See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/229362621
Carbon Nanoparticles from Laser Pyrolysis Article in Carbon · December 2002 DOI: 10.1016/S0008-6223(02)00195-1
CITATIONS
READS
87
77
5 authors, including: Nathalie Herlin Boime
Cecile Reynaud
CEA, Atomic and Alternative Energy co…
Atomic Energy and Alternative Energie…
258 PUBLICATIONS 2,392 CITATIONS
204 PUBLICATIONS 3,474 CITATIONS
SEE PROFILE
SEE PROFILE
Jean-Noël Rouzaud Ecole Normale Supérieure de Paris 224 PUBLICATIONS 4,490 CITATIONS SEE PROFILE
All content following this page was uploaded by Cecile Reynaud on 01 April 2015. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document and are linked to publications on ResearchGate, letting you access and read them immediately.
Carbon 40 (2002) 2775–2789
Carbon nanoparticles from laser pyrolysis ´ Aymeric Galvez a,b , Nathalie Herlin-Boime b , *, Cecile Reynaud b , Christian Clinard a , ¨ Rouzaud a Jean-Noel a
` Divisee ´ ( CRMD), CNRS-Universite´ d’ Orleans ´ , 1 B rue de la Ferollerie ´ ´ , 45071 Orleans Cedex 2, Centre de Recherche sur la Matiere France b ´ , DSM-CEA Saclay, Bat 522, Laboratoire Francis Perrin ( URA CEA-CNRS 2453) Service des Photons, Atomes et Molecules 91191 Gif/Yvette Cedex, France Received 27 November 2001; accepted 8 June 2002
Abstract Carbon nanoparticles synthesised by laser pyrolysis of hydrocarbons in a flow reactor have been investigated as a function of laser power. Samples are cross-characterised by high resolution transmission electron microscopy (HRTEM) and infrared (IR) spectroscopy. Nanoparticles appear highly aromatic in character in all the experimental conditions explored here. As the flame temperature in the interaction zone increases, the nanoparticles evolve drastically from poorly organised, highly hydrogenated samples toward turbostratic concentric particles of carbon. The multiscale organisation of the samples and its evolution with the synthesis parameters are quantitatively determined and correlated to IR properties through an original development of HRTEM image analysis. The multiscale organisation does not reduce to the classical view of assemblies of basic structural units (BSU) of aromatic bricks. More refined models are proposed where non-stacked aromatic layers play a noticeable role and lead to a better understanding of the samples optical properties. Possible contribution to a better understanding of carbon cosmic dust is discussed from an astrophysical point of view. 2002 Elsevier Science Ltd. All rights reserved. Keywords: B. Laser irradiation; C. Infrared spectroscopy; Transmission electron microscopy; D. Structure, texture
1. Introduction The multiscale organisation (structure, microtexture, texture) of carbons resulting from pyrolysis of organic precursors at moderate temperature (,1400 8C) is not yet fully understood, despite numerous works carried out from half a century. These carbons are frequently met in nature (for instance terrestrial coals or cosmic dust) or synthesised for industrial applications (blast furnace cokes, adsorbent carbons, . . . ). They are often inadequately described as amorphous or graphitic. A better knowledge is required to understand their properties. Studies based on X-rays techniques and transmission electron microscopy (TEM), have introduced the concept of basic structural units (BSU) [1,2]. Such carbons were described as elemental bricks constituted by the stacking of a few aromatic layers of nanometric size. The BSU size was usually considered as quasi-constant, in the entire *Corresponding author. Fax: 133-1-6908-8707. E-mail address:
[email protected] (N. Herlin-Boime).
carbonisation temperature range. It must be noticed that only BSU are sufficiently organised as stacks of graphene layers to give the 002 diffracted beams used in X-ray diffraction (XRD) or in the dark-field and lattice fringe TEM modes. Consequently, these techniques are blind for a more or less important fraction of possible single layers. Such models appear as relatively simplistic and more realistic ones have to be searched for. With the help of XRD coupled with small angle X-ray scattering, new models were published, as the castle of cards proposed by Dahn et al. [3], where the layers are mainly single below 1000 8C, but tend to be stacked above. This model has however the main disadvantage to involve clearly nonconnected aromatic sheets. Such a carbon would be soluble in heavy solvent and it is not the case. Azuma [4] proposed a model where the layers are described as bent and contain ‘well stacked’ and ‘not well stacked’ parts. Efforts are presently made to quantify the high resolution TEM (HRTEM) images obtained on poorly organised carbons. A methodology for analysis of 002 lattice fringes images and its application to combustion-derived
0008-6223 / 02 / $ – see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S0008-6223( 02 )00195-1
2776
A. Galvez et al. / Carbon 40 (2002) 2775–2789
carbons was proposed by Shim et al. [5]. We developed quantitative analysis to quantify the growth of the aromatic structure during carbonisation stage on poorly-organised carbons [6–8]. To improve the pertinence of these studies, the synthesis of references chars in well-defined conditions is required. Laser pyrolysis of pure and simple hydrocarbons appears as a very convenient method to reach this aim. The infrared laser pyrolysis (IRLP) is a highly versatile method for the production of a wide range of nanopowders [9–16]. This process is inherently clean because homogeneous nucleation occurs in a well-defined reaction zone without interaction with inner walls. The physical and chemical properties of the particles can be adjusted by changing the molecular precursors and the synthesis parameters. The resulting powders are fine (10–100 nm), with a spherical shape, more or less agglomerated and nearly monodispersed in size. Recently, we have shown that IRLP of hydrocarbons leads to the synthesis of hydrogenated aromatic nanoparticles [17]. A first correlation between synthesis conditions, optical properties and organisation was presented: the higher the reaction temperature, the lower the intensity of the infrared bands, and the more organised the carbon materials. These carbon nanoparticles appear as convenient models for carbon cosmic dust because their formation process is based on homogeneous nucleation from hot molecular species, and similar mechanisms are postulated for the formation of carbon cosmic dust [18,19]. Their IR spectra exhibit similarities with the spectra observed in space, the so-called unidentified infrared bands (UIB), but some discrepancies are difficult to understand [17]. Moreover, the changes in the IR spectra with the pyrolysis temperature are not compatible with a structure made of unconnected BSU of constant size. Clearly, more insights into the relation structure–IR properties are needed. In this paper, our main objective is to contribute to the understanding of carbons resulting from pyrolysis of organic precursors at moderate temperature (,1400 8C). Several series of carbon-based nanoparticles have been obtained by IRLP of gaseous hydrocarbons with well controlled experimental conditions, at increasing flame temperature (T f ). The as-formed samples have been characterised mainly by IR spectroscopy and HRTEM, in order to establish a relationship between synthesis parameters, optical properties and multiscale organisation. The quantification of this relationship has been obtained by developing a new analysis method for HRTEM images. At last, a comparison with other carbon nanoparticles is presented and the astrophysical relevance of these synthetic model materials is briefly discussed. 2. Experimental
2.1. Synthesis The IRLP method is based on the interaction of an IR
laser beam with a gaseous precursor which absorbs the laser radiation, leading to the formation of a flame. Reactant gases are heated and decompose, causing aggregates to nucleate and grow rapidly. The small reaction volume and the ability to maintain steep temperature gradients allow to control the nucleation, the growth rate, and the residence time. The experimental set-up has been previously described [9]. The reactant gases flow crosses orthogonally the beam of a continuous wave CO 2 laser emitting at 10.6 mm and delivering a power in the 100–1000 W range. The overlap between the laser beam and the reactant gases flow defines a small zone usually called a wall-less reactor where the reaction is confined by a coaxial argon flow which carries the nanopowders away from the reaction zone. The powders are deposited on a silicon substrate located right in the stream, far enough from the reaction zone in order to avoid interactions with the flame. Several hydrocarbon (C 4 H 6 (butadiene), C 2 H 4 (ethylene) or C 3 H 4 (propadiene)) with an absorption band near 10.6 mm have been tested as precursors. In order to evaluate the best candidate, the threshold for flame apparition was measured as a function of the reactant flow. C 4 H 6 appears to be the best precursor because a reaction occurs even at very low laser power, then ensuring the largest utilisation domain with the largest T f range. Moreover, we have checked that the IR spectra of the samples do not depend significantly from the precursor, much less than the physical parameters such as T f and residence time. Also, IR spectroscopy of the gas mixture coming from the reaction zone shows the formation of molecules such as CH 4 , C 2 H 4 and C 2 H 2 whatever the precursor. In all cases C 2 H 2 is the major gaseous product. These observations indicate that the choice of the precursor is not a crucial parameter and C 4 H 6 will be the only one used in the examples presented hereafter. The main experimental parameters which can be controlled are: the reactor pressure, the laser power and the flow rate. All these parameters influence directly T f . In this paper, the effect of laser power is studied in the range 250–800 W. Total pressure (10 5 Pa) and C 4 H 6 flow rate (136 cm 3 / min) have been kept constant. The typical value of the residence time in the reaction zone is around 10 ms. During each experiment, T f was measured by a vanishing pyrometer working in the range 700–3000 8C. By this method, series of hydrogenated carbon nanoparticles have been produced in quantities (milligram range) sufficient for optical and TEM studies.
2.2. Characterisation IR spectroscopy, optical and electron microscopy have been used to characterise the IRLP samples. Also, the mean thickness of the deposits and eventual irregularities has been measured by profilometry with an accuracy of a few micrometers (Veeco dektak 3030st profilometer). In a few cases, the H / C atomic ratio has been measured by
A. Galvez et al. / Carbon 40 (2002) 2775–2789
elastic recoil detection analysis (ERDA). Some samples have been checked by nuclear magnetic resonance (NMR). Transmission IR spectra were recorded using an FTIR spectrometer (Perkin-Elmer 2000) at a resolution of 4 cm 21 . The absorptivity a (cm 21 ) can be deduced from IR absorbance and thickness measurements. The absorptivity determination allows direct comparison between IR spectra of deposits obtained with different conditions of reaction. The IR spectra are presented here after subtraction of the silicon substrate contribution, but without continuum subtraction. To avoid oxidation of the samples, easily detected by IR spectroscopy, the experimental system and the glove-box used for sample storage, are kept under argon atmosphere. The multiscale organisation, characteristic of carbons [6,20], is studied over more than six orders of magnitude by using successively eye (resolution about 0.1 mm), optical microscopy (resolution about 1 mm), scanning electron microscopy (SEM) which gives morphological information in the 100 mm–10 nm range and TEM (resolution in the lattice fringe mode: 0.14 nm) (Fig. 1). Optical microscopy (OM), performed directly on the deposits in the reflection mode and with natural light, allows a textural characterisation at the micrometric scale. The pseudo-relief of the images obtained by SEM was used to interpret the 2D images given by OM and by TEM. Structure (organisation at the atomic scale) and microtexture (spatial distribution of the aromatic layers) can be directly observed by TEM, due to its high resolution mode allowing to image the profile of the aromatic sheets. The single (non stacked) layers can be detected through an absorption contrast; this contrast is also involved in the imaging of single aromatic planes [21]. The TEM observations have been carried out by using a Philips CM20 microscope operating at 200 kV (resolution 0.144 nm). For TEM observations, a few milligrams of the sample were ultrasonically dispersed in anhydrous ethanol. A drop of this suspension was deposited on a TEM grid covered with an holey amorphous carbon film. Only the isolated particles (,50 nm) lying across the holes were studied in order to avoid the background noise due to electrons scattered by the supporting film and to limit superimpositions due to the TEM image formation.
3. Analysis of the high resolution TEM images A procedure to achieve quantitative analysis of the HRTEM images was developed for this work [8]. It is based on: (1) background noise reducing performed through a filtration of the Fourier Transform (FT) of the raw HRTEM images; (2) thresholding and binarisation of the filtered image; (3) skeletonization; and (4) computerised extraction of quantified structural and microtextural data from the skeletons. These steps are described hereafter. (1) An area of the raw HRTEM image, representative of
2777
the sample, is digitised with a scanner, and stored as a 5123512 pixels image. A graphite sample observed in the same conditions and digitised with the same parameters was used to accurately calibrate the magnification and to obtain a direct correspondence between the pixel scale and the nanometer scale. Then FT was carried out and the diffraction patterns was calculated from the FT modulus. Such patterns contain spots due to the diffraction by pairs of parallel fringes, with a given orientation and, near the central spot, diffusion due to single layers. However, some pairs of fringes in the 2D image does not correspond necessarily to a BSU, some of them could be formed by two parallel layers in the 3D object but placed at two different positions. Such artefacts can be detected when the interfringe spacing is smaller than the graphite interlayer spacing (0.335 nm), or larger than 0.7 nm, i.e. when the Van der Waals forces can be considered as negligible. The more amorphous the carbon, the more intense the central diffusion, whereas BSU with a concentric orientation lead to a well-defined ring. Thus, the filtering was performed by using ring-shaped masks; the dimensions of the mask were evaluated according to the carbon organisation to eliminate the majority of the periodicities without physical sense (as discussed above) as well as the central spot. An inverse Fourier transform was then applied from the filtered modulus and the unmodified phase, and a filtered image is obtained with a strongly reduced background noise and more contrasted fringes. (2) In this filtered image, the brightness of each pixel is represented by 256 intensity levels (grey scale), where zero represents black and 255 represents white. To reduce possible artefacts due to remaining intensity variations, the grey scale image is processed by setting grey level thresholds and a binary image is then obtained thanks to a top-hat mathematical operation. Such operation is not free of risks; for instance, poorly contrasted fringes (planes not exactly under the Bragg angle) could be eliminated. (3) The resulting image is skeletonised until each fringe is 1 pixel width. To avoid artefacts which could be introduced during the operations 2 and 3, filtered images are systematically compared with the raw one. Each individual fringe is considered as an independent object from which an area can be determined and therefore the fringe length. The fringes smaller than 0.25 nm were eliminated, because it is the size of a single aromatic ring. For a graphite crystal, all the carbon atoms are gathered within planar and perfect aromatic layers (graphene layers), and the compactness of the planes is maximum (A–B stacking, leading to the smallest d 002 interlayer spacing, i.e. 0.3354 nm). To specify such evolution, the proportion of aromatic layers in a carbon sample was evaluated from the measurement of the image area covered by fringes. This proportion was calibrated by the value obtained for a graphite sample in the same conditions. This parameter, called fringe density, is consequently equal to 1 for graphite, 0 for amorphous carbon, and increases from 0 to 1 for more and more ordered turbostratic carbons
2778
A. Galvez et al. / Carbon 40 (2002) 2775–2789
Fig. 1. Multiscale organisation of a carbon deposit. Example of a high T f dusty deposit. (a) Binocular image (millimetric scale); (b) optical microscopy image (micrometric scale); (c) scanning electron microscopy image (micrometric scale); (d) low magnification transmission electron microscopy image (submicrometric scale); (e) high resolution transmission electron microscopy image (nanometric scale); (f) selected area electron diffraction pattern of the particle e.
A. Galvez et al. / Carbon 40 (2002) 2775–2789
(4) From such skeletons, we developed a computerised interface in order to obtain structural and microtextural data. The length of each individual fringe was measured, and an average fringe length L is obtained. L corresponds to a mean aromatic layer size. The amount of single fringes was also determined. To specify the presence of BSU (i.e. stacked layers), sets of parallel layers were defined according to adjustable parameters as the tolerance angle of misorientation (here chosen as6108), the tortuosity of the fringe (here chosen as 6108) and the maximum interfringe spacing (here chosen as 0.5 nm). When an as defined BSU is detected, its interfringe spacing, d 002 , its height, Lc , and its length, La , were measured. Then, the corresponding mean data were calculated for all the BSU. As far as the microtexture is concerned, the angle of each individual fringe was determined vs. an orientation arbitrarily chosen as reference. We chose to determine the distribution of the angles on a given area representing a quarter of spherule. From these data, histograms of orientation can be built and the microtextural improvement from randomly oriented layers toward a concentric orientation can be followed.
2779
obtained. This shows the good stability of the experimental system and allows the determination of the growth rate of the samples. T f and the growth rate of the deposits are reported on Fig. 2 as a function of the laser power. A flame appears when the laser power is higher than 300 W and T f increases with the laser power. As shown in Fig. 2, it is easy to obtain carbon samples in the T f range 800– 1400 8C. Profilometry measurements have been possible up to 550 W, before the deposits become too dusty. Naked eye observations give information on the aspect of the sample, showing a noticeable evolution according to T f . The low T f samples are very bright and smooth, whereas the highest T f samples appear to be composed of fine dust which sticks to the substrate. The low T f IRLP carbon samples are hydrogenated: for T f of about 1000 8C, (H / C) at obtained by ERDA is about 0.7; this ratio decreases at higher T f to reach 0.3 at about 1250 8C. NMR measurements reveal that the samples are always strongly aromatic in character, whatever T f . Only a very small shoulder due to sp 3 carbon can be detected in the spectra. The different samples used in the following IR and HRTEM analysis are referenced in Table 1.
4. Results
4.2. IR properties of the deposits 4.1. General properties of the IRLP nanoparticles Samples have been obtained with deposition times between 15 s and 5 min depending on the growth rate. At the macroscopic scale, the surface of the deposits had two types of morphology: smooth films or dusty deposits. In the case of dusty deposits, profilometry measurements are difficult or even not possible. When possible, the thickness and weight of the deposits have been measured as a function of the deposition time, and a linear dependence is
The IR spectra of IRLP samples synthesised at different T f ’s are presented in Fig. 3. The spectra are composed of large bands superimposed on a continuum decreasing towards longer wavelength. The continuum increases as T f increases. All the bands can be assigned to CH and CC bonds, at positions well established for hydrogenated carbon compounds (see for example Refs. [22,23]). The main bands can be observed into three groups: A first group, in the range 2700–3400 cm 21 , corre-
Fig. 2. Variation with the laser power of T f (open circles s, right-hand scale) and deposition rate (full squares j, left-hand scale) at constant gas flow (136 cc / min) and pressure (10 5 Pa).
A. Galvez et al. / Carbon 40 (2002) 2775–2789
2780
Table 1 Operating power of the cw CO 2 laser during the synthesis of the samples used in Figs. 1–10 Sample
Laser power (W)
Flame temperature (8C)
A1 A2 A3 A4 A5 A6 A7 A8
285 320 355 390 425 500 605 780
820 a 900–950 1000–1050 1050–1100 1095–1145 1175–1225 1250–1300 1350–1400
a
Extrapolated value (Fig. 2).
sponds to the C–H stretching vibrations. The band at 3049 21 cm attributed to aromatic compounds is always the most 21 intense, the C–H stretching modes (2967 and 2915 cm ) of aliphatic compounds are much weaker. 21 A second group of bands, between 1000 and 2000 cm , forms a composite massif due essentially to the in-plane vibrational modes of aromatic CC and CH bonds. The 21 most emergent bands at 1600 and 1440 cm are due to aromatic CC stretching, symmetric and antisymmetric ring modes, respectively. Both are infrared-allowed here, due to a high level of non-symmetric substitution in the aromatic units as expected from an irregular network of aromatic units. The in-plane bending CH modes are responsible for 21 the bump at 1160 cm . With increasing T f , the IR 21 intensity at 1300 cm , very low initially, increases compared to other bands of the massif. A third group of bands is found, between 600 and 1000 21 cm , characteristic of the aromatic C-H out-of-plane bending modes. The large intensity of these bands shows
the strong aromatic character of the deposits. These modes are very sensitive to the ring substitution and give rise to three main components according to the number of adjacent hydrogen atoms on a ring. The band at 752 cm 21 corresponds to three and four adjacent H, the feature exhibiting two maxima at 839 cm 21 and 815 cm 21 is due to two adjacent H, and the band at 885 cm 21 is attributed to lone H. The corresponding summation bands can be found in the 2000–1650 cm 21 region Finally, minor bands can be observed in the spectra, due to alkyne groups at 3290 and 2103 cm 21 , and to mono and di-substituted aromatic rings as well as olefin groups in the 1200–400 cm 21 region. These minor bands decrease noticeably when T f increases. For the low T f samples, up to 360 W (about 1050 8C), absolute IR intensity can be measured. For these samples, an attempt to quantify the IR spectra evolution with T f has been made through the variations of the absolute IR intensity taken at selected wavelengths after background subtraction (Fig. 4). This simple procedure has been preferred to a complete deconvolution, which, due to the number of superimposed bands, does not give an univocal result. The 3049 and 2927 cm 21 positions have been chosen because they give information on the amount of hydrogen (aromatic and aliphatic) in the deposit. Also, the aromatic to aliphatic ratio in the IRLP samples can be estimated from the ratio between these peaks. The 1300 cm 21 wavelength has been chosen because it is attributed to the aromatic network extent [24]. Fig. 4 also presents the evolution of the ratio of the intensities at 885 and 752 cm 21 . This ratio can be measured in a more extended range than absolute intensities and is reported in Fig. 4 up to 550 W of laser power. This ratio allows to quantify the increase of the contribution of lone H as compared to the others when T f increases and gives indications on the
Fig. 3. IR spectra of as-formed nanoparticles as a function of T f for samples A1, A3 and A7, respectively.
A. Galvez et al. / Carbon 40 (2002) 2775–2789
2781
Fig. 4. Evolution with nominal laser power of the intensity of IR bands at 3049 cm 21 (open squares h), 2927 cm 21 (open circles s) and 1300 cm 21 (open triangles ^) and ratio 885 / 752 cm 21 (full circles d).
substitution degree of the aromatic units, and therefore on their size. The observed decreasing intensity of the aromatic and aliphatic C–H bands with laser power is in good agreement with the decrease in H content measured by ERDA. No significant evolution of the aromatic to aliphatic ratio is observed in these samples. The intensity of the band at 1300 cm 21 slightly increases, which seems to indicate an extension of the carbon network. The 885 cm 21 / 752 cm 21 intensity ratio increases with T f which corresponds also to aromatic units of increasing size. These indications will be compared to TEM measurements in the discussion.
4.3. Multiscale organisation of the nanoparticles To describe the texture, i.e. the deposit organisation at the micrometric scale, we coupled OM and SEM images (Fig. 5) with low magnification bright field TEM images (Fig. 6) for a series of samples obtained at different T f ’s. At this scale, all the samples are texturally homogeneous. The deposits synthesised at low T f show a smooth surface, without irregularities and grains (Fig. 5a and d) confirmed by the profilometry measurements. For a high T f synthesis, the surface of the deposit is formed of small agglomerates (,3 mm), homogeneous in size (Fig. 5c). For the intermediate T f ’s, more or less extended surface irregularities are observed (Fig. 5b). The coarse mosaic texture observed by OM corresponds to a ‘cauliflower texture’ imaged by SEM and due to agglomerates, a few mm to about 10 mm in size (Fig. 5b and e). For higher T f ’s, the cauliflower texture vanishes and a granular texture develops (Fig. 5c and f); this corresponds to the transition toward micrometric aggregates made of nanometric round-shaped particles. The TEM resolution (about 1 nm) allows to image the microtexture (Fig. 6). At low T f , the continuous and
homogeneous carbon deposit observed by OM and SEM (Fig. 5a and d) remain similar at the TEM scale (Fig. 6a): large (.1 mm) and homogeneous fields are observed. When T f increases, aggregates of more and more clearly individualised round particles are detected (Fig. 6b). At the highest T f (1350–1400 8C), nanometric particles are well individualised and spherical (Fig. 6c); their aggregation, in a chain-like manner, is responsible for the occurrence of the fine granular texture imaged by OM (Fig. 5c). While the extent of the agglomerates decreases with T f , the diameter of the spherical particles does not significantly changes. TEM images show that most of the particles have a diameter around 50 nm. In Fig. 7, are presented the raw and skeletonised HRTEM images of the samples. From the quantitative data extracted from the numerised images, the proportion of single layers, the number of layers by stacks, the size distribution and the mutual orientation of the aromatic layers can be drawn (Figs. 8–10). At the lowest T f , the deposit is made of short, randomly oriented, single aromatic layers. L, the average size is around 0.5 nm (Fig. 8) whereas the fringe density is less than 0.2 (Fig. 9). The proportion of single layers represents more than 65% of the fringes (Fig. 9). Only few stacks of layers can be detected clearly as BSU. Such BSU are very small (La [0.3 nm, Lc [0.6 nm) and the mean interlayer spacing is very large (d 002 [0.5 nm) (Fig. 8). This amorphous-like nature is strengthened by the electron diffraction mode: only broad, diffuse and faint 10. and 11. bands are visible; no 002 reflections, due to layers stacking, are detected. With increasing T f (up to 1400 8C), a pure polyaromatic carbon with a concentric microtexture forming spherules (Fig. 7b–d) develops. The mean fringe size increases up to 0.9 nm (Fig. 8) and the fringe density up to 0.5 (Fig. 9).
2782
A. Galvez et al. / Carbon 40 (2002) 2775–2789
Fig. 5. Textural evolution of the deposit as a function of T f as seen by optical microscopy (a, b, c) and SEM (d, e, f) for sample A2 (a and d), sample A4 (b and e) and sample A8 (c and f), respectively.
The number of fringes larger than 0.5 nm increases considerably and the maximal size shifts from 1.5 to 4.75 nm (Fig. 10). The proportion of single layers significantly decreases to about 10% (Fig. 9) and most of the planes are now stacked to form clearly visible BSU. The structural parameters increase, although the reached averaged values of La and Lc remain very small (respectively 0.7 and 0.5 nm) as presented in the Fig. 8. The ‘mean BSU’ of such carbons can thus be described as stacks of two (sometimes 3, up to 5) polyaromatic structures made of a few aromatic rings; d 002 is still very large (0.45 nm), i.e. far from the graphite value (0.3354 nm). The structure of the spherules is turbostratic, as demonstrated by the electron diffraction pattern (Fig. 1f): presence of a 002 ring and hk bands, without hkl reflections (as 11.2), which are characteristic of the triperiodic order [2,25]. The concentric orientation
of the aromatic layers progressively improves (Fig. 10) since the angles distribution tends to 908 and is responsible for the individualisation of the carbon spherules. These particles tend to stick together to form more or less compact clusters. It must be noticed that neighbouring spherules can be strongly connected, as shown by the existence of common aromatic layers.
4.4. FTIR–HRTEM correlation From TEM and FTIR results, a strong evolution of the materials with increasing T f is observed, which can be attributed to the growth and / or crosslinking of polyaromatic layers. Such evolution was directly imaged by HRTEM and quantified by image analysis. As hydrogen atoms are grafted on the boundaries of the aromatic
A. Galvez et al. / Carbon 40 (2002) 2775–2789
2783
to a distribution of IR intensity ratio. This fact can be understood regarding the wide range of experimental conditions explored here in terms of T f and residence time.
5. Discussion
5.1. Structural models
Fig. 6. Morphology of carbon particles (low magnification TEM images) as a function of T f for sample A2 (a), A5 (b) and A8 (c), respectively.
structures, both growth and cross-linking are accompanied by a dehydrogenation, as shown by the decrease of the H / C atomic ratio, detected by ERDA elemental analysis. FTIR data presented in Fig. 4 and discussed above also indicate that dehydrogenation increases with increasing T f . To verify the coherence of these data, the mean size of the fringes, L, was plotted vs. the intensity ratio of the IR bands 752 cm 21 / 885 cm 21 (Fig. 11). In this figure are plotted the data discussed in the present paper together with data obtained at increasing residence time but constant T f and reported in another paper [26]. A linear relationship between TEM and FTIR data can be clearly deduced from this figure. A certain dispersion around the linear fit is observed, the same mean fringe length leading
The carbon nanopowders synthesised by laser pyrolysis of small hydrocarbons molecules appear to form a family of highly aromatic compounds. The presence of clearly identified polyaromatic layers, even randomly oriented, does not permit to define any of them as stricto sensu amorphous. Their molecular organisation evolves continuously as the pyrolysis T f increases in the studied range from 900 to 1400 8C, from poorly organised, highly hydrogenated samples toward turbostratic concentric particles of carbon. However, the HRTEM image analysis developed here gives a view different from the classical models of BSU [1,2,25] where disordered carbons are described as assemblies of well-defined elemental bricks of well-stacked layers assumed to be poorly connected pairs of coronene. Indeed, the new HRTEM image analysis allows to specify the large amount of single (non-stacked) layers, their usual tortuosity, the weak number of stacked layers, the usual shift of the neighbouring stacked layers and the large interlayer spacing. In fact, a large proportion of layers must be only partially stacked, as clearly shown in Fig. 8 where L, the mean fringe length, is always much greater than La , the BSU average length. Also, L increases more rapidly than La with T f . La evolves from 0.3 up to less than 0.5 nm, while L jumps from 0.5 up to 0.9 nm. This quasi invariance of the BSU size is in agreement with the previous BSU description. Interestingly, this BSU size appears to be very close to the most probable layer length in our samples, as observed in Fig. 10. This equivalence allows to establish a coherent correlation between previous BSU description and the present data. The strong difference in the most probable fringe length and the mean fringe length is a consequence of the very extended length distribution revealed by the up-to-date HRTEM image analysis. It explains also the persistence of the 752 cm 21 band in the infrared spectra in all our samples. Finally, both the high interlayer spacing (Fig. 8) and the low fringe density compared to graphite (Fig. 9) indicate that the IRLP samples are far from being graphitised material. As the samples are solid carbon particles insoluble in various solvents (acetone, benzene, . . . ), the aromatic structure cannot be a physical aggregation of polyaromatic hydrocarbons, but does form a cross-linked network. Such network is responsible for the multiscale organisation of the nanoparticles. Due to the synthesis and handling conditions, the only possible heteroatom is hydrogen. As shown by the IR study, H atoms are either grafted on
2784
A. Galvez et al. / Carbon 40 (2002) 2775–2789
Fig. 7. Analysis of high resolution TEM images of carbon particles as a function of T f : raw images (on the left) and skeletonised images (on the right). (a) sample A2; (b) sample A5; (c) sample A6; (d) sample A8.
aromatic layers boundaries or included in other molecular groups. These later can be grafted on the aromatic layers boundaries or form cross-links between aromatic units. As
shown by NMR and FTIR analysis, aliphatic groups are present in very small quantities and the samples are qualified as quasi pure sp 2 compounds. Therefore, for the
A. Galvez et al. / Carbon 40 (2002) 2775–2789
2785
Fig. 8. Evolution of structural data extracted from skeletonised HRTEM images as a function of T f : L, average length of the fringes (open circles s); La , average length of the BSU (full squares j), Lc , average height of the BSU (full triangles m), and d 002 , mean interlayer spacing (stars *).
less organised samples, the zones exempt of fringes have to be viewed as mainly made of olefin groups. To describe the samples in terms of number of carbon atoms (NC ) and aromatic rings (NR ), it is possible to use the relationships existing between these numbers and the size (D) in the case of isometric aromatic layers NC 5 3 / 2(1 1 D/d)2 NR 5 (D/ 2a)2 where a is the CC distance in an aromatic ring (a50.142 nm) and d the size of one ring (d5œ3a50.246 nm). D is expressed in nm. The second relation is exact only when D 4 a, but is nevertheless a good approximation. Of
course, these relationships do not hold exactly for nonisometric forms and we have to keep in mind that they give only mean values and that the aromatic units can take complex forms, more or less elongated or jagged-edged. Such forms are in fact strongly suggested by the IR spectra, since the intensity of the 752 cm 21 band remains large even at high T f . This band cannot be found in isometric aromatic layers, such as coronene since it corresponds to 3 or 4 adjacent H on a ring. Thus, at low T f , carbon nanoparticles are mainly made of small aromatic units with in average three rings (around 14 C atoms). More than 95% of the layers are formed by less than 12 rings (40 C atoms), and the largest layer is made of 28 rings. These layers are frequently single (more than 65%) or stacked by two. They are usually connected,
Fig. 9. Fringe density (full squares j, left-hand scale) and percentage of single layers (open circles s, right-hand scale) from skeletonised HRTEM images as a function of T f .
2786
A. Galvez et al. / Carbon 40 (2002) 2775–2789
Fig. 10. Histogram of the fringe length and orientation (the distribution was determined on an area representing a quarter of spherule) from skeletonised HRTEM images for samples A2 (a), A5 (b) and A8 (c).
but maintained in a complete random orientation by bridges made of sp 2 carbon atoms and possible small aliphatic groups; such random orientation of the layers is responsible for the isometric microtexture and texture. High T f carbon nanoparticles are formed by much larger units (32 C atoms and 10 rings in average). Around 30% of the units contain more than 40 C atoms (12 rings), and for a few percents of the units, the number of C atoms is even as large as 125 and up to 550, which corresponds to 50 up to 250 rings. As the fringe density increases, much less atoms are needed as bridges between the aromatic units. Stacking is now possible and is energetically favoured (only 10% of single layers). The stacks are constituted of two to five piled-up layers, but some layers are only partially stacked. These partial stacks correspond to the BSU [1,2]. The layers are no longer randomly oriented and a concentric microtexture develops.
5.2. Comparison with other carbons To our knowledge, few studies [15,16] are dealing with hydrogenated carbon nanoparticles obtained by laser pyrolysis. In the study of Xian-Xin Bi, nanoparticles are
obtained from an ethylene–benzene mixture with a small percentage of Fe(CO) 5 . The experimental conditions were 300 W cw laser power, pressure 0.5 10 5 Pa, nozzle diameter 6 or 9 mm and C 2 H 4 flow rate 300 cm 3 / min. The T f was not measured but was estimated to 1000–1200 K. These experimental conditions appear close to ours, the main differences are, the residence time in the reaction zone (four to nine times higher in their case) and the precursor. The as-formed particles, imaged by TEM [16], are agglomerated, nearly amorphous, spherical with an average diameter around 30 nm. HRTEM shows the formation of small graphene layers near the surface of a few particles. The spacing between graphene layers is in the range 0.38–0.42 nm, smaller than the d 002 measured in our study (d 002 .0.45 nm). This difference can be explained by the different measurement techniques, because, in our case, the mean value of interlayer spacing is obtained over the whole spherule (the interlayer spacings are known to be larger in the inner part than in the external part [28]). The H / C ratio is about 0.1 and slightly decreases when the residence time in the flame increases. Thus, the nanoparticles obtained by Xian-Xin Bi et al. [16] seem similar to ours obtained at high T f . This is because the residence time in their case is longer, a fact which compensates the lower T f . Nanoparticles have also been produced with cw or pulsed CO 2 laser from a SF 6 / C 2 H 2 mixture [15]. Again, HRTEM data shows a structural similarity with our particles which confirms that the precursor is not fundamental. With the pulsed laser, particles are amorphous, while some stacks of BSU are detected in the particles obtained with the cw laser. Evolution of IR spectra of powders obtained in pulsed conditions as a function of SF 6 / C 2 H 2 ratio (i.e. T f ) has been reported [15]. The IR spectra were obtained using the KBr pellet method, and show evidence of particles oxidation. Similarities with our deposits exist as shown by the presence of the characteristic aromatic bands. The relative intensity of the 885 cm 21 band relative to the 752 cm 21 band is in their case higher than in our case. This seems to indicate a lower content of hydrogen in their case. It can be due to a higher T f due to the presence of SF 6 in the reactants mixture, which strongly absorbs the laser. The main discrepancy comes from the intensity of the aliphatic signature (2927 cm 21 ) always higher than the aromatic one (3030 cm 21 ) in their study, whereas it is the reverse in our case. This difference could be due to different handling of the samples. IRLP carbon nanoparticles appear particularly similar to others carbon nanoparticles grown also in the gas phase, without substrate, but in very different devices: shock tube [29], electrical arc [30,31]. This suggests similar growth mechanisms. IRLP deposits obtained at low T f are close to those of sputtered thin carbon films which are classical references of amorphous carbons [32,33]. Such carbons can be described as subnanometric single aromatic layers more or less connected to each other by numerous defects
A. Galvez et al. / Carbon 40 (2002) 2775–2789
2787
Fig. 11. Correlation between the 885 / 752 cm 21 ratio deduced from FTIR spectra and L, the average size of the fringes. Linear fit of the experimental points (full line) and its extrapolation (dotted line) are also plotted.
preventing layers stacking. However, the IR spectra indicate here a high level of hydrogenation of the aromatic units which is not the case of the sputtered thin carbon films. At the highest T f , the IRLP samples develop the characteristic onion-like microtexture of carbon blacks [27], classically obtained by incomplete combustion of hydrocarbons. However, IRLP samples studied here never develop BSU as large as the BSU of nanometric size, frequently described in synthetic anthracene-based cokes [34] and in natural anthracites [35]. In particular, the mean interlayer spacing is in these cases usually smaller than 0.38 nm, whereas it remains higher than 0.45 in the IRLP samples.
5.3. Astrophysical application The FTIR spectra of the IRLP nanoparticles were found interesting from an astrophysical point of view [17] because IR bands attributed to aromatic carbon species are observed by astronomers in an ubiquitous manner. However, the precise attribution of these unidentified IR bands (UIB) is still controversial. Without developing all the arguments here (see Reynaud et al. [26] and references therein), let us say that the main difficulty comes from the fact that the UIB carriers must exhibit the good spectrum but contain less that around 100–200 carbon atoms [36]. An encouraging point in the IR spectra of the IRLP particles is that most of the UIB are present [17], and that their aromatic character is more pronounced than the one developed by other earth analogues such as hydrogenated
amorphous carbon (HAC) or quenched carbonaceous composites (QCC) [37–39] which show a too strong aliphatic character. However, strong discrepancies with UIB remain in the spectra of IRLP particles which are the too high intensity of two bands (752 and 1450 cm 21 ), and the too low intensity of another band (1300 cm 21 ). These discrepancies were attributed to a too high hydrogenated level in the IRLP samples, and correlated to a too small size of the polyaromatic units [17]. The coupling between HRTEM and IR data obtained in this work (see Fig. 11) gives quantitative data which could be useful to determine the minimum size of the polyaromatic units for improving the agreement with UIB. The extrapolation of the linear fit of the data of Fig. 11, up to the limit case where the 752 cm 21 band disappears, corresponds to a mean polyaromatic layer of 2.2 nm, thus involving at least 60 rings, and around 150 carbon atoms. Interestingly such size is in the range admitted for the UIB carriers. However, due to the very extended size distribution of the IRLP nanoparticles towards the long-size edge as well as the cross link between the aromatic units, the true number of carbon atoms could be much higher, as high as several hundreds. The development of extended layers is responsible for the strong increase of the continuum in the IR spectrum of the high T f sample (see Fig. 3), and increasing T f to improve the IR spectra is not relevant. Therefore, alternative solutions have to be found. One strategy could be to synthesise high T f nanoparticles, with onion-like structure and weakly hydrogenated, but with much lower diameter than in the present experiments
2788
A. Galvez et al. / Carbon 40 (2002) 2775–2789
(decreasing from 50 to less than 5 nm). Such very small sizes are expected to limit the growth of extended layers and can be obtained in a pyrolysis reactor when particles are extracted perpendicularly to the gas flow [40].
[3] [4] [5] [6] [7]
6. Conclusion Relationships between synthesis conditions, optical and multiscale properties (structure, microtexture and texture) of carbon nanoparticles, synthesised by gas phase laser pyrolysis of small hydrocarbons, have been established owing to a new HRTEM image analysis procedure. Measurement of the fringe lengths (i.e. the aromatic layers extent) and of their mutual orientation (from random to concentric orientation) were performed and mean structural models have been proposed. One of the most striking results is the presence of non-stacked (single layers) or partially stacked layers in the carbon nanoparticles. The most probable layer length has been found to be systematically shorter than the average layer length, and to correspond nearly to the BSU size (length of the coherent domain made of stacked layers). This observation could explain the origin of the concept of BSU assemblies to describe the low-temperature carbons, in spite of the more complicated structure revealed here. The structural transition, due to T f increase, from an amorphous-like carbon, i.e. free of large aromatic structures, to a polyaromatic turbostratic carbon was quantitatively specified. The evolution of the amount of single layers, the growth of the aromatic layers, their progressive stacking and their reorientation to give microtexturally concentric spheres were determined. A relationship between the mean layer size and the intensity ratio of the 885 cm 21 on the 752 cm 21 IR bands was found. Such relationship may enable to use optical properties of carbon nanoparticles to specify some of their structural data, and to give indications for the synthesis of carbon (nano)particles as earth analogues of carbon cosmic dust.
[8] [9] [10] [11] [12] [13] [14] [15] [16]
[17]
[18] [19] [20]
[21]
[22]
Acknowledgements Support from the French National Program ‘Physique et chimie du milieu interstellaire’ of CNRS-CNES-CEA is gratefully acknowledged. The authors are very thankful to Aline Brunet-Bruneau for ERDA measurements.
[23]
[24] [25]
[26]
References [1] Franklin RE. Cristallite size in graphitizing and nongraphitizing carbons. Proc R Soc 1950;209:196–218. [2] Oberlin A. In: Thrower PA, editor, Chemistry and Physics of Carbon, vol. 22, New York: Marcel Dekker, 1989, pp. 1–143.
[27] [28]
Dahn JR, Xing W, Gao Y. Carbon 1997;35:825–30. Azuma H. J Appl Cryst 1998;31:910–6. Shim HS, Hurt RH, Yang NYC. Carbon 2000;38:29–45. ´ ´ Galvez A. PhD Thesis. Orleans: University of Orleans, 22 octobre 1999. Clinard C, Rouzaud JN, Pellenq RJM. Analyse d’image de ´ haute resolution en Microscopie Electronique par Transmis´ In: sion. Application au cas de carbones microtextures. ´ ` en Genie ´ ´ ´ Visualisation-ImageRecents Progres des Procedes, ´ ` ‘VIM 2001’, Nancy 17–20 Mai 2001, Modelisation, Congres vol. 15, 2001(78), pp. 29–34. Rouzaud JN, Clinard C. Fuel Proc Technol 2002;77:229–35. Cauchetier M, Croix O, Luce M. Adv Ceram Mater 1988;3:548–52. Cauchetier M, Croix O, Herlin N, Luce M. J Am Ceram Soc 1994;77:993–8. Ehbrecht M, Faerber M, Rohmund F, Smirnov V, Stelmach O, Huisken F. Chem Phys Lett 1993;214:34–8. Herlin N, Armand X, Musset E, Martinengo H, Luce M, Cauchetier M. J Eur Ceram Soc 1996;16:1063–73. Boulanger L, Andriot B, Cauchetier M, Willaime F. Chem Phys Lett 1995;234:227–32. Voicu I, Armand X, Cauchetier M, Herlin N, Bourcier S. Chem Phys Lett 1996;256:261–8. Schnaiter M, Henning Th, Mutschke H, Kohn B, Ehbrecht M, Huisken F. Appl J 1999;519:687–96. Xiang-Xin Bi, Jagtoyen M, Endo M, Das Chowdury K, Ochoa R, Derbyshire FJ, Dresselhaus MS, Eklund PC. J Mater Res 1995;10(11):2875–84. Herlin N, Bohn I, Reynaud C, Cauchetier M, Galvez A, Rouzaud JN. Astronomy and Astrophysics 1998;330:1127– 35. Frenklach M, Feigelson E. Appl J 1989;341:372–84. Cherchneff I. Astrophysics and space. Science 1995;224:379–82. Rouzaud JN, Duval B, Leroy L. In: Lahaye J, Ehrburger P, editors, Fundamental issues in control of carbon gasification reactivity, Dordrecht: Kluwer Academic Publishers, 1991, pp. 257–68. Rouzaud JN, Duber S, Clinard C, Pusz S. Optical properties of porous carbons. Berlin: Extended Abstracts of the International Conference. Carbon 2000, 9–14 juillet 2000, pp. 187–188. Dischler B, Bubenzer A, Koidl P. Solid State Commun 1983;48:105–8. Lin-Vien D, Colthup NB, Fateley WG, Grasselli JG. The handbook of infrared and Raman characteristic frequencies. New York: Academic Press, 1991. Balanzat E, Bouffard S, Bouquerel A, Devy J, Gate C. Nucl Instr Methods Phys Res B 1996;116:159–63. ´ Oberlin A, Goma J, Rouzaud JN. Techniques d’etude des ´ ´ structures et textures (microtextures) des materiaux carbones. J Chim Phys 1984;81:701–10. Reynaud C, Guillois O, Herlin-Boime N, Rouzaud JN, Galvez A, Clinard C, Balanzat E, Ramillon JM. Spectrochim Acta A 2001;57(4):797–814. Heidenreich RD, Hess WH, Ban LL. J Appl Cryst 1968;1:1. Douce F, Djebaili-Chaumex N, Paillard C, Clinard C, Rouzaud JN. Berlin: Extended Abstracts of the International Conference on Carbon 2000, 9–14 juillet 2000, pp. 1063– 1064.
A. Galvez et al. / Carbon 40 (2002) 2775–2789 [29] Douce F, Djebalai-Chaumeix N, Paillard CE, Clinard C, Rouzaud JN. Proceedings of the 4th International Conference on Internal Combustion Engine. Capri: Experiments and Modelling, ICE99, 1999, 12–16 September 1999;251. ´ [30] Clinard C, Rouzaud JN, Delpeux S, Beguin F, Conard J. J Phys Chem Solids 1994;55:651–7. [31] Rouzaud JN, Galvez A, Clinard C, Beny JM. Extended Abstracts. Eurocarbon 98, Strasbourg 5–9 juillet 1998;:763– 4. [32] Mclintock IS, Orr JC. In: Walker PL and Thrower PA (editors), Chemistry and physics of carbon, vol. 11, New York: Marcel Dekker, 1973:243. [33] Goma J, Oberlin A. Thin Solid Films 1980;65:221–32. [34] Rouzaud JN, Oberlin A. Carbon 1989;27:517–29.
View publication stats
2789
[35] Rouzaud JN, Oberlin A. Chapter 17. In: Charcosset H, Nickel-Pepin-Donat B, editors, Advanced methodologies in coal characterization, Amsterdam: Elsevier, 1990. [36] Boulanger F, ASP Conference Series 132, Star formation with the infrared space observatory, Yun JL and Liseau R (editors), San Francisco: 1998; ASP 15. [37] Colangeli L, Mennella V, Palumbo P, Rotundi A, Bussoletti E. Astronomy and Astrophysics Supplement Series 1995;113:561–77. [38] Sakata A, Wada S, Onaka T, Tokunaga A. Appl J 1987;320:63–7. [39] Scott A, Duley WW. Ap J 1996;472:L123–5. [40] Ehbrecht M, Ferkel H, Smirnov V, Stelmakh O, Zhang W, Huisken F. Rev Sci Instrum 1995;66(7):3833–7.
