Chemical Mapping of Pharmaceutical Cocrystals Using TerahertzSpectroscopic ImagingDanielle M. Charron,† Katsuhiro Ajito,* Jae-Young Kim, and Yuko Ueno

NTT Microsystem Integration Laboratories, NTT Corporation, 3-1 Morinosato Wakamiya, Atsugi, Kanagawa 243-0198 Japan

*S Supporting Information

ABSTRACT: Terahertz (THz) spectroscopic imaging is apromising technique for distinguishing pharmaceuticals ofsimilar molecular composition but differing crystal structures.Physicochemical properties, for instance bioavailability, aremanipulated by altering a drug’s crystal structure throughmethods such as cocrystallization. Cocrystals are molecularcomplexes having crystal structures different from those oftheir pure components. A technique for identifying the two-dimensional distribution of these alternate forms is required.Here we present the first demonstration of THz spectroscopicimaging of cocrystals. THz spectra of caffeine−oxalic acidcocrystal measured at low temperature exhibit sharp peaks,enabling us to visualize the cocrystal distribution in nonuni-form tablets. The cocrystal distribution was clearly identified using THz spectroscopic data, and the cocrystal concentration wascalculated with 0.3−1.3% w/w error from the known total concentration. From this result, THz spectroscopy allows quantitativechemical mapping of cocrystals and offers researchers and drug developers a new analytical tool.

Terahertz (THz) spectroscopy is a useful technique forchemical recognition based on both molecular and crystal

structure fingerprinting. Occupying the part of the electro-magnetic spectrum between microwaves and infrared radiation,the THz region corresponds to the frequency range between0.1 and 10 THz (3−333 cm−1). Absorption of THz waves isdominated by weak intermolecular and intramolecular inter-actions within molecular networks, such as hydrogen bondingand van der Waals forces. Separation of intermolecular andintramolecular modes for assignment of absorption bands1−3 isboth challenging and ambiguous, making it difficult to isolatemolecular and crystal structures. It is useful, however, that thesemodes are highly collective, involving vibrations in multipleregions of a molecule.4 Thus, THz spectra are more stronglydependent on molecular arrangement and intermolecularinteractions than on molecular composition. For this reason,closely related chemicals such as polymorphs can easily bedistinguished by THz spectroscopy.5−8 Similarly, cocrystals canbe differentiated from mixtures of their pure components.Cocrystals are solid molecular complexes of two or moreneutral components bound together in stoichiometric amountsby noncovalent intermolecular interactions.9−11

Cocrystallization plays a similar role to polymorphism inpharmaceutical development. Pharmaceutical cocrystals containan active pharmaceutical ingredient (API). In its pure form theAPI is itself crystalline, but after cocrystallization the APImolecules are interspersed within the new crystal structure.This new structure is comprised of the API molecules and oneor more accessory component molecules. These additionalcomponents are called coformers, and their selection plays a

determinant role in dictating the physicochemical properties ofthe cocrystal such as solubility, solution stability, andprocessability.12 Physicochemical properties of crystallinematerials are dependent not only on their molecular speciesbut on their crystal structures. By adjusting synthesisconditions, cocrystals may take multiple polymorphic formsowing to their conformational flexibility.13 In the pharmaceut-ical industry, the purpose of crystallizing cocrystals andpolymorphs is to improve drug properties in order to enhancebioavailability without altering the chemical nature, and thusbioactivity, of an API. Differences in crystal structure betweencocrystallized and pure pharmaceuticals, therefore, affect thedosage response of a drug. Furthermore, chemical imagingmethods are required for observing and evaluating thedistribution of cocrystallized pharmaceuticals in tablets.Chemical mapping has been used to distinguish genuine and

counterfeit drugs, as API distribution reflects tablet processingconditions.14 In addition, spatial nonuniformity affects drugrelease kinetics and therapeutic efficiency;15 therefore, chemicalmapping is important in process monitoring and control.Techniques such as Raman microscopy and near-infraredspectroscopic imaging are advantageous for chemical mappingnear the tablet surface; however, strong scattering precludesbulk imaging. THz radiation is lower in energy and so is capableof penetrating pharmaceutical tablets16,17 and biological

Received: October 30, 2012Accepted: January 28, 2013Published: January 28, 2013

Letter

pubs.acs.org/ac

© 2013 American Chemical Society 1980 dx.doi.org/10.1021/ac302852n | Anal. Chem. 2013, 85, 1980−1984

samples,18 enabling nondestructive characterization of the bulkchemical distribution. It has also been shown that quantitativeanalysis is possible with THz spectroscopy.19−21 Previous THzspectroscopic imaging analysis has focused on the comparisonof absorption images at different THz frequencies to identifylarge regions of desired chemicals within nonuniform tablets6

and to distinguish homogeneous tablets comprised of differentchemicals.7,22,23 Moreover, low temperature measurement hasbeen shown to be advantageous for chemical mapping sincemeasured THz spectra exhibit sharper peaks due to narrowingin the distribution of populated vibrational modes.11

Here we present the first demonstration of chemical mappingof cocrystal distributions. Imaging test tablets includingnonuniform caffeine−oxalic acid cocrystal powders wereprepared and investigated by THz spectroscopic imaging. Thecocrystal of the API caffeine and the coformer oxalic acid waschosen, as the crystal structure is simple and known inpharmaceutical research.24 Spectral analysis was performed toquantitatively determine the distribution of caffeine−oxalicacid, identifying the cocrystal as separate from its purecomponents caffeine and oxalic acid. The total cocrystalconcentration was determined with between 0.3 and 1.3% w/w error from the known concentration.

■ EXPERIMENTAL SECTIONCocrystal Synthesis. Synthesis of caffeine−oxalic acid

cocrystal was performed following a modified version of apreviously reported solution precipitation technique.20 A 2:1molar ratio of caffeine (Tokyo Chemical Industry Co., Ltd.)and anhydrous oxalic acid (Wako Pure Chemical Industries,Ltd.) dissolved in chloroform and methanol (Kanto ChemicalCo., Inc.) was heated at 330 K for 20 min under reflux. Thesolution was allowed to cool to room temperature, thenrotoevaporated for 15 min. A white, cotton-like crystal wasextracted by vacuum filtration and then dried under lowvacuum overnight.Powder X-ray Diffraction (PXRD). Cocrystallization was

verified by PXRD using a Rigaku RU-300 rotating-anodegenerator (0.154 nm Cu source, 40 kV voltage, 200 mAfilament emission). Powdered cocrystal and caffeine sampleswere scanned from 7 to 30° 2θ with a step size of 0.02° 2θ.Raman Spectroscopy. Cocrystallization was also con-

firmed by Raman spectrometry (inVia Reflex Raman micro-scope, Renishaw) using a 532 nm line laser-diode continuous-wave laser (H39969, Renishaw) at 1 mW power. Spectra ofpowdered samples were collected at room temperature, with anobjective lens (×100, NA 1.85) providing a spectral resolutionof 1 cm−1.Tableting. Polyethylene was selected as a test tablet matrix

material because of its low absorption and the absence of largepeaks in its terahertz spectrum. The cocrystal was ground topowder and mixed with polyethylene (Aldrich) to dilute to 0, 5,10, 15, and 20% cocrystal by weight. The homogeneous samplemixtures were compressed at 40 kg/cm2 for 2 min to formtablets 10 mm in diameter, 1.5−2.0 mm thick. Homogeneoustablets of 20% caffeine or oxalic acid in polyethylene weresimilarly prepared. Two nonuniform samples were made byarranging a mixture of cocrystal and polyethylene on one sideof a tablet, with either pure polyethylene or a mixture ofpolyethylene, caffeine, and oxalic acid on the opposite side.THz-TDS. Cocrystallization was confirmed by THz time-

domain spectroscopy (THz-TDS) using a commercial spectro-scope (THz-TDS 2004, AISPEC) with a 13 fs near-infrared

pulse laser (Fusion, Femtolasers). Spectra of the 0, 5, 10, 15,and 20% cocrystal, 20% caffeine, and 20% oxalic acid tabletswere collected at 120 K in a vacuum chamber and averagedover 32 acquisitions. All spectra were normalized for tabletthickness, and the pure polyethylene spectrum was subtractedfrom each sample spectrum.

THz-Imaging. The THz-imaging setup is shown in Figure1. It consists of a THz-TDS system using a 9 fs near-infrared

pulse laser (Integral Pro, Femtolasers), with the samplechamber mounted on a translational stage. A detaileddescription of the apparatus was reported elsewhere.7 Acryostat was also added to the sample chamber of the THz-imaging system. Images with a step size of 200 μm and spatialresolution of about 600 μm at 1 THz were taken, rastering overa 12 mm × 12 mm scanning area. The two nonuniform tabletsand the 0, 10, and 20% cocrystal tablets were imaged. Allimages were collected at the lowest preset temperatureachievable with the THz-imaging system, 120 K under vacuum,and took approximately 7 h to complete. The THz-imagingtemperature was confirmed by THz-TDS cocrystal peakposition temperature calibration.Samples were set into a metal holder mounted to a quartz

plate so that the metal formed a ring around each tablet. Pixelscorresponding to the metal sample holder were removed fromthe image data. Each image was also normalized for tabletthickness. This and all further spectral analysis was completedusing MATLAB (MathWorks).

■ RESULTS AND DISCUSSIONTHz-TDS Cocrystal Spectra. Cocrystallization was readily

confirmed by THz-TDS. Caffeine−oxalic acid exhibits threestrong absorption peaks (νi) at 1.35, 1.43, and 1.56 THz ataround 120 K, whereas caffeine and oxalic acid exhibit noabsorption peaks within the 1−2 THz range studied (Figure2a). The oxalic acid spectrum compares well with a previouslyreported THz spectrum.25 Cocrystallization was verified byPXRD (Figure S-1 in the Supporting Information). Thecaffeine and cocrystal diffraction patterns are similar to reportedmeasurements.26 These results are also supported by Ramanspectroscopy (Figure S-2 in the Supporting Information). Thecocrystal Raman spectrum exhibits a distinct new peak at 899cm−1 that is not observed in either the pure caffeine or oxalicacid spectra. Caffeine and oxalic acid Raman spectra are similarto reported spectra.27,28

Figure 2b illustrates the changes in cocrystal spectral shapemeasured from 298 K down to 120 K. The result clearlyindicates that terahertz spectra measured at low temperaturesexhibit sharper peaks. A blue shift is also observed withdecreasing temperature.

Figure 1. Diagram of the terahertz imaging system.

Analytical Chemistry Letter

dx.doi.org/10.1021/ac302852n | Anal. Chem. 2013, 85, 1980−19841981

THz-TDS was used to verify observed trends in THz-imaging and any spectral analysis assumptions made. Eachimage pixel corresponds to only one spectral acquisition,whereas the THz-TDS spectra are averaged over multipleacquisitions, providing much higher quality spectra. For the 5,10, 15, and 20% cocrystal THz spectra at 120 K (Figure 2c), thetrend in optical density at ν3 is well approximated by a linearpolynomial (Figure S-3 in the Supporting Information, y = ax +b, a = 0.026, b = 0.050). Attenuation of THz waves is thereforedependent on cocrystal concentration, satisfying the Beer−Lambert Law for absorbance. From this result it is expectedthat the THz-imaging system can be calibrated linearly and thatTHz spectroscopy may be used for quantitative analysis.THz Mapping System Performance. The most intense

spectral peak ν3 will be used to identify the cocrystal because itoffers the highest signal-to-noise ratio and is separate from ν1and ν2. THz images of the 0, 10, and 20% cocrystal tablets(Figure 3) were used to calibrate the imaging system forcocrystal concentration. The ν3 peak transmittance wasaveraged across all pixels within the tablet region for eachcalibration standard image. Transmittance as a function ofsample concentration provided a linear calibration (Figure S-4in the Supporting Information, y = ax + b, a = −0.942 ± 0.109,b = 66.0 ± 1.20). Error is due primarily to nonuniformity incocrystal concentration within the calibration standard imagesand due to fluctuation in THz pulse intensity. No shift in peakfrequency was detected across the images, so it is unlikely thatvariance observed is due to a change in temperature. The 10%cocrystal sample was slightly damaged along the edge prior toimaging; however, the average transmission 54.5 cm−1 has astandard deviation of only 1.16 cm−1, comparable to the 20%cocrystal tablet.Nonuniform Cocrystal Mapping. Figure 4a shows visible

light photographs of the two nonuniform sample tablets.Regions of concentrated chemicals are indicated in Figure 4b,though their distribution cannot be seen in the photographs orin the THz pulse intensity images (Figure 4c). THz spectraldata was used to map these unknown regions. Cocrystal pixelswithin the two nonuniform tablets were identified by thepresence of ν3. A frequency range is used for identificationinstead of the absolute peak frequency to account for high noisein the image spectra and local temperature fluctuations.Presence of this peak was determined within ν3 ± 0.5Δν3THz. Within this range, the minimum transmittance is assumedto be the peak intensity only if the intensities at the twoneighboring frequencies are comparable to each other. If this isnot true, the minimum transmittance within the studied range

is assumed to be due to noise, and instead the absolutetransmittance at ν3 is taken as the peak intensity.The cocrystal concentration (Figure 4d) for each pixel

previously identified as containing cocrystal in the twononuniform tablets was determined using the THz-imagingconcentration calibration calculated previously. The absoluteerror in the calculated concentration due to variation in thehomogeneous calibration standards is given in Figure 4e.Propagated error is proportional to the calculated concen-tration, not to the concentration-dependent signal-to-noiseratio of ν3.For the tablet containing a nonuniform distribution of

caffeine−oxalic acid in polyethylene, the total cocrystalconcentration was found to be 5.25% by weight, only 0.352%more than the known concentration. Similarly, for the tablet

Figure 2. (a) Terahertz spectra of caffeine (dashed line), oxalic acid (dotted line), and their cocrystal (solid line) measured at 120 K. Theconcentration of each tableted chemical is 20%. (b) Temperature-dependent THz spectra of caffeine-oxalic acid cocrystal measured at 298 K (red),220 K (green), 170 K (blue), and 120 K (black). (c) Concentration-dependent THz spectra of caffeine-oxalic acid cocrystal measured at 120 K.Tableted cocrystal concentrations are 5% (black), 10% (blue), 15% (green), and 20% (red).

Figure 3. Chemical mapping of (a) 20%, (b) 10%, and (c) 0%caffeine−oxalic acid cocrystal calibration standard tablets using theTHz-imaging system set to 120 K. Left, THz pulse amplitude image(dark blue corresponds to metal sample holder); right, transmittanceat ν3 peak.

Analytical Chemistry Letter

dx.doi.org/10.1021/ac302852n | Anal. Chem. 2013, 85, 1980−19841982

containing cocrystal in caffeine, oxalic acid, and polyethylene,the total cocrystal concentration was calculated to be 8.61%, anoverestimate of 1.24%. Overestimation may be due to error inthe concentration calibration or to fluctuations in temperatureand THz pulse intensity during imaging. In addition, at theinterface between the tablet and the metal sample holder, THzradiation is scattered independent of frequency due to thechange in density between the two materials. Therefore,transmittance near the edge of the imaged tablet may be lowerthan without this halo effect.For both nonuniform tablets, the cocrystal was successfully

mapped based on its spectral information as separate from thediluent polyethylene and its pure precursor components,caffeine and oxalic acid. For more practical investigation oftablets in the pharmaceutical market, various kinds of tabletmatrix materials should be tested. Furthermore, more complexdistributions of cocrystals should be analyzed, includingevaluation of smaller-scaled domains and differences betweensurface and inner regions of tablets. Assembling a THz

spectrum database of cocrystals and other excipients willallow us to analyze each component quantitatively, even ifseveral kinds of chemicals are mixed in the focal spot of theterahertz beam.29

To improve the spatial resolution beyond the opticaldiffraction limit, near-field THz imaging techniques withsubwavelength resolution have been proposed.30 Recentprogress in near-field imaging has demonstrated spatialresolution down to λ/1000 (150 nm at 2 THz)31,32 whichmay allow THz imaging of pharmaceutical crystal powders. Inaddition, the spatial resolution in the z-direction of our systemis over 2 mm; therefore, no z-direction information is includedin the THz mapping data. The use of confocal imagingtechniques33,34 in the THz mapping system would provide z-direction chemical distribution in tablets. The use of frequencytunable THz lasers instead of THz pulse beams is a veryimportant factor in reducing measurement time for practicalapplications by reducing the delay-line scanning time. Suchsuitable lasers for terahertz spectroscopic imaging systems arenow under development.

■ CONCLUSIONS

We have shown the first report demonstrating THz chemicalmapping of pharmaceutical cocrystals. Cocrystallization is amethod used to manipulate the physicochemical properties ofdrugs. Accurate chemical mapping is important for drug designand development. THz spectra of caffeine-oxalic acid cocrystalmeasured at low temperature exhibit sharp peaks, enabling usto visualize the distribution of cocrystal in tablets. For thenonuniform tablets, the cocrystal caffeine−oxalic acid wassuccessfully mapped based on its spectral information asseparate from its pure precursor components, caffeine andoxalic acid. Quantitative analysis using a linear concentrationcalibration was also shown to be accurate for determining thetotal cocrystal concentration. This result broadens theprospective applications of THz spectroscopy in pharmaceuticalchemical imaging.

■ ASSOCIATED CONTENT

*S Supporting InformationX-ray diffraction patterns and Raman spectra of powderedsamples and calibration plots for cocrystal concentration usingTHz-TDS and THz-imaging. This material is available free ofcharge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Phone: +81-46-240-3565. Fax: +81-46-240-4041. E-mail: ajito.katsuhiro@lab.ntt.co.jp.

Present Address†University of Waterloo, 200 University Ave. W., WaterlooOntario, N2L 3G1 Canada.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

We thank Ms. Maro Yamaguchi of WDB for her kind assistancewith the THz measurements and Dr. Ho-Jin Song, Mr. MakotoYaita, Dr. Naoya Kukutsu, and Mr. Osamu Kagami of NTT fortheir encouragement.

Figure 4. Chemical mapping of caffeine−oxalic acid cocrystal innonuniform tablets with the THz-imaging system set to 120 K. Left,cocrystal in polyethylene; right, cocrystal in polyethylene with caffeineand oxalic acid. (a) Photograph of imaging tablet, (b) diagram ofregions of high sample concentration within tablet, (c) THz pulseamplitude image, (d) cocrystal fraction by weight at ν3 peak for eachidentified pixel, and (e) absolute error in calculated cocrystalconcentration. Approximate tablet position is outlined by a dashedblack line in parts d and e.

Analytical Chemistry Letter

dx.doi.org/10.1021/ac302852n | Anal. Chem. 2013, 85, 1980−19841983

■ REFERENCES(1) Ueno, Y.; Rungsawang, R.; Tomita, I.; Ajito, K. Chem. Lett. 2006,35, 1128−1129.(2) Ueno, Y.; Ajito, K. Anal. Sci. 2007, 23, 803−807.(3) Williams, M.; True, A.; Izmaylov, A.; French, T.; Schroeck, K.;Schmuttenmaer, C. Phys. Chem. Chem. Phys. 2011, 13, 11719−11730.(4) Castro-Camus, E.; Johnston, M. Chem. Phys. Lett. 2008, 455,289−292.(5) King, M.; Buchanan, W.; Korter, T. Anal. Chem. 2011, 83, 3786−3792.(6) Ajito, K.; Ueno, Y.; Song, H.-J.; Tamechika, E.; Kukutsu, N. ECSTrans. 2011, 35, 157−165.(7) Ajito, K.; Ueno, Y.; Song, H.-J.; Tamechika, E.; Kukutsu, N. Mol.Cryst. Liq. Cryst. 2011, 538, 33−38.(8) Taday, P. F.; Bradley, I. V.; Arnone, D. D.; Pepper, M. J. Pharm.Sci. 2003, 92, 831−838.(9) Friscic, T.; Jones, W. J. Pharm. Pharmacol. 2010, 62, 1547−1559.(10) Schultheiss, N.; Newman, A. Cryst. Growth Des. 2009, 9, 2950−2967.(11) Shen, Y.; Upadhya, P.; Linfield, E.; Davies, A. Appl. Phys. Lett.2003, 82, 2350−2352.(12) Qiao, N.; Li, M.; Schlindwein, W.; Malek, N.; Davies, A.;Trappitt, G. Int. J. Pharm. 2011, 419, 1−11.(13) Schultheiss, N.; Henck, J.-O. Role of co-crystals in thepharmaceutical development continuum. In Pharmaceutical Salts andCo-crystals; Wouters, J., Quere, L., Eds.; Royal Society of Chemistry:Cambridge, U.K., 2012; pp 110−125.(14) Kwok, K.; Taylor, L. J. Pharm. Biomed. Anal. 2012, 66, 126−135.(15) Franch-Lage, F.; Amigo, J.; Skibsted, E.; Maspoch, S.; Coello, J.Int. J. Pharm. 2011, 411, 27−35.(16) Zeitler, J. A.; Shen, Y. C.; Baker, C.; Taday, P. F.; Pepper, M.;Rades, T. J. Pharm. Sci. 2007, 96, 330−340.(17) Russe, I. S.; Brock, D.; Knop, K.; Kleinebudde, P.; Zeitler, J. A.Mol. Pharm. 2012, 9, 3551−3559.(18) Oh, S.; Kang, J.; Maeng, I.; Suh, S.-J.; Huh, Y.-M.; Haam, S.;Son, J.-H. Opt. Exp. 2009, 17, 3469−3475.(19) Ueno, Y.; Rungsawang, R.; Tomita, I.; Ajito, K. Anal. Chem.2006, 78, 5424−428.(20) Nguyen, K.; Frisccaron, T.; Day, G.; Gladden, L.; Jones, W. Nat.Mater. 2007, 6, 206−209.(21) Taday, P. F. Trans. R. Soc. London A 2004, 362, 351−364.(22) Watanabe, Y.; Kawase, K.; Ikari, T.; Ito, H.; Ishikawa, Y.;Minamide, H. Opt. Commun. 2004, 234, 125−129.(23) Kawase, K.; Ogawa, Y.; Watanabe, Y. Opt. Express 2003, 11,2549−2554.(24) Trask, A.; Motherwell, W.; Jones, W. Cryst. Growth Des. 2005, 5,1013−1021.(25) King, M.; Korter, T. J. Phys. Chem. A 2010, 114, 7127−7138.(26) Aher, S.; Dhumal, R.; Mahadik, K.; Ketolainen, J.; Paradkar, A.Pharm. Dev. Technol. 2013, 18, 55−60.(27) Hedoux, A.; Decroix, A.; Guinet, Y.; Paccou, L.; Derollez, P.;Descamps, M. J. Phys. Chem. 2011, 115, 5746−5753.(28) Jeziorowski, H.; Moser, B. Chem. Phys. Lett. 1985, 120, 41−44.(29) Ueno, Y.; Ajito, K.; Kukutsu, N.; Tamechika, E. Anal. Sci. 2011,27, 351−356.(30) Hunsche, S.; Koch, M.; Brener, I.; Nuss, M. C. Opt. Commun.1998, 150, 22−26.(31) Blanchard, F.; Doi, A.; Tanaka, T.; Hirori, H.; Tanaka, H.;Kadoya, Y.; Tanaka, K. Opt. Exp. 2011, 19, 8277−8284.(32) Chen, H.-T.; Kerstinga, R.; Cho, G. C. Appl. Phys. Lett. 2003, 83,3009−3011.(33) Salhi, M. A.; P.a, I.; Koch, M. J. Infrared Millimeter TerahertzWaves 2010, 31, 358−366.(34) Cumis, U. S.; Xu, J.-H.; Masini, L.; Degl’Innocenti, R.; Pingue,P.; Beltram, F.; Tredicucci, A.; Vitiello, M. S.; Benedetti, P. A.; Beere,H. E.; Ritchie, D. A. Opt. Exp. 2012, 20, 21924−21931.

Analytical Chemistry Letter

dx.doi.org/10.1021/ac302852n | Anal. Chem. 2013, 85, 1980−19841984