Polymorphism is an issue troubling numerous scientific fields. A phenomenon where molecules can arrange in different orientations in a crystal lattice, polymorphism in the field of organic photovoltaic materials can dramatically change electronic properties of these materials. Rubrene is a benchmark photovoltaic material showing high carrier mobility in only one of its three polymorphs. To use rubrene in devices, it is important to quantify the polymorph distribution arising from a particular crystal growth method. However, current methods for characterizing polymorphism are either destructive or inefficient for batch scale characterization. Lattice phonon Raman spectroscopy has the ability to distinguish between polymorphs based on low frequency intermolecular vibrations. We present here the addition of microscopy to lattice phonon Raman spectroscopy, which allows us to not only characterize polymorphs efficiently and nondestructively through Raman spectroscopy but also concurrently gain information on the size and morphology of the polymorphs. We provide examples for how this technique can be used to perform large, batch scale polymorph characterization for crystals grown from solution and physical vapor transport. We end with a case study showing how Raman microscopy can be used to efficiently optimize a green crystal growth method, selecting for large orthorhombic crystals desired for rubrene electronic device applications.
Polymorphism in molecular crystals impacts progress in a number of fields, including synthetic chemistry, pharmacology, and materials science.1,2 It occurs when a molecule can form different combinations of intermolecular interactions, leading to different arrangements of the crystal lattice.3 Significant research has been placed into understanding how and why molecules pack in these specific arrangements.4,5 Understanding the interactions that direct molecular packing can also help predict which polymorphs are most likely to be formed.6 With infinite combinations of packing arrangements available for any single molecule, computationally screening through each potential arrangement for a thermodynamically stable form is an active area of research, taking significant time and resources.7,8 While improvements have been made in computational polymorph prediction, it is still necessary to screen for polymorphism within a sample to verify or test predictions.9 The more polymorphs are observed experimentally, the more we learn about the identity of intermolecular interactions that direct crystal packing. The key to both the understanding and prediction of polymorphism is being able to efficiently characterize any and every observed polymorph.
Single-crystal x-ray diffraction (SC-XRD) has long been the dominant method for characterizing a crystal structure.2,10 This method requires large, high quality crystals with few defects so that exact atom positions can be determined within the crystal lattice, thereby assigning the crystal structure. Crystals of this quality can be difficult to obtain, and a single measurement can take many hours, meaning that many crystals are excluded from analysis.9 Thus, thoroughly characterizing a molecule’s polymorphism through this method can be a challenge, given that SC-XRD is generally not used to assign a crystal structure to an entire batch of crystals. If multiple polymorphs exist in a batch, each crystal would need to be analyzed individually to find any additional polymorphs, a process that could take many days. Powder x-ray diffraction (PXRD) is often used for distinguishing between polymorphs, having advantages over SC-XRD in that a powder sample contains numerous crystallites; therefore, multiple polymorphs in a batch can be detected.11 However, obtaining quantitative information on the polymorphic ratio from PXRD can be difficult. During sample mounting, crystals may order in what is known as preferred orientation,12 reducing intensity of reflections required for data processing. This problem can be surmounted by grinding the sample, which can cause phase changes, solvent loss, decomposition, and other issues that can alter the polymorph ratio.13 As PXRD involves powder samples, no information is derived on the shape or size of a single crystal in the sample. The community needs a quantitative polymorph characterization method that is efficient, nondestructive, provides information on the shape and size of polymorphs present in a sample, and requires less sample restrictions than commonly used methods.
Raman spectroscopy offers an alternative method of polymorph characterization that meets these goals of nondestructive analysis, batch scale characterization, and few sample restrictions. Raman spectroscopy often probes intramolecular vibrations, which generally occur in the range of 500–3000 cm−1. Because polymorphs share the same molecular structure, they are also expected to have extremely similar intramolecular vibrations. As solid-state packing is driven by intermolecular interactions, significant differences in intermolecular vibrations (<150 cm−1) are expected for different polymorphs.14 Raman spectroscopy has been used to probe these intermolecular modes, termed lattice phonons, and thereby distinguish polymorphs based on these modes.15,16 This technique has expanded far past simple polymorph assignment to applications including identifying different polymorphs within the same crystal,14,17 mapping grain boundaries in crystals,18 and monitoring crystallization in pharmaceuticals.19
Our specific interest in lattice phonon Raman spectroscopy lies in polymorph characterization of organic photovoltaic (OPV) materials. The electronic properties of these materials have long shown a dependence on solid-state packing,20–22 highlighting the importance of efficient polymorph identification following the crystallization process. Lattice phonon Raman spectroscopy has been used to distinguish polymorphs of popular OPV materials such as tetracene,23 pentacene,24 rubrene,25 and more.26 Rubrene [Fig. 1(a)] is a benchmark organic semiconductor with three well-characterized polymorphs. However, only the orthorhombic polymorph of rubrene [Fig. 1(b)] displays high charge carrier mobility in single-crystal devices.27,28 We have synthesized an entire library of rubrene derivatives in attempts to alter this orthorhombic packing arrangement and further increase the carrier mobility of rubrene.29 As we cannot predict polymorphism for these derivatives, we require an efficient polymorph characterization method to screen an entire batch of crystals before integrating them into devices. Lattice phonon Raman spectroscopy offers the advantage of large-scale screening of these derivatives to find the optimal orthorhombic packing.
Here, we use lattice phonon Raman spectroscopy to solve the problem of batch characterization of polymorphs through our combination with Raman microscopy. Adding Raman microscopy to batch characterization collects new information on crystal size across different polymorphs, providing data not readily accessible through SC-XRD or PXRD. Our method expands upon lattice phonon Raman spectroscopy previously used to characterize individual crystals. By mapping the size and shape of individual crystals, we can more efficiently optimize growth procedures for different applications. Integrating rubrene into single-crystal devices requires large crystals of a specific polymorph. This technique of Raman microscopy is able to assess the number of crystals within a specific growth method, in addition to assigning the polymorph. This makes Raman microscopy a more application-based technique as we can use it to put an upper bound on the number of “usable” crystals grown by a specific technique. In addition, we provide a case study in which we efficiently optimize a green solution crystallization method to grow large, orthorhombic rubrene crystals by tracking the polymorph and size distribution through lattice phonon Raman microscopy.
For recrystallizations, we used commercial rubrene (98% + purity) purchased from Sigma-Aldrich with no further purification. We used two different methods to grow crystals from solution. The first was reported to give mixtures of triclinic and monoclinic crystals through liquid–liquid diffusion, although, as described in the section titled “Results and Discussion,” we observed only triclinic crystals.31 For this method, we layered methanol (MeOH) over a saturated solution of rubrene in dichloromethane (DCM). Crystals typically began to form after 3 h and grew for three days. We used three different ratios of DCM to MeOH for crystallization as described later. Previous studies reported polymorphic mixtures of rubrene through a second solution growth method.28 We followed a similar procedure using 0.2 mg/ml rubrene concentration in ethanol (EtOH) instead of the reported 1-propanol. We also grew rubrene crystals by this method using two additional concentrations: 0.5 and 1.0 mg/ml concentrations. We sealed the solutions and heated just above the boiling point (85 °C) for 10 h and then allowed to cool to room temperature for a growth period of fourteen days. As rubrene is prone to photobleaching in the presence of oxygen, we grew and stored crystals from all the described solution methods in the dark. We took no additional measures to limit exposure to air.
We grew crystals from physical vapor transport (PVT) in a horizontal apparatus, similar to previous reports.30 To remove impurities, we heated the chamber overnight to 400 °C. We placed a sample boat with 15 mg rubrene inside the chamber. We sealed the chamber and put it under an argon flow of 100 ml/min at 0.15 MPa. Rubrene sublimed at 300 °C, and a thermocouple created a temperature gradient of 50 °C over 16 in. with glass collection pieces spaced in 1-in. intervals, beginning 5 in. from the sample boat. Crystals grew for 4.5 h before we allowed the system to cool to room temperature overnight under Ar.
We used single-crystal x-ray diffraction to confirm polymorphic assignments of different crystal morphologies. We used a Bruker Photon II CMOS diffractometer with molybdenum radiation (λ = 0.710 73 Å) at low temperature (125 K) to collect unit cells of a single crystal from each growth method and compared the unit cells to structures in the Cambridge Structural Database for orthorhombic (QQQCIG11)30 and triclinic (QQQCIG14)31 rubrene polymorphs. We collected powder diffraction patterns for polymorph mixtures on a PANalytical X’Pert diffractometer with cobalt radiation (λ = 1.789 010 Å) at room temperature. We scanned over a 2θ range from 3 to 35° with a 120 s dwell time and integrated data using EXPO2014. We made polymorph assignments from simulated powder diffraction patterns produced with the Mercury software.
Lattice phonon Raman spectroscopy
We obtained spectra on a home-built setup. A 633 nm He:Ne laser (ThorLabs, HNL210L) reflected off a 633 nm BragGrate bandpass filter (OptiGrate) passed through a 30R/70T beam splitter (Edmund Optics 43-434). The beam entered an inverted microscope (Olympus IX73), where a 10× (Olympus Ach, NA = 0.25) or 60× (Plan L, NA = 0.70) objective focused 8 mW of power onto the sample. The beam passed back through the objective and beam splitter in a backscattering geometry and through three 633 nm BragGrate notch filters (OptiGrate). A lens (100 mm focal length) focused the beam into a monochromator (PI, Acton SP2500) with a 1200 grooves/mm grating (PI 750 nm blaze) and a CCD detector (PIXIS 100BX, 7515-4002). We calibrated the spectrometer before each measurement using chloroform (Fisher). The addition of a ThorLabs fast XY scanning stage (MLS203) and sample holder (MLS203P10) allowed for Raman imaging. Our setup allows for a maximum of 6 μm spatial resolution, as determined by a scanning razor edge measurement, and 70 μm axial resolution, as determined by measuring the Raman signal across a chloroform/glass boundary.
To prepare rubrene samples for Raman imaging, we suspended the entire collection of crystals grown from one technique in cold MeOH. We then dropped aliquots of the suspension onto a glass slide to evenly disperse the crystals. We obtained optical images of the samples with a CCD camera (Leica, DFC450) on an upright microscope (Leica, M165 C). We selected four regions of each sample with adequate crystal coverage for imaging. For crystals grown from solution, we imaged over 2 × 2 mm2 regions with a step size of 10 μm. For crystals grown by PVT, we imaged over 0.6 × 0.6 mm2 or 1 × 1 mm2 regions with a step size of 3 or 5 μm, respectively. Lattice phonon Raman spectra for each pixel had an exposure time of 0.1 s. We used MATLAB for image analysis (code available upon request). A low-pass frequency filter set to one-thousandth the sample frequency removed noise from each spectrum. We calculated the minimum intensity found for each dataset at each pixel on the CCD and subtracted them from the data to baseline correct for systematic noise. We used nonlinear least squares fitting to fit each peak between 15 and 55 cm−1 to a Gaussian line shape. We separated peaks into two bins based on the center frequency of the peak (25–36 cm−1 for expected orthorhombic peaks and 41–50 cm−1 for expected triclinic peaks; see results for details). We converted each bin to a binary image and used the “bwboundaries” function to determine the boundaries of each crystal. To determine the area of each crystal, the “polyshape” function fits each boundary to an n-dimension polynomial. We set the requirement for a crystal to >4 pixels, limiting the minimum area for a measurable crystal to 100 μm2 for solution-grown crystals or 25 μm2 for PVT crystals. We plotted the final Raman images in MATLAB by filling in the boundaries of each crystal. A comparison of the filled and unfilled plots is located in the supplementary material (Fig. S4). For Figs. 4(h)–4(j), the exposure time was increased to 0.3 s to detect thin crystalline sheets. The three BragGrate notch filters were adjusted to increase Raman signal, resulting in more bleed through of the fundamental 633 nm beam. This prevented detection of the 32 cm−1 mode. Instead, we fit the other three phonon modes unique to the orthorhombic polymorph (102 cm−1, 116 cm−1, and 136 cm−1 modes).
To assign vibrational motions to the lattice phonon Raman spectra, we optimized the geometry of three rubrene dimer pairs for the orthorhombic polymorph and two for triclinic (Fig. S1) to serve as a representation of the unit cell. We performed all density functional theory calculations on Gaussian16 using the B3LYP/6-31G(d) level of theory.32,33 Initial coordinates came from the published crystal structures (orthorhombic QQQCIG11 and triclinic QQQCIG14), and we froze the coordinates of the carbon atoms for optimization. By convention, we applied a scale factor of 0.962 to the calculated Raman frequencies;34 however, we did not aim to match the intensities between computed and experimental Raman signals. We averaged the frequencies between the dimers for comparison to experimental results. The coordinates for the optimized geometries and a table of scaled and unscaled frequencies for each system are provided in the supplementary material.
RESULTS AND DISCUSSION
To develop Raman imaging for polymorph characterization, we first confirmed the polymorph assignments of our rubrene crystals. Three polymorphs of rubrene have been published in the Cambridge Structural Database. We followed the growth procedures associated with crystallization for the triclinic,31 monoclinic,31 and orthorhombic30 polymorphs. However, we found no monoclinic crystals formed from the solution method suggested.31 Other solution methods only gave mixtures of orthorhombic and triclinic polymorphs, which we will discuss later. Because we found the triclinic and orthorhombic forms to be the most prevalent, we focus on them here. Unit cell constants obtained through SC-XRD from the standard triclinic and orthorhombic crystals, grown through published methods,28 confirmed the polymorph assignments of our samples. That we did not obtain the monoclinic polymorph here further highlights the unpredictability of polymorphism and emphasizes the necessity for an improved polymorph screening method. Therefore, a batch characterization method is needed to confirm the crystal structure, even when grown using previously published procedures.
We then collected lattice phonon Raman spectra for a crystal of each polymorph to determine if the polymorphs could be easily distinguished solely based on their lattice phonons. The polymorphs give two distinct sets of lattice vibrations that can be assigned to the triclinic and orthorhombic forms based on previous studies.25,32 Figure 2 shows the lattice phonon Raman spectra and frequencies for both polymorphs, which we offset for comparison. The polymorphs appeared to share only one mode at 72 cm−1, which was assigned from DFT calculations (Table I) to a π-stacking motion of the tetracene cores, similar to previous studies.35,36 Each polymorph also had an additional π-stacking motion (orthorhombic, 102 cm−1 and triclinic, 119 cm−1). Both polymorphs had a vibration related to a scissoring motion of the side phenyl rings. These motions occurred at vastly different frequencies (orthorhombic, 32 cm−1 and triclinic, 91 cm−1). A motion for the triclinic polymorph involving both phenyl scissoring and π-stacking was assigned to the 44 cm−1 mode. The orthorhombic polymorph also contained a unique motion at 116 cm−1, which we assigned to a twisting motion of the tetracene core. The final 136 cm−1 mode for the orthorhombic polymorph was not able to be assigned with DFT calculations of the dimer. Instead, a correlative frequency at 137 cm−1 was calculated from a single molecule. We chose the lowest frequency mode (orthorhombic 32 cm−1 and triclinic 44 cm−1) to use as a basis for Raman imaging due to the large frequency difference between the modes for each polymorph (12 cm−1) and the lack of interference from other nearby modes.
|.||Triclinic .||Orthorhombic .|
|Mode .||Frequency (cm−1) .||Description .||Frequency (cm−1) .||Description .|
|1||44||π-stack/phenyl scissor||32||Phenyl scissor|
|.||Triclinic .||Orthorhombic .|
|Mode .||Frequency (cm−1) .||Description .||Frequency (cm−1) .||Description .|
|1||44||π-stack/phenyl scissor||32||Phenyl scissor|
The triclinic crystals grown from liquid–liquid diffusion produced the largest and highest quality crystals in addition to a mixture of morphologies. Because this crystal growth method was originally reported to give a mixture of polymorphs,31 we first used our imaging setup to scan over these different morphologies to support our SC-XRD assignment of triclinic across the batch (Fig. S2). Despite variations in Raman intensity of the 44 cm−1 mode across the crystal, we were able to see only the presence of the 44 cm−1 mode and no modes related to other polymorphs. This confirmed that both the block and needle shaped crystals grown from solution were triclinic crystals. To properly characterize the crystals grown from this method, it was necessary to use Raman imaging. Previous literature reported this method to grow two different polymorphs and we, therefore, expected to find two polymorphs.31 Because we observed multiple crystal morphologies, we initially assumed multiple polymorphs formed. While we still were able to provide polymorph assignment to a larger sample size than SC-XRD, we did not characterize every crystal, only a representative sample, and therefore, some implicit bias is involved in the selection of the imaged regions. However, if we relied on SC-XRD to characterize crystal batch, we would have had to look at a crystal of each morphology individually, a process that could take days depending on the number of different morphologies. Instead, Raman imaging was able to confirm the polymorph identity over multiple morphologies in a matter of hours.
We then sought to prove what additional information we could gain from Raman imaging beyond simple polymorph recognition. A highlight of Raman imaging is that it is a nondestructive method, keeping the crystals in their original shape and size. Techniques such as SC-XRD and PXRD often require cutting or grinding of the sample and can, therefore, no longer comment on the size of any crystal. In addition to losing valuable information, this makes it hard to determine if the polymorph ratio is a function of the crystal size. To prove that our technique can adequately characterize different crystal sizes, we deliberately attempted to alter the crystal size of solution-grown triclinic crystals by altering the solvent ratio for crystallization.
Figure 3 shows three Raman Figs. 3(c)–3(e) and optical Figs. 3(f)–3(h) images of batches of rubrene crystals grown from three different solvent ratios. Lattice phonon Raman imaging first confirms that these samples are composed entirely of the triclinic rubrene polymorph by plotting an image of the intensity of the 44 cm−1 mode characteristic of triclinic crystals. Second, by mapping the boundary of each area containing the 44 cm−1 mode, we can calculate the area of each crystal. We recognize that the top right crystal in Fig. 3(c) is rotated compared to Fig. 4(f). As we do not adhere the crystals to a slide, there is a possibility of crystals shifting between the collection of the optical and Raman images.
The histogram in Fig. 3(a) shows the number of crystals formed from each solvent ratio over a size distribution of 500 to 8 × 104 μm2. Looking at the lowest ratio of rubrene (in DCM) to MeOH (1:13), we see a large size distribution of crystals, as opposed to the higher concentrations (1:6 and 1:3.5), which have many crystals with an area below 500 μm2. Figure 3(b) highlights the changing crystal size as the ratio of rubrene to MeOH is lowered. The 1:13 solvent ratio produced nearly half (47%) of the crystals above 1 × 104 μm2, whereas the other solvent ratios had over 66% of their crystals below 1 × 104 μm2. By demonstrating a clear trend in crystal area with the solvent ratio, we show how optimization of a crystal growth method to increase crystal area is more efficiently accomplished through lattice phonon Raman imaging, as it allows for confirmation of both polymorph identity and crystal area for multiple crystals at the same time in a single technique.
One of the most common methods for growing rubrene crystals, especially for electronic applications, is through PVT due to the large orthorhombic needles produced.37 When growing crystals by PVT, the location at which crystals are collected can greatly affect the crystal morphology as illustrated in Fig. 4(a). In addition to needles, crystals from PVT can result in plates or thin sheets, as shown in the optical images in Figs. 4(b)–4(d). We show in Fig. 4 that all of these morphologies can be confirmed as orthorhombic crystals. Both the plates and needles are clearly distinguished by the intensity of the 32 cm−1 mode used to create the Raman images [Figs. 4(e)–4(g)]. The thin sheet-like crystal grown on the second collection piece posed an additional challenge as it would likely be too thin to produce an x-ray diffraction pattern or Raman scattering. However, on the edges of the crystal, we were able to detect the 32 cm−1 mode, suggesting that these sheets belong to the orthorhombic polymorph. To further increase the Raman signal and confirm this assignment, we increased the exposure time of the beam and adjusted our setup to maximize the other phonon modes. The Raman images in Figs. 4(h)–4(j) were generated using the other three unique phonon modes for the orthorhombic polymorph (102, 116, and 136 cm−1). Each of these additional images shows Raman signal throughout the crystalline sheet, which further confirm the orthorhombic assignment of this crystal.
Based on the images of the triclinic crystals grown from solution and orthorhombic crystals from PVT, we found this imaging technique to be successful in characterizing polymorphs grown from preparations previously optimized for a single polymorph growth. Previous studies have observed multiple polymorphs within the same rubrene crystal during sublimation growth; however, this was not something we observed.25 To test this technique’s ability to distinguish between two polymorphs in the same image, we took crystals previously characterized as triclinic or orthorhombic and combined them in a known mass ratio of 2:3 orthorhombic to triclinic [Fig. 5(a)]. Raman imaging was able to distinguish between both polymorphs within the same sample [Fig. 5(b)]. To demonstrate the clear definition between the two polymorphs, we plotted representative pixels for each of the polymorphs with no offset or background subtraction, as shown in Fig. 5(c).
We then compared these results to PXRD, which is often used to characterize polymorphism. Figure S3 shows the simulated powder patterns for each polymorph and the experimental powder pattern for our 2:3 mass mixture. Simulated reflections for each of the polymorphs are visible in the mixture, confirming the presence of both polymorphs. PXRD can also determine the polymorphic ratios through the intensity of reflection characteristic of each polymorph. However, due to preferred orientation of the crystals, many of the reflections are not visible in our mixture and the intensities of the visible reflections have been altered. PXRD assumes a random orientation of crystalline powder, but during preferred orientation, certain crystal faces are oriented in a direction more likely to reflect. For example, in our sample, the plates are likely to stack, and therefore, many of the faces of the crystals have the same orientation and are no longer random. Without a mixture of orientations, some of the reflections will not be present because the orientation does not allow for those reflections. The estimated polymorph ratio from PXRD is 32:1 orthorhombic to triclinic (Fig. S3, Table SXIII), but because of preferred orientation, we can only use PXRD here to confirm that both polymorphs are present in the sample. This further exemplifies the need for a polymorph characterization method that is not sensitive to shape, size, or crystal orientation.
While mapping a 2D area of crystals through Raman imaging does not allow for direct comparison to the mass ratio, we gain additional information about the polymorph composition not accessible through SC-XRD or PXRD. By mapping the area of each crystal, we are able to determine the number of crystals in the sample. Raman imaging identified 18 orthorhombic crystals and 14 triclinic crystals in this image (Fig. 5, Table II), giving a polymorph ratio of 56.2% orthorhombic and 43.8% triclinic. Because Raman imaging is a nondestructive method, we retain the individual crystals during characterization and, therefore, count the number of crystals belonging to each polymorph. As only the orthorhombic polymorph is used in single-crystal devices,38 it is important to know how many orthorhombic crystals of a given size are grown from a particular method.
|.||Orthorhombic .||Triclinic .|
|Average crystal size (μm2)||109||49|
|Number of crystals||18||14|
|% of polymorph||56.2%||43.8%|
|.||Orthorhombic .||Triclinic .|
|Average crystal size (μm2)||109||49|
|Number of crystals||18||14|
|% of polymorph||56.2%||43.8%|
We have shown that lattice phonon Raman microscopy can determine the polymorphic ratio of a mixed sample. To provide an example of the usefulness of this method, we present a case study showing how Raman imaging can be used to optimize crystal growth conditions for a desired application. We demonstrate here the optimization of a green solution growth method for large, orthorhombic single crystals. Earlier, we discussed the growth of a single polymorph from solvent mixtures of DCM/MeOH, resulting in large triclinic crystals. We avoided the use of halogenated solvents when searching for a new growth method in order to find a more environmentally conscious procedure. As EtOH is widely regarded as a green solvent,39 we began by crystallizing rubrene in ethanol at low concentrations. Concentrations of rubrene below 0.2 mg/ml in EtOH were too dilute to form crystals. The 0.2 mg/ml concentration of rubrene in EtOH gave mixtures of both the triclinic (48%) and orthorhombic (52%) polymorphs, as shown in Figs. 6(b) and 6(c). Increasing the concentration from 0.2 mg/ml rubrene in EtOH to 0.5 mg/ml gave majorly orthorhombic crystals [97%, Fig. 6(b)]. Imaging over multiple regions showed the presence of trace triclinic needles [Fig. 6(d)]. As a pure orthorhombic mixture was the desired product, we increased the concentration further to 1.0 mg/ml [Fig. 6(e)]. Imaging scans over multiple regions showed only orthorhombic plates [Fig. 6(b)]. The ability to screen over large areas of the crystalline batch identified triclinic impurities in the 0.5 mg/ml crystal batch that may have previously been overlooked. The image analysis also revealed that the orthorhombic crystals from the pure polymorph batch (1.0 g/ml) were, on average, larger than from the other two methods. The size distribution in Fig. 6(a) shows a larger distribution of sizes from the 1.0 mg/ml sample and overall more crystals larger than 18 × 104 μm2. Further increasing the concentration led to powder samples. Of the three concentrations, the combination of larger crystal size and polymorph purity makes the 1.0 mg/ml EtOH growth method the most ideal for growing large, orthorhombic crystals.
We have shown here the process of optimizing a crystal growth method using lattice phonon Raman microscopy. Without the use of a batch characterization method such as lattice phonon Raman imaging, the triclinic impurities from the 0.5 mg/ml sample may never have been detected. Because we were able to view these impurities, we further altered the growth concentration, resulting in a polymorphically pure sample and overall larger crystals, which would be more useful for applications such as single-crystal electronic devices.
There is a strong need for a polymorph characterization method that allows for efficient, large-scale, nondestructive analysis. SC-XRD, the standard method for characterizing the crystal structure, allows only for characterization of a single crystal and cannot efficiently analyze multiple crystals required for thorough polymorph characterization. PXRD has the ability to recognize different polymorphs but is a destructive technique and, as a result, we lose valuable information on the size and morphology of the crystals. We have shown here that Raman spectroscopy can overcome these drawbacks by differentiating polymorphs through intermolecular vibrations or lattice phonons. Raman spectroscopy is nondestructive in nature, contrary to x-ray crystallography methods. The addition of microscopy to this technique greatly increases the efficiency over SC-XRD, allowing for imaging over batches of crystals with quantitative characterization of crystal area and polymorph identity. We demonstrated how this technique can be used to optimize a crystal growth method, selecting for not only a specific polymorph but also a specific crystal size. The dual analysis of the crystal size with polymorph characterization can be applied to a variety of other materials for which the knowledge of the polymorph ratio would be important, materials ranging from other OPVs to pharmaceuticals.
See the supplementary material for full details on phonon mode assignments—including optimized coordinates, frequencies, and vibrational mode animations—along with PXRD patterns and additional Raman images of triclinic crystals.
We would like to thank Dr. Lee Penn and Maetzin Cruz-Reyes for powder diffraction analysis and Dr. Victor Young, Jr., and the X-Ray Crystallographic Laboratory at the University of Minnesota for resources and instrumentation. We acknowledge the Minnesota Supercomputing Institute (MSI) at the University of Minnesota for resources. This work was supported by DOE under Grant No. DE-SC0018203 and by the National Science Foundation Graduate Research Fellowship under Grant No. 00074041.
Conflict of Interest
The authors have no conflicts to disclose.
The data that support the findings of this study are available from the corresponding author upon reasonable request.