Scalability of Pharmaceutical Co-Crystal Formation by Mechanochemistry in Batch

Graphical Abstract

Upscaling the mechanochemical synthesis of co-crystals of active pharmaceutical ingredients (APIs) is a sustainable approach for more environmentally friendly production pathways. Several kilograms of co-crystals could be synthesized in industrial eccentric vibrating mills with a very high conversion rate.

Abstract

The development of mechanochemistry is considerably growing. Benign by design, this technology complies with several principles of green chemistry, contributing to the achievement of the United Nations Sustainable Development Goals (UN SDGs) and the European Green Deal objectives. Herein, we report the use of mechanochemical processes in batch to prepare kilogram-scale of the Active Pharmaceutical Ingredient (API): Ibuprofen-Nicotinamide (rac-IBP:NCT) co-crystal in an industrial eccentric vibration mill. This scenario shows a sustainable approach to the industrial up-scaling of pharmaceutical co-crystals by a solvent-free mechanochemical process in batch. The quantitative assessment of the greenness of the mechanochemical process against the Twelve Principles of Green Chemistry was performed using the DOZN 2.0 Green Chemistry Evaluator.

Introduction

One of the major problems in the preparation of co-crystals in solution is the different solubility of the solid blends.¹ Mechanochemical methods allow to circumvent this issue, allowing to access co-crystals otherwise impossible to be obtained.^2-7 Despite their broad application, mechanochemical methods for co-crystal synthesis are mainly limited to small-scale preparation. The scalability of co-crystal preparation in the absence of solvents is a valuable approach to disclose the full potential of mechanochemistry as a key enabling technology to sustainable industrial synthesis.^8-11 In this regard, the greenness of mechanochemical processes to prepare an Active Pharmaceutical Ingredient (API) (both in batch and in continuous) was quantitatively assessed following the Twelve Principles of Green Chemistry,¹² while Life Cycle Assessment (LCA) studies demonstrated its low environmental impact (e. g. waste economy and a reduction of CO₂ emissions up to 85 %) and a shrinking of the costs estimated to 12 % less.¹³

We previously reported the mechanochemical preparation in batch mode¹⁴ (on multigram-scale by ball-milling) and continuously¹⁵ (with a space time yield of 6.8×10³ kg m⁻³ day⁻¹ by twin screw extrusion), of several APIs, also including World Health Organisation (WHO) essential medicines,¹⁶ such as the antiepileptic phenytoin^{17, 18} the antibacterial agent nitrofurantoin,^{15, 19, 20} the analgesic paracetamol²¹ and the antibiotic silver sulfadiazine.²²

However, around 90 % of preclinical compounds APIs in the development pipeline are poorly soluble and present bioavailability challenges.^{23, 24} Therefore, the industrial interest in the preparation of pharmaceutical co-crystals as marketable drugs is growing, together with the need of sustainable methods for their clean manufacturing, to fulfill the United Nations Sustainable Development Goals (UN SDGs) and the European Green Deal objectives. For these reasons, the preparation of pharmaceutical materials, including co-crystals, by mechanochemical methods is a valuable and sustainable approach,^{4, 8} also considering some of the peculiarities of this powder-based technology, allowing: i) scalability,^{8, 10, 11, 25, 26} ii) generally higher productivity compared to solution based procedures, and iii) to circumvent issues related to solubility, sometimes encountered in solution synthesis and crystallisation, allowing to access compounds otherwise impossible to be obtained,^{2-7, 27} or not displaying the suitable (e. g. tabletability) and biopharmaceutical (e. g. dissolution rate) properties.

In the case of Ibuprofen (IBU), an analgesic drug listed as WHO essential medicine,¹⁶ displaying poor aqueous solubility, bioavailability,²⁸ and thermal lability,^{29, 30} which affect its therapeutical efficiency. Co-crystallisation with the WHO essential medicine¹⁶ and food additive nicotinamide (NIC), a highly water-soluble form of vitamin B3 (niacin), modified the physicochemical properties of IBU but not its therapeutic effect, leading to a IBU:NIC co-crystal with a solubility 7.5 times higher than IBU.³¹

The preparation of IBU:NIC in batch by grinding (e. g. manually in a mortar with pestle),^{32, 33} and by ball-milling (neat^34-36 or liquid-assisted grinding conditions⁷), and continuous by twin-screw extrusion³⁷ (TSE) or hot melt extrusion^38-40 (HME) were previously reported. The mechanochemical co-crystallisation by extrusion,⁴¹ delivered IBU:NIC co-crystal in amounts comprised between 0.025 and 0.2 kg h⁻¹, with better results obtained upon heating at 90 °C with a screw-speed of 20–50 rpm. However, batch mechanochemical methods were limited to laboratory scale preparation delivering IBU:NIC co-crystal in limited amount up to 1.0 g (after 40 min at 50 Hz using stainless steel as grinding media).³⁴ Therefore, despite the growing interest in continuous flow operations, the vast majority of fine chemical, pharmaceutical and biopharmaceutical industries operate in batch,⁴² not only for synthesis but also on the formulation side (e. g. for powder mixing, granulation, coating etc.). Therefore, the development of large-scale batch mode for mechanochemical processes is still an appealing alternative to continuous flow mechanochemistry. Moreover, a major challenge in the production of co-crystals of APIs is the transfer of processes in the gram-scale from laboratory preparations to the kg-scale for commercial production.

Herein, the unprecedented kg-scale synthesis of the API rac-Ibuprofen-nicotinamide (rac-IBU:NIC) co-crystals by mechanochemistry in batch is reported in an industrial eccentric vibratory mill,⁴³ paving the way for an unprecedented approach to the mechanochemical preparation of pharmaceutical materials in batch, by retooling and retrofitting an equipment already used for the industrial processing of materials other than pharmaceuticals. The manufacturing of rac-IBU:NIC occurred without generation of waste and with improved green metrics and reduced environmental impact, assessed by DOZN 2.0 tool,^{12, 44} in comparison with large-scale processes by solvent-free in continuous by HME.^{39, 45}

Results and Discussion

Screening of process parameters for the laboratory-scale synthesis of rac-IBU:NIC. (S)-ibuprofen is the active isomer in the body,⁴⁶ however, the pharmaceutical product presents both forms. Therefore, several ibuprofen co-crystals were previously prepared and characterized, including rac-IBU:NIC, for which the single-crystal structure was also reported,^{32, 47} showing the formation of an heterosynthon between the carboxylic group of the ibuprofen and the pyridine ring of nicotinamide.

The preparation of rac-IBU:NIC on a gram scale was investigated by grinding an equimolar amount of IBU and NIC under different conditions, providing different mechanical stresses (Tables 1 and 2). Indeed, the mechanochemical reaction was carried out in mixer (with horizontal or vertical oscillation), shaker (undergoing angular harmonic displacement in the vertical plane simultaneously with rotation in the equatorial plane),⁴⁸ and planetary ball mills, to compare their performances and to find the optimum process conditions. The influence of milling time (up to 2 h), the milling speed (up to 50 Hz or 800 rpm), the nature of the grinding media (i. e. stainless steel or zirconium oxide), the volume size (up to 500 mL) and geometry of the milling jars (cylindrical or egg-shape), the size and number of milling balls (up to 24 with variable diameters) were studied with the purpose to understand the influence of mechanochemical process parameters to drive the co-crystal formation to full completion.

Table 1. rac-IBU:NIC batch synthesis in vibrating or shaker ball mills in different process conditions and at different scales.^[a]

Entry	Milling device	Grinding Media/Jar volume (mL)	Speed (Hz)	Time (min)	Recovery (%)
1	mixer^[b]	SS^[c]/22	23	15	n.d.^[d]
2				30	n.d.^[d]
3				60	–
4		ZrO2^[e]/10	30	60	92
5	mixer^[f]	ZrO2^[f]/10	30	60	93
6			50	30	88
7		SS^[g]/10	30	60	87
8			50	30	94
9	shaker^[h]	ZrO2^[h]/55	14	60	63

• [a] General reaction conditions: 400 mg of an equimolar amount of rac-IBU (251 mg, 1.22 mmol) and NIC (149 mg, 1.22 mmol) were ball-milled in the specified conditions using 2 balls (Ø 10 mm), except when otherwise specified; [b] An horizontal mill with screw closure cylindric-shape jars was used; [c] SS=Stainless Steel, m_ball=4.07 g, m_tot=8.15 g [d] n.d.=not determined, the conversion was not complete; [e] m_ball=3.59 g, m_tot=7.18 g [f] A vertical mill with snap closure egg-shape jars was used, m_ball=3.03 g, m_tot=6.06 g; [g] SS=Stainless Steel, m_ball=4.01 g, m_tot=8.02 g; [h] The reaction was performed in a SPEX 8000 shaker-mill on 6.09 mmol scale using equimolar amount of rac-IBU (1.257 g) and NIC (0.744 mg) and 10 balls (Ø 10 mm, m_ball=3.0 g, m_tot=30.0 g).

Table 2. rac-IBP:NIC batch synthesis in planetary ball mills, in different process conditions and at different scales.^[a]

Entry	Milling device	Grinding Media/Jar volume (mL)	Speed (rpm)	Time (min)	Recovery (%)
1	P7^[a]	SS^[b,c]/12	400	15	n.d.^[d]
2			400	60	n.d.^[d]
3			600	60	n.d.^[d]
4			800	120	n.d.^[d]
5		ZrO2^[e]/12	600	60	n.d.^[d]
6	P7P^[f]	ZrO2^[f]/45	800	120	71
7	P5^[g]	SS^[b,g]/500	300	30	–

• [a] General reaction conditions: 400 mg of an equimolar amount of rac-IBU (251 mg, 1.22 mmol) and NIC (149 mg, 1.22 mmol) were ball-milled in a planetary ball mill, in the specified conditions using 2 balls (Ø 10 mm), except when otherwise specified. P7, P7P and P5 indicate different planetary milling devices used. [b] SS=Stainless Steel; [c] m_ball=4.07 g, m_tot=8.15 g; [d] n.d.=not determined, the conversion was not complete; [e] m_ball=3.03 g, m_tot=6.06 g; [f] The reaction was performed on 6.09 mmol scale using equimolar amount of rac-IBU (1257 mg) and NIC (744 mg) and 10 balls (Ø 10 mm, m_ball=3.03 g, m_tot=30.3 g); [g] The reaction scale was 0.152 mol. 50 g of an equimolar amount of rac-IBU (31.4 g) and NIC (18.6 g) were ball-milled with 24 balls (Ø 20 mm, m_ball=31.5 g, m_tot=757 g).

In a first set of experiments (Table 1), the kinetics of the mechanochemical reaction were studied during 15, 30 and 60 min in a vibrating (horizontal) ball-mill using stainless steel (entries 1–3 and Figures S1 and S2, ESI), or zirconium oxide grinding media (entry 4), by milling 400 mg of equimolar amount of rac-IBU and NIC. rac-IBU:NIC was formed already after 15 min, however, the full conversion of the reactants was obtained after 60 min milling (entries 3 and 4), as demonstrated by monitoring ex situ the mechanochemical reaction by thermal analyses by Differential Scanning Calorimetry (DSC) and by powder X-ray diffraction (PXRD) measurements (Figures S1 and S2, ESI). Indeed, DSC thermograms recorded after 15 min reaction, showed two endothermic peaks, at 76.5 °C and 90 °C, identified respectively as the melting of precursor rac-IBU and the rac-IBU:NIC co-crystal. The signal corresponding to the melting of nicotinamide (expected at 131.3 °C, as determined by measuring its melting point by both DSC method and in the capillary), was absent, due to the solubilization of nicotinamide in molten ibuprofen (formation of a peritectic)⁴⁹ (Figure S1, ESI). After 60 min milling, the pure co-crystal rac-IBU:NIC was formed, as shown by DSC thermogram presenting a single and characteristic signal of melting endotherm at 90 °C^{7, 35} (Figure S1, ESI). PXRD experiments and thermal analysis confirmed the purity of the rac-IBU:NIC co-crystal prepared by mechanochemistry, and easily recovered as a homogeneous white powder by scratching it out of the jar with a spatula.

In a comparative experiment (Table 1, entries 3 vs 4), it was demonstrated that the reaction outcome was independent on the milling load, the milling media and the milling speed, the only parameter influencing the reaction being the milling time. Indeed, when performing the reaction at 30 Hz, using egg-shape jars (entries 4 and 7), providing a different type of friction/shearing/compression forces with respect to cylindrical shape ones (entry 3 vs entries 4 and 7), and independently on the milling media used, the results were comparable and the reaction was completed within 1 h (entries 4 and 7). In contrast, the reaction was completed within 30 min by increasing the degree of mixing at a milling speed at 50 Hz (entries 5 and 8). Worth of note is that the co-crystal was quantitatively obtained after 1 h milling, even when processing a much larger amounts of powder (ca. 2.0 g) at a lower milling frequency of 14 Hz (entry 9) in a shaker mill, the last one characterized by an elliptical movement of the jar. However, in this case, the product was stuck to the wall of the jar, and the amount of rac-IBU:NIC recovered was lower (63 %) than the other cases.

The DSC thermograms recorded for each rac-IBU:NIC co-crystal obtained in the milling conditions reported in Table 1, showed identical thermal profiles, and excellent recovery percentages in most of the cases (Table 1). Nevertheless, the results reported in Table 1 seemed to suggest that the successful outcome of rac-IBU:NIC mechanochemical synthesis is mainly dependent on the stress frequency⁵⁰ that can be modulated by a suitable combination of operational frequency (14, 23, 30 or 50 Hz) and ball numbers (2 or 10, 10 mm diameter) (Table 1). PXRD analyses also showed that, whatever the milling conditions were, the same crystalline solids (no amorphization) were always obtained and the formation of IBU:IBU, and/or NIC:NIC dimers⁵¹ sometimes observed by HME due to external energy input such as heating, was never observed.

With standard laboratory mixer and shaker mills up to 2.0 g rac-IBU:NIC co-crystal can be synthesized. Nevertheless, in view of scalability purposes on kg-scale, the mechanochemical reaction was also investigated in planetary ball-mills equipped with jars having the size to process higher amounts of powders (Table 2) compared to the most common laboratory mixer and shaker mills (Table 1).

Preliminary experiments performed by ball milling equimolar amount (400 mg) of rac-IBU and NIC in stainless steel or zirconium oxide jars, at different reaction times and milling speed (400, 600 and 800 rpm), always showed traces of reagents, as demonstrated by DSC analyses of the powders (Table 2, entries 1–5). Better results were obtained in reaction conditions with increased stress frequency (Table 2, entries 6 and 7), independently of the milling speed and the milling media, leading to pure rac-IBU:NIC co-crystal, confirmed by DSC and PXRD analyses. Worth to note is the unprecedented preparation of 50 g of pure and crystalline rac-IBU:NIC in only 30 min at 300 rpm (Table 2, entry 7, Figure S3a and S3b, ESI).

Therefore, the direct upscaling to kg amount of rac-IBU:NIC in an eccentric vibratory mill⁴³ was expected to be promising.

Kg-scale mechanochemical synthesis of rac-IBU:NIC in an eccentric vibrating mill.

Industrially used for mineral processing (e. g. for fine grinding and pulverisation of raw materials) and metallurgical industry, in the chemical and pigment market sector and for recycling purposes (e. g. to mill cemented carbides), eccentric vibratory mills use and unexpressed potential for the preparation of pharmaceutical materials is still unexplored.

Unlike conventional vibrating or planetary ball mills, eccentric vibratory mills cause elliptical, circular and linear vibrations, usually leading to throughput increase by a factor 2, because of a substantial increase of the acceleration of the grinding media.⁴³

Therefore, the mechanochemical kg-scale synthesis of rac-IBU:NIC co-crystal was carried out in industrial eccentric vibration mills (Figure S4 and video, ESI), providing an all-in-one combination of the mechanical stress obtained separately by three different specific equipment (mixer, shaker and planetary mills). Therefore, equimolar amount of rac-IBU and NIC were ball-milled in two different eccentric vibratory miIls (Figure S4, ESI), having identical horizontal (11–12 mm) and vertical (7–8 mm) vibration widths (Table 3) but differed in the type of milling vessel: in stainless steel (entry 1) or aluminium oxide (entry 2).

Table 3. Kg-scale mechanochemical synthesis of rac-IBP:NIC co-crystal in eccentric vibratory mills, in different process conditions.^[a]

Entry	Milling device^[b]	Milling Media	mball/ mtot (g)/(kg)	input/ output (kg)	Recovery (%)^[c]
1	ESM-236-1bs	SS^[d]	260/76^[e]	2.8^[f]/2.63^[c]	94
2	ESM-234-1bk	Al2O3	77/15^[g]	1.4^[h]/1.14^[c]	81

• [a] General reaction conditions: equimolar amount of rac-IBU and NIC were ball-milled during 150 min in an eccentric vibratory mill filled with grinding balls with a diameter of 40 mm, in the specified process conditions; [b] For the type of milling devices, please see Figure S4, ESI; [c] not-optimised data; [d] SS=Stainless Steel; [e] the total volume occupied by stainless steel grinding balls was ca. 15 L; the free volume in the grinding vessel was 5 L; [f] the reaction scale was 8.517 mol, input (2.8 kg): rac-IBU (1.757 kg) and NIC (1.043 kg), output: 2.63 kg; [g] the total volume occupied by Al₂O₃ grinding balls was ca. 6.5 L; the free volume in the grinding vessel was 2.5 L; [h] the reaction scale was 4.261 mol, input (1.4 kg): rac-IBU (0.879 kg) and NIC (0.521 kg), output : 1.14 kg.

The two experiments were conducted with a stoichiometric 1 : 1 ibuprofen/nicotinamide mixture, corresponding to a total amount of powder (input) of 2.8 kg (entry 1) or 1.4 kg (entry 2) respectively. The milling process proceeded for 150 min and both reactions were monitored by ex situ PXRD (Figure 1) and DSC (Figure 2) analyses. Samples were collected while pausing the process at 15, 60, 120 and 150 min, each. Full conversion of the starting materials was achieved with an output of 2.63 kg (entry 1, for stainless steel vessel) and 1.14 kg (entry 2, for Al₂O₃ vessel) respectively of pure rac-IBU:NIC co-crystal.

Figure 1

Normalized X-ray diffraction ratio (PXRD) pattern following the rac-IBU:NIC co-crystal formation in eccentric vibration mills (black: Al₂O₃, blue: steel) and the respective simulated patterns from single-crystals structure of starting materials (red).

Figure 2

Normalized DSC measurements following the rac-IBU:NIC co-crystal (m.p. 90 °C) formation in eccentric vibration mills (black: Al₂O₃, blue: steel) and the respective starting materials IBU (m.p. 75–77 °C) and NIC (m.p. 132 °C) (red).

The X-ray diffraction patterns in Figure 1 show that in both cases most of the 1 : 1 rac-ibuprofen/nicotinamide mixture was converted into the co-crystal, which can be seen from the intense and characteristic diffraction peak of the co-crystal at 3.1° (2θ),^{37, 47, 49, 52} clearly displaying the existence of the new phase. When using stainless vessel, the diffraction peak at 6.1° (2θ)^{37, 49, 52} associated with pure rac-ibuprofen is still present after 15 min milling, but it becomes smaller with increasing milling time (at 60 min), until it is completely gone and could not be observed after 120 min. Similarly, no significant diffraction intensities of nicotinamide at 14.8°, 25.3°, 25.7° and 27.2° (2θ) could be observed after 120 min. However, the mechanochemical reaction carried out using Al₂O₃ as grinding media was slightly slower compared to the process making use of stainless steel vessel. Even if Al₂O₃ is a harder material than stainless steel, its lower density (3950 kg m⁻³) compared to stainless steel (7500–8000 kg m⁻³⁾, can be responsible for this outcome. Based on the relationship of kinetic energy and mass, it is expected that a body with lower mass but equal volume should transfer a smaller amount of energy during a single collision. Accordingly, the lower density of Al₂O₃ compared to steel could be responsible for the prolonged conversion. Indeed, after 120 min, a very weak diffraction peak of ibuprofen can still be detected at 6.1° (2θ), but this signal is completely gone after 150 min.

Regardless of the milling conditions used (Table 3), no further crystalline or non-crystalline (e. g. ibuprofen dimers⁵¹) forms were detected other than the rac-IBU:NIC co-crystal, obtained in quantitative yield, at room temperature. In this regard, the mechanochemical process in batch is more advantageous that its continuous flow counterpart by HME,^{37, 39} for which the full conversion of the starting materials requires: i) processing the mixture at 90 °C (barrel temperature), ii) needing both high degree of mixing (modulated by the screw geometries), and high shearing (modulated by the screw speed), and iii) the use of a liquid binder (usually water), to mitigate the formation of ibuprofen dimers.⁵¹

The better reactivity observed in batch compared to HME could be explained by an increased energetic mechanical stress applied to the powders, combining simultaneously intense stress energy by mixing, high shearing and energetic impacts (due to the intrinsic characteristic of eccentric vibratory mills),⁴³ with high stress frequency and specific process conditions designed (e. g. milling speed and time, milling balls filling degree, number and diameter of milling balls, choice of the milling media, geometry and volume of the vessel, etc.).^{53, 54}

It is worth to note that no amorphization occurred during the milling process, overcoming one of the main challenges for the large scale preparation of pharmaceutical materials by mechanochemical procedures.^{38, 55} The rac-IBU:NIC co-crystal was identical to that one obtained from solution or other solvent-free methodologies, as evidenced by PXRD diffractograms^{47, 49, 52} and by solid-state NMR analyses.⁵⁶ The ¹³C CP MAS NMR spectra of the reactants as well as the rac-IBU:NIC co-crystal from the steel mill shown in Figure 3 are in good agreement with spectra from the literature,⁵⁶ thus demonstrating that a pure crystalline product can be generated via scale up in eccentric vibrator mills.

Figure 3

¹³C CP MAS NMR spectra of rac-IBU:NIC co-crystal synthesized in eccentric vibration mill (steel) and the respective starting materials rac-IBU and NIC (*=spinning side bands).

Moreover, in contrast with solution-based methods, the preparation by milling presents the additional advantage of not requiring any purification procedure (e. g. filtration), with a high (and not optimized) recovery percentage (Table 3) of up to 94 % without any further optimization of the whole process, where the recovery was determined by the ratio of input to output mass. With respect to the here reported single batch experiments, the considered losses in recovery would be compensated by thorough cleaning procedures or simple reuse of the preconditioned vessels in series of batch syntheses.

Therefore, the results above indicate that the 1 : 1 rac-ibuprofen/nicotinamide mixtures of 2.8 kg (steel) and of 1.4 kg (Al₂O₃) were converted into the corresponding co-crystal after only 120–150 min of milling (Table 3). Thermal analyses by DSC (Figure 2) of both experiments support these observations and conclusions, as the endothermic peaks (melting points) displayed for Ibuprofen, nicotinamide and the 1 : 1 co-crystal at 75–77 °C, 130 °C and 90 °C respectively.

An aspect often neglected when preparing pharmaceutical materials by mechanochemical methods is related to the analyses of residual metal impurities due to abrasion/leaching of the milling media during the process. To the best of our knowledge, only one report quantified the concentration of residual metals after ball-milling in a stainless-steel jar by ICP-MS.⁵⁷ Therefore, considering the stringent quality regulations for manufacturing pharmaceuticals, and in the perspective of giving an additional insightful approach towards a future implementation of mechanochemical processes at industrial level for preparing pharmaceuticals,^{8, 10, 11, 15} the samples prepared in process conditions detailed in Table 3 were analyzed by ICP-OES. In addition, the starting materials were analyzed as a reference to the processed samples, and the obtained results are listed in Table 4.

Table 4. Metal impurities determined by ICP-OES after ball milling according to Table 3.

Sample	Al (ppm)	Cr (ppm)	Fe (ppm)	Ni (ppm)
IBU	13.1	5.8	44.6	11.7
NIC	11.3	5.4	39.0	8.0
Al2O3 150^[a]	95.8	6.9	95.0	9.7
SS 60^[b]	129.7	25.9	679.8	16.1
SS 120^[c]	119.3	33.0	772.2	19.0

• [a] Sample was processed for 150 min with reaction parameters according Entry 2 Table 3. [b] Sample was processed for 60 min with reaction parameters according Entry 1 Table 3. [c] Sample was processed for 120 min with reaction parameters according Entry 1 Table 3.

It can be stated that traces of the elements investigated (Al, Cr, Fe and Ni) can already be found in the starting materials. However, the results of the processed samples clearly show that abrasion of the grinding media has taken place, with the determined amounts being in the lower ppm range and the contamination of the organic products being far below the limits of the daily permitted load. Assuming that the co-crystal daily dose is 2400 mg, this would be equivalent to a nickel daily dose of close to 24 μg and thus below maximum permitted daily dose of 220 μg/day for oral administration prescribed by the ICH Guidelines.^{58, 59} It should be emphasized that the abrasion of the stainless steel grinding media is significantly higher than that of the samples processed with alumina grinding media. This effect is due to the higher wear resistance of alumina. Thus, the use of alumina as grinding media brings two effects to play. The lower density of alumina compared to stainless steel reduces the conversion rate, as seen in the DSC experiments. At the same time, however, much lower contamination of the organic products was observed due to the higher abrasion resistance. These results highlight once again that the mechanochemical method shows a high potential for industrial scale, even taking into account the strict regulations on metal impurities in pharmaceutical materials.

Assessment of the greenness of the mechanochemical process in batch and comparison with other processes

The use of green metrics applied to mechanochemical processes is quite recent.⁶⁰ In particular, mechanochemistry is intertwined with the 12 Green Chemistry Principles, highlighting their close connection,⁸ which makes mechanochemical methodologies advantageous over conventional synthetic routes based on the use of solvents.⁶¹ However, the 12 principles are only conceptual and do not provide a quantitative framework. While various approaches to quantifying greener processes and products have been proposed, there is no unifying set of metrics in place.⁶⁰ We previously reported the use of the free, web-based scoring matrix DOZN^TM 2.0 tool^{44, 62} providing a unified set of metrics to quantitatively assess the greenness of mechanochemical processes against the 12 Principles of Green Chemistry, applied to the preparation of the World Essential Medicine (WHO) nitrofurantoin¹² and unsymmetrical N-aryl-5,6-unsubstituted-1,4-dihydropyridines (DHPs).⁶³ In the case of co-crystal preparation, to the best of our knowledge, green chemistry metrics were never calculated, despite that their preparation intrinsically complies with the second principle of Green Chemistry, accounting for 100 % atom economy,⁶⁴ being all the atoms of the two reagents found into the final product.

Based on DOZN^TM 2.0 tool, the 12 Green Chemistry principles are organized into three major groups: i) Group 1 (G1, it includes principles # 1, 2, 7, 8, 9 and 11) accounting for improved resources, ii) Group 2 (G2, it includes only principle # 6) assesses the energy efficiency of the process, and iii) Group 3 (G3, for principles # 3, 4, 5, 10 and 12) which evaluates reduced human and environmental hazards. The individual greenness scores calculated for each principle are aggregated in a normalized value, referred to as aggregate score, accounting of the greenness of the process on a scale of 0–100, with 0 being the most desirable.

Therefore, DOZN^TM 2.0 greenness scores were generated for the synthesis of rac-IBU:NIC by ball milling process in batch (by eccentric vibratory mill) in comparison with continuous flow mechanochemical process by HME upon heating^{39, 45} (Table 5).

Table 5. Comparative scores for the preparation of rac-IBU:NIC in batch (by eccentric vibratory mill) vs. in continuous (by HME).^{39, 45}

Entry	Solvent-free	Reactor	T(°C)/t	Input (kg)/	DOZN 2.0 Scores
	Process	Media	(min)	Output (kg)	Improved resource use	Increased energy efficiency	Reduced human and environmental hazards
					G1	G2	G3	Aggregate
1^[a]	Batch	SS^[b]	r.t^[c]/120	2.80/2.60	0.28	0.00	0.51	0.0158
2^45	Continuous	SS^[b]	80/120^[d]	0.72^[e]/0.67^[f]	0.29	0.08	0.51	0.0176
3^39	Continuous	SS^[b]	90/120^[d]	0.10^[g]/0.094^[f]	0.28	0.15	0.51	0.0188

• [a] Reaction in an eccentric vibratory ball-mill, the general reaction conditions are given in Table 3; [b] SS=stainless-steel; [c] r.t. (=room temperature) refers to the operational temperature of the mill; [d] To compare with the batch process in entry 1, it is assumed that the process operates in continuous during 120 min; [e] equimolar amount of rac-IBU (0.452 kg) and NIC (0.268 kg) were processes at a feed rate of 6 g min⁻¹ with a screw speed of 50 rpm; [f] To compare with the data reported in Table 3 (entry 1), it is assumed the same 94 % recovery of the rac-IBU:NIC co-crystal; [g] equimolar amount of rac-IBU (0.063 kg) and NIC (0.037 kg) were processes at a feed rate of 0.05 kg h⁻¹ with a screw speed of 30 rpm.

All the three processes display aggregate scores below 1 (Charts C1-C3, ESI): the scores for Group 1 and 3 were practically identical, because they use the same reactants in equimolar amounts of raw materials rac-IBU and NIC. These scores quantify the amount of waste produced, the severity of waste and their level of hazards. The differences in scores arise in Group 2 (energy efficiency): the mechanochemical process in batch received a zero score due to no deviation from ambient conditions with regard to both temperature and pressure. Indeed, HME required some heating (Table 5, entries 2 and 3) to afford a full conversion of the raw materials into rac-IBU:NIC, thus generating a higher score for Group 2, and a linear increase of the aggregate scores with the increase of the process temperature. In this regard, the preparation of rac-IBU:NIC co-crystal by eccentric vibratory mill results greener and more economic in terms of electrical energy consumed, calculated to 1.815 kWh, with a release CO₂ calculated to 0.726 kg.⁶⁵

Conclusions

This study used various mechanochemical setups on a gram-scale for the synthesis of the pharmaceutical co-crystal rac-IBU:NIC, while evaluating the most influential parameters. This was followed by an up-scaling to industrial relevant kg-scale using an eccentric vibration mill and assessing the final purity with respect to the high standards of the pharmaceutical industry.

The number of patents where mechanochemistry is industrially used is rapidly growing,⁶⁶ witnessing an acceleration towards a more extensive industrial adoption, in market sectors where this technology results still unexploited, such as the pharmaceutical industry.^{8, 10, 11, 25}

Experimental Section

Materials: rac-Ibuprofen ((RS)-2-(4-(2-methylpropyl)phenyl)propanoic acid) was provided by BASF Company. Nicotinamide (99 %) was obtained from Alfa Aesar.

X-ray powder diffraction (PXRD). For samples in Table 1, entries 4–9, and Table 2, entry 6: PXRD patterns were collected on a PANalytical X'Pert Pro Automated diffractometer equipped with an X'celerator detector in Bragg-Brentano geometry, using Cu-Kα radiation (λ=1.5418 Å) without monochromator in 3–40° 2θ range (step size 0.033°; time/step: 20 s; Soller slit 0.04 rad, anti-scatter slit: , divergence slit: ; 40 mA*40 kV). For samples in Table 1, entries 1–3, Table 2, entry 7 and Table 3: data was measured on a STOE Stadi P diffractometer equipped with a sealed long-fine-focus Cu-tube running at 40 kV/40 mA, a short collimator, a curved Ge (111) monochromator to yield pure Cu−Kα1 (λ=1.54056 Å) radiation, and a linear PSD (position sensitive detector filled with CH_4/Ar (90 : 10). The sample was packed in a 0.7 mm borosilicate glass capillary and measured in the range of 2–70° (2θ) within 1 : 20 h. The intensity of obtained diffraction data was normalized to the max. peak for comparison and only selected data are shown.

Solid-State NMR (SS NMR). The solid-state ¹³C NMR experiments were performed on a Bruker Avance III HD 500WB spectrometer using a double-bearing 4 mm MAS probe (DVT BL4) at a resonance frequency of 125.8 MHz. The experimental conditions for ¹³C CP MAS NMR for rac-IBU:NIC co-crystal and rac-IBU were as follows: 10 kHz spinning rate, 5 s recycle delay, 4096 scans, 3 ms contact time, 4.4 μs ¹H π/2-pulse, and high-power proton decoupling (SPINAL-64). The ¹³C NMR spectra of NIC sample was recorded with following conditions: 10 kHz spinning rate, 120 s recycle delay, 512 scans, 3 ms contact time, 3. μs ¹H π/2-pulse, and high-power proton decoupling (SPINAL-64). The ¹³C chemical shift was referenced with respect to neat tetramethylsilane (TMS) in a separate rotor. All spectra were collected at 298 K.

Thermal analyses. DSC measurements (Table 1, entries 4–9, and Table 2, entry 6) were performed with a Netzsch DSC 200F3, at the heating rate of 2 °C min⁻¹. Samples (3–5 mg) were placed in hermetic aluminum pans. All the experiments were carried out in duplicate to ensure reproducibility of the experimental data. Yields refer to pure isolated materials. DSC measurements (Table 1, entries 1–3, Table 2, entry 7 and Table 3) were performed by using a Mettler Toledo TGA/DSC 1 with a gas controller GC 200 using an argon flow of 50 mL min⁻¹. Samples of 5–12 mg were placed in a 100 μl aluminium crucible and heated from 30 °C to 200 °C with a rate of 10 K min⁻¹. The intensity of obtained DSC data was normalized to the max. peak for comparison and only selected data are shown.

Melting points (Tables 1 and 2) were measured in an open capillary on a Stuart SMP50 automatic melting point apparatus to confirm the values measured by Differential Scanning Calorimetry (DSC).

Elemental analysis: Inductively coupled plasma-optical emission spectrometry (ICP-OES) measurements were performed with a SPECTROGREEN DSOI instrument equipped with UVPlus optics (165 – 770 nm and optimized Rowland Circle Alignment) and carried out in crossflow. The samples were dissolved in a solution of nitric acid (67 %; 2.5 ml) and hydrochloric acid (37 %; 7.5 ml) under an argon atmosphere using an Anton Paar Multiwave 5000. The samples were heated to 90 °C and then to 180 °C. After cooling to room temperature, the solution was transferred to tared 50 ml centrifuge tubes and made up to 50 ml with deionized water. As standard solutions, various volumes of commercial standards (1 g L⁻¹) of Berndkraft were mixed and filled up to 50 ml.

MM200 mixer mill with horizontal movement (Table 1, entries 1–4 and Figures S1 and S2, ESI): 400 mg of an equimolar amount of rac-IBU (251 mg, 1.22 mmol) and NIC (149 mg, 1.22 mmol) were ball-milled in: i) (entries 1–3) a~22 mL stainless steel jar with 2 balls (SS, Ø 10 mm, m_ball=4.07 g, m_tot=8.15 g) for 30 min. Then a sample for PXRD and DSC was taken and milling continued for another 30 min; or ii) (entry 4) in a 10 mL zirconium oxide jar with 2 balls (ZrO₂, Ø 10 mm, m_ball=3.59 g, m_tot=7.18 g) during 60 min. The rac-IBU:NIC co-crystal was recovered by scratching the white powder out of the jar.

Mixer-mill with vertical movement (Table 1, entries 5–8): 400 mg of an equimolar amount of rac-IBU (251 mg, 1.22 mmol) and NIC (149 mg, 1.22 mmol) were ball-milled in a 10 mL jar (SS or ZrO₂), with 2 balls (Ø 10 mm; for ZrO₂: m_ball=3.03 g, m_tot=6.06 g; for SS: m_ball=4.01 g, m_tot=8.02 g) at 30 Hz for 60 min (entries 5 and 7) or at 50 Hz for 30 min (entries 6 and 8). The rac-IBU:NIC co-crystal was recovered by scratching the white powder out of the jar.

SPEX 8000 shaker-mill with elliptical movement (Table 1, entry 9): 2.0 g of an equimolar amount of rac-IBU (1.257 g, 6.09 mmol) and NIC (0.744 mg, 6.09 mmol) were ball-milled in a 55 mL ZrO₂ jar with 10 balls (ZrO₂, Ø 10 mm, m_ball=3.0 g, m_tot=30.0 g) at 14 Hz for 60 min. The rac-IBU:NIC co-crystal was recovered by scratching the white powder out of the jar.

Fritsch P7P planetary mill (Table 2, entry 6): 2.0 g of an equimolar amount of rac-IBU (1.257 g, 6.09 mmol) and NIC (0.744 mg, 6.09 mmol) were ball-milled in a 45 mL ZrO₂ jar with 10 balls (ZrO₂, Ø 10 mm, m_ball=3.03 g, m_tot=30.3 g) at 800 rpm for 2 h. The rac-IBU:NIC co-crystal was recovered by scratching the white powder out of the jar.

Fritsch P5 planetary mill (Table 2, entry 7): 50 g of an equimolar amount of rac-ibuprofen (31.4 g, 0.152 mol) and nicotinamide (18.6 g, 0.152 mol) were ball-milled in a~500 mL hardened steel jar with 24 balls (hardened steel, Ø 20 mm, m_ball=31.5 g, m_tot=757 g) at 300 rpm for 15 min. Then a sample for PXRD and DSC was taken and milling continued for another 15 min to extract another sample taken and finally milled for additional 30 min to obtain the last sample. The rac-IBU:NIC co-crystal was recovered by scratching the white powder out of the jar.

Scale-up experiments with eccentric vibratory mill⁴³ (Table 3, Figure S4, ESI): Single batches of equimolar rac-ibuprofen/nicotinamide mixtures were premixed and milled for 150 min, while samples of 5–10 g were taken every 15 min leaving a break of ~3 min and the residue of the sampling output jack was transferred back to the Mill. a) (entry 1) input 2.8 kg: rac-ibuprofen (1.758 kg, 8.512 mol) and nicotinamide (1.756 kg) were milled in a Siebtechnik eccentric vibrating mill ESM 236–1bs (steel~15 L, 40 mm steel balls, ~1 L/76 kg balls, each ball~260 g, free volume~5 L) ESM; b) (entry 2) input 1.4 kg: rac-ibuprofen (0.879 kg, 4.256 mol) and nicotinamide (0.520 kg, 4.256 mol) were ball-milled in a ESM 234–1bk (ceramic Al₂O₃~6.5 L, 40 mm Al₂O₃ balls, ~6,5 L/15 kg balls, each ball~77 g, free volume~2.5 L). Recovery of the sample: The milling vessel has a perforated disk at the bottom, covered with a second disk hampering the solid to go out of the vessel, At the end of the milling process the second disc is removed and the sample is shaked/milled again during less than 5 min. The whole material falls down in any kind of container.

Further information on the eccentric vibrating mills can be obtained at http://www.siebtechnik-tema.com/eccentric-vibrating-mills/.

Acknowledgments

The authors would like to thank BASF Pharma Solutions and Midas Pharma for providing sufficient quantities of rac-ibuprofen and Siebtechnik Company (Mülheim an der Ruhr, Germany) for the opportunity to perform mechanochemical batch experiments on a kg scale. The authors would also like to thank Steffen Reichle for joining the milling session at Siebtechnik GmbH (Germany, http://www.siebtechnik-tema.com/eccentric-vibrating-mills/) and visual documentation. The authors thank Florian Baum for ICP-OES measurements. F.W. acknowledges funding from the International Max Planck Research School for Interface Controlled Materials for Energy Conversion (IMPRS-SurMat). E.C. is grateful to Julien Fullenwarth (Institut Charles Gerhardt de Montpellier, ICGM, France), for the access to SPEX shaker-mill 8000. E.C., A.P. and D.V. acknowledge Région Occitanie (France) for the Pre-Maturation 2020 – MECH-API grant (ESRPREMAT – 00262) and BetaInnov (www.beta-innov.com) for the gift of a planetary milling equipment.

This article is based upon work from COST Action CA18112 Mechanochemistry for Sustainable Industry,^[67–68] supported by COST (European Cooperation in Science and Technology).^[69] COST (European Cooperation in Science and Technology) is a funding agency for research and innovation networks, helping to connect research initiatives across Europe and enable scientists to grow their ideas by sharing them with their peers. This boosts their research, career and innovation. www.cost.eu. E.C., M.F., F.G., E.P. and D.V. acknowledge IMPACTIVE (Innovative Mechanochemical Processes to synthesize green ACTIVE pharmaceutical ingredients),^[70] the research project funded from the European Union's Horizon Europe research and innovation programme under grant agreement: No. 101057286. Open Access funding enabled and organized by Projekt DEAL.

Conflict of interests

There are no conflicts of interest to declare.

Open Research

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.