Solid State Amorphization of Î²-Trehalose: A Structural Investigation

Article pubs.acs.org/crystal

Solid State Amorphization of β‑Trehalose: A Structural Investigation Using Synchrotron Powder Diffraction and PDF Analysis Pierre Bordet,*,†,‡ Aleksei Bytchkov,§ Marc Descamps,⊥ Emeline Dudognon,⊥ Erik Elkaïm,∥ Pauline Martinetto,†,‡ William Pagnoux,†,‡ Agnieszka Poulain,# and Jean-François Willart⊥ †

Université Grenoble Alpes, Inst NEEL, F-38000 Grenoble, France CNRS, Inst NEEL, F-38000 Grenoble, France § Institut Laue-Langevin, 6 rue Jules Horowitz, BP 156, F-38042 Grenoble, France ⊥ UMET, Universite Lille 1, F-59655 Villeneuve d’Ascq, France ∥ Synchrotron SOLEIL, l’Orme des Merisiers Saint-Aubin, 91192 Gif-sur-Yvette Cedex, France # European Synchrotron Radiation Facility, BP 220, F-38043 Grenoble Cedex, France ‡

ABSTRACT: We have investigated the amorphization of β-trehalose by high energy milling using in situ and ex situ synchrotron powder diffraction and PDF analysis. From this analysis we show that amorphization takes place through a two-phase process involving an amorphous and a long-range ordered phase. The proportion and coherent domain size of the latter rapidly decrease with milling time until the whole sample appears amorphized. The PDF describing the local structure of the amorphous phase after two hours of milling is very close to that of a sample quenched from the liquid, and seems to continue to evolve for longer milling times. Their differences with the PDF expected for a rigid THL molecule confirm the existence of a conformational disorder of the torsion angles from the glycosidic linkage between the two cycles forming the molecule.

■

INTRODUCTION Mechanical milling processes are widely used in the pharmaceutical industry to reduce the size of the particles.1,2 This size reduction generally improves the flowability of powders, the dosage accuracy, the compaction, and the solubility.1,3 However, milling is also known to induce solidstate transformation and, in particular, amorphization of crystalline powders.2,4,5 Such an amorphization can have some positive effect on the efficiency of a drug as it generally increases its solubility and thus its biodisposability. However, it also gives rise to some physical and chemical stability issues associated with the unstable nature of the amorphous state.6 As a result, a perfect understanding of the physical mechanisms involved in the solid-state amorphization process upon milling appears essential to both foresee and control these transformations. This requires capture of the pattern of structural evolution which drives the crystalline state into a disordered amorphous state. However, most structural investigation techniques, like powder X-ray diffraction, rapidly fails to probe a material as its crystalline long-range order disappears. In this paper, we present a pair distribution function (PDF) analysis of the entrance of a crystalline material into the amorphous state upon milling. The pair distribution function G(r) (PDF) is based on the Fourier transform of total scattering powder diffraction data to direct space. It yields the probability of finding two atoms separated by a distance r in an isotropic sample7 and thus provides multiscale information © XXXX American Chemical Society

about the atomic arrangement which cannot be obtained from conventional powder X-ray diffraction analysis. In the version we use,8 the PDF is defined as G(r ) = 4πr[ρ(r ) − ρ0 ] =

2 π

∫Q

Q max

Q [S(Q ) − 1] sin Qr d Q

min

where ρ(r) is the microscopic pair density, ρ0 the average pair density, S(Q) the structure function and Q = 4π sinθ/λ. Due to the second term in this expression which subtracts the average pair density, the G(r) function displays oscillations about 0 reflecting the presence or absence of interatomic distances at the corresponding r value. At distances above the particle size or coherent structural domain size, the microscopic pair density becomes equal to the average pair density and the PDF vanishes. Study of the PDF at short distances yields the local structure, and this technique was used for a long time to investigate amorphous materials and liquids. PDF analysis has now found many other applications3 in the structural study of nanoparticles, short-range ordered materials or compounds underReceived: May 2, 2016 Revised: June 28, 2016

A

DOI: 10.1021/acs.cgd.6b00660 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

consequences on the intramolecular interatomic distance distribution and then the PDF.

going local structural distortions. One of the most striking properties of the PDF is that it can be obtained experimentally with the same accuracy whatever the crystallization state of the sample. The usefulness of this unique property was recently demonstrated in studies of molecular compounds recrystallization.9,10 Here, we will make use of this advantage to follow the structure and the microstructure of a molecular compound as a function of high energy milling time tmill, using the same PDF analysis method, from the pristine crystalline to the fully amorphized state, and compare it to classical crystallographic analysis of the same sets of data. Fully or partly amorphized active pharmaceutical ingredients (APIs) are known to often be subject to recrystallization effects. For example, Bragg peaks from their powder diffraction patterns become progressively more intense and less broad as a function of conservation time after milling,11 reflecting the progressive recovery of the crystalline state. This effect may depend on the conditions of conservation of the sample (relative humidity, temperature, etc.). Obviously, the results of ex situ experiments may be biased to a large extent by this effect. In order to investigate it, we have carried out two synchrotron powder diffraction experiments (one ex situ and one in situ) and compared the evolution of the sample microstructures in both cases. These experiments were carried out on the β-trehalose compound. β-trehalose (β-THL) is the stable crystalline form of α-α′ trehalose, a disaccharide made of two pyranose rings joined by a glycosidic linkage12 (Figure 1).

■

EXPERIMENTAL SECTION

Sample Preparation and Characterization. Crystalline anhydrous β-THL (purity ≥ 99.5%) was purchased from Acros Organics and was used without any further purification. Ex situ ball-milling was performed in a high energy planetary mill (Pulverizette 7 − Fritsch) at room temperature. We used ZrO2 milling jars of 45 cm3 with seven balls (Ø = 1.5 cm) of the same material. One gram of β-THL was placed in the planetary mill, corresponding to a ball/sample weight ratio of 75:1. The rotation speed of the solar disk was set to 400 rpm. We took care to alternate milling periods (typically 20 min) with pause periods (typically 10 min) in order to limit the overheating of the sample. A set of samples with increasing milling times was prepared. Each sample taken out from the mill was immediately submitted to a characterization by differential scanning calorimetry (DSC), or placed into a sealed Lindemann glass capillary for further investigations using X-ray diffraction. A quenched sample was prepared by heating in a capillary commercial β-THL powder just above its melting point and pouring it on a glass plate at room temperature. DSC measurements were performed with a TA Instruments Discovery calorimeter from room temperature up to above the βTHL melting point, using approximately 3−5 mg of sample. All the experiments were performed with a heating rate of 5 °C/min. The sample was placed in an open aluminum pan (container with no cover) and was flushed with highly pure nitrogen gas. Temperature and enthalpy readings were calibrated using pure indium at the same scan rates and with the same kind of pans used in the experiments. Laboratory powder X-ray diffraction (XRPD) experiments were performed just after sample preparation with a Panalytical XPert Pro diffractometer in Debye−Scherrer geometry (λCuKα = 1.5418 Å), equipped with an Xcelerator linear position sensitive detector. The samples were contained into Lindemann glass capillaries (Ø = 0.7 mm). All analyses of the X-ray diffraction data were performed with the FP_Suite software.21 Microstructural parameters were obtained thanks to the determination of the diffractometer resolution function using the measurement of a NAC sample.22 The samples were studied by field effect scanning electron microscopy (FESEM) using a Zeiss Ultra+ microscope. The powder without any preparation was dispersed on a carbon Scotch tape. High Resolution Synchrotron Powder Diffraction and PDF Analysis. The experiment was carried out at the high resolution powder diffraction (HR-XRPD) station of the CRISTAL beamline at synchrotron SOLEIL on the samples with tmill = 0, 30, 45, 120, and 180 min and a sample quenched from the melt. The pristine sample with tmill = 0 was taken from another part of the same commercial powder pot. The synchrotron measurements took place 6 days after the samples were prepared. A wavelength of 0.5441 Å selected by a Si(111) monochromator was used. Powder diffraction data were collected at room temperature using Ø = 0.7 mm capillary samples in parallel beam Debye−Scherrer geometry with the 21 Si(111) crystal multianalyzer stage.23 In order to reach a high enough maximum value of Q = 4.π.sinθ/λ and to improve counting statistics at high Q for the PDF data collection, the latter consisted of four detector scans up to 2θ = 125° starting from 0°, 30°, 60°, and 90°, respectively. These scans were summed and binned to a 0.002° step size for the Rietveld refinements. To determine the Instrument Resolution Function and the signal from sample environment, the diffractograms of a NAC sample and an empty capillary were recorded under the same conditions. To further increase counting statistics, all diffractograms used for the PDF analysis were rebinned to a 0.016° step size. The PDFGetX2 software24 was used to calculate the PDFs from the powder diffractograms. The environmental signal was subtracted using the empty capillary data. Corrections for polarization, absorption effects, and Compton and inelastic scattering contributions were applied before Fourier transforming the data using a Qmax value of 20.5 Å. A Lorch damping function was used in order to decrease the effect of statistical noise at high Q.

Figure 1. Sketch of the α-α′ trehalose molecule. The numbering is from the structure of β-trehalose of ref 12. The two Ψ and Φ torsion angles of the glycosidic linkage are marked.

Trehalose (THL) is a well-known molecule for its protective properties in Nature and is used as a stabilizing excipient in the pharmaceutical industry.13 It is known as a model compound for the study of the amorphous state, which can be prepared by various techniques,14 including quenching from the liquid phase, high energy milling,15 dehydration of the dihydrate form,16 and spray drying.17 Previous investigations18 have shown that the milling times to reach amorphization by high energy milling (several hours) are compatible with the requirements of in situ experiments, and the recrystallization process may occur over sufficiently long time scales for ex situ measurements, both at large experimental facilities. A further interest of THL is the ability of this molecule to present a configurational disorder in the liquid and glassy states by varying the two torsion angles of the glycosidic linkage in a quite large range. This was observed in the glassy state using 13 C SS-NMR19 and proposed from ab initio calculations.15 Thus, one aim of our PDF analysis of β-THL samples amorphized by different methods (milling and quenching) was to observe directly this effect which must have strong B



Article

Figure 2. Left: DSC curves of β-THL samples after increasing milling times and quenched liquid (5×). Right: corresponding XRPD patterns using a laboratory diffractometer with λCuKα In Situ High Energy Synchrotron Powder Diffraction during Milling. The experiment was carried out at the ID15B beamline of the ESRF. We took advantage of the very high energy (λ = 0.14259 Å) and highly collimated X-ray beam to perform real-time diffraction data collection, while the powder sample was submitted to high energy milling at room temperature within the vessel of a modified oscillating Retsch MM400 ball mill.25,26 About 0.6 g of commercial β-THL was placed in the Perspex vessel with a single ZrO2 ball (Ø = 10 mm, ball/ sample weight ratio of 9:1) and the oscillation frequency was set to 30 Hz. The mill was operated continuously for 3 h with 20 min interruptions after each hour of milling to avoid heating of the sample. The X-ray beam (size 0.4 × 0.4 mm) was aligned across the vessel so as to maximize the amount of sample powder allowed to diffract. The diffracted signal was collected with a 10 s/image exposure time (plus a ∼2.4 s dead time between two consecutive images) on an amorphous Si flat-panel Perkin-Elmer detector located at 1.25 m from the sample and was azimuthally integrated using Fit2D.27 The scattering signal from the empty vessel was then subtracted and the diffractograms were normalized to compensate for the evolution of the direct beam intensity, and the rather randomly varying amount of diffracting powder and absorption from the ZrO2 ball. The Instrument Resolution Function was determined from the measurement of a CeO2 standard with the mill ball removed. The evolution of the sample microstructure was obtained by sequential Rietveld refinements using the FP_Suite. The diffraction signal from the ball was found to consist of two different ZrO2 phases with tetragonal (S.G.: P42/nmc) and monoclinic (S.G.: P21/c) symmetries, which were treated with LeBail refinement, whereas the β-THL phase was treated with the Rietveld method. For the latter phase, the scale factor, cell parameters, and a parameter describing isotropic broadening effects of particle size origin were refined.

the eruption of water from microinclusions and readily disappears after the shortest milling time. Thermogravimetric analysis has clearly shown that the water loss responsible for the shouldering is close to 1% of the total sample mass. On the other hand, the milled samples show several additional features: An exothermic peak attributed to a recrystallization starts to be detectable around 140 °C after only 10 min of milling. This peak then becomes increasingly larger and strongly shifts toward higher temperatures for longer milling times. This increasing recrystallization is a clear indirect signature of the progressive amorphization which occurs during the milling process. After 5 h of milling the enthalpy of crystallization becomes stationary and its comparison with the enthalpy of melting of crystalline β-THL indicates that the amorphization by milling is total. However, the crystallization still shifts toward high temperatures for longer milling which reveals an increase of the stability of the amorphous material against crystallization. A melting peak whose evolution can be divided into two stages. From 0 min to 1 h, it progressively decreases, broadens, and shifts strongly toward the low temperatures. This evolution is so pronounced that after 1 h of milling, the melting peak becomes hardly detectable. From 1 to 10 h, we observe the inverse behavior. The melting peak develops, sharpens, and shifts toward the high temperatures to retrieve the characters of the nonmilled crystalline β-THL. A Cp jump characteristic of a glass transition whose onset is found to be constant and located at Tg = 120 °C indicates that the amorphous THL which forms during the milling procedure has a glassy character. It is already detectable for tmill = 10 min, increases until tmill = 1 h, and then remains stationary upon further milling. The Cp jump amplitude of the sample milled 10 h appears to be close to that of the quenched liquid, which is further proof that the crystalline material has been totally amorphized during the milling process. The corresponding XRPD patterns show a concomitant decrease of intensity and broadening of the Bragg peaks with the progressive appearance of a diffuse scattering halo. They

■

RESULTS Calorimetry and Laboratory XRPD. Crystalline β-THL was milled from 0 min to 10 h and the obtained samples were characterized by DSC and XRPD. These DSC curves and corresponding XRPD data are displayed on Figure 2. The DSC curve of the nonmilled β-THL shows only a complex melting endotherm with a maximum at ∼209 °C. The low temperature shoulder on this endotherm is attributed to C



Article

can be described as the sum of a “crystalline” long-range ordered (LRO) phase characterized by the existence of Bragg peaks, and an “amorphous”, short-range ordered (SRO) phase giving rise to the diffuse halo. At this stage, one can assume that both phases represent two different structural arrangements of the same THL molecules. At tmill = 2 h, Bragg peaks are hardly detectable, and for longer times none can be observed. In all samples, only the β-THL phase could be detected in the crystalline fraction. In order to estimate the proportion of crystalline and amorphous phases, the diagrams were normalized at high angles in a region where no Bragg peak can be detected. The diagram of the pristine crystalline powder was then Rietveld refined with the background described as an 18 term Chebyshev polynomial. This refined background, due to scattering from the sample environment (air, capillary, etc.) and self-scattering from the β-THL compound was subtracted to the normalized diagrams of all other samples. These diagrams were then used for Rietveld refinements for tmill up to 2 h. The atom positional parameters from ref 12 were kept fixed, an overall atomic displacement parameter (a.d.p.) and a peak broadening parameter measuring the size of the coherent structural domains were refined, together with the cell parameters and the scale factor. For tmill = 2 h, the size parameter could not be refined and was fixed at the value obtained for tmill = 1 h, due to weakness of the peak to background ratio. For each sample, the proportion of the crystalline phase to the whole sample was obtained from the ratio of areas integrated between 2θ = 5° and 40° above the refined background curve and the full powder pattern, respectively. Table 1 summarizes the evolution of the crystalline fraction and microstructural parameters for the refined samples. These

HR-XRPD Experiment at CRISTAL-SOLEIL for Ex Situ Milling. Figure 3 shows the HR-XRPD diffractograms recorded at the CRISTAL beamline for increasing milling times, together with those for the pristine sample and a sample quenched from the melt.

Figure 3. HR-XRPD patterns of β-THL for increasing milling times (downward) and quenched sample (bottom) obtained at the CRISTAL-SOLEIL synchrotron beam line. The contribution from the environment has been subtracted. The intensity of the Crystal pattern is divided by 2. Arrows mark the strongest reflections from THL-dihydrate.

Table 1. Fraction (LRO) and Coherent Domain Size (Size) of the Long-Range Ordered Phase for β-THL as a Function of Milling Timea Rietveld (lab data)

Rietveld (CRISTAL data)

PDF (CRISTAL data)

tmill (min)

LRO (%)

size (Å)

LRO (%)

size (Å)

LRO (%)

size (Å)

0 5 10 20 30 45 60 120

100 67(2) 55(2) 36(2) 21(2) 20(3) 10(2) 3(3)

2386(1) 402.4(4) 329.0(4) 279.6(5) 241.0(9) 215(1) 183(2) 183b

100 18(2) 8(2) -

∞ 262(1) 195(1) -

100 19(9) 10(3) 4(7)

∞ 279(14) 191(14) 68(8)

Again, one can observe the progressive evolution of the Bragg peaks toward lower intensities and larger breadths and the increase of an amorphous halo. The instrumental background obtained by measuring the signal from an empty capillary was subtracted to all diagrams. The mass fraction of the LRO phase was estimated in a similar way as described above for laboratory data, from the ratio of areas integrated between 2θ = 2° and 40° above the refined background curve and the full powder pattern, respectively. Bragg peaks from the THL-dihydrate can be seen on the pattern of the pristine sample (arrows on Figure 3). THL-dihydrate is known to remain in its crystalline form without presenting amorphization effects in ball milling experiments at room temperature.28 Therefore, THL-dihydrate is not expected to be present in the amorphous fraction of the milled samples, and except for the pristine sample (which was from a separate batch and was probably contaminated with water), no corresponding Bragg peak can be observed on the diffractograms shown in Figure 3. Thus, the milled sample were not subject to water contamination. Rietveld refinements of the diffractograms allowed to study the evolution of the structure and microstructure of the LRO β-THL phase as function of milling time. For tmill = 0, the peak profiles had to be described using anisotropic size and strain parameters, while for longer tmill, the

a

Data from Rietveld refinement of laboratory XRPD (left), HR-XRPD at CRISTAL-SOLEIL (center), and PDF analysis of the same CRISTAL-SOLEIL data (right). bThis parameter was fixed to the tmill = 60 min value.

data clearly confirm that high energy milling leads to amorphization via a two-phase process. After only 5 min of milling, already 33% of the sample mass contributes to the diffuse scattering halo: the LRO and SRO phases are coexisting in the sample. At the same time, the size of the coherent structural domains of the LRO phase decreases to become smaller than 200 Å at tmill = 60 min. For longer times, the crystallographic analysis using the Rietveld method fails due to the small LRO phase proportion and domain size. D



Article

size effect was found to dominate and a single isotropic size broadening parameter was sufficient to satisfactorily fit the patterns, as for the laboratory data. This had the advantage of allowing the comparison with the results of the PDF refinements (see below), in which the treatment of isotropic particle size effects is incorporated, but not the one for stress− strain or anisotropic effects. The Rietveld refinements were carried out for tmill = 0 to 45 min, after which the Bragg peaks were too weak and broad for a reasonable analysis. The starting structural parameters from ref 12 were refined using soft distance and angle restraints for tmill = 0 and 30 min, and kept fixed for tmill = 45 min. In all three cases, the cell parameters, overall a.d.p., and peak broadening parameters were refined. The background was described by linear interpolation of a set of 2θ points with refined intensities. For the pristine sample, THL-dihydrate was introduced as a secondary phase using the data from ref 32; its refined proportion was 2.09(1) wt %. The Rietveld plots for the pristine sample and the one milled for 30 min are shown as examples in Figure 4.

Table 2. Structural Parameters and Agreement Factors from the Rietveld Refinements of β-THL Samples for Increasing Milling Timesa a (Å) b (Å) c (Å) β (Å) V (Å3) Boverall (Å2) Rwp RBragg χ2 a

tmill = 0

tmill = 30 min

tmill = 45min

13.00236 (2) 8.26375 (2) 6.79961 (1) 98.3581 (2) 722.831 (3) 1.29 (2) 12.5 5.58 2.67

12.9973(5) 8.2623(3) 6.7992(3) 98.341(3) 722.43(5) 1.70(8) 9.55 3.45 1.01

12.993(2) 8.263(1) 6.799(1) 98.34(1) 722.2(2) 1.2(2) 11.5 8.21 1.44

Data obtained at the CRISTAL-SOLEIL beam line.

found for the laboratory data refinements (see Table 1 and Figure 7). On one hand, this result is expected because the same samples were analyzed in the two experiments. However, since the laboratory measurements were made just after sample preparation, whereas the synchrotron experiment was carried out about 6 days later, it indicates that recrystallization effects can be neglected at this time scale. This is probably due to the relatively high value of the glass transition temperature for this compound. As stated above, for longer times, the crystallographic analysis using the Rietveld method fails due to the too small LRO phase proportion and coherent domain size. Moreover, this technique provides no indication about the SRO phase structure and coherent length. To obtain this information, we have used the PDF analysis of the same data as described in a following part. In Situ Milling Experiment at ID15B-ESRF. The evolution of the β-THL diffraction patterns during in situ milling can be seen in Figure 5 after normalization and subtraction of the signal from the empty Perspex milling vessel. The quite fast decrease of the Bragg peak intensities and concomitant raise of the diffuse scattering halo are obvious. This observation already qualitatively confirms the two-phase model proposed from the analysis of the CRISTAL experiment

Figure 4. Rietveld refinement plots of two β-THL samples. Data obtained at the CRISTAL-SOLEIL beam line. Top: pristine sample, bottom: sample milled for 30 min. Tick marks for the pristine sample represent Bragg peak positions of β-THL (top) and THL-dihydrate (bottom).

The LRO phase proportion and coherent domain size are reported in Table 1, while the cell parameters, overall a.d.p.’s, and agreement factors are shown in Table 2. The structure does not display any significant evolution upon milling. The cell parameters remain very close between tmill = 0 and 45 min, as seen in Table 2. By comparison of the refined structures for tmill = 0 and 30 min, the average distance between the positions of the same atom in the two refinements was 0.08(4) Å, with a maximum value at 0.17(16) Å. The LRO phase proportions and coherent domain sizes display a quite similar evolution as

Figure 5. Evolution of the β-THL diffraction pattern recorded at the ID15B-ESRF beam line during in situ milling. E



Article

the evolution of the phase fraction, obtained from the relative decrease of the scale factor, and coherent domain size of the LRO phase as function of milling time, respectively. The LRO phase could be observed and refined during the entire experiment, indicating that full amorphization of the sample was not reached, in contrast to the previous experiments. This can be attributed to the smaller milling efficiency of the oscillatory mill used here, as compared to the planetary ball mill used in ex situ experiments. To compare the efficiencies of both types of instruments, the values of LRO phase fraction vs milling time (Table 1) determined previously at CRISTAL-SOLEIL and using laboratory data on samples milled ex situ using the planetary mill are reported at the top left of Figure 7. Both experiments using the planetary mill yield similar results, whereas for the ID15B in situ experiment using an oscillatory mill, the decrease of the LRO phase proportion is much slower. Using dimensionless reduced time as usually done when studying solid state kinetics,29 a coarse estimate of the relative efficiencies of the two types of mill can be obtained as shown at the top right of Figure 7, for which the time scale for the different experiments were scaled to t(20%), the time at which the phase fraction has decreased to 20% of its starting value. This gives a reasonable agreement between the evolutions of the LRO phase proportion in both cases, with a time scale of approximately 3 between the planetary and oscillatory mill. This suggests that both milling techniques imply similar mechanisms of microstructure modification, based on chocks of particles with the milling ball and vessel. It also

above. Despite the strong contamination from the ZrO2 ball diffraction signal, the scale factor, cell parameters, and isotropic size parameters of the β-THL LRO phase fraction could be followed throughout the 884 patterns representing 3 h of milling by sequential Rietveld refinements. The crystal structure was considered fixed and described with positions and a.d.p.’s taken from ref 12. No noticeable evolution of the cell parameters or structure could be observed during the entire experiment. A typical Rietveld refinement plot obtained after 45 min of in situ milling is shown in Figure 6. Figure 7 represents

Figure 6. Rietveld refinement plot obtained for β-THL after 45 min of in situ milling at the ID15B beamline. Tick marks are from top to bottom: β-THL, ZrO2 tetragonal phase, ZrO2 monoclinic phase.

Figure 7. Comparison of the long-range ordered phase fraction (top) and coherent domain sizes (bottom) obtained by Rietveld refinement for in situ oscillatory milling at ID15B (gray dots) and ex situ planetary milling at CRISTAL (green disks) and laboratory (blue squares). The results from the PDF fits are shown as red squares. In the right panels, the times have been scaled by t(20%), the time at which the phase fraction has decreased to 20% of its starting value. F



Article

Figure 8. PDFs from data recorded at the CRISTAL-SOLEIL beam line for β-THL. Left: the PDF of the pristine sample up to 250 Å. Right: comparison of the PDFs recorded for the pristine and quenched samples up to 30 Å.

adjacent molecules. Above, only pairs belonging to different molecules contribute to the PDF. Clearly, in Figure 8, oscillations of the PDF are present in the whole range for the pristine sample, while for the quenched one the oscillations become hardly detectable after 6−7 Å, which yields an estimate of the coherence length of the structural arrangement. Since this distance is smaller than the longest intramolecular distance in β-THL (∼9.5 Å), this observation confirms the amorphous nature of the quenched sample, i.e., no correlation of position between adjacent molecules can be observed on average. Below ∼4 Å, both PDFs are remarkably identical, indicating that the two compounds share the same atomic arrangement in this distance range. Here, as shown above, the PDFs are dominated by intramolecular distances and even more by distances between atoms belonging to the same pyranose ring. The common structure of these groups is thus maintained during the melting/quenching process, which is expected since the pyranose groups are highly stable and should not be influenced by crystal packing and molecular conformation. Between 4 and 7 Å, clear discrepancies between the two PDFs are visible. In this range, intermolecular distances become more important and it is difficult to decide whether these discrepancies originate from modifications of the intramolecular arrangement between the crystalline and amorphous state, or to the disappearance of the intermolecular contribution to the PDF for the quenched sample. To answer this question, we have calculated the PDF of a single THL molecule from the atomic positions of ref 12. To do this, we computed the Debye function of the single molecule using the debyer software30 and then used the PDFGetX2 software to obtain the PDF from the Debye function. This theoretical PDF is compared to the experimental one for the quenched sample in Figure 9. Here again, both PDFs are very similar up to ∼4 Å, after which obvious differences appear in positions and intensities of the peaks, indicating that the set of intramolecular distances above ∼4 Å is significantly different in the quenched sample than in the crystalline phase. Moreover, the oscillations of the calculated PDF remain much more prominent up to above 8 Å. The fading of the PDF of the quenched sample 2−3 Å before the longest intramolecular distances suggests that the two glucose units of the molecule may be rather randomly oriented with respect to each other in the quenched state. This is reminiscent of 13C SS-NMR observations of a quite wide distribution of the torsion angles for THL in the glassy state.19,20 The existence of such a sizable flexibility of the

shows that ex situ experiments can lead to an accurate representation of the effect of milling on the LRO phase proportion, at least for β-THL and provided the experiment is carried out quickly after milling was performed. The same time scaling factor was used to compare the average domain sizes for in situ and ex situ milling at the bottom of Figure 7. The size decrease is very fast: it becomes smaller than 500 Å after ∼5 min of milling, then tends more slowly toward ∼200 Å. Since all ex situ data were measured within this slowly decreasing part of the size evolution, the effect of time contraction is less sensitive here than for the LRO phase fraction shown above. Nevertheless, the evolution of the domain size appears to be very similar for both types of mills, confirming our previous observations for the phase fraction. Pair Distribution Function Analysis. Figure 8 (left) shows the experimental PDF obtained at the CRISTAL beamline for the pristine β-THL sample up to 250 Å. On the right side of the same figure, the same PDF is compared to that of the quenched sample up to 30 Å. Due to the tremendous resolution of the instrument, no decay of the PDF due to the Fourier transform of the instrumental resolution is observed even for very long distances. This allows performing a multiscale analysis of the data, depending on the distance range considered. Because of the complex molecular structure, it is not possible to assign a peak of the PDF to a single bond, but salient features can be empirically fitted using Gaussian functions to identify their origin using distances calculated from the known structure. For example, the first peak at 1.46 Å is composed of C−O first neighbor bonds at 1.42−1.46Ǻ and C−C bonds at 1.52−1.54Ǻ . The second peak at 2.42 Ǻ is made of second nearest neighbor C−O and C−C distances ranging from 2.34 to 2.54 Ǻ and the third one of O−O and C−O distances between 2.70 and 3.06 Ǻ . The large majority of these bonds originate from pairs of atoms belonging to a single hexagonal pyranose ring, with fewer additional distances around 2.36 and 2.83 Ǻ from pairs of atoms engaged in or close to the linkage between these rings. The longest pair distance in a ring is 6.3 and 9.5 Ǻ for the whole molecule. The first few intermolecular contacts appear at about 2.7 Ǻ through hydrogen bonds between O3, O6, and O11 atoms of neighboring molecules. So, below 4 Ǻ the PDF is largely dominated by atom pairs from a single pyranose ring. Between 4 and 9 Ǻ , it consists in a combination of intramolecular distances between atoms of the two pyranose rings and intermolecular distances between G



Article

signal progressively vanishes so that the last peak F at ∼8 Å no longer exists. The characteristic features of the PDF of the quenched sample are systematically closer to the average data, which try to mimic the effect of conformational disorder, than to those of the single THL molecule. This trend is a clear indication that the conformation of the THL molecule is indeed disordered in the quenched phase and our PDF data provide a direct observation of this effect. Further study including quantitative structural refinement is in progress. In Figure 10 is shown the evolution of the PDFs from the same experiment with respect to milling time, together with the patterns from the quenched and pristine samples. At first glance, one can observe that the PDFs remain almost identical up to ∼4 Å, after which the amplitude of the oscillations gradually decreases with increasing tmill. This is consistent with the two-phase model proposed above, where an “amorphous” SRO phase of very small coherence domain size starts to exist even for short milling times and then increases in proportion, coexisting with a “crystalline” LRO phase for which the proportion and coherence domain size progressively decrease. Therefore, the observed total PDF could be modeled by summing the PDFs of the SRO and LRO phases. We have built such an empirical model by using the experimental PDFs of the quenched and pristine samples. The experimental PDFs of the milled samples were then tentatively described as

Figure 9. Comparison of the PDFs calculated for a single THL molecule from ref 12 (green line), the PDF obtained by a weighted average of the calculated PDFs of conformationally disordered THL molecules by varying the torsion angles of the glycosidic linkage (red line), and the experimental PDF of the quenched sample (blue points).

molecule is consistent with the observation by PDF analysis of a rigid structure for distances up to ∼4 Å and an increasing disorder for longer intramolecular distances. A quantitative test of this effect would require a fitting of the PDF to a structural model, but this is not feasible due to the complexity of the molecule and the fact that the PDF results from a distribution of conformations. In order to obtain a semiquantitative check of this assumption, we have simulated the PDFs for a single molecule as above, but with the two torsion angles of the glycosidic linkage varying from 50° to 100° by steps of 10°. These PDFs were then averaged according to the distribution probability reported in ref 19 from 13C SS-NMR data. The average is compared to the PDF calculated from the structure of ref 12 in Figure 9. Although we cannot expect from such a calculation a perfect agreement with the experimental data, it is interesting to observe the changes in the PDF brought about by the weighted averaging with respect to the experimental PDF of the quenched compound. With respect to the PDF calculated for a single molecule in the crystalline phase, the averaged PDF is identical below ∼3.2 Å, where only distances inside glucose units are present. At higher distances, differences between the PDFs occur, and can be compared by looking at the features marked by letters on Figure 9. The positions and intensities of the A, B, C, and D peaks obtained by Gaussian fitting are reported in Table 3. One can see that peak A becomes larger, peak B becomes smaller and is slightly shifted to the left, peak C becomes smaller and a shoulder appears on its left side, peak D is shifted to the right, and then toward larger distances the

Gm − calc = s × [sm × E(D) × Gc + (1 − sm) × Gq ]

where Gm‑calc is the calculated PDF for the milled sample, Gc and Gq are the experimental PDFs of the pristine crystalline and quenched samples, s and sm are the experimental scale factor and the LRO phase fraction, and E(D) is an envelope function representing the effect of a spherical particle of diameter D (i.e., the isotropic coherent domain size) on the PDF of the LRO phase. 31 Note that this empirical method based on experimental data obtained in the same way does not depend on instrumental parameters. The fit of Gm‑calc to the experimental PDFs yielded for each sample the fraction and the coherent domain size of the LRO phase, reported in Table 1. Figure 11 shows a typical fit with the sample milled for 30 min. Figures 7 compares the evolution of the phase fraction and coherent domain size of the LRO phase as determined by diffraction (laboratory and synchrotron data) and PDF. The two techniques yield quite similar results, confirming the validity of our two-phase analysis of the PDF. Then, subtracting the determined LRO phase (= sm × E(D) × Gc) from the total PDF yielded the PDF of the SRO phase for the milled samples. The evolutions of these PDFs are shown on Figure 12 compared to that of the quenched sample. In all cases, the PDFs do not extend after 6−7 Å, thereby confirming the amorphous nature of these phases and supporting the two-phase model used for the analysis. The PDFs of the milled samples are quite similar to that of the quenched sample up to r ≈ 4 Å. It is interesting to note that for tmill = 2 h, the whole PDF is quasi identical to that of the quenched sample, whereas for tmill = 3 h, discrepancies start to appear between 4 and 6 Å, although these three samples are all fully amorphized as seen from Figure 12. This can be related to the difference existing between the HR-XRPD patterns of the 3 h milled sample and of the quenched sample (Figure 3). For the milled sample there is a surplus of diffuse scattering intensity between Q = 1.4 and 2 Å−1, which is also a 2θ range where the pattern of the crystalline phase shows several intense

Table 3. Positions (Left) and Intensities (Right) of Characteristics Features Observed on the PDFs of Amorphous Trehalose Reported in Figure 9, Obtained by Gaussian Fitting. PDF feature

quench sample

single molecule from cif

A B C D

3.7175(9); 2.23(1) 4.218(1); 1.36 (1) 4.946 (1); 0.866 (6) 5.6068(7); 0.447(3)

3.734(1); 1.89(3) 4.253(2); 3.21(5) 5.003(1); 1.17(1) 5.43(3); 1.6(8)

weighted average of conformational disorder 3.734(1); 4.228(1); 4.965(1); 5.533(2);

2.37(3) 2.56(2) 0.89(2) 0.69(1) H



Article

Figure 10. Evolution of the PDFs of β-THL recorded at the CRISTAL-SOLEIL beam line as function of milling time.

R=

1 n−1

n

⎛ Xi − X̅ ⎞⎛ Yi − Y ̅ ⎞ ⎟⎟ ⎟⎜⎜ ⎝ σx ⎠⎝ σy ⎠

∑⎜ i=1

where X and Y are two distributions to be compared and X̅ and σx represent the mean and standard deviation of the X distribution. For the distance range 1 ≤ r ≤ 10, the values of R obtained for the quenched-tmill = 2 h and quenched-tmill = 3 h couples of PDFs were 0.995 and 0.916, respectively. The values close to 1 obtained in both cases indicate that the main structural features are similar for all three samples. However, the smaller correlation for tmill = 3 h clearly indicates that structural modifications at the molecular level take place even at long milling times. Elucidating the nature or these modifications would require the refinement of the PDFs including atomic positions and has not been undertaken yet. These results suggest that, as for the quenched sample, amorphization by high energy milling leads to a state where the torsion angle of the glycosidic linkage binding the two pyranose rings groups of the THL molecules present a marked disorder, leading to the fading of the PDF several angstroms before the largest intramolecular distance of ∼9.5 Å. The local structural arrangement is identical to that of the quenched sample for tmill = 2 h, but seems to continue evolving at longer milling times. This may be related to the evolutions of recrystallization and melting temperatures observed by DSC, while both phenomena are absent in the case of a quenched liquid. It would be very interesting to pursue these experiments by investigating the PDFs of samples with longer milling times, looking for structural modifications at the mesoscopic scale. It could also be very interesting to follow the evolution of the PDFs during the milling of a previously quenched sample.

Figure 11. Fit to the PDF of the sample with tmill = 30 min, using the model based on the weighted sum of the pristine and quenched samples PDF’s. Observed PDF: blue circles, calculated PDF: red line, difference: green line.

■

DISCUSSION The crystallographic and PDF analyses both indicate that amorphization by high energy milling can be described fairly accurately with a two-phase model, in which as a function of milling time, a LRO phase decreases in proportion and coherent domain size and a SRO phase increases in proportion to finally represent the whole sample volume. However, the evolutions of these two phases are markedly different. The possibility to extract the PDF of the SRO phase allowed us to demonstrate its amorphous character, with interatomic correlations not exceeding the molecule size in the whole range of tmill. This amorphous phase is formed from the

Figure 12. PDFs of the SRO phase of β-THL obtained by subtracting the contribution of the LRO phase, as a function of milling time. The PDF of the quenched sample is shown for comparison.

Bragg peaks. To quantify the similarities of these PDFs, we compared those measured at tmill = 2 h and 3 h to that of the quenched sample using the Pearson product-momentum correlation R. This indicator has already proven its usefulness in similar context4,5 and is defined as I



Article

beginning of the milling process, so that amorphization cannot only be described as a progressive increase of average defect density, or reduction of coherent length inside the grains. It is logical to propose that this amorphous phase is primarily produced at the grain surface where the chocks are more efficient, so that the general picture of the sample would be a distribution of LRO grains surrounded by an amorphous layer, in a core−shell fashion. The evolution of the LRO microstructure is complex, with a concomitant reduction of core size and crystallinity. The size reduction can be interpreted as due to the growth of the amorphous layer thickness, which also leads to the observed decrease of the LRO phase proportion. At the same time, the coherent length of the LRO phase inside the grain cores decreases due to an increase of the defect density brought about by the chocks from the milling process. We have determined the evolution of the coherent domain size of the LRO phase by Rietveld and PDF refinements of milled samples. This is a measure of the distance over which the structure remains well enough ordered to produce interferences of the wave scattered by individual atoms (diffraction in reciprocal space) or oscillation about the average pair density (PDF in real space). It is important to remark that this parameter may not reflect the core size, but may be smaller in the case where the defect density inside the particle cores is high. Ultimately, the complete amorphization will be reached either by the growth of the amorphous shells to the center of the grains and disappearance of the cores, or by the decrease of the coherent domain size due to very large defect densities inside the remaining cores, until it reaches the molecule dimensions. It is worth noting that the PDF analysis cannot distinguish these two cases, although for the latter, the nature of the SRO phases in the outer shells and inside the remaining cores are different. Which process dominates will depend on many factors, including the grain size of the starting material. As seen in the SEM images of a sample milled for two hours in Figure 13,

Figure 14. Tentative model of the evolution of a grain substructure upon milling.

phase forming the shell and the LRO phase forming the core. With increasing tmill, the thickness of the SRO phase shell increases, and consequently the core volume becomes smaller. The defect density inside the core increases with milling time. The powder sample could be viewed as a collection of such grains, bathed in more or less agglomerated nanograins of the amorphous phase. Beyond this microstructural study, combined Rietveld refinement and PDF analysis were able to provide information about the structures of the LRO and SRO phases. This is particularly interesting for the case of THL, reported as being a flexible molecule, with a distribution of torsion angles from the glycosidic linkage having a half width as large as 17° in the glassy state.19 In the whole accessible range for Rietveld refinement of HR-XRPD data, we observe no significant modification of the average structure. Therefore, the molecular conformation in the LRO phase does not appear to be noticeably modified. On the other hand, the comparison with the quenched sample indicates that (1) the glucose units are rigid and keep the same atomic arrangement in the quenched and crystalline phases and (2) the PDF of the quenched phase results from the average of various conformations of the THL molecule and thus provides a direct observation of the molecular flexibility. Let us now compare the PDFs from the amorphous phases obtained by quenching and by high energy milling. Some differences are indeed expected between the two materials, since the milled compounds present a typical glass transition and a recrystallization peak in their DSC curves, while the quenched phase only presents a glass transition and gives no sign of recrystallization or melting. To explain this effect, it has been proposed that nanocrystalline grains survive in the milled samples, which serve as germs to launch the crystallization process.15 As an example, the PDFs of the quenched phase and the sample milled for 2 h are indeed almost indistinguishable (Figure 12). First, this indicates that their structures are the same: the conformational disorder observed in the quenched phase is also present in the milled sample. It also indicates that if present, nanocrystalline germs should exist in proportions well below 1% of the sample volume, as they are undetectable in the PDF. Another hypothesis is that the remaining cores of the grains, although highly defective and displaying a correlation length on the order of the molecule dimension, retain memory of the LRO phase structural arrangement. Heating the sample could help cure the defects inside these grain cores and promote the recrystallization process. Avoiding recrystallization is a major issue for the use of amorphous materials by the pharmaceutical industry. If the above hypothesis is verified, suppressing the recrystallization process in milled samples could be obtained by ensuring the complete disappearance of the grain cores, by controlling the initial grain

Figure 13. SEM image of a THL sample after 2 h of milling.

it is remarkable to observe that grains showing typically crystalline shapes and sizes of several micrometers are still present in a sample which is seen as completely amorphized by all other techniques. A sketch summarizing our interpretation of the microstructural evolution of sample grains as function of milling time is shown in Figure 14. Due to the milling process, the grains of the starting sample (tmill = 0) will transform to a core−shell type with the SRO J



■

ACKNOWLEDGMENTS The authors acknowledge financial support from the Agence Nationale de la Recherche through the MIPHASOL project (grant number ANR-12-BS08-0008-01).

size, the milling energy, and the temperature. In this respect, it is worth noting that samples milled for 100 h do not display recrystallization effects on heating.15 Probably, for such extensively treated samples, the grain cores have finally completely lost the memory of the original crystalline arrangement. Up to now, almost all investigations of milled organic samples have been carried out ex situ. Despite experimental precautions, it is always difficult to know to what extent the sample has evolved in the time lapse between milling and the actual characterization experimentor even during the latter and what does this evolution consist of. In the case of milled βTHL, our comparison of an in situ and an ex situ synchrotron diffraction experiment indicates that the LRO phase proportion and coherent domain size do not seem to be markedly affected by recrystallization effects over several days. It is worth noting that the present study only deals with relatively long milling times, starting at 30 min of HEM where the sample is close to full amorphization. It may be recalled that in many cases, the early stage of the amorphization kinetics derived from DSC (amplitude of the heat capacity jump at Tg) differs widely from the ones given by the XRPD and NMR measurements (see, for example, ref 33 for the case of lactose). It can be suspected that in the early stage of the milling process two different qualities of amorphous materials could be produced which would be undistinguishable structurally, but would have different molecular dynamics, as detected by DSC. In order to answer this question, the present work will be completed by systematic DSC and diffraction measurements for short tmill values where the microstructural evolution is much faster. However, this was not the main objective of the present study.

■

REFERENCES

(1) Loh, Z. H.; Samanta, A. K.; Sia Heng, P. W. Overview of milling techniques for improving the solubility of poorly water-soluble drugs. Asian J. Pharm. Sci. 2015, 10, 255. (2) Brittain, H. G. Effects of mechanical processing on phase composition. J. Pharm. Sci. 2002, 91, 1573. (3) Murdande, S. B.; Pikal, M. J.; Shanker, R. M.; Bogner, R. H. Solubility advantage of amorphous pharmaceuticals: I. A thermodynamic analysis. J. Pharm. Sci. 2010, 99, 1254. (4) Willart, J. F.; Descamps, M. Solid State Amorphization of Pharmaceuticals. Mol. Pharmaceutics 2008, 5, 905. (5) Zimper, U.; Aaltonen, J.; McGoverin, C.; Gordon, K.; KrauelGoellner, K.; Rades, T. Quantification of Process Induced Disorder in Milled Samples Using Different Analytical Techniques. Pharmaceutics 2010, 2, 30. (6) Graeser, K. A.; Patterson, J. E.; Rades, T. Physical stability of amorphous drugs: Evaluation of thermodynamic and kinetic parameters. J. Pharm. Pharmacol. 2008, 60, 116. (7) Proffen, T.; Billinge, S. J. L.; Egami, T.; Louca, D. Structural Analysis of Complex Materials Using the Atomic Pair Distribution Function − a Practical Guide. Z. Kristallogr. - Cryst. Mater. 2003, 218, 132−143. (8) Egami, T.; Billinge, S. J. L. Underneath the Bragg peaks: structural analysis of complex materials, 2nd ed.; Elsevier: Amsterdam, 2012. (9) Terban, M. W.; Cheung, E. Y.; Krolikowski, P.; Billinge, S. J. L. Recrystallization, Phase Composition, and Local Structure of Amorphous Lactose from the Total Scattering Pair Distribution Function. Cryst. Growth Des. 2016, 16, 210−220. (10) Davis, T.; Johnson, M.; Billinge, S. J. L. Toward Phase Quantification at the Nanoscale Using the Total Scattering Pair Distribution Function (TSPDF) Method: Recrystallization of Cryomilled Sulfamerazine. Cryst. Growth Des. 2013, 13, 4239−4244. (11) Stieger, N.; Liebenberg, W. Recrystallization of Active Pharmaceutical Ingredients, in Crystallization - Science and Technology, Andreeta, M., Ed.; 2012. Dujardin, N.; Willart, J-F.; Dudognon, E.; Hédoux, A.; Guinet, Y.; Paccou, L.; Chazallon, B.; Descamps, M. Solid state vitrification of crystalline α and β -D-glucose by mechanical milling. Solid State Commun. 2008, 148, 78−82. (12) Jeffrey, G. A.; Nanni, R. The crystal structure of anhydrous alpha,alpha-trehalose at −150 degrees. Carbohydr. Res. 1985, 137, 21− 30. (13) Crowe, L. M.; Reid, D. S.; Crowe, J. H. Is Trehalose Special for Preserving Dry Biomaterials? Biophys. J. 1996, 71 (4), 2087−2093. (14) Willart, J. F.; Descamps, M.; et al. Direct crystal to glass transformations of trehalose induced by milling, dehydration and annealing; AIP Conference Proceedings, Sendai, 2008. (15) Willart, J. F.; De Gusseme, A.; Hemon, S.; Odou, G.; Danede, F.; Descamps, M. Direct Crystal to Glass Transformation of Trehalose Induced by Ball Milling. Solid State Commun. 2001, 119 (8−9), 501− 505. (16) Willart, J. F.; De Gusseme, A.; Hemon, S.; Descamps, M.; Leveiller, F.; Rameau, A. Vitrification and Polymorphism of Trehalose Induced by Dehydration of Trehalose Dihydrate. J. Phys. Chem. B 2002, 106 (13), 3365−3370. (17) Surana, R.; Pyne, A.; Suryanarayanan, R. Effect of preparation method on physical properties of amorphous trehalose. Pharm. Res. 2004, 21 (7), 1167−1176. (18) Caron, V. Mécanosynthèse et vitrification à l’état solide d’alliages moléculaires, PhD thesis, Université Lille1, 2006. (19) Zhang, P.; Klymachyov, A. N.; Brown, S.; Ellington, J. G.; Grandinetti, P. J. Solid-State 13C NMR Investigations of the Glycosidic Linkage in Alpha-Alpha’ Trehalose. Solid State Nucl. Magn. Reson. 1998, 12 (4), 221−225.

■

CONCLUSION By using Rietveld and PDF analysis of laboratory and synchrotron powder diffraction data, we have investigated the evolution of microstructure and local structure upon amorphization of β-THL by high energy milling. We confirm the existence of a strong conformational disorder in the glass phase for a quenched sample, and show that a similar disorder is also present in the SRO phase of milled samples. The evolution of the microstructure with milling time can be described by a twophase model in which core−shell grains are formed: an amorphous phase similar to the one obtained by quenching forms a shell around a core consisting of a LRO phase. Due to the progressive increase of defect density, the coherence length of the latter decreases with milling, to eventually become similar to that of the amorphous shell. At that point, both phases become indistinguishable with our techniques. It is tempting to propose that recrystallization might process by curing the defects in the remaining grain cores, explaining why an “amorphous” phase obtained by high energy milling usually recrystallizes upon heating (except for extremely long milling times), whereas one obtained by quenching, though apparently structurally similar, does not. Similar experiments as function of temperature are being analyzed to investigate this matter.

■

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: +33 (0)4 76 88 74 24. Notes

The authors declare no competing financial interest. K



Article

(20) Lerbret, A.; Bordat, P.; Affouard, F.; Descamps, M.; Migliardo, F. How Homogeneous Are the Trehalose, Maltose, and Sucrose Water Solutions? An Insight from Molecular Dynamics Simulations. J. Phys. Chem. B 2005, 109 (21), 11046−11057. (21) Rodríguez-Carvajal, J. Recent Advances in Magnetic Structure Determination by Neutron Powder Diffraction. Phys. B 1993, 192, 55−69. (22) Courbion, G.; Ferey, G. J. Na2Ca3A12F14: A New Example of a Structure with “Independent F-” -A New Method of Comparison between Fluorides and Oxides of Different Formula. J. Solid State Chem. 1988, 76, 426−431. (23) Hodeau, J.-L.; Bordet, P.; Anne, M.; Prat, A.; Fitch, A. N.; Dooryhee, E.; Vaughan, G.; Freund, A. K. Nine-Crystal Multianalyzer Stage for High-Resolution Powder Diffraction between 6 and 40 keV. Proc. SPIE 1998, 3448, 353−361. (24) Qiu, X.; Thompson, J. W.; Billinge, S. J. L. PDFgetX2: A GUIDriven Program to Obtain the Pair Distribution Function from X-Ray Powder Diffraction Data. J. Appl. Crystallogr. 2004, 37, 678−678. (25) Halasz, I.; Jones, W.; Štrukil, V.; Halasz, I.; et al. Real-Time insitu Powder X-Ray Diffraction Monitoring of Mechanochemical Synthesis of Pharmaceutical Cocrystals. Angew. Chem., Int. Ed. 2013, 52 (44), 11538−11541. (26) Halasz, I.; Kimber, S. a J.; Beldon, P. J.; Belenguer, A. M.; Adams, F.; Honkimäki, V.; Nightingale, R. C.; Dinnebier, R. E.; Frišcǐ ć, T. In-situ and Real-Time Monitoring of Mechanochemical Milling Reactions Using Synchrotron X-Ray Diffraction. Nat. Protoc. 2013, 8 (9), 1718−1729. (27) Hammersley, A. P.; Svensson, S. O.; Hanfland, M.; Fitch, A. N.; Hausermann, D. Two-Dimensional Detector Software: From Real Detector to Idealised Image or Two-Theta Scan. High Pressure Res. 1996, 14, 235−248. (28) Willart, J. F.; Dujardin, N.; Dudognon, E.; Danède, F.; Descamps, M. Amorphization of sugar hydrates upon milling. Carbohydr. Res. 2010, 345, 1613−1616. (29) Shalaev, E. Y.; Byrn, S. R.; Zografi, G. Single-Phase and Heterophase Solid-State Chemical Kinetics of Thermally Induced Methyl Transfer in Tetraglycine Methyl Ester. Int. J. Chem. Kinet. 1997, 29, 339−348. (30) http://code.google.com/p/debyer (31) Kodama, K.; Iikubo, S.; Taguchi, T.; Shamoto, S. Finite Size Effects of Nanoparticles on the Atomic Pair Distribution Functions. Acta Crystallogr., Sect. A: Found. Crystallogr. 2006, A62, 444−453. Howell, R. C.; Proffen, T.; Conradson, S. D. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 73, 094107. (32) Brown, G. M.; Rohrer, D. C.; Berking, B.; Beevers, C. a.; Gould, R. O.; Simpson, R. The Crystal Structure of A,α-Trehalose Dihydrate from Three Independent X-Ray Determinations. Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem. 1972, 28 (11), 3145−3158. (33) Caron, V.; Willart, J-F.; Lefort, R.; Derollez, P.; Danède, F.; Descamps, M. Solid state amorphization kinetic of alpha lactose upon mechanical milling. Carbohydr. Res. 2011, 346 (16), 2622−2628.

L


Solid State Amorphization of Î²-Trehalose: A Structural Investigation

Recommend Documents