Why nilutamide forms a glass easily, and why its glass nucleates best far below the glass transition

Background and motivation

Amorphous drugs are used to raise the transient solubility and dissolution rate of poorly water-soluble compounds, but they are metastable and tend to crystallize during manufacture and storage, so formulators need to understand what drives that crystallization (Hancock and Zografi 1997, ref 1; Bhugra and Pikal 2008, ref 2) C1. When a melt is cooled without crystallizing it becomes a supercooled liquid, and on further cooling the molecules lose mobility until the system falls out of equilibrium at the glass transition temperature, $T_g$; most pharmaceutical molecules form "fragile" liquids whose viscosity rises steeply near $T_g$ (refs 3–5) C2. Because mobility in the glass is low, the glass is held to be more resistant to crystallization than the supercooled liquid (Zhou et al. 2008, ref 6) C2.

Crystallization needs two steps, nucleation and growth, and the two depend on temperature in different ways. Growth speeds up with temperature as mobility rises, then slows near the melting point $T_m$ as the driving force vanishes; Naito and Miura proposed that the growth maximum "typically" lies at 0.94 $T_m$ (refs 7, 8) C3. The nucleation rate also passes through a maximum, at a lower temperature, and the size of the two rates and the overlap of their temperature ranges decide how easily a compound forms a glass (ref 9) C3. A practical consequence has been the advice that storing an amorphous material about 50 °C below $T_g$ prevents crystallization, since molecules there should be too immobile to nucleate or grow (Hancock, Shamblin and Zografi 1995, ref 10) C4.

That advice had already been questioned. Vyazovkin and Dranca found that indomethacin glasses stored 55 °C below $T_g$ gave a small melting endotherm on reheating, a sign that nuclei had formed in the cold glass (ref 11), and Okamoto and Oguni reported an organic compound whose nucleation zone lay well below $T_g$ (ref 12) R34·C4. On the other hand, Andronis and Zografi placed the nucleation maxima of two indomethacin polymorphs at or above $T_g$ (ref 13) R32·C4. Whether, and why, a drug might nucleate faster deep in the glass than above $T_g$ was therefore open.

The paper takes up nilutamide, a small molecule used to treat prostate cancer, with one rotatable bond between its nitrophenyl and imidazolidinedione rings, one N–H hydrogen-bond donor and three acceptors (two carbonyls and a nitro group) (Figure 1) M1. The authors had found that it nucleates a metastable polymorph faster at temperatures far below $T_g$, and they set out to describe this behaviour, to explore why the metastable form nucleates at low temperature, and to measure the growth rates of the different polymorphs (Introduction). Two polymers, poly(acrylic acid) (PAA) and poly(vinylpyrrolidone) K12 (PVP), were added in part of the work to see how hydrogen-bonding additives change growth M2·M3.

Background concepts

Nucleation zone and the two-step method. A nucleation zone is the temperature range in which nuclei form at an appreciable rate. Nuclei are far too small to see, so they are usually detected indirectly. One route, used here with DSC, is to cool a melt to a chosen temperature, hold it briefly and reheat it: if nuclei formed on cooling, crystals grow from them during reheating and give an exotherm; lowering the end temperature run by run maps where nucleation starts (Naito and Miura, ref 8; Oguni, ref 18) P1. The second route is the two-step method of James (ref 15): nuclei are produced by holding the glass at one temperature, then grown at a higher temperature where growth is fast but new nucleation is slow, and finally counted under the microscope P7. Both methods report nucleation through a grown crystal, so they work only where growth at the development temperature is fast, which is the case for nilutamide.

Polymorphs, monotropy and their thermodynamics. Polymorphs are crystal forms of the same molecule with different packing. When the higher-melting form also has the higher enthalpy of fusion, the pair is monotropic: the higher-melting form is the more stable at every temperature below melting, and the other can only convert into it (the thermodynamic rules of Burger and Ramberger, ref 19) E2. A metastable form often crystallizes first and converts later; Ostwald's rule of stages names this sequence C44. If one polymorph nucleates on the surface of another, this is called cross-nucleation R28.

Conformers and tautomers. A molecule with a rotatable bond can adopt several conformers, which differ in energy. In nilutamide the relevant coordinate is the angle between the two rings, and a single-molecule density functional calculation can map the energy as a function of that angle P10. In solution crystallization it has been proposed that crystallization, and in particular growth, becomes unfavourable when the crystal contains a high-energy conformer (refs 29–31) C49. Polymorphs built from different conformers are called conformational polymorphs; the metastable ibuprofen polymorph that nucleates far below $T_g$ is one (refs 25, 26) R33. Tautomers differ in where a hydrogen sits; in a keto–enol shift an N–H hydrogen moves to a carbonyl oxygen to give an O–H C47.

What the spectra report. In infrared spectra a hydrogen bond lowers the stretching wavenumber of the donor N–H and of the acceptor group, so shifts to lower wavenumber are read as stronger or more hydrogen bonding, and shifts to higher wavenumber as weaker bonding C16. In 13C solid-state NMR (ssNMR), the two methyl carbons on the imidazolidinedione ring give two separate signals near 25 ppm (a doublet) when their environments differ and a single signal (a singlet) when they are equivalent; the authors use simulated spectra of conformers to relate this to the ring angle R25·R27.

Methods in detail

The study has four strands that run side by side: locating where the melt and glass nucleate, measuring how fast each crystal form grows, characterizing the forms and their conversion, and testing whether molecular conformation can explain the pattern. The figure below shows how they connect.

Strand 1: where and what the melt nucleates

Problem. A melt that does not crystallize on cooling or reheating could be slow to nucleate, slow to grow, or both, and a glass can still nucleate where nobody looks. The authors needed to find at which temperatures nilutamide nucleates and which crystal form it produces there.

What they did. In a TA Q2000 DSC, about 5 mg of nilutamide in a pin-holed hermetic pan was heated to 160 °C, just above melting, cooled at 15 °C/min to a chosen temperature, held for 1 min and reheated at 10 °C/min; the end temperature was lowered run by run (Figure 3) P1. Because the bulk glass nucleated very slowly but the free surface much faster, the counting experiments used an exposed surface: about 3 mg was melted and cooled between a glass slide and a coverslip, the coverslip was removed, and the film was held for 30 min on a hot stage at temperatures from −30 to 60 °C, then for 1 min at 65 °C so that any nuclei grew large enough to see; the nuclei were counted, and each value is the mean of 5 experiments (Figure 5) R1·P7. Above 60 °C the film started to flow, which ended the series R9. To learn which form nucleates, around 50 DSC runs were made at each of two temperatures: the melt was cooled to −30 or 60 °C, held 30 min and reheated at 10 °C/min to 120 °C, and the product was identified by Raman spectroscopy P8.

Why it works, and its limits. All three experiments see nuclei only after they have grown, which nilutamide allows because its crystals grow quickly once above $T_g$ C39. The count in Figure 5 is a number of nuclei per sample after a fixed time, not a rate per area and time, and the observed area is not given; the 60 °C hold also lies only 5 °C below the 65 °C growth step X7.

Strand 2: how fast each form grows

What they did. For growth, 3–5 mg of powder was melted at 170 °C between two cover slips and quenched to room temperature; because the glass does not nucleate easily, the edge was seeded with the polymorph of interest, and the advancing crystal front was photographed at intervals on a Linkam hot stage under a polarizing microscope P2·P3. Crystal length increased linearly with time, and its slope is the growth rate; at the lowest temperatures the front became irregular and the rate was taken from the area crystallized, measured with ImageJ as in earlier work by the group (Rumondor 2009, ref 14) E1·P3. Each rate is a triplicate; the form was checked by XRPD after full crystallization P3. Growth was measured for the stable form and the new metastable form I between 30 and 100 °C (Figure 4), and for the stable form with 10% molar PAA or with PVP between 60 and 100 °C (Figure 2) A2·A4. The PVP level in Figure 2 and the way the drug–polymer melts were made are not printed X12.

Polymer interactions. To relate any growth inhibition to hydrogen bonding, nilutamide and each polymer were dissolved in ethanol at a 1:1 molar ratio (per monomer unit) and spin coated on ZnSe discs, and their FTIR spectra (128 scans, 400–4000 cm−1) were compared with that of pure amorphous nilutamide P4. Nilutamide has one donor and three acceptors, so at least one carbonyl is, in the authors' words, "theoretically" available for the acid OH of PAA, whereas PVP offers a good acceptor and no donor C15·M2·M3.

Strand 3: characterizing the forms and their conversion

What they did. The form that nucleated at −30 °C was first made in bulk by reheating melt-quenched material, lightly triturated and examined by XRPD; later bulk attempts produced a second metastable form, form II P12·R15. The three forms were compared by XRPD (5–30° 2θ), ATR-FTIR, Raman at 785 nm and 13C ssNMR; form II could not be made in enough quantity for ssNMR, because it was difficult to nucleate and converted quickly P6·P4·P5·P9·R25. A 1H–15N heteronuclear correlation experiment, mentioned only in the discussion, tested whether form I still carries its N–H hydrogen R26. Conversion of each metastable form to the stable form was followed by Raman peak ratios (1351 and 1373 cm−1 for form I, 1768 and 1780 cm−1 for form II) during isothermal holds at 70–90 °C (form I) and 60–80 °C (form II) (Figure 13) P5. The sample form used for these kinetics is not stated X14.

Strand 4: conformers

What they did. With SPARTAN'08 at the B3LYP/6-31G* level, the authors rotated the bond between the rings of the keto form from 0 to 180° in 20° steps, starting with the O15 carbonyl oxygen on the same side as the CF3 group, and computed the energy of each conformer; they also simulated the IR spectra of the keto and enol forms and the 13C NMR spectra of the conformers P10. These are single-molecule, gas-phase calculations P10. The computed energies were set against the conformer found in the published crystal structure of the stable form (Trasi, Fanwick and Taylor 2012, ref 21), in which the phenyl ring is rotated about 54.4° from the imidazolidine ring R19.

Experimental analysis

Main result 1: fast growth, yet a glass. As supplied, nilutamide melted with an onset of around 153 °C and an enthalpy of fusion of about 98 J/g; the melt cooled at 15 °C/min to 0 °C gave a glass with a $T_g$ onset of 32 °C, and this glass did not recrystallize on reheating up to the melting temperature M1·R2. The authors read this as nucleation and/or growth being slow on the DSC time scale and, by the group's classification of crystallization tendency (Baird et al. 2010, ref 16), place nilutamide among slowly crystallizing, good glass formers ("appears to be") C5. Growth, however, is not slow. Seeded stable-form crystals grew at about 0.03 µm/s at 60 °C and about 1.5, 5, 15 and 38–40 µm/s at 70, 80, 90 and 100 °C, and the stable-form data of Figures 2 and 4 agree R3·R8. The authors call this rate "fairly rapid" compared with other organic compounds measured in their laboratory or reported elsewhere, and note that acetaminophen, with similar rates, recrystallizes readily on reheating; the glass-forming ability of nilutamide therefore "appears" to come from a low nucleation tendency C8·C9. The comparison values themselves are not shown in the paper X8.

Between 70 and 60 °C the stable-form growth rate fell by about 50-fold (from about 1.5 to about 0.03 µm/s), and the growth front changed from smooth to rough (Figure 2 insets); Raman microscopy showed the same crystal form at both temperatures R3·R4. The authors attribute the drop to the change in growth morphology ("appears to be due to"), a behaviour reported for other compounds (ref 17) C7.

Main result 2: nucleation peaks far below $T_g$. In the DSC screen (Figure 3), reheating after cooling to end temperatures from about 40 °C down to −20 °C showed no crystallization exotherm; only the run cooled to −30 °C, around 60 °C below $T_g$, crystallized on reheating, with an exotherm starting at about 88 °C and peaking at about 97 and 104 °C, followed by melting at 141 °C, well below the melting point of the starting material R5·R6. The DSC holds confirm the pattern on a larger number of runs (Table A) R16.

Table A. Outcome of around 50 DSC runs per temperature: melt cooled to the hold temperature, held 30 min, reheated at 10 °C/min; form identified by Raman (text values) R16.

At −30 °C the few stable-form outcomes came from samples that had already nucleated and partly crystallized during cooling, at about 100–115 °C, so the authors conclude that form I "nucleates preferentially" at low temperature ("These results suggest") R16·C24. On the open glass surface the counts after 30 min trace a U shape (Table B): many nuclei at −30 °C, few between −20 and 40 °C, and a smaller rise at 50–60 °C, roughly 20–30 °C above $T_g$ R9·C13.

Table B. Number of nuclei per sample after 30 min at the hold temperature and 1 min at 65 °C, mean of 5 (read from Figure 5, approximate; at −30 °C the error bars span about 7 to 21) R9.

Two caveats belong next to Table B. The PAA series does not follow pure nilutamide everywhere: at −20 °C it gives about 9.0 nuclei against about 2.3, and from 30 to 60 °C it is consistently higher, although the authors write that PAA "did not seem to have much effect" R9·C21·X5. And the numbers are counts after a fixed 30 min: melt-quenched samples kept a few hours at room temperature nucleated both the metastable and the stable form, so the low counts between −20 and 40 °C describe a short time scale only R17·X16.

Main result 3: the cold nuclei are a new metastable form. Form I melts at 141 °C with an enthalpy of fusion of about 66 J/g, against around 153 °C and about 98 J/g for the stable form; the higher-melting form has the higher enthalpy, so the authors classify the pair as monotropic, with form I metastable at all temperatures (ref 19) R6·M1·C11. Triturated form I heated in the DSC showed an exotherm and then melting of the stable form, as expected for a monotropic pair R6·C11. From the melting data the authors calculate a free energy difference of about 3.4 kJ/mol at 25 °C R7.Card check: the relation and molar mass behind this value are not printed; a simple estimate from the printed melting data gives about 3.5 kJ/mol, close to the stated value (X13). Form I is slightly yellow, the stable form colourless R6. Form II, found during bulk preparation, has an XRPD pattern distinct from both (Figure 8) R15·C23.

Main result 4: the two forms swap growth speed. Figure 4 compares the growth of the stable form and form I from 30 to 100 °C (Table C) R8.

Table C. Radial growth rates of the stable form and metastable form I (read from the log axis of Figure 4, approximate; points at 45 and 55 °C omitted) R8.

The stable form grows 3–4 times faster from 70 to 100 °C, as is common for polymorph pairs (Kestur and Taylor 2013, ref 20), while form I grows faster at 60 °C and below, about 30-fold at 30 °C R8·C12. The crossover therefore lies between 60 and 70 °C, which is how the discussion puts it; the results section says "below 60 °C" C12·X3. The authors call this pattern unusual, unreported for other polymorphic systems "to the best of our knowledge" C12.

Main result 5: the forms bond differently. Table D collects the spectral fingerprints the authors use R20·R22·R25·R10.

Table D. Spectral fingerprints of the three crystal forms and the glass. Values without "about" are from the text; the others are read from Figures 6 and 11 (approximate) R10·R20·R22·R25.

In the stable crystal (Figure 10, from the published structure) the N–H donates a hydrogen bond to a ring carbonyl of the next molecule, building chains, and the second carbonyl, next to the N–H, is free; the authors assign the 1777 cm−1 band "most likely" to the free carbonyl and the stronger 1718 cm−1 band to the bonded one R19·R20·C26. The simulated gas-phase spectrum also has two carbonyl bands, but there they are the asymmetric and symmetric stretches of an isolated molecule, not bonded and free groups R21·C27. In form I the N–H band is much broader and lower, which "could indicate" stronger or more extensive hydrogen bonding, yet the carbonyls shift slightly upward; the authors therefore propose that the N–H is "potentially" bonded to another electron-rich group, "most likely" the nitro group, whose symmetric stretch falls from 1368 to 1350 cm−1 R20·C28. Nitro groups are weak acceptors in solution, but stronger nitro hydrogen bonds are known in crystals, for example in N-(4-nitrophenyl)-n-hexyl-urethane (refs 22, 23) C29. The unique Raman peak of form I at 1329 cm−1, in the C–NO2 region, is read as further support R22·C31. Form II, with its N–H band slightly higher, "appears" to have weaker hydrogen bonds than the stable form R20·C30. The glass has its N–H and C=O bands above those of the crystal, which the authors read as weaker hydrogen bonds in the amorphous state ("suggested") R10·C16.

Main result 6: the conformer of the stable crystal is not the cheapest one. The DFT scan (Figure 14) gives two energy minima, about 0.3 kJ/mol apart, with the most stable conformer at a ring angle of 30°, and a barrier of about 11–12 kJ/mol near 90° R24. The stable crystal holds the rings at about 54.4°, a "higher energy conformer", which the authors say is "presumably" stabilized by packing in the crystal; on the drawn curve this angle lies roughly 4 kJ/mol above the minimum, a reading of the plot that the paper does not state R19·R24·C34.Card check: the 30° minimum and the 0.3 kJ/mol difference do not fall on the 20° grid of computed points, and the paper does not say how they were located (X9). Simulated NMR spectra give a 25 ppm doublet for the 60° conformer and a singlet at 0, 40, 80, 120, 140 and 160°; the stable form, at 54°, shows the doublet, in "excellent agreement" R27·R25·C53. Form I and the glass show a singlet R25. Single crystals of form I could not be grown, from solution (where the stable form always appeared) or from the melt, so its structure is unknown R29.

Main result 7: form I converts slowly, form II fast. Table E summarizes Figure 13 R23.

Table E. Raman-monitored conversion of the metastable forms to the stable form (read from Figure 13, approximate) R23.

Main result 8: PVP binds nilutamide and slows its growth; PAA does neither. Table F compares the stable-form growth rates and the FTIR evidence R3·R11·R12·R13.

Table F. Polymer effects. Growth rates read from Figure 2 (log axis, approximate; PAA at 10% molar, PVP level not printed); FTIR positions read from Figure 7 or stated in the text R3·R11·R12·R13.

PVP slowed growth at every temperature and most strongly at the lowest one (about 50-fold at 70 °C, about 5-fold at 100 °C), while PAA lowered it only marginally from 70 to 100 °C and not at all at 60 °C, where its point lies above the pure drug R3·C14·X4. In the FTIR spectra, PVP shifted the drug N–H and its own carbonyl to lower wavenumbers, "indicating" nilutamide–PVP hydrogen bonds stronger than those between nilutamide molecules, which is "possible because" PVP is a better acceptor than nilutamide's own groups; PAA left the drug bands almost unchanged, so any drug–PAA hydrogen bond is comparable to the drug–drug ones R12·C18·R11·C17. The authors conclude that interaction strength "correlates well" with the impact on growth, in line with other studies C19. The $T_g$ moved by only 1 and 3 °C, and the effect of PVP on nucleation could not be measured because nuclei could not be grown to a visible size in its presence R13·C20·R14·C22.

Further observations. Amorphous nilutamide exposed to high humidity crystallized at the surface; on heating, these crystals lost their birefringence around 90 °C, after which the stable form grew and melted at 155 °C (Figure 9), and a spin-coated film exposed to high humidity gave a distinct FTIR spectrum (not shown) R18. The authors "suspect" an unstable hydrate that is "likely" to dehydrate with loss of long-range order; it could not be made in bulk because it converted quickly to the anhydrous stable form C25·R18. At 30 °C the metastable form was seen to nucleate on the stable form (cross-nucleation; no data shown) R28.

Discussion

Two nucleation zones and an inverted growth order. The authors call the large separation between the temperature ranges that favour nucleation and growth the most interesting feature of nilutamide C36. As a small molecule with few rotatable bonds it would be expected to crystallize readily, and its growth rate at a given undercooling is comparable to that of flutamide, its structural analogue, and acetaminophen; both of those form glasses easily but crystallize on reheating above $T_g$ even at fast heating rates, and their nucleation zones lie above $T_g$ (refs 7, 8, 24) C36·R30. Nilutamide instead nucleates reluctantly above $T_g$ and much more readily far below it C37. A compound usually has one nucleation zone, but with several polymorphs each may have its own, as the γ and α forms of indomethacin, whose nucleation maxima lie at about $T_g$ and about 20 °C higher (ref 13) ("it is possible") C38·R32. From the need to cool at least 60 °C below $T_g$ the authors place the nucleation zone of the metastable forms there ("suggesting"), and from the second rise in the counts about 30 °C above $T_g$ they infer a second zone; it is "likely" that the two zones belong to the metastable and the stable forms C40·C41. The DSC holds, with far more crystallization at −30 °C than at 60 °C, are presented as support C42. Ibuprofen behaves similarly: a metastable conformational polymorph nucleates about 60 °C below $T_g$ and, not consistently, above it (refs 25, 26), and Oguni proposed that nucleation far below $T_g$ is controlled by beta relaxation rather than viscosity (ref 12) C43·R33.

Is form I just a precursor? If form I nucleated at every temperature and converted quickly to the stable form above $T_g$, as in Ostwald's rule of stages, the "stable-form zone" would be an artefact. The authors judge this "somewhat unlikely": form I needs about 24 h to convert at 70 °C, far longer than the 1 min at 65 °C used to grow the nuclei, so "it appears reasonable to conclude" that the stable form nucleates directly at higher temperatures, "albeit somewhat reluctantly" C44·R23. Note by the AI writer (not discussed in the paper): the argument uses form I only; form II converts far faster (complete in about 30 min at 80 °C and about 3 h at 60 °C), and the nuclei counted on the surface were never identified, so the assignment of the upper zone to the stable form rests on the 60 °C DSC holds alone X6.

Why should cold favour form I? The authors list three ideas. First, cracks forming on cooling could start nucleation at the new surfaces through mechanical stress, as postulated for ibuprofen (ref 27) C45. Second, nucleation of one form or the other could follow the population of conformers or tautomers in the liquid, which changes with temperature C46. For tautomers they consider a keto–enol equilibrium shifting to the enol at low temperature; a strongly hydrogen-bonded O–H in an enol-containing form I would also explain its broad, low N–H band, and solid-state keto–enol tautomerism is known for barbiturates after cryomilling (ref 28) C47. The 1H–15N correlation experiment, however, shows that the nitrogen in form I keeps its hydrogen, which excludes the enol R26·C48. That leaves conformers, and the authors put the pieces together as follows C50·C51.

The supporting observations are the growth-rate crossover between 60 and 70 °C, the cross-nucleation of the metastable form on the stable form at 30 °C and the preference for form I at low temperature, which together "suggest" that "molecular conformation in the liquid may play a role" C50·R8·R28. The authors add that crystallization from solution becomes unfavourable when the crystal holds a high-energy conformer (refs 29–31), and that flutamide, whose crystal contains the low-energy conformer, nucleates readily at 20–30 °C C49·C52. For form I they infer a nearly planar conformer: its yellow colour "suggests" conjugation between the rings, and together with the 25 ppm singlet this places the ring angle below 40° or around 160°, both low-energy regions of the scan C54·C55. The glass also shows the singlet, "suggesting" that its conformation is closer to form I than to the stable form C56. The abstract puts the conclusion as "may explain the slow nucleation from the supercooled melt as well as the cross-over in the growth rates", and the conclusions as "may be due to" C51.

Implications for formulation. The authors' own point is that faster nucleation far below $T_g$, leading to a metastable polymorph, is "unusual" because "low temperature storage is generally considered to promote the stability of amorphous systems" C57. Note by the AI writer (not discussed in the paper): for a drug that behaves like nilutamide, a cold excursion (freezer storage, cold shipping, cryomilling, freeze-drying) could leave metastable nuclei in a glass that looks unchanged, and these would grow quickly once the product is warmed above $T_g$. A DSC cool–reheat screen at moderate temperatures would not reveal this; a screen would need to go far below $T_g$, include holds longer than 30 min and identify the form that appears. On the polymer side, the data show that an acceptor polymer that hydrogen-bonds to the drug N–H (PVP) slows growth much more than a donor polymer (PAA), but whether PVP also suppresses the cold nucleation was not measured C19·C22.

Limitations

Limitations stated by the authors. The effect of PVP on nucleation "could not be easily studied", because PVP slowed growth so much that nuclei could not be grown to a visible size C22·R14. Form I could not be grown as single crystals, from solution or from the melt, so its conformation is inferred indirectly; form II could not be made in enough quantity for ssNMR R29·R25. The hydrate is "suspected" and could not be prepared in bulk C25. Surface counting stopped at 60 °C because the film began to flow R9.

The AI writer's own judgement (1): what is measured and what is inferred. The phenomenon itself is well supported: crystallization only after cooling to −30 °C in the DSC screen, the highest nuclei count at −30 °C and 88% form I in about 50 holds at −30 °C point the same way, and form I is a distinct polymorph by DSC, XRPD, FTIR, Raman and ssNMR R5·R9·R16·R6·R15·R20·R22·R25. The growth crossover rests on triplicate seeded measurements over more than five decades R8. The weaker parts are the inferences built on top. The surface counts are numbers per sample after 30 min, not rates, with n = 5, large error bars at −30 °C and no stated area, and the nuclei were not identified X7·X6. The conformer mechanism has no measured conformer population in the liquid at any temperature; it rests on a single-molecule gas-phase scan, a colour, and one NMR signal of a form whose structure is unknown X15·X10. Mechanical stress and beta relaxation, which the authors themselves raise, were not examined for nilutamide X15.

The AI writer's own judgement (2): places where the text and the data differ. Several statements do not match the plotted data exactly. PAA is said to reduce growth and to leave nucleation nearly unchanged, but its growth point at 60 °C lies above the pure drug and its nuclei counts are higher at −20 °C and from 30 to 60 °C X4·X5. The results section puts the growth crossover "below 60 °C", while the plotted points and the discussion put it between 60 and 70 °C X3. The ssNMR methods name form II, but the spectra are of form I X1. The Raman bands used for kinetics are 1351 and 1373 cm−1 in the methods and 1350 and 1374 cm−1 in the results, a difference that does not matter X2. The comparison of nilutamide's growth rate with other compounds, which carries the argument that nucleation, not growth, controls glass formation, is stated without data X8. The PVP level in the growth experiments and the composition behind the $T_g$ changes are not given, so the polymer effects can be compared only qualitatively X12.

Conclusion

The paper shows that nilutamide, a fast-growing but reluctantly nucleating compound, forms a glass because nuclei are scarce, and that its glass nucleates best not near $T_g$ but around −30 °C, about 60 °C below it, where a previously unknown metastable polymorph, form I, appears C9·C10·R16. A second, weaker nucleation zone above $T_g$ is "likely" that of the stable form, and the growth rates of the two forms cross between 60 and 70 °C C41·R8. Spectroscopy shows that form I bonds its N–H differently, "most likely" to the nitro group, and the authors propose, with "may", that the scarcity of the stable crystal's higher-energy conformer in the cold liquid explains both the slow nucleation of the stable form and the crossover C28·C51. For formulation the lesson the authors draw is that cold storage, usually trusted to protect a glass, can do the opposite for some drugs C57.