Isomorphism, disorder, and hydration in the crystal structures of racemic and single-enantiomer carvedilol phosphate

DOI: 10.1021/cg100209v

Isomorphism, Disorder, and Hydration in the Crystal Structures of

Racemic and Single-Enantiomer Carvedilol Phosphate

2010, Vol. 10

2713–2733

Frederick G. Vogt,*^,†Royston C. B. Copley,^‡Ronald L. Mueller,^†,¥Grant P. Spoors,^†

Thomas N. Cacchio,^†Robert A. Carlton,^†Lee M. Katrincic,^§James M. Kennady,^§

Simon Parsons, and Olga V. Chetina^#

^†Preclinical Development, GlaxoSmithKline plc., P.O. Box 1539, King of Prussia, Pennsylvania 19406,

^‡Molecular Discovery Research, GlaxoSmithKline plc., New Frontiers Science Park, Third Avenue,

Harlow, Essex CM19 5AW, U.K., ^§Preclinical Development, GlaxoSmithKline plc., 5 Moore Drive,

Research Triangle Park, North Carolina 27709, School of Chemistry, The University of Edinburgh,

West Mains Road, Edinburgh, Scotland EH9 3JJ, U.K., and ^#Chemistry Department, University of

Durham, South Road, Durham DH1 3LE, U.K.

Received February 9, 2010; Revised Manuscript Received April 8, 2010

ABSTRACT: Understanding the crystalline structure of racemic carvedilol phosphate hemihydrate presents several challenges

that were overcome using a combination of single-crystal X-ray diffraction, solid-state NMR (SSNMR), and other analytical

techniques. Initial attempts to obtain a crystal structure were hampered by difficulties with twinning and problematic disorder in

the final refinements. Multinuclear SSNMR analysis localized the disorder to portions of the molecule near the chiral center. As

a result, single-enantiomer carvedilol phosphate was prepared and was found to crystallize in a phase that was isomorphous with

the racemate, while SSNMR spectra of the single enantiomers did not contain the disorder observed in the racemate. The single-

crystal X-ray structure of the (R)-enantiomer was solved and used as a starting point to successfully progress the solution of the

disordered racemic crystal structure. Thermal analysis and construction of a phase diagram, along with crystallographic

and spectroscopic analysis, found the crystal structure of the racemate to be a solid solution of (R)- and (S)-enantiomers, with

the conformation of the molecule adjusting to fit. The crystal structures show the stoichiometry of the both the racemate and

(R)-enantiomer to be a hemihydrate. The phase isomorphically dehydrates below relative humidity values of 1% and above

temperatures of 125 ꢀC as assessed by water vapor sorption studies, powder X-ray diffraction, and SSNMR. Single-crystal

diffraction detected significant changes in the unit cell dimensions as the phase dehydrated, which was related to the visual

appearance of opacity in a single crystal of the (R)-enantiomer. The mechanism of water incorporation was further probed

spectroscopically via exchange with deuterium, ¹⁷O-, and ¹⁸O-labeled water; the results suggest that dehydration and

rehydration likely proceed via narrow tunnels in the crystal structure, combined with the formation of fissures in the crystal.

²H SSNMR experiments showed that the water does not engage in solid-state jump motion even at higher temperatures.

Introduction

potentially occur in different states showing a well-defined

stoichiometric relationship between the water and the parent

molecule, while in other cases, the water content of some solid

hydrates can be varied as a nonstoichiometric fraction by

changing the surrounding humidity and temperature.^6-10

Unlike the usually stoichiometric isolated site hydrates, non-

stoichiometric hydrates usually contain stacks or chains of

water molecules in the crystal structure, and solvation and

desolvation processes occur readily with little or no change to

the crystal structure. The water channels can fill up to reach a

maximum stoichiometry, or the crystal structure can expand to

accommodate more water.⁵Various designations have been

given to these types of systems: the basic labels of “variable”,

“non-stoichiometric”, “tunnel”, or “channel” hydrates, the

more exact terms of “isomorphic solvate” (or desolvate),⁶or

The development of a pharmaceutical as an oral dosage

form requires control over solid-state properties, inparticular,

the solid phase of the active ingredient, including its salt,

polymorph, and solvation state.^1,2The solid phases present in

a dosage form or produced by a manufacturing process can be

a criticalattributeiftheyaffectdissolution andbioavailability,

product stability and shelf life, and manufacturing perfor-

mance.³Most of the chiral pharmaceuticals currently in

development are progressed as single enantiomers or single

diastereomers. However, in the case of racemic mixtures of

enantiomers, the two enantiomers have several options and

can, for example, crystallize either as conglomerates, in which

domains of a particular isomer crystallize separately, or as a

crystalline racemate containing both enantiomers in close

proximity in the unit cell.⁴The solid form of a racemic drug

molecule may be further complicated by the existence of

hydrates, a common occurrence in organic molecular solids.⁵

Like nonsolvated forms, hydrates can be polymorphic and

exist in forms with different molecular conformations, pack-

ing motifs, and supramolecular interactions. Hydrates may

the simple addition of “ xH₂O” to the end of a molecular

3

formula. In the characterization of racemic and single enantio-

mer structures as well as isomorphic desolvation, tools such as

single-crystal X-ray diffraction (SCXRD), powder X-ray dif-

fraction (PXRD), solid-state NMR (SSNMR), and gravimetric

vapor sorption (GVS) as well as thermal methods are invaluable

to gain a structural understanding of macroscopic properties

such as the nature of hydration and dehydration processes.^5-10

In the present work, the solid-state structure of carvedilol

phosphate hemihydrate is studied. The free base of racemic

*Corresponding author. E-mail: Fred.G.Vogt@gsk.com.

^¥Present address: West Pharmaceutical Services, 101 Gordon Drive, Exton,

PA 19341.

r

2010 American Chemical Society

Published on Web 05/03/2010

pubs.acs.org/crystal

2714 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

carvedilol exhibits β- and R₁-receptor blocking activity and

antioxidant effects and is used for the treatment of high blood

pressure and congestive heart failure, while its enantiomers

block different receptors.^11-13The chemical structure of the

phosphate salt of carvedilol (Scheme 1), with the atomic

numbering scheme used here, is shown below:

(S)-enantiomers of carvedilol were resolved by preparative chiral

liquid chromatography (LC) using a scaled-up version of the

analytical method described below. Samples of I, I-(R), and I-(S)

were prepared by charging a 250 mL flask with 6.2 g of carvedilol

free base, 55 mL of acetone, and 1.55 mL of water. The mixture was

stirred and heated to 45 ꢀC and became clear after 10 min. The clear

solution was allowed to cool to room temperature, and 1.8 g of 85%

H₃PO₄and 5.9 mL water were simultaneously added. With gentle

stirring, the solution became cloudy after 15 min. After 1 h, the

material was filtered, washed with acetone, and dried in an oven at

50 ꢀC at reduced pressure (0.8 atm), for a typical yield of 95%.

Various ratios of the (R)- and (S)-enantiomers were prepared

as phosphate salts in a similar manner using appropriate initial

ratios of the enantiomers and racemate. Grinding experiments were

performed with a Restch MM200 mixer mill using 200 mg of solid

powder and 3 drops of methanol. The sample was ground using a

stainless steel ball at a frequency of 30 Hz for 20 min. Powders

of I were dehydrated in a vacuum oven at temperatures of 65-70 ꢀC.

Deuterium exchange was carried out by rapidly rehydrating a

vacuum-dried sample of I in a humidity chamber containing

a saturated solution of NaCl in D₂O (Sigma-Aldrich, St. Louis,

MO, USA) at 75% RH.²⁰Samples that were preconditioned in

a 75% RH H₂O chamber prior to deuterium exchange were also

used. Analogous procedures were used to exchange ¹⁷O and

¹⁸O-labeled water (Cambridge Isotope Laboratories, Andover,

MA, USA) into I.

Scheme 1

Carvedilol contains a chiral center (C16) and is shown as a

racemate in I. The (R)- and (S)-enantiomers of carvedilol

phosphate hemihydrate, discussed extensively in the following

sections, are referred to as compounds I-(R) and I-(S). The

separation of the enantiomers has been accomplished using

liquid chromatography methods.^14,15

Characterization of Purity and Structure. A chiral LC method was

used to assess the enantiomeric content of the materials using an

Agilent 1100 LC system (Agilent Technologies, Palo Alto, CA,

USA). The chiral method used a ChiralPak AD column (250 ꢀ

4.6 mm, 5 μm particle size), an isoscratic mobile phase consisting of

82% hexanes, 12% ethanol, 6% methanol, and 0.1% trifluoroacetic

acid and UV detection at 250 nm. The chemical purity of batches of I

used in this work was confirmed to be greater than 98% w/w using

an achiral LC method and an Agilent 1100 LC system. This method

used a Supelco pKb-100 column (150 ꢀ 4.6 mm, 5 μm particle size),

a mobile phase A with 80% water and 20% acetonitrile, a mobile

phase B with 75% acetonitrile and 25% water, both with 0.05% v/v

trifluoroacetic acid, a flow rate of 1.5 mL/min and UV detection at

225 nm. Mobile phase B was ramped from 2 to 7% over 10 min, then

to 86% over 18 min, prior to a return to 2% for 10 min. The chemical

purity of the I-(R) and I-(S) was also tested using this method, and

was greater than 95%. The purity of each sample was also con-

The work described here begins with attempts to determine

the crystal structure of racemic I, with the purpose of obtain-

ing the structure and confirming the expected protonation

state and also understanding the nature of the crystalline

water. After crystallizing several forms that did not represent

the bulk material, the correct solid-state form was finally

obtained, but the crystal structure solution was highly com-

plex, involving both twinning and serious disorder that was

difficult to explain. Because of these challenges, a variety of

additional structural studies were initiated. As part of these

structural studies, the disorder was further characterized by

SSNMR, which led directly to the preparation of single-

enantiomer phosphate salts. These were quickly recognized

as being isomorphous¹⁶with racemic I and were subsequently

solved by SCXRD and were then used to rationalize and

complete the structure refinement of the racemate. Thermal

methods were used to study the relationship between the

enantiomer structures and the racemate, leading to the finding

that the (R)- and (S)-domains in the racemate were best

described not as conglomerates but as a solid solution. The

occurrence of solid solutions in organic molecular crystals

(which are also referred to as pseudoracemates) is relatively

uncommon compared to the more frequent cases of racemate

structures and conglomerate formation (racemic mixtures).^17,18

The hydration state of I and its enantiomers, which were

found to be hemihydrates by SCXRD, was also of interest,

particularly because of the unusual ability of this phase to

undergo reversible isomorphic dehydration under extreme

conditions. Using complementary methods, including GVS,

SSNMR, Raman spectroscopy, and X-ray diffraction analyses,

the water of hydration was found to be unusually tightly bound

but still capable of undergoing a process similar to isomorphic

desolvation. The phases can exchange a significant amount of

water with surrounding atmospheric water even under ambient

conditions. The evidence collected indicates that water ex-

change into the crystal structure is facilitated by surprisingly

narrow channels and the effects of unit cell deformation.

1

firmed by H solution-state NMR experiments in DMSO-d₆solu-

tion using a Bruker DRX-700 spectrometer equipped with a

broadband inverse probe. Solution-state NMR assignments were

made using standard 1D and 2D NMR experiments.²¹Solution-

state NMR spectra were referenced using an internal tetramethyl-

silane (TMS) standard for ¹H and ¹³C experiments, an external 90%

formamide standard for ¹⁵N experiments, and an external 85%

H₃PO₄standard for ³¹P experiments.²²Optical rotation at 25 ꢀC

([R]_D) was measured at þ7ꢀ for the (R)-enantiomer and -11ꢀ for the

(S)-enantiomer phosphate salts using a 241 MC polarimeter

(Perkin-Elmer, Waltham, MA, USA).

Solid-State NMR Spectroscopy. Solid-state NMR experiments

were performed on Bruker Avance 400 and 500 spectrometers

operating at a ¹H frequency of 399.87 MHz (9.4 T) and 500.13

MHz (11.7 T) respectively, equipped with variable temperature

systems including a BCU-05 chiller, a liquid nitrogen heat exchan-

ger and a liquid nitrogen boil-off gas system. Bruker MAS-II rate

controllers were used to control spinning speeds to within (2 Hz of

the set point, and zirconia MAS rotors were used. ¹³C spectra were

externally referenced to TMS using hexamethylbenzene.²³The ¹⁵

N

spectra were externally referenced to a sample of NH₄Cl.^{24 1}H

spectra were referenced by addition of a small amount of liquid

TMS to the sample and repeating the analysis, while ³¹P spectra

were referenced to liquid 85% H₃PO₄in a similar manner.^{22 13}

C

cross-polarization spectra were obtained with a Bruker 4-mm triple

resonance probe tuned to ¹H, ¹³C, and ¹⁵N frequencies and spinning

at an MAS rate υ_rof 8 kHz. The sample was restricted to the center

of the 4-mm volume rotors to maximize RF homogeneity. A linear

power ramp from 75 to 90 kHz was used on the ¹H channel to

enhance CP efficiency.²⁵Spinning sidebands in 1D ¹³C spectra were

Experimental Section

Preparation of Materials. Racemic carvedilol free base was

synthesized using previously reported procedures.¹⁹The (R)- and

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2715

eliminated by a five-pulse total sideband suppression (TOSS)

sequence with a 243-step phase cycle, with a total of either 1944 or

3888 scans acquired for the spectra shown.²⁶Proton decoupling was

performed at an RF power of 105 kHz using the SPINAL-64 pulse

sequence.²⁷Edited ¹³C spectra containing only quaternary aromatic

signals were obtained using dipolar dephasing (also known as

nonquaternary suppression or interrupted decoupling) during the

TOSS period and three additional rotor periods using a shifted echo

pulse sequence.²⁸1D ¹⁵N NMR spectra were obtained with a Bruker

7-mm double resonance probe using the basic CP-MAS pulse

sequence with a 5-ms contact time and a 10-s relaxation delay,

and using a ¹H decoupling power level of ∼65 kHz. A total of 4096

transients were averaged for the ¹⁵N spectra. Single-pulse ¹H

experiments were performed using a Bruker 2.5-mm double-reso-

nance probe tuned to ¹H and ³¹P frequencies and spinning at υ_r=

35 kHz. A 2.5-μs excitation pulse and a 30-s relaxation delay were

used and a total of 32 transients were acquired. ³¹P MAS and CP-

MAS spectra were acquired on the 2.5 mm probe using a 2.5-μs

excitation pulse and a 2 ms contact time, respectively. 2D rotor-

displacement parameters. The I-(R) structure has two molecules in

the asymmetric unit (Z⁰= 2), as discussed in detail below, and the

numbering scheme was expanded accordingly beyond that shown in

Scheme 1. Atoms C51 to C68 in one of the two molecules in the

asymmetric unit were disordered over two orientations, with refined

complementary site occupancy factors. The atomic coordinates

were refined with restraints to maintain essentially common bond

lengths and angles. The atoms of the major component were refined

anisotropically, while the minor component employed refined iso-

tropic terms. In the structure of I, with Z⁰= 1, atoms C1 to C18 were

disordered over three orientations with refined site occupancy

factors. Geometric restraints were applied to ensure a chemically

sensible and comparable result for each component. An isotropic

parameter was refined as a free-variable for the non-hydrogen

atoms in each component. For both structures, hydrogen atoms

were generally included in calculated positions, with riding coordi-

nates and an isotropic displacement parameter as a multiple of that

of the bonded non-hydrogen atom. Hydrogen atoms bonded to

nitrogen and oxygen atoms in the ordered parts of each structure

had their positions refined with suitable distance restraints and

employed an isotropic atomic displacement parameter. No hydro-

gen atom could be located for the hydroxyl group in the various

disordered components found in I. Additional information on the

single crystal diffraction studies may be found in the crystallo-

graphic information files deposited with the Cambridge Structural

Database (CSD).³⁵

The dehydration and rehydration studies on I-(R) were per-

formed in situ on the SMART 6000 diffractometer using the same

crystal as used for the structure analysis. Dehydration was carried

out at 295 K using the nitrogen stream, which is near a 0% RH

environment. Rehydration was achieved by redirecting the flow of

the nitrogen stream, exposing the crystal to the ambient laboratory

conditions of 24 ꢀC ( 3 ꢀC and 40% ( 10% RH while inside the

diffractometer enclosure. The time-lapse footage was generated

using the Microsoft Windows Movie Maker program (V5.1, Win-

dows XP) from images taken with the diffractometer microscope.

The time between images was calculated and incorporated into the

movie, with 1 s in the movie representing 1 min in real time (see

Supporting Information).

1

synchronized H double-quantum broadband back-to-back (DQ-

BABA) MAS experiments were performed with 2.5-mm probes at

υ_r= 35 kHz using two rotor periods of double-quantum excitation

and two rotor periods of reconversion.²⁹A double-quantum POST-

C7 experiment was performed using ³¹P RF field strengths of 70

kHz with υ_r= 14 kHz and a CP preparation period.³⁰2D CP

heteronuclear correlation (CP-HETCOR) and heteronuclear multi-

ple quantum coherence based on J-coupling (MAS-J-HMQC)

experiments between ¹H and ¹³C or ³¹P nuclei made use of

frequency-switched Lee-Goldburg (FSLG) homonuclear decou-

pling at 100-105 kHz with υ_rranging from 12.5 to 15 kHz.^{31,32 1}

H

T₁measurements were performed using saturation recovery

sequences with a 100-pulse saturation comb via both direct ¹H

observation and indirect ¹³C detection.³³The sample chamber

temperature for MAS experiments was maintained at 273 K using

a chiller except where otherwise noted, to minimize the effects of

frictional heating. Static ²H NMR spectra were obtained at 9.4 T

with a Bruker wide-line probe using a quadrupolar echo pulse

sequence.³⁴The samples were loaded into the center of 5-mm

glass tubes using Teflon spacers and caps. Static ¹⁷O NMR

spectra were obtained using a spin echo sequence at 11.7 T using

a Bruker wide-line probe and a similar experimental arrange-

ment.

PXRD patterns were obtained under ambient conditions using

X’Pert Pro diffractometers equipped with X’Celerator Real Time

Multi-Strip (RTMS) detectors (PANalytical B. V., Almelo, The

Netherlands). For flat-plate reflection PXRD analysis, samples

were flattened onto a zero-background silicon holder and run

immediately after preparation. A continuous 2θ scan range of 2ꢀ

X-ray Diffraction. Single crystals of I-(R) and I were prepared by

seeding saturated solutions of the compounds in aqueous THF and

aqueous methanol, respectively. The crystal and molecular struc-

tures were determined from three-dimensional X-ray diffraction

data. Crystals were flash frozen and held at the data collection

temperature of 90(2) K using an Oxford Cryosystems 600 Series

Cryostream. All diffraction measurements were made using a

Bruker-AXS SMART 6000 diffractometer with graphite mono-

˚

to 70ꢀ was used with a Cu-KR (1.5418 A) radiation source and a

generator voltage of 45 kV and current of 40 mA. A step size of

0.0084ꢀ 2θ was used and individual patterns required 30 min to

obtain. The goniometer radius was 240 mm. Soller slits (incident

and diffracted) were set to 0.01 radians and the divergence slit

(incident) and antiscatter slits (incident and diffracted) were set to

1/2ꢀ. Flat-plate reflection samples were rotated at 25 rpm. Capillary

PXRD patterns were also collected under ambient conditions using

an X’Pert Pro diffractometer equipped with an incident beam

elliptical focusing mirror and X’Celerator RTMS detector with

the same conditions as above, except for the following changes.

The powder sample was filled into a 0.7-mm borosilicate glass

capillary. Soller slits (incident and diffracted) were set to 0.02

radians. An effective scan rate of 0.0625ꢀ 2θ per min was used.

The capillary spinner was rotated at 300 rpm. Variable humidity

(VH) and variable temperature (VT) PXRD patterns were mea-

sured on an Advance D8 diffractometer using similar conditions

(Bruker AXS, Madison, WI, USA). A stainless steel sample holder

and an Anton-Paar TTK450 variable temperature/humidity stage

were used. Temperature was controlled to within (2 ꢀC and

humidity was controlled to within (2% RH. Evaporated liquid

nitrogen was used to cool the stage to <100 K for low-temperature

experiments. Samples on the Advance D8 system were not rotated

and measurement of each pattern required 30 min.

˚

chromated Cu-KR radiation (λ = 1.54178 A) from a normal focus

sealed tube source. The instrument was controlled using the

SMART software (V5.632). The distance between the crystal and

the detector was 5.0 cm. Frames were collected with the detector set

in two positions (2θ = -40.00ꢀ and -108.00ꢀ). The low angle

exposure times were 6 and 12 s for I-(R) and I respectively and 24 s

for the high 2θ detector position in both cases. Each frame covered

0.20ꢀ in ω. Data integrations were carried out using the SAINT

program (V6.45A), with refining box sizes. For the twinned crystal

of I, two orientation matrices were used, with common unit cell

dimensions for both domains. The twin law relating the orientations

of the two domains was readily obtained from the integration

software. The final unit cell parameters were taken from the

integration output. A Gaussian face-indexed absorption correction

was applied to the data from I-(R) using the SHELXTL program

(V5.12). The data were also subsequently scaled using the SADABS

program (V2.1). The absorption correction for I was carried out

using the TWINABS program (V1.05).

The structures of I-(R) and I were solved by direct methods and

refined by full-matrix least-squares procedures that minimized the

PXRD data were refined and analyzed using the Pawley refine-

ment procedures in the Materials Studio Reflex software package,

version 4.2 (Accelrys, San Diego, CA, USA). Additional PXRD

data analysis including background subtraction was performed

2 2

function Σw(F_o²- F_c) using the Bruker-AXS SHELXTL software

package (V6.12). For both structures, the coordinates of ordered

non-hydrogen atoms were freely refined, using anisotropic atomic

2716 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

Figure 1. (a) ¹³C CP-TOSS spectra (υ_r= 8 kHz) of I, I-(R), and I-(S), obtained at 273 K and 9.4 T. (b) ¹³C CP-TOSS spectra of the same

samples obtained under the same conditions but with dipolar dephasing. Key features in the enantiomer spectra (denoted with arrows) are

indicative of disorder in the isomorphous racemate as discussed in the text.

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2717

Figure 2. (a) ¹⁵N CP-MAS spectra (υ_r= 5 kHz) of samples of I, I-(R), and I-(S) obtained at 273 K and 9.4 T. (b) ¹H MAS spectra (υ_r= 35 kHz)

of the same samples under the same conditions.

using the HighScore Plus software package, version 2.1 (PANalytical

B. V., Almelo, The Netherlands).

32 scans were averaged at a resolution of 2 cm^-1for each of the

spectra, requiring approximately 10 min. Spectra were obtained

using a single-bounce attenuated total reflectance (ATR) accessory

with a diamond window. Samples were covered with a glass slide

and pressed against the window for analysis.

Other Physical Methods. GVS experiments were performed using

a DVS-1000 instrument (Surface Measurement Systems, Ltd., UK).

Samples were studied over a humidity range of 0% to 90% RH at

25 ꢀC. Each humidity step was made if less than 0.0025% weight

change occurred over 10 min, with a maximum hold time of 3 h.

Differential scanning calorimetry (DSC) was carried out on a Q1000

system (TA Instruments, New Castle, DE, USA). A N₂flow rate of

50 mL/min was used. Sample sizes ranged from 0.5 to 3.0 mg, and

the heating rate was 10 ꢀC/min. Samples were run in aluminum

pans, and to minimize the effects of hydration prior to the

Vibrational Spectroscopy. Raman spectra were obtained with a

Thermo Nicolet FT-Raman 960 spectrometer equipped with a Nd:

YAG near-infrared laser (1064 nm wavelength), an InGaAs detec-

tor, and a CaF₂beamsplitter (ThermoFisher, Madison, WI, USA).

The laser power was 500 mW. Samples were analyzed directly in

their glass vials, which were sealed immediately upon their removal

from either humidity chambers or the drying oven. A total of 256

scans were averaged at a resolution of 2 cm^-1for each of the Raman

spectra (requiring approximately 16 min). Laser-induced sample

decomposition and fluorescence were not observed.

Infrared spectra were obtained using a Perkin-Elmer Spectrum

One spectrometer equipped with a DTGS (deuterated triglycine

sulfate) detector (Perkin-Elmer, Waltham, MA, USA). A total of

2718 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

Figure 3. (a) ³¹P CP-MAS spectra (υ_r= 25 kHz) of samples of I, I-(R), and I-(S) obtained at 273 K and 9.4 T. (b) Reflection PXRD patterns,

measured at 298 K and ambient humidity with a flat-plate sample holder, of samples of I, I-(R), and I-(S).

experiment, the DSC sample was stored in its aluminum pan in the

75% RH chamber for the six-week equilibration period. As soon as

the chamber was opened, a lid was pressed on the pan and the

sample was analyzed. Thermogravimetric analysis (TGA) was also

performed on the samples using a TA Instruments Q500 system,

again using a 10 ꢀC/min heating rate. TGA sample sizes were in the

range of 5-20 mg. Coulometric Karl Fischer (KF) titrations were

performed at room temperature using a Metrohm 756 KF titrator

and Hydranal Coulmat reagent (Reidel-deHaan). The analysis was

started within a minute of opening each vial. The system was

checked before and after the analysis using a Hydranal 0.101%

suitability sample traceable to NIST SRM 2890. All reported results

were the average of two titrations. Surface area measurements using

the Brunauer-Emmett-Teller adsorption method with nitrogen

gas were performed with a Tristar instrument (Micromeritics,

Norcross, GA, USA), using the average of six experiments.

study and was chronologically the first work attempted.

Because of the complex disorder and twinning found for I,

this structure determination presented a significant challenge

and could not initially be completed. SSNMR methods were

applied in an attempt to support the crystal structure analysis

and learn the nature of the disorder; from this work came the

valuable observation of unusual disorder in the racemate

structure, suggesting further exploration of the single enan-

tiomers. To illustrate this finding, the ¹³C CP-TOSS spectra

of I, I-(R) and I-(S) are shown in Figure 1a. Spectral assign-

ments have been made using the solution-state NMR assign-

ments given in the Supporting Information as a starting

point, with more detailed assignments made using the dipolar-

1

dephased spectra in Figure 1b and H-¹³C CP-HETCOR

data (discussed below). The spectra shown in Figure 1 im-

mediately suggest that the crystal structure of the racemate is

isomorphous with that of the single enantiomers but con-

tains additional disorder manifested by peak broadening

(denoted by arrows in the figure). The ¹³C assignments

Results and Discussion

Initial Solid-State NMR and PXRD Studies. Determina-

tion of the crystal structure of I was the initial goal of this

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2719

indicate that the disorder is present in the vicinity of a

number of positions, most notably C4 and C11. Since many

of these positions were near to the chiral center in carvedilol,

it was hypothesized that the removal of the disorder con-

tribution from the racemic chiral center might enable a better

SCXRD result. Furthermore, the dipolar-dephased spec-

trum indicated that either the number of molecules in the

asymmetric unit (Z⁰) was greater than 1, or that disorder

was resulting in multiple carbon environments because of

the clear assignment of C4 and C11. Up until this point,

attempts at structural solutions on the twinned racemate

crystal had centered on Z⁰= 1 structures with significant

disorder.

Additional comparisons of the I, I-(R), and I-(S) samples

were made using other relevant NMR-active nuclei. The ¹⁵N

CP-MAS spectra of these samples are compared in Figure

2a. Again, through their similarity, the spectra show indica-

tions of isomorphism between the structures. A significant

splitting observed for the N9 position in the enantiomers is

obscured in the spectrum of the racemate, suggesting dis-

1

order involving this position. H spectra of these samples,

compared in Figure 2b, show the expected similarity but do

not offer any further insight into the disorder, as the resolu-

tion of the ¹H spectra even at υ_r= 35 kHz is insufficient to

elucidate these effects. However, a deshielded signal at 13.0

ppm is clearly observed, which is useful for confirming

structural features and will be further discussed below. The

³¹P CP-MAS spectra of the I, I-(R), and I-(S) samples, shown

in Figure 3a, indicate that the disorder reaches the phosphate

counterion in the structure. The sample of I was analyzed

1

using a ¹³C-detected H saturation-recovery T₁experiment

to ensure all peaks had the same ¹H T₁value within the error

of the experiment; a uniform ¹H T₁of 7.5 ( 0.2 s was

observed across the spectrum, offering no evidence of phase

separation.³³The PXRD patterns of the three materials, in

Figure 3b, show no detectable difference except for minor

intensity variations caused by preferred orientation. The

combined data from multiple NMR spectra and PXRD thus

indicate that the I, I-(R), and I-(S) crystal structures are

isomorphous with one another, with the structure of I

containing disorder elements near the chiral center and

phosphate group but also extending throughout the mole-

cule. Single crystals of I-(R) were then grown and analyzed

by SCXRD, as discussed in the next section. Further

SSNMR analysis of the detailed structural features of I is

revisited in a later section.

SCXRD Analysis of the (R)-Enantiomer. A crystal of I-(R)

that was a colorless plate with dimensions of 0.28 ꢀ 0.28 ꢀ

0.12 mm was successfully studied, and the asymmetric unit

was found to contain two cations and two dihydrogen

phosphate anions, together with half of two water molecules.

The half occupancy results from the water molecules being

positioned on crystallographic 2-fold axes and consequently

being shared between two asymmetric units. In the remain-

der of this paper, the two independent cations, anions, and

water molecules will be referred to as A and B, as shown in

Figure 4a,b. An expanded numbering scheme beyond that

given in Scheme 1 is used for the second molecule in the

asymmetric unit of this Z⁰= 2 structure, as shown in

Figure 4b. The diffraction study allowed the unambiguous

assignment of absolute stereochemistry. The chiral centers in

both independent cations were found to have the (R)-con-

figuration. The final R₁for the I-(R) structure was 3.53%

and a summary of the study is presented in Table 1.

Figure 4. (a) A view of a cation A, anion A, and water molecule A

from the single enantiomer crystal structure of I-(R). (b) A view of a

cation B, anion B, and water molecule B from the same structure,

showing only the major component of the disorder for cation B and

showing the expanded numbering scheme. Anisotropic atomic

displacement ellipsoids for the non-hydrogen atoms are shown at

the 50% probability level. Hydrogen atoms are displayed with an

arbitrarily small radius. Only half of the water molecule shown is in

the same asymmetric unit as the cation and anion.

Several aspects of the I-(R) crystal structure are note-

worthy. The fused ring system and part of the central linker

of cation B were modeled as being disordered over two

positions as shown in Figure 5a. The site occupancy factor

associated with the major component was 0.784(5). The need

for a second orientation was recognized when the ordered

model could not be used to explain the electron density

associated with C65⁰. For the other disordered atoms, the

minor component was initially disguised by the anisotropic

atomic displacement parameters, with corresponding atoms

being relatively close together. Plane normals for the fused

ring systems in the major and minor components are inclined

at just 7.6(3)ꢀ. The conformation of the linker in cation A

does not match that observed for either component of the

disorder in cation B. Nonetheless, the effect of the linker is

very similar in each case, and the crystal structure can be

2720 Crystal Growth & Design, Vol. 10, No. 6, 2010

Table 1. Summary of the SCXRD Analysis of I and I-(R)

Vogt et al.

structure

racemic I

[C₂₄H₂₇N₂O₄]^þ(H₂O₄P)^-0.5 H₂O

(R)-enantiomer I-(R)

[C₂₄H₂₇N₂O₄]^þ(H₂O₄P)^-0.5 H₂O

moiety formula

empirical formula

formula weight

temperature (K)

wavelength (A)

crystal size

3

C₂₄H₃₀N₂O_8.50

513.47

90(2)

P

C₂₄H₃₀N₂O_8.50

P

513.47

90(2)

1.54178

0.28 ꢀ 0.28 ꢀ 0.12 mm

colorless plate

monoclinic

˚

1.54178

0.36 ꢀ 0.32 ꢀ 0.04 mm

colorless plate

monoclinic

crystal habit

crystal system

space group

unit cell dimensions

C2/c

C2

˚

a

b

c

R

β

γ

26.6624(7) A

12.2211(3) A

˚

16.3184(4) A

5067.7(2)

8

1

1.346

1.419

2168

90ꢀ

26.7679(6) A

12.2428(3) A

˚

16.2786(4) A

5069.3(2)

8

2

1.346

1.418

2168

90ꢀ

107.6220(15)ꢀ

90ꢀ

108.1488(12)ꢀ

90ꢀ

3

˚

volume (A )

molecules per cell (Z)

molecules in the asymmetric unit (Z⁰)

calculated density (Mg/m³)

absorption coeff, μ (mm^-1

F₀₀₀

)

theta range (ꢀ)

index ranges

3.48 to 73.14

-32 e h e 31

0 e k e 15

0 e l e 19

5075

0.0000

96.4

semiempirical

0.9454 and 0.6291

5075/188/348

1.447

R₁= 0.1039

wR₂= 0.3203

R₁= 0.1128

wR₂= 0.3295

not applicable

2.86 to 73.26

-32 e h e 32

-13 e k e 14

-15 e l e 20

24604

9105

0.0131

98.0

integration

0.8537 and 0.6852

9105/163/767

1.042

R₁= 0.0353

wR₂= 0.0960

R₁= 0.0357

wR₂= 0.0964

0.001(13)

measured reflections

independent reflections

R(int)

coverage of independent reflections (%)

absorption correction

min/max transmission

data/restraints/parameters

goodness of fit on F²

final R indices for

I > 2σ(I) data

final R indices for

all data

absolute structure parameter

largest diff peak and hole

-3

˚

1.080 and -0.416 e A

0.390 and -0.226 e A

described as pseudocentrosymmetric, with the rings of the

independent cations, the anions, and the water molecules

being related by approximate inversion symmetry as seen at

the center of Figure 5b. This symmetry breaks down in the

linker since only one enantiomer is present. The intermole-

cular interactions associated with the corresponding inde-

pendent cations, anions, and water molecules are broadly

related by the pseudoinversion symmetry. However, one

important difference relates to the interactions of the hydro-

xyl groups in cation A and cation B. For the former, a strong

and direct interaction to O33 is observed. Since the anions

are related by the approximate inversion symmetry but the

chiral center and the attached hydroxyl group are not, an

analogous interaction is not observed for cation B. Instead,

the hydroxyl group of both components has a longer,

bifurcated interaction with O82 and O83. The lack of a more

direct interaction may explain why cation B is disordered in

this region.

axis. With four hydrogen bonding interactions in total,

the water can be thought of as being strongly bound in

the structure. The significance of the strongly bound water

will be discussed in conjunction with SSNMR and GVS

measurements in a later section. As expected, given the

pseudosymmetry, the other water molecule has an ana-

logous environment between cation B and its symmetry

equivalent.

In summary, the untwinned I-(R) crystal contained dis-

order that was easily modeled, allowing a clear indication of

the atomic connectivity, chemical composition, and hydra-

tion state, all of which are of value in solving the racemate

structure. The structure has Z⁰= 2, in full agreement with

the splitting observed in the ¹³C and ¹⁵N SSNMR spectra for

N9, C4, C11, and other positions. The strong splitting seen

for the C11 position is traceable to the conformational

change between cation A and B near to this atom, as seen

in Figure 4. For example, the C11-C4-O14-C15 torsion

angle for cation A is -175.54(18)ꢀ; the equivalent C61-C54-

O64-C65 torsion angle for the major component of the

disordered cation B is -129.1(3)ꢀ. The lack of a splitting in

the ³¹P spectrum and for the N19 signal in the ¹⁵N spectrum is

likely related to the approximate inversion symmetry found in

the I-(R) structure. A PXRD pattern simulated from the

crystal structure and subjected to Pawley refinement agrees

with the pattern measured on powders, as shown in the

Supporting Information, indicating that the crystal analyzed

here was representative of bulk material.³⁶

The two independent anions are arranged as a dimeric unit

about a pseudoinversion center, linked by the O-H

hydrogen bonds shown in Figure 5b. The dimeric units are

O

3 3 3

themselves linked by further O-H O hydrogen bonds,

3 3 3

generating chains of anions that run in the direction of

the crystallographic c-axis. Within a chain, the order of

the anions can be described as AABBAABB..., while the

P31 P31, P31 P81, and P81 P81 separations are

3 3 3

˚

4.0302(8), 4.2835(5), and 4.0420(8) A, respectively. One

water molecule in the I-(R) crystal structure donates a

˚

hydrogen bond to O29 [2.775(2) A] and accepts one from

˚

N19 [2.872(2) A] of cation A, with two further symmetry

related interactions being generated by virtue of the 2-fold

Analysis of Twinning in the SCXRD Data of Racemic I.

With a full understanding of the I-(R) crystal structure,

efforts were redirected toward understanding the twinned

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2721

Figure 5. (a) Illustration of the two components of the cation B disorder in the crystal structure of I-(R). The major component is shown with a

solid bond type. The ordered portion of cation B (apart from N69) has been omitted for clarity. (b) Hydrogen-bonding observed for the crystal

structure of I-(R), highlighting the pseudocentrosymmetric packing arrangement. Hydrogen bonds are shown as dashed lines. Hydrogen atoms

(except NH and OH) and the minor component of the cation B disorder have been omitted for clarity.

crystal of I. The twinning was apparent from the room

temperature diffraction pattern, which was characterized

by split peaks, particularly at high resolution. This pattern

was indexed on the basis of two orientation matrices related

by the following twin law:

from examination of systematic absences. Both of these

features are indicative of twinning.³⁷

At low temperature, derivation of the twin law can be

accomplished by analysis of the unit cell dimensions.³⁸Let

a_o, b_o, and c_obe the basis vectors of the orthorhombic cell,

and a_m, b_mand c_mbe the basis vectors of the monoclinic cell.

Possible twin laws correspond to 2-fold rotations about a_oor

b_o(R_aoand R_bo): c_oneed not be considered as this corres-

ponds to b_m, where there is a 2-fold axis due to the mono-

0

1

0

- 1

0

0:93

0

- 1

@

A

clinic symmetry. The triple matrix products (M^-1R_aoM and

At 90 K, splitting of the peaks was not observed, and

a single orientation matrix was sufficient to index all

the diffraction spots. The monoclinic unit cell obtained at

90 K can be transformed to a metrically orthorhombic

-1

M

terms of the monoclinic cell as:

R M) allow these 2-fold rotations to be expressed in

bo

0

1

0

1

- 1

0

- 1

0

- 1

0

1

0

- 1

0

1

0

- 1

˚

unit cell, with approximate dimensions of a = 16.31 A,

˚

@

A

@

A

and

˚

b = 50.79 A, c = 12.22 A, R = β = 90ꢀ, and γ = 89.96ꢀ using

the matrix M:

0

1

0

1

Note the similarity of the second of these matrices to that

derived from the room-temperature data set above. The off-

diagonal term (-2a_mcos β_m/c_m) is essentially unity at 90 K

(0.99 from the diffractometer integration software) but

significantly nonintegral at room temperature, explaining

why the diffraction spots were clearly split under ambient

conditions.

SCXRD Structure of Racemic I. As with I-(R), the crystal

of I studied was also a colorless plate (0.36 ꢀ 0.32 ꢀ 0.04 mm).

The asymmetric unit was found to contain a single cation,

0

- 2

0

1

- 1

0

@

A

Merging the transformed data set in point group mmm

yielded R_int= 0.0829. Merging the data as originally indexed

in point group 1 1 2/m yielded R_int= 0.0357. The merging in

2/m was clearly superior, indicating that the crystal system is

monoclinic. However, the merging statistics in mmm is only

slightly worse. In neither case was the space group obvious

2722 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

unit cell dimensions. The I-(R) structure was described as

pseudocentrosymmetric, and the structure of I has essen-

tially been modeled as a true centrosymmetric variation of

this. Indeed, instead of using direct methods, the structure

of I could also be solved by simply starting from the

coordinates of the ordered molecule from the I-(R) struc-

ture and applying the correct space group symmetry and

origin shift.

In terms of the twinning and disorder present, the struc-

ture of I is far more complex than I-(R), as reflected by its

relatively high R₁value of 10.39%. In the crystal structure of

I-(R), the fused ring system and part of the central linker of

cation B were modeled as being disordered over two posi-

tions. For the crystal structure of I, three similar components

of disorder (henceforth referred to as 1, 2, and 3) had to be

modeled and are depicted in Figure 6b. The site occupancy

factors for the three components were 0.368(3), 0.319(7), and

0.313(8), respectively. It was not possible to locate the

hydroxyl hydrogen atom for any of the cation’s components.

Alternative models of the disorder were also obtained during

this study, including a model in which only a small subset of

the atoms in the fused ring system and the central linker were

modeled as being disordered. Although this approach is

simpler and required fewer refinement parameters, much of

the physical significance of the disorder was lost. The use of

three components avoids this and also ensured that realistic

chemical environments were maintained for the disordered

atoms. The orientation of the fused ring system is similar in

each of the three components. Considering how the plane

normals for the fused rings in the components are inclined

relative to one another, the largest angular deviation, bet-

ween 1 and 2, is just 7.8(5)ꢀ. An effectively identical finding

was noted for cation B in the crystal structure of I-(R). The

absolute configuration modeled for component 1 is different

to that of 2 and 3. This finding implies disorder of the two

enantiomers over the cation sites. Although the model does

not have an equal distribution of the two enantiomers on

each site, the (R)- and (S)-cations are present in equal

proportions overall. This is because an equally disordered

but inverted arrangement is generated elsewhere by space

group symmetry.

The largest peak in the Fourier difference map at the end

of the refinement (1.08 e A^-3) was believed to be associated

˚

with yet another cation component, with the peak corre-

sponding to a new position for O14. With a much lower

occupancy than the other three components, the possible

new component appeared to contribute little to the overall

electron density and also proved difficult to refine. As such, it

was not included in the final structural model. Extensive

efforts to resolve the cation disorder by changing the space

group were unsuccessful. Slightly lower R₁values could be

obtained using subgroups but only at the cost of introducing

additional refinement parameters. With pseudosymmetry

being a large issue in these subgroups and with no clear

advantage being demonstrated, the C2/c space group was

retained.

Figure 6. (a) A view of a cation, anion, and water molecule from

the crystal structure of racemic I. Half of the water molecule shown

is in the same asymmetric unit as the cation and anion. The

hydrogen atom associated with O17 was not located. Only the first

component of the C1-C18 disorder is shown, with these atoms

being refined isotropically. Anisotropic atomic displacement ellip-

soids for the remaining non-hydrogen atoms are shown at the 50%

probability level and hydrogen atoms are displayed with an arbi-

trarily small radius. (b) The three components of the cation dis-

order. Each component is shown with a different bond type

(1 - solid; 2 - open; 3 - dashed). Only component 1 is labeled.

Hydrogen atoms, apart from those associated with the chiral

center, have been omitted, as has the ordered portion of the

cation, apart from N19.

a dihydrogen phosphate anion, and half a water molecule,

as shown in Figure 6a. As before, the half-occupancy of the

water molecule results from it being positioned on a crystal-

lographic 2-fold axis. The unit cell dimensions for I-(R) and

I at 90 K are closely related and compared in Table 1. In

making this comparison, it is essential to make sure that in

both cases, C-centered cells are being considered. This is

because the alternative I-centered cells have very similar

The phosphate anion was modeled in only one orientation.

While this represents a simplification of the true arrange-

ment given the results of the ³¹P SSNMR experiment, efforts

to refine the anion as a multicomponent system, to match the

variety shown by the cation, led to complex models with few

advantages. As previously reported for the structure of I-(R),

the anions are arranged in chains. Within these chains there

are two independent P P separations, 4.039(2) and

3 3 3

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2723

˚

4.284(2) A, similar to the three distances found for I-(R).

Despite the complexity of the racemic structure, it was

possible to identify most of the hydrogen bonds and the

water environment, which are closely related to that seen in

I-(R). Indeed, most of the hydrogen bonds identified can be

easily related to corresponding interactions for the structure

of I-(R). Since the hydroxyl hydrogen atoms for the various

components of the cation disorder of I were not located, no

hydrogen bonds involving this group could be identified.

However, close O O contacts indicate possible interac-

3 3 3

tions. There appears to be more than one bonding option

available to these hydrogens, which, along with the packing

requirements, may help to explain the disorder observed in

the structure.

The SCXRD analysis indicates that the crystal structure of

racemic I contains intimately mixed (R)- and (S)-enantio-

mers in which the conformation of the molecule has adjusted

to fit. The significance of the interaction between the two

enantiomers is discussed further in the following sections. To

verify that bulk samples of I were representative of the crystal

structure determined here, a Pawley refinement was per-

formed against a high-resolution experimental capillary

PXRD pattern of I and was found to be in agreement, with

the results shown in the Supporting Information.

SSNMR Structural Analysis of Racemic I. The SSNMR

data obtained for I was closely examined for effects that

would confirm the SCXRD structure and disorder model.

To verify that key elements of the crystal structures of I and

I-(R) matched with available SSNMR data, additional 2D

1

SSNMR analysis was performed. A H DQ-BABA homo-

nuclear correlation spectrum is shown in Figure 7a.²⁹This

spectrum helps facilitate comparison of ¹H dipolar interac-

tions for the distinctive signal at 13.0 ppm with the SCXRD

structure. The interpretation of the ¹H DQ-BABA data

˚

confirms three short (∼2.5 A) hydrogen-hydrogen interac-

tions in the crystal structure of I, primarily involving H34

and H35 (as N9H is not involved in sufficiently short

interactions). To confirm the assignment of the deshielded

¹H signal at 13.0 ppm to N9H, H34, and H35, two ¹H-³¹P

CP-HETCOR experiments³¹with mixing times of 100 and

500 μs were performed; the results obtained using the latter

mixing time are shown in Figure 7b. A strong correlation is

seen between P31 and the deshielded ¹H signal at 13 ppm; this

interaction arises primarily from the short interactions with

˚

H34A and H35A of 2.0 A. This was confirmed by compari-

son to the ¹H-³¹P CP-HETCOR experiment performed

with a 100 μs contact time (not shown); in this spectrum only

the correlation between P31 and the H signals at 13.0 ppm

1

was detected. The assignment of the other proton signals

in Figure 7b was made in conjunction with the SCXRD

structure of I and the ¹H-¹³C CP-HETCOR spectrum

of I obtained with a 500 μs contact time (see Supporting

Information).³¹No intermolecular dipolar correlations use-

ful for confirming aspects of the crystal structure could

be located in the ¹H-¹³C CP-HETCOR spectrum of I,

although the spectrum did assist with the ¹H and ¹³C assign-

ments given here including confirmation of the assignment of

N9H.

1

Figure 7. (a) H DQ-BABA spectrum (υ_r= 35 kHz) of a sample

of I obtained at 11.7 T. (b) ¹H-³¹P CP-HETCOR spectrum (υ_r=

15 kHz) obtained at 9.4 T using a 500 μs contact time. (c) Build-up of

DQ coherence as a function of ³¹P DQ-POST-C7 mixing time

(obtained at 9.4 T), showing experimental values (points) and a

simulated line (see text). All spectra were obtained at 273 K.

The ¹H-¹³C CP-HETCOR spectrum of I, combined with

1

the H-¹³C MAS-J-HMQC spectrum (see Supporting In-

formation) provides useful evidence of an edge-on aromatic

π-stacking interactions through ¹H shielding effects.^39,40

One notable effect involves H3A and the ring defined by

C5-C6-C7-C8-C12-C13 through a ¹H-¹³C correlation

at 105.5 ppm and 5.5 ppm, the latter being significantly

shielded relative to the ¹H shift of H3 in solution (6.69 ppm),

as is often observed in π-stacked aromatic organic and

2724 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

pharmaceutical crystals.^41-43The distance between the cen-

˚

troid of this ring and H3A was 3.1 A in the major component

of the disordered structure, explaining the shielding effect on

the ¹H shift of H3. The neighboring H2 signal, which appears

at 7.29 ppm in solution, also shows a similar but less

pronounced shielding trend to 6.2 ppm, as it has a greater

distance to the ring centroid. Another strong edge-on aro-

matic π-stacking interaction is also observed through a

¹H-¹³C correlation at 110.2 ppm and 4.5 ppm, which is

assigned to an interaction between H25 and the ring defined

by C1-C2-C3-C4-C11-C10, with a H25A-to-centroid

˚

distance of 2.7 A (measured using the major component of

the disordered model). Again, this proton signal is signifi-

cantly shielded relative to its value in solution (6.96 ppm).

The observation of these aromatic interactions in both the

crystal structure and SSNMR data thus help confirm the

crystal structure of I.

To confirm that the phosphorus atoms were near in space

to one another and arranged in near-linear chains as seen in

the crystal structure of both I and I-(R), a ³¹P DQ-POST-C7

experiment (with a ¹H-³¹P preparation period and ¹H

decoupling) was conducted on a sample of I during which

the DQ excitation period was incremented.³⁰The results of

this experiment are plotted in Figure 7c, where the excitation

of DQ coherence is compared to a theoretical curve simu-

˚

lated for a 4.0 A P P internuclear distance using an

3 3 3

empirical expression that has been developed for phosphate

networks in glasses.⁴⁴Although a full analysis of the multi-

spin DQ excitation behavior expected from linear chains of

phosphorus atoms is beyond the scope of this work, the

results shown in Figure 7c are a useful confirmation of the

proximity of the phosphorus atoms in the structure of I.

Finally, the observation of disorder in the ¹³C spectra also

serves as a check of the crystal structure, as disordered

positions in the SCXRD structure should correlate with

disordered positions in the SSNMR spectra. This is the case

here; for example, in Figure 1a,b, positions C4, C15, C16,

and C21 (all of which are near to the chiral center) are seen to

be broadened in the SSNMR spectra; these same positions

are the most disordered in the final crystal structure, as was

originally noted at the start of the efforts to produce the

single enantiomer salts. The ¹³C signal for C11 observed

using dipolar dephasing shows the most structure, with two

clearly resolved peaks, the more deshielded of which shows

evidence of additional unresolved splitting. As was the case

with the crystal structure of I-(R), the structure observed for

the C11 position is likely related to conformational changes

involving O14 and its attached chain, as seen in Figure 5b.

The signal for C11 suggests at least three significant compo-

nents for the disorder model, in agreement with the three-

component model used here for the crystal structure of I.

However, as it was not possible to fully resolve the three

components of the disorder modeled in the crystal structure

of I in any of the multinuclear SSNMR spectra, even at 11.7

T, no direct comparison with the relative occupancies in the

SCXRD structure could be made.

Figure 8. (a) DSC thermograms of racemic I, I-(R), and I-(S).

Exothermic events are plotted in the downward direction. (b) Dia-

gram constructed from samples crystallized with different enantio-

meric input ratios, showing the peak temperature of the melt plotted

against the mole fraction of the (S)-enantiomer in the solid solution as

measured by chiral LC (see text). (c) Expansions of the ¹³C dipolar-

dephased spectra of the samples in (b), obtained at 273 K and 9.4 T.

occurs at 144.8 and 144.9 ꢀC for I-(R) and I-(S), respectively,

compared to 156.5 ꢀC for I. The broad endotherm between

60 and 125 ꢀC for all three samples corresponds to loss of

water of hydration, discussed in more detail below. Given the

similar purity values determined for these materials, the

melting point behavior is assigned to the formation of a solid

solution of the enantiomers in I.^17,18The heats of fusion for

the enantiomers were 37.4 and 36.5 J/g for I-(R) and I-(S),

respectively, compared to 46.9 J/g for I, suggesting a stabi-

lization energy of approximately 10 J/g for I over its en-

antiomers. A series of crystallizations were carried out using

Thermal Analysis. DSC studies of I, I-(R), I-(S), and

mixtures of different mole fractions of the enantiomers

were also performed to determine whether I crystallized

as a racemic mixture (single enantiomer conglomerates)

or as a solid solution containing intimately mixed (R)- and

(S)- enantiomers.^17,18The DSC traces for samples of I-(R)

and I-(S) are compared with that of I in Figure 8a. The peak

temperature of the melting endotherm for the enantiomers

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2725

mixtures of the (S)-enantiomer and racemic carvedilol as

input material, to obtain mixtures of I and I-(S). The melting

points of these mixtures are plotted in Figure 8b against the

actual molar ratios determined by chiral LC, and match the

trend expected for a solid solution.¹⁸The DSC results thus

demonstrate that racemic I exists as a solid solution (also

referred to as a pseudoracemate) as opposed to phase-

separated conglomerates of (R)- and (S)- domains.¹⁸This

can be anticipated given their similar unit cells and crystal

structures, but as solid solutions are rarely encountered,^45-49

DSC analysis is necessary to confirm this structural

feature.¹⁸

In Figure 8c, expanded regions of the dipolar-dephased

SSNMR spectrum are shown for the crystallized samples

with different molar ratios. As the excess of the (S)-enantio-

mer increases, the C4, C11, and several other positions show

broadening behavior characteristic of additional disorder.

This illustrates the molecular level effect of extra (S)-enan-

tiomer in the intermediate mole ratio solid solutions, creating

domains of the (S)-enantiomer and disrupting the stabilizing

disorder model seen in the racemate. The full ¹³C CP-TOSS

and ³¹P CP-MAS spectra of these samples show similar but

less pronounced effects (see Supporting Information).

An equimolar sample of I-(R) and I-(S) was also produced

by solvent-drop grinding (SDG) with methanol, using the

same enantiomer samples that melted at ∼145 ꢀC, in an

attempt to produce a solid solution in a different manner to

the direct crystallization approaches used above. The use of

SDG to produce crystalline solid solutions has been pre-

viously demonstrated.⁵⁰The DSC thermogram for this

sample yielded a peak temperature of 153.8 ꢀC and an

enthalpy of fusion of 45.4 J/g for the melting endotherm;

the ¹³C SSNMR spectra of this sample showed additional

broadening in the disordered regions associated with the

chiral center but otherwise resembled the spectra of I pro-

duced by crystallization (see Supporting Information).

This sample is thus likely a partially mixed solid solution,

in which the enantiomers have not reached their optimal

low-energy structure because of the more limited mixing

process afforded by the SDG technique in comparison to

solution crystallization.

Figure 9. GVS isotherms obtained at 25 ꢀC for a powder sample of

racemic I. (a) Isotherm obtained with 10% RH steps from ∼0% RH

(dry nitrogen) to 90% RH. (b) Isotherm obtained with 1% RH steps

from ∼0% RH (dry nitrogen) to 12% RH. The water is not fully

removed at the lowest RH achieved in this experiment; higher

temperatures and reduced pressures are needed to remove more of

the water, although the process remains fully reversible up to the

melting temperature.

bolstered by the lack of any detectable peak shift changes in

1

the deshielded region of the H SSNMR spectra of I when

1

compared to I-(R), since H chemical shifts are sensitive to

minor changes in hydrogen bond geometry.⁵³

GVS, PXRD, and SSNMR Studies of Dehydration. The

second interesting feature of the crystal structure of I,

namely, the nature of the water in the structure, also required

investigation using a multidisciplinary approach. The theo-

retical water level is 1.8% w/w for the hemihydrate found by

SCXRD; however, the water in powder samples could be

readily lowered to 1.2% w/w by vacuum drying at slightly

elevated temperatures (e.g., 65 ꢀC) and quickly analyzing

samples by Karl Fischer titration (KFT). However, some

rehydration likely occurred during sample weighing and

transfer to the KFT titration media, so that the reported

value of 1.2% w/w should be considered to be an upper limit

with the actual moisture level being lower. GVS was used to

assess the change in water content of I with humidity. The

isotherm in Figure 9a, which is a typical Type I isotherm,^2,54

shows a feature at 10% relative humidity (RH) that indicates

a significant change in hydration state. Steps of 10% RH

were used to obtain these data. A more precise determination

of the critical RH was obtained by stepping the GVS

isotherm in 1% RH increments, as shown in Figure 9b; this

indicates a critical RH of approximately 1% RH. Significant

water loss is observed between 10% and 1% RH. Between

10% and 90% RH, the water content of I can reversibly

The results of the DSC, SSNMR, and SCXRD studies

show that the nearly isomorphous nature of the single enan-

tiomer and racemate structures allows for formation of a

solid solution. Solid solutions of organic enantiomers^45-47

and partial solid solutions^48,49are not commonly encoun-

tered; conglomerate and racemic crystal structures are far

more frequently observed.^17,18From the crystallographic

perspective, the occurrence of the solid solution can be easily

rationalized, given that the C2/c space group of the crystal

structure of I incorporates inversion centers not found in the

C2 space group of the enantiomer; comparison of the

structures suggests that the I-(R) structure would prefer to

be centrosymmetric but cannot easily do so because of its

uniform chirality. Additionally, the crystal structure of I may

simply satisfy intermolecular interactions in a more energe-

tically favorable way, thus leading to enthalpy-based stabi-

lization of the free energy. Since the crystal structure of the

solid solution of I shows only a slight increase in density

relative to that of the I-(R) structure (systems which obey

Wallach’s rule usually show a more significant increase

in density),^51,52it appears that greater satisfaction of inter-

molecular interactions, such as van der Waals forces, drives

the increase in free energy in I over I-(R). This argument is

2726 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

Figure 10. (a) Flat-plate PXRD patterns of I at ambient conditions,

after 48 h at 0% RH at room temperature and after 2.5 h at 125 ꢀC,

with affected regions denoted by arrows. (b) PXRD patterns of I

showing the reflection at ∼14 deg 2θ as a function of drying time

(under dry nitrogen) over a 48 h period at room temperature,

highlighting the continuous nature of the dehydration process in

powders.

Figure 11. (a) ¹³C CP-TOSS spectra of a sample of I at 125 ꢀC and

the vacuum-dried sample with 1.2% w/w water content compared

to a spectrum of I stored under ambient conditions. Arrows denote

changes observed in the spectra. (b) Expanded FT-Raman spectra

of a dehydrated sample of I (vacuum-dried to <1.2% w/w water)

used for the SSNMR experiment compared to the same batch at

ambient conditions, acquired with 2 cm^-1resolution. Two diag-

nostic bands for the dehydration process are noted with arrows.

The off-scale band at 1628 cm^-1was used to normalize the spectra.

Raman spectra of the same samples sealed within a zirconia

MAS rotor are also shown, confirming the integrity of the samples.

The zirconia rotor produces strong Raman signals at 643, 464,

320, 262, and 148 cm^-1, outside of this region (see Supporting

Information).

change over a 0.4% w/w range. Above 80% RH, the change

in slope seen in the GVS isotherm offers evidence of surface

water uptake. The lack of hysteresis observed suggests that

the dehydration is a reversible hydration state change, as

phase changes usually exhibit hysteresis. An indistinguish-

able GVS isotherm was obtained for a sample of I-(R), with a

similar critical RH of about 1% RH. DSC and TGA experi-

ments on samples of I showed a relatively high final de-

hydration temperature of 125 ꢀC. As seen in Figure 8a, DSC

experiments detected a slightly higher peak dehydration

temperature of I-(R) and I-(S) relative to I (100 ꢀC versus

92 ꢀC), providing evidence that water is more tightly held in

the single enantiomer phases.

PXRD measurements on I were performed at elevated

temperatures and a range of relative humidity settings,

including the acquisition of patterns over several days as

dry nitrogen was blown over the sample. No changes were

detected in the PXRD pattern over a humidity range of

10%-90% RH. However, PXRD patterns obtained at

125 ꢀC and at ambient temperature after two days of drying

under nitrogen show a number of changes, which are

depicted in Figure 10a. The similarity of the patterns

suggests that only minor structural changes have occurred.

Over the course of the two-day drying period, acquisition

of PXRD patterns showed a continuous displacement of

diffraction peaks to both higher and lower values of 2θ.

An example of this is depicted in Figure 10b, where the

change in the feature at approximately 14ꢀ 2θ is plotted

with the drying time. A similar continuous shift in peak

positions and shapes, including the feature at 14ꢀ 2θ, was

also observed with increasing temperature up to 125 ꢀC.

The PXRD results thus indicate that the dehydration

process in powders in I is continuous and devoid of any

discrete transition.

The structural aspects of the continuous dehydration

process were further investigated using FT-Raman and ¹³C

SSNMR spectroscopy. ¹³C CP-TOSS spectra were collected

at elevated temperature and low humidity. A vacuum-dried

sample of I was quickly packed into a rotor in a low-humidity

environment and analyzed. As shown in Figure 11a, only

minor changes are observed in the spectra. The changes are

confined to the regions assigned to carbons that are near to

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2727

Figure 12. An illustration of molecular packing in the I-(R) crystal structure viewed in the direction of (a) the crystallographic a-axis, (b) the

crystallographic b-axis, and (c) the crystallographic c-axis. The cations are colored green, the anions are purple, and the water molecules are red,

with the darker and lighter of each color representing A and B, respectively. The disorder in the B cation has been omitted for clarity. Narrow

channels are visible in the structure; the larger and more obvious of these run along the b- and c-axes (in the latter, the water molecules are in a

staggered arrangement).

the water in the crystal structure of I. Only subtle changes are

observed upon dehydration in the FT-Raman spectra in

Figure 11b, in agreement with the SSNMR and PXRD

results. FT-Raman spectra were also collected through the

wall of the zirconia MAS rotors used to produce the SSNMR

spectra in Figure 11a, confirming that the packing of the

rotor had resulted in minimal rehydration. The combined

SSNMR and Raman results thus indicate minimal changes

to the molecular structure upon dehydration of the crystal

structure.

Both the racemic and single-enantiomer crystal structures

show water molecules contained in narrow tunnels running

through the structure, as depicted for the crystal structure

of I-(R) in Figure 12. The presence of tunnels suggests that

the solvent can be removed without significantly disrupting

the structure of I-(R). Extensive experience with channel

hydrates has shown that the water can be both weakly or

tightly bound and the lattice can either expand or not expand

upon hydration or dehydration.^2,5-10The size of the chan-

nels in Figure 12 combined with the relatively short water

2728 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

Figure 13. (a) Changes in unit cell dimensions during dehydration of a single crystal of I-(R) as measured by X-ray diffraction. Error bars are

shown for each measurement. The dashed lines represent the value found upon returning to ambient conditions after heating and drying.

(b) The change in appearance of the crystal of dehydrated I-(R) as it was rehydrated under ambient laboratory conditions. The time after the

start of the rehydration is shown under each image as minutes:seconds. A video of the changes can be found in the Supporting Information.

hydrogen bonds immediately suggests that the water will not

be easily removed, in agreement with a low critical RH and a

high dehydration temperature from the GVS and TGA

results.

cell dimensions of this crystal were monitored periodically.

Although the water content of the crystal was not directly

determined as part of the experiment, the results indicated

that this process substantially dehydrated the crystal. The

changes in the unit cell dimensions are plotted in Figure

13a. The most informative metric for following the dehy-

dration was the β angle, which decreased by over two

degrees. The change in the β angle with time was not linear,

slowing throughout the dehydration. Changes in the other

unit cell constants were less diagnostic. Rehydration of the

dehydrated crystal was rapid under ambient laboratory

conditions, completing in less than 30 min. The rehydration

process could also be followed by changes in the single-

crystal diffraction pattern occurring over this period (see

Supporting Information).

SCXRD Studies of Dehydration. To obtain additional

understanding of the dehydration effects seen in I, the single

crystal of I-(R), originally flash-frozen to 90 K for the

crystal structure analysis, was studied in situ on the dif-

fractometer in a series of dehydration and rehydration

experiments that lasted approximately seven weeks. Pow-

ders of I-(R) show DSC, TGA, and GVS results similar to

those of I and thus the behavior of the high-quality crystal

of I-(R) should be useful as a surrogate for I. The first

dehydration was carried out at room temperature using the

dry nitrogen gas stream over a period of 18 days. The unit

Article

The changes in β angle explain the results of the PXRD

Crystal Growth & Design, Vol. 10, No. 6, 2010 2729

analysis at low humidity and high temperature, although the

SCXRD study allows for a far more accurate determination

of the trends in unit cell parameters than indexing and

refinement of PXRD patterns. Pawley refinement of the

PXRD patterns of dehydrated samples of I confirmed the

same trend toward lower β angles for the racemic material

upon drying. Interestingly, rehydration of the crystal of I-(R)

could also be followed visually as shown in Figure 13b (see

the Supporting Information for a time-lapse video of this

process). The opaque areas seen during rehydration are

believed to result from “fissures” created in the crystal, as

β and the other unit cell dimensions rapidly change. The

opaqueness starts at the outer boundaries of the crystal,

moves through the whole crystal, and then clears from the

edges, presumably as a result of a healing process within the

crystal. On the basis of this visual evidence, the rehydration

occurs via the smaller faces on the crystal and not through

the main face. This corresponds to water entering the

crystal in directions approximately perpendicular to the

crystallographic a-axis (the main flat face of the crystal is

the (001) face. In the view shown in Figure 13b, the a-axis of

the unit cell is approximately perpendicular to the page,

while the b-axis runs approximately parallel to the attached

fiber. The opacity effect observed in the images is thus

consistent with the narrow and tortuous channels observed

in the packing plots in Figure 12, in that these views suggest

water ingress and egress should occur perpendicular to the

a-axis of the unit cell. Although 1D opacity effects are far

more common,⁵⁵a similar 2D opacity effect along the edges

of a crystal has been reported for manganese(II) formate

dihydrate.^2,56,57

Throughout the single crystal study, it was found that the

dehydration and rehydration processes were totally reversi-

ble, based on the unit cell constants determined before and

after several cycles. Furthermore, heating the hydrated form

to 373 K (100 ꢀC) led to dehydration in less than 20 min

without any clear visual signs of change to the crystal. The

unit cell constants after the rapid dehydration were consis-

tent with those obtained from the two-step process of

dehydrating at 295 K (22 ꢀC) and then heating to 373 K.

This observation supports the presence of a single phase at

373 K, independent of the method of dehydration. After

seven weeks of temperature and humidity changes, the

crystal of I-(R) remained intact, with the only damage being

superficial surface peeling and a minor crack.

Figure 14. (a) ¹⁷O static spin-echo SSNMR spectrum (obtained at

11.7 T and 273 K) of a sample exposed to 75% RH H₂¹⁷O (40%

enriched) after preconditioning at 75% RH H₂O for 5 days. No

signal is obtained after postconditioning this sample at 75% RH

H₂O for another 5 days, illustrating the reversibility of this process.

(b) IR spectra showing the OH stretching region of a sample

exposed to 75% RH H₂¹⁸O (99% enriched) for different times after

preconditioning at 75% RH H₂O for 5 days. The spectrum of a

sample that was vacuum-dried prior to exposure to 75% RH H₂¹⁸O

is shown for comparison. After 96 h, the material was re-exposed to

75% RH H₂O, and the IR band reverted back to its original position

(not shown), confirming that water exchange is reversible.

fissures offer an explanation for the hydration behavior of I,

exchange experiments were conducted with isotopically la-

beled water to gain a more detailed understanding of water

exchange via SSNMR experiments and vibrational spectro-

scopy.^10,58,59The ability of whole water molecules to ex-

change into the structure from the vapor phase was probed

using H₂¹⁷O and H₂¹⁸O. An exchange experiment was con-

ducted on vacuum-dried I that was subsequently exposed to

75% RH H₂¹⁷O (40% enriched) for 5 days. The ¹⁷O SSNMR

spectrum of this sample, shown in Figure 14a, confirmed that

whole water molecules may exchange under these condi-

tions. A line shape with a 42 kHz width at its base, consistent

with a second-order quadrupolar broadened interaction, is

observed. The breadth of this signal is consistent with a C_q

(quadrupolar coupling constant) of approximately 7-8 MHz,

typical of a bound hydrate.^59-62This second-order line

shape is broadened and does not match the line shape

expected from a single ordered oxygen site.⁶³Because of

the disorder in the SCXRD structure of I and broadening

observed for the water signal in the ²H MAS data (discussed

below), attempts to obtain a more detailed fit of the ¹⁷O line

shape were not attempted. To follow the uptake of water

dynamically, IR spectroscopy was employed. The ATR IR

Additional variable temperature studies on the dehy-

drated form suggested the presence of a constant phase

between the temperatures of 166 and 328 K. At 166 K, a

distinct but wholly reversible phase transition occurred.

The low temperature unit cell appeared to be triclinic,

suggesting 12 independent molecules in the asymmetric

unit. This temperature region was not explored on powder

samples (the lowest temperature reached on a powder

2

sample was 173 K in the H SSNMR study discussed in

the next section). The low-temperature dehydrated phase

of I-(R) was not investigated further; powders of I were

studied by PXRD at liquid nitrogen boil-off temperatures

(<100 K) and did not show evidence of a similar transi-

tion under these conditions.

2

Exchange with H, ¹⁷O, and ¹⁸O-labeled Water and Ana-

lysis of Water Incorporation and Dynamics. Although the

presence of tunnels in the lattice, the observed unit cell

changes, and the optical changes potentially associated with

2730 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

spectra as a function of exposure time to 75% RH H₂¹⁸O

(99% enriched) are shown in Figure 14b, for a sample that

was pre-equilibrated at 75% RH H₂O. The IR results show

that ¹⁸O-labeled water exchanges rapidly into the structure

under these conditions. The peak maximum of the OH

stretching vibration shifts by 6 cm^-1after 1 h of exposure

and does not shift any further, as seen in comparison to the

spectrum after 96 h of exposure. Subsequent re-exposure of

the sample exposed to 75% RH H₂¹⁸O for 96 h to 75% RH

H₂O led to a complete reversal of the shift in IR band

position to its initial position, confirming the reversibility

observed by ¹⁷O SSNMR. The spectra in Figure 14b are

compared to the spectrum of a sample that was vacuum-

dried prior to exposure to 75% RH H₂¹⁸O; the shift in the

band maximum was 15 cm^-1for this sample, in agreement

with the theoretical maximum for an asymmetric water

stretching vibration based on reduced mass changes.

The IR results with and without preparatory vacuum

drying suggest that the water exchange, although facile,

cannot access the inside regions of the particles. The powder

used in this study was composed of irregularly shaped plates

with a median (X₅₀) particle diameter of 12 μm, and no

evidence of unusually high porosity was observed as the

specific surface area (SSA) was 1.8 m²/g (measured by the

Brunauer-Emmett-Teller adsorption method with nitro-

gen gas). Assuming a spherical particle, the surface area per

particle would be approximately 450 μm², offering ample

surface area for exchange. The evanescent wave in diamond

ATR IR penetrates approximately 0.5 to 2 μm into the

surface of a particle at 3500 cm^{-1 64}

Thus, the partial

.

exchange seen by comparing the 6 cm^-1shift to the 15

cm^-1shift is consistent with substantial exchange of water

molecules in the outer regions of the particles. The vacuum

drying procedure clearly enables access to deeper regions in

the particles, consistent with the proposed fissures seen by

optical and unit cell changes upon dehydration. This also

helps explain the behavior of the GVS isotherm, in that

minor changes in water content are observed from outer-

surface effects until the critical RH is reached, at which point

the exchange process is greatly enhanced.

Figure 15. (a) ²H MAS spectra (obtained at 9.4 T) of a sample of I

stored at 75% RH D₂O for five days and after drying (insets show

expansions of the centerband and þ3 sideband). No evidence of

surface water is observed (which can be distinguished in a sample

exposed to 100% RH D₂O for 6 h by the observation of a peak at

∼5 ppm in the centerband spectrum, denoted by arrows). (b) FT-

Raman spectra (obtained at 2 cm^-1resolution) of I obtained from

different exposure times to 75% RH D₂O after preconditioning at

75% RH H₂O for 5 days, showing the continuous nature of the

deuterium exchange process over the course of several hours.

Solid powders of I were also subjected to D₂O exchange, to

study deuterium transport in the material and the presence of

surface water. A sample of I was vacuum-dried and then

exposed to 75% RH D₂O, while a second sample, pre-

equilibrated at 75% RH H₂O, was also placed into the

(see Supporting Information). Only a slight peak shift was

observed for the N19 signal in the ¹⁵N spectrum obtained

with a 2 ms contact time, consistent with a distant isotope

effect on hydrogen bonding but not direct exchange of the

NH₂^þprotons. This confirms that no appreciable exchange

has occurred at N9 or N19, while leaving open the possi-

bility of exchange at O17, the oxygen atoms of the anion

2

75% RH D₂O chamber. The H MAS SSNMR spectra of

these two samples are indistinguishable as seen in Figure 15a,

indicating that initial drying has no effect on the deuterium

incorporation in I. Evidence of surface water is not seen in

either spectrum but can be seen when additional exposure to

100% RH is performed as shown in Figure 15a by the

appearance of a peak at ∼5 ppm in the centerband. This

confirms that the mass change in the GVS isotherm in the

vicinity of 75% RH still involves bound water and not

surface water.

1

and the water molecule itself. Next, the H spectra of the

deuterium-exchanged sample were compared with the ini-

tial sample of I. A loss of intensity was observed for the

peaks at 13.0 and 8.0 ppm (see Supporting Information).

2

This corresponds to the appearance of intensity in the H

spectrum, as confirmed by the expanded plots of spinning

sidebands in Figure 15a. On the basis of the observed

chemical shifts, the deuterium exchange can be definitively

linked to water, as confirmed by the ¹⁷O SSNMR results

and the P31 dihydrogen phosphate anion protons (via the

The structural location of the deuteration in the crystal

structure of I can be followed using several approaches.

First, the multinuclear SSNMR spectra of the deuterium-

exchanged sample were compared to that of the initial

sample of I. Various minor changes in CP intensity (both

increasing and decreasing peak intensities) were observed

in a short contact time (1 ms) ¹³C spectrum, while a 45%

loss of CP intensity was seen in the short contact time

(500 μs) ³¹P spectrum of the deuterium-exchanged sample

2

¹H and H peak at 13.0 ppm and the slight intensity loss

seen in the ³¹P CP spectra). The presence of deuteration can

be ruled out at N9 and N19, but deuteration at O17 cannot

be ascertained from the SSNMR spectra.

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2731

methods.^10,59,65,66Figure 16a shows the ²H static spectra

obtained over a 200 K temperature range. A simulated

spectrum obtained by fitting the lowest temperature (173 K)

pattern to a single-site deuterium line shape with ν_Q= 225

kHz and η = 0.0 is also shown. The lack of a peak near to

0 kHz or any other changes indicates that there is no sig-

nificant water motion over a 200 K temperature range,

although there is evidence of a slight increase in intensity at

0 kHz at 373 K, indicating that motion may begin near the

dehydration temperature. This lack of water motion is highly

atypical given the normally facile jump motion of water in

hydrates.^10,59,65,66The hydrogen bonding network for the

water molecule may be the cause of this; as shown in

Figure 16b, the water strongly interacts with two donors

and two acceptors.

Combining the IR, SSNMR, and SCXRD results, the

dehydration, rehydration, and exchange processes in I and

its enantiomers may be viewed as follows. Below the critical

RH of ∼1%, the material dehydrates by first losing moisture

through surfaces perpendicular to the crystallographic

a-axis, leading to unit cell changes and particularly a sig-

nificant distortion of the β angle of nearly 2ꢀ. This results in

optical opacity of the crystal, possibly from fissures created

by the large change in unit cell parameters, which start at the

surface, work through the crystal but ultimately disappear.

The effect, a continuous disruptive process, is associated

with improved access to deeper cells and more complete

water exchange. No evidence was found of any major

differences in the dehydration and rehydration process; they

appear to be reversible images of one another. Significant

exchange of whole water molecules occurs at the surface at

75% RH. There is no evidence that the 0.4% w/w change

observed between 10 and 90% RH is due to surface water;

hence it is likely that the water vapor activity leads directly to

water exchange processes at the surface. A facile tunnel

process is unlikely, as evidenced by the partial exchange

obtained in the H₂¹⁸O-exchange IR experiments and the

narrow tunnels in the structure, but may contribute to the

dehydration process, possibly across a limited number of

unit cells. Although the fissure mechanism offers an expla-

nation for the incorporation of water deep into the poly-

crystalline material, in combination with limited water

ingress through the narrow tunnels seen in the packing

diagrams of I, alternative possibilities do exist. Molecular

motion could also provide a mechanism for rapid water

incorporation without the presence of large channels, as

was recently proposed for clarithromycin.⁶⁷It is also possible

that molecular motion allows for facile water incorporation

at the particle surface and may be enhanced by the observed

unit cell distortions under vacuum conditions, allowing for

deeper water exchange.

Figure 16. (a) Static ²H NMR spectra (obtained at 9.4 T) of D₂O-

exchanged I obtained over a 200 K temperature range. A simulated

pattern is shown at the bottom, obtained by fitting against the 173 K

experimental pattern, with ν_Q= 225 kHz and η = 0.0. (b) The water

environment of one of the water molecules from the crystal structure

of I-(R). The water lies on a crystallographic 2-fold axis and engages

in hydrogen bonds (dashed lines) to two symmetry equivalent

cations, creating an unfavorable environment for jump motion.

The environment of the second water molecule (not shown) is nearly

identical but is instead hydrogen-bonded to two symmetry equiva-

lents of the other independent cation.

The D₂O incorporation process was also followed dyna-

mically using Raman spectroscopy. The results of exposure

of two samples of the same powder used in the aforemen-

tioned SSNMR experiments are shown in Figure 15b. D₂O

uptake is monitored by FT-Raman via the shifting of

two bands in the spectrum, from initial values of 916 and

764 cm^-1to final values of 906 and 769 cm^-1. No evidence of

an OD stretching vibration was observed in the Raman

spectra of the final D₂O exchanged sample, most likely

because this band was broadened by hydrogen bonding

(a strong, broad band was observed in the IR spectra). The

conversion was observed by FT-Raman to be complete after

5 days as evidenced by no further change in the spectra. The

deuterium exchange process thus occurs on a slower time

scale than exchange of whole water molecules, presumably

because of limited motion inside the hydrated structure.

From the bulk perspective, the hydration state behavior of

I is similar to that of an isomorphic desolvate, with dehydra-

tion occurring as a continuous process in concert with

significant structural adjustments. Among the isomorphic

desolvates and channel hydrates published in the literature,

we are not aware of an example with a critical RH as low as

that of I.^5-10The situation contrasts with that of typical

isomorphic desolvates, where water easily moves along

tunnels (potentially crossing many unit cells) to escape via

a particular crystal face.^2,5-10,54,55Despite its underlying

structural complexity, powders of I behave in many respects

like a tightly bound channel hydrate with a typical type I

isotherm.

2

Since a large part of the H signal can be ascribed to

exchange of the half mole of water, the motion of the water

molecules can be further investigated. Water is known to

commonly execute jump motions (such as jumps about its

C₂-axis and tetrahedral jumps) in a wide variety of inorganic

and organic hydrate systems.^64,65This type of dynamic

behavior is readily observed by static ²H SSNMR, which is

also amenable to a wider temperature range than MAS

2732 Crystal Growth & Design, Vol. 10, No. 6, 2010

Vogt et al.

Conclusions

for assistance with water content determinations. Professor

ꢀ

Gerard Coquerel (University of Rouen) and Dr. Ludovic

Renou (Pharmorphix) are thanked for insightful discussions.

Racemic carvedilol phosphate (I) was found to exist as a

solid solution of its enantiomers, which is a relatively infre-

quent occurrence in organic molecular crystals. Efforts to

solve its crystal structure were hindered by weakly diffracting

twinned crystals, from which data refinement was difficult.

Multinuclear solid-state NMR analysis showed indications of

disorder in the vicinity of the chiral center. Single enantiomer

phosphate salts were therefore prepared and found to be

isomorphic with the racemate using SSNMR and other

techniques but lacking the disorder element detected in the

SSNMR spectra of the racemate. A crystal of I-(R) did not

exhibit twinning and yielded good diffraction data that were

readily solved with an R₁of <5%. The structure of the single

enantiomer enabled a better understanding of the related

racemic crystal structure. It is unusual for the single-enantiomer

and racemic crystals of a compound to have such similar unit

cell dimensions and crystallize as isomorphous phases, allowing

for formation of a solid solution. That the racemate yields

twinned crystals, whereas the enantiopure form does not even

though the metric requirements for twinning are met, is also

worthy of note. It may be possible that the development of twin

domains in this system depends on the presence of both

enantiomers; however, this cannot be confirmed at the present

time. The increased disorder seen by SSNMR in I is associated

with the formation of the solid solution, as confirmed by

thermal analysis.

Supporting Information Available: A movie file illustrating the

rehydration process for a single crystal of I-(R), crystallo-

graphic information files for the crystal structures of I and

I-(R), and additional PXRD, NMR, and SSNMR data. This

material is available free of charge via the Internet at http://

pubs.acs.org.

References

(1) Bernstein, J. Polymorphism in Molecular Crystals; Oxford Univer-

sity Press: New York, 2002.

(2) Byrn, S. R.; Pfeiffer, R. R.; Stowell, J. G. Solid-State Chemistry of

Drugs; SSCI: West Lafayette, IN, 1999.

(3) Lee, S. L.; Raw, A. S.; Yu, L. In Drugs and the Pharmaceutical

Sciences; Informa Healthcare: New York, 2008; Vol. 178, pp 571-

583.

(4) Eliel, E. L.; Wilen, S. H.; Mander, L. N. Stereochemistry of Organic

Compounds; Wiley: New York, 1993.

(5) Morris, K. R.; Rodriguez-Hornedo, N. In Encyclopedia of Phar-

maceutical Technology; Swarbrick, J.; Boylan, J. C.; Eds; Marcel

Dekker: New York, 1993; Vol. 7, pp 393-440.

(6) Stephenson, G. A.; Groleau, E. G.; Kleeman, R. L.; Xu, W.;

Rigsbee, D. R. J. Pharm. Sci. 1998, 87, 536–42.

(7) Khankari, R. K.; Law, D.; Grant, D. J. W. Int. J. Pharm. 1992, 82,

117–27.

(8) Stephenson, G. A.; Stowell, J. G.; Toma, P. H.; Xu, W.; Pfeiffer,

R. R.; Byrn, S. R. J. Pharm. Sci. 1997, 86, 1239–44.

(9) Apperly, D. C.; Basford, P. A.; Dallman, C. I.; Harris, R. K.;

Kinns, M.; Marshall, P. V.; Swanson, A. G. J. Pharm. Sci. 2005, 94,

516–23.

Both I-(R) and I were observed to be hemihydrates by

SCXRD but can undergo isomorphic dehydration at low

humidity and high temperature and take up additional water

at higher humidity. The water is situated in narrow tunnels

that run the length of the crystals. The dehydration and

rehydration processes were found to be associated with a

distortion of the unit cell, in turn leading to fissures that

facilitate additional dehydration of the crystals or enhance

molecular mobility. The critical RH of racemic I was found to

be <1% RH, which is extremely low in comparison to other

known isomorphic desolvate systems. Exchange experiments

with oxygen-labeled water were used to show that water

molecules can appreciably exchange into bound positions in

the lattice at ambient conditions and that enhanced exchange

(10) Vogt, F. G.; Dell’Orco, P. C.; Diederich, A. M.; Su, Q.; Wood, J. L.;

Zuber, G. E.; Katrincic, L. M.; Mueller, R. L.; Busby, D. J.;

DeBrosse, C. W. J. Pharm. Biomed. Anal. 2006, 40, 1080–88.

(11) Tenero, D.; Boike, S.; Boyle, D.; Ilson, B.; Fesniak, H. F.; Brozena,

S.; Jorkasky, D. J. Clin. Pharmacol. 2000, 48, 844–53.

(12) Stoschitzky, K.; Koshucharova, G.; Lercher, P.; Maier, R.; Sakotnik,

A.; Klein, W.; Liebmann, P. M.; Lindner, W. Chirality 2001, 13,

342–46.

(13) Almeida, D. R. P.; Gasparro, D. M.; Pisterzi, L. F.; Torday, L. L.;

Varro, A.; Papp, J. G.; Penke, B.; Csizmadia, I. G. J. Phys. Chem A

2003, 107, 5594–5610.

(14) Chilmonczyk, Z.; Ksycinska, H.; Aboul-Enein, H. Y.; Lee, W.-J.

J. Liq. Chromatogr. Relat. Technol. 2001, 24, 2505–12.

(15) Lamprecht, G.; Gruber, L.; Stoschitzky, K.; Lindner, W. Chroma-

tographia 2002, 56, S25–S29.

(16) In this work, isomorphs are defined as nearly identical crystal

structures of materials that have different chemical compositions,

including enantiomers and dehydrated forms. These types of

situations are also referred to as isostructural forms in the litera-

ture. See Bernstein, J. Polymorphism in Molecular Crystals; Oxford

2

is obtained with preparatory vacuum drying. H SSNMR

experiments showed that the water molecules are unable to

execute typical jump motions in the crystal lattice over a

temperature range of 200 K, which is also unusual for a

hydrate system and together with the narrow tunnel structure,

offers an explanation for the low critical RH observed.

The use of complementary techniques in the study of this

complex crystal system enabled both the solution of the

racemic structure (the marketed form of this drug substance)

and the understanding of the nature of disorder and water in

the crystal, allowing for deeper insight into the structure

than what would have been possible from the individual

techniques. The results illustrate the benefits of applying a

multidisciplinary approach to complex solid form science

questions.

ꢀ

University Press: New York, 2002; p 271; Kalman, A.; Parkanyi, L.;

Argay, G. Acta Crystallogr. 1993, B49, 1039-1049. The definition

of isomorphic desolvate used here and in ref 6 to describe desolva-

tion with minimal structural changes is also consistent with this

nomenclature.

(17) Datta, S.; Grant, D. J. W. Nat. Rev. Drug Discovery 2004, 3, 42–57.

(18) Mitchell, A. G. J. Pharm. Pharm. Sci. 1998, 1, 8–12.

(19) Brook, C. S.; Chen, W.; Dell’Orco, P. C.; Katrincic, L. M.; Louvet,

A. M.; Oh, C. K.; Spoors, G. P. PCT Int. Appl. 2004, WO

2004002419, 53 pp.

(20) Nyqvist, H. Int. J. Pharm. Tech. Prod. Manuf. 1983, 4, 47–48.

(21) Braun, S.; Kalinowski, H. O.; Berger, S. 150 and More Basic NMR

Experiments, 2nd ed.; New York: Wiley-VCH, 1998.

(22) Harris, R. K.; Becker, E. D.; Cabral de Menezes, S. M.; Goodfellow,

R.; Granger, P. Pure Appl. Chem. 2001, 73, 1795–1818.

(23) Earl, W. L.; Vanderhart, D. L. J. Magn. Reson. 1982, 48, 35–54.

(24) Ratcliffe, C. I.; Ripmeester, J. A.; Tse, J. S. Chem. Phys. Lett. 1983,

99, 177–180.

Acknowledgment. The authors thank Dr. Leo Hsu

(GlaxoSmithKline) for performing the chiral preparative

chromatographic separation of racemic carvedilol and Pro-

fessor Judith Howard (University of Durham), in whose

laboratories one of us (O.V.C.) prepared the crystals of I.

The authors alsothankMr. Jerome Socha (GlaxoSmithKline)

(25) Metz, G.; Wu, X.; Smith, S. O. J. Magn. Reson. A 1994, 110, 219–

227.

(26) Antzutkin, O. N. Prog. NMR Spectros. 1999, 35, 203–266.

(27) Fung, B. M.; Khitrin, A. K.; Ermolaev, K. J. Magn. Reson. 2000,

142, 97–101.

Article

Crystal Growth & Design, Vol. 10, No. 6, 2010 2733

(28) Opella, S. J.; Frey, M. H. J. Am. Chem. Soc. 1979, 101, 5854–

5856.

(29) Feike, M.; Demco, D. E.; Graf, R.; Gottwald, J.; Hafner, S.; Spiess,

H. W. J. Magn. Reson. A 1996, 122, 214–221.

(30) Hohwy, M.; Jakobsen, H. J.; Eden, M.; Levitt, M. H.; Nielsen,

N. C. J. Chem. Phys. 1998, 108, 2686–2694.

(47) Chion, B.; Lajzerowicz, J.; Bordeaux, D.; Collet, A.; Jacques, J.

J. Phys. Chem. 1978, 82, 2682–2688.

(48) Renou, L.; Morelli, T.; Coste, S.; Petit, M.-N.; Berton, B.;

Malandain, J.-J.; Coquerel, G. Cryst. Growth Des. 2007, 7, 1599–

1607.

(49) Wermester, N.; Aubin, E.; Pauchet, M.; Coste, S.; Coquerel, G.

Tetrahedron: Asymmetry 2007, 18, 821–831.

€

(31) van Rossum, B. J.; Forster, H.; de Groot, H. J. M. J. Magn. Reson.

1997, 124, 516–519.

(32) Lesage, A.; Sakellariou, D.; Steuernagel, S.; Emsley, L. J. Am.

(50) Vogt, F. G.; Vena, J. A.; Chavda, M.; Clawson, J. S.; Strohmeier,

M.; Barnett, M. E. J. Mol. Struct. 2009, 932, 16–30.

(51) Wallach, O. Liebigs Ann. Chem. 1895, 286, 90–143.

(52) Brock, C. P.; Schweizer, W. B.; Dunitz, J. D. J. Am. Chem. Soc.

1991, 113, 9811–9820.

(53) Jeffrey, G. A.; Yeon, Y. Acta Crystallogr. 1986, B42, 410–413.

(54) Authelin, J. R. Int. J. Pharm. 2005, 303, 37–53.

(55) Perrier, P. R.; Byrn, S. R. J. Org. Chem. 1982, 47, 4671–4676.

(56) Eckhardt, R. C.; Flanagan, T. B. Trans. Faraday Soc. 1964, 60,

1289–1298.

(57) Clarke, T. A.; Thomas, J. M. J. Chem. Soc. 1969, 2227–30.

(58) Ahlqvist, M. U. A.; Taylor, L. S. Int. J. Pharm. 2002, 241, 253–

261.

(59) Vogt, F. G.; Brum, J.; Katrincic, L. M.; Flach, A.; Socha, J. M.;

Goodman, R. M.; Haltiwanger, R. C. Cryst. Growth Des. 2006, 6,

2333–2354.

Chem. Soc. 1998, 120, 13194–13201.

(33) Zumbulyadis, N.; Antalek, B.; Windig, W.; Scaringe, R. P.;

Lanzafame, A. M.; Blanton, T.; Helber, M. J. Am. Chem. Soc.

1999, 121, 11554–11557.

(34) Davis, J. H; Jeffrey, K. R.; Bloom, M.; Valic, M. I.; Higgs, T. P.

Chem. Phys. Lett. 1976, 42, 390–394.

(35) The crystal structure data for I-(R) and racemic I have been

deposited with the Cambridge Crystallographic Data Center under

deposition numbers CCDC 775400 and 775401. This material can

be obtained free of charge via www.ccdc.cam.ac.uk/data_request/

cif, by emailing data_request@ccdc.cam.ac.uk, or by contacting

the Cambridge Crystallographic Data Centre, 12, Union Road,

Cambridge CB2 1EZ, UK; fax: þ44 1223 336033.

(36) Pawley, G. S. J. Appl. Crystallogr. 1981, 14, 357–361.

(37) Herbst-Irmer, R.; Sheldrick, G. M. Acta Crystallogr. 1998, B54,

443–449.

(60) Lemaitre, V; Smith, M. E.; Watts, A. Solid State NMR 2004, 26,

215–235.

(61) Zhang, Q. W.; Zhang, H. M.; Usha, M. G.; Wittebort, R. J. Solid

(38) Parsons, S. Acta Crystallogr. 2003, D59, 1995–2003.

(39) Hunter, C. A.; Lawson, K. R.; Perkins, J.; Urch, C. J. J. Chem.

Soc., Perkin Trans. 2001, 2, 651–669.

(40) Johnson, C. E.; Bovey, F. A. J. Chem. Phys. 1958, 29, 1012–1014.

(41) Brown, S. P.; Spiess, H. W. Chem. Rev. 2001, 101, 4125–4155.

(42) Vogt, F. G.; Katrincic, L. M.; Long, S. T.; Mueller, R. L.; Carlton,

R. A.; Sun, Y. T.; Johnson, M. N.; Copley, R. C. B.; Light, M. E.

J. Pharm. Sci. 2008, 97, 4756–4782.

(43) Vogt, F. G.; Clawson, J. S.; Strohmeier, M.; Edwards, A. J.; Pham,

T. P.; Watson, S. A. Cryst. Growth Des. 2009, 9, 921–937.

(44) Schmidt auf der Gunne, J.; Eckert, H. Chem.;Eur. J. 1998, 4,

1762–1766.

State NMR 1996, 7, 147–154.

(62) Ueda, T.; Nakamura, N. Z. Naturforsch. 2000, 55a, 362–368.

(63) Butler, L. G. In ¹⁷O NMR Spectroscopy in Organic Chemistry;

Boykin, D. W., Ed.; CRC Press: Boca Raton, FL, 1991; pp 2-17.

(64) Harrick, N. J. Internal Reflection Spectroscopy; Wiley: New York,

1967.

(65) Weiss, A.; Weiden, N. In Advances in Nuclear Quadrupole Reso-

nance; Smith, J. A. S., Ed.; Heyden & Son Ltd.: London, UK, 1980; Vol.

4, pp 149-248.

(66) Barnes, R. G. In Advances in Nuclear Quadrupole Resonance;

Smith, J. A. S., Ed.; Heyden & Son Ltd.: London, UK, 1974; Vol. 1,

pp 335-355.

(45) Coquerel, G. Enantiomer 2000, 5, 481–498.

(46) Huang, J.; Chen, S.; Guzei, I. A.; Yu, L. J. Am. Chem. Soc. 2006,

128, 11985–11992.

(67) Tian, J.; Thallapally, P. K.; Dalgarno, S. J.; Atwood, J. L. J. Am.

Chem. Soc. 2009, 131, 13216–13217.

Article Doi

Isomorphism, disorder, and hydration in the crystal structures of racemic and single-enantiomer carvedilol phosphate

DOI: 10.1021/cg100209v

Source and publish data:

Authors:

Article abstract of DOI:10.1021/cg100209v

Full text of DOI:10.1021/cg100209v

Products guided by the article

R&D Labs maybe for 374779-45-0

Relevant to this article

Hot Product