Solvent Vapor Annealing of Amorphous Carbamazepine Films for Fast Polymorph Screening and Dissolution Alteration.

Solubility enhancement and thus higher bioavailability are of great importance and a constant challenge in pharmaceutical research whereby polymorph screening and selection is one of the most important tasks. A very promising approach for polymorph screening is solvent vapor annealing where a sample is exposed to an atmosphere saturated with molecules of a specific chemical/solvent. In this work, amorphous carbamazepine thin films were prepared by spin coating, and the transformation into crystalline forms under exposure to solvent vapors was investigated. Employing grazing incidence X-ray diffraction, four distinct carbamazepine polymorphs, a solvate, and hydrates could be identified, while optical microscopy showed mainly spherulitic morphologies. In vitro dissolution experiments revealed different carbamazepine release from the various thin-film samples containing distinct polymorphic compositions: heat treatment of amorphous samples at 80 °C results in an immediate release; samples exposed to EtOH vapors show a drug release about 5 times slower than this immediate one; and all the others had intermediate release profiles. Noteworthy, even the sample of slowest release has a manifold faster release compared to a standard powder sample demonstrating the capabilities of thin-film preparation for faster drug release in general. Despite the small number of samples in this screening experiment, the results clearly show how solvent vapor annealing can assist in identifying potential polymorphs and allows for estimating their impact on properties like bioavailability.

Introduction

The search for polymorphic crystal forms is necessary to identify favorable properties for various fields.[1−6] For active pharmaceutical ingredients (APIs), polymorphs should remain unchanged over the shelf-life time[7,8] and be compatible with excipients and processes, and their production has to be very reproducible.[9] As such, many screening experiments are applied to generate and discover new, potentially best suited and stable polymorphs.

Many different approaches exist which all aim at generating different processing pathways for finding new polymorphs. Bulk crystallization experiments employ seeding,[10] different solvents,[5,6,11] degree of saturation, and varying temperature, among others, to enable the production of new polymorphs which in other terms just represent the fact that the energy landscape is screened.[12] Even the usage of structure prediction methods is becoming more important as potential new polymorphs might be more easily accessible in-silico but with the prediction rate exhausting physical capabilities.[2,12−14]

In recent years, optimization in thin-film technologies enables addition of new functionalities to surfaces which may assist in further stages or can even be used directly within application. As the simplest technique, the usage of the drop casting technique in terms of printing yields high quality deposition of inks for newspaper, organic electronic materials for solar cells,[15] and transistors.[16] Even personalized medication could become reality soon.[17−21] Spin coating, dip coating, and vacuum deposition enable similar modifications to surfaces. Further, these deposition techniques are capable of adjusting the polymorphic form as well as the morphology; e.g., three different polymorphs of DH6T can be obtained by process parameter variation.[22] Phenytoin can be processed so that at least nine different morphologies can be set without changing the polymorph,[23] and even new polymorphs were found by applying such thin-film techniques.[24−26] The dissolution studies on the new phenytoin polymorph even showed to be of faster nature compared to the other form which again demonstrates the importance of polymorph screening in pharmaceutical research in general.

Very often, solution processing of drug molecules enables achieving homogeneous thin films at a solid substrate surface. Depending on the processing conditions the films might be crystalline or amorphous. Employing faster processing like present in spin coating, amorphous phases frequently remain after solvent removal as found in samples of acetometaphene,[8] clotrimazole,[27−29] or phenytoin.[24,26,30] Having such an amorphous film, crystallization routes distinct from standard bulk solution processes can be tested. For instance, mechanical stressing using atomic force microscopy[30] might be employed for understanding crystallization in more detail. Another more common approach is temperature treatment:[26] higher temperatures favor molecular diffusion, facilitating nucleation, and finally resulting in crystallization. In a similar manner, solvent vapor annealing (SVA) is an excellent tool for changing the environmental parameters for crystal growth processes.[24,28,31] Here, a sample is exposed to a solvent vapor of any chemical composition, similar to isothermal calorimetry experiments employing defined humidity levels and dynamic vapor sorption experiments. In the vapor phase solvent molecules interact with the molecules in the drug film which eventually facilitates specific crystallization into a defined polymorphic form (or morphology). Hereby the solvent–vapor interaction strength might be estimated using the Hansen-solubility parameters.[28] Using such a solvent vapor annealing treatment, the second polymorphic form of phenytoin was also accessible, while up to recently this polymorph was just accessible using drop casting under very defined conditions (temperature and solute concentration). SVA on clotrimazole thin films was unable to generate new polymorphs,[28] but strong changes in the morphology/crystal habit could be obtained. Having a chance to change only the habit (size and shape) provides the possibility to understand its role in the overall dissolution performance, i.e., how different facets or surface areas change the release from thin films.

In this work, it is demonstrated how the solvent vapor process can be utilized to deliver information on a potentially better sample state with faster dissolution properties. As the model substance here, carbamazepine (CBZ) is used. In general, carbamazepine is an anticonvulsant used to treat epileptic seizures and nerve pain such as trigeminal neuralgia. CBZ is a Biopharmaceutics Classification System Class II molecule, possessing low solubility and high permeability. The poor dissolution performance has most likely the biggest influence on the bioavailability, driving the search for ideal formulations providing enhanced properties and finally resulting in a higher bioavailability. Deriving a better understanding for handling this problem might remain one target in research as many (∼40%) new chemical entities identified in screening are prone to fail due to their low water solubility. As one of the best ways to improve this situation, polymorph screening studies are advised. A simple change in the molecular arrangement (each polymorph is distinct) hereby might strongly change the lattice free energy reflected in a better solubility.

CBZ is a wonderful material to study such effects in more detail as it is prototypical for drug molecules with its hydrogen bonding potential. Especially, the formation of multiple readily accessible polymorphic forms[32] makes it an interesting candidate for fundamental research, like structure prediction,[33,34] the screening process in general, or for this work. Fast processing of CBZ results in amorphous films of sufficient stability so that different film treatments can be tested, inducing deviating polymorphic forms and morphologies. Using grazing incidence X-ray diffraction[35,36] and optical microscopy the impacts of the various treatments on the solid state properties are elucidated and discussed. Dissolution experiments assist in the identification of candidates for possible applications of potentially high bioavailability.

Results

Layer Thickness Dependent Morphology

Spin coating carbamazepine solutions onto silicon wafers initially results in the formation of dry, homogeneous, and optically transparent layers which are amorphous. Upon storage at ambient conditions, such samples crystallize eventually. The initial CBZ concentration in solution has thereby a profound impact on the initial layer thickness. At the highest concentration employed in this study (16 mg/g), a film thickness of 157 nm was determined by X-ray reflectivity (data not shown). Halving the concentration value, so that a solution of 8 mg/g is used, the thickness reduces to 74 nm. Diluting the solution further the thickness reduces, whereby a minimum layer thickness of 3 nm was obtained at a concentration of 0.5 mg/g. The different thicknesses reflect the fact that the amount of CBZ for the crystallization process is different in each sample. In Figure , optical micrographs of various samples 48 h after thin-film fabrication are depicted. For the lowest concentration, i.e., the thinnest film, few individual crystals randomly distributed at the silica surface are found. The needle-like shapes are about 100 μm long and are some micrometers wide. As the layer thickness increases to 8 nm, the spatial density of the needles increases, while leaving the size (almost) unaffected. In addition, a few smaller, dot-like structures appear, most likely of crystalline character. Reaching 17 nm, some interconnection with respect to each other exists, typical for the initial state of spherulitic growth. Here the crystals are accompanied by some vacant areas. While the film after deposition constitutes a homogeneous dense layer, nucleation and further crystal growth come at the expense of smaller crystallites, leaving these areas depleted. This is generally referred to as Ostwald ripening.[37] At 38 nm thickness, extended spherulitic structures with several hundreds of micrometer diameter emerge.

Figure 1

Optical micrographs of crystallized CBZ thin films as a function of the layer thickness obtained from different solute concentration after storage under ambient conditions.

It takes an 8 mg/g CBZ–THF solution to achieve a thickness of 74 nm which can provide a crystalline layer covering the entire substrate surface. In this sample, the spherulitic character remains but shows significantly increased dimensions. Individual spherulite branches are absent as the individual branches pack densely. At 158 nm thickness, the most obvious difference is the change in color, which results from a change in the crystal thickness compared to the 74 nm sample. In the spherulite centers, the color is blueish, while in the other areas more orange/yellowish colors appear. Homogeneously spread, there are also some darker areas, which are likely the result of the dot-like structures noticed in thinner films being developed into more extended structures.

Solvent Vapor Annealing

For solvent vapor annealing and subsequent dissolution experiments, the thickest carbamazepine amorphous films were used (i.e., samples were prepared only from a 16 mg/g CBZ–THF solution). This way, a dense layer can (mostly) be maintained during solvent vapor treatment, and a larger carbamazepine fraction is available for diffraction and dissolution experiments. Depending on the treatment of the samples, drastic changes in the CBZ morphology occur. In Figure , optical micrographs of the various samples are summarized after their individual treatments. (For the sake of completeness the very same spots are also examined under crossed polarizers, and the data are provided in the Supporting Information section.) Hereby, the samples are ranked by their dissolution rate, starting from those of slowest release (EtOH vapor sample) to the highest, which is the sample only heat treated at 80 °C directly after sample preparation.

Figure 2

Various samples after heat treatment at different temperatures or being exposed to solvent vapor for 48 h. Sequence of images corresponds to their respective dissolution rate, ranging from lowest (top left) to highest (bottom right).

The different treatments of the samples results in the appearance of each sample being different, whereby similarities between individual samples exist. In all samples the crystalline CBZ coated the entire surface, and only in the sample just heat treated at 80 °C is the crystalline film disrupted; thus, some vacant areas exist. This sample shows clear spherulitic structures with distinct boundaries in between. In the disrupted area less densely packed spherulites exist. The xylene sample appears similarly homogeneous. A closer look under crossed polarizers weakly shows the presence of a Maltesian cross typical for spherulitic growth (see Supporting Information).

The sample treated at 45 °C and 75% RH as well as the samples exposed to IPA or THF vapor reveal two distinct types of spherulites, one appearing blueish in color and the other similar to the one observed in the xylene-treated sample and appears brownish. The sample exposed to EtOH vapor reveals the largest blueish spherulites of all samples investigated in this study, nearly covering the entire surface. Besides this, structures with an extension of even some micrometers are present. As these areas appear dark under crossed polarizers, they could be amorphous or consist of a crystal phase which is not birefringent. Similar structures also appear in most of the other samples but with their extension and frequency being smaller. The samples exposed to acetonitrile vapor or heat treated at 50 °C show brownish and blueish spherulites whereby the amount or frequency of these two structures seems to be vice versa.

Crystal Structure

In order to investigate the crystal structure/polymorphic form, grazing incidence X-ray diffraction (GIXD) measurements were performed. In such an experiment, the shallow incidence angle allows measuring diffraction from very thin films.[35,36,38] An exemplary map is shown in Figure a for the sample after acetonitrile vapor annealing. Having an area detector, this allows identifying Debye–Scherrer type ring structures at different separations from the beam center (at around pixels x = 764, y = 824). In general, concentric rings in the pattern result from crystals behaving like a powder without preferred orientation (texture) whereby each ring corresponds to a defined net plane distance. (In GIXD measurements, the rings are disrupted in the lower half by the presence of the substrate so that only ring segments can be measured.) In all samples, the mostly powder-like character prevailed. Only in some samples a weak preferred orientation exist, which means that some crystal contact planes occur slightly more frequently along a certain direction than others. Figure b shows an example, which depicts the results of a sample heat treated at 50 °C. In the middle of the image, located at pixels x = 764 and y = 640, higher intensities are detected. In a GIXD image, this region corresponds to information from packing nearly parallel to the sample surface; i.e., the corresponding net plane is parallel to the substrate surface. The observed textures found for the individual samples are summarized in Table . The two sharp peaks in the pattern at pixels x = 400, y = 580 and x = 1100, y = 580, respectively, stem from the substrate; the silicon wafer is a 001 cut, making these peaks the 111 reflections diffracting into these very positions.

Figure 3

Grazing incidence X-ray diffraction maps of the carbamazepine film after solvent vapor annealing using acetonitrile (a) and another film treated at 50 °C (b). Both patterns share a common horizontal axis for clarity.

Table 1

Summary of the Samples, Their Polymorphic Forms, Identified Texture, and the Rate Parameter Determined from the Dissolution Profiles

	Form
sample	I	II	III	IV	hydrate	texture	a (rate param. eq 1)
powder			X			none	0.02
EtOH				X	X	none	0.11
45 °C – 75% RH	X				X	002 (hyd)	0.13
IPA	X				X	none	0.14
THF		X			solvate	110 (II)	0.21
xylene	X	X				none	0.42
ambient			X	X		510 (IV)	0.42
acetonitrile	X	X				none	0.44
50 °C	X					140 (I)	0.48
80 °C			X			none	1.86

For the sake of polymorph identification, these 2-dimensional data are integrated to yield a representation of intensity as a function of the absolute scattering vector q (information on the scattering direction is lost, and only information on the d-spacing is extractable). Often this is referred to as a powder plot. These integrated diffraction curves are provided in Figure . Independent of the sample, all curves contain peaks showing that each preparation route induced crystallization in the films which might already be expected from the microscopy images. At small scattering vectors the spectra contain a high diffraction background which reduces for larger scattering vectors. Such background behavior is due to air scattering from the primary beam, causing X-rays to bypass the beam stop. This behavior is identical for all samples and does not disturb the diffraction signal from the crystalline fraction of the samples, thus being of no further importance for the data analysis. On top of this background, distinct Bragg peaks are visible. From their position, the polymorphic form(s) of the respective sample is deduced by comparison with theoretical peak positions. The naming convention of the individual polymorphs here follows a previous report[32] whereby the list of the individual unit-cell parameters is provided in the SI.

Figure 4

Powder plots extracted from the grazing incidence X-ray diffraction patterns. Data are shifted for clarity. Order of appearance compared to dissolution results and microscopy images.

For example, using the example of the AN-treated sample of Figure a, the peak at q = 0.35 Å–1 (d spacing 17.95 Å) is explained by the 2–10 peak of CBZ form II. The peak at q = 0.67 Å–1 (d spacing 9.38 Å) results from the 0–12 reflection of the CBZ form I structure, whereas the peak at 0.62 Å–1 (d spacing 10.13 Å) is again explained by form II being the 300 reflection. This means this sample consists of two forms, i.e., forms I and II. In a similar fashion, the CBZ phases of the different samples are identified. The evaluations of all curves from Figure are summarized in Table .

The results reveal that many of the samples consist of multiple polymorphs. Only heat treatment at higher temperature (50 and 80 °C) results in solely forms I–III being present. Form III is also found for the as-delivered powder, i.e., the polymorph which is provided by the supplier (see SI and Table ). Further, the formation of solvates is noted on the exposure of EtOH or IPA or after storage at elevated humidity. Characteristic of carbamazepine form II are, in most cases, empty channels, in which, however, THF can be incorporated.[39,40] In terms of lattice parameters, this structure is very similar to the standard form II, which is why they cannot be differentiated from the experimental data.

Dissolution

As an estimate for the bioavailability of CBZ from these different samples, dissolution studies were performed. For the sake of comparability, each sample was measured under identical conditions. Starting with the sample stored at 40 °C and high humidity of 75%, the release of CBZ over time increases as the time progresses (see blue curve in Figure a). After 10 min, nearly 75% of the material is released from the silica surface. Another 15 min is required, i.e., a total time of 25 min, to dissolve all of the CBZ from the sample surface into the surrounding media. In comparison to that, storing the sample at ambient conditions or at an elevated temperature of 50 °C results in a quicker CBZ release so that already after 4 min about 75% of the entire drug amount is released. After this the ambient stored sample is slightly faster in its release, while the 50 °C one seems to have a reduced dissolution rate so that it takes around 30 min for the entire material to be dissolved. The sample heat treated at 80 °C releases most of the drug within 3 min, from which it can be assumed that this sample shows an immediate release.

Figure 5

Carbamazepine release as a function of time for samples treated under the presence of only water (a) and under different organic solvent vapors (b). Both diagrams share a common abscissa.

Using different organic solvents for the vapor treatment, the dissolution profiles change. EtOH vapor-treated samples show rather slow drug release, taking about 10 min to release 70% and another 35 min (total 45 min) to achieve 100% release. By using IPA or THF for the SVA process the initial CBZ release can be increased, but for all of these samples the maximum is reached only after 30 min. It can be expected that 100% release is achieved slightly faster than for the EtOH sample, but the low amount of data points at this time frame does not allow clarification. Treating samples with xylene or AN results in very fast releases of nearly 90% in 5 min but still slower than the sample treated at 80 °C.

To gain numbers for comparison, a regression fit to the dissolution data was performed. A lot of different mathematical descriptions aim to provide some understanding of the dissolution behavior.[41] As only one layer is at hand and also sink conditions are chosen, the situation is sufficiently described by a homogeneous model (often referred to as first-order release). The formula used for the evaluation is given bywhereby the concentration C as a function of time t is equal to the maximum concentration (Cs) minus an exponential decay over time with the exponent representing the fastness of drug release. For the evaluation of our samples, Cs is kept at 100% for all samples so that the only fitting parameter is a. The values for the various samples are tabulated in Table , ranging from 0.11 for the sample exposed to the EtOH vapor, representing a slow dissolution, up to 1.86 for the sample heat treated at 80 °C which showed the fastest dissolution.

A dissolution experiment was also performed for the pristine CBZ powder as obtained from the supplier. Hereby, some of the powder was introduced into a vessel containing the same dissolution medium as used for the thin-film samples, and the increase in CBZ concentration over time was determined. Similar to the other samples, the amount of material released (dissolved) increased with progressing time. The drug release was about 5 times slower compared to the EtOH sample which was the sample with the slowest CBZ release from a thin-film surface after treatment. An X-ray diffraction experiment shows that the pristine powder sample obtained from the supplier was of pure form III (data provided in the SI).

Discussion

CBZ possesses a rather complex phase behavior with a lot of different polymorphic forms resulting from sample preparation using varying experimental conditions. Within this study, the anhydrous forms I, II, III, and IV were found in different compositions. Indications of formation of form V were absent. Form V might be accessible from amorphous thin films of CBZ using different solvents, different temperatures, or combinations thereof or even a transformation from one of the other polymorphic phases. As the method used is very versatile, expanding these experiments is simple in order to explore additional forms.

Using experimental parameters that are easily accessible by tuning the temperature and the water content in the surrounding, our approach is directly adaptable for any manufacturing or drug formulation. In many areas, especially in pharmaceutical manufacturing, the usage of organic solvents should be limited or is even permitted. However, in an experiment, where harm to the environment or living organism can be prevented, such screening processes can be simply extended to more and/or even toxic solvents which might increase the chance of finding more (new) polymorphs or even leading to solvates. Following the procedures described in this work, only some of the amorphous films could be transferred into the crystalline state containing only a single polymorphic form. Likely, more optimization can provide monomorphic films. Recently it was demonstrated that SVA on crystalline samples can also induce changes of a crystalline polymorphic form, which might be associated with a solid–solid transition.[31] A temperature treatment step can also help to reduce the amount of one species on account of the other(s) if this one is less stable. Combinations of different solvents within one exposure step might provide a sufficient environment to achieve more specific crystallization, but as such adjustments correspond more to process optimizations rather than a screening process this has not been of interest here.

Spin coating is very versatile in means of drug layer thickness, so that the carbamazepine amount can be adjusted easily. In this study, the focus was put on film thicknesses in which the crystalline CBZ starts covering the entire sample surface. As noticeable in Figure , thinner films still provide information on the CBZ crystallization, but in these cases, the solvent–API–substrate interactions are more complicated. Large areas of the substrate surface are exposed, which is often referred to as dewetting or depletion of the circumjacent area of a growing crystal, which can cause the appearance of different morphologies. In our spin-coated samples a maximum amount of CBZ of about 100 μg would be achievable. Above this amount, crystallization is often initiated rapidly on account of being close to saturation concentration, leaving no time for the SVA process to be performed on amorphous CBZ. For most drugs this small amount is likely too low for therapy which might reduce the usage of spin coating in an upscale process. Employing other thin-film technologies like drop casting (dropping the solution onto a substrate followed by evaporation of the solvent), dip coating (withdrawal of a substrate from a solid solution under defined velocities), or a simple vacuum deposition process,[24,30]the amount of CBZ on the surface might be strongly increased. As far as drop casting is concerned, much more material can be deposited, but the process velocity is limited which also limits amorphous layer formation. Dip coating, on the other hand, can be utilized to grow crystals of defined textures, which can tune the dissolution properties further. Recently also jet spraying of materials was demonstrated to achieve amorphous layers containing a large amount of drugs.[42,43]

Drug dissolution is complex and according to the Noyes–Whitney[44] equation depends on the diffusion coefficient, boundary layer thickness, maximum solubility, and the drug concentration in the surrounding and the surface area. In fact, all our samples could be described by a first-order release, reflecting the fact that the underlying dissolution mechanism is, to the level of accuracy within our experiment, unaffected. Many experiments demonstrate that the polymorphic form impacts the dissolution behavior, as the different lattice and surface energies, among others, suggest that an altered interaction with the drug film takes place. In our experiments, the sample heat treated at 80 °C resulted in spherulitic structures homogeneously distributed over the entire surface, whereby only form III could be identified. The dissolution of this sample proceeded rapidly and was in fact the fastest of all samples investigated. Especially, the comparison with the as-obtained powder shows a 2 orders of magnitude faster release. The reason for this is the surface area being larger; a widespread thin film has a much bigger surface compared to some large powder particles even though it contains the same amount of material. Also, the microscopic roughness is very likely much larger compared to bulk grown crystals, and a lot of pores might exist in spin-coated samples which in addition enhance the surface area significantly. Having now a tool at hand that enables spreading drugs over large areas, this is another well-suited application for identifying potential limitations in the drug release on account of size reduction.

There are several treatments found which cause amorphous–crystalline transitions which then result in the drug release being very similar. This involves storing the sample at 50 °C or at ambient conditions and exposure to AN or xylene vapors. Surprisingly, the sample stored at ambient conditions has a completely different composition in terms of polymorphic form: forms III and IV are identified. In the other three samples, clearly form I exists, while also some form II is found in the AN and xylene sample. From the microscopy images, the samples appear different, at least in the amount of the individual structures being present. This would allow concluding that the change in the polymorphic form might change the dissolution properties, but this is counteracted by changes in the morphology. Likely, the deviation on the microscale in terms of surface area or even in the boundary layer formation can explain their similar dissolution profiles. Nevertheless, the experiment shows that four different routes are at hand that can provide samples of very comparable dissolution performance, so that fabrication optimization is not limited to a specific route. Nevertheless, the dissolution performance is worse when compared to the sample heated to 80 °C. While their surfaces seem to be similarly coated, the morphology might differ at the microscopic scale. At 80 °C more energy for diffusion is at hand which often allows growing structures in directions independent of the substrate surface which can even lead to crystal growth perpendicular to the substrate surface rather than parallel to it. Such a behavior would then lead to the solid state of the 80 °C sample being more porous; i.e., it has a lower density which provides easier access for the dissolution medium, and thus faster dissolution is observable.

The sample exposed to THF vapor shows a somehow intermediate release rate with the rate constant a being around 0.2 which is smaller than the one previously discussed (∼0.4–1.8). The X-ray investigation shows that this sample contains form II of CBZ, but also some of the crystals are solvates made from THF and CBZ. While this finding is very interesting in a screening process and for fundamental understanding, the usage of this very vapor might not be justified in any application as harm to the patient might occur.

There are several samples which contain hydrates besides anhydrous CBZ phases. In general, hydrate forms are expected to be less soluble in an aqueous dissolution media as they already contain water, and thus the free energy released when starting to interact with water is less for hydrates compared to anhydrate forms. This behavior is also clearly noticeable for samples investigated here, with samples containing hydrate forms having a values (from eq ) of about half compared to those of the anhydrates. This is in agreement with previous findings of dissolution rate reduction of CBZ on the incorporation of H2O into the crystal structure.[45]

Grazing incident X-ray diffraction is an excellent tool as it provides information on very thin films in terms of polymorphic forms but also on the texture. Texture results in crystals contacting the substrate surface with preferred crystal planes. This might be due to the processing conditions or due to the interaction of the substrate with the drug, inducing specific molecular arrangements during crystallization. In this study, only some of the samples revealed preferred orientations; e.g., form I in the 50 °C treated sample reveals a preferable contact with the 140 plane. Most of the other CBZ forms show arbitrary contact with silica surfaces in accordance with other recent reports. Likely, using substrates of stronger surface–CBZ interaction might allow changing this behavior as in the case of CBZ on top of crystalline iminostilbene templates.[46] As a preferred crystal orientation results in specific facets being in contact with the dissolution medium, this might be used to reach a desired drug release profile.

Conclusion

Having only nine samples prepared, the screening using solvent vapor annealing experiments is very effective in the finding of polymorphs. Here we found four different anhydrous forms of the five CBZ forms known, hydrate formation, and even a solvate. The finding in this work suggests that by using SVA processes an understanding in the polymorphic landscape of a given material can be established quickly and with a very small set of samples. While the deconvolution of the crystal morphology, polymorphic form, and texture might remain difficult, especially for this complex set of samples containing CBZ, possible routes for a much faster drug release can already be identified using information from this approach.

Experimental Methods

Carbamazepine (CBZ) was purchased from AGFA Pharma and used without further treatment. Acetronitrile (AN), 96% ethanol (EtOH), isopropanol (IPA), tetrahydrofuran (THF), and xylene were purchased from different suppliers in spectroscopic grade and used as received. Solutions of carbamazepine were prepared in THF and stirred prior to usage. Polished silicon wafers with a native silicon oxide layer (silica) were purchased from Siegert Wafers (Germany) and cut into 2.5 × 2.5 cm2 pieces. Prior to usage, the substrates were cleaned in acetone and ethanol and finally rinsed with Milli-Q water.

Sample preparation was performed in a two-step process. First, samples were spin coated onto a piece of silicon wafer (see Figure a). For this, a drop of approximately 100 μL of CBZ solution was placed onto the substrate followed by a continuous sample rotation around its surface normal at a rotation speed of 17 rps for 30 s. This results in the formation of a dry, homogeneous, and amorphous drug layer, as confirmed by X-ray diffraction experiments. Subsequently, selected samples were kept under ambient conditions and others at 50 or 80 °C for 48 h. Another set of samples was exposed to vapors of different chemical composition (see Figure b), typically referred to as solvent vapor annealing (SVA). In this, the amorphous samples are enclosed in a desiccator together with an excess amount of solvent. As the solvent vessel is not sealed against the sample, a saturated vapor develops within the desiccator which then interacts with the sample. For our purpose, all vapors are formed at ambient conditions (∼1 atm, 22 °C) and either consist of AN, EtOH, IPA, THF, or xylene, respectively. Each sample was exposed to just one solvent vapor. After 48 h the respective sample was stored at ambient conditions until further experiments were performed.

Figure 6

Scheme of sample preparation, depicting the preparation of amorphous carbamazepine films (a) and the subsequent solvent vapor annealing step (b) transferring the films to the crystalline state.

Morphological investigations were performed on an Olympus BX51 microscope equipped with polarizers, and images were taken with a standard digital camera in reflection mode. Crystallographic information was obtained by X-ray diffraction under nearly grazing incidence conditions (GIXD), where also thin layers as used in this study provide sufficient diffraction signal to obtain structural information (using standard θ/2θ scans, such samples provide only little or even no diffracted intensities due to small diffraction volumes).[35] The measurements were performed at the XRD1 beamline[47] at the Elettra synchrotron, Trieste (Italy). Data were collected for a wavelength of 1.4 Å using a Pilatus-2 M detector from DECTRIS (Switzerland) by integrating for 2 min. Data acquisition took place while rotating the sample around the axes perpendicular to the film surface by 90° while maintaining constant grazing conditions to improve counting statistics. Due do the detector construction, detector gaps exits so that two images per sample needed to be recorded; the second image was taken at slightly elevated detector position. During data processing using the in-house developed software GIDVis, data from these images are merged and transferred from pixel space to reciprocal space using standard procedures.[35,36] To identify the individual polymorphs, diffracted intensity data were integrated along constant q values (∼1/d-spacing), yielding a graphical representation analogous to classic powder patterns. The data are compared to the calculated powder patterns of polymorphs with fully known crystal structures (Cambridge crystal structure database identifier: CBMZPN03, 11, 12, 14 and 16)[32] using the software packages Mercury,[48] PowderCell,[49] and GIDVis.

In vitro dissolution testing was performed in 20 mL of Milli-Q water. As the amount of CBZ on each surface was low, standard dissolution apparatuses like an USP Apparatus 2 are improper. As only relative differences are of interest, a custom-made setup was used as introduced previously.[25,26] For this, a sample was placed in a glass container filled with the dissolution media and gently shaken at 100 rpm at room temperature. An amount of 1 mL of the dissolution media was withdrawn at predetermined times for UV–vis absorption measurements using a NanoPhotometer from Implen Gmbh, Germany, and fed back after measurement leaving the dissolution medium volume constant over the run of the experiment. The dissolved CBZ amount was determined at 211 nm using standard Quartz cuvettes. Each data point is an arithmetic mean of three identical samples/measurements. Error bars are omitted for sake of clarity.