Dick Wieboldt

Aligned semiconducting single-walled carbon nanotubes (s-SWCNTs) are expected to outperform silicon as the next generation of integrated circuits. Greater utilization of polarized Raman spectroscopy is proving beneficial for efficient characterization of alignment in carbon nanotube films. Here, we present the results of how polarized Raman imaging can be used to effectively characterize alignment in large regions of aligned s-SWCNT films.

	Semiconducting single-walled carbon nanotubes (s-SWCNTs) have been a promising material over the past decade and will continue to have a significant role in high-speed and low-powered semiconductor electronics. Recently, one of the major challenges in SWCNT electronics has been overcome-hierarchical organization of the s-SWCNTs (1). Ordered films of s-SWCNTs possess exceptional electronic properties including increased hole mobility and packing density compared to nonaligned s-SWCNT films (1). Using the novel yet simple method of floating evaporative self assembly (FESA), arrays of aligned s-SWCNTs with exceptional electronic purity can be deposited as seen in Figure 1 (1). These s-SWCNTs are likely to become a better alternative to silicon in the next generation of integrated circuits.

					Figure 1: (a) Polarized optical image of s-SWCNT array fabricated using FESA. The white box highlights one stripe of randomly aligned s-SWCNTs shown by the darker orange color. The black box highlights one lighter colored stripe corresponding to aligned s-SWCNTs. (b) Inset shows scanning electron microscope image of randomly aligned s-SWCNTs (top) and aligned s-SWCNTs (bottom).

	Various techniques used to study the molecular and electronic structure of aligned s-SWCNTs include absorption spectroscopy, photoluminescence, scanning electron microscopy, and atomic force microscopy (7). With the recent advances in Raman imaging, Raman spectroscopy is now among the most efficient methods in the analysis of carbon nanomaterials. In Figure 2, a typical Raman spectrum of s-SWCNTs is shown. The presence or absence of peaks along with the peak positions and relative peak intensities provides a wealth of information about various types of carbon nanotube (CNT) materials. For instance, the position of the radial breathing mode (RBM, near 160 cm

) allows the nanotube diameter and subsequently the chirality to be determined. The D band indicates the presence of structural defects; the intensity ratio between the D and G peaks provides a method to qualitatively compare the defect density between CNT samples. The G band distinguishes whether the CNTs are semiconducting or metallic. Furthermore, the position of the bands can be used to quantify the strain and doping in CNT films. A more comprehensive review of the theory behind Raman imaging of carbon nanotubes can be found in an article by Dresselhaus and colleagues (7).

					Figure 2: Raman spectrum of an s-SWCNT film with the D, G, M, and G’ bands labeled, in addition to the Si peak from the Si–SiO

 substrate. The inset shows the RBM of the s-SWCNT film along with another peak from the Si substrate.

	Polarized Raman spectroscopy can provide additional information to characterize the orientation and the quality of the alignment of s-SWCNT films. When the long axis of the s-SWCNT stripe is in parallel with the incident laser polarization (0°), there is a large enhancement of the Raman signal (1–6). As the sample is rotated away from this alignment, the signal decreases until it reaches a minimum at 90° (2–4). By extracting the maximum peak intensity (G-band used because of high signal-to-noise ratio) of the cosine

 fit (0°) and dividing it by the minimum peak intensity (90°), we obtain a ratio known as the 

. For well-aligned s-SWCNT films, the optical anisotropy has a value of ≥10 (2,4). Assuming the orientational distribution of nanotubes follows a standard Gaussian distribution, this value correlates to a Gaussian distribution with a sigma (σ) of 17°, as can be seen in Figure 3 (2). Thus, from a standard Gaussian distribution, a sigma of 17° indicates that the majority of nanotubes are aligned within ±17° of one another over the film area measured (2).

					Figure 3: Raman optical anisotropy plotted as a function of sigma (2). This is based on the assumption that within the CNT arrays, the nanotube orientational distribution follows a standard Gaussian distribution (2).

	The most efficient method to capture the data necessary to assess the quality of the alignment involves simply collecting two Raman images with rotating polarization optics. One Raman image with the vertically polarized incident laser and vertical analyzer orientation (V_V) aligned parallel to the SWCNT long axis. This measurement yields the strongest Raman peak intensity. The second measurement is collected with the horizontally polarized incident laser and horizontal analyzer orientation (H_H) with the alignment remaining the same (no rotation of the sample). This measurement produces the weakest Raman peak intensity. These measurements are equivalent to the 0° (V_V) and 90° (H_H) measurements mentioned previously. Thus, from the ratio of the V_V and H_H peak intensities (G-band used), we are able to calculate the aforementioned optical anisotropy and determine the quality of the alignment in the s-SWCNT films.

	A secondary method to determine the alignment characteristics of the s-SWCNTs involves plotting the Raman signal intensity as a function of the angle between the Raman laser polarization and the long axis of the s-SWCNT stripes. This approach expands upon the previous method by using a rotational stage to physically rotate the sample at a set polarization while providing the polarized Raman data at various angles between 0° and 90°; additionally, it serves to confirm the direction of the orientation in the CNTs. This method, however, is a more time-consuming approach and the mechanical nature of the setup yields a lower degree of accuracy.

	Carbon nanotubes were deposited on the silicon substrate via floating evaporative self-assembly (FESA) (1). For this technique, a substrate is suspended vertically in a water bath and an ink, which consists of polymer-sorted s-SWCNTs in a solution of 0.75 wt% ethanol in chloroform (prepared according to the method outlined by Joo, and colleagues [1]) is dosed at the substrate–water–air interface. The silicon substrate is uniformly withdrawn from the water at a rate of 5 mm/min. As the s-SWCNT ink evaporates, the interface between air and water also moves along the substrate via a slip-stick motion, which creates one strip of aligned s-SWCNTs (~20–50 mm) and one stripe of randomly oriented s-SWCNTs (~30–100 mm).

	Raman spectroscopy measurements were made using a Thermo Scientific DXR2xi Raman Imaging microscope. The data were collected using a 532-nm laser with a 50x/0.5 BD LMPlan FI Olympus objective. The following settings were used in the Raman spectroscopy experiments: 4 mW power at the sample, a 50-µm pinhole, 300 nm to 1.0 µm image pixel size, a 90-Hz exposure time, and 4 scans per image. The sample was mounted on a rotational stage with the long axis of the CNT stripes appearing vertically.

	The alignment of the s-SWCNT films was characterized using both methods mentioned above: automated control of polarization orientation and manual rotation of the sample using a rotational stage. To determine the quality of that alignment using polarization optics rotation, one polarized Raman image was collected in the V_V configuration and another in the H_H configuration. A rotational stage was used before image collection to maximize the Raman signal in the V_V configuration with the long axis of the s-SWCNT films. After the images were collected, two average areas were analyzed: an aligned stripe of s-SWCNTs and a stripe of randomly oriented s-SWCNTs. Figure 4 shows results from both of these areas. Figure 4a shows the G band (V_V and H_H polarization) of the average spectra from randomly oriented s-SWCNT areas whereas in Figure 4b it shows the average spectra from aligned s-SWCNT areas (V_V and H_H polarization). As expected, the optical anisotropy measured in the stripe containing randomly aligned s-SWCNTs is ~1, which corresponds with random alignment. The optical anisotropy measured in the stripe containing aligned s-SWCNTs is ~22, corresponding with aligned s-SWCNTs. From Figure 3 we can determine how well the s-SWCNT stripes are aligned; an optical anisotropy value of 22 corresponds to a σ of 12, thus indicating that the majority of nanotubes are aligned within ±12° of one another over the area measured. This is significantly greater than the σ of ~10 seen in literature for similarly well-aligned CNT films, thus indicating exceptional alignment in s-SWCNTs fabricated using the the FESA method.

					Figure 4: Polarization Raman spectrum of the G-band intensity as a function of s-SWCNT orientation and polarization orientation (V_V or H_H) using automated polarization: (a) G-band intensity of randomly oriented s-SWCNTs with V_V and H_H polarization orientations, (b) G-band intensity of aligned s-SWCNTs with V_V and H_H polarization orientations. The strong signal and high optical anisotropy (~22) indicate exceptional alignment in the s-SWCNTs.

	Additionally, to further confirm and supplement data already collected, polarized Raman images were collected at 15° intervals from 0° to 90° using a rotational stage in the V_V configuration. Before collecting these images, the rotational stage was used to maximize the Raman intensity of the G-band with the long axis of the SWCNT oriented parallel to the polarization of the incident light, thus providing the signal for the 0° position. Figure 5b illustrates the change in the Raman signal intensity as the sample is rotated from 0° to 90° with respect to the polarization of the incident light. Figure 5a shows the results from an area of unaligned s-SWCNTs; Figure 5b shows an aligned area. Note the difference in scale for the intensity axis when comparing the two figures. Quantitatively, the optical anisotropy in Figure 5b can be calculated and yields a value of ~22 wheras in Figure 5a, the optical anisotropy was calculated to be ~1. An optical anisotropy of ~22, again, indicates exceptional alignment in these SWCNT films compared to those found in literature. This data further confirms that the FESA fabrication technique yields well-aligned s-SWCNT stripes.

					Figure 5: Polarization Raman spectrum of the G-band with angle dependence of both aligned and randomly oriented s-SWCNT areas. (a) Randomly oriented s-SWCNTs. The variation in intensity with angle is small, ranging from 125–180 a.u. This variation is due to imperfect sample positioning repeatability. (b) Aligned s-SWCNTs. The strong signal intensity indicates a high degree of alignment in the stripe of aligned s-SWCNTs.

	From the polarized Raman images using the V_V and H_H orientations, we are able to produce an “orientation” image from the ratio of those Raman images. This spectroscopic image not only allows us to visualize the areas containing well aligned s-SWCNT films, but also provides a quantitative measurement of the extent of orientation in these areas.

	Image contrast for the Raman chemical images shown in Figure 6 is generated by the intensity of the G-band in spectra collected at each pixel. From Figure 6a, we can infer that the red areas have the highest degree of alignment. Subsequently, the blue and black areas have the lowest degree of alignment. By collecting a second image at crossed polarization, Figure 6b, we now have enough information to calculate the alignment of the s-SWCNTs.

					Figure 6: Polarized Raman imaging on arrays of s-SWCNTs with various polarization orientations: (a) laser polarization aligned with s-SWCNT axis (V_V), (b) laser polarization perpendicular to s-SWCNT axis (H_H), (c) orientation image produced by dividing the spectroscopic arrays from (a) and (b) polarization images. The areas of higher intensity represent areas of well aligned s-SWCNTs whereas areas of low intensity represent areas of randomly oriented s-SWCNTs.

	The image contrast in Figure 6c is generated from a calculation of optical anisotropy at each pixel. Average optical anisotropy values observed were between 1.0 and 22.0 corresponding to the nonaligned and exceptionally aligned portions of the image. It indicates areas of subtle variation in alignment quality and with a careful comparison of Figures 6a and 6c, it reveals areas of exceptional alignment. This orientation image, enabled by polarized Raman imagery, provides a detailed map of the quality of SWCNT alignment, gives us a better understanding of our experimental process, and can be further used to optimize process parameters in future fabrications of aligned arrays of s-SWCNTs.

	Polarized Raman spectroscopy is a powerful analytical tool for the analysis of molecular orientation. Here, the alignment of the s-SWCNT films fabricated via FESA was confirmed through two methods of polarized Raman: automated control of polarization orientation and manual sample rotation using a rotational stage. The high quality of the alignment was characterized by calculating the optical anisotropy, which was found to be ~22; this correlates with an exceptionally high degree of alignment. Additionally, using the polarization Raman images, an orientation image was produced which visually indicates the exact areas of alignment and quality of the alignment in those areas. With the growing research and development in fabricating aligned CNT films, polarized Raman imaging serves as a quick and informative tool to characterize alignment in them.

	Additional supporting information is available on-line at: 

www.spectroscopyonline.com/supplementary-information-polarized-raman-spectroscopy-aligned-semiconducting-single-walled-carbon-n

	The authors declare no competing financial interest. This work was partially supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1256258 (A.M. and K.J.). This work was also supported by the National Science Foundation CMMI-1462771.

	This version of this article includes a number of significant corrections to the text and to two of the figures (Figures 1 and 5) compared to the version that appeared in print in June 2016. This version should be considered the official version.

		Y. Joo, G. Brady, M. Arnold, and P. Gopalan, 

		Q. Cao, S.J. Han, G.S. Tulevski, Y. Zhu, D.D. Lu, and W. Haensch, 

		X. Li, L. Zhang, X. Wang, I. Shimoyama, X. Sun, W.S. Seo, and H. Dai, 

		J. Hwang, H. Gommans, A. Ugawa, H. Tashiro, R. Haggenmueller, K.I. Winey, J.E. Fischer, D. Tanner, and A. Rinzler, 

		C.L. Pint, Y.-Q. Xu, S. Moghazy, T. Cherukuri, N.T. Alvarez, E.H. Haroz, S. Mahzooni, S.K. Doorn, J. Kono, and M.Pasquali, 

Physics Reports-Review Section of Physics Letters

 is with Thermo Fisher Scientific and the Department of Electrical and Computer Engineering at the University of Wisconsin-Madison in Madison, Wisconsin. 

 are with the Department of Materials Science and Engineering at the University of Wisconsin-Madison. Direct correspondence to: 

Articles

Aligned semiconducting single-walled carbon nanotubes (s-SWCNTs) are expected to outperform silicon as the next generation of integrated circuits. Greater utilization of polarized Raman spectroscopy is proving beneficial for efficient characterization of alignment in CNT films. Here, we present the results of how polarized Raman imaging can be used to effectively characterize alignment in large regions of aligned s-SWCNT films.

Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes

	This information is supplementary to the article “Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes” that was published in the June 2016 

Raman Technology for Today’s Spectroscopists

	Optical Anisotropy Calculation in Figure 4

	One way of using the Raman imaging data is to obtain an average reading for a larger area of the sample compared to a single point measurement. The boxed areas in Figure 4 represent areas where the Raman spectral measurements were averaged to produce a single Raman spectrum. The boxed area in Figure 4 on the left is an area of unaligned SWCNTs; the boxed area on the right is an area of aligned SWCNTs.

	Each boxed area in Figure 4 is 11 x 47 µm (517 sq. µm) with spectra collected at a spatial resolution (pixel size) of 0.3 µm; this comes out to ~1800 individual spectra. Therefore, each boxed area produces a single Raman spectrum representing the averaged sample area. We calculated the peak area of the G-Band from the averaged spectrum collected at each polarization configuration (V_V, H_H) and took the ratio. A correction factor was applied to compensate for the reduced optical throughput in the H_H configuration. This correction factor is calculated from the silicon peak at 519 cm

 Raman shift which is not sensitive to polarization.

	Therefore, the average optical anisotropy value for the aligned s-SWCNTs is ≈22 and for randomly oriented s-SWCNTs is ≈1.4.

	The V_V and H_H polarization images were exported from OMNICxi as grayscale TIFF images having the same intensity scale. This preserved the relative numerical values of each pixel between the two images. The TIFF images were imported into ImageJ where the subsequent orientation image was created by dividing the V_V image by the H_H image (3). The resulting image was scaled by a factor of 0.882 to correct for the lower instrument throughput caused by additional polarization optics (half wave plate) when collecting the H_H image.

		A. Mashal, D. Wieboldt, K. Jinkins, and M.S. Arnold, 

This information is a supplementary to the article “Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes” that was published in the June 2016 Spectroscopy supplemental issue Raman Technology for Today’s Spectroscopists (1).

Supplementary Information: Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes

The needs of the lithium-ion battery customers can be segmented into in situ and ex situ modes of analysis. In situ analysis allows researchers to follow changes in a battery cell during its charge and discharge cycles. Recent improvements in Raman sensitivity enable these changes to be imaged on a dynamic time scale. The same techniques offer significant improvements for ex situ analysis of battery components by providing more thorough sampling than traditional single-point measurements. Examples are presented showing the advantages of both modes of analysis for lithium-ion electrodes.

	The use of Raman spectroscopy for the analysis of battery materials has been around for years. During the 1960s, researchers used Raman to elucidate many of the fundamental spectral features of the minerals and inorganic materials widely used in battery research today (1,2). Raman is a good fit for these materials because many of the characteristic vibrational and rotational modes occur in the low-wavenumber region of the spectrum typically accessible only by far-infrared (IR) measurements. In that day and age, both Raman and far-IR measurements were time consuming and difficult experiments.

	Advances in instrumentation have greatly increased the ease-of-use of Raman making it a much more approachable technique. New areas of application ensued such as the exploding interest in rechargeable lithium (Li)-ion batteries. Many researchers are involved and have published careful studies of materials specifically related to Li-ion batteries as well as next generation batteries. A review article by Baddour-Hadjean (3) published in 2010 is an excellent resource for those wishing to get up to speed in this field. The focus of this article is on the in situ and ex situ applications of Raman spectroscopy as it pertains to battery research.

Analysis Techniques: In Situ Versus Ex Situ

is used to describe experiments where the battery components are studied in an assembled cell under operating conditions. Think of in situ as a window on the case of a battery that lets you see the chemistry of what goes on when you charge and discharge a battery. There are very few commercially available cell designs compatible with spectroscopic measurements. Researchers have resorted to building their own cells to meet the needs of their experimental apparatus. Examples of such designs have been published along with experimental results (4–10).

	While in situ cells provide valuable information, their use is generally targeted at research and development of batteries. After a formulation is designed, a candidate battery is scaled up through pilot production to actual product samples. At this stage of development, researchers are most interested in characterizing failure modes and gaining a better understanding of performance differences. For example, they want to learn what makes one production run work better than another, and why one battery fails yet its siblings from the same batch work fine.

	To answer these questions, researchers carefully disassemble a battery cell so the individual components can be examined. This type of analysis is termed 

 because the battery components are removed from the operating battery cell. The goal is to prepare the samples for analysis in as close to a native state as possible.

	Battery disassembly for ex situ analysis is carried out in an inert environment such as an argon-filled glove box to protect the battery components from moisture and oxidation. For example, the anode, separator, and cathode sandwich must be carefully separated and rinsed to remove excess electrolytes.

	After the samples are prepared, they must be kept in an inert environment to protect against changes during analysis. When space is available, instruments are installed inside the glove box so the samples can be analyzed. In most cases, the sample must be removed from the glove box and transferred to an external instrument for analysis. This is where an ex situ transfer cell becomes a key component of the workflow. It preserves the inert environment around the sample so it can be studied.

From Single Point Measurements to Raman Imaging

	The majority of published research on Li-ion battery in situ Raman work is based on single-point measurements acquired over time during charge–discharge cycles. An example is the excellent work done by Kostecki's group at Lawrence Berkeley National Lab (11).

	Single-point measurements can be misleading because there is no way of knowing if the sampled point is representative of the entire electrode. It is important to make multiple measurements to be sure. Because the Raman signals are weak, it takes many minutes to generate enough signal-to-noise ratio at each measurement point. A complete, multipoint experiment can be quite time consuming to complete.

	Today, Raman imaging is a viable alternative that lets you quickly make thousands of measurements over an area of the electrode rather than just single-point measurements. Each pixel in a Raman image is a complete Raman spectrum. This technique provides confidence in understanding if changes are heterogeneous or hot spots.

	The following experimental results demonstrate the flexibility of using Raman spectroscopy for both in situ and ex situ analysis of Li-ion batteries and their components.

	Graphite is widely used as an anode material for rechargeable Li-ion batteries.

	During the Li-ion battery charging cycle, positively charged Li

 ions move from the cathode through the electrolyte, across a separator into the anode to balance the flow of electrons in the external circuit (Figure 1). This process of Li

 ions entering the graphitic structure of the anode is called 

. Intercalation causes changes in the anode structure-primarily a swelling of the graphite structure.

 ions balance electrons during the charge and discharge cycles of a Li-ion battery.

	The experimental setup for this example consists of a DXRxi Raman imaging microscope (Thermo Scientific) and an ECC-Opto-Std optical electrochemical cell (EL-CELL). This cell enables the investigation of batteries in a so-called sandwich configuration where the working electrode material is placed under a sapphire (Al

) window. Electrode material (graphite powder in this example) is spread onto a copper grid serving as the current collector. This working electrode is sandwiched from below, with a glass fiber separator soaked with the electrolyte solution and lithium metal as the counter electrode.

	The Raman beam from the microscope objective impinges onto the backside of the working electrode material through the sapphire window (Figure 2). The advantage of investigating the backside of the electrode is that the pathway for the Raman beam is minimized, allowing the use of high magnification objectives and optimizing the spectra quality. The drawback is the gradient of lithiation concentration along the depth of the electrode. Accordingly, the electrode must be charged very slowly to minimize this unwanted gradient.


	Figure 2: Experimental setup for the in situ example showing the electrochemical cell mounted on the stage of a Raman imaging microscope.

	The graphite electrode was cycled at a constant rate of approximately 0.06C. The C-rate is measure of how rapidly a battery is charged-discharged. This rate of 0.06C corresponds to 33 h for a full charge-discharge cycle between 1.5 and 0.005 V against Li/Li

. Raman imaging was carried out during the initial 480 min of the charging (lithiation) process only.

	Raman spectra were collected over a 30 µm × 30 µm area at 1 µm pixel spacing using 2 mW of 532-nm laser excitation, a 0.01-s exposure time for each pixel, and 50 scans per image. Higher laser powers or longer exposure times resulted in burning of the graphite and boiling of the electrolyte.

	A Raman image is a hyperspectral data set with each pixel in the image being a complete Raman spectrum. Using a variety of spectral processing techniques, this hyperspectral Raman data generates image contrast pertaining to specific chemical features. It is this capability that visualizes minute differences within a sampled area. By collecting a sequence of Raman images, we now have the ability to monitor changes in both space and time. As mentioned earlier, a variety of chemical images can be created from each data set showing changes within the sampled area. Alternatively, the Raman spectral data within each data set can be averaged to produce a single spectrum for each time slice. In this mode, the Raman imaging data set is used as a means of homogenizing any differences in the electrode area. This average spectrum represents a single-point measurement yet each point represents a 30-µm square compared with the typical 1-µm sample area from a standard Raman microscope.

	In Figure 3, the three-dimensional (3D) view (bottom left) shows changes in the Raman spectrum as a function of time over 8.3 h (1–500 min). During this time, the battery cell is in the charging (lithiation) process only. This portion of the electrochemical cycle is shown in the lower right of Figure 3.


	Figure 3: Different views rendered from the time lapse hyperspectral Raman data provide a wealth of experimental information.

	The spectrum of graphite exhibits a prominent peak at 1580 cm

mode (G band). At potentials between 0.42 and 0.31 V (specific charge 33 and 45 mAh/g), the band gradually disappears along with the simultaneous emergence of a peak centered at 1590 cm

. This peak shift is attributed to the Li

 ions intercalated into the graphite structure. This is more easily seen in the center, two dimensional (2D) Raman image. The inset shows Raman spectra before and after the change.

	Toward the end of the charge cycle at 8.3 h (496 min), where the voltage is less than 0.15 V (specific charge greater than 146 mAh/g), a strong Raman band centered at 154 cm

 begins to appear. This Raman band has not been previously reported so its assignment is not conclusive. Strong Raman bands in this region have been attributed to TiO

	The type of views shown in Figure 3 are "spectrum-centric" because they show changes in the Raman spectra captured at different times during a time-based analysis. Figure 4 shows another way of exploring the same Raman imaging data set from an alternative "image-centric" point-of-view. Here, we are not as interested in the Raman spectrum itself, but rather its use as a tool to enhance differences within the image (image contrast).


	Figure 4: Raman images from different time slices in the graphite lithiation experiment.

	In Figure 4, Raman images are presented in which the image contrast is generated by multivariate curve resolution (MCR) analysis. In this case, MCR finds the differences not only within each image but also across the entire time sequence. A different color is assigned to each resolved component. This is analogous to the use of dyes in biological fluorescence imaging, which tag different parts of a cell. Each image is from the same 30-µm-square portion of the anode. The blue MCR component is indicative of the 1580 cm

 band; and red represents carbon black, a conductivity enhancer.

	It can be challenging to visualize the information content with such a massive wealth of data. Figure 4 shows just three frames to demonstrate this type of analysis. The changes are easier to grasp using a time lapse viewer of the complete time sequence (video clips are available at 

Ex Situ Example: Characterization of Li-Ion Anodes

	After safety concerns, a leading area of interest in Li-ion battery research is understanding the cause of performance degradation over time. Research indicates the solid electrolyte interphase (SEI) layer formed on the surface of the electrode is key to performance. The SEI layer is formed by deposition of organic and inorganic compounds during the first several charge-discharge cycles (12). It stabilizes the electrode from further decomposition and promotes reversible capacity. Because of the complexity of the SEI layer, results from any and all analytical techniques contribute to an incremental understanding of its behavior.

	As you might expect, it is a messy business to extract electrodes from a used battery so the SEI can be studied. It takes great care to prepare the sample and preserve its integrity for ex situ analysis. This is usually achieved by working in an argon-filled glove box to prevent sample degradation because of atmospheric exposure. A transfer cell with a window is used to seal the sample in the inert argon environment, and remove it from the glove box for analysis using a Raman microscope.

	For this example, anode samples from a disassembled Li-ion battery were cut and mounted in a transfer cell (Thermo Scientific) so that a cross-section of the anode could be imaged. Raman spectra were collected over a 76 µm × 160 µm area at spatial resolution of 1.0 µm per pixel using a DXRxi Raman imaging system (Thermo Scientific). Laser power at the sample was 2.0 mW at 532 nm with a 0.2-s exposure time and four image scans. A 50× long working distance, 0.5 NA microscope objective (Olympus) was used to focus through the transfer cell window.

	Figure 5 is a micrograph of the anode cross section. The copper current collector is in the center with anode material coated on both surfaces. The Raman image created from the spectral differences shown by the inset Raman spectra is superimposed. The Raman image clearly shows that the coating on one side of the copper current collector is dominated by carbon black (red) whereas the other side has a much greater density of the active graphite phase (blue).


	Figure 5: Cross-section view of a Li-ion battery anode. The Raman image indicates a difference in the anode coating on each side. Inset Raman spectra are color coded to the areas in the Raman image.

	This example demonstrated the advantage of Raman imaging over traditional single-point measurements. The major differences in the two coatings could easily have been missed by single-point measurements depending on where the points were measured.

	The high sensitivity of Raman imaging is a benefit for Li-ion battery analysis. In situ Raman imaging techniques show the spatial distribution of phase changes in electrodes over time. This is a capability that was not possible with single-point measurements using traditional Raman microscopy. Ex situ Raman imaging measurements give results with a higher degree of confidence compared to single points.

 is a senior scientist in the Research and Advanced Development group at Thermo Fisher Scientific in Madison, Wisconsin. 

 is an application specialist for Central Europe in the Molecular Spectroscopy division of Thermo Fisher Scientific GmbH in Dreieich, Germany. 

 is with EL-CELL GmbH in Hamburg, Germany. Direct correspondence to: 

	(3) R. Baddour-Hadjean and J.P. Pereira-Ramos, 

	(5) P. Novák, D. Goers, L. Hardwick, M. Holzapfel, W. Scheifele, J. Ufheil, and A. Wursig, 

	(7) E. Markevich, V. Baranchugov, G. Salitra, D. Aurbach, and M. Schmidt, 

	(8) Y. Luo, W.B. Cai, X.K. Xing, and D.A. Scherson, 

	(9) T. Gross, L. Giebeler, and C. Hess, 

	(10) K. Hongyou, T. Hattori, Y. Nagai, T. Tanaka, H. Nii, and K. Shoda, 

	(11) J. Lei, F. McLarnon, and R. Kostecki, 

 Lithium Ion Batteries - New Developments, 

I. Belharouak, Ed. (ISBN: 978-953-51-0077-5, InTech, 2012). Available at: 

http://www.intechopen.com/books/lithium-ion-batteries-new-developments/electrolyte-and-solid-electrolyte-interphase-layer-in-lithium-ion-batteries

The needs of the lithium-ion battery customers can be segmented into in situ and ex situ modes of analysis. In situ analysis allows researchers to follow changes in a battery cell during its charge and discharge cycles.

Techniques for Raman Analysis of Lithium-Ion Batteries

Modern Raman systems provide advanced "smart" algorithms that considerably aid in the optimization of the instrument's operating parameters. Inevitably, however, there are those who wish to understand the instrument's operation beyond its automated operation and to learn more about the interaction of various instrument settings. This article presents best practices for those who would like to better understand Raman spectrometer parameters.  

Some key aspects of dispersive Raman systems will be presented and advice on best practices is provided to enable optimum results, including laser power, aperture, exposure time versus number of exposures, sample focus, and background handling. Concluding will be a discussion on cosmic ray rejection, photobleaching, alignment, and calibration. 

 Raman signal strength is directly proportional to the power of the Raman laser (milliwatts) exciting the sample. It should be clear that the more laser power used, the larger the Raman signal will be. This is why the first best practice is to use full laser power whenever possible. One drawback to this recommendation is some samples will burn when exposed to full laser power. Usually, such samples are dark in color or have an absorption band close to the excitation wavelength. With valuable samples, dial down the laser power and check for sample damage. Increase the laser power exponentially until comfortable there is no sample damage. 

Just as important is laser power density at the sample. This can be maximized through the use of what is known as a high brightness laser, which allows tighter focus and improves Raman scatter "yield" for a given laser power. Sometimes there is a fine line between maximizing signal and burning a sample. Accurate measurement and very fine control of laser power at the tenths of milliwatts level is desirable. Some of the latest Raman systems on the market measure exact laser output and allow this fine control, making them ideal for samples such as carbon nanotubes and surface-enhanced Raman spectroscopy (SERS) experiments that are very sensitive to laser power. 

Another key hardware consideration is the instrument aperture. The aperture controls how much Raman signal passes into the spectrograph and onto the detector. Its primary use is to achieve spectral resolution. In some Raman microscope systems, it also enables confocal operation. As a general rule, use the largest aperture whenever possible. For example, a 50-μm slit. 

Apertures are typically slits or pinholes. They range in size from small, 10–20 μm to large, 50–100 μm. The larger the aperture, the greater the Raman signal admitted into the spectrograph and the larger the signal appears in the spectrum. 

Smaller apertures (10–25 μm) yield the rated spectral resolution of an instrument. Using a larger aperture (50–100 μm) will degrade the spectral resolution slightly. However, the loss in spectral resolution is often insignificant. A quick comparison of spectra collected with each size aperture will help with the decision. The choice of aperture for the application requires some experimentation.  

Figure 1 shows an acetaminophen tablet measured with 532-nm excitation. The difference in spectral resolution is minimal.  

Pinhole apertures admit the least amount of light. Pinholes are needed only when confocal operation with the Raman microscope is required. The choice here is relatively simple: if confocal operation is not needed, use slits. With a large, bulk sample, always use slits as confocal capability is not required anyway. 

One thing to keep in mind when comparing instrument specifications is the absolute size of an aperture. The size depends upon the optical design. A 10-μm aperture in one design might be equivalent to a 25-μm aperture in another. The size alone does not dictate spectral resolution. 

Sometimes higher resolution is needed to adequately identify samples. For example, distinguishing between different polymorphs or looking at ring breathing modes in carbon nanotubes. 

The Raman spectrum of l-cystine (Figure 2) is a good example to show the benefit of high spectral resolution. l-Cystine is sometimes used to measure Raman spectrometer resolution because it has several bands that are very close together, such as the two peaks at 76.8 and 65.7 cm

 Raman shift. There is also a shoulder on the peak at 110 cm

  that requires high resolution to detect.  

The main point to remember is a large aperture can provide a spectrum in cases in which the Raman signal is very weak with a minor loss of resolution. If there is adequate Raman signal, then by all means, use the smaller aperture will achieve the best spectral resolution. 

Despite the perception that the higher the spectral resolution the better, again there are tradeoffs and most practical Raman analyses including identification, detecting peak shifts, and revealing fine structure can be achieved at 4 and even 8 cm

  resolution. These resolutions provide an ideal balance between analytical information and signal to noise. 

Exposure time versus number of exposures: 

Modern Raman instruments enable the control of data acquisition by adjusting the exposure time and the number of exposures added together for a final spectrum. For best results, start by maximizing the exposure time. 

Exposure time is completely analogous to a photographic camera. Just as a longer exposure gives you a better picture in dim light, you get a better spectrum from a weak Raman scatterer. 

Spectroscopy is static. As such, for scientific measurements, it is common practice to record multiple exposures and combine them. This has been done for years by astronomers, physicists and chemists. 

Given the two options, longer exposure time versus more exposures, which is best? 

Deciding to choose a longer exposure time or use more exposures comes down to signal-to-noise ratio. Raman signals are measured in counts per second (cps). Using a longer exposure time or averaging many spectra will not increase the signal — just reduce the noise. The previous topics (laser power and aperture) explain how to increase the Raman signal. So the question comes down to which does a better job of reducing noise in the final spectrum.  

It turns out the main sources of noise in modern dispersive Raman systems are read noise and shot noise (1/f noise). Read noise is the noise introduced when the charge in the CCD is digitized and converted to a spectrum. Shot noise is noise proportional to the intensity of the signal (either Raman or fluorescence). Most CCD detectors themselves are very quiet and do not contribute significantly to noise when collecting signal during exposure. 

For a quiet sample, one with little shot noise, it makes sense that for a given measurement time, two long exposures produces less noise than many, short exposures. There are fewer read operations. Figure 3 is a good example of this effect.  

The sample used for Figure 3, silicon with 532-nm laser excitation, is collected for a total measurement time of 1 min. The number of exposures and exposure time are varied to produce the same total measurement time. The top trace (green) is the shortest exposure time (60 exposures × 1 s); the bottom trace (red) is the longest exposure time (2 exposures × 30 s). It is clear the longer exposure time yields lower noise. 

For samples with a fluorescence background, the difference is less pronounced. The reason being the shot noise caused by the fluorescence signal now contributes more to the spectrum noise level than other sources of noise. 

Figure 4 shows a Raman spectrum of sugar using a 1-min total measurement time and 532-nm excitation. One spectrum uses a long exposure time (2 × 30 s); the other uses a larger number of scans (10 × 6 s).  

The inset shows an enlargement of the baseline from 2608–2484 cm

  Raman shift. The 2 × 30 s spectrum shows a slight reduction in noise but the difference is negligible. 

For samples with lots of Raman signal, it does not make a huge difference what combination of exposure time and number of exposures is selected. It is weak Raman scatterers where these guidelines really pay off.  

Throughout this discussion, two exposures have been used. It is possible to use one exposure, but it is necessary to consider a main source of interference for integrating detectors like CCDs: cosmic rays.  

Recently, Raman systems have become smarter and incorporate many of these guidelines in software. For example, Thermo Scientific (Madison, Wisconsin) Raman systems include an auto-exposure feature. Auto-exposure selects the optimum exposure time and number of exposures for the sample. The signal-to-noise ratio is selected along with the maximum collection time. The software determines the data collections parameters to give the best spectrum without saturating the detector. 

Whether using a Raman microscope or a conventional Raman sampling accessory (nonmicroscope), the sample must be placed at the correct focus. It is critical that the excitation laser focus be coincident with the Raman collection optics. There are some Raman techniques, such as spatially offset Raman spectroscopy (SORS) (1), which rely on an offset of these beam paths. For the purposes of this discussion, traditional Raman spectroscopy is assumed. Most modern instruments handle this by using a 180° sampling geometry. This uses the same optics to excite and observe the sample. Even with this design, it is still important that the laser and spectrograph be aligned optimally to the beam path. 

When using visible wavelength lasers (514, 532, 633 nm), the visual focus and Raman focus are very close to being the same distance from the lens. It is sufficient to simply focus on the sample visually (microscope eyepieces or video camera) and collect data. 

When using longer wavelength lasers (780 nm and longer), there is a significant difference between the visual sample focus and the focus giving best Raman signal. (IR wavelengths focus further from the lens than visual ones.) After visually focusing on the sample, it is necessary to "maximize" the Raman signal. This is best done by watching the Raman signal while adjusting the focus or, optionally, using autofocus. 

Autofocus adjusts the sample position to attain the maximum Raman signal from the sample. For a Raman microscope, the sample needs to be placed on the stage and visually focused on the sample. The automated stage option is required which includes a 

For conventional Raman systems (nonmicroscope), an autofocus capability is even more useful because there is no visual focus reference. The sample is placed on the sample holder and the collect started. The system will locate the position for maximum Raman signal. Some algorithms enable the modification of the autofocus behavior of the system.  

Background signals sometimes interfere with autofocus, for example, when attempting to measure a sample at the bottom of a plastic well plate. The plastic material also produces a Raman signal. Depending upon the intensity of the analyte Raman signal, the plastic signal can "fool" the autofocus operation. To eliminate this problem, some systems provide the option to ignore a background interference. The software lets you identify a reference spectrum and subtract it before performing the autofocus operation.  

The final best practice recommendation involves background correction. This is somewhat of a misnomer because Raman is a scattering experiment with, ideally, no background. The "background" is actually the dark signal from the CCD detector. Dark signal is the product of dark current and exposure time. If either of these are changed, the background changes. 

Dark current involves residual electrons in the detector and depends primarily upon temperature. As the CCD cools, the dark current is reduced beneficially but the quantum efficiency, QE, is also reduced negatively. This is why most Raman systems utilize a cooled detector (CCD). Some CCD cameras are capable of operating down to liquid nitrogen temperature. This offers no significant reduction in noise for Raman because the QE is reduced at these colder temperatures. 

A good way to correct for the dark signal is to store background spectra measured at different exposure times while keeping the detector temperature constant. This is a slight inconvenience because every time the exposure time is exchanged, a new background is required. Fortunately, the software in most modern systems use data caching to handle this smoothly. 

OMNIC software (Thermo Scientific) measures backgrounds across a range of exposure times. A mathematical model is constructed that is used to predict the background response for any exposure time. This enables any exposure time to be used without having to collect new backgrounds. 

Cosmic rays are natural events detected by sensitive integrating detectors such as a CCD. They are random, unpredictable events. They typically show up as spikes in Raman spectra or backgrounds. They can also appear as broad singularities. 

Most systems provide cosmic ray rejection as an option during data collection. It is a good idea to always use this option, particularly for long exposures lasting more than a few seconds. While cosmic rays are infrequent, it is not unusual to get one during a 30–60 s exposure. 

Cosmic ray rejection requires two exposures. The two exposures are compared to detect and repair events. This is why many systems require a minimum of two exposures for each collection. Even if cosmic rays are not detected, the second exposure is advantageous because it reduces noise via signal averaging. 

Today's Raman spectrometers are more capable than ever before. The seeds of innovation in filter, laser, and CCD technology have produced a crop of instruments that are fast, sensitive, and robust. This is good news because scientists are constantly bombarded with challenging problems that require the top performance from their instruments.

Dick Wieboldt

Articles

Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes

Supplementary Information: Polarized Raman Spectroscopy of Aligned Semiconducting Single-Walled Carbon Nanotubes

Techniques for Raman Analysis of Lithium-Ion Batteries

Understanding Raman Spectrometer Parameters