Why acetaminophen forms a glass, and why a polymer that slows its crystal growth need not slow its nucleation

Background and motivation

Many poorly water-soluble drugs are given in the amorphous form because its higher energy gives a faster dissolution rate and a higher transient solubility than the crystal; the price is a constant driving force back to the crystal, and crystallization during processing or storage can make a product fail (Hancock and Zografi 1997, ref 3) C1. The common remedy is to mix the drug with a compatible polymer into an amorphous solid dispersion. Polymers are "thought to" help in two ways: they raise the glass transition temperature (Tg) of the blend, which lowers drug mobility, and they form specific drug–polymer interactions that disrupt the self-assembly of drug molecules (refs 4–7) C2. Hydrogen bonding has received most attention: polymers with strong hydrogen-bond donors tend to stabilise drugs with good acceptors and vice versa (Van Eerdenbrugh and Taylor 2010, ref 8; Telang et al. 2009, ref 9), and the strength of the drug–polymer hydrogen bond affects how well a polymer stops crystal growth from the supercooled liquid (Kestur and Taylor 2010, ref 10) C3.

Crystallization needs two steps, nucleation and crystal growth, and the two depend on temperature in different ways (refs 15, 16) C5. Nucleation is generally fastest at a lower temperature than growth (refs 16, 17). If the two temperature zones overlap, a melt crystallizes as it cools; if nucleation happens well below the zone of fast growth, the melt can be quenched to a glass. Such a glass has been suggested to contain "quenched-in" nuclei that cannot grow at low temperature but grow when the glass is reheated (Baird et al. 2010, ref 11; Paladi and Oguni 2003, ref 18) C5·C6. Baird et al. used this picture to sort organic molecules into three classes, at a cooling rate of 20 °C/min and a heating rate of 10 °C/min: class I compounds crystallize on cooling, class II compounds form a glass on cooling but crystallize on reheating, and class III compounds crystallize in neither step (ref 11; Figure 1) C7. The class of a given compound can change with the rates, but the picture is useful to reason about where nucleation and growth happen C7.

Acetaminophen (APAP) is a clear class II case: it forms a glass even at slow cooling but "always recrystallizes on heating, even at extremely high heating rates" (Baird PhD thesis 2011, ref 22) R1. A few studies had measured how polymers change crystal growth rates from amorphous drugs (Kestur et al. 2010, ref 19; Kestur et al. 2011, ref 20; Cai et al. 2011, ref 21), but the effect of polymers on nucleation in amorphous organic systems was, in the authors' words, "not at all well understood" C8. That matters for polymer selection: a polymer that slows growth but not nucleation, or the reverse, protects only half of the process.

The paper therefore has two goals C9. The first is to test the hypothesis that class II compounds such as APAP form glasses because their nucleation and growth zones are separated in temperature. The second is to compare polymers of different chemistry, at the same loading, for their effect on crystal growth and, separately, on nucleation, and to relate the effects to drug–polymer hydrogen bonding.

Background concepts

Nucleation and growth. Nucleation is the formation of new crystal nuclei in the supercooled liquid or glass; crystal growth is the enlargement of existing crystals. Classical crystallization theory gives each process a maximum rate between Tg and the melting temperature Tm. Nucleation needs a large thermodynamic driving force and is therefore favoured at large supercooling; growth needs molecular mobility and is favoured closer to the melting point. Hence the growth maximum lies closer to Tm and the nucleation maximum further from it (Okui 1990, ref 16; Figure 1) C35.

The nucleation rate equation. In the Discussion the authors quote the classical expression for the nucleation rate (Mullin 2001, ref 34), printed without a number but called "equation 3" in the text E1:

Here $J$ is the nucleation rate (number of crystals per ml per min), $A$ a frequency factor, $k$ the Boltzmann constant, $T$ the absolute temperature, $T_m$ the melting temperature, $\Delta H_f$ the heat of fusion, $T_r=T/T_m$ the reduced temperature and $\Delta T_r=1-T_r$ the reduced supercooling; $\Delta G^{\prime }$ is the activation energy for the molecular motion needed for nucleation, and the symbol list gives $\gamma$ as the interfacial energy between crystal and melt E1. In plain terms: the first term is the thermodynamic barrier to forming a nucleus, which shrinks as supercooling grows, and the second is the kinetic barrier of molecular motion, which grows as the liquid cools; their opposite temperature dependences produce a maximum in $J$ C48. Note by the AI writer (not discussed in the paper): the printed expression has $r$ where the symbol list defines $\gamma$, and a plus sign with an uppercase $K$ in the kinetic term; for the two terms to oppose each other as the text says, the kinetic term must enter with a minus sign, so these look like typesetting errors X1. The paper uses the equation only to say that the polymer effects on nucleation are "not particularly straightforward" to explain with it, partly because it contains no term for molecular recognition (Weissbuch et al. 2003, ref 35) C49.

The nucleation zone, as measured here. The authors use the term for the range of temperatures through which a melt must be cooled for enough nuclei to form that it recrystallizes on reheating. Its limits come from a DSC test (described below) and depend on the cooling and heating rates, so they are comparative, not absolute C10·C13.

Spherulites and radial growth rate. APAP crystallizes from its melt between cover slips as spherulites, radial polycrystalline aggregates, of polymorph form III; at a fixed temperature the spherulite radius grows linearly with time, and the slope is the radial growth rate (DiMartino et al. 1997, ref 23; Burley et al. 2007, ref 24) R2·R3·R4.

Polymorphs of APAP. APAP has three polymorphs R2. Crystalline APAP as received melts at 169 °C (onset, Table 1), form II melts at 156 °C, and the text gives 134 °C as the melting point of form III R9.a·R10·M1. Form III converts to form II on heating, an enantiotropic transition C20.

Hydrogen bonding seen by FTIR. When a group such as N–H, O–H or C=O takes part in a hydrogen bond, its stretching band moves. A shift to lower wavenumber usually means a stronger bond than in the pure material, a shift to higher wavenumber a weaker one (Andrews et al. 2010, ref 26; Matsumoto and Zografi 1999, ref 27) C24. APAP has two donors, the phenolic OH (the stronger) and the amide NH, and two acceptors, the amide C=O (much the better) and the phenolic oxygen C32. Among the polymers, PVP and PVPVA carry a very strong acceptor (the pyrrolidone C=O), PAA a very strong donor (the carboxylic acid OH), and HPMC, HPMCAS and Eudragit E100 weaker donor and acceptor groups (Van Eerdenbrugh and Taylor 2011, ref 31) C32.

Methods in detail

The study has one drug and seven polymers. The polymers were chosen for their hydrogen-bonding chemistry: PAA (poly(acrylic acid), two grades, Mw 1800 and Mv 450,000), PVP K12 (poly(vinylpyrrolidone), Mw 2,000–3,000), PVPVA64 (PVP-vinyl acetate, Kollidon VA64, Mw 45 000–47 000), PVPh (poly(vinylphenol), Mw 1,500–7000), HPMC (hypromellose, substitution type 2910), HPMCAS (HPMC acetate succinate, type AS-MF, Mw 18,000) and Eudragit E100 M2·M3·M4·M5·M6·M7·M8·M9. Most experiments used 10% w/w polymer; growth rates were also measured at a 10% molar ratio, and the nucleation zone also at 15% w/w A1·A2·A3. The work splits into four kinds of measurement, each targeting one part of the crystallization process.

Sample preparation

APAP and polymer (500 mg in total) were premixed with mortar and pestle and then cryomilled for 10 min in a 6750 freezer mill to mix them thoroughly P1. For the growth, DSC and nucleation experiments the mixtures were melted and quenched, so these start from a glass made from the melt; only the FTIR films were made from solution (below) P2·P4·P6·P7·P8. Besides the 10% and 15% w/w mixtures, APAP was mixed with PVP, PAA and PVPh at a 9 : 1 molar ratio with respect to the polymer repeat unit P1. The reason, as the authors explain, is that at equal weight a polymer with a lighter repeat unit supplies more functional groups; comparing on a molar basis isolates the effect of the number of groups C12. This was possible only for polymers with simple repeat units, which excluded the cellulosics and the copolymers C12.

Crystal growth rates by hot-stage microscopy

Three to five milligrams of powder were melted at 185 °C between two cover slips, quenched to room temperature and then held on a Linkam hot stage at one temperature under a polarizing microscope; time-lapse photographs followed a single growing spherulite P2. Its radius rose linearly with time (least-squares R² above 0.99), and the slope was taken as the radial growth rate; each sample was measured in triplicate P2·R4. The Methods give the temperature range as 70–110 °C; the figures also contain points at 50, 60, 120 and 130 °C, for which replication is not stated P2·X2. Raman spectroscopy of APAP crystallized between quartz cover slips confirmed the expected form III R3. Figure 3 shows what the measurement looks like at 80 °C: a pure APAP spherulite grows visibly in 15 s, whereas one in a 10% PAA sample is still smaller after 180 s R5.

Thermal analysis

DSC (TA Q2000, nitrogen purge 50 ml/min) gave the onset Tg of each glass on reheating at 10 °C/min after the melt had been cooled at 15 °C/min to −20 °C, and the onset melting temperature and melting enthalpy of physical mixtures of crystalline APAP with 10% w/w polymer at 10 °C/min P4·P5. The same reheating runs (Figure 6) show at which temperature the glass recrystallizes and which polymorphs appear R10.

The nucleation zone by DSC

Nucleation is hard to observe directly, so the authors inferred it from thermal history C10. About 5 mg in a pinholed hermetic pan was melted at 170 °C, cooled at 15 °C/min to a chosen end temperature, held for 1 min and reheated at 10 °C/min; an exotherm on reheating means the sample recrystallized P6. The logic, shown schematically in Figure 7(a), runs as follows C10. If the melt is not cooled far enough, too few nuclei form during the cycle and nothing crystallizes on reheating; the highest end temperature that still gives an exotherm is the upper limit of the nucleation zone. Cooling further produces more nuclei, and more nuclei make the sample recrystallize at a lower temperature (Schroers et al. 1999, ref 25); once the peak temperature stops falling, further cooling adds no nuclei, and that end temperature is the lower limit. Figure 7(b) shows a real run for APAP with 10% PVPVA: after cooling to 70 °C no exotherm appeared, after 60 °C the exotherm peaked at 115 °C, after 30 °C at 93 °C, and below 30 °C the peak no longer moved, so the zone is 30–60 °C R11·R12.h. The authors point out two limits of the method: the zone depends on the cooling and heating rates, so only comparisons between systems are meaningful, and the shape of the exotherm also reflects how fast the crystals grow, so the method gives no relative nucleation rates C13.

Counting nuclei on the hot stage

To compare how many nuclei form, about 3 mg was melted between coverslips, cooled at 15 °C/min to −30 °C so that every sample passed through its whole nucleation zone, and reheated at 10 °C/min to 90–100 °C so that the nuclei grew to visible size; the growth spots in a given volume were then counted, and the average of five experiments was used P7. The number of nuclei "was assumed to be proportional" to the nucleation rates met during this non-isothermal treatment C11. Each sample passes through its zone twice, once on cooling and once on reheating, so the count includes nuclei formed in both steps P7.

Hydrogen bonding by FTIR

Because earlier work had shown that about 20–50% w/w polymer is needed to see band shifts, the FTIR samples contained 30% w/w polymer and 70% w/w APAP P8. They were dissolved in ethanol or 1 : 1 ethanol–dichloromethane, spin-coated onto ZnSe discs at 2,000 rpm, and measured with 128 scans over 400–4000 cm−1; spectra were compared with those of crystalline form II (crystallized from an amorphous film) and amorphous APAP P8. Spectra are shown for PAA (Figure 10) and PVP (Figure 11); the results for PVPVA, PVPh and the other polymers are given in the text only R18·R19·R20.

Experimental analysis

Result 1: acetaminophen nucleates where its crystals barely grow

Pure APAP had a very narrow nucleation zone, 50–60 °C: it recrystallized on reheating only if it had first been cooled to 60 °C or lower R12.a·C36. Its growth rate, in contrast, is fast only well above that range. Read from Figures 4 and 5, the radial growth rate of pure APAP is about 97 µm/s at 130 °C, 58 at 110 °C, 45 at 100 °C, 19 at 90 °C, 5.3 at 80 °C and 0.92 at 70 °C, and falls to about 0.23 µm/s at 60 °C and 0.04 µm/s at 50 °C R6.a·R8.a. Between 110 and 70 °C alone the rate drops by "nearly two orders of magnitude", which the authors attribute to the sharp loss of molecular mobility on cooling R6·C14. At the hot end, growth levels off near 110 °C as the driving force shrinks towards the form III melting point (134 °C); above 110 °C form II appeared and grew from the liquid R7·C14.

The authors combine these facts into an explanation of class II behaviour C37·C38. Near 60 °C, where nucleation starts, growth is "presumably" too slow to give a detectable exotherm, and in the DSC example of Figure 7 crystallization becomes detectable only from about 80 °C on reheating C37·R11. So the nuclei, formed "primarily" at 50–60 °C during cooling and reheating, grow only when the reheated sample passes about 80 °C, which produces the recrystallization exotherm C38. The hypothesis of separated nucleation and growth zones therefore "appears to be reasonable" for APAP C39. In the Conclusions they add that the glass made by cooling the melt is "furthered suggested" to contain quenched-in nuclei, and that such nuclei "may" lead to observable crystallization during long storage C40·C41.

Note by the AI writer (not discussed in the paper): the 80 °C threshold is argued for pure APAP but read from the APAP–PVPVA run of Figure 7; in Figure 6 the exotherm of pure APAP begins at about 73 °C and peaks at about 86 °C X10·R10.a. And because both the DSC test and the nuclei count pass the zone on cooling and again on reheating, nuclei formed on cooling were not separated from those formed on reheating; the quenched-in nuclei are an inference X9.

Result 2: every polymer slows growth, by very different amounts

All seven polymers kept the spherulitic shape and lowered the growth rate at every temperature tested (Figure 4) R4·R6. On a weight basis PAA was the most effective, more than one order of magnitude below APAP at any temperature; PVP and PVPVA also reduced growth substantially; HPMCAS, E100 and HPMC gave "very similar" growth rates and were "clearly less effective" C15. Table A gives the values read from Figure 4.

Table A. Radial growth rate of spherulites (µm/s) with 10% w/w polymer, read from Figure 4 (approximate; log axis; mean of triplicates). *The PVPVA symbol overlaps PVP or PVPh at these temperatures. R6

Three points stand out. The ranking is the same at every temperature: PAA lowest, then a middle group of PVP, PVPVA and PVPh, then HPMC, E100 and HPMCAS closest to APAP R6. The effect is large: at 80 °C, PAA 450 K grows about 26 times slower than pure APAP, HPMCAS only about 3 times slower R6.a·R6.b·R6.i. And molecular weight matters for PAA on a weight basis: the authors call the 450 K grade "slightly better" than the 1800 grade C15. Note by the AI writer: in Figure 4 the 450 K grade grows about half as fast as the 1800 grade between 80 and 110 °C, which is more than "slightly" X17.

Result 3: per functional group, the inhibitors are equal

When PAA, PVP and PVPh were compared at a 10% molar ratio, the picture changed (Figure 5) R8. All three polymers, and both PAA grades, gave growth rates within a narrow band, about 5 to 17 times below pure APAP between 50 and 100 °C; the authors describe this as "a similar level of growth inhibition", and the two PAA grades "did not differ" R8·C16.

Table B. Radial growth rate (µm/s) at a 10% molar ratio of polymer repeat unit, read from Figure 5 (approximate; log axis). R8

Read together with Result 2, this is the authors' main argument for a functional-group effect: on a weight basis PAA, with the lightest repeat unit and so the most hydrogen-bonding groups per gram, is the best inhibitor, but per group it is no better than PVP C33. Note by the AI writer: the paper does not give the weight fractions of the molar series; with standard formula weights they are about 5 wt% PAA and 8 wt% PVP or PVPh, less than the 10 wt% of Figure 4 X14.

Result 4: Tg and melting do not explain the ranking

Table C. Transcribed from Table 1 (DSC at 10 °C/min; Tg of the glasses, Tm and ΔHf of physical mixtures; standard deviation of Tg and Tm below 1 °C; the meaning of the values in parentheses is not stated). R9

The polymers raised Tg by only 1.7 to 4.2 °C R9. The authors take the single, slightly higher Tg as evidence that each polymer is miscible with APAP ("confirms" in the Results, "indicating" in the Discussion) C17. The rise is about the same for strong and weak growth inhibitors: HPMC (+3.4 °C) and E100 (+3.0 °C) slowed growth little, PVPh (+2.8 °C) a lot R9·R6. So the Tg rise is "likely to contribute" to slower growth but "cannot explain" the differences between polymers C18. The text summarises the other polymers as raising Tg by 2–3 °C; Table 1 gives 1.7 °C for HPMCAS and 3.4 °C for HPMC X4. The melting onset barely moved (166.7–168.6 °C against 169 °C), so the degree of supercooling at a given temperature was about the same in every system R9·C19. The melting enthalpy fell from 180 J/g to 124–155 J/g, most with PVP and PVPVA; the authors attribute the drop beyond dilution to heat released by favourable drug–polymer interactions in the melt ("most likely") R9·C43. Note by the AI writer: dilution by 10% polymer alone would give 162 J/g, so every mixture lies below the dilution value X16.

Result 5: the best growth inhibitors are the polymers that hydrogen-bond to APAP

The FTIR spectra first set a reference (Figure 9). In crystalline form II the NH stretch is sharp at 3331 cm−1 and the OH stretch broad at 3162 cm−1, so the OH group forms the stronger hydrogen bond, as the single-crystal structure confirms (NH to phenolic O 2.939 Å, OH to carbonyl O 2.712 Å) R14·R15·C25. In amorphous APAP the bands broaden and merge into one NH/OH band near 3301 cm−1, and the carbonyl band moves to 1655 cm−1 with a shoulder at 1636 cm−1; the authors read the shoulder as "most likely" stronger phenol–carbonyl bonds and the main peak as weaker NH–CO bonds R14·C26.

Table D. FTIR band positions from the text and Figures 9–11. "Other polymers" are not named; by elimination they are HPMC, HPMCAS and E100. R14·R16·R17·R18·R19·R20

With PAA the combined NH/OH band moved up to 3320 cm−1, which the authors read as partial disruption of APAP–APAP hydrogen bonds; the PAA acid carbonyl moved from 1705 to 1714 cm−1, "suggesting" that the acid dimers of PAA are broken; and the growing 1636 cm−1 shoulder means, "it appears", that the acid OH of PAA bonds the amide carbonyl of APAP R16·C27. With PVP the NH/OH band moved down to 3283 cm−1 and the pyrrolidone carbonyl from 1679 to 1657 cm−1, a clear sign that the APAP donors bond the single strong acceptor of PVP R17·C28. PVPVA gave a smaller NH/OH shift, as expected for a polymer with fewer pyrrolidone groups, but the same carbonyl shift; since the phenol is the better donor, it is "likely" the group bonded to the pyrrolidone carbonyl in both polymers R18·C28. PVPh moved the NH/OH band slightly up with no carbonyl change, which the authors expect because its phenol resembles that of APAP R19·C29. The remaining polymers showed no noticeable shifts, "indicating" that their interactions with APAP are no more favourable than drug–drug or polymer–polymer contacts R20·C30.

Putting Results 2, 3 and 5 together, the authors conclude that more favourable drug–polymer hydrogen bonding "was thus found to correlate well" with growth inhibition: the strongest inhibitors (PAA, PVP, PVPVA) are the polymers with the clearest spectral changes, and the site differs, the vinylpyrrolidone carbonyl bonding the APAP phenol and the PAA acid bonding the APAP amide carbonyl C31. They go on to suggest that drug–polymer hydrogen bonding "should be considered as major contributor" to the stabilisation of amorphous dispersions C34. Note by the AI writer: PVPh does not fit neatly; it slowed growth about as much as PVP at 70–90 °C and as much as PAA and PVP on a molar basis, yet its spectrum changed little, and the paper does not discuss it X6. The FTIR films also contained three times more polymer than the kinetic samples and were made from solution X15.

Result 6: nucleation follows a different order

Table E. Transcribed from Table 2: DSC nucleation zones and the average number of nuclei after one cooling–reheating cycle on the hot stage (5 runs; values in parentheses not defined). "No crystallization" stands for a dash in the table, explained in the text. R12

The nuclei counts reorder the polymers R12. HPMC (9.3), PVPh (9.2), PVPVA (12.3) and E100 (13.1) did not differ significantly from pure APAP (10.8), in the authors' judgement, although no statistical test is reported R12·C23·X13. PVP K12 reduced the count "to some extent", to 5.2 R12.f·C23. HPMCAS was the best nucleation inhibitor, with 2.5 nuclei, and PAA "was observed to promote nucleation", with 26 R12.c·R12.e·C23. The text describes these as reductions "to around a third" and "nearly three times more" nuclei; the table values correspond to about a quarter and about 2.4 times X5. So the polymer that was least effective against growth, HPMCAS, was the most effective against nucleation, and the polymer most effective against growth, PAA, increased nucleation C46.

Note by the AI writer (not discussed in the paper): the count adds up all nucleation while the sample is inside its nucleation zone, and the zones differ in width, 10 °C for APAP against 40 °C for PAA. At the stated rates a PAA sample spends about four times as long in its zone as pure APAP, so per second in the zone PAA would not exceed APAP. The reading that PAA promotes nucleation depends on treating the count per cycle as a rate X8·C11. The PAA grade used for the nucleation work is also not stated; Figure 8 labels the curve "PAA 450K" X11.

Result 7: polymers move and widen the nucleation zone

The DSC zones in Table E show that most polymers broadened the zone and moved it down R12·C22. PAA had the largest effect, lowering the upper limit by 30 °C and the lower limit by 60 °C, to −10–30 °C; HPMC (10–60 °C) and HPMCAS (10–45 °C) extended the zone far below that of APAP; E100 (40–60 °C) and PVP (30–50 °C) had, in the authors' words, no substantial effect on the nucleation temperatures, and PVPh "appeared to increase" the upper limit to 70 °C C22. At 15% w/w, PVP, PAA and HPMCAS samples did not crystallize at all in this test, while HPMC, E100, PVPh and PVPVA still did R12. Figure 8 sketches zones and relative rates together, as peaks against temperature; the authors call it "semi-quantitative", and its heights are not proportional to the counts in Table 2 R13·X12. Note by the AI writer: the PVP zone lies 10–20 °C below that of APAP, a shift similar to the one the Discussion counts for HPMCAS X18.

The zone matters for storage, the authors argue: storage inside a zone leads to high nucleation, so although HPMCAS gives fewer nuclei than PVP in this test, at room temperature, which lies inside the HPMCAS zone but outside the PVP zone, an HPMCAS sample "could potentially" accumulate more nuclei over time C47.

Result 8: polymers delay the polymorph change on reheating

The reheating thermograms of Figure 6 show a second effect of the polymers R10. Pure APAP recrystallized in an exotherm peaking at about 86 °C, first as form III, which converted at once to form II (a small hump after the exotherm), and form II melted at 156 °C R10.a·C20. With polymers the recrystallization peak moved to about 94–110 °C, and with PAA it became weak and broad, which the caption attributes to slower growth R10·C21. The form III to form II transition was delayed: with PVPVA a separate exotherm appears near 124 °C, with PVP no clear event, and with HPMC, HPMCAS, E100 and PAA an endotherm near 135–140 °C, read as form III melting, is followed by the crystallization of form II R10·C20. The PVPh sample is not shown in Figure 6 R10.

Discussion

How the authors connect the findings

The argument has two halves. The first explains why APAP is a class II compound; the second explains why polymers act differently on the two steps of crystallization. The flow below follows the authors' reasoning; the green and red cases are the outcomes for a formulation (fewer or more crystals) C35·C36·C37·C38·C31·C46·C50·C51.

For growth the authors have a consistent story. In supercooled APAP, the polymers that inhibited growth most were those that formed the strongest hydrogen bonds with the drug, and PAA and PVP inhibited equally when the number of hydrogen-bonding groups was the same C31·C33. The rise in Tg, small and similar for all polymers, "cannot explain" the differences C18. This agrees with earlier studies that linked drug–polymer hydrogen bonding with lower growth rates (Kestur and Taylor 2010, ref 10; Van Eerdenbrugh and Taylor 2011, ref 31) C44.

For nucleation the authors are explicit that they "can only speculate" C51. Earlier work on polymers and nucleation was thin: PVP, HPMC and HPMCAS had very similar effects on felodipine nucleation, but at one temperature only, which the authors think may not hold at other temperatures (Konno and Taylor 2006, ref 32); PVP, HPMC and HPMCP reduced nitrendipine nucleation (Miyazaki et al. 2011, ref 33) R21·R22·C45. For PAA they propose molecular recognition: it is "therefore possible" that PAA "may" exert a local ordering effect on APAP and so enhance nucleation C50. The analogies they cite are that poly(methyl methacrylate) and poly(n-butyl methacrylate) direct APAP to different polymorphs when it crystallizes from solution (Lopez-Mejias et al. 2011, ref 36), and that PAA acts as a template for the nucleation of calcium carbonate polymorphs (Kotachi et al. 2006, ref 37) R23·R24. For HPMCAS they borrow a finding from solution crystallization: additives with some favourable interaction with the crystallizing compound but a larger molecular size tend to lengthen nucleation induction times (Anwar et al. 2009, ref 38), and HPMCAS has the largest monomer unit of the polymers used R25·C51. Note by the AI writer: FTIR found no noticeable interaction between APAP and the remaining polymers, which include HPMCAS, so the "favourable interaction" part of this analogy has no support in the paper's own data X7.

The authors' general conclusion is that "the effect of polymers on nucleation behavior is not predictable", because it shows no direct correlation with any easily identifiable property of the system, in contrast with growth C52. A polymer that inhibits growth "does not necessarily" reduce nucleation; PAA shows that a polymer can inhibit growth and promote nucleation, and HPMCAS the reverse C53. Hence, in the Conclusions, "it is apparent that polymers can stabilize amorphous drugs formulations by selectively influencing either nucleation or crystal growth or both" C53.

What it suggests for formulation

The authors draw three practical points. First, DSC screening of drug–polymer dispersions is harder to read than it seems: when a dispersion shows no recrystallization exotherm, that "could indicate" reduced nucleation, reduced growth or both C54. Second, the temperature dependence of nucleation matters as much as the rate: storage inside the nucleation zone gives high nucleation, so a polymer such as HPMCAS, whose zone includes room temperature, could build up more nuclei on storage than PVP, whose zone does not C47. Third, investigating whether a polymer acts on nucleation or on growth is important "for choosing the right polymer" for robust amorphous systems C55.

Note by the AI writer (not discussed in the paper): in this data set the best growth inhibitor was also the polymer with the most nuclei per cycle, so a polymer selected only from growth rates or hydrogen-bonding strength could leave more nuclei in a product. The results also invite combining one polymer that inhibits growth with one that inhibits nucleation; the paper does not test this. And at 10% w/w the Tg rise was too small to matter, so here the chemistry of the polymer, not its effect on Tg, decided growth inhibition.

Limitations

Limits the authors state. The nucleation zone "depends on the experimental parameters used and can change depending on the rates of cooling and heating"; its purpose is to compare systems, not to give absolute values, and the DSC procedure gives no relative nucleation rates because the exotherm also reflects growth C13. The nuclei counts rest on the assumption that the number of nuclei is proportional to the nucleation rates met during the cycle C11. The classical nucleation equation does not explain the polymer effects and does not encompass molecular recognition C49, and the mechanisms proposed for PAA and HPMCAS are, in the authors' word, speculation C50·C51.

The AI writer's own judgement.

• The growth data are the strongest part: direct, triplicate isothermal measurements with linear radius–time plots and large, consistent differences between polymers R4·R6. The molar series (Figure 5) is a useful independent test of the functional-group argument R8.
• The nucleation data are semi-quantitative. The counts per cycle depend on the width of each zone and on the time spent in it, so they are not rates per unit time; the scatter is large (for example 26 (7.6) for PAA against 10.8 (3.1) for APAP, n = 5) and no statistics are given X8·X13.
• Quenched-in nuclei are inferred: the protocols pass the zone twice and do not separate nuclei formed on cooling from those formed on reheating X9.
• The hydrogen-bond correlation has one unexplained case, PVPh, and rests on FTIR of 30% w/w films made from solution, not the 10% w/w melt-quenched samples X6·X15.
• Several details disagree between text and figures or tables: the growth temperature range (70–110 °C in the Methods, 50–130 °C in the figures), the growth of APAP above 110 °C ("remained the same" against a continued rise in Figure 4), the nuclei ratios, the Tg rises, the typesetting of the nucleation equation, and the PAA grade used for the nucleation work X2·X3·X5·X4·X1·X11.
• Only one drug was studied, at one cooling and one heating rate, without isothermal storage or humidity; whether the separation of nucleation and growth inhibition holds for other drugs is open.

Conclusion

The paper makes two contributions. First, for acetaminophen it supports the view that a class II compound forms a glass because its nucleation zone (50–60 °C by DSC) lies at temperatures where its crystals grow slowly, while fast growth occurs only from about 80 °C upwards; the authors find this hypothesis "reasonable" and suggest, without showing it directly, that the glass holds quenched-in nuclei C39·C40·R12.a·R8.a. Second, it shows for seven polymers that inhibiting crystal growth and inhibiting nucleation are different properties C53. Growth inhibition followed drug–polymer hydrogen bonding, with PAA and PVP the strongest inhibitors and PAA and PVP equal per hydrogen-bonding group C31·C33. Nucleation did not: HPMCAS, a weak growth inhibitor, gave the fewest nuclei, PAA, the strongest growth inhibitor, gave the most and moved the nucleation zone down to −10–30 °C C46·C22·R12.

For formulation, the message is that a polymer cannot be judged from one measurement. A missing DSC exotherm, a low growth rate or strong hydrogen bonding each describe only part of how the polymer protects the amorphous drug, and the nucleation zone relative to the storage temperature is a separate question C54·C47·C55.