal, something PXRD of the whole sample cannot do when one form is a minor fraction C47.

Contribution 2: growth rates over the whole Tg–Tm range

From the time-lapse images, crystal size was measured in Elements software, plotted against time in Prism, and the growth rate taken as the slope P2. Rates were obtained for β, γ and δ and for the three α morphologies separately (Figure 7) R12. The number of crystals measured per temperature and the meaning of the error bars are not stated, and the paper does not say how the films were brought to $T_{cc}$ X9.

Contribution 3: following the conversions

Three kinds of experiment show how the forms turn into one another. A sample first crystallized at 190 °C, containing α, β and γ, was heated at 10 °C/min under the microscope, with Raman spectra of each region (Figure 9) R15. A small area that, after crystallization at 150 °C, held only δ nuclei was heated in the same way (Figure 10) R17. Isothermal holds followed the β-to-α front at 170 °C, with Raman and PXRD measured after cooling (Figure 8), and at 230 °C (Figure S1), and the γ-to-β change at 190 °C (Figure S3) R13·R14·R16. In parallel, the glass was heated in the DSC (TA Q2000) at 10 °C/min, and samples heated at the same rate were quenched at 170 and 236 °C for PXRD (Figure 3B) P3·R5. To separate β melting from the recrystallization that follows it, a glass was annealed at 150 °C for 20 min, which gives β with very little δ, and then heated at 100 °C/min (Figure S4) R21. A solvent-mediated conversion at room temperature was also run, but its data are not shown R23·P8.

Contribution 4: ranking α and β thermodynamically

Pure γ and δ could not be prepared, and both convert quickly, so only the α–β pair was ranked C49. α for this purpose was made by cooling the melt at 10 °C/min, β by holding the melt (made at 280 °C) at 190 °C for 20 min; their DSC melting endotherms (Figure 12) gave $T_m$ and $\Delta H_m$ P3·R20. These values were put into the heat-of-fusion rule and the Hoffman equation (Figure 13) E3·E1·P7. The methods paragraph and the Figure 2 caption describe the DSC programme slightly differently (end temperature 280 or 285 °C, hold 2 or 1 min, cooling to 25 or 20 °C), and the heating rate for Figure 12 is not stated X5.

Experimental analysis

What the DSC of the glass shows

The as-received crystal melts sharply at 271.3 °C with $\Delta H_m$ = 55.9 kJ/mol; cooled at 10 °C/min, the melt crystallizes from 248.9 °C (exotherm peak 246.7 °C in Figure 2) with $\Delta H_c$ = 48.7 kJ/mol, and on reheating it melts again at 271.5 °C as α, as XRPD confirmed R1. The glass behaves very differently (Figure 3A). After the glass transition at 108 °C come a sharp exotherm and a broad one, then a tiny endotherm immediately followed by a small exotherm, and finally the melting of α R3·M1.

Table A. DSC events of the heated glass (Figure 3A and text) and their interpretation by the authors R3·R4·C31·C33·C34·C35.

PXRD of samples quenched during heating supports the first two assignments: at 170 °C the pattern is that of β, and at 236 °C, near the end of the broad exotherm, peaks of α have appeared but several β peaks remain, so the β-to-α change was incomplete and, judging by eye, happened at the surface of the β crystals R5·C33. The authors explain why the δ-to-β step leaves no DSC signal: "probably due to the low amount present" C32. For the pair of peaks at 250 °C they write that "it is likely that the endothermic melting is offset by the exothermic crystallization event" C34. Heating a β-rich sample at 100 °C/min made the β melting peak (near 252 °C, read from Figure S4) "much more visible", but it was still followed at once by a recrystallization exotherm and then α melting near 274 °C R21·C36.

Note by the AI writer (not discussed in the paper): the enthalpy of the final α melting peak is printed as 51.0 kJ/mol on p.6 and as 53.8 kJ/mol on p.18 for the same event, and the as-received α melts with 55.9 kJ/mol, which equals the 114 J/g of Table 2 X1.

Four polymorphs in six morphologies

Isothermal crystallization of the supercooled liquid between 120 and 270 °C produced six morphologies, which Raman microscopy sorted into four polymorphs (Figure 4) R7·R8·C7. The panel letters printed inside Figure 4 do not match the caption and some text references; the table below follows the letters in the figure X4.

Table B. The six morphologies as labelled in Figure 4 R7. Figure 5 and the text give somewhat different ends for several ranges, for example β growth to about 240 °C and γ to 190 °C R9·X3.

The three α morphologies gave identical Raman spectra, also side by side in one sample at 200 °C (Figure S2), and the same melting point, which first "suggests" and later, in the authors' words, confirms one crystal structure R8·C8. The other morphologies gave clearly different spectra, mainly between 1550 and 1700 cm⁻¹ (Table 1) R18. The ring C=C stretching band sits at 1597 cm⁻¹ in amorphous, α, β and δ, with a second, weaker band or shoulder at 1608–1613 cm⁻¹; γ shows two strong bands at 1600 and 1610 cm⁻¹. The C=O stretching band is at 1640 cm⁻¹ in the amorphous form and α, at 1640 and 1655 cm⁻¹ in β, and at 1624 cm⁻¹ in δ R18·C30. The authors read the shifted C=O bands as "suggesting different intermolecular interactions between the C=O and N-H groups" in the different crystal structures, and the spectra as "strong evidence" for a tetramorphic system C29·C28.

Table C. PXRD peaks and DSC data of the solid forms (Table 2; Cu Kα, λ = 1.542 Å); δ peaks from the text, p.22 R19·R11.

The as-received α gives the same diffractogram as the crystalline vemurafenib described by Shah et al. (ref 17) C39. A patent describes two further solid forms made from solvent, Form A, which "appears" to be a solvate, and Form B, made by drying Form A; the authors state that neither their peaks nor their melting points match any form found here C40. Note by the AI writer (not discussed in the paper): Form B melts at 270.9 °C, within 0.1 °C of α, and four of its six listed peaks lie within 0.2° of α peaks, so its distinction from α rests on the peaks at 5.9 and 23.7° and on the enthalpy of fusion X14.

The temperature decides the mix

Which forms appear depends strongly on the crystallization temperature (text, Figure 6A) R10·R11:

Table D. Forms present after isothermal crystallization, from the text and Figure 6A R10·R11·C59.

The abstract sums this up as two or three of the four polymorphs in each sample between 120 and 240 °C, α or β typically dominating, and only α at higher temperature C59. Note by the AI writer (not discussed in the paper): by the text's own statements, a sample crystallized at 165 °C should contain β only, since no α forms at 165 °C or below, no δ was seen at 165 °C and γ nucleates only from 170 °C; Figure 7 nevertheless plots a δ growth rate at 165 °C X16·X3. On the printed scale of Figure 6A, the 120 °C pattern also shows a strong unassigned peak near 5.3° (present at 130 and 150 °C too) and peaks near 13.6, 15.1 and 19.2° at α positions, while the four δ-specific peaks of the text are hard to resolve; the Raman spectrum remains the stronger evidence for δ X15.

Growth rates

Figure 7 is the quantitative core of the paper. The values below are read from it and are approximate R12.

Table E. Crystal growth rates in µm/s, read from Figure 7 (approximate); — means no point plotted. The number of crystals and the error type are not stated R12·X9.

Three patterns stand out. First, β grows very slowly below 150 °C, faster above it, and jumps between 180 and 190 °C to rates that exceed those of every other form up to 230 °C, the highest temperature at which β was measured R12·C21. Second, α-needles follow a bell-shaped curve with the maximum at 250 °C (about 106 µm/s, in line with the 106 ± 4 µm/s of the earlier study) and a sharp fall towards the melting point R12·R25. The authors explain the shape by a driving force that rises and a molecular mobility that falls on cooling ("can be readily explained"), note that Naito's 0.94 $T_m$ rule predicts 240 °C, "close to" the observed 250 °C, and find it "apparent" that vemurafenib grows very rapidly compared with other drug-like molecules C17·C18·C19. The α-spherulites, measured at 170–200 °C, follow the same rising trend as the needles C16. Third, γ, δ and the brown feathers barely grow: γ reaches about 4 µm/s at 190 °C, δ stays near zero, and brown feathers advance at about 1 µm/s or less, with 0.3 ± 0.1 µm/s measured on β at 220 °C R12·R6.

How the forms convert

All conversions run towards α, through β C54:

• γ to β. In a sample crystallized at 190 °C and heated at 10 °C/min, γ turned into a new phase once the temperature exceeded 200 °C, and the conversion was complete around 220 °C (images at 190, 212 and 222 °C, Figure 9). The new phase melted at 254–255 °C together with the original β, "indicating" the same melting point, and Raman showed it to be β. Held at 190 °C, the same conversion needed 60 min (Figure S3). "It can be concluded" that γ is metastable with respect to α and β at 190 °C R15·R16·C24·C25.
• δ to β. A δ-only area grown at 150 °C turned into β at around 220 °C on heating; Figure 10 shows a β front sweeping through δ between 219 and 225 °C. The δ melting point could therefore not be measured, and δ is metastable to α and β over 170–225 °C R17·C27.
• β to α in the solid state. At 170 °C a dark brown α front grew out of a diamond spherulite into the surrounding β within 5–25 min; Raman of the front matched α, and the PXRD pattern after 60 min showed α peaks added to those of β (Figure 8). The same front appeared at 230 °C (Figure S1). The brown feathers "always" started at an interface between α and β, and β regions without α contact did not transform. The authors call this change very slow, "presumably due to a high activation barrier" R13·R14·R6·C14·C11.
• β to α through the melt. On heating, β crystals that had not converted melted at about 250 °C, and α-needles grew from the edges of α-diamond spherulites (Figure 9D, E). In the authors' reading, the melting β supplies drug for the needles R15·C34. Note by the AI writer (not discussed in the paper): the separate β sample of Figure 12 shows no recrystallization after melting, so this route is seen where α crystals are already present or at fast heating (Figure S4) X11.
• β to α in solvent. At room temperature, β converted to α in a solvent; the data are not shown R23.

Thermodynamics of α and β

In the text, α melts at 271 °C with $\Delta H_m$ = 53.8 kJ/mol and β at 248 °C with $\Delta H_m$ = 33.3 kJ/mol; in Figure 12 the peaks are at 271.5 and 247.9 °C R20. α has both the higher melting point and the larger enthalpy of fusion, so by the heat-of-fusion rule the two forms are monotropically related; the authors find this "consistent with" β always converting to α between 25 and 250 °C and with the exothermic β-to-α change C50. The Hoffman curves (Figure 13) place α below β at every temperature studied; read from the figure, the free energy of the crystal relative to the liquid near 115 °C is about −11.0 kJ/mol for α and about −6.3 kJ/mol for β, and the two curves cross only at a virtual transition temperature of 319 °C, above both melting points R22·C51. Note by the AI writer: these numbers reproduce from Eq. 1 with the printed inputs when temperatures are converted to kelvin (318.7 °C for the crossing). Figure 13 plots the negative of $\Delta G_v$, and the crossing moves to 313.6 °C if the as-received α value of 55.9 kJ/mol is used instead of 53.8 E1·X17·X1. Either way the pair is monotropic.

The authors' overall ranking is α > β > γ or δ (abstract) C55. γ and δ are below β because both turn into β spontaneously; their order relative to each other "could not be established" (Figure 5 caption) C57.

Discussion

Why the least stable form can win: the authors' kinetic reading

The central puzzle is that α has the larger driving force for crystallization at every temperature (Figure 13), yet β dominates the outcome between 120 and 190 °C R22·R10. The authors resolve it with nucleation and growth zones: each polymorph nucleates well only in its own temperature window and grows fast only in its own window, and the overlap of these windows, and the size of the rates within them, decide which forms appear at a given temperature and in what amounts ("This can be explained by considering that different polymorphs will have different nucleation and growth zones") C42. Read window by window, the argument runs as follows.

The authors' chain from nucleation and growth windows to the crystallization outcome C42·C46·C47·C48·C26.

For α, the authors argue that its nucleation rate must be high near 250 °C, the temperature of fastest growth, because the melt crystallizes spontaneously on cooling (Figure 2). Since the nucleation maximum usually lies below the growth maximum (Okui, ref 20), and since α never appears at 165 °C or below, "it appears that the favorable nucleation zone for the α-form is between approximately 170 and 270 °C" C43. For β, "nucleation appears to be the limiting step": few β nuclei form above 150 °C, but they grow so fast that β fills large regions below 200 °C; above 200 °C β nucleation becomes less favourable, and although β still grows faster than α there, α "starts to be the predominant form due to more favorable nucleation" C46. γ is the textbook case of a form that "nucleates rapidly but grows slowly", so it stays a small fraction that PXRD barely sees C47. δ "exhibits a fast nucleation rate" at low temperature, but grows extremely slowly and is engulfed by β C48·C26. In the DSC run of the glass, crystallization peaks near 160 °C, where β grows fastest, so β "would be expected" to dominate C44.

The authors' conclusion is that "the cold-crystallization of amorphous vemurafenib is controlled by factors other than the thermodynamic driving force" ("These observations indicate"), in line with literature observations that metastable forms are more probable at larger undercooling (Ishihara, ref 30) C52·C53. The authors frame this against Ostwald's law of stages, in which forms appear one after another from least to most stable, and against the view that concomitant polymorphs share similar nucleation rates, citing instead Yu's point that a slow-nucleating but fast-growing form can dominate; in vemurafenib the forms crystallize together, "due to the balance between nucleation and growth rate of the different forms" C37·C45.

Note by the AI writer (not discussed in the paper): the growth half of this argument is measured, the nucleation half is not. The paper reports nucleation only as microscope impressions ("low nucleation density", "erratic", "high nucleation density"), and the α nucleation zone is inferred from a cooling DSC run and a cited general rule. The reading is plausible, but a nucleation rate or nucleus count per form and temperature would be needed to test it X8.

Morphology of α and the role of undercooling

α grows as shiny "diamond" spherulites at 170–200 °C and as smooth needles with a sharp melt interface above 200 °C; above 210 °C the spherulites no longer appear R7·R10. The authors relate this to Jackson's account (ref 19): at small undercooling organic crystals often grow faceted or spiky, at larger undercooling isotropically as spherulites, a change attributed to kinetic roughening C15. The two morphologies lie on one growth-rate trend, and both differ from the brown feathers, which form inside solid β rather than from the liquid and grow far more slowly C16·C12.

What it suggests for formulation

The authors' own conclusion for practice is general: the three new polymorphs "highlight the rich polymorphic landscape that can be accessed by evaluating crystallization in supercooled liquids", and the results underscore the complexity of evaluating the crystallization of amorphous drug forms C56. They also note that vemurafenib, with its fast growth, "might be anticipated" to be unstable as an amorphous form, yet it is marketed as an ASD C4·R26.

Note by the AI writer (not discussed in the paper): three practical points follow for amorphous vemurafenib, all for the pure drug, since no polymer was tested. First, the crystals that appear when the amorphous drug crystallizes above $T_g$ are likely to be metastable β (or δ close to $T_g$), not α; a stability test that looks only for α peaks could miss them, and γ was found by Raman microscopy although it was too minor for PXRD R10·C47. Second, the form depends on temperature: β-rich product at 150–190 °C, α above 200 °C R10. Heat exposure during processing could therefore leave different forms from those formed during storage, although this paper does not test either situation. Third, growth near 100 µm/s means that a single nucleus can cross a 100 µm region of pure drug in about a second at 210–250 °C R12; this is why crystal growth, not only nucleation, has to be suppressed to keep this drug amorphous.

Limitations

Limits the authors state. Phase-pure γ and δ were difficult to obtain and both convert rapidly, so "only the thermodynamic relationship between α-form and β-form could be evaluated in more detail" C49. Neither γ nor δ has a melting point, because both turn into β first C23·C27. The stability relationship between γ and δ "could not be established" C57. α is called the most stable form "at least at the higher temperatures studied", with room-temperature stability supported by a solvent-mediated conversion whose data are not shown C38·R23.

The AI writer's own judgement (1): the evidence for four forms. α and β are well established: distinct Raman and IR bands, distinct PXRD patterns, separate DSC melting peaks and a conversion observed directly R18·R19·R20·R13. γ and δ rest on morphology, Raman spectra and their conversion to β; Table 2 has no PXRD peaks, melting point or enthalpy for either, γ was not seen by PXRD at all, and the δ peaks come from a sample that also contained β X6. On the printed scale, the 120 °C pattern of Figure 6A cannot confirm the δ peak list and shows peaks at α positions X15. "At least a tetramorphic system" is therefore a fair statement for α and β plus two Raman-defined forms whose structures are not known C41. The conclusion's ranking "followed by the β, γ and δ forms" reads as if γ were more stable than δ, which the Figure 5 caption says could not be established X7.

The AI writer's own judgement (2): numbers that do not agree across the paper. The enthalpy of fusion of α is given as 55.9, 53.8 and 51.0 kJ/mol in different places, and the 319 °C virtual transition holds only for 53.8 kJ/mol X1. Melting temperatures of β range from 247.9 to 255 °C depending on method and sample X2. Several temperature ranges differ between Figure 4, Figure 5 and the text, the lowest δ temperature (110 °C) lies below the lowest stated crystallization temperature (120 °C), and the Figure 4 caption labels two panels D X3·X4. Growth rates are given without the number of crystals or the error type; the β error bar at 190 °C spans about 45–94 µm/s X9. The β sample used for its melting data was crystallized at 190 °C, where α and γ also form, and its melting endotherm may overlap recrystallization, so its 33.3 kJ/mol could be a lower bound X10·X12. None of these changes the qualitative picture (four forms, β kinetically favoured, α stable, α–β monotropic), but the ends of the temperature ranges should be read as approximate to about ±10 °C.

The AI writer's own judgement (3): scope. The work concerns the pure drug, heated from a melt-quenched glass, between 120 and 270 °C, in the dry state A1·P1. Crystallization below $T_g$, in water or in a polymer matrix, the conditions that matter for an ASD in storage or dissolution, were not examined, and the source is the accepted manuscript rather than the final edited version.

Conclusion

Lu and Taylor show that amorphous vemurafenib, crystallized from its supercooled liquid between 120 and 270 °C, produces four polymorphs: the known α, in three morphologies, and three new metastable forms, β, γ and δ, which appear two or three at a time depending on temperature C54·C59. The metastable forms convert towards α through β, by slow solid–solid fronts that start only where β touches α and, near 250 °C, through the melt C14·C34. α and β are monotropic, with α the stable form at all temperatures C50·C51. Even so, β dominates between 150 and 190 °C because it grows fastest there, which led the authors to conclude that the crystallization outcome is set by nucleation and growth kinetics rather than by thermodynamic driving force; the growth rates support this directly, the nucleation part only qualitatively C52·R12·X8.