Using Milling to Explore Physical States: The Amorphous and Polymorphic Forms of Sulindac.

This article shows how milling can be used to explore the phase diagram of pharmaceuticals. This process has been applied to sulindac. A short milling has been found to trigger a polymorphic transformation between form II and form I upon heating which is not seen in the nonmilled material. This possibility was clearly demonstrated to result from crystalline microstrains induced by the mechanical shocks. A long milling has been found to induce a total amorphization of the material. Moreover, the amorphous fraction produced during milling appears to have a complex recrystallization upon heating which depends on the milling time. The investigations have been mainly performed by differential scanning calorimetry and powder X-ray diffraction.


INTRODUCTION
Mechanical milling of powders is a usual process used in the course of drug formulation to reduce the particle size. However this process may also change the physical state of the end product [1][2][3] , leading sometimes to an amorphization [4][5][6][7][8][9][10] and sometimes to a polymorphic transformation [11][12][13][14][15] . It appears, that amorphizations mainly occur when milling is performed below the glass transition temperature (Tg) of the material while polymorphic transformations mainly occur when milling is performed above Tg [13,16,17] . These transformations are expected to be governed by a competition between a mechanical disordering process and a thermally activated recrystallization.
This competition explain that most materials only amorphize when they are milled below Tg since it is in the glass transition temperature range that the molecular mobility -which drives the crystallization -decreases the most rapidly. In some cases, the milling does not induce any apparent structural changes, but the microstructural changes it induces (size of crystallites, crystalline defects and lattice distortions) can strongly modify the physical stability of the material. This can lead for instance to polymorphic transformations upon heating [18] which do not occur in the non-milled crystal. Transformations induced byor triggered bymilling have thus a strong repercussion on both the physical stability and the bioavailability of drugs [24] . That's why they have to be fully understood to be perfectly controlled [25,26] . Moreover, they can also be used in a positive way to investigate further the phase diagram (polymorphism and relative stability of polymorphs) of pharmaceutical materials.
Sulindac (C20 H17 F O3 S) is a nonsteroidal anti-inflammatory agent with a rich polymorphism. Three crystalline polymorphs were already reported: Forms I (CCDC number DOHREX01) and II (CCDC number DOHREX) which have respectively the space groups P21/c [27] and Pbca [28] , and form III whose structure has not yet been determined. The melting temperature and the melting enthalpy of form I have been found to be T m I = 187°C and H m I = 66 J/g and those of form II has been found to be T m II = 186°C and H m II = 85 J/g. According to burgers' laws [29] , since the highest melting point is associated to the lowest enthalpy of melting, form I and II are expected to form an enantiotropic system. This point, is also expected from the solubility curves of forms I and II in ethyl acetate which have been determined by Tung et al. [30] in the temperature range [6°C ; 54°C]. By extrapolation to higher temperatures, these curves are expected to cross around 160°C which suggests an inversion of the physical stability of forms II and I at this temperature. However, up to now, the transition between these two forms was never directly observed in the solid state suggesting that form II can be easily overheated above its expected transition temperature toward form I.
The objective of this paper is to investigate further the phase diagram of sulindac using original thermo-mechanical treatments. We will show, in particular, how the transition II→I can be easily triggered by a short milling process of form II and the causes of this triggering will be discussed in details. It will also be shown that a long milling leads to a total solid state amorphization of sulindac, which then shows a complex recrystallization pattern upon heating. were used to achieve good resolution and good thermal conductivity.

EXPERIMENTALS
Powder X-ray diffraction (PXRD) experiments were performed with a PanAlytical X'PERT PRO MPD diffractometer (λCuKα = 1.5418 Å for combined K1 and K2) equipped with an X'celerator detector (Almlo, The Netherlands). Samples were placed into Lindemann glass capillaries (Ø = 0.7mm) and installed on a rotating sample holder to avoid any artifacts due to preferential orientations of crystallites. Thermal treatments at a given temperature have been systematically carried out in the DSC device, using heating parameters identical to those used for the DSC investigations.
All samples were thus heated at 5°C/min (or 0.5°C/min when specified) at a given temperature, annealed at this temperature for 15 min to allow any transformation to reach completion, and then cooled to 20°C at 20°C/min. The samples were then removed from the DSC pan, placed in a Lindeman capillary and analysed by PXRD.
This protocol guaranties a very accurate control of the sample temperature and a perfect coherence between DSC and PXRD data. All X-ray diffraction patterns have thus been recorded at room temperature (RT), which has the advantage of avoiding any shift of Bragg peaks due to temperature changes and makes it possible the direct comparison of all diffraction patterns of the paper.
Thermogravimetric analysis (TGA) were performed with a Q500 TGA from TA Instruments (Guyancourt, France). Samples were placed in open aluminum pans, and the furnace was flushed with a highly pure nitrogen gas (50 mL/min). The temperature reading was calibrated using the Curie points of alumel and nickel, while the mass reading was calibrated using balance tare weights provided by TA Instruments. All TGA scans were performed at 5°C/min. Figure 1 shows the X-ray diffraction pattern of commercial sulindac (black line). It The melting temperature and the melting enthalpy are respectively Tm = 186°C and Hm = 85 J/g. These values are consistent with those already reported in the literature [31] . The TGA scan (5°C/min) of the crystal is also reported in the insert of

Amorphization by mechanical milling
The crystalline form II of sulindac has been milled in a planetary mill during 10 hours.
The X-ray diffraction pattern of the milled material is reported in figure 1. can also promote the crystallization as molecular mobility is higher at the surface than in the bulk [18][19][20] . The X-ray diffraction pattern of the milled sample, recorded at room temperature after its recrystallization at 115°C is reported in figure 1. The recrystallization was achieved in the DSC device by heating (5°C/min) the milled materials up to 115°C, annealing at this temperature for 15 min and cooling (20°C/min) to 20°C (see section 2 for details). The diffractogram shows many well defined Bragg peaks which are characteristic of the polymorphic form I of sulindac [27] (space group: monoclinic P21/c, CSD: refcode DOHREX01 (code CCDC: 637252)). No trace of the initial form II can be detected which indicates that the recrystallization of amorphous sulindac obtained by milling the crystal takes place entirely towards the form I.
The amorphization kinetics of sulindac upon milling has been investigated by PXRD and DSC. Figure 3a shows X-ray diffraction patterns recorded after different milling times, ranging from 0 to 10 hours. One can note a rapid decrease of the Bragg peak intensity accompanied by the development of an underlying broad halo which reflects the progressive amorphization of the material. The evolution is also marked by a strong broadening of the Bragg lines. This reflects a microstructural evolution of the remaining crystalline fraction which can be attributed to a strong crystallite size reduction and to the development of lattice strains upon milling. The disappearance of Bragg peaks is quite rapid so that after two hours of milling the milled material appears to be totally Xray amorphous. No further modification of the PXRD pattern can be observed for longer milling times. Figure 3b presents the heating DSC scans (5°C/min) of sulindac recorded after different milling times, ranging from 0 to 10 hours. It shows the progressive development of an exothermic recrystallization which reflects the increasing amorphization with milling time. This recrystallization slightly shifts towards the high temperatures. It ranges from 75°C to 95°C after 5 min of milling while it ranges from 90°C to 105°C after 10 hours of milling. This shift reveals an increasing physical stability of the amorphized fraction as milling progresses. It can be due to an evolution of the structural short range order in the amorphized fractions during milling. We may think, in particular, that the short range order of freshly amorphized fractions shows a reminiscence of the initial crystalline order which progressively vanishes upon further milling. Such an evolution would make the recrystallization upon heating increasingly difficult. The shift of the recrystallization could also be due to the decrease in the number of remaining crystallites as milling progresses. These crystallites are likely to facilitate the recrystallization of the amorphous fraction through a seeding effect, so that their gradual disappearance makes recrystallization more and more difficult.
The thermograms also show a Cp jump around 75°C whose amplitude increases in parallel with that of the enthalpy of recrystallization. It corresponds to the glass transition of the amorphized fraction. However, for short milling times, the Cp jump is truncated by the recrystallization exotherm which starts before the end of the glass transition. This effect is particularly pronounced for short milling times so that the glass transition cannot be clearly detected for milling times shorter than 10 minutes. The recrystallization toward form II for short milling times could be due to a seeding effect of the amorphized fractions by the initial form II not yet amorphized by the milling.

Recrystallization of milling induced amorphous sulindac
The latter disappears progressively during the milling so that the recrystallization is expected to be less and less oriented by the seeding effect. When the sample is totally amorphized, there are no more crystallites of form II and the recrystallization occurs entirely towards form I. The recrystallization toward form II could also be due to an evolution of the short range order of the amorphous fraction during milling. It is possible that the crystallites that have just been amorphized have a local order which exhibits a reminiscence of the crystalline order of the initial phase (form II) which could facilitate the recrystallization toward this phase. This reminiscent order is expected to progressively disappear upon further milling so that the recrystallization occurs no longer toward form II.

Polymorphic transition II→I upon heating
Interestingly, the thermograms recorded after short milling times ( figure 3) show transiently an additional endotherm around 160°C which is not observed in the nonmilled sulindac. It develops between 0 and 5 minutes of milling and then decreases for longer milling times to finally totally disappear after 60 minutes of milling. The origin of this peak has been investigated on the sample milled for 5 minutes for which the endotherm is maximum. Figure 5 shows X-ray diffraction patterns of this sample recorded at RT, just after milling and after heating at 120°C and 160°C (see section 2 for details). The black diffractogram has been recorded just after milling. It shows small One can also wonder why the transformation II→I occurs in the very shortly milled material while it is not detected in the absence of milling. Two main features distinguishing the milled and non-milled materials can be responsible for this difference.
Upon heating, just before the transformation II→I, the material milled for 5 minutes is made of defective crystallites of form II not yet amorphized, mixed with non-defective crystallites of form I and II arising from the recrystallization of the amorphous fraction previously generated by the milling. On the contrary, the non-milled material is free of form I and crystallites of form II are not defective. These two essential differences suggest two possible scenarios which can trigger the transformation II→I upon heating: -In the scenario 1, we might think that the traces of form I developing during the recrystallization of the milling induced amorphous fraction could act as seeds to trigger the transformation II→I at higher temperature. This could explain that the transformation does not occur in the non-milled material which is free of form I.
-In the scenario 2, the polymorphic transformation could be facilitated by the presence of crystalline defects in the milled form II which increases its Gibbs energy. It could also be promoted by the crystallite size reduction which increases the specific surface of the powder where nucleation and growth phenomena are known to be much faster than in the bulk [23] . To test the scenario 2, we have produced a sample made of strongly defective crystallites of form II without any crystallite of form I. Such a sample cannot be obtained simply by milling form II at RT as a noticeable amount of amourphous sulindac would be unavoidably produced. It has thus has been obtained by milling the crystalline form II at high temperature (130°C) for 5 minutes. This high temperature milling has been performed by equilibrating the milling jar containing the sample at 130°C in an external oven and then by milling the material at RT for 1 minute using the hot milling jar. The two previous steps were repeated 5 times to reach an effective milling time of 5 minutes at high temperature. The decrease of the milling jar temperature during the one minute milling stage at RT has been followed using an infrared thermometer. It appears that the temperature drop within one minute is about 3-5°C so that it can be estimated that the material has been milled at a temperature ranging between 125°C and 130°C. The pronounced that that milling time is long. This clearly proves that the transformation II→I upon heating is triggered by the defects generated by the milling process.
Since the transformation II→ I is triggered by the damages induced by the mechanical chocks, the evolution of the microstructure of form II upon both milling and heating has been analyzed in details. Figure 8 shows the X-ray diffraction patterns of sulindac form II recorded before milling (in black) and just after a 10 minutes milling at 130°C (in red).
Clearly, the Bragg peaks characteristic of form II (e.g. at 2 = 9.67°) are broader after milling. This broadening generally reflects both a crystallite size reduction and / or the development of micro strains in the crystalline lattice. A detailed analysis of the diffractions lines by Rietveld refinement (comparison between a simulated diffraction pattern and the experimental one) has been performed using MAUD software [37] on each whole PXRD pattern (see ref [38] for more details). The diffractometer set-up contribution to broadening of the diffraction peaks has been determined using a NAC

CONCLUSION
In this paper, we have studied the effects of milling on the structure of the crystalline form II of sulindac and on its evolutions upon subsequent heating. We have shown that, during milling, the fraction of crystalline form II decreases while that of amorphous sulindac increases so that a total amorphization occurs in less than 10 h of milling. Upon heating, the amorphous fraction recrystallizes toward forms I and II,

ACKNOWLEDGEMENTS
This project has received funding from the Interreg 2 Seas programme 2014-2020 cofunded by the European Regional Development Fund under subsidy contract 2S01-

DATA AVAILABILITY
The raw/processed data used in this paper are not available on line.     The black curve was recorded just after milling. The red and blue curves were recorded after heating (5°C/min) the milled material to respectively 120°C and 160°C, annealing at these temperatures for 15 min, and cooling (20°C/min) to 20°C. This thermal treatment was performed in the DSC device as explained in the "Experimentals" section.