How heating loosens drug–drug hydrogen bonds in crystals, glasses and liquids of seven dihydropyridines

Background and motivation

Amorphous materials behave in a complex way with temperature and time. A liquid cooled below its melting point is a supercooled liquid; cooled further below the glass transition temperature ($T_g$) it becomes a glass. The supercooled liquid can be treated as an equilibrium state, metastable only with respect to the crystal, but the glass is not in configurational equilibrium, and the glass transition seen on heating is a kinetic event spread over a temperature range rather than a thermodynamic transition. Near $T_g$ the specific volume begins to rise faster on heating, because the thermal expansion coefficient of the liquid is considerably larger than that of the glass, which is close to that of the crystal. The authors note that these volume differences will affect intermolecular distances and hence intermolecular interactions (McKenna 1989, ref 1; Ediger 1996, ref 2). C4

For formulation the interest is practical. Amorphous drugs, especially in solid dispersions, can raise bioavailability, but chemical and physical instability has limited their use, and how amorphous molecules associate with each other is thought to shape both the pure amorphous phase and its interaction with other components (Chiou 1971, ref 3; Serajuddin 1999, ref 4; Hancock 1997, ref 6). C5 In a companion paper the same authors compared hydrogen bonding in seven amorphous 1,4-dihydropyridine calcium channel blockers with their crystals (Tang 2002, ref 7). In these molecules the hydrogen bond forms between the N–H of the dihydropyridine ring and an acceptor group on a neighbouring molecule. C10 In amorphous felodipine the main NH band lies at 3330 cm−1, lower than in the crystal, which the companion study read as stronger average hydrogen bonding in the glass, and a second band at 3420 cm−1 was assigned to N–H groups that are not hydrogen bonded. R5·C13 In crystalline felodipine the N–H forms a relatively weak hydrogen bond (Fossheim 1986, ref 12). C12

What was missing was the temperature dependence. Infrared spectroscopy is well established for following hydrogen bonds as a function of temperature, and vibrational spectroscopy had been applied to crystalline and amorphous materials, but, as the authors put it, "little data exists that compares the two phases" (Wolkers 1998, ref 8; Lutz 1992, ref 14; Söderholm 2000, ref 15). C7 The authors expected the interactions in the amorphous phase to change with temperature differently from those in the crystal, and they were particularly interested in the passage from glass to supercooled liquid, where hydrogen bonds might "abruptly weaken"; they add, "Indeed, we find this behavior". C6

The paper therefore extends the companion study in one direction: it heats the same seven compounds, as crystals and as glasses, from 0 to 200 °C inside an attenuated total reflection (ATR) infrared cell, follows the position of the NH stretching band, and relates the changes to the glass transition and melting measured by differential scanning calorimetry (DSC). A1·A2·A3·P1·P3

Background concepts

Glass, supercooled liquid and the reduced glass transition temperature. Below $T_g$ the amorphous solid is a glass; between $T_g$ and the melting temperature $T_m$ it is a supercooled liquid; above $T_m$ it is the melt, which is the same liquid whether it was reached from the crystal or from the glass. The paper uses the DSC $T_g$ and $T_m$ to mark where each state begins and ends, which it calls "necessary for meaningful interpretation of the spectroscopic data". C9 Table I also reports the ratio

In plain terms: the glass transition temperature divided by the melting temperature, both in kelvin. The table note states that the ratio was calculated after converting to kelvin, and recomputation shows that the onset $T_g$ was used: for felodipine, (43.7 + 273.15)/(142.9 + 273.15) = 0.762, printed as 0.76. E2·R1.a A value near 0.7 is typical of pharmaceutical materials (Fukuoka 1989, ref 11). C8

The NH stretching band as a hydrogen-bond probe. A hydrogen bond pulls on the N–H, lowers its stretching frequency and broadens the band. The paper therefore reads a shift of the NH band to higher wavenumber as a weaker hydrogen bond and a shift to lower wavenumber as a stronger one; the abstract says such an increase "was taken to indicate" weakening. C11 In amorphous felodipine two NH bands are visible: the hydrogen-bonded band near 3330 cm−1 and a small free-NH band near 3420 cm−1, whose areas can be followed separately. R5

The wavenumber–temperature correlation. Following Lutz and van der Maas, the paper summarises each state by the change of band position per degree:

In plain terms: $\Delta \nu$ is the shift in band position, so the WTC is the slope of band position against temperature, in cm−1 °C−1, obtained here by a straight-line fit for each state. E1 A positive WTC on heating is typical of hydrogen-bonded systems and means the bond weakens as temperature rises (Lutz 1994, ref 13; Lutz 1992, ref 14). C24 For crystalline sugars the WTC depends mainly on how the O···O distance of the hydrogen bond changes with temperature, which in turn depends on how much the material expands and on how the groups are oriented (ref 13). C25 A short hydrogen bond also changes its strength more than a long one for the same change in length (Slootmaekers 1991, ref 10). C27

Attenuated total reflection. In ATR the infrared beam is reflected inside a crystal, here a single-reflection diamond, and probes a thin layer of the sample pressed against it. No KBr pellet is needed and sample handling is minimal, which suits heating a glass or powder in place. P1

Enthalpy and entropy of fusion. The enthalpy of fusion is the energy difference between liquid and crystal at the melting point; the entropy of fusion is the corresponding gain in disorder and is always positive. The paper compares both with how much the NH band shifts on melting. C30·C31

Methods in detail

The study has four parts: a thermal map of the seven compounds by DSC; variable-temperature infrared spectra of each compound as crystal and as glass; a band-area analysis of amorphous felodipine; and two comparisons across compounds, one at melting and one at the glass transition. The flow below shows how they connect.

Part 1: a thermal map of the seven compounds

Problem. To say whether a spectrum was taken from a glass, a supercooled liquid, a melt or a crystal, the authors need $T_g$ and $T_m$ for each compound. C9

What they did. About 5 mg of each crystalline powder was weighed into an aluminium pan with a vented lid and heated at 10 °C min−1 under dry nitrogen (50 ml min−1) in a Mettler Toledo 820 DSC calibrated with indium; melting temperatures and heats of fusion came from this scan. Glasses were made in situ by quenching the melt. $T_g$ was taken from a second scan after heating to 10 °C above $T_g$ to erase the thermal history, and onset, midpoint and end point were reported. Samples were analysed in duplicate. P3 Each crystal showed one endotherm, its melting, and each glass a stepwise change in heat capacity. R2

Part 2: variable-temperature ATR infrared spectra

Problem. The hydrogen bond of interest has to be followed in place while the sample is heated through its glass transition and melting, without KBr and with little handling. P1

What they did. Spectra were collected on a Biorad FTS 575C spectrometer (KBr beamsplitter, Peltier-cooled DTGS detector) with a Golden Gate single-reflection diamond ATR accessory whose temperature controller covers 0–200 ± 1 °C, at 4 cm−1 resolution and 128 scans, with the optics and accessory purged with dry nitrogen; a polystyrene reference monitored wavenumber accuracy. Freshly prepared glass or crystalline powder was loaded at room temperature (22 °C), the cell was heated or cooled to the analysis temperature, and the spectrum was recorded. P1 For felodipine, Figure 1 shows crystalline spectra at 25, 40, 80, 120, 160 and 200 °C and amorphous spectra at 0, 20, 40, 80, 120, 160 and 200 °C; the peak-position plots (Figures 3 and 4) use steps of 5–10 °C below 100 °C. P1·A3 The paper does not say whether one sample was stepped through the series or a fresh sample used at each temperature, nor how long each temperature was held. X12 Samples heated to the highest temperatures were rescanned after cooling and showed no visible degradation, and duplicates for "a number of samples" were reproducible. P2·R3

From spectra to numbers. The position of the NH band was plotted against temperature, and for each state a straight line was fitted by linear regression; its slope is the WTC (equation above). The error given is the standard deviation of the regression, and every fit had a correlation coefficient above 0.98. P5·E1·R9 The authors also read from these plots the onset, midpoint and end of the glass-transition event as it appears in the infrared data. P6 The temperature range of each fit is not stated. X2

Part 3: free and bonded NH in amorphous felodipine

What they did. For amorphous felodipine, the bonded (about 3330 cm−1) and free (about 3420 cm−1) NH bands were separated by peak fitting at each temperature, and each area was normalised to the area of a band at 775 cm−1 used as an internal reference. P4·R5 The band shapes and constraints of the fit are not described.

Why it helps. The band position tells how strong the average hydrogen bond is; the free-NH area tells whether the number of unbonded groups changes. A larger free-NH area at higher temperature would mean that some hydrogen bonds break rather than merely lengthen. C14

Part 4: two comparisons across compounds

At melting. For each crystalline compound the paper takes the NH shift on melting,

In plain terms: the NH band position of the liquid ($l$) minus that of the crystal ($c$). A positive value means weaker hydrogen bonding in the melt, a negative value stronger. E4 This shift is plotted against the heat of fusion and the entropy of fusion (Figure 5). P8 The paper does not state the temperatures at which the two band positions were read, nor how the entropy of fusion was obtained; the heats of fusion are not tabulated. X6

At the glass transition. For each glass the total NH shift on passing from glass to supercooled liquid was obtained by extrapolating the straight lines of the glass and of the supercooled liquid and taking the difference in band position at the beginning and end of the $T_g$ event. This shift is plotted against the calorimetric $T_g$ (Figure 6) and compared with the heat-capacity step $\Delta C_p$ of Table I. E3·P7·C37

Experimental analysis

Main result 1: similar molecules, different glass transitions. Table A gives the DSC data. The compounds have similar structures and molecular weights (346.3–479.5), yet their $T_g$ midpoints span 13.4–44.6 °C and their melting points 124.5–171.0 °C. R1 The text says "up to 30°C" for both; that fits $T_g$ (a spread of 31.2 °C) but not $T_m$, which spans 46.5 °C. C8·X1 The ratio $T_g/T_m$ is 0.71–0.73 for five compounds and 0.76 for felodipine, which the authors call "reasonably constant at approximately 0.72". R1·C8 Nicardipine has no melting point and no crystalline data in either table, and the paper does not say why. R1.e·X13

Table A. DSC data of the seven compounds (from Table I; Tg/Tm in kelvin with the onset Tg). R1

Main result 2: crystal, glass and liquid respond differently (felodipine). Figure 3 is the central figure. Crystalline felodipine keeps its NH band near 3367 cm−1 at 30 °C, creeping up to about 3370.7 cm−1 at 140 °C, before it drops to about 3347 cm−1 on melting (values read from Figure 3, approximate). R7 The authors read the small rise as a small loss of hydrogen-bond strength ("the data indicate"), and the lower band of the melt as shorter average hydrogen bonds in the liquid than in the crystal, with a much wider spread of bond lengths and angles shown by the broad band ("The data suggest"). C17·C18 Amorphous felodipine starts much lower, near 3331 cm−1 at 0–25 °C and about 3333 cm−1 at 40 °C; it rises steeply between 40 and 65 °C (to about 3338.7 cm−1), then more slowly to about 3354 cm−1 at 200 °C, continuing smoothly through $T_m$, where it joins the melt of the crystal. Above $T_m$ the spectra of both starting materials are identical. R7 The steep part coincides with the glass-transition region of Table I, "although the calorimetrically determined Tg event is much sharper". C19

Main result 3: the same ordering in all seven compounds. Table B gives every WTC. In each compound the liquid WTC is the largest, 0.106–0.165 cm−1 °C−1, against 0.056–0.110 for the glasses and 0.007–0.075 for the crystals. R9 The authors conclude that hydrogen bonding is "most sensitive to temperature" in the supercooled liquid and liquid, and they state that, for all compounds, the WTC of the supercooled liquid was identical with that of the melt. C1·C23 Bands of groups without hydrogen bonds are a useful control: a CH stretching band moved by less than 5 cm−1 in total, with a small negative WTC, which the authors call typical for vibrations subject only to van der Waals forces. R8·C20

Table B. WTC of the NH band (cm−1 °C−1, ± standard deviation of the regression) and the glass-transition region read from the infrared data (from Table II). R9

Readings of the plotted points agree with most of Table B: for amorphous felodipine a straight line through 0–40 °C gives about 0.058 and one through 65–200 °C about 0.107, and the crystalline isradipine and nisoldipine points between 20 and 100 °C give about 0.006 and 0.026. Crystalline felodipine is the exception: its points rise by about 3.4 cm−1 between 30 and 140 °C, a slope near 0.03, against 0.010 ± 0.001 in the table; even a fit over 30–90 °C gives about 0.014. X2

Main result 4: a step through the glass transition. Isradipine and nisoldipine (Figure 4) repeat the felodipine pattern: little change as a glass, then a clear change of slope near $T_g$. Amorphous isradipine rises from about 3329 cm−1 at 0 °C to about 3342.5 cm−1 at 100 °C, where crystallization begins, with changes in several bands and the NH band moving towards the crystalline position (about 3346–3348 cm−1). R10 Amorphous nisoldipine rises from about 3333 to about 3344.5 cm−1 at 90 °C and then, as it crystallizes after its glass transition, falls to about 3323 cm−1 at 120 °C, on top of the crystal band. R11 Nimodipine and nifedipine also crystallized on heating, at 100 and 70 °C. R12 The authors' reading is that hydrogen-bond strength "decreases most rapidly as the material undergoes the glass transition", and they find it "reasonable to conclude" that the change is directly related to the glass-transition event; the abstract adds that the effect "was more pronounced for some samples than for others". C2 They see this as unsurprising, since $T_g$ marks a co-operative structural change. C36

Two qualifications come from the paper's own data. For nicardipine and nimodipine, the two lowest-$T_g$ compounds, no distinct change of slope near $T_g$ could be seen. R12 And the infrared step is much broader and mostly later than the calorimetric one (Table C). R9·R1·X8

Table C. The glass transition seen by DSC and by the NH band (from Tables I and II). R1·R9·R12

Main result 5: free N–H groups appear above Tg. In amorphous felodipine the free-NH band grows relative to the bonded band as temperature rises (Figure 1b). R5 In the peak-fitted areas (Figure 2, values approximate), the free-NH area stays near 0.0013–0.0016 up to 30 °C and then rises to about 0.0024 at 50 °C, 0.0038 at 90 °C and 0.0065 at 200 °C. The bonded area is about 0.047–0.048 up to 35 °C, rises to about 0.053 at 50 °C and then falls to about 0.038 at 200 °C. R6 The authors say both changes occur "just above" $T_g$ and that the amount of free NH "appears to increase somewhat"; the fall of the bonded area is "consistent with" a smaller absorption coefficient at higher temperature and/or fewer bonded groups. C14·C15 The text describes the bonded area as "initially fairly constant" and then decreasing; the figure shows a rise before the fall. X3

Main result 6: melting and the entropy of fusion. Figure 5 plots the NH shift on melting of the six crystalline compounds. The shifts run from about −24 cm−1 to about +44 cm−1 (read from Figure 5). Against the heat of fusion (about 29.7–38.7 kJ mol−1) the points scatter, and the authors see "no simple relationship"; against the entropy of fusion (about 71.4–96.0 J K−1 mol−1) they fall near a straight line, and "a reasonably linear correlation appears to exist (R² = 0.96)". R13·C30·C31 Felodipine has the smallest entropy of fusion and is the only compound whose hydrogen bonds become stronger on melting. C31·R7 The authors conclude that "Our results suggest" the extent of hydrogen bonding in the liquid plays a role in the entropy difference between crystal and liquid. C32 The points are not labelled; matching each pair to the melting points of Table I identifies them, so that, for example, isradipine shows about +5 cm−1 and nimodipine about +44 cm−1. X5

Main result 7: higher Tg, larger change at Tg. Figure 6 plots the calorimetric $T_g$ against the NH shift across the glass transition. The shift rises from about 0.5 cm−1 for glasses with $T_g$ near 13–14 °C to about 5.2 cm−1 for $T_g$ near 45 °C, with one point (about 2.5 cm−1, 35.6 °C) above the drawn line (read from Figure 6). R14 The authors write "There appears to be a reasonably linear relationship" and call it "a possible relationship": higher-$T_g$ compounds change more in hydrogen bonding through $T_g$, and lower-$T_g$ compounds are more similar in glass and supercooled liquid. C35 "Somewhat surprisingly", they found no correlation between the decrease in hydrogen-bond strength and $\Delta C_p$. C37 Nor does $T_g$ follow molecular weight: the two heaviest compounds, nicardipine (479.5) and nimodipine (418.5), have the two lowest $T_g$ values. R1·C33 Because the companion study found about the same average hydrogen-bond strength in all seven glasses, the authors attribute the $T_g$ differences "most likely" to molecular geometry or other noncovalent interactions. C34

Controls and checks. The CH band, which forms no hydrogen bond, barely moves; the authors use it to show that bands of non-hydrogen-bonding groups are much less sensitive to temperature. R8·C20 Rescans after cooling from the highest temperatures showed no visible degradation. R3 Above $T_m$ the melt gives the same spectrum whether it started as crystal or glass, "as expected". R7

Discussion

The authors' explanation. The paper ties the results together with one physical picture, summarised below. Each step is the authors' reasoning; thermal expansion, molecular mobility and configurational freedom were not measured. C16·C26·C28·C3

The authors say the main differences between phases "can be ascribed" to thermal expansion, while other factors, such as how short the hydrogen bond is, also contribute, particularly in crystals. C26·C27 The extra free NH above $T_g$ is put down, "presumably", to the increase in molecular mobility. C28 At melting, the change in hydrogen bonding is linked to the entropy gained, not the energy. C32

What the glass gains from its hydrogen bonds, according to the authors. In the discussion the authors draw several consequences that this paper does not test. The more extensive hydrogen-bond network of the glass "presumably contributes" to its greater physical stability and reduced molecular mobility. C38 The lower fraction of free N–H in the glass "suggests" greater short-range orientational order there (Leheny 1996, ref 16). C39 Two populations of molecules, bonded and free, "may well contribute" to the non-exponential relaxation often seen in organic glasses (Shamblin 2000, ref 17). C40 The free fraction itself might be more chemically reactive and less physically stable ("It can be speculated"). C29 Finally, temperature-dependent changes in molecular association are "likely to be important" both for the stability of amorphous materials and for forming molecular-level mixtures in melt extrusion and solid-dispersion manufacture. C41

Scope. The results concern seven pure drugs from one chemical family, each with a single dihydropyridine N–H donor, heated in an ATR cell between 0 and 200 °C under dry nitrogen. A1·A3·P1 No polymer or other excipient was present, and four of the glasses crystallized during heating (nifedipine at 70 °C, isradipine, nimodipine and nisoldipine near 100 °C), so the amorphous series above those temperatures no longer describes a single amorphous phase. R10·R11·R12·X12

Note by the AI writer (not discussed in the paper). Two points may matter for formulation work. First, drug–drug hydrogen bonds in these glasses change little below $T_g$ but loosen quickly above it; if a polymer has to replace those bonds during hot processing, the temperature above $T_g$ may matter as much as the mixing time. Second, the same ATR approach, applied to a drug–polymer dispersion, could show whether drug–polymer hydrogen bonds follow the same crystal-like, glass-like and liquid-like regimes. Neither point was tested here.

Limitations

Limitations stated by the authors. The authors note that the calorimetric $T_g$ event is "much sharper" than the step in the infrared data. C19 For nicardipine and nimodipine no distinct change of slope near $T_g$ could be seen, and there were not enough glass data for their WTC. R12·R9 The loss of bonded-NH area may reflect a temperature-dependent absorption coefficient rather than fewer hydrogen bonds ("and/or"). C15 The data for non-hydrogen-bonding bands are "not shown". R8 The explanations are worded with hedges: "can be ascribed", "Presumably", "most likely", "It can be speculated", "may well". C26·C28·C34·C29·C40

The AI writer's own judgement (1): how the numbers were obtained. Several reading choices are not stated, and some printed statements do not match the data. The temperature range of each WTC fit is not given; for crystalline felodipine the plotted points imply a slope near 0.03 cm−1 °C−1 over 30–140 °C, about three times the tabulated 0.010 ± 0.001. X2 The text describes the bonded-NH area as constant and then falling, whereas Figure 2 shows a rise to about 0.053 at 50 °C first. X3 The text gives a variation of "up to 30°C" in melting points, while Table I spans 46.5 °C. X1 The $T_g$ marked on Figure 4b is the nisoldipine onset, not the midpoint used on Figures 3 and 4a. X4 The infrared glass-transition region is 12–25 °C wide and its midpoint lies up to 10.5 °C above the DSC midpoint; since the heating protocol of the ATR cell (step size, hold time, one sample or fresh samples) is not described, part of this offset may come from the measurement rather than the material. X8·X12 Nicardipine has no crystalline data and no reason is given. X13

The AI writer's own judgement (2): how far the correlations carry. Figures 5 and 6 are not labelled by compound; the labels can be inferred from Table I (melting points and $T_g$ values), but they remain an inference. X5 The heats of fusion are not tabulated and the derivation of the entropy of fusion is not stated, although the plotted pairs agree with dividing the heat of fusion by $T_m$. X6 Five of the six heats of fusion lie within about 37.3–38.7 kJ mol−1, so among those compounds the entropy differs mainly through $T_m$; the entropy correlation cannot be separated from a dependence on $T_m$, and the felodipine point spans most of the range. X15 In Figure 6, the two lowest points have $T_g$ values that only nimodipine and nicardipine have, yet those two showed no infrared step and no glass WTC, and how their shift was obtained is not stated; one further point lies about 7 °C above the line. X9·X10 The statement of "no correlation" between the change at $T_g$ and $\Delta C_p$ is not shown: pairing the Table I $\Delta C_p$ values with the Figure 6 shifts gives a positive correlation (r ≈ 0.77, an extractor calculation on plot readings). X7 The phase-level conclusion, that the liquid is far more temperature-sensitive than the glass or crystal, rests directly on Table II and is consistent across all seven compounds; the cross-compound relations rest on six or seven points each. R9·R13·R14

Conclusion

This short paper turns the NH stretching band of seven dihydropyridine calcium channel blockers into a thermometer of their hydrogen bonds. Its main contribution is direct: for every compound, hydrogen bonds weaken on heating least in the crystal and the glass and most in the supercooled liquid and melt, with WTC values of 0.007–0.075, 0.056–0.110 and 0.106–0.165 cm−1 °C−1. R9·C1 In amorphous felodipine, isradipine and nisoldipine the band steps up through the glass-transition region, which the authors read as an abrupt weakening of hydrogen bonds at $T_g$; in amorphous felodipine more N–H groups become free above $T_g$. R7·R10·R11·R6·C2·C14 Two cross-compound relations follow: the NH shift on melting tracks the entropy of fusion (R² = 0.96), and the shift across $T_g$ is larger for higher-$T_g$ compounds ("a possible relationship"). R13·C31·R14·C35 The authors frame the whole as "greater configurational freedom" from crystal to glass to liquid and suggest, without testing it, that these changes in association matter for amorphous stability and for mixing in melt extrusion and solid dispersions. C3·C41