Introduction

Our interest in the solid-state chemistry of methoxybenzoic acids, also known as anisic acids, Fig. 1, comes from an ongoing project dealing with pharmaceutical co-crystals. The best co-formers for the synthesis of these new solid forms, which are of interest in the pharmaceutical and food industries, must be selected. Anisic acids are excellent candidates. These compounds occur naturally in Pimpinella anisum and are used as flavouring agents, according to the European Food Safety Authority [1] and the World Health Organization [2]. The most widely employed isomer is p-anisic acid, which is also used as a preservative of fragrance in cosmetic formulations [3–6]. Some thermodynamic data are available for the three isomers [7], and p-anisic acid is recommended as an organic reference material for temperature calibration in differential scanning calorimetry [8]. Crystalline structures of o-anisic acid [9] and of p-anisic acid [10] are already available in literature. Very recent data, however, with a CCDC number superior to ours, could be found for m-anisic acid [11] but without any reference to polymorphism.

Fig. 1
figure 1

Anisic acid isomers

Full size image

It is known that approximately one-third of organic compounds exhibit polymorphism [12], “Polymorphism is a solid crystalline phase of a given compound resulting from the possibility of at least two crystalline arrangements of the molecules of that compound in the solid state” [13]. Conformationally flexible molecules have more degrees of freedom than rigid molecules, such that a greater scope for polymorphism might be expected with them. m-Anisic acid is a small molecule that has some degree of conformational flexibility. This property sometimes leads to conformational polymorphism, the occurrence of different conformers in distinct polymorphic structural modifications or more rarely to conformational isomorphism, the occurrence of different conformers in the same crystal structure [14–16].

Being a GRAS flavouring substance [17], m-anisic acid aroused our attention as an interesting potential co-former in co-crystallization studies. Based on the absence in the literature of resolved crystal structures, thermal behaviour studies, spectroscopic data or computational studies on this system, research on the polymorphism of m-anisic acid is likely to be relevant both when it is used alone and also for a correct co-crystal synthesis interpretation. In this paper, an investigation of m-anisic acid polymorphism is undertaken using thermal analysis (DSC and PLTM), infrared spectroscopy, X-ray diffraction and DFT calculations.

Materials and methods

Materials

3-Methoxybenzoic acid was acquired from Sigma-Aldrich (m-anisic acid, 99 %, lot 13920MC-435A). The commercialized compound was slightly brownish and was purified by sublimation at 75 °C under reduced pressure (P = 103 Pa). The resulting compound was white and crystalline. Samples were also prepared by crystallization from solutions and by grinding. In crystallization experiments, typically 10–20 mg of m-anisic acid was used in 2 mL of analytical grade solvents. After filtration onto a Petri dish, the solvent was evaporated, either at room temperature or at 2 °C. A Retsch MM400 mill with a 10 mL stainless steel grinding jar and two 7 mm diameter stainless steel balls per jar was used for grinding. A total mass of about 50 mg was ground for 30 min at a frequency of 15 Hz.

Polarized light thermal microscopy (PLTM)

The solids were characterized by PLTM using a Linkam hot stage system, model DSC600, with a Leica DMRB microscope and a Sony CCD-IRIS/RGB video camera. A Real-Time Video Measurement System software from Linkam was used for image analysis. The images were obtained by combined use of polarized light and wave compensators, using 200× magnification.

Differential scanning calorimetry (DSC)

The studies were performed on a PerkinElmer Pyris1 power compensation calorimeter with an intracooler cooling unit at −25 °C [ethylene glycol/water, 1:1 (v/v), cooling mixture]. The samples, mass ~1.5 mg, were hermetically sealed in 30 μL aluminium pans, and an empty pan was used as reference. A 20-mL min−1 nitrogen purge was employed. Temperature calibration [18, 19] was performed with the high-grade standards biphenyl (CRM LGC 2610, T fus = 68.93(3) °C, benzoic acid (CRM LGC 2606, T fus = 122.35(2) °C), indium (Perkin-Elmer, x = 99.99 %, T fus = 156.60 °C) and caffeine (Mettler Toledo calibration substance, ME 18 872, T fus = 235.6(2) °C). Enthalpy calibration was performed with indium Δfus H = 3286(13) J mol−1 [18].

Infrared spectroscopy (FTIR)

Spectra of the solids were recorded at room temperature with the KBr pellet technique using a ThermoNicolet IR300 FTIR spectrometer, resolution 1 cm−1, 32 scans and a PerkinElmer Spectrum 400 FTIR/ATR diamond/zinc selenide plate, resolution 1 cm−1, 32 scans.

Single-crystal X-ray diffraction (SXD)

A crystal of the title compound obtained by crystallization in water solution at room temperature, with a plate habit and having approximate dimensions of 0.38 mm × 0.18 mm × 0.08 mm, was glued on a glass fibre and mounted on a Bruker Apex II diffractometer. Diffraction data were collected at room temperature 293(2) K using graphite monochromated Mo Kα radiation (λ = 0.71073 Å). Data reduction was performed with APEX II [20]. Lorentz and polarization corrections were applied. Absorption correction was applied using SADABS [21].

The crystal structure was solved with direct methods using SHELXS-97 programme [22] and refined on F 2 by full-matrix least squares with SHELXL-97 programme [22]. The anisotropic displacement parameters for non-hydrogen atoms were applied. Carboxylic hydrogen atoms were located in a difference Fourier map, and their coordinates were refined with U iso(H) = 1.5U eq(O) and the O–H bond lengths restrained to 0.82 Å. The other hydrogen atoms were placed at calculated positions and refined with isotropic parameters as riding atoms using SHELXL-97 defaults. The final least-squares cycle was based on 3521 observed reflections [I > 2 σ(I)], 207 variable parameters, converged with R = 0.0395 and wR = 0.1049. Supplementary data have been deposited at the Cambridge Crystallographic Data Centre (CCDC No. 985352).

X-ray powder diffraction (XRPD)

The single crystals were powdered thoroughly using a pestle and mortar to prepare a polycrystalline sample. The sample powder was sifted with a 63 μm sieve, and glass capillaries (0.5 mm diameter) were filled with the powdered specimens. The samples were mounted on an ENRAF-NONIUS powder diffractometer (equipped with a CPS120 detector from INEL), and data were collected for 15 min to 1 h using Debye–Scherrer geometry. Monochromatized Cu K α1 radiation was used (λ = 1.5406 Å). Silicon was chosen as an external calibrant. Samples were heated/cooled by a hot/cold nitrogen gas stream (Oxford Cryosystems, series 600) at 6 °C min−1. One data collection of 48 h at room temperature was also performed for the Rietveld refinement.

Ab initio calculations

The geometry optimizations were performed using the Firefly QC package [23], which is partially based on the GAMESS (US) source code [24], starting from the experimental X-ray geometries of molecules A and B. The calculations were performed within density functional theory (DFT) using Becke three-parameter Lee–Yang–Parr (B3LYP) for exchange and correlation, which combines the hybrid exchange functional of Becke [25, 26] with the correlation functional of Lee, Yang and Parr [27]. The calculations were performed with the Pople’s ‘triple split’ 6-311G(d,p) basis set, which includes a set of p-polarization functions for the H atoms and a set of d-polarization functions for the C and O atoms. Each self-consistent field calculation was iterated until a ∆ρ of less than 10–5 bohr–3 was achieved. The final equilibrium geometries at the minimum energy had a maximum gradient in internal coordinates of 10–5 hartree bohr–1 or hartree rad–1. At the end of these geometry optimizations, Hessian calculations were carried out to guarantee that the final structures correspond to true minima using the same level of theory as in the geometry optimizations. DFT calculations have also been performed of the single point energy for the experimental X-ray geometries of the two independent molecules A and B, using the method described above (B3LYP functional, 6-311G(d,p) basis set), to determine the energy difference between the two geometries in the crystal.

Results and discussion

Crystal structure of polymorph I

Good quality single crystals allowed the determination of a crystalline structure by single-crystal X-ray diffraction, hereafter identified as polymorph I (Table 1). During the preparation of our manuscript, another group determined almost simultaneously (CCDC No. 985637) the crystal structure of polymorph I [11]. Within experimental error, the two structures are equal.

Table 1 Crystal data and structure refinement parameters for m-anisic acid polymorph I
Full size table

Polymorph I crystallizes in the space group P21/n, with two molecules (A and B) in the asymmetric unit (Fig. 2). In 2000, Steiner [28] indicated the existence of systems with Z′ > 1 as a relatively rare occurrence for organic molecules (10.8 % of CSD structures out of 2000). A recent inspection to the CCDC database [CSD version 5.34 updates (Nov 2012)] shows the existence of 56950 compounds (organic and organometallic), with Z′ > 1, out of 630919 compounds (9.0 %). The origin of this phenomenon (the presence of multiple molecules in the asymmetric unit) is still a matter of much discussion within the scientific community [29, 30]. The molecules are essentially planar, with the carboxyl and the methoxy groups almost in the plane of the benzene ring. The maximum deviations from the least-squares planes of molecules A and B are 0.029(1) Å for the O1A atom and 0.040(1) Å for the C8B atom. Selected structural parameters can be seen in Table 2.

Fig. 2
figure 2

Hydrogen-bonded dimer of symmetry-independent molecules A and B of m-anisic acid polymorph I. Ellipsoids are drawn at the 50 % probability level

Full size image
Fig. 3
figure 3

Packing diagram of m-anisic acid polymorph I viewed along the b-axis. H bonds are drawn as dashed lines

Full size image
Table 2 Comparison of selected geometrical parameters for m-anisic acid polymorph I as determined by X-ray diffraction and from DFT geometry optimizations for the symmetry-independent molecules A and B
Full size table

The bond lengths in the carboxyl group are similar, suggesting some charge delocalization between the two C7–O bonds. This may indicate a small degree of rotational disorder in the carboxyl group, a situation that is common in carboxylic acids. However, the residual charges located near the O1 atoms are small (about 0.16 e Å−3), and for this reason, it was not considered necessary to refine a disorder model.

The orientation of the methoxy group is quite different in the two symmetry-independent molecules. Considering Newman projections along the C3–O3 bonds, in molecule A atom C8 eclipses ring atom C2 (ϕC2–C3–O3–C8 = −1.4(2)°), whereas in molecule B, it eclipses the ring atom C4 (ϕC2–C3–O3–C8 = −176.75(11)°).

The symmetry-independent molecules A and B are hydrogen bonded in pairs through the carboxyl groups (see Fig. 3; Table 3), forming rings of descriptor R 22 (8) according to Etter’s graph-set theory [31, 32]. The molecules in these dimers are almost coplanar. The crystal structure is build up from dimers in two different orientations which are approximately perpendicular to each other, with the angle between the least-squares planes being 83.8° (see Fig. 4). The crystal packing is also stabilized by a weak C–H⋅⋅⋅π interaction where the C4A–H4A bond points towards the edge of the aromatic ring of molecule B, with a H⋅⋅⋅π i distance of 3.60 Å and a C–H⋅⋅⋅π angle of 149° [symmetry code: (i) x − 1, y, z].

Table 3 Hydrogen bonds parameters of m-anisic acid polymorph I
Full size table
Fig. 4
figure 4

A view down the a axis of the unit cell of m-anisic acid polymorph I with the molecules coloured by symmetry equivalence: green molecule A, blue molecule B. (Color figure online)

Full size image

Fingerprint plots

Hirshfeld surfaces and fingerprint plots [33, 34] are powerful tools for visualizing and exploring intermolecular interactions. This is important when the same molecule is in different crystalline environments, as in the case of polymorphism of molecular materials or of crystal structures with more than one molecule in the asymmetric unit (Z′ > 1).

The intermolecular interactions of the two independent molecules were analysed by the two-dimensional fingerprint plots [34] derived from Hirshfeld surfaces [33], using CrystalExplorer version 3.1 [35]. 2D-fingerprint plots were generated from the d i and d e pairs measured on each individual spot of the calculated Hirshfeld surface. The two-dimensional fingerprint plots for the independent molecules A and B are presented in Fig. 5. The fingerprint plots have the same basic features, with both showing the presence of a pair of long sharp spikes, a characteristic of strong hydrogen bonds, in this case the O–H⋅⋅⋅O hydrogen bonds. The two polymorphs feature a diffuse region of blue points between the hydrogen bond spikes, which is characteristic of the cyclic hydrogen bond dimer motif.

Fig. 5
figure 5

2D fingerprint plots for the independent molecules A and B in the crystalline structure of m-anisic acid polymorph I

Full size image
Fig. 6
figure 6

Illustrative DSC curves of the first and second heating runs of m-anisic acid (m = 1.51 mg, β = 10 °C min−1)

Full size image

The main difference between the two fingerprint plots is the relative spread of data points at high d i and d e in molecule B, which is related to the different orientation of the methoxy group, which points away from the ring in this structure.

Ab initio calculations

To get some insight on the influence of the intermolecular interactions on the molecular geometries, DFT calculations of the equilibrium geometries were performed for molecules A and B in vacuum. These DFT calculations closely reproduce the solid-state geometry of the two molecules (Table 2). The agreement between the experimental and calculated bond lengths and valence angles is very good, except for the C7–O bond lengths where the differences may be explained by the charge delocalization due to a small degree of rotational disorder in the carboxyl group, as mentioned above.

The energy difference between molecules A and B is just 4.90 kJ mol−1, with the B conformation being the more stable. This energy difference is comparable to the energies involved in rotations about single bonds [36]. Intermolecular interactions can easily account for this relative low conformational energy difference since hydrogen bond strengths are generally estimated to be in the range of 2–63 kJ mol−1 [37]. The molecular geometric parameters most likely to be affected by these interactions are the torsion angles around single bonds, as is the case of these two conformers.

Evidence of a new polymorphic form

Thermal behaviour

Polymorph I, acquired commercially and purified by sublimation, was studied in heating/cooling cycles by DSC, Fig. 6 and Table 4. In the first heating run, polymorph I melts at 105.4 °C, and no other events are observed. A supercooling value of about 25 °C, typical for organic compounds [38], is registered on cooling at a scanning rate of 10 °C min−1. The resulting solid submitted to another heating run melts at 94.0 °C—melt crystallization gives rise to a new solid form, polymorph II.

Table 4 Thermodynamic parameters from DSC curves obtained on heating and cooling runs of m-anisic acid
Full size table

After melt crystallization in a DSC pan, a second heating run performed 15 days after shows a small peak of form I, and 6 months later shows that about 10 % of polymorph II was transformed into polymorph I, while 2 years later a complete transformation was observed, Fig. 7.

Fig. 7
figure 7

DSC heating curves showing the ageing effect on polymorph II of m-anisic acid

Full size image

The effect of milling on the solid samples is an essential step in the green synthesis of co-crystals as it may induce polymorphic transitions. The knowledge of the occurrence of these transitions in the pure components is very important for the correct interpretation of co-crystal synthesis data. For m-anisic acid, the milling process (mechanical milling or even using a pestle and mortar) induces the polymorphic transition of polymorph II into polymorph I.

X-ray powder diffraction

A Rietveld refinement [39] was performed with Fullprof software [40] using a powder diffractogram collected at room temperature over a 48-h period. The overall parameters, such as cell parameters, 2θ zero, scale factor, full-width at half-maximum and asymmetry parameter, were allowed to refine in the range of 5°–45°, with a final Bragg reliability factor of 13.5 %. The agreement between experimental and calculated patterns (see Fig. 8) confirms that the material obtained by crystallization in water solution at room temperature consists only of polymorph I.

Fig. 8
figure 8

Results from a least-squares fit using the Rietveld method: experimental powder diffraction pattern (green); simulated diffraction pattern of polymorph I using the Rietveld method (dark-blue); difference between the observed and calculated intensities (red) and corresponding Bragg positions (blue). (Color figure online)

Full size image

After this confirmation of the polymorph composition of the powdered sample, we conducted XRPD heating/cooling experiments. A first diffractogram was collected at 20 °C for 1 h (Fig. 9). The capillary was then heated (β ~ 6 °C min−1) until the melting of the sample at around 110 °C. After melting, the sample was cooled to 75 °C at a rate β ~ 6 °C min−1, and XRPD data collection was conducted for 30 min at this temperature (Fig. 9). The sample was further cooled until 50 °C at the same rate, and data collection was performed at this temperature for 15 min. The diffractogram was very similar to the one obtained at 75 °C. Although the heating rates for XRPD and DSC are different, the diffractogram shown at 75 °C provides evidence of a new solid form, which we ascribe to polymorph II.

Fig. 9
figure 9

Powder X-ray diffractograms of m-anisic acid polymorph I (blue crystallized in water at room temperature collected at 20 °C) and II (red cooled from the melt to 75 °C). (Color figure online)

Full size image

Fourier transform infrared spectroscopy

As milling, even for KBr pellet preparation, induces polymorphic transition from form II to form I, a less aggressive technique was used, FTIR/ATR, to obtain the infrared spectra of both polymorphs (see Fig. 9).

As expected in the infrared spectra of polymorphic forms, some spectral differences in the intensity and position of the bands are observed, particularly in those bands associated with functional groups involved in hydrogen bonds.

The spectroscopic and theoretical studies of m-anisic acid polymorph I [41, 42] allow the assignment of some selected bands of polymorph II identified in Fig. 10 by their wavenumbers. These are presented in Table 5.

Fig. 10
figure 10

FTIR/ATR spectra of m-anisic acid polymorphic forms I and II

Full size image
Table 5 m-Anisic acid polymorphs I and II: assignment of infrared bands identified in Fig. 10
Full size table

Several empirical equations relate the enthalpy, ΔH, of hydrogen bonds to the shift of the bands relatively to the free molecule band frequency values [43–45]. In m-anisic acid, the blue shift of the hydroxyl group out of plane bending mode, γ(OH), is a good indicator of the strength of the hydrogen bonds. Applying the Rozenberg equation (−ΔH = 0.67 × 10−4 × [(γ(ΟΗ)H)2 − (γ(ΟΗ)0)2]) [45], the value −35 kJ mol−1 is obtained for polymorph I and −33 kJ mol-1 for polymorph II. These are in agreement with the presence of strong hydrogen bonds in both polymorphs. In these calculations, γ(ΟΗ)0 was taken as 596 cm−1. This frequency was obtained from ab initio calculations performed on the isolated B molecule. The calculated vibrational frequencies were scaled by the recommended factor of 0.9613 [46].

According to the empirical rules developed by Burger and Ramberger [47, 48], it is possible to classify this system as a monotropic one. We note that polymorph II melts at T fus = 94.0 °C with enthalpy value of Δfus H = 19.8 kJ mol−1, these values are smaller than those of polymorph I that melts at T fus = 105.4 °C with enthalpy of Δfus H = 23.2 kJ mol-1. Infrared spectra show that the hydrogen bond network of polymorph I is stronger than that of polymorph II. Finally, the conversion of polymorph II into polymorph I after some elapsed time reveals that the most stable form at room temperature is polymorph I.

Solid samples obtained by crystallization from solutions

Crystallization from several solvents (water, ethanol, ethyl acetate, n-hexane) induced by solvent evaporation under different temperature conditions (room temperature and 2 °C) and also by vapour diffusion (n-hexane into ethanol solution) was carried out. The resulting solids were analysed by DSC and PLTM for the identification of the polymorphic form, and the results are presented in Table 6. Although the best crystals were analysed by SXD, all the single crystals of polymorph II were too thin for a data collection.

Table 6 m-Anisic acid solid forms obtained in the crystallization experiments
Full size table

Considering the thermal analysis results, special attention should be made to the DSC curve obtained for the sample crystallized from ethanol at room temperature, Fig. 11. The presence of polymorph II concomitantly with a small amount of polymorph I leads to this DSC curve where melting of polymorph II is followed by a small crystallization event and afterwards by melting of polymorph I. A second heating cycle of these samples shows the normal melting of polymorph II.

Fig. 11
figure 11

DSC curves of the first and second heating runs of m-anisic acid crystalized in ethanol solution at room temperature (m = 1.22 mg, β = 10 °C min-1)

Full size image

Further information was obtained by PLTM, a good technique to follow polymorphic behaviour [49–52]. Polymorph I (crystals obtained in water at room temperature) and II (crystals obtained in ethanol at room temperature) were joined in the same microscopic field, and their thermal behaviour was observed, Fig. 12.

Fig. 12
figure 12

PLTM images (a) and DSC curve (b) of heating run of a m-anisic acid mixture of polymorph I and II obtained by crystallization in water and ethanol, respectively. (β = 10 °C min−1, ×200 and m = 1.20 mg)

Full size image

The images reveal melting of polymorph II at 95 °C and melting of polymorph I at 105 °C without observation of crystallization of the liquid. This experiment reveals that the crystallization process of the liquid in contact with solid form I is kinetically rather slow, even though it is thermodynamically favourable under these conditions. However, in DSC conditions for the same sample, melting of polymorph I is the major event observed.

Conclusions

An investigation on the polymorphism of m-anisic acid, a GRAS flavouring substance, was undertaken. The presence of two conformers in polymorph I unit cell is a rare case of conformational isomorphism. A new solid form, polymorph II, was also identified and characterized. Thermal analysis data, infrared spectra and time evolution allow ascribing a monotropic relationship between polymorphs I and II. Polymorph I is the stable form under ambient conditions.

Supplementary material

Crystallographic data for the structure reported in this paper have been deposited at the Cambridge Crystallographic Data Center number, CCDC 985352. Copies of this information may be obtained free of charge on application to CCDC, 12 Union Road, Cambridge CB2 1EZ, UK (fax: +44 1223 336 033; e-mail: deposit@ccdc.cam.ac.uk or http://www.ccdc.cam.ac.uk).