Why three different polymers slow felodipine nucleation equally: Tg and hydrogen bonding are not the deciding factors

Background and motivation

An amorphous drug dissolves faster and can reach higher concentrations than its crystal, which is why poorly water-soluble drugs are often formulated as amorphous solid dispersions. The catch is physical stability: the amorphous form is thermodynamically unstable and may recrystallize during storage, which erases the dissolution advantage. Mixing the drug at the molecular level with a polymer carrier, most often PVP, PEG or a cellulose derivative such as HPMC, is the standard way to keep it amorphous (Sekiguchi and Obi 1961; Chiou and Riegelman 1971; Serajuddin 1999). C1

By 2006 two families of explanations competed for why such polymers work. The first is mobility through the glass transition temperature, Tg. Tg separates a regime of high molecular mobility from one of low mobility, and a polymer whose Tg is much higher than the drug's raises the Tg of the mixture, an effect called antiplasticization; the stabilizing effect of high-Tg polymers had been attributed to it (Van den Mooter et al. 2001; Oksanen and Zografi 1990). C2 The second is specific interaction: ion–dipole interactions and hydrogen bonds between drug and polymer had been seen in stable dispersions, together with disruption of the hydrogen-bond pattern that the drug uses in its crystal (Taylor and Zografi 1997; Tang et al. 2002; Miyazaki et al. 2004). C4

Neither explanation was complete. Crystallization had been reported well below Tg (Yoshioka et al. 1994). Matsumoto and Zografi (1999) found that PVP inhibited indomethacin crystallization at levels too low to change Tg, and Khougaz and Clas (2000) saw the crystallization tendency of a drug fall in PVP dispersions even when the dispersion Tg was below that of the drug. C3·R21 On the interaction side, Miyazaki et al. (2004) found polyacrylic acid a better inhibitor of acetaminophen crystallization than PVP at equal Tg and credited stronger drug–polymer interactions, while stabilization had also been seen in systems where hydrogen bonding is chemically impossible (Van den Mooter et al. 2001). R21·C4 The authors' summary is that the polymer properties needed to inhibit crystallization from the amorphous state "are not fully understood". C5

The study was designed to separate the two explanations. The model drug, felodipine, is a dihydropyridine with a single NH hydrogen-bond donor. The three polymers have different glass transition temperatures (PVP 167 °C, HPMC 140 °C, HPMCAS 114 °C) and different hydrogen-bond acceptor groups (Table 1: a pyrrolidone carbonyl in PVP; methoxyl and hydroxypropyl groups in HPMC; methoxyl, acetyl, hydroxypropyl and succinoyl groups in HPMCAS). C6·M2·M3·M4 The authors measured the nucleation rate of felodipine in thin, optically transparent films containing each polymer, in the absence of moisture, and characterized the same compositions by differential scanning calorimetry (DSC) and infrared (IR) spectroscopy. A1·P2·P4·P7 If antiplasticization were the main mechanism, PVP should inhibit best; if drug–polymer hydrogen bonding were, the polymer with the strongest bond should inhibit best.

Background concepts

Nucleation rate and what controls it

Crystallization from an amorphous solid has two steps: nuclei of the crystal must form (nucleation), and they must then grow. This paper measures only the first step. The nucleation rate is the number of new nuclei that appear per unit volume and time. In the discussion the authors write it with the classical expression of Turnbull and Fisher (Eq. 8): E8

In plain terms: the rate $I$ falls exponentially with the sum of two barriers. $\Delta G^{*}$ is the thermodynamic barrier to forming a nucleus: it balances the free-energy cost of creating a new interface ($\Delta G_s$, unfavourable) against the bulk free-energy gain of turning amorphous material into crystal ($\Delta G_v$, favourable, the driving force). $\Delta G_a$ is the kinetic barrier, the activation energy for a molecule to cross from the amorphous phase onto the nucleus, which is related to diffusivity or to the reciprocal of viscosity. $k$ is Boltzmann's constant, $T$ the absolute temperature and $A$ a constant. E8 The authors note that this theory describes homogeneous nucleation, whereas nucleation in their films is "most likely heterogeneous"; they use it to sort possible polymer effects into a driving-force part and a kinetic part, not to fit data. C33

Driving force in a mixture: the Flory–Huggins activity

Mixing a drug with a polymer lowers the drug's free energy (through the entropy of mixing and any favourable heat of mixing) while leaving the crystal's free energy unchanged, so $\Delta G_v$, and with it the driving force for nucleation, becomes smaller. C34 For a polymer–small-molecule mixture the drug's activity is given by the Flory–Huggins equation (Eq. 9): E9

In plain terms: $a$ is the activity of the drug (component 1), ${\Phi }_1$ and ${\Phi }_2$ are the volume fractions of drug and polymer, $x_{12}$ is the ratio of the molecular volumes, and $\chi$ is the Flory–Huggins interaction parameter, which measures how far the pair departs from ideal mixing. The first two terms are the entropy of mixing. The lower the activity, the smaller the driving force to crystallize. The paper prints no $\chi$ for felodipine with any of its polymers. E9

Glass transition, antiplasticization and ideal mixing

A miscible drug–polymer mixture shows one Tg between those of its components. Two equations predict where that Tg lies if the components mix ideally. The Gordon–Taylor equation (Eq. 2) is E2

In plain terms: the mixture Tg is a weighted average of the component Tgs $T_{g1}$ and $T_{g2}$, weighted by their weight fractions $w_1$ and $w_2$, with the polymer's weight scaled by a constant $K$. The equation assumes the components are miscible and their free volumes add. E2 The authors estimate $K$ from the densities $\rho$ and Tgs of the two components (Eq. 3): E3

In plain terms: $K$ grows when the drug's density and Tg are high relative to the polymer's. The densities used were 1.33 g/cm³ for amorphous felodipine and 1.25 g/cm³ for PVP (both measured by helium pycnometry), and 1.29 and 1.19 g/cm³ for HPMCAS and HPMC (from the manufacturer). E3·M1·M2·M3·M4 The Couchman–Karasz equation has the same form as Eq. 2, with $K$ taken from the heat-capacity steps $\Delta C_p$ at the Tg of each component (Eq. 4): E4

In plain terms: the component with the larger heat-capacity step at Tg pulls the mixture Tg toward its own value. The measured $\Delta C_p$ values were 0.33 J/g/K for felodipine, 0.20 for PVP, 0.31 for HPMCAS and 0.25 for HPMC. E4·M1·M2·M3·M4 A measured Tg below these ideal-mixing lines is a "negative deviation", which the literature associates with like–like interactions (drug–drug, polymer–polymer) being stronger than drug–polymer ones (Brekner et al. 1988). C32

Note by the AI writer (not discussed in the paper): the broken Gordon–Taylor lines drawn in Figure 5 are reproduced by Eqs. 2–3 only if Tg is entered in °C when computing $K$ (for example $K$ = 0.280 for felodipine–PVP, giving about 54.5 °C at 25 wt% PVP, as plotted). The rule behind Eq. 3 uses absolute temperature, which gives $K$ = 0.767 and a predicted 69.0 °C at 25 wt% PVP. The printed Couchman–Karasz constant (Eq. 4) is the inverse of the usual weighting, and neither orientation reproduces the solid lines in Figure 5. X3·X4·E3·E4

Enthalpy relaxation as a measure of mobility below Tg

A glass stored below its Tg slowly loses excess enthalpy as its molecules rearrange; on reheating in the DSC this lost enthalpy is recovered as an endotherm. The fraction of the glass that has relaxed after an ageing time $t$ is (Eq. 5) E5

In plain terms: $\Delta H_t$ is the enthalpy recovered after ageing for $t$, and $\Delta H_{\infty }$ is the most that could ever be recovered at that ageing temperature. E5 The maximum is estimated from the heat-capacity step (Eq. 6): E6

In plain terms: the further below Tg the glass is aged (ageing temperature $T_a$), and the larger its heat-capacity step, the more enthalpy it can lose. E6 The decay of the unrelaxed fraction is fitted with the Kohlrausch–Williams–Watts (KWW) function, printed as Eq. 7: E7

In plain terms: the unrelaxed fraction decays as a stretched exponential with an average relaxation time $\tau$ and a parameter $\beta$ (between 0 and 1) describing how broad the distribution of relaxation times is. The paper prints the minus sign inside the power; the standard KWW form, and the only one that reproduces the paper's $\tau$ and $\beta$ against Figure 6, is $\exp (-{(t/\tau )}^{\beta })$. E7·X5 A longer $\tau$ means slower molecular motion in the glass.

Reading felodipine's infrared bands

Felodipine's NH group is its only hydrogen-bond donor, and the position of the NH stretching band shifts to lower wavenumber when the hydrogen bond it forms gets stronger (Tang et al. 2002; Jeffrey 1997). R19 In crystalline felodipine the NH bonds weakly to a carbonyl group of a neighbouring molecule (Fossheim 1986); in amorphous felodipine NH-to-carbonyl bonds also form and are on average stronger than in the crystal. R19 The paper's own spectra show this: the crystal has its NH band at 3373 cm⁻¹ and one carbonyl band at 1696 cm⁻¹, while the amorphous drug has its NH band at 3341 cm⁻¹ and a carbonyl doublet at 1701 and 1682 cm⁻¹, assigned earlier to carbonyl that is not hydrogen-bonded ("free") and carbonyl that is hydrogen-bonded to an NH. R10·R19 When a polymer offers its own acceptor groups, two things can change: the NH band moves (stronger or weaker bond than drug–drug), and the free carbonyl band grows as drug carbonyls lose their NH partners.

Methods in detail

The study has one central measurement, the nucleation rate in dry films, and three supporting characterizations of the same drug–polymer pairs. The flow below follows the order of the paper.

Measuring nucleation rates in dry, thin films

The problem. The authors wanted the nucleation rate of the drug in the absence of any moisture, which calls for thin, optically transparent samples that can be kept dry and examined repeatedly under a microscope. P1·P2

What they did. Felodipine and polymer were dissolved together in a 1:1 mixture of dichloromethane and ethanol and dropped onto a glass cover slip spinning at about 2500 rpm; the solvent evaporated as the solution spread, and the film was heated to 90 °C for several minutes to remove residual solvent. All preparation took place in a glove box purged with nitrogen at a relative humidity below 10%. P1 The films were stored in desiccators over phosphorus pentoxide (0% RH) at 22 °C and taken out for about 5 minutes at each time point; the same films were returned and reused. P2 Under a polarized light microscope, birefringent crystallites were counted in 12 areas of each film at each time, with a 10× objective; nucleation that sometimes started preferentially at the film edges was excluded, and each composition was run in triplicate. P2·R18 Polymer levels were 0, 3, 7, 10, 15, 20 and 25% w/w. A2 Freshly prepared films were transparent and showed no birefringence, which the authors take as evidence that they were amorphous. R1·C7

To turn counts per area into counts per volume, the authors used the depth of field of the objective (Eq. 1): E1

In plain terms: the slice of film that is in focus is thicker for longer wavelengths $\lambda$ and smaller numerical apertures NA; the second term adds the contribution from the detector's resolution $e$ at lateral magnification $M$; $n$ is the refractive index of the medium. With $\lambda$ = 550 nm, $n$ = 1.000 (air), $e$ = 14 µm and $M$ = 10×, the authors obtained $D_{\mathrm{t}\mathrm{o}\mathrm{t}}$ = 0.0144 mm. E1·P2 The nucleation rate is then the slope of the number of sites per volume against time (Andronis and Zografi 2000). C12

Note by the AI writer (not discussed in the paper): the numerical aperture is not printed, but NA = 0.25, a common value for a 10× objective, reproduces 0.0144 mm exactly. The methods say the density per volume was obtained by multiplying the density per area by the depth of field; to obtain a density per volume one must divide, so the word "multiplying" is read here as a slip. E1·X6

Glass transition temperature of bulk dispersions

For thermal analysis, dispersions were made in bulk: drug and polymer in the same 1:1 dichloromethane–ethanol solvent, solvent removed by rotary evaporation at 60 °C, then several hours under vacuum. P3 About 5 mg was sealed in an aluminium pan and run at 20 K/min in a modulated DSC calibrated with indium and benzophenone; the onset Tg was taken from the second heating, after heating to 20 K above Tg to erase thermal history, in triplicate. P4 The compositions were 0, 7, 15, 25, 50, 70, 85 and 100 wt% polymer, compared with the Gordon–Taylor and Couchman–Karasz lines described above. A3·P5

Enthalpy relaxation

Dispersions with a 9:1 felodipine:polymer weight ratio were prepared the same way, aged for 0–38 hours at 2 °C, about 40 °C below their Tg (a temperature chosen to avoid crystallization), cooled to −20 °C and heated at 20 °C/min to measure the recovered enthalpy. P6·C9 No felodipine melting peak appeared, so no crystallization occurred during these runs. R17 The unrelaxed fraction was fitted with the KWW function by nonlinear regression. P6·E7

Infrared spectroscopy

For IR, films were spin-coated onto zinc sulfide discs in the dry nitrogen glove box; 128 scans were collected at 4 cm⁻¹ resolution over 6000–400 cm⁻¹ with the optics and sample compartment purged with dry nitrogen. P7 Bulk samples in KBr discs gave similar spectra, but spin coating removed solvent and moisture more easily and was preferred. C10 Spectra were recorded for crystalline and amorphous felodipine, each polymer alone, and dispersions with 10–70% polymer; for HPMCAS and HPMC the height ratio of the 1701 and 1682 cm⁻¹ carbonyl peaks was also plotted at 3, 7, 15 and 25%. P7·A4

What the design adds. Three polymers are compared at matched weight fractions, the nucleation rate is measured directly in films kept at 0% RH, and both candidate explanations, Tg and hydrogen bonding, are measured on the same drug–polymer pairs, so that each can be set against the nucleation data. A1·A2·P2·P4·P7

Experimental analysis

Nucleation: a large drop with the first few percent of polymer

Films of felodipine alone, clear at first, developed roughly spherical birefringent crystallites within days at 0% RH and 22 °C (Figure 1), and the same happened in every composition, so nucleation could be followed across the whole series. R1·C7 Figure 3 shows how a rate is obtained: for felodipine alone the number of sites per volume rises from zero to about 3.5 × 10¹⁰ m⁻³ over 5 days, and the slope of the fitted line is taken as the steady-state nucleation rate. R3·C12

Note by the AI writer (not discussed in the paper): the y-axis unit of Figure 3 is printed as "x10⁰ m⁻³"; only a ×10¹⁰ multiplier gives rates of the order shown in Figure 4, so it is read that way here. Even so, the Figure 3 line implies about 7.5 × 10⁴ m⁻³ s⁻¹, somewhat below the 1.2–1.35 × 10⁵ m⁻³ s⁻¹ shown for felodipine alone in Figure 4; Figure 3 is called "an example" and is not identified as the data behind Figure 4. The points at 0.8–1.2 days also lie below the fitted line, so an early lag cannot be excluded. X7·X8

Figure 4 gives the main result. The rates, read from its logarithmic axis, are listed in Table A. R4

Table A. Nucleation rate of amorphous felodipine (m⁻³ s⁻¹) in spin-coated films at 22 °C and 0% RH, n = 3 (read from Figure 4, approximate). At 0% the three symbols are the same polymer-free control and overlap. R4

Three features stand out. First, the drop is large and front-loaded: 3% w/w of any polymer lowers the rate from about 1.3 × 10⁵ to about 2–3 × 10³ m⁻³ s⁻¹, which the authors describe as "around 1.5 orders of magnitude", and 25% PVP brings it to about 3 × 10², "approximately 2.5 orders" below felodipine alone. R4·C13 The next 22% of polymer therefore adds only about one further decade. R4 Second, the three polymers overlap: from 3 to 15% their rates lie within a factor of about 1.5 of each other and the error bars overlap. R4 Third, the scatter grows at the highest levels: at 20% the HPMCAS point sits about three-fold below PVP and HPMC (about 1.8 × 10² against 5.5 × 10²), with error bars from about 70 to 1.05 × 10³, and at 25% only the PVP point can be seen in the figure. R4·X9 The authors' reading is that "within experimental error, there was no difference" in stabilizing ability at any concentration. C14

Glass transition: only PVP antiplasticizes in the nucleation range

Every bulk dispersion showed a single Tg, between those of the pure components and unchanged by repeated cycling through Tg, which the authors take to mean that the mixtures are miscible over the whole composition range. R5·C8 The values are in Table B. R6

Table B. Onset Tg (°C) of bulk felodipine–polymer dispersions, n = 3 (read from Figure 5, approximate; pure polymer values from the text). R6·M1·M2·M3·M4

With PVP, Tg rises steadily, by about 10 °C at 25% (from about 44 to 55 °C); the authors call PVP "the best antiplasticizing agent". R6·C25 With HPMCAS and HPMC, Tg stays at the felodipine value up to 25% and rises only above about 30% w/w, consistent with earlier observations on co-lyophilized mixtures (Shamblin et al. 1998). R6·C15 Over the 0–25% range used for nucleation the three systems therefore span about 10 °C in Tg. C26

Note by the AI writer (not discussed in the paper): at 7% PVP the Tg reading (about 44 °C) equals that of felodipine, so the statement that PVP raises Tg "at all concentrations" holds from about 15% upward. X15

Against the ideal-mixing lines, the PVP data lie close to the Gordon–Taylor line, while the HPMCAS and HPMC data fall below both lines, most at mid-range compositions; at 50 wt%, for example, the HPMCAS and HPMC Tgs are about 50 and 49 °C against Gordon–Taylor values of about 63 and 68 °C read from the plotted lines. R6·R7·C16 The authors find no sign of a proposed link between the size of this deviation and the difference between the component Tgs (Shamblin et al. 1998). C17

Molecular mobility below Tg

The enthalpy relaxation data for 9:1 dispersions aged at 2 °C (Figure 6) are summarized in Table C. R8·R9

Table C. Enthalpy relaxation at 2 °C: unrelaxed fraction read from Figure 6 (approximate) and KWW parameters from the text (fits with R² above 0.9). R8·R9

The PVP dispersion relaxed fastest and the HPMCAS dispersion slowest. R9 The paper reports these values without discussing them, and reports no relaxation time for felodipine alone. X13

Note by the AI writer (not discussed in the paper): the order of relaxation times (PVP shortest) is opposite to the order of dispersion Tg (PVP highest), and without a polymer-free reference the data cannot show whether any of the polymers slowed the mobility of felodipine. X13

Hydrogen bonding: all three bond, PVP most strongly

The IR results are summarized in Table D. R16·R11·R13·R14·R15

Table D. Felodipine NH and carbonyl bands; dispersion NH positions from Table 2 (70% w/w polymer) and the text. R10·R16·R11·R12·R13·R14·R15

With PVP (Figure 7a), the NH band broadens and grows a component on its low-wavenumber side: at 30% PVP the band has two poorly resolved maxima near 3345 and 3290 cm⁻¹, and with more PVP the lower one grows at the expense of the upper one. R11 The authors read the downward shift as a hydrogen bond between felodipine and PVP that is stronger than those in amorphous felodipine, and the remaining 3345 cm⁻¹ shoulder, still present at 70% PVP, as a population of drug molecules that keep bonding to each other. C18·C19 The carbonyl region (Figure 7b) is dominated by the intense PVP carbonyl band, so no clear change could be read there. R12·C22

With HPMCAS (Figure 8a), the NH band moves slightly up, to about 3355 cm⁻¹ at 70%, which the authors read as a slightly weaker interaction than in amorphous felodipine. R13·C20 With HPMC (Figure 9a) it stays at about 3341–3345 cm⁻¹ and broadens slightly. R14 An unchanged NH position could mean either no interaction or one of the same strength; the carbonyl region decides between them. With both cellulosics the free carbonyl peak at 1701 cm⁻¹ grows relative to the bonded peak at 1682 cm⁻¹ as polymer is added (Figure 10), from a ratio of about 1.07–1.10 for the drug alone to about 2.15 with 70% HPMCAS and about 1.61 with 70% HPMC. R15 The authors write that this "strongly suggests" that felodipine's NH–carbonyl bonds are broken and replaced by bonds from the NH to acceptor groups of the polymer, increasingly so with more polymer. C21 Their ranking of bond strength is PVP > HPMC > HPMCAS, with the caveat that not every NH group bonds to polymer and that the extent of drug–polymer bonding could not be extracted from the spectra. C29·C30

Note by the AI writer (not discussed in the paper): the captions of Figures 8 and 9 say that the percentages are weight percentages of PVP; the figures show HPMCAS and HPMC dispersions. In the PVP spectra a second band near 3225–3240 cm⁻¹, weak in amorphous felodipine and distinct at 40–70% PVP, is not assigned or discussed. X2·R11

Crystal morphology

Figure 2 compares crystallites grown at 22 °C from felodipine alone and from films with 3% of each polymer. Alone, felodipine grows a uniform, circular spherulite with a smooth boundary to the amorphous film. With polymer the crystallites have irregular edges and carry dark points ("droplets") that form ring-like patterns on the crystals and line their edges, clearest with HPMCAS. R2 The authors attribute the dark points to domains of pure polymer or of a polymer-rich phase that separated from the dispersion once the drug crystallized, and read the irregular outline as polymer hindering the circular extension of growth. C11 Neither the composition of the droplets nor any growth rate was measured. X14

Discussion

Neither Tg nor hydrogen-bond strength explains the result

The results "clearly show", in the authors' words, that the three polymers inhibit crystallization of amorphous felodipine by reducing the nucleation rate. C23 Set against Table B and Table D, the equal rates in Table A rule out the two simple explanations for this system. HPMCAS and HPMC left Tg at the felodipine value up to 25% yet lowered the nucleation rate as much as PVP, which raised Tg by about 10 °C. The authors conclude (i) that an increase in Tg relative to the drug "was not required" to lower the nucleation rate, in line with Matsumoto and Zografi (1999) and Khougaz and Clas (2000), and (ii) that the nucleation rate did not correlate with the Tg of the dispersion; their conclusions extend this to the Tg of the pure polymer. C26·R21 A higher Tg "would be anticipated" to help, but the results "suggest that it is not the only factor" in the physical stability of solid dispersions. C27

The same holds for hydrogen bonding. PVP bonds more strongly to felodipine than felodipine bonds to itself, HPMC about as strongly and HPMCAS slightly less, yet all three slow nucleation equally, so "although the strength of the drug–polymer interaction varies, there is no correlation" with the effect on nucleation. C31·R16 This differs from Miyazaki et al. (2004), who credited stronger interactions for the better performance of polyacrylic acid over PVP with acetaminophen. C31·R21 The authors add a general point: some hetero (adhesive) interaction between drug and polymer is needed for any one-phase mixture, so the specific role of drug–polymer interactions in crystallization inhibition is hard to isolate, although some studies point to it (Aso et al. 2002; Fujii et al. 1991). C28

The Tg data also say something about mixing. The negative deviations of the cellulosic dispersions from the ideal-mixing lines are read, following Brekner et al. (1988), as a sign that the felodipine–HPMCAS and felodipine–HPMC interaction energy is lower than the sum of the drug–drug and polymer–polymer energies, whereas felodipine–PVP mixes more ideally; the authors "suggest" this and note that the IR data, with the weakest interactions in the system with the greatest negative deviation, support it. C32

Note by the AI writer (not discussed in the paper): two points limit this reading. The largest deviation in Figure 5 belongs to HPMC (at 70 wt% about 57 °C against a plotted line of about 87 °C; HPMCAS about 63 against 78 °C), while the weakest bond belongs to HPMCAS, so the IR support holds only for the cellulosics as a group against PVP. And because the plotted Gordon–Taylor lines use Tg in °C in $K$, on the usual kelvin basis PVP also lies about 14 °C below its line at 25 and 50 wt%; the ranking of the deviations (PVP smallest) is the same either way. X10·X3

How the authors explain equal inhibition

To explain why three different polymers act alike, the authors go back to Eq. 8 and consider each way a polymer could change the two barriers. The chain below sets out their argument.

Driving force. In a molecular mixture the drug's free energy falls, so $\Delta G_v$ and the driving force fall too. C34 Using Eq. 9 with literature data for "reasonably comparable systems", PVP with sucrose or water (Hancock and Zografi 1993; Zhang and Zografi 2001), the drug's activity is lowered significantly only when the polymer volume fraction approaches a value greater than 0.5. R20·E9 At 3–25% polymer the driving force should therefore be almost unchanged, and this route "probably cannot account" for the lower nucleation rate. C35 The argument rests on systems other than felodipine; no Flory–Huggins parameter or activity was measured here. C35

Dilution. If the polymer acted only as a diluent of the drug, the nucleation rate would vary in direct proportion to the volume fraction of polymer; 3% polymer would then lower it by only a few percent, not by 1.5 orders of magnitude, so this is "clearly not the case". C37·R4 (The text cites Figure 9 at this point; the nucleation rates are in Figure 4. X1)

Coupling through specific interactions. Drug–polymer bonds "might be expected" to slow drug diffusion by coupling the motion of the drug to the polymer, as PVP does for sucrose, but drug molecules bonded to the polymer and drug molecules bonded to each other coexist (Table D), so the authors call this "hard to assess". C38·R11

Polymer accumulation at the crystal front. A miscible polymer is unlikely to be built into the drug's crystal lattice, so when the drug crystallizes the polymer is rejected and must diffuse away into the amorphous matrix (Stein 1994). Below Tg this diffusion is slow, so polymer can pile up at the crystal–amorphous interface. Such a polymer-rich layer "would impede" transport of drug to the crystal, because the local drug concentration is lower and the local Tg higher, and it "may also" raise the energy penalty for creating new surface, the $\Delta G_s$ term of Eq. 8. A similar mechanism had been proposed to explain why the extent of indomethacin crystallization with low PVP levels depended on particle size (Crowley and Zografi 2003). C39

From these considerations the authors find it "reasonable to speculate" that the polymers increase the magnitude of $\Delta G_a$, with a consequent decrease in nucleation and growth rates. C40 After the first large drop with a small amount of polymer, the effect "appears" to depend on the amount of polymer and to be relatively independent of polymer type, "consistent with" the polymers acting as a physical impediment to drug transport at the nucleus–matrix interface, with the volume fraction of polymer as the controlling parameter. C41 Figure 2 offers "some evidence": the uniform spherulitic growth of the pure drug is disrupted by the polymers, and the small droplets on the crystallites are "thought" to be a polymer-rich phase, an observation that "clearly ... requires further investigation". C42·R2

Where the result applies

The conclusions are stated for dry conditions: the three polymers are equally effective "in the absence of moisture", at a given weight percentage, for one drug, felodipine, in thin films at 22 °C. C14·A1·P2 The polymer range for nucleation was 3–25% w/w, and the equal-effectiveness statement is made for these three polymers specifically. A2·C41

Note by the AI writer (not discussed in the paper): for formulation this suggests that, for felodipine in dry storage, a few percent of any of these polymers gives most of the protection against nucleation, so the choice among PVP, HPMCAS and HPMC at low loading could rest on other properties such as dissolution behaviour, water uptake or processing, none of which was tested. Under humidity the polymers differ strongly in water uptake, and the result may not carry over.

Limitations

Limits the authors state

• Nucleation in the films is "most likely heterogeneous", while Eq. 8 is the homogeneous theory. C33
• Not all felodipine NH groups bonded to polymer, and "no comment can be made" about the extent of drug–polymer interaction, which could not be extracted from the spectra. C30
• The effect of specific interactions on drug diffusivity is "hard to assess". C38
• The droplet observation "requires further investigation". C42
• For PVP, interference from the polymer's own carbonyl band prevented any reading of the felodipine carbonyl region. C22

The AI writer's own judgement

What holds. The central observation is strong: a one- to two-and-a-half-decade fall in nucleation rate with every polymer, measured in triplicate under well-controlled dry conditions. R4 The contrast with Tg is clean, because HPMCAS and HPMC left Tg unchanged up to 25% yet slowed nucleation as much as PVP. R6·R4 For the cellulosics, the NH shift and the growth of the free carbonyl band agree. R13·R14·R15 The printed KWW parameters reproduce the plotted relaxation data. E7

What weakens it.

• "Equally effective" rests mainly on 3–15%: at 20% the polymers differ about three-fold inside wide error bars, and at 25% only PVP is visible in Figure 4. On a log scale with these error bars, a two- to three-fold difference between polymers could go unseen. X9
• The comparisons cross sample sets: Tg was measured on bulk rotary-evaporated dispersions, nucleation on spin-coated films heated to 90 °C, and the hydrogen-bond ranking on films with 70% polymer, where nucleation was not measured. At 3–25% the cellulosic NH bands move by only a few cm⁻¹. X12·X11
• Absolute nucleation rates depend on an area-to-volume conversion described as multiplication rather than division, and the unit of Figure 3 is misprinted; the Figure 3 slope is lower than the Figure 4 value for the same drug. X6·X7
• The ideal-mixing lines of Figure 5 use a °C-based Gordon–Taylor constant and a Couchman–Karasz form that cannot be reproduced from Eq. 4, so the statement that PVP mixes ideally depends on that choice. X3·X4
• The kinetic-barrier mechanism is argued, not measured: there are no growth rates, no analysis of the droplets, no measurement of polymer at the crystal interface and no drug diffusivity; the driving-force argument rests on PVP–sucrose and PVP–water data. X14·C35·C40
• The enthalpy relaxation data have no polymer-free control and are not used in the argument. X13
• Minor source errors: Figure 9 cited for nucleation data, Figures 8 and 9 captions naming PVP, a sign misprint in Eq. 7, and references 8 and 19 being the same paper. X1·X2·X5·X16

Conclusion

This paper set out to find which polymer property, a higher Tg or stronger drug–polymer hydrogen bonding, lets a polymer slow the crystallization of a molecularly dispersed amorphous drug. For felodipine in dry films at 22 °C, the answer was neither: PVP, HPMCAS and HPMC each cut the nucleation rate by about 1.5 orders of magnitude at 3% w/w and about 2.5 orders at 25%, and did so equally within experimental error, although only PVP raised Tg in that range and the hydrogen-bond strength ranked PVP > HPMC > HPMCAS. R4·R6·R16·C14·C26·C31 Having argued that the driving force barely changes and that dilution is far too weak, the authors speculate that the polymers raise the kinetic barrier to nucleation, perhaps through a polymer-rich layer that builds up at the nucleus surface, with the size of the effect set by the amount of polymer rather than its chemistry. C35·C37·C39·C40·C41 That mechanism remains a hypothesis in this paper; what the data establish is that, for this drug and these polymers in dry conditions, Tg and bond strength did not rank the inhibitors. C40·X14