CicarboTM Graphene Characterization and Analysis
In spite of its simple atomic structure, graphene has not only been notoriously difficult to make, but equally difficult to characterize. Most standard characterization techniques are pushed to their instrumental limitations to quantify the structure, size distribution, and relevant physical properties of various grades of graphene particles. Currently, no standards exist within the industry for quantifying the relevant properties of different grades of graphene, precisely for the two reasons cited above: graphene has been difficult to produce at the industrial scale and has proven very hard to characterize accurately. Indeed, ASTM International has not yet even begun to develop protocols for graphene characterization and analysis, and it is only within the past few years that any attempts have been made at establishing a basis of nomenclature for denoting various grades of graphene [Carbon, vol. 65, p. 1, 2013].
At Celtig, we have devoted a great deal of effort to characterizing our products as completely as possible. Below, we present a sampling of information derived from hundreds of hours spent studying our materials using the most sophisticated techniques currently available for such analyses. Characterizations have been performed primarily in-house, but with independent verification from several external entities, as noted below. Detailed technical data sheets summarizing the results of these analyses can be found using the menu to the right for various grades of CicarboTM graphene. Besides simply presenting the results of these analyses, we will also devote some time to discussing the various characterization techniques as well as the advantages and limitations associated with each one. Only by understanding the mechanical limitations of the various instruments is it possible to understand why the physiochemical properties of graphene can exhibit such broad ranges of values from one published journal article to the next. In so doing, we can hopefully clear up some of the confusion surrounding contradictory characterizations of this material and also obtain a certain level of understanding as to why graphene has proven so difficult to analyze.
The primary physical property of graphene is the layer number, since its amazing properties are directly related to the number of layers that the material possesses—the fewer the better. Accordingly, the best indicator of the quality of a particular grade of graphene is the layer number, but, unfortunately, this is usually the most difficult quantity to determine. Although TEM (and sometimes SEM) can be used “to visualize” graphene particles (see below), it is not a reliable method for determining layer number, both because the layer numbers cannot readily be counted accurately and the sample size of particles that can be examined under TEM is too small to form a statistically significant set. Consequently, the only analysis technique that can (semi-)accurately determine the nanoflake layer number is Atomic Force Microscopy. AFM can be used to determine rough estimates of both particle-size and layer-number distributions.
An AFM is usually operated in “tapping mode,” in which a conducting, electronically-sensitive nanoscale cantilever is systematically moved over the sample grid, measuring the electric potential response at each grid point and converting this into a height of a particle relative to the sample substrate. AFM does not visualize the particles per se, although it does provide height profiles of the nanoparticles that can have a 3-d effect which approximates the particles’ shapes; however, it is important to keep in mind that a particle might have an uneven surface topology (like a crumpled sheet of paper) resulting in AFM images that appear to have holes in them which are merely regions that are relatively closer to the sample substrate baseline than the highest peaks on the particle’s topological landscape.
Below is the AFM image of a flat single-layer graphene flake with a planar dimension of about 4 μm and a thickness of roughly 0.4 nm from the Materials Research and Innovation Laboratory (MRAIL) at the University of Tennessee in Knoxville using an NTEGRA Prima SPM. This is approximately the theoretical thickness of a single layer of graphene, 0.34 nm. The virtual image on the left can be used to estimate the particle’s planar dimensions using the scales on the x and y axes. The height trace in the right figure shows the particle thickness (relative to the substrate) in the x direction along the tracer line (it is very faint!) located at y = 4.0 μm in the image on the left. At this y location, the particle is about 2 μm in the planar x direction.
To obtain an accurate layer number count, much experience is necessary to understand the limits of the AFM instrument, which can vary from machine to machine as well as from user to user. The theoretical distance between two graphene layers is 0.335 nm, which, together with the layer thickness, can be used in principal to determine the layer number of an AFM imaged particle. However, as discussed by Nemes-Incze et al. [Carbon, vol. 46, p. 1435, 2008], the measured height actually depends on the way in which the particle interacts with the substrate on which it is deposited. They measured the step height from one layer of graphene to another under a wide range of conditions and found that each additional layer could add as much as 1 nm to the total particle height as measured by AFM. As such, accurate layer counts require expert operators who are able to minimize these errors and control the precision of the measured step height. In MRAIL, the measured AFM step height in moving from one layer to the next is approximately 0.6 nm.
Below are several more typical AFM images of Cicarbo graphene taken by MRAIL and by researchers at Jilin University in Changchun, China, consisting of graphene flakes with one to four layers and of planar dimensions ranging from 100 nm to 1 μm. The traces represent the height of the flakes at the locations indicated in the images by the faint lines.
Thermogravimetric analysis is a simple technique wherein chemical and physical properties of a substance can be measured by increasing the temperature of the sample under a prescribed ramping function. For graphene, the primary use of TGA is to determine the relative proportions of carbon and inorganic constituents within the material; i.e., the effective purity of the graphene. Cicarbo graphene is continuously monitored using TGA, typically with a ramping function of 50 °C/min beginning at room temperature and increasing to 900 °C. To the right is a typical profile of the weight percentage of EG016 Cicarbo graphene as a function of ramping time, where the maximum temperature of 900 °C is achieved after about 18 minutes of operation. At this temperature, all the organic components within the graphene should be combusted, and if the sample were pure carbon, the weight percentage of the remaining sample would be 0.00. Any number above this indicates the amount of inorganic impurity within the sample, such as ash content. As evident from the TGA profile, Cicarbo graphene grade EG016 is approximately 99.8% pure carbon. Other grades of Cicarbo graphene contain > 99.5% (MG016) and > 95.5% (NCG015) carbon, respectively.
Raman spectroscopy is a common, economical spectroscopic technique that is used to observe the atomic vibrational, rotational, and other low frequency dynamic modes in materials. It is a very useful tool in materials characterization laboratories that is used to provide a wavenumber spectrum from which a sample’s constituent molecular or atomic elements can be identified. It makes use of inelastic scattering, or Raman scattering, of monochromatic light from a laser in the visible, near infrared, or near ultraviolet spectra. The laser light interacts with molecular vibration or other excitation mechanisms within the sample, resulting in the energy of the laser’s photons being shifted up or down the frequency spectrum. This shift in energy provides information about the dynamic vibrational modes within the material.
Although RS is a reliable, fast, and easy way to identify the constituent elements and chemical bond networks within materials, as with all other characterization methods, graphene once again proves to be the exception to the rule. Much research has been directed at studying the Raman spectrum of single and multi-layered graphene, but practical problems remain; that is to say, the very distinctive Raman spectrum of graphene very quickly reduces to that of graphite as the layer number increases. Indeed, graphene of 5 or more layers is essentially indistinguishable from the spectrum of many-layered graphite.
Taken from Quantum Frontiers (https://quantumfrontiers.com/2013/09/06/graphene-gets-serious/).
With reference to the RS diagram above, and the Raman spectra below, there are three primary peaks in the Raman spectrum of graphene. The D peak arises due to one or both of the following: defects in the structure of graphene or edge effects arising from the laser beam striking the edge of a sample. (Basically, these are the same, since defects present holes, and thus edges, to the laser spot.) A typical Raman spectrometer utilizes a laser with a spot size of about 1 μm in diameter. If the laser beam overlaps the edge of a sample, the D peak is evident in the resulting spectrum. For CVD grown graphene that is defect free, if the laser strikes the center of the sample, the D peak is not present, but for powdered graphene that contains many particles smaller than the diameter of the laser spot, the D peak will always be present.
The locations and relative heights of the G and 2D peaks provide clear evidence of the layer count of a graphene sample, but only up to 4 or 5 layers, as discussed above. In single-layer graphene, the G peak occurs at approximately 1587 cm-1 and the 2D peak at 2700 cm-1 [Ferrari et al., Phys. Rev. Lett., vol. 97, art. 187401, 2006; Ferrari, Solid State Comm., vol. 143, p. 47, 2007]. The ratio of the height of the G peak to that of the 2D peak provides the most recognizable feature of single-layer graphene; i.e., this ratio is roughly 0.5 for pristine graphene. As the layer number increases, the ratio increases, being approximately 1 for bilayer graphene, about 1.2 for few-layer graphene and roughly 2.0-2.5 for 4 or 5-layered graphene all the way up to many-layered graphite. The D’ peak is also related to defects and edge effects in the sample, and the G/D’ peak ratio usually hovers around 5, with lower values tending to indicate sample defects and higher values indicating edge effects.
The Raman spectra below were taken at the Center for Nanophase Materials Science (CNMS) at Oak Ridge National Laboratory in Tennessee using a RS at a wavelength of 514 nm. The upper image is the Raman spectrum of EG016 Cicarbo graphene and the lower image is of a common graphite. The relevant peak locations and peak height ratios are displayed in the figures. One can see clearly the trends of the Raman shifts discussed above with increasing layer number. For instance, the G peak of the Cicarbo graphene sample shifts downward from 1587.7 cm-1 as the layer number is increased, shortly reaching the upper limit after 4 or 5 layers that is the same for many-layered graphite, 1581.7 cm-1. G/2D = 1.15 for the Cicarbo graphene sample, but approximately 2 for graphite. In all respects, the Raman spectrum of 4 or more layered graphene is indistinguishable from that of graphite.
In short, the slight variations in the Raman spectrum of graphene and graphite make it difficult for even an experienced user to analyze quickly various samples for layer count and quality. The relevant factors are many and the small shifts in the peaks that occur with layer number are difficult to be obtained by a casual user with any degree of confidence. It can be done, of course, if the operator is willing to devote the necessary patience to the task; however, there is one quick analysis that RS is extremely well suited for: distinguishing between a true graphene and a reduced graphene oxide being marketed as graphene.
Graphene oxide also has a distinctive Raman spectrum, as shown below. However, as pointed out by Sobon et al. [Optics Express, vol. 20, p. 19463, 2012], the Raman spectra of graphene oxide (GO) and reduced graphene oxide (RGO) are practically indistinguishable from each other. The only minor difference is that the RGO has a more distinct peak at about 2700 cm-1, corresponding to the 2D peak in graphene. The Raman spectra below, taken at CNMS, clearly display the similarities in the two spectra. The characteristic features of these spectra are associated with the oxygen atoms in the chemical network; obviously, the RGO contains much the same atomic framework as the GO, which for graphene should be completely devoid of oxygen content. (The RGO used in this analysis was purchased as “graphene” from an international corporation, but RS quickly revealed its true nature.)
In summary, although Raman spectroscopy is not a particularly ideal analysis tool for counting graphene layers, or even in distinguishing few-layer graphene from bulk graphite, it is very quick and easy to apply it to determining whether a particular material is truly graphene or merely a reduced graphene oxide.
Transmission electron microscopy (TEM) is a microscopy technique in which a beam of electrons is transmitted through a thin sample and experiences interference due the scattering of some of the electrons from the atoms within the material. In an optical microscope, a light source is used to image a sample material, whereas in a TEM, the electron beam acts as the light source. The beam is focused on the sample and as the electrons pass through it, some of the electrons are scattered in random directions whereas the unscattered electrons are collected on a screen, thus creating a shadow image of the sample. TEMs have significantly higher resolution than optical microscopes, down to a few Angstroms, because the wavelength of the electrons (i.e., the de Broglie wavelength) is much smaller than the wavelength of light. As a consequence, a TEM can examine the fine atomic structure of materials, down to single rows of atoms.
Below are images of Cicarbo Graphene (MG016) on a lacy carbon grid at various length scales, as obtained using a Zeiss Libra 200MC TEM at the Advanced Microscopy and Imaging Center (AMIC) at the University of Tennessee in Knoxville. The images are typical of few-layer graphene, which often appears in a folded structure when deposited on a substrate. Focusing down to smaller length scales allows imaging of the fine atomic details of the samples.
Although TEM can provide fairly clear images of atomic structure, it is not well suited for making determinations for particle layer counts because of the small number of sample particles that can be examined within a reasonable period of time. AFM is much more suitable for making layer count determinations. However, TEMs can also output the scattering patterns directly, and these can give accurate ideas of layer number at particular points within the sample. In the snapshots below, one can see single-layer (upper left), bilayer (upper right), trilayer (lower left), and multi-layer (lower right) graphene nanoparticles. Single-layer graphene displays the hexagonal packing structure of the graphene honeycomb, with a single white dot indicating the placement of the atoms on the hexagon. Bilayer graphene displays two white dots at each hexagonal position, owing to the displacement of the carbon atoms between the layers; i.e., the carbon atoms in one layer are stacked above the holes on the layer below it. Trilayer graphene will exhibit several dots, whereas multi-layer graphene will display many points, which ultimately become of sufficient number to make continuous rings for bulk graphite (not shown).
A scanning electron microscope (SEM) is similar to the TEM in that it uses a focused beam of electrons, possessing a much smaller wavelength than light waves, to irradiate a sample and scan its surface. The interaction of the electrons in the beams with the sample’s atomic constituents provides reconstructed images of the material topology, which can be visualized as actual images of the sample. Because of the small wavelength of the electron beam, the resolution of an SEM is extremely high, and because there are no lenses involved (because there is no light source), highly tunable electromagnets are used to control the resolution at a level of very fine detail, as well as from a larger working distance to image a larger portion of the sample than can be managed by a TEM.
Below are some typical images of few-layer and multi-layer graphene materials obtained using a Zeiss Auriga SEM at the Advanced Microscopy and Imaging Center at the University of Tennessee. The images display particle snapshots at various resolution qualities. Note in the figure on the far right the peeling back of the atomic layers of the graphene sheet.
Like other characterization techniques, graphene can be a problem when analyzing with an SEM. Although in principle one should be able to obtain clear images of few-layer graphene, in practice this can be hard to achieve. The primary problem is that imaging few-layer graphene with an SEM requires lowering the electron voltage (1-3 keV is preferred) because thin graphene sheets can be transparent for a high energy electron beam because of its high penetration depth [Xie and Spallas, Microscopy and Microanalysis, vol. 19, suppl. 2, p. 370, 2013]. Therefore, unlike TEM, layer counts are virtually impossible using an SEM, and it is also difficult to get a particle count as well due to transparency issues.
Dynamic light scattering is a characterization technique that can be used to determine the size distribution of particles in solution, and indirectly the specific particle surface area if the true particle density is known and reasonable assumptions can be made about the particle shape. In suspension, the particles move in random directions due to thermally induced fluctuations, with particles of different sizes moving at different rates. [See the schematic to the right: https://en.wikipedia.org/wiki/Dynamic_light_scattering.] These can be analyzed as a function of scattering intensity of incident light since smaller particles fluctuate more rapidly and hence decorrelate over a shorter period of time. Typically, a monochromatic laser light source is passed through the sample and the light is scattered in all directions as it strikes the particles. The light thus diffracted can experience either constructive or destructive interference, leading to light and dark regions once it is collected and projected onto a screen producing a speckle pattern, as shown in the figure below. This process is replicated over many small time intervals and analyzed using scattering theory to determine the particle size distribution. When the particle size is much smaller than the wavelength of the incident light, Rayleigh scattering theory is used to determine the scattering particle size, whereas the more complicated Mie theory is used otherwise. However, these theories assume that the scattering particle is spherical, so if the particle has a different shape, the theory provides the size of a sphere that moves in the same manner as the scattering particle; hence there is some inherent uncertainty in a non-spherical particle’s actual dimensions.
A typical speckle pattern obtained during a dynamic light scattering experiment.
DLS is a relatively simple apparatus to use most of the time; however, graphene can cause uncommon problems with many analytical instruments. In the case of DLS, the particles to be studied are typically suspended in water, which has a well-characterized refractive index. Unfortunately, graphene is hydrophobic and will not form a stable dispersion in pure water. (If it does, you probably have a reduced graphene oxide that is being marketed as graphene!) This is critical sense measurements are made over a long period of time, and any precipitation will change the concentration and distribution of the particles within the solvent, rendering the results meaningless. Hence a solvent is necessary that disperses and stabilizes the particles and possesses a refractive index that is approximately that of water; therefore, the dispersing medium must be chosen very carefully. Celtig is currently working to develop an ASTM standard protocol for DLS measurements of graphene dispersions.
A typical particle-size distribution on a particle-number basis is shown in the output from a DLS measurement performed by MRAIL at the University of Tennessee using a Malvern Zetasizer Nano ZS operated at a wavelength of 632 nm using a dispersing medium comprised of a mixture of acetone and water. (Simple organic solvents such as acetone and ethanol have refractive indices almost identical to water and both disperse and stabilize graphene particles.) The analysis reveals that over 90% of the particles in this EG016 Cicarbo graphene sample have (an assumed planar) dimension lying within the range of 200 to 300 nm. Using the layer-number distribution provided by complementary AFM analyses (i.e., particle thickness ranges from 0.5 to 5.0 nm, corresponding to 1 to 5 layers) and assuming that the particles possess disc-like shapes, the specific surface area of this sample can be estimated as approximately 600 m2/g.
XRD is a very common laboratory technique used to identify the atomic and molecular structure of crystalline materials. In use, a beam of X-rays irradiates a sample, causing the incident X-rays to diffract in various specific directions due to the organizational structure and atomic patterning of the substance. The angles and intensities of the diffracted beams are measured and related to the density of electrons within the crystalline material, from which the positions of the constituent atoms can be deduced as well as the nature of any chemical bonds between them.
As is the case with Raman Spectroscopy, XRD is not particularly useful in determining the quality of graphene powders and films, although it does provide a rapid means of determining whether a sample is a true graphene, graphene oxide (GO), or a reduced graphene oxide (RGO). Graphite will show a number of peaks, with the two most dominating being those associated with the 002 and 004 planes located at angles 26º and 54º, respectively. The first peak corresponds to the graphite interlayer spacing of about 0.34 nm, whereas the second is associated with the graphene plane. Hence for a pristine monolayer of CVD grown graphene, the first peak should not be present and only the 004 peak should be evident. For graphene oxide, the dominant peak appears at roughly 10º, dwarfing any other peaks present in the spectrum. In the case of RGO, the peak at 10º shifts increasingly toward 26º as the quality of the reduction increases; i.e., as the oxygen content and defect degree decrease.
XRD patterns of graphene oxide, reduce graphene oxide, and graphite [Cui et al., Chem. Commun., vol. 47, p. 12370, 2011].
Unfortunately, unless one has a pristine monolayer sample of graphene, the 002 peak will always dominate the XRD spectrum. Hence for graphene powders composed of micron-sized particles of 1 or more layers, there is very little difference between the XRD patterns of graphene and graphite to exploit in order to characterize the material. Hence XRD is essentially limited to determining qualitatively whether or not a sample is GO, RGO, or some graphitic form of carbon of undetermined layer count.