Thermal Investigations on Carbon Nanotubes by Spectroscopic Techniques

: Carbon nanotubes (CNTs) thanks to their unique physical properties have been employed in several innovative applications particularly for energy storage applications. Certain technical features of carbon nanotubes, such as their remarkable speciﬁc surface, mechanical strength, as well as their electron and thermal conductivity are suitable for these applications. Furthermore, in order to produce a device, thermal treatment is needed and for this reason the trend of thermal decomposition of the tubes plays a key role in the integration process. The main purpose of this work was to characterize the thermal behavior of CNTs. In particular, we show the ﬁndings of an experimental study on CNTs performed by means of Fourier Transform InfraRed and Raman spectroscopy investigations. The collected FTIR and Raman spectra were analyzed by using two innovative procedures: spectral distance (SD) and wavelet cross correlation (XWT). From both analyses, a relaxation temperature value emerged of T = 206 ◦ C, corresponding to a relaxation inﬂection point. Such a system relaxation phenomenon, occurring in the ﬁber CNTs, could be connected with the decay of the mechanical properties due to a decrease in the alignment and compaction of the ﬁbers.


Introduction
The growing request for renewable energy sources has accelerated the development of novel techniques used to produce new devices for energy thermal storage. Nanotechnology can represent an important solution to these problems, as it can instigate the synthesis of advanced materials with improved performances, to be used in various applications, such as energy storage, molecular biology, and water treatment. Carbon nanotubes (CNTs) are considered to be among the most important nanostructured materials due to their particular chemical composition: indeed, they show a high correlation between structure and properties, that holds for different structures [1,2]. Since their discovery in the early 1990s, the electro-mechanical properties of carbon nanotubes have been extensively studied. Thanks to their important properties (such as good elasticity and mechanical strength, good electronic and thermal properties, and high specific surface), CNTs have been used effectively for numerous applications such as energy storage devices, photovoltaic devices, chemical sensors, and actuators [3][4][5][6]. CNTs represent monodimensional carbon allotropic forms, in which the carbon atoms present are arranged in such a way as to create a cylindrical shape. Carbon materials can be divided based on their atomic bond (sp, sp 2 , and sp 3 hybridizations) [7][8][9].
Furthermore, as regards the preparation of CNTs, among the numerous synthesis methods developed [10][11][12][13][14][15][16], laser ablation and graphite arc discharge are those most used to produce CNTs from the vapor phase. The carbon nanotubes obtained from these two synthesis methods have good characteristics but are characterized by the presence of defects linked to impurities. The problem of the synthesis methods is related to the high temperatures used [17,18], which result in considerable In order to improve the electronic properties of CNTs, different methods of functionalization of carbon nanotubes have recently been developed by replacing the carbon atoms on the walls with other elements such as nitrogen N [33][34][35][36] and boron B [37][38][39][40][41].
It is well known that because of their properties, such as high flexibility, mechanical strength, aspect ratio, as well as high thermal and high electrical conductivity, CNTs are employed for many applications, to mention a few; CNTs catalyst supports, CNTs conductive properties and CNTs energy storage. Regarding this latter application, thanks to their intrinsic properties, CNTs are the favorite materials for electrodes in batteries and capacitors and find applications in various fuel cell components [42]. Furthermore, due to the high capacity of heat and electrical conductivity and mechanical properties, CNTs are considered to be efficient electron field-emitters. As for the techniques concerned for these systems, both the FTIR method and Raman spectroscopy have been employed [43][44][45][46][47]. It is well known that FTIR and Raman techniques are non-destructive and not invasive. In particular, FTIR allows nanoscale structure changes to be revealed while Raman spectroscopy obtains information about their interaction with adsorbate molecules.
For the characterization of CNTs thermal properties, two innovative methods have been applied. The first, Spectral Distance, refers to the deviation from the spectrum at the lowest temperature, of T = 25 • C [48,49]. The second method, Cross Correlation Wavelet (XWT) by means of the coefficient of XWT (r XWT ), is based on the similarity degree of several spectra, taking as reference spectrum at the lowest temperature, of T = 25 • C [50][51][52]. Both the procedures reveal the same result.

Materials and Methods
The investigated materials were the CNTs, purchased from Sigma Aldrich, held at different temperatures. More specifically, the CNTs were placed with a metal spatula inside glass containers, which in turn were placed on a heating plate. They were heated, by means of an electric hot plate, to T = 25 • C, 50 • C, 100 • C, 150 • C, 200 • C, 250 • C, 300 • C, 350 • C, 400 • C, and 450 • C.
Such a technique turns out to be a complementary technique to other methods such as Raman spectroscopy and other neutron-based spectroscopies [64][65][66][67][68][69][70]. In this experimental work, CNTs spectra were registered by means of an ATR-FTIR Vertex 70 V spectrometer (Bruker Optics, Ettlingen, Germany) equipped with a platinum diamond ATR, working in the Mid-IR range from 4000 to 400 cm −1 and in the temperature range of 25-450 • C.
The main characteristic of the employed VERTEX 70v spectrometer (Bruker Optics, Ettlingen, Germany) is the RockSolidTM interferometer (Bruker Optics, Ettlingen, Germany), equipped with gold-coated optics and a 30 • angle of incidence in order to maximize efficiency and sensitivity and to minimize polarization effects.
For each spectrum, in order to obtain a good signal-to-noise ratio, 64 repetitive scans were accumulated with an average resolution of 4 cm −1 , in the 400-4000 cm −1 spectral range. As far as the Raman spectroscopy was concerned, Raman spectra were obtained by means of the Spectrometer BRAVO (Bruker Optics, Ettlingen, Germany); the laser source delivers two different excitation wavelengths, i.e., 785 nm and 1064 nm. Such a characteristic is especially important since, due to fluorescence, the analysis of raw materials by Raman spectroscopy is often difficult. The BRAVO spectrometer manages the Sequentially Shifted Excitation (SSETM) procedure, a patented fluorescence moderation process which allows a broad range of raw materials to be measured.
The investigated spectral range was 300-3200 cm −1 in the temperature range of 25-450 • C; spot size value was 10-15 µm at a magnification value of 10× lens.
Concerning the interpretation of spectra and the data analysis, preprocessing of the data was performed by means of the Bruker OPUS software. The following processing steps were performed: baseline treatment to decrease the differences between the spectra due to the shift of the baseline smoothing treatment to reduce instrument noise first derivative treatment to correct baseline shift second derivative to better distinguish the spectral characteristics spectra normalization to correct the path length dissimilarity and to diminish the disparities among single measurements.
Then, the data were processed by means of Matlab environment.

Spectral Distance
For the Spectral Distance evaluation, we consider the following expression (1): Here, A(ω) is the absorbance at frequency ω; T r represents room temperature that coincides with the reference temperature, that is the lowest value of temperature, i.e., T = 25 • C and finally ∆ω is the frequency resolution. This formula furnishes the value of the deviation from the reference spectrum (T = 25 • C) for each investigated temperature. Since the data points arrange themselves in a sigmoidal trend, the following model fit (2) was adopted: Here A is the sigmoidal amplitude, B denotes the steepness, T 0 represents the temperature value at the point of inflection, and C − DT indicates a linear background contribution [71,72]. This approach was applied both on registered IR and Raman spectra.

Wavelet Cross-Correlation
To evaluate the similarity that exists among the collected spectra at different temperatures, r XWT was measured. Such a method has found many applications in different fields, such as meteorology, neutron spectroscopy, financial science, and also in geological and geophysical sciences [73][74][75][76]. In general, Wavelet Transform (WT) breaks down a signal y(t) as sum of the scaled (through the positive parameter s) and shifted (through the parameter u) mother wavelets Ψ s,u by the following Equation (3): and, from this latter, one can express the Wavelet Spectrum (WS) by means of the following Equation (4): Then, starting from these two equations, we can define the Wavelet Cross-Correlation (r XWT ) as follows by means of Equation (5): where WT 1 (s, u) and WT * 2 (s, u) represent, respectively, the Wavelet Transform of the first and second spectrum and * is the complex conjugate, while WS 1 and WS 2 are the first and the second wavelet spectrum. The value of r XWT is between −1 and +1 which signifies that there is either a negative or a positive correlation [77,78]. Also in this case, from the evaluation of r XWT for each temperature, there results sigmoidal behaviour that can be analyzed with the model fit of Equation (2). Figure 1 shows, as an example, a comparison between the IR intensity as a function of wavenumber of CNT at T = 25 • C and at T = 450 • C in the Mid-IR spectral range, i.e., 400-4000 cm −1 . The characteristic peaks are induced by S-H bonds that are centered around 2327 cm −1 and 2113 cm −1 . As can be seen, by increasing the temperature, these characteristic peaks do not change significantly while in the right-hand side region of the infrared spectrum, called the fingerprint region, which covers the spectral range 1450-500 cm −1 , there are some relevant changes. In particular, a peak around 1143 cm −1 and 868 cm −1 appears that can be attributed to C-O stretching and C-C stretching.

Raman Spectroscopy
The Raman spectrum of pure CNTs has several characteristic bands such as the D band at 1300 cm −1 which can be attributed to the presence of defects in the structure of the CNTs, and the G band at 1575 cm −1 which originated from tangential vibration modes of the bonds in carbon atoms with sp 2 hybridization. The D and G bands are typical Raman peaks of CNTs and the peak at 2668 cm −1 represents the 2D band of carbon nanotubes (CNTs) [79,80]. The G band at 1575 cm −1 represents the phase vibration of the graphite lattice and the D band at about 1300 cm −1 is related to the disorder of the graphite edges. The G band is composed only of sp 2 type C-C bonds, corresponding to the concentric cylinders of the graphene layers, while the D band is attributed to vibrations of carbon atoms with dangling bonds in the graphite structure (carbon atoms having sp 3 hybridization similar to diamond) [81][82][83]. Figure 2 shows, as an example, a comparison between Raman intensity versus wavenumber for the two investigated samples, i.e., CNT at T = 25 °C and CNT at T = 450 °C in the range 1250-3000 cm −1 . As above mentioned, the characteristic bands of CNT are present. As can be seen, by increasing the temperature up to T = 450 °C, several changes occur in the spectrum, expecially in the 2500-2750 cm −1 region.  Figure 3a,b shows the spectral distance (SD) behavior with respect to temperature for CNTs, respectively for the recorded IR and Raman spectra, together with the fit of the model (brown line).

Raman Spectroscopy
The Raman spectrum of pure CNTs has several characteristic bands such as the D band at 1300 cm −1 which can be attributed to the presence of defects in the structure of the CNTs, and the G band at 1575 cm −1 which originated from tangential vibration modes of the bonds in carbon atoms with sp 2 hybridization. The D and G bands are typical Raman peaks of CNTs and the peak at 2668 cm −1 represents the 2D band of carbon nanotubes (CNTs) [79,80]. The G band at 1575 cm −1 represents the phase vibration of the graphite lattice and the D band at about 1300 cm −1 is related to the disorder of the graphite edges. The G band is composed only of sp 2 type C-C bonds, corresponding to the concentric cylinders of the graphene layers, while the D band is attributed to vibrations of carbon atoms with dangling bonds in the graphite structure (carbon atoms having sp 3 hybridization similar to diamond) [81][82][83]. Figure 2 shows, as an example, a comparison between Raman intensity versus wavenumber for the two investigated samples, i.e., CNT at T = 25 • C and CNT at T = 450 • C in the range 1250-3000 cm −1 . As above mentioned, the characteristic bands of CNT are present. As can be seen, by increasing the temperature up to T = 450 • C, several changes occur in the spectrum, expecially in the 2500-2750 cm −1 region.

Raman Spectroscopy
The Raman spectrum of pure CNTs has several characteristic bands such as the D band at 1300 cm −1 which can be attributed to the presence of defects in the structure of the CNTs, and the G band at 1575 cm −1 which originated from tangential vibration modes of the bonds in carbon atoms with sp 2 hybridization. The D and G bands are typical Raman peaks of CNTs and the peak at 2668 cm −1 represents the 2D band of carbon nanotubes (CNTs) [79,80]. The G band at 1575 cm −1 represents the phase vibration of the graphite lattice and the D band at about 1300 cm −1 is related to the disorder of the graphite edges. The G band is composed only of sp 2 type C-C bonds, corresponding to the concentric cylinders of the graphene layers, while the D band is attributed to vibrations of carbon atoms with dangling bonds in the graphite structure (carbon atoms having sp 3 hybridization similar to diamond) [81][82][83]. Figure 2 shows, as an example, a comparison between Raman intensity versus wavenumber for the two investigated samples, i.e., CNT at T = 25 °C and CNT at T = 450 °C in the range 1250-3000 cm −1 . As above mentioned, the characteristic bands of CNT are present. As can be seen, by increasing the temperature up to T = 450 °C, several changes occur in the spectrum, expecially in the 2500-2750 cm −1 region.  Figure 3a,b shows the spectral distance (SD) behavior with respect to temperature for CNTs, respectively for the recorded IR and Raman spectra, together with the fit of the model (brown line).  Figure 3a,b shows the spectral distance (SD) behavior with respect to temperature for CNTs, respectively for the recorded IR and Raman spectra, together with the fit of the model (brown line). The SD analysis procedure was applied to the whole investigated spectral range for each temperature spectrum.

Spectral Distance Evaluation
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 11 The SD analysis procedure was applied to the whole investigated spectral range for each temperature spectrum. Each data point represents the value of the deviation of the spectrum at a given investigated temperature from the respective spectrum measured at 25 °C. From the evaluation of the SD applied for each spectrum as a function of temperature an increasing sigmoidal trend was obtained. In order to extract quantitative information, a curve fit through Equation (2) was performed for the two profiles. From this analysis it emerged that the curve, for the IR spectra, presents a point of inflection equal to = 206.55 °C and for the Raman spectra presents a point of inflection equal to = 206.28 °C. Such a method furnishes the same value of inflection point for both the spectroscopic techniques employed.

Wavelet Cross-Correlation Analysis
From the value of , by means of Equation (5), for each investigated temperature of the registered IR and Raman spectra, it emerges that the values arrange themselves in decreasing sigmoidal behavior. In this case, a curve fit model based on Equation (2) was also employed for the two profiles and an inflection point at 206.19 °C for the IR spectra and an inflection point at 206.15 °C for the Raman spectra were obtained. Figure 4a,b reports the trend of XWT versus temperature for CNTs, respectively of the recorded IR and Raman spectra, in the temperature interval from 25 °C to 450 °C together with the model fit of Equation (2) (blue line). This approach, like the previous one, provides the same value of inflection point for both the spectroscopic techniques employed.  Each data point represents the value of the deviation of the spectrum at a given investigated temperature from the respective spectrum measured at 25 • C. From the evaluation of the SD applied for each spectrum as a function of temperature an increasing sigmoidal trend was obtained. In order to extract quantitative information, a curve fit through Equation (2) was performed for the two profiles. From this analysis it emerged that the curve, for the IR spectra, presents a point of inflection equal to T 0 = 206.55 • C and for the Raman spectra presents a point of inflection equal to T 0 = 206.28 • C. Such a method furnishes the same value of inflection point for both the spectroscopic techniques employed.

Wavelet Cross-Correlation Analysis
From the value of r XWT , by means of Equation (5), for each investigated temperature of the registered IR and Raman spectra, it emerges that the values arrange themselves in decreasing sigmoidal behavior. In this case, a curve fit model based on Equation (2) was also employed for the two profiles and an inflection point at 206.19 • C for the IR spectra and an inflection point at 206.15 • C for the Raman spectra were obtained. Figure 4a,b reports the trend of XWT versus temperature for CNTs, respectively of the recorded IR and Raman spectra, in the temperature interval from 25 • C to 450 • C together with the model fit of Equation (2) (blue line). This approach, like the previous one, provides the same value of inflection point for both the spectroscopic techniques employed.
From the performed SD analysis, both for the IR and Raman obtained spectra, it emerges that the obtained relaxation temperature values, i.e., the T 0 value of Equation (2), are very close for the two techniques. This value detects the decay of the mechanical properties and coincides with the oxidation of amorphous carbonate. Furthermore, one should also take into account that the tubes begin to oxidize at 200 • C under partial oxygen and that a greater quantity of oxygen increases the reaction rate at lower temperatures and causes oxidation of amorphous carbon at 200 • C.
Although the obtained relaxation temperature values are similar, the SD fit curve rises to~2.3 in Figure 3a, while to~3.5 in Figure 3b. A specular behavior is registered in Figure 4, where the fit curve drops to~0.2 in Figure 4a, while to~0.4 in Figure 4b. The amplitude difference can be attributed to the different technique sensitivities; in particular IR and Raman techniques couple with a different system quantity, i.e., dipole moment and polarizability tensor, respectively. sigmoidal behavior. In this case, a curve fit model based on Equation (2) was also employed for the two profiles and an inflection point at 206.19 °C for the IR spectra and an inflection point at 206.15 °C for the Raman spectra were obtained. Figure 4a,b reports the trend of XWT versus temperature for CNTs, respectively of the recorded IR and Raman spectra, in the temperature interval from 25 °C to 450 °C together with the model fit of Equation (2) (blue line). This approach, like the previous one, provides the same value of inflection point for both the spectroscopic techniques employed.  In this experimental work, the investigation of thermal behavior for CNTs was evaluated by means of Fourier Transform Infrared analysis and then continued by means of two innovative approaches, i.e., SD and XWT. From both techniques, a relaxation temperature value of about T = 206 • C was obtained, corresponding to the inflection point of the curves. Such a value was obtained both for the IR and Raman data. This relaxation temperature can be linked to the same relaxation phenomenon, which generally occurs in fiber CNTs. This relaxation process is connected with decay of the mechanical properties and is due to a decrease in the alignment and compaction of the fibers. Furthermore, the value obtained coincided with the oxidation of the tubes that occurred at 200 • C. This took place because the amount of oxygen present allowed the speed rate to be increased at the lowest temperature, thus causing oxidation of the amorphous carbonate at 200 • C even though the greatest loss occurred above 420 • C.

Conclusions
The paper displays the findings of experimental research on CNTs performed by means of FTIR and Raman investigations. The collected FTIR and Raman spectra were analyzed by means of two innovative procedures: Spectral Distance and Wavelet Cross Correlation. The first method, i.e., the SD approach, makes reference to spectral deviation at the lowest temperature, T = 25 • C. The second method, i.e., the XWT is based on the evaluation of the similarity degree of the spectra, always taking as reference spectrum the lowest temperature, as T = 25 • C. From both these two analyses, a relaxation temperature value emerges, corresponding to the relaxation inflection point, of T = 206 • C for the IR and Raman spectra. More precisely, the two innovative methods, SD and XWT, show not only the presence of a relaxation temperature occurring in the fiber CNTs and which identifies the decay of the mechanical properties, but also the obtained value of 206 • C coincides with the oxidation of the amorphous carbonate.