J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Raman Spectroscopy for Quantitative Analysis in the Pharmaceutical Industry

José Izo Santana da Silva de Jesus1, Raimar Löbenberg2, Nádia Araci Bou-Chacra1

1.Department of Pharmacy, Faculty of Pharmaceutical Sciences, University of São Paulo, Professor Lineu Prestes Av 580, Cidade Universitária, 05508-000 São Paulo, SP, Brazil. 2.Faculty of Pharmacy and Pharmaceutical Sciences, Katz Group-Rexall Centre for Pharmacy & Health Research, University of Alberta, Edmonton, Alberta, Canada.

Received, July 24, 2019; Revised, October 27, 2019; Accepted, January 3, 2020; Published, February 27, 2020.

ABSTRACT - Raman spectroscopy is a very promising technique increasingly used in the pharmaceutical industry. Due to its development and improved instrumental versatility achieved over recent decades and through the application of chemometric methods, this technique has become highly precise and sensitive for the quantification of drug substances. Thus, it has become fundamental in identifying critical variables and their clinical relevance in the development of new drugs. In process monitoring, it has been used to highlight in-line real-time analysis, and it has been used more commonly since 2004 when the Food and Drug Administration (FDA) launched Process Analytical Technology (PAT), integrated with the concepts of Pharmaceutical Current Good Manufacturing Practices (CGMPs) for the 21st Century. The present review presents advances in the application of this tool in the development of pharmaceutical products and processes in the last six years. ______

INTRODUCTION solvents; it is easy to use, and the analyses can be without any preparation, even penetrating primary Raman and Krishnan discovered Raman packaging, such as polymeric materials and spectroscopy in 1928, and this technique has gone transparent glasses. In addition, portable types of several breakthroughs between the 1930s and equipment are available, and the analytical results 1950s. Currently, it is one of the leading analytical can be obtained in seconds (1,8). techniques among spectroscopies (1–3). This Therefore, this review presents the advances of technique, based on light scattering during Raman spectroscopy application in the quantitative monochromatic radiation exposure to samples, analyses in the product development cycles and as involves molecular vibration of the chemical a tool in process analytical technology (PAT). structures of a substance (1,4). The monochromatic laser beam illuminates the Main variants of Raman spectroscopy samples resulting in scattered light, from photon- In addition to the conventional Raman dispersion molecule interactions. Each photon occurs in a phenomenon, several studies, over recent decades, different vibrational mode, and this frequency have made possible the discovery of other events, difference refers to the separation of the vibrational which allowed the development of several variants energy level of the molecules. The dispersion of this spectroscopy (1,4,6). The difference signals origins from the inelastic movement frequency of incident radiation and the intensity of between the incident monochromatic radiation and the dispersion signals originates Stokes and anti- vibrational molecular motions, providing a unique Stokes lines, as shown in Figure 1. These signature for each substance (1,4–6). phenomena define the basic principle for the In the pharmaceutical industry, Raman variants presented in Table 1 (5,9). The spectroscopy is an excellent tool for identifying instrumental versatility of Raman spectroscopy counterfeit drugs (1,7). In addition, it is suitable in includes interferometric dispersive systems, single product development and the real-time monitoring channel multichannel detection systems, and of productive processes through Process Analytical microscope coupled systems (Table 1). Several Technology (PAT), according to the current laser systems are available for concepts of Pharmaceutical Current Good ______st Manufacturing Practices (CGMPs) for the 21 Corresponding Author: Nádia Araci Bou-Chacra, Century (1,8). Department of Pharmacy, Faculty of Pharmaceutical Sciences, Raman spectroscopy has essential advantages University of São Paulo, Professor Lineu Prestes Av 580, due to its non-invasive feature; it does not use Cidade Universitária, 05508-000 São Paulo, SP, Brazil; E- mail: [email protected]

24 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

A. Vibrational energy levels

E1

Stokes Rayleigh Anti-Stokes Virtual energy levels Energy

Vibrational energy levels

E 0 B. Rayleigh

Stokes Anti-Stokes Intensity Intensity

1000 0 - 1000 -1 Wavenumbers/cm Figure 1. Energetic transitions involved in Raman scattering, A. More intense (blue shifted) Stokes bands, and less intense ( shifted) anti-Stokes bands. Rayleigh line to higher wavenumbers (green shifted), without bands of Raman spectra, B. excitation and filtering units for spectral Conventional Raman spectroscopy requires purification. Typically, the characteristics of the high sample concentration and it is affected by devices vary depending on the size and relative fluorescence. FT-Raman can resolve such spectral cost of its components and accessories (5). interference (11). Furthermore, fluorescence The diversity of the types of device interference also can be reduced by exciting the configurations allows amplifying the sensitivity of near infrared laser sample as Nd-YAG at 1064 nm the technique and allows a variety of applications, (5). ranging from the quantification of drug substances For samples at low concentrations, Resonance in tablets and mixtures of powders to multiplex Raman Spectroscopy (RRS) increases the samples (3,10–17), samples in low concentrations scattering sensitivity. In such cases, the samples (18,19), molecular traces (20–22), samples inside must have electronic transitions close to the laser packs (23) and materials in nanometric scale line´s frequency. Thus, the incident radiation (Table 1) (24,25). frequency coincides with an electronic transition of

25

J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020 the molecule, and as a result of this combination, a (PLS) method (10,38,39), described by Wold in more intense Raman spectrum is obtained (18,19). 1966 (40). Furthermore, for multicomponent In the Surface-enhanced Raman spectroscopy matrices, classical least squares (CLS), alternating (SERS), the sample is adsorbed on a colloidal least squares (MCR), principal component metal surface or by nanostructures such as regression (CRP), and principal component nanotubes typically made of gold (Au), silver (Ag) analysis (PCA) are commonly used (41). or copper (Cu) (20–22). This approach improves For the development of the calibration models, the intensity of signals and also extinguishes carefully selected representative samples should be fluorescence. used, which generally require qualification by Another variant, the Tip-enhanced Raman independent reference analytical procedures (36). spectroscopy (TERS), provides high chemical The selection of these models depends on the type sensitivity for surface molecular mapping with a of data and analytical objectives. In some cases, to nanoscale spatial resolution (24,25). This variant achieve better precision of the quantitative method, may help researchers to characterize and to it is necessary to compare several multivariate develop nano-based strategies in innovative models. Thus, it is possible to detect errors and therapy. Furthermore, Micro-Raman presents a allow selecting the appropriate model (41–43). wide area of illumination, which allows In 2015, the FDA (48) released the Analytical overcoming the challenges of quantifying Procedures and Methods Validation for Drugs and polymorphic mixtures (14). Biologics Guidance for Industry. This guidance All these variants were carefully developed was harmonized with ICH to complement the aiming to overcome the limitations of the guidance Q2(R1) Validation of Analytical conventional Raman spectroscopy. By using Procedures: Text and Methodology (49). The chemometric treatments, it is possible to improve guidelines provide the validation criteria to achieve analytical models for the quantification of drugs a successful calibration model to support the substances and excipients in multicomponent quantitative analysis in pharmaceutical formulas (10,16,26), as presented in Table 1. applications. According to these official documents, to Raman calibration model define a robust calibration model for quantification The interpretation of a complex spectrum of it is required the evaluation of the different sources samples requires the use of chemometric of variability and includes them in the samples. calibration models. is the use of These variabilities must be identified to predict and mathematical and statistical methods in chemical control them during method validation and routine data, allowing the acquisition and extraction of application (36). essential information regarding the components of The performance characteristics for the the formulation, as shown in Figure 2 (27). For this, spectroscopic methods is similar to the it is necessary to apply a univariate or multivariate conventional ones and include the evaluation of analysis. accuracy, precision (repeatability and intermediate The univariate analysis contemplates alone precision), specificity, linearity, range, and one variable at a time. Therefore, it uses a specific robustness (48,49). Besides, it is essential to use a band intensity of the spectra (28). The multivariate referential method to validate the Raman analysis is a set of techniques that allows statistical spectroscopy method, usually the high- analysis of data collected with more than one performance liquid chromatography method variable (29–32) such as the wavelengths and (HPLC). The reference values are necessary to spectral interactions (33). determine the method accuracy (36,44). Due to the complexity and enormous amount In Table 2, the statistical parameters for the of information acquired by Raman spectroscopy, it selection and prediction of the calibration model is uncertain whether univariate treatments are are presented. These are in accordance to the main sufficient to select as the calibration model. performance characteristics established by the (48) Consequently, it is necessary to adopt multivariate and ICH (49). Bias refers to accuracy, Q2, Qr and analyses to understand and determine the relevance R2 assessing robustness, RMSEC, RMSECV and of the data originating from multiple variables (34– RMSEP refers to the quantitative performance of 37). the calibration models. The method most commonly used in the Bias represents the systematic error and refers multivariate analysis is the partial least squares to accuracy, and this avoids the acceptance or

26 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Table 1. Main variants of Raman spectroscopy for quantification

Variants Brief description Equipment Configuration Advantages Application Ref

Inelastic scattering effects responsible Non-destructive; Little or no Conventional for generating different frequencies from API quantification and Conventional equipment. sample preparation; (10) Raman scattering the incident monochromatic light beam characterization of tablets. Short analysis time. used to irradiate the sample.

Measure all wavelengths The spectrum is obtained from the Fourier transform- Multiplexing spectrometer simultaneously, improved Fourier transformed signal of the API quantification in powder Raman (FT- system, such as a Michelson sensitivity compared with the single (3,11) interfering light in a Michelson-type mixtures. Raman) interferometer. channel spectrometers and less optical interferometer. fluorescence interference.

Complex mixture containing ‘Unidirectional’ mirror permitting the Greater efficiency of laser photons Transmission different constituents at varying transfer of photons from one side and Photon diode or in the sample; improved signal-to- Raman concentrations and quantitative (12,16) acting as a reflector for photons "unidirectional mirror". noise ratio; improves the level of spectroscopy (TRS) analysis of pharmaceutical bilayer influencing it from the other side. quantification accuracy. tablets.

Can investigate samples with low Resonance Raman The incident photon energy approaches concentrations of single spectroscopy Raman microscope. Samples with low concentrations. (18,19) the electron transition energy. constituents or many different (RRS) constituents. Raman scattering by molecules Sensitive and reliable method Surface enhanced adsorbed on rough metal to determine the concentration of Amplification of electromagnetic fields Quantifies highly complex samples Raman surfaces or by nanostructures target molecules in unknown (20–22) generated by the excitation of localized and target molecules in unknown spectroscopy such as nanotubes typically systems; Detect trace organic and surface plasmons. systems. (SERS) made of gold (Au) or silver inorganic analytes in different (Ag). media in nanogram level.

27

J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 1 Continued

Variants Brief description Equipment Configuration Advantages Application Ref

Spatially Dedicated Raman microscopes Isolation of chemically rich spectral data offset Diffuse Raman scattering from a region away or conventional equipment from sub surfaces, substructures, layers Samples inside packs and samples Raman from the laser excitation. Uses the energy of with a coupled microscope, and through other types of barriers. (9,23) for medical diagnostics. spectroscopy vibrational motion from within a molecule. which use optical fibers. Quantifies chemical markers. Readings (SORS) Including handheld equipment. can penetrate container.

Coherent Inverted microscope with a Label-free, chemically specific signal, anti-Stokes Non-linear optical imaging techniques, by laser-scanning confocal scan- fast data-acquisition time and inherent Characterizes raw materials, Raman means of non-linear probing of molecular head and photomultiplier tube (13,15) non-destructive “confocal”- like tablets and powder mixtures. spectroscopy vibrational resonances. (PMT) and GaAsP hybrid imaging. (CARS) (HyD) photodetectors.

Energy difference between the pump and Visible stimulated Raman Stimulated Stokes photons matches the vibrational scattering microscope with Increased sensitivity, because it is free Characterizes multiple excipients Raman energy of the target molecules; the coherently dual-output laser system, a of limitations from labeling and distributions in tablets. Mapping (17) Scattering induced vibrational transition absorbs one Stokes beam, and a pump applicable to spot reduction effect. of samples as cells and tissues. (SRS) photon in the pump beam and gains one beam. photon in the Stokes beam. Tip- Imaging involves measuring the signal at each Spatial resolution required for nanoscale enhanced Raman spectroscopy coupled pixel location of the corresponding scanning characterization. Characterization of materials at (24,25) Raman confocal microscope, and a probe microscope image during the raster Ability to spectroscopically map nanoscale. scattering scanning probe microscope. scan of the surface. surfaces. (TERS) Raman Acquire more representative spectra of Micro The sample is illuminated using laser light the sample mixtures; The wide area Quantification of polymorphic Raman spectrophotometer spectroscopy and an objective lens with a disc point size of illumination and/or reduced particle size mixtures, API content in tablets (14) coupled microscopes. (micro- 1μm in diameter or less. allows for more accurate measurements and powder mixtures. Raman) of polymorphs. API, Active Pharmaceutical Ingredient.

28 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020 rejection of an inaccurate or capable method, For example, in the color tablet coating analysis, respectively (44,45). Q2 represents the predictive the results depend on the model applied, but some capacity of the multivariate models, demonstrating colors lead to larger or smaller calibration errors the robustness evaluation; negative values indicate (45). RMSECV is appropriate to test if the that the model is uncertain (10,46). Qr represents calibration model is well fitted according to the variation usually from instrumental fault, and R2 is current data (35,36). RMSEP is the predictions of one of the most applied statistical parameters to the calibration model; to satisfactory values, it is determine sources of variation of the models essential to consider matrix variations when (36,39). comparing the predictive models (35,36,47). In addition, RMSEC, RMSECV, and RMSEP Therefore, these statistics e/or chemometric refer to the prediction results of the validation parameters (Table 2) allowing to make the method, the values of these parameters are assertive decision in Raman method validation. dependent on the characteristics of the samples.

1660,40 1612,00 1521,21

Figure 2. Raman spectra obtained by BRAVO Handheld Raman Spectrometer (Bruker Corporation, Billerica, MA, EUA). The analysis shows the ability of Raman spectroscopy to characterize specific bands of substances in the multi- component formula. The peaks arising at wavelengths 1660, 1392, 867, and 495 represents the antiretroviral zidovudine (blue spectrum). The red line (red spectrum) refers to lamivudine, commonly associated antiretroviral with zidovudine, for which the bands with intensity at wavelengths 1243, 798, and 780 are the most characteristic. The purple spectrum represents the mixture of excipients (microcrystalline cellulose, croscarmellose sodium, colloidal silicon dioxide, magnesium stearate, Opadry™ white YS-1-7003). Green spectrum represents the powder mixture of the formula components (lamivudine + zidovudine 150mg + 300mg); even in the complex mixture, it is possible to distinguish the main peaks of each drug and differentiate them from the excipients bands.

29 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Table 2. Main statistical parameters for prediction multivariate analysis

Statistical Reference Description Formula Meaning Range References parameters value

Represents the systematic error, is the average value of ŷ is the prediction value, y is the Nearly (44,45) Bias the difference between reference value and i is the test 0 - 1 zero. predicted and measured set sample. values. PRESS: prediction error sum of Predictive Relevance: squares, the sum of squared Represents the fraction of differences between predicted Q² the total variation of the 0 - 1 0 - 1 (10,46) and observed Y -data. response that can be SS: residual sum of squares of predicted by the model. the previous components. Sum of squared reconstruction error: xi is the sample spectrum, ti is determines whether a sample the latent variable scores for the Q 0 - 1 0 - 1 (36,39) spectrum is different due to sample spectrum, and P is the an un-modeled source of model loading. variance.

Coefficient of determination: is the proportion of the SSRegression: sum of squares R² variance in the dependent regression. 0 - 1 1 (36,39) variable predictable from the SSTotal: total sum of squares. independent variable(s).

Root mean square error of Calibration: Is the yi and ŷi as the known and Are determined by Depends corresponding measure for calculated mass of coating the measurement on the RMSEC the calibration model fit. (45) suspension in sample i and n as error in the reference Represents the proximity of the number of samples. reference values. value. the data of calibration model and samples data.

30 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 2. Continuance

Statistical Reference Description Formula Meaning Range References parameters value

Are Root mean square error of cross- ŷ and y represent determined validation: Prediction errors are vectors of predicted and by the Depends on the RMSECV calculated for the samples left reference values, measurement (35,36) reference value. out as the difference between respectively, and n the error in the prediction and reference value. number of samples. reference values.

Are Root mean square error of ŷ and y represent determined prediction: deployed to quantify vectors of predicted and by the Depends on the RMSEP the uncertainty in all future reference values, measurement (35,47) reference value. predictions of the calibration respectively, and n the error in the model. number of samples. reference values.

Spectroscopy Raman in pharmaceutical development In pharmaceutical development, a quality by design (QbD) approach is describe its current applications in the synthesis of drug candidates. In the essential to identify the particularities of drug substances and excipients for early developmental studies, this technique helps characterize the active new formulations (50–53). This concept allows risk management in the pharmaceutical ingredient (API) and understand drug-excipient interactions development of new medicines (54–58). in the formulation, besides allowing for evaluating the critical quality Raman spectroscopy allows for identifying critical variables and their attributes of pharmaceutical products (CQAs) (62,63). clinical relevance related to quality and safety for patients (26,50,59,60). These attributes play a fundamental function in the quality, safety, and Table 3 presents 14 studies in the last five years, related to the use of this efficacy of pharmaceuticals, directly influencing the bioavailability of drugs technique in the product development phase. In these studies, 23 drug (63–65). In this sense, 57% of the studies presented in Table 3 evaluated the substances from 10 therapeutic classes were evaluated. crystalline and polymorphic forms of the compounds in mixtures of powders Protasova and colleagues (61) demonstrated the applicability of Raman and commercial products. spectroscopy for monitoring reactions in solid phases (Table 3). They

31 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Three different devices were compared: Furthermore, Walker and colleagues (70) used benchtop confocal microscope coupled the FT-Raman variant and Low-Frequency Raman spectrophotometer (micro-Raman1), portable Spectroscopy to identify disorders during the spectrophotometer (macro-Raman1) and benchtop milling of the L-tryptophan and indomethacin spectrophotometer (macro-Raman2) for the drugs individually and in the binary mixture (1: 1 quantification of three polymorphs of mebendazole molar ratio). Analysis revealed that the blend in mixtures (Table 3). Partial least squares changed more rapidly and more thoroughly regression models (PLS) were developed obtaining compared to the separate components. Such RMSEP values of 1.68%, 1.24% and 2.03% (w/w) phenomena indicated possible favorable for polymorphs A, B, and C, respectively, using the intermolecular interactions between the two macro Raman analysis. This macro presented substances. These evaluations are critical for better performance, due to the configurations of forecasting and planning large-scale production this equipment, which provides more widely processes. illuminated area laser, allowing getting more Interactions among the components and reproductive and representative spectra (14). degradation products can be detect by Raman Similarly, for the quantification of crystalline under adverse conditions such as temperature and amorphous forms of warfarin sodium (Table 3) (71,72). SERS was employed (Table 1) for the using PCL and PCR, the obtained values were: R2 quantification of ofloxacin and for monitoring its = 0.993, RMSEC = 2.60 and Bias = 0.007; and R2 stability after several forced degradation processes. = 0.993, RMSEC = 2.61 and Bias = 0.019 for PLS The studies revealed method capacity for and PCR, respectively (66). The predictive results quantifying this drug in the presence of its for the quantification of the fraction of crystalline degradation product, simultaneously (73). (theoretical 10%) and fraction amorphous In addition, in the development stage, nine (theoretical 90%) were 15.32 ± 2.63% and 84.68 ± different applications supported Raman 2.63%, respectively, for PLS and 15.32 ± 2.61% analysis (Table 3). This software considers the and 84.68 ± 2.61, respectively, for PCR. These main variables and statistical parameters analyses are critical; especially in the case of applicable to multivariate analyses. Thus, they formulation development with low therapeutic have a fundamental role in modeling, prediction, index drugs, in which any change in these forms and optimization of the chemometric models. over the life cycle of a product, can affect the Therefore, they allow minimizing the exhaustive performance of the drug adding risk to the patient work for application and understanding of these (66). calculations and formulas, in the development and The understanding of the crystalline forms of routine analysis. the drug substance and the interactions with the In pharmaceutical development, a quality by excipients are of fundamental importance for design (QbD) approach is essential to identify the formulations in different presentations, mainly particularities of drug substances and excipients in those exposed to humidity (53,64,65). Three of the complex mixtures and in multi-component 13 studies presented in Table 3 monitored the formulas and their clinical relevance. Raman transition of the crystalline form (cocrystallization) spectroscopy, combined with multivariate models allowing their in-line quantification. These allows identifying the main critical variables, such demonstrated that the Raman technique is a as interactions between the drug substances and the reliable tool to solve challenges for the other components and changes in the crystalline quantification of highly hygroscopic drugs form of the drug. Thus, Raman spectroscopy plays (53,67,68). a keys role in risk management in the development Table 3 shows the particle size monitoring in 2 of new medicines. of 13 studies. The mean particle size of micronized components in the development of ebastine tablets Raman spectroscopy in process analytical (6.25 wt. %) was calculated using polystyrene technology microspheres as standard size (4.9±0.4, 9.8±0.5 The CGMPs for the 21st Century (59), harmonized and 15.8±0.6μm). The particle size and the forms with the concept of QbD addressed by ICH (50), of the components of the formulation remained allows for understanding, continuous improvement unchanged throughout the tablet development and allows for reducing the variabilities in the process (28). Similarly, a Raman probe was used critical stages of the process, ensuring higher to evaluate changes in the size and shape of drug quality to the final product (74). substance particles, generated from the friction in the powder mixing process (69).

32 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Table 3. Raman spectroscopy in pharmaceutical development

Drug Therapeutic Critical Quality Calibration Excipients Variants Equipment Manufacturer Data Analysis Software References Substance Classification Attributes Model T64000 (micro- Raman1); Binary and Horiba – TacTicID-GP The Unscrambler X 10.3 ternary Polymorph Jobin Yvon; Mebendazo Raman (macro PLS, MSC, (CAMO) and Anthelmintic mixtures quantification in B&W (14) le microscopy Raman1); SNV, WLS MATLAB R2010a (Math containing mixtures. Tek.; RamanStation Works). polymorphs. Perkin-Elmer 400F (macro Raman2) Surface- Confocal Regression Forced Horiba Jobin enhanced Raman analysis and Commercial degradation Yvon, Ofloxacin Antibiotic Raman spectrometer validation Not mentioned (73) dosage forms. procedures in eye Bensheim, spectroscopy LabRam using ICH drop and tablets. Germany. (SERS) HR800 guidelines Anhydrous lactose, pregelatinized Quantification of starch, Unscrambler Crystalline/Amor RamanRXN2TM Kaiser Optical magnesium X software (version10.1; Warfarin Anticoagulant phous API in Raman Multi-Channel System Inc., stearate, PLS, PCR Camo Software Inc., (66) sodium s mixtures (placebo spectroscopy Raman Ann Arbor, hydroxypropyl Woodbridge, and sample Analyzer Michigan. cellulose, New Jersey). matrices). methanol and deuterated water (99.9%).

In-line monitoring PLStoolbox 6.2 Anticonvulsan Solvent (ethyl of (Eigenvector Research Carbamaze B&W Tek, t, Psychotropic acetate) for cocrystallization Raman i-Raman BWS PCA, PLS, Inc., Wenatchee, WA, pine/Nicoti Inc., Newark, (68) and crystallization process and spectroscopy 415-785H MCR-ALS USA); Matlab®2011a namide DE, USA. Neurotropic reaction. quantification in (Mathworks Inc., Natick, ternary mixture. MA, USA).

33 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 3. Continuance

Drug Therapeutic Critical Quality Calibration Data Analysis Excipients Variants Equipment Manufacturer References Substance Classification Attributes Model Software

N- acetylbenzyl Intermolecular WITec confocal WITec WITec Instruments Analysis of amine and interactions in Confocal CRM alpha 300 software Antiepileptic Not mentioned Corp., Knoxville, crystallogra (71) N-benzyl- Polycrystalline Raman Imaging Raman (Project TN. USA. phic data. ethanethioa samples. microscope FOUR 4.1). mide

Lactose, carmellose calcium, crystalline Univariate cellulose, ISys 4.0 CI Raman analysis and hydroxypropyl Raman ARAMIS, Horiba, software Ebastine Antihistamine Particle sizing. microscope: Raman (28) cellulose, microscopy Kyoto, Japan. (Malvern LabRAM chemical magnesium stearate Instruments). images. and light anhydrous silicic acid.

Low-frequency Monitoring of (LF) Ethanol, cocrystals: the THz-Raman® Raman Ondax, Inc., CA, HoloGRAMS Furosemide microcrystalline crystalline-form Probe System, Antihypertensi spectroscopy USA and Kaiser Calibration 4.1 (Kaiser /Nicotinami cellulose and conversion in TRPROBE and (53) ve and Optical Systems, using sulfur. Optical de hydroxypropyl suspension or Raman RXN1 Conventional Inc., MI, USA. Systems, Inc.). cellulose. fluidized bed System Raman granulation. spectroscopy Unscrambler FT- L- X 10.3 Identifying Raman/Low- Multi-RAM FT- Bruker Optics, tryptophan Anti- (CAMO Not mentioned disorder during Frequency Raman Ettlingen, PCA, SNV (70) Indomethaci inflammatory Software AS, Milling. Raman spectrometer Germany. n Oslo, Spectroscopy Norway).

34 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 3. Continuance

Drug Therapeutic Critical Quality Calibration Data Analysis Excipients Variants Equipment Manufacturer References Substance Classification Attributes Model Software

MatLab (Mathworks, Kollidon® 25 Natick, MA, (PVP), Quantification of Kaiser Optical Transmission Kaiser USA) and PLS Anti- magnesium the solid-state Systems (Ann Celecoxib Raman RXN1 PLS Toolbox 8.1.1 (72) inflammatory stearate and properties of drugs Arbor, MI, spectroscopy Microprobe (Eigenvector sodium in tablets. USA) Research Inc. chloride. Manson, WA, USA).

Normalized relative peak H O/ethanol areas and 2 Polymorphism and Lamivudin solution for Fourier Transform FT-Raman Thermo Nicolet heating Anti-viral polymorphic Not mentioned (64) e recrystallizati Raman spectrometer 960, USA. times with transformation. on. single exponential functions. PLS tool box Croscarmello 6.2 se, Lactose Monitoring (Eigenvector monohydrate, crystalline phase i-Raman Research Inc., B & W Tek, lipid-lowering Magnesium transition of API Raman BWS415-785H. PCA, PLS, Wenatchee, Ezetimibe Inc., Newark, (67) compounds stearate, and quantification in spectroscopy (Portable MCR-ALS WA, USA) and DE, USA. Cellulose, Powder of the equipment) Matlab®2011a and mixture and tablets. (Mathworks Povidone. Inc., Natick, MA, USA).

35 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 3. Continuance

Drug Therapeutic Critical Quality Calibration Data Analysis Excipients Variants Instrument Manufacturer References Substance Classification Attributes Model Software Choice of Sulfonyl functional Monitoring Reaction chloride resin Raman groups and the Not mentioned reactions on solid Not mentioned Not mentioned Not mentioned (61) molecular and Merrifield spectroscopy implementation phases. resin. of a suitable model system. Comparison of 15 Lactose the measured (Minitab, anhydrous, particle spectra State College, Kaiser Optical microcrystalli Particle size with reference PA, USA), and BMS- Raman Kaiser Raman Systems Inc., Not mentioned ne cellulose distributions in library spectra SigmaPlot (69) 823778-03 microscopy system Ann and powder mixtures. and version 12.5 Arbor, USA magnesium classification of (Systat software stearate. individual Inc., Chicago, components. USA). API: Active Pharmaceutical Ingredient. MCR: Multivariate Curve Resolution. MCR-ALS Multivariate Curve Resolution with Alternating Least Squares. MSC: Multiplicative Scatter Correction. PCA: Principal Component Analysis. PLS: Partial Least Squares. SNV: Standard Normal Variate. WLS: Weighted Least Squares.

Thus, in 2004 the FDA published the Guide to Process Analytical processes. Therefore, it has essential advantages as PAT tools, meeting Technology (PAT), defined as a risk-based scientific approach, aimed at regulatory requirements (33,46,53,77). supporting innovation and efficiency in pharmaceutical development, Riolo and colleagues (78) highlighted the importance of Raman manufacturing, and quality (75). spectroscopy monitoring in the early part of a production process. The Real-time PAT analysis can be in-line/online, and it is carried out acquisition of initial data is fundamental for statistical treatment and directly on the production line, in critical stages as in the process of mixing, identification of variabilities, aiming to determine the outcome of powder granulation, and drying. At-line, rapid analyses are carried out, after the mixtures and ensure the manufacture of safe and effective medicines. critical steps of the process, for example, spectroscopy and disintegration Additionally, other two studies of 13 (Table 3) presented real time assay (50,76). monitoring of the powder mixing process. The drug substance quantification occurred during the process, which allowed for evaluating the mixing Raman spectroscopy allows for identifying changes in physicochemical profiles, as well as identifying the outcome of these, ensuring their properties during the relevant unitary operations from the pharmaceutical uniformity (78–80). point of view. Such an approach enables real-time understanding of these

36 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Two of three of these studies used the same The granulation process was monitored in 5 of model of the phantom (PhAT: Pharmaceutical 13 studies (Table 3) using the partial least squares Area Testing) with the same condition of excitation (PLS) calibration model, which was applied in and diameter of point for quantification of the 100% of the cases, although two studies also used drugs, although with different types of equipment other methods such as MCR, PCA, and MLR. (79,80). These examples show that some With reference to these models, the MCR accessories and devices can be applicable for model was the only predictive model not to exceed different equipment configurations, including the established acceptance limits of 10% (relative different manufacturers, as in a study performed by Bias) for the monitored mixtures containing 17.5, Netchacovitch and colleagues (33). In this study, a 22.5, 25.0, 27.5 and 32.5% (w/w) of metoprolol. Kaiser Optical Systems probe coupled to a Perkin The calibration model for the quantification of Elmer device was used. Table 3 revealed that 85% drug substances during the hot melt extrusion of the studies (11 of 13) used the equipment of the process presented results of accuracy, precision (95 manufacturer Kaiser Optical Systems. of the 100 measurements below the predefined A single-channel device model, used for acceptance limits of 10%) and robustness (46). measurements through the glass wall of the mixer, The calibration models should consider the allowed acetyl salicylic acid detection (1.1% w/w) concentration of the drugs in the powder mixture, in the evaluation of an aspirin mixture with Avicel the time of exposure of the sample to radiation, and PH-101, at a 50-rpm profile. The monitored the volume of the sample. This volume shall not variable was the drug substance particle size. The overstep three times the unit dose range results did not identify profile variation in the (33,36,46,84). mixture during the process. However, an increase Following the development of the appropriate in the time to obtain a homogeneous mixture was PLS calibration model, Harting and Kleinebudde observed, through the evaluation of peak-to-peak (84) demonstrated RMSEP of 0.59% and 1.5%, noise of the spectra. This behavior was dependent respectively, for the quantification of ibuprofen on the drug substance concentration as they added and diclofenac sodium. A difference of 6% in the the drug (79). concentration of diclofenac was observed due to Unlike the previous study, a multichannel not cleaning the granulator between the tests. This device was used to quantify drug substances in demonstrates the ability of the technique to identify samples of powders and tablets employing variability and possible challenges in the simultaneous monitoring (PhAT probes) of the processes, such as equipment setup failure. blend. The performance characteristics of the Like the previous study, a univariate and a method revealed, coefficients of determination multivariate model (including all wavelengths and 2 2 (R c = 0.9695, R cv = 0.9605), errors (RMSEC [% their interactions) were developed for w/w] = 0.8295, RMSECV [% w/w] = 0.9446, quantification of itraconazole in line, for RMSEP [% w/w]) = 0.8246 and Biascv [% w/w] = monitoring the hot-melt extrusion process. The 0.0091, respectively. The real-time monitoring of validation for the two calibration models occurred; the powder mixture allowed for indicating possible however, the multivariate approach was more problems such as an insufficient amount of powder accurate due to including a more significant in the mixer and changes in the concentration of the number of variables in the model (33). Such an API (80). approach may justify the use of multivariate In addition, the Raman technique can be models in 100% of the studies presented in Table suitable in online monitoring of the gel mixing 3. process. Predictive results for method performance Five of 13 studies (Table 3) were developed for (R2 = 0.973; Q2 = 0.973; RMSECV = 0.0418% monitoring of the coating nuclei stages. Recently, w/w) were obtained even in a low concentration. Korasa and Vrečer (85) monitored the pellet On three different days, RMSEP values of 0.0255, coating process using an in-line probe. The results 0.0235 and 0.0381% (w/w) for three validation sets were R2 = 0.9970, RMSEP = 0.5998 and slope of were obtained (81). These results evidence the the regression line of 0.9463, which showed high capability of the technique to evaluate the mixtures consistency between observed and predicted of different pharmaceutical forms (solid, semi- values. The technique not only contains a tool to solid, and liquid). In addition, the Raman technique determine the amount of spray coating, but it is also allows for quantifying samples with high water able to assess the thickness of the coating (85). content, demonstrating its viability in the Similarly, an MCR calibration model for the monitoring granulation unit operation (67,82,83). quantification of multilayer film coating was developed. The model allowed to predict the

37 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020 thickness of films during the process and showed promoted an increase in dissolution rate and may better performance compared to the PLS increase bioavailability. In addition, it presents regression model, more frequently applied, as greater bioadhesiveness to the biological revealed in Table 3 (86). membranes compared to the drug in the Kim and Woo (87) evaluated coat weight gain micrometric scale. Such advantages potentially in-line by quantitative analysis of the film layer. improve the safety and efficacy of the drugs, They accurately monitored the endpoints of the allowing dose reduction (89–92). process with an aim weight gain of 3% (w/w). In recent decades, there has been a significant Image analysis was also performed on the coating rise in the development of drug nanocrystals. As of layer, using a Raman microscope showing an 2017, more than 80 drug nanocrystal applications increase of thickness. were submitted to the FDA (90). The critical Fluorescence from pigments or certain attribute characteristics of drug nanocrystals such excipients may interfere with spectral analyses as size, particle-size distribution and mean particle (87), although, two studies reported the sensitivity diameter, along with morphological analyses are of the technique for monitoring the processes of fundamental to evaluate crystalline forms and to colored coatings (Table 3). In the first study, in- optimize size-dependent properties (90,93,94). line monitoring using multivariate models (PLS Due to the versatility of Raman technology and and SBC) and a univariate analysis model were the interface advantage of different analytical tools developed for endpoint determination of the with this technique, they can lead to an expansion coating process of colorless suspension. The into the field of pharmaceutical nanotechnology method evidenced to have high predictive power (1). Surface Raman Enhanced Scattering (SERS) (R2 = 0.9996 Q2 RMSEC/% = 0.85 and RMSEP/% and Tip-enhanced Raman spectroscopy (TERS) = 0.96) and all spectral variation were linear over (Table 1) enhance the sensitivity of the technique time (88). and enable the evaluation of nanoscale substances In the second study, the authors monitored six (21,24,25,95,96). colored coatings, which after 30 min of the Low-frequency Raman spectroscopy was process, there was no change, which revealed its applied to evaluate the crystalline state of endpoint. Such an approach allowed for reducing nanocrystals; the technique was able to assess the this step by 20 minutes (from 50 to 30 minutes) polymorphic form of furosemide after nano (45). pulverization (89). In addition, this technique The ability of the continuous verification allowed chemical and spatial quantification in the presented in these studies allows a greater process of mixing cellulose nanocrystals in understanding of the variabilities of the primary thermoplastics (97). pharmaceutical unit operations. The potential to A method for quantitative analysis of alginate overcome the challenges of quantifying, in-line, nanocarriers loaded with curcumin in hydrogels samples in the presence of classical interferents presented standard deviation results of SD equal to from the pharmaceutical industry, such as 0.34, 0.50 and 0.96% (w/w) and relative error of fluorescent and highly hygroscopic components, is 3.21, 0.42 and 0.85% for concentrations of 2.11, presented. Also, the ability to determine the 5.43 and 10.48% (w/w), respectively, using the endpoints of blending processes increases patient PLS predictive model. Due to low interference safety, reduces process time and open the way for from samples with high water content, the continuous production and reduction of lead time precision of the method demonstrated the release, one of the significant challenges of the effectiveness of the technique for the pharmaceutical industries. quantification at low concentrations of nanocrystals (98). Raman spectroscopy in pharmaceutical Besides, Raman spectroscopy Raman nanomaterials spectroscopy was employed to characterize Improvements in the technology used to obtain graphite oxide nanoparticles to aid the nanocrystals have presented opportunities never administration of photothermally-controlled drugs achieved before to facilitate the administration of (99). It allowed evaluating the penetration capacity medicines. This approach efficiently increases the of caffeine and propylene glycol nanocrystals in surface area of the drug with the biological media gel form applied topically to swine (100). and increases the saturation solubility. This Furthermore, it enabled the broader verification of cellular interactions in cytotoxicity assays (101).

38 J Pharm Pharm Sci (www.cspsCanada.org) 23, 24 - 46, 2020

Table 4. Raman spectroscopy in pharmaceutical process measurements

Sample Drug Unit Operating Accessories and Calibration Application/Critical Instrument Manufacturer References characteristic Substance Operations Principles apparatus Model Quality Attributes Aspirin, PhAT probe, Kaiser Optical aspartame, excitation laser Powder Convection Raman RXN1 Systems Inc., In situ monitoring of Avicel PH-101 Blending 785 nm and a spot PLS (79) blending Mixing TM Ann Arbor, MI, powder blending and size diameter of 6 USA. sodium nitrate. mm. Jobin Yvon Nd:YAG blue Horiba Raman signal Diffusion laser at 473.1 nm, Powder Blending LabRam Horiba, Kyoto, of the Confidential Blending and attached to an Mixing monitoring (78) blending and Mixing confocal Japan. component of (Tumble) Olympus BX40 Raman interest microscope. spectrometer PhAT probe, excitation laser Blending, Wet Powder Kaiser Optical 785 nm spot size API content in Anhydrous granulation granulation RamanRxn2TM blending and Systems, Ann diameter of 6 mm PLS blended powder and (80) caffeine and and melt Hybrid tablet Arbor, USA. and a Kaiser tablets in real-time tableting extrusion transmission accessory. A CCD detector Gel (2% w/w) Kaiser Optical In-line quantitative and a fiber-optic and confidential Convection RamanRxn2TM Systems, Ann determination (API) Mixing PhAT probe. PLS (81) Suspension information Mixing Spectrometer Arbor, MI, in a gel and Excitation laser (0.09% w/w) USA. suspension 785 nm. PhAT probe, Raman Ibuprofen 50 Twin-screw Kaiser Optical excitation laser RXN2™ In-line continuous Granules and diclofenac Granulation wet Systems, Ann 785 nm and a spot PLS (84) Hybrid API quantification sodium granulation Arbor, USA. size diameter of 6 Analyzer mm. Raman probe (Kaiser Optical API content in real- Hot-melt RamanStation Perkin Elmer, Systems Inc., MI, time Extrudates Itraconazole Granulation PLS (33) extrusion 400F MA, USA. USA). Two- during a Hot-Melt dimensional CCD Extrusion process detector.

39 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Table 4. Continuance

Sample Unit Operating Accessories Calibration Application/Critical Drug Substance Instrument Manufacturer References characteristic Operations Principles and apparatus Model Quality Attributes Fiber-optic Kaiser Optical Raman Dynisco API content during Extrusion Hot-melt Raman Rxn1™ Systems, Ann Metoprolol tartrate Granulation probe. PLS, MCR pharmaceutical hot- (46) mixtures extrusion spectrometer Arbor, MI, Excitation laser melt extrusion USA. 785 nm. 785 nm Wet High- excitation laser, Shear Kaiser Optical a CCD detector Assessment and Theophylline Granulation/ RamanRxn2 PLS, PCA, Tablet Granulatio Systems, MI, and PhAT prediction of tablet (77) anhydrate Unit Dosing Analyzer MLR n/ USA. probe properties Tableting (backscattering geometry). PhAT probe using a PLS, MCR, Fluid-bed RamanRXN2TM Kaiser Optical Pellets Acetylsalicylic acid Coating diode laser MCR-ALS, Coating thickness (86) coating analyzer Systems, USA). operating at 785 SNV, SVD nm PhAT probe Kaiser Optical RamanRXN1TM with 785 nm Prediction of sprayed Systems, Inc, Film Spectrometer / excitation laser quantity, coating Coated pellets Diclofenac Sodium Coating USA / Tornado PLS (85) Coating HyperFluxTM PRO of 6 mm thickness, and Loss Spectral Plus Raman diameter / 785 on drying. System, USA nm laser. a) 785nm diode laser and WAI Coating excipients: a) Kaiser a) in-line monitoring probe having a detackifier, TiO2, a) Raman Optical Inc., for coating weight spot size of 6 Kolliphor® SLS, Pan spectrometer MI, USA. gain Tablet Coating mm PLS (87) talc and aqueous Coating b) Raman b) Nanophoton b) imaging analysis b) objective film former that is microscope Corporation, for coating layers and lens TU Plan polyvinyl alcohol Osaka, Japan. thickness Fluor 20x/NA0.45. Drug-free cores - Kaiser Optical PhAT probe, Endpoints of coating Coating suspensions Gas RamanRXN2™ PLS, MCR, Tablet Coating Systems, Ann excitation laser processes for colored (45) comprising Suspension Analyzer SBC, UV Arbor, USA. 785 nm spot tablets polymers and

40 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

pigments and/or size diameter of dyes 6 mm.

PhAT probe, Placebo and Kaiser Optical excitation laser Endpoint Gas RamanRXN2™ PLS, SBC, Tablet caffeine Coating Systems, Ann 785 nm spot determination of a (88) Suspension Analyzer UV cores Arbor, USA. size diameter of tablet coating process 6 mm. API: Active Pharmaceutical Ingredient. MCR: Multivariate Curve Resolution. MCR-ALS Multivariate Curve Resolution with Alternating Least Squares. MLR: Multiple Linear Regression. NIPALS: Non-Linear Iterative Partial Least Squares. PCA: Principal Component Analysis. PCS: Principal Component Analysis. PLS: Partial Least Squares. PhAT: Pharmaceutical Area Testing. SBC: Science-Based Calibration. SNV: Standard Normal Variate. SVD: Singular Value Decomposition. WAI: Wide Area Illumination. UV: Univariate analysis.

Collectively, these studies highlight the prospects of routine analysis of formulas of various therapeutic classes. Thus, it has become essential to materials at the nanometer scale, mainly, considering the possibility of identify critical variables and their clinical relevance in the development of auxiliary Raman spectroscopy and even replacing destructive techniques new drugs in the QbD environment. such as transmission electron microscopy, X-ray diffraction, and In process monitoring, this technique has been established for off-line photoluminescence spectroscopy, commonly used for size analysis and measurements, but it is in real-time (in-line) control based on PAT, where it nanocrystal characteristics. has advanced exponentially. In this sense, it allows quantification in powder mixtures under adverse conditions, as in the case of blends with high water FINAL CONSIDERATIONS content, semi-solid and liquid mixtures. It also enables the collection of an enormous amount of data, compared to the conventional technique by high- The application of Raman spectroscopy has evolved due to its increasing performance liquid chromatography. These data allow a greater instrumental versatility. This evolution allowed different configurations of understanding of variability in the pharmaceutical process, as well as the the instrument, significantly improving its sensitivity. In combination with continuous verification of its critical stages and unit operations. chemometric models, this technique makes it possible to investigate a wide Furthermore, it will enable researchers to overcome the challenges of variety of samples, quickly and non-destructively. quantifying, in-line, fluorescent interfering samples such as colored tablets. In the development of pharmaceutical products, this technique has been Therefore, Raman spectroscopy has become a potent tool for consolidated as a reliable quantitative analytical method in the evaluation of pharmaceutical applications, integrated into the concept of Good the main critical quality attributes. Raman spectroscopy has been able to Manufacturing Practices of the 21st Century. Recently, the use of Raman overcome classic challenges of the pharmaceutical industry, such as the spectroscopy has advanced in pharmaceutical nanotechnology, standing out quantification of substances at low concentrations in complex mixtures, for quantifying materials on the nanoscale scale. including the presence of their degradation products and in multicomponent

41 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

ACKNOWLEDGEMENTS 11. Dzsaber S, Negyedi M, Bernáth B, Gyüre B, Fehér T, Kramberger C, Pichler T, Simon F. A Fourier We appreciate the support of Jim Hesson of transform Raman spectrometer with visible laser Academic English Solutions.com, and Ana Paula excitation. J. Raman Spectrosc 2015;46:327-332. Peinado of Bruker do Brasil Ltda., which provided 10.1002/jrs.4641. 12. Griffen JA, Owen AW, Matousek P. Development us the BRAVO Handheld Raman Spectrometer. of Transmission Raman Spectroscopy towards the in line, high throughput and non-destructive REFERENCES quantitative analysis of pharmaceutical solid oral dose. Analyst 2015;140:107-12. 1. Paudel A, Raijada D, Rantanen J. Raman 10.1039/c4an01798f. spectroscopy in pharmaceutical product design. 13. Ojarinta R, Saarinen J, Strachan CJ, Korhonen O, Adv Drug Deliv Rev 2015;89:3–20. Laitinen R. Preparation and characterization of 2. Raman C V, Krishnan KS. The Optical Analogue multi-component tablets containing co-amorphous of the Compton Effect. Nature 1928;121:377–8. salts: Combining multimodal non-linear optical 10.1038/121711a0. imaging with established analytical methods. Eur J 3. Zhu X, Xu T, Lin Q, Duan Y. Technical Pharm Biopharm 2018;132:112-126. development of raman spectroscopy: From 10.1016/j.ejpb.2018.09.013. instrumental to advanced combined technologies. 14. Paiva EM, da Silva VH, Poppi RJ, Pereira CF, Appl Spectrosc Rev 2014;49:64–82. Rohwedder JJR. Comparison of macro and micro 10.1080/05704928.2013.798801. Raman measurement for reliable quantitative 4. Depciuch J, Kaznowska E, Zawlik I, Wojnarowska analysis of pharmaceutical polymorphs. J Pharm R, Cholewa M, Heraud P, Cebulski J. Application Biomed Anal 2018;157:107-115. of Raman Spectroscopy and Infrared Spectroscopy 10.1016/j.jpba.2018.05.010. in the Identification of Breast Cancer. Applied 15. Wu T, Chen K, Zhao H, Zhang W, Li Y, Wei H. Spectrosc 2016;70:251–63. Flexible dual-soliton manipulation for coherent 10.1177/0003702815620127. anti-Stokes Raman scattering spectroscopy. Opt 5. Bumbrah GS, Sharma RM. Raman spectroscopy – Express 2018;26:22001-22010. Basic principle, instrumentation and selected 10.1364/OE.26.022001. applications for the characterization of drugs of 16. Zhang Y, McGeorge G. Quantitative Analysis of abuse. Egypt J Forensic Sci 2016;6:209-215. Pharmaceutical Bilayer Tablets Using 10.1016/j.ejfs.2015.06.001. Transmission Raman Spectroscopy. J Pharm Innov 6. Netchacovitch L, Thiry J, De Bleye C, Chavez PF, 2015;10:269–280. 10.1007/s12247-015-9223-8. Krier F, Sacré PY, Evrard B, Hubert P, Ziemons E. 17. Bi Y, Yang C, Chen Y, Yan S, Yang G, Wu Y, Vibrational spectroscopy and microspectroscopy Zhang G, Wang P. Near-resonance enhanced label- analyzing qualitatively and quantitatively free stimulated Raman scattering microscopy with pharmaceutical hot melt extrudates. J Pharm spatial resolution near 130 nm. Light Sci Appl Biomed Anal 2015;113:21–33. 2018;7. 10.1038/s41377-018-0082-1. 10.1016/j.jpba.2015.01.051. 18. Villegas Borrero NF, Clemente da Silva Filho JM, 7. Wang H, Boraey MA, Williams L, Lechuga- Ermakov VA, Marques FC. Silver nanoparticles Ballesteros D, Vehring R. Low-frequency shift produced by laser ablation for a study on the effect dispersive Raman spectroscopy for the analysis of of SERS with low laser power on N719 dye and respirable dosage forms. Int J Pharm 2014;469:197- Rhodamine-B. MRS Advances 2019;4:723-731. 205. 10.1016/j.ijpharm.2014.04.058. 10.1557/adv.2019.157. 8. Lakhwani GR, Sherikar OD, Mehta PJ. 19. Höhl M, Roth B, Morgner U, Meinhardt- Nondestructive and rapid concurrent estimation of Wollweber M. Efficient procedure for the paracetamol and nimesulide in their combined measurement of preresonant excitation profiles in dosage form using Raman spectroscopic technique. UV Raman spectroscopy. Rev Sci Instrum 2017;88. Indian J Pharm Sci 2013;75:211–6. 10.1002/9781118971147.ch20. 9. Ember KJI, Hoeve MA, McAughtrie SL, Bergholt 20. Chruszcz-Lipska K, Jaworska A, Szczurek E, MS, Dwyer BJ, Stevens MM, Faulds K, Forbes SJ, Baranska M. (-)-R-mevalonolactone studied by Campbell CJ. Raman spectroscopy and ROA and SERS spectroscopy. Chirality, regenerative medicine: a review. NPJ Regen Med 2014;26:453-61. 10.1002/chir.22288. 2017;2. 10.1038/s41536-017-0014-3. 21. Saleh TA, Al-Shalalfeh MM, Al-Saadi AA. 10. Casian T, Reznek A, Vonica-Gligor AL, Van Graphene Dendrimer-stabilized silver Renterghem J, De Beer T, Tomuță I. Development, nanoparticles for detection of methimazole using validation and comparison of near infrared and Surface-enhanced Raman scattering with Raman spectroscopic methods for fast computational assignment. Sci Rep 2016;6. characterization of tablets with amlodipine and 10.1038/srep32185. valsartan. Talanta 2017;167:333-343. 22. Zong C, Xu M, Xu LJ, Wei T, Ma X, Zheng XS, 10.1016/j.talanta.2017.01.092. Hu R, Ren B. Surface-Enhanced Raman

42 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Spectroscopy for Bioanalysis: Reliability and 132. 10.5487/TR.2018.34.2.127. Challenges. Chem Rev 2018;118:4946-4980. 35. Faber NM, Song XH, Hopke PK. Sample-specific 10.1021/acs.chemrev.7b00668. standard error of prediction for partial least squares 23. Ellis DI, Eccles R, Xu Y, Griffen J, Muhamadali H, regression. Trends Analyt Chem 2003;22:330-334. Matousek P, Goodall I, Goodacre R. Through- 10.1016/S0165-9936(03)00503-X. container, extremely low concentration detection of 36. Hossain MN, Igne B, Anderson CA, Drennen JK. multiple chemical markers of counterfeit alcohol Influence of moisture variation on the performance using a handheld SORS device. Sci Rep 2017;7. of Raman spectroscopy in quantitative 10.1038/s41598-017-12263-0. pharmaceutical analyses. J Pharm Biomed Anal 24. Sonntag MD, Pozzi EA, Jiang N, Hersam MC, 2019;164:528-535. 10.1016/j.jpba.2018.10.022. Duyne RP Van. Recent Advances in Tip-Enhanced 37. Lavine BK. Special Issue: Chemometrics. Appl Raman Spectroscopy. J Phys Chem Lett Spectrosc 2018;72:339. 2014;5:3125–30. 10.1021/jz5015746. 10.1177/0003702818761619. 25. Su W, Kumar N, Krayev A, Chaigneau M. In situ 38. Brouckaert D, Uyttersprot JS, Broeckx W, De Beer topographical chemical and electrical imaging of T. Development and validation of an at-line fast and carboxyl graphene oxide at the nanoscale. Nat non-destructive Raman spectroscopic method for Commun 2018;9. 10.1038/s41467-018-05307-0. the quantification of multiple components in liquid 26. Kornecki M, Strube J. Process Analytical detergent compositions. Anal Chim Acta Technology for Advanced Process Control in 2016;941:26–34. 10.1016/j.aca.2016.08.050. Biologics Manufacturing with the Aid of 39. Wold S, Sjöström M, Eriksson L. 2 2001;58:109- Macroscopic Kinetic Modeling. Bioengineering 130. 10.1016/S0169-7439(01)00155-1. 2018;5. 10.3390/bioengineering5010025. 40. Wold H. Estimation of principal components and 27. Hopke PK. The evolution of chemometrics. Anal related models by iterative least squares. Chim Acta 2003;500:365-377. 10.1016/S0003- Multivariate Analysis 1966:1391–420. 2670(03)00944-9. 41. Farkas A, Vajna B, Sóti PL, Nagy ZK, Pataki H, 28. Kuriyama A, Ozaki Y. Assessment of Active Van Der Gucht F, Marosi G. Comparison of Pharmaceutical Ingredient Particle Size in Tablets multivariate linear regression methods in micro- by Raman Chemical Imaging Validated using Raman spectrometric quantitative characterization. Polystyrene Microsphere Size Standards. AAPS J Raman Spectrosc 2015;46:566–76. PharmSciTech 2014;15:375–87. 10.1208/s12249- 10.1002/jrs.4672. 013-0064-9. 42. Fransson M, Johansson J, Sparén A, Svensson O. 29. Petty RE, Laxer RM, Lindsley CB, Wedderburn Comparison of multivariate methods for LR. Textbook of pediatric rheumatology. 7th ed. quantitative determination with transmission Philadelphia: Saunders, 2016. Raman spectroscopy in pharmaceutical 30. Grimnes S, Martinsen ØG. Bioimpedance and formulations. J Chemom 2010;24:674–80. Bioelectricity Basics. 3rd ed. Cambridge, 10.1002/cem.1330. Massachusetts: Academic Press, 2015. 43. Singh VD, Daharwal SJ. Development and 31. Nosyk B, Montaner JSG, Colley G, Lima VD, Chan validation of multivariate calibration methods for K, Heath K, Yip B, Samji H, Gilbert M, Barrios R, simultaneous estimation of Paracetamol, Enalapril Gustafson R, Hogg RS; STOP HIV/AIDS Study maleate and hydrochlorothiazide in pharmaceutical Group. The cascade of HIV care in British dosage form. Spectrochim Acta A Mol Biomol Columbia, Canada, 1996-2011: A population-based Spectrosc 2017;171:369–75. retrospective cohort study. Lancet Infect Dis 10.1016/j.saa.2016.08.028. 2014;14:40–9. 10.1016/S1473-3099(13)70254-8. 44. Müller J, Knop K, Wirges M, Kleinebudde P. 32. Long FH. Multivariate Analysis for Metabolomics Validation of Raman spectroscopic procedures in and Proteomics Data. In: Haleem J I, Timothy D V, agreement with ICH guideline Q2 with considering editor. Proteomic and Metabolomic Approaches to the transfer to real time monitoring of an active Biomarker Discovery. Cambridge, Massachusetts: coating process. J Pharm Biomed Anal Academic Press, 2013, p. 299–311. 10.1016/B978- 2010;53:884-94. 10.1016/j.jpba.2010.06.016. 0-12-394446-7.00019-4. 45. Barimani S, Kleinebudde P. Monitoring of tablet 33. Netchacovitch L, Thiry J, De Bleye C, Dumont E, coating processes with colored coatings. Talanta Cailletaud J, Sacré PY, Evrard B, Hubert P, 2018;178:686–97. 10.1016/j.ejpb.2017.05.011. Ziemons E. Global approach for the validation of 46. Saerens L, Segher N, Vervaet C, Remon JP, De an in-line Raman spectroscopic method to Beer T. Validation of an in-line Raman determine the API content in real-time during a hot- spectroscopic method for continuous active melt extrusion process. Talanta 2017;171:45–52. pharmaceutical ingredient quantification during 10.1016/j.talanta.2017.04.060. pharmaceutical hot-melt extrusion. Anal Chim 34. Açikgöz G, Hamamci B, Yildiz A. Determination Acta 2014;806:180–7. 10.1016/j.aca.2013.11.020. of ethanol in blood samples using partial least 47. Sparén A, Hartman M, Fransson M, Johansson J, square regression applied to surface enhanced Svensson O. Matrix effects in quantitative raman spectroscopy. Toxicol Res 2018;34:127- assessment of pharmaceutical tablets using

43 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

transmission raman and near-infrared (NIR) as reliable PAT tool. Biotechnol Prog 2018;34:659- Spectroscopy. Appl Spectrosc 2015;69:580–9. 670. 10.1002/btpr.2635. 10.1366/14-07645. 57. Bostijn N, Dhondt W, Vervaet C, De Beer T. PAT- 48. Food and Drug Administration (FDA). Analytical based batch statistical process control of a Procedures and Methods Validation for Drugs and manufacturing process for a pharmaceutical Biologics. Silver Spring, MD: FDA, 2015. ointment. Eur J Pharm Sci 2019;136. Available at: 10.1016/j.aca.2018.02.007. https://www.fda.gov/media/87801/download. 58. Webster TA, Hadley BC, Hilliard W, Jaques C, Accessed April 27, 2019. Mason C. Development of generic raman models 49. International Conference on Harmonisation of for a GS-KOTM CHO platform process. Technical Requirements for Registration of Biotechnol Prog 2018;34:730-737. Pharmaceuticals for Human Use (ICH). Validation 10.1002/btpr.2633. of Analytical Procedures: Text and Methodology 59. Food and Drug Administration (FDA). Q2(R1). 2005. Available at: Pharmaceutical CGMPs for the 21st Century - A https://www.ich.org/fileadmin/Public_Web_Site/I risk-based approach. Silver Spring, MD: FDA; CH_Products/Guidelines/Quality/Q2_R1/Step4/Q 2004. Available at: 2_R1__Guideline.pdf. Accessed April 27, 2019. https://www.fda.gov/media/77391/download. 50. International Conference on Harmonisation of Accessed Juny 10, 2019. Technical Requirements for Registration of 60. Wang X, Esquerre C, Downey G, Henihan L, Pharmaceuticals for Human Use (ICH). O’Callaghan D, O’Donnell C. Assessment of infant Pharmaceutical Development Q8(R2). 2009. formula quality and composition using Vis-NIR, Available at: MIR and Raman process analytical technologies. https://www.ich.org/fileadmin/Public_Web_Site/I Talanta 2018;183:320–8. CH_Products/Guidelines/Quality/Q8_R1/Step4/Q 10.1016/j.talanta.2018.02.080. 8_R2_Guideline.pdf. Accessed May 20, 2019. 61. Protasova I, Heissler S, Jung N, Braese S. 51. Kushner J, Langdon BA, Hiller JI, Carlson GT. Monitoring Reactions on Solid Phases with Raman Examining the impact of excipient material Spectroscopy. Chemistry 2017;23:8703–11. property variation on drug product quality 10.1002/chem.201700907. attributes: A quality-by-design study for a roller 62. Dharani S, Rahman Z, Barakh Ali SF, Afrooz H, compacted, immediate release tablet. J Pharm Sci Khan MA. Quantitative estimation of phenytoin 2011;100:2222-39. 10.1002/jps.22455. sodium disproportionation in the formulations 52. Kushner J, Langdon BA, Hicks I, Song D, Li F, using vibration spectroscopies and multivariate Kathiria L, Kane A, Ranade G, Agarwal K. A methodologies. Int J Pharm 2018;539:65-74. quality-by-design study for an immediate-release 10.1016/j.ijpharm.2018.01.005. tablet platform: Examining the relative impact of 63. Yu LX, Amidon G, Khan MA, Hoag SW, Polli J, active pharmaceutical ingredient properties, Raju GK, Woodcock J. Understanding processing methods, and excipient variability on Pharmaceutical Quality by Design. AAPS J drug product quality attributes. J Pharm Sci 2014;16:771-83. 10.1208/s12248-014-9598-3. 2014;103:527-38. 10.1002/jps.23810. 64. Du Y, Zhang H, Xue J, Tang W, Fang H, Zhang Q, 53. Otaki T, Tanabe Y, Kojima T, Miura M, Ikeda Y, Li Y, Hong Z. Vibrational spectroscopic study of Koide T, Fukami T. In situ monitoring of cocrystals polymorphism and polymorphic transformation of in formulation development using low-frequency the anti-viral drug lamivudine. Spectrochim Acta A Raman spectroscopy. Int J Pharm 2018;542:56-65. Mol Biomol Spectrosc 2015;137:1158-63. 10.1016/j.ijpharm.2018.03.008. 10.1016/j.saa.2014.08.128. 54. Aksu B, De Beer T, Folestad S, Ketolainen J, 65. Izutsu K, Koide T, Takata N, Ikeda Y, Ono M, Lindén H, Lopes JA, de Matas M, Oostra W, Inoue M, Fukami T, Yonemochi E. Rantanen J, Weimer M. Strategic funding priorities Characterization and Quality Control of in the pharmaceutical sciences allied to Quality by Pharmaceutical Cocrystals. Chem Pharm Bull Design (QbD) and Process Analytical Technology 2016;64:1421-1430. 10.1248/cpb.c16-00233. (PAT). Eur J Pharm Sci 2012;47:402-5. 66. Rahman Z, Mohammad A, Akhtar S, Siddiqui A, 10.1016/j.ejps.2012.06.009. Korang-Yeboah M, Khan MA. Chemometric 55. Li MY, Ebel B, Paris C, Chauchard F, Guedon E, Model Development and Comparison of Raman Marc A. Real-time monitoring of antibody and 13C Solid-State Nuclear Magnetic Resonance- glycosylation site occupancy by in situ Raman Chemometric Methods for Quantification of spectroscopy during bioreactor CHO cell cultures. Crystalline/Amorphous Warfarin Sodium Fraction Biotechnol Prog 2018;34:486-493. in the Formulations. J Pharm Sci 2015;104:2550-8. 10.1002/btpr.2604. 10.1002/jps.24524. 56. Santos RM, Kessler JM, Salou P, Menezes JC, 67. Farias MA dos S, Soares FLF, Carneiro RL. Peinado A. Monitoring mAb cultivations with in- Crystalline phase transition of ezetimibe in final situ raman spectroscopy: The influence of spectral product, after packing, promoted by the humidity of selectivity on calibration models and industrial use excipients: Monitoring and quantification by

44 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

Raman spectroscopy. J Pharm Biomed Anal Beer T, Vervaet C, RemonJP. Assessment and 2016;121:209-214. 10.1016/j.jpba.2016.01.008. prediction of tablet properties using transmission 68. Soares FLF, Carneiro RL. In-line monitoring of and backscattering Raman spectroscopy and cocrystallization process and quantification of transmission NIR spectroscopy. Asian Journal of carbamazepine-nicotinamide cocrystal using Pharmaceutical Sciences 2016;11:547–58. Raman spectroscopy and chemometric tools. 10.1016/j.ajps.2016.04.004. Spectrochim Acta A Mol Biomol Spectrosc 78. Riolo D, Piazza A, Cottini C, Serafini M, Lutero E, 2017;180:1–8. 10.1016/j.saa.2017.02.045. Cuoghi E, Gasparini L, Botturi D, Marino IG, 69. Gamble JF, Hoffmann M, Hughes H, Hutchins P, Aliatis I, Bersani D, Lottici PP. Raman Tobyn M. Monitoring process induced attrition of spectroscopy as a PAT for pharmaceutical drug substance particles within formulated blends. blending: Advantages and disadvantages. J Pharm Int J Pharm 2014;470:77–87. Biomed Anal 2018;149:329–34. 10.1016/j.ijpharm.2014.04.028. 10.1016/j.jpba.2017.11.030. 70. Walker G, Römann P, Poller B, Löbmann K, 79. Allan P, Bellamy LJ, Nordon A, Littlejohn D, Grohganz H, Rooney JS, Huff GS, Smith GPS, Andrews J, Dallin P. In situ monitoring of powder Rades T, Gordon KC, Strachan CJ, Fraser-Miller blending by non-invasive Raman spectrometry with SJ. Probing Pharmaceutical Mixtures during wide area illumination. Journal of Pharmaceutical Milling: The Potency of Low-Frequency Raman and Biomedical Analysis 2013;76:28–35. Spectroscopy in Identifying Disorder. Mol Pharm 10.1016/j.jpba.2012.12.003. 2017;14:4675–84. 80. Nagy B, Farkas A, Gyürkés M, Komaromy-Hiller 10.1021/acs.molpharmaceut.7b00803. S, Démuth B, Szabó B, Nusser D, Borbás E, Marosi 71. Śmiszek-Lindert WE, Chełmecka E, Lindert O, G, Nagy ZK. In-line Raman spectroscopic Dudzińska A, Kaczmarczyk-Sedlak I. Towards a monitoring and feedback control of a continuous better comprehension of interactions in the twin-screw pharmaceutical powder blending and crystalline N-acetylbenzylamine and its sulphur tableting process. Int J Pharm 2017;530:21–9. analogue N-benzyl-ethanethioamide. IR, Raman, 10.5650/jos.ess13082. DFT studies and Hirshfeld surfaces analysis. 81. Bostijn N, Hellings M, Van Der Veen M, Vervaet Spectrochim Acta A Mol Biomol Spectrosc C, De Beer T. In-line UV spectroscopy for the 2018;201:328-338. 10.1016/j.saa.2018.05.021. quantification of low-dose active ingredients during 72. Edinger M, Knopp MM, Kerdoncuff H, Rantanen the manufacturing of pharmaceutical semi-solid J, Rades T, Löbmann K. Quantification of and liquid formulations. Anal Chim Acta microwave-induced amorphization of celecoxib in 2018;1013:54–62. 10.1016/j.aca.2018.02.007. PVP tablets using transmission Raman 82. Griffen J, Owen AW, Matousek P. Development of spectroscopy. Eur J Pharm Sci 2018;117:62-67. Transmission Raman Spectroscopy towards the in 10.1016/j.ejps.2018.02.012. line, high throughput and non-destructive 73. El-Zahry MR, Lendl B. Structure elucidation and quantitative analysis of pharmaceutical solid oral degradation kinetic study of Ofloxacin using dose. Analyst 2015;140:107–12. surface enhanced Raman spectroscopy. 10.1039/c4an01798f. Spectrochim Acta A Mol Biomol Spectrosc 83. Mazurek S, Szostak R. Quantification of active 2018;193:63-70. 10.1016/j.saa.2017.12.007. ingredients in pharmaceutical suspensions by FT 74. Holm P, Allesø M, Bryder MC, Holm R. Q8(R2). Raman spectroscopy. Vibrational Spectroscopy ICH Quality Guidelines. In: Andrew T, David E, 2017;93:54-64. 10.1016/j.vibspec.2017.10.003. Raymond W N, editor. ICH Quality Guidelines: An 84. Harting J, Kleinebudde P. Development of an in- Implementation Guide. Hoboken: John Wiley & line Raman spectroscopic method for continuous Sons, Inc. 2017, p. 535-577. API quantification during twin-screw wet 10.1002/9781118971147.ch20. granulation. Eur J Pharm Biopharm 2018;125:169– 75. Food and Drug Administration (FDA). Guidance 81. 10.1016/j.ejpb.2018.01.015. for industry: PAT — a framework for innovative 85. Korasa K, Vrečer F. A study on the applicability of pharmaceutical development, manufacturing, and multiple process analysers in the production of quality assurance. Rockville, MD: FDA; 2004. coated pellets. Int J Pharm 2019;560:261-272. Available at: 10.1016/j.ijpharm.2019.01.069. https://www.fda.gov/media/71012/download. 86. Hisazumi J, Kleinebudde P. In-line monitoring of Accessed Juny 15, 2019. multi-layered film-coating on pellets using Raman 76. Esmonde-White KA, Cuellar M, Uerpmann C, spectroscopy by MCR and PLS analyses. Eur J Lenain B, Lewis IR. Raman spectroscopy as a Pharm Biopharm 2017; 114:194-201. process analytical technology for pharmaceutical 10.1016/j.ejpb.2017.01.017. manufacturing and bioprocessing. Anal Bioanal 87. Kim B, Woo YA. Coating process optimization Chem 2017;409:637–49. 10.1007/s00216-016- through in-line monitoring for coating weight gain 9824-1. using Raman spectroscopy and design of 77. Peeters E, Tavares da Silva AF, Toiviainen M, Van experiments. J Pharm Biomed Anal 2018;154:278– Renterghem J, Vercruysse J, Juuti M, Lopes JA, De 84. 10.1016/j.jpba.2018.03.001.

45 J Pharm Pharm Sci (www.cspsCanada.org) 23, xxx - xxx, 2020

88. Barimani S, Kleinebudde P. Evaluation of in–line for Surface-Enhanced Raman Spectroscopy under Raman data for end-point determination of a Different Conditions. J Nanomater 2013;2013. coating process: Comparison of Science–Based 10.1155/2013/790323. Calibration, PLS-regression and univariate data 96. Xu X, Li H, Hasan D, Ruoff RS, Wang AX, Fan analysis. Eur J Pharm Biopharm 2017;119:28-35. DL. Near-field enhanced plasmonic-magnetic 10.1016/j.ejpb.2017.05.011. bifunctional nanotubes for single cell bioanalysis. 89. Barbosa SF, Takatsuka T, Tavares GD, Araújo Adv Funct Mater 2013;23:4332–8. GLB, Wang H, Vehring R, Löbenberg R, Bou- 10.1002/adfm.201203822. Chacra NA. Physical–chemical properties of 97. Lewandowska AE, Eichhorn SJ. Quantification of furosemide nanocrystals developed using rotation the degree of mixing of cellulose nanocrystals in revolution mixer. Pharm Dev Technol thermoplastics using Raman spectroscopy. J 2016;21:812-822. Raman Spectrosc 2016;47: 1337-1342. 10.3109/10837450.2015.1063650. 10.1002/jrs.4966. 90. Chen M-L, John M, Lee SL, Tyner KM. 98. Miloudi L, Bonnier F, Bertrand D, Byrne HJ, Perse Development Considerations for Nanocrystal Drug X, Chourpa I, Munnier E. Quantitative analysis of Products. AAPS J 2017;19:642-651. curcumin-loaded alginate nanocarriers in hydrogels 10.1208/s12248-017-0064-x. using Raman and attenuated total reflection 91. Davis BM, Pahlitzsch M, Guo L, Balendra S, Shah infrared spectroscopy. Anal Bioanal Chem 2017; P, Ravindran N, Malaguarnera G, Sisa C, Shamsher 409:4593-4605. 10.1007/s00216-017-0402-y. E, Hamze H, Noor A, Sornsute A, Somavarapu S, 99. Chen J, Liu H, Zhao C, Qin G, Xi G, Li T, Wang Cordeiro MF. Topical Curcumin Nanocarriers are X, Chen T. One-step reduction and PEGylation of Neuroprotective in Eye Disease. Sci Rep 2018;8. graphene oxide for photothermally controlled drug 10.1038/s41598-018-29393-8. delivery. Biomaterials 2014;35:4986-95. 92. Tyner KM, Zheng N, Choi S, Xu X, Zou P, Jiang 10.1016/j.biomaterials.2014.02.032. W, Guo C, Cruz CN. How Has CDER Prepared for 100. Mujica Ascencio S, Choe CS, Meinke MC, Müller the Nano Revolution? A Review of Risk RH, Maksimov G V., Wigger-Alberti W, Assessment, Regulatory Research, and Guidance Lademann J, Darvin ME. Confocal Raman Activities. AAPS J 2017;19:1071-1083. microscopy and multivariate statistical analysis for 10.1208/s12248-017-0084-6. determination of different penetration abilities of 93. Doǧan I, Van De Sanden MCM. Direct caffeine and propylene glycol applied characterization of nanocrystal size distribution simultaneously in a mixture on porcine skin ex using Raman spectroscopy. J Appl Phys 2013;114. vivo. Eur J Pharm Biopharm 2016;104:51-8. 10.1063/1.4824178. 10.1016/j.ejpb.2016.04.018. 94. Doğan İ, van de Sanden MCM. Characterization of 101. Kuku G, Saricam M, Akhatova F, Danilushkina A, Nanocrystal Size Distribution using Raman Fakhrullin R, Culha M. Surface-Enhanced Raman Spectroscopy with a Multi-particle Phonon Scattering to Evaluate Nanomaterial Cytotoxicity Confinement Model. J Vis Exp 2015. on Living Cells. Anal Chem 2016;88:9813–20. 10.3791/53026. 10.1021/acs.analchem.6b02917. 95. Hong S, Li X. Optimal Size of Gold Nanoparticles

46