PCCP PAPER. Microwave plasma enabled synthesis of free standing carbon nanostructures at atmospheric pressure conditions. 1.

Transcription

1 PAPER Cite this: Phys. Chem. Chem. Phys., 2018, 20, Microwave plasma enabled synthesis of free standing carbon nanostructures at atmospheric pressure conditions N. Bundaleska, * a D. Tsyganov, a A. Dias, a E. Felizardo, F. M. Dias, a M. Abrashev, b J. Kissovski b a and E. Tatarova a J. Henriques, a Received 23rd March 2018, Accepted 26th April 2018 DOI: /c8cp01896k rsc.li/pccp An experimental and theoretical study on microwave (2.45 GHz) plasma enabled assembly of carbon nanostructures, such as multilayer graphene sheets and nanoparticles, was performed. The carbon nanostructures were fabricated at different Ar CH 4 gas mixture composition and flows at atmospheric pressure conditions. The synthesis method is based on decomposition of the carbon-containing precursor (CH 4 ) in the hot microwave plasma environment into carbon atoms and molecules, which are further converted into solid carbon nuclei in the colder plasma zones. By tailoring of the plasma environment, a controlled synthesis of graphene sheets and diamond-like nanoparticles was achieved. Selective synthesis of graphene flakes was achieved at a microwave power of 1 kw, Ar and methane flow rates of 600 sccm and 2 sccm respectively, while the predominant synthesis of diamond-like nanoparticles was obtained at the same power, but with higher flow rates, i.e and 7.5 sccm, respectively. Optical emission spectroscopy was applied to detect the plasma emission related to carbon species from the hot plasma zone and to determine the main plasma parameters. Raman spectroscopy and scanning electron microscopy have been applied to characterize the synthesized nanostructures. A previously developed theoretical model was further updated and employed to understand the mechanism of CH 4 decomposition and formation of the main building units, i.e. CandC 2, of the carbon nanostructures. An insight into the physical chemistry of carbon nanostructure formation in a high energy density microwave plasma environment is presented. 1. Introduction Carbon based materials are of great interest for academia and industry owing to their abundance, stability and relative environmental friendliness. Recently a renewed interest has been generated as new advanced carbon nanostructures are being introduced, bringing new prospects for applications. One of the most recent additions of great technological interest is graphene, a few layers thick variant of graphite. Characterized by its hexagonal structure, where each carbon atom is bonded to three others, graphene can be idealized as an infinite and perfect two-dimensional molecule. 1 Much research has been conducted ever since on the synthesis and engineering of graphene as well as on its structure and fundamental properties The extraordinary properties of graphene give it an enormous potential for application. Free-standing graphene sheets have been used as batteries, fuel cells, solar cells and supercapacitors elements, among others, by many research groups. 6,11 13 Furthermore, a Instituto de Plasmas e Fusão Nuclear, Instituto Superior Técnico, Universidade de Lisboa, Lisbon, Portugal. neli.bundaleska@ist.utl.pt b Faculty of Physics, University of Sofia, 1164, Sofia, Bulgaria free-standing carbon nanoparticles have attracted much research interest due to their unique chemical and physical properties Nano or dust particles have been observed in fusion plasmas, processing plasmas used for thin film deposition, etc. Recently they have elicited interest for a wide range of applications such as use as building units for nano-assemblies, the synthesis of novel nanocomposite materials, drug delivery etc. Multiple processes have been reported for carbon nanostructure production, corresponding to so-called top-down or bottom-up approaches. Mechanical exfoliation of natural or synthetic graphite into a mixture of single and few layer graphene platelets, for example, is a representative example of a top-down approach. Thermal exfoliation of graphite oxide and wet chemistry reduction techniques that employ graphite oxide as a precursor and a reducing agent such as hydrazine or urea and additives to eliminate oxygen groups are widely used to produce graphene sheets. 19 Plasma etching technology is another well-established top-down method in the production of different nanostructures, such as 3D cone-like structures and nanodiamonds. 2 Bottom-up approaches are based on the direct growth of nanostructures from carbon precursors such as hydrocarbons, Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018

2 alcohols etc. These approaches include epitaxial growth, chemical vapor deposition (CVD), vacuum graphitization of silicon carbide substrates, etc. Acommondrawbackofthesemethodsisasignificant lack of controllability in the shape, size, and the structure synthesized. Note that one of the main challenges of conventional, i.e., chemical routes, is the very limited, or lack of, control of the assembly process. 2 6 Clearly, beyond the mainstream of widely used chemical methods, plasma methods possess many distinctive features that place them uniquely and very competitively in the carbon nanostructure synthesis domain. Plasma systems comprise thermal and chemical reactor functions, as well as catalytic properties. Plasma assisted growth of nanostructures can be achieved without using catalysts due to plasma s unique ability to activate the surface, thus creating favorable conditions for nucleation and growth processes. The main advantage is the very high and extremely controllable energy density in the plasma reactor, which allows effective control over the energy and material fluxes for growing nanostructures, via proper reactor design and tailoring of the plasma environments in a synergistic way. 6 Furthermore, plasma methods are very versatile, with different types of nanostructures being synthesized using one and the same plasma reactor. The approaches used to generate free standing materials are not usually focused on plasma methods. Apart from the creation of islands and clusters by sputtering techniques, plasma-based methods have found a niche application in thin film technology, opening ways to other techniques as the ones based on colloids, chemical exfoliations, etc. for particles and self-standing sheets synthesis. However, after decades of the application of PECVD for the production of thin films, today there are an increased number of applications around the deposition of nanoparticles in thin films (nanocomposites). 18 In particular, polymer nanocomposites are of great interest since they possess enhanced properties compared to clean polymers. 25 To the best of our knowledge, only a couple of examples in the literature using plasmas at atmospheric pressure conditions focus on the self-organization of carbon nanostructeres. Besides, only a few assembly paths have been tested and much has still to be done to achieve controlled assembly of free-standing carbon nano-architectures. Revealing the relation between the plasma specific features and specific structural qualities of assembled nanostructures, which determine their properties, is a major challenge to be faced. Such an understanding would help to direct the power into targeted reactions and precursor species and to control the energy exchange pathways in the plasma. Numerical modeling can provide the required approach towards management of material and energy fluxes for growing nanostructures. However, a comprehensive model of plasma-based synthesis of carbon nanostructures, describing thermodynamics, plasma kinetics and chemistry as well as the processes of the formation of solid nuclei and the growth and interaction of plasma species with solid surfaces, is not yet available due to its complexity. Nevertheless, simplified models related, for example, to plasma chemistry, can give information for precursor flux and composition, while plasma thermodynamics benefits energy fluxes. Examples of existing modeling approaches are analytical models, zero-dimensional (0D) chemical-kinetics simulations, fluid models (which describe the chemical kinetics as well as fluid dynamics), Boltzmann transport equation solvers, and Monte Carlo (MC) models, summarized in ref. 26. A review on modeling the performance of the PECVD synthesis of carbon nanotubes can be found in ref. 27. Our work extends the scope of previous efforts in this field by using methane as a carbon precursor to produce freestanding carbon nanostructures using microwave plasmas driven by surface waves at atmospheric pressure conditions. To this end, theoretical modeling and the experimental results of carbon nanostructure synthesis using a substrate-free, microwave argon plasma driven procedure/process are presented. The method, as described in previous work, involves injecting a carbon-precursor through a plasma environment, where decomposition processes via collisions and intensified radical chemistry take place and carbon atoms and molecules are created. The transport of species into the colder zone of the reactor results in the formation of solid carbon nuclei. The main stream of carbon nuclei is gradually withdrawn from the hot plasma region into the outlet plasma stream, where flowing carbon nanostructures assemble and grow. By tailoring the microwave plasma environment, a selective synthesis of free-standing graphene platelets and nano-particles was achieved. A theoretical model similar to those presented in ref was developed to describe methane decomposition in the microwave argon plasma environment and the formation of solid carbon nuclei. Given the fact that both thermodynamic and kinetic factors have a significant impact on the nucleation process, the model includes a set of nonlinear spatially dependent differential equations describing plasma thermodynamics and chemical kinetics. The transport of carbon species into the colder zones of the reactor and carbon nuclei formation are analyzed. The influence of the microwave power, the hydrogen percentage in the background gas mixture and background Ar flow rate on the concentration of the main outlet decomposition products as well as on the formation of C and C 2 in the active hot plasma zone is analyzed. Morphological and structural qualities of the synthesized nanostructures are determined by applying scanning electron microscopy (SEM) and Raman spectroscopy. The present study is a continuation of the one presented in ref. 31, in which ethanol was used as a carbon precursor. In this work methane is used instead. The aim is to clarify the mechanisms leading to the generation of building units (C 2,C) considering two specific sets of externally controlled, operational parameters. These two sets of operational parameters can be considered as two extremes in terms of the predominant synthesis of sheets (sp 2 carbons) and diamond-like particles (sp 3 carbons) as experimentally evidenced. 2. Experimental part 2.1. Experimental setup and diagnostic techniques A surface-wave driven plasma, which is a special type of microwave plasma, was used to synthesize carbon nanostructures. The experimental setup is shown in Fig. 1. A microwave generator (magnetron) provides microwaves at 2.45 GHz with This journal is the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20,

3 Fig. 1 (A) Experimental setup, (B) cross-section of the field applicator. a maximum power of 2 kw, which travel along a waveguide system. The system consists of a water-cooled circulator, directional couplers, tree-stub adjusters, a movable short-circuit and a surfaguide, which is used as a field applicator. 34,35 About 95% of the incident microwave power P is absorbed by the plasma. A quartz tube (1.5 cm inner and 1.8 cm outer diameters) is placed vertically and perpendicularly to the wider wall of the waveguide. The background gas, argon is injected into the quartz tube, where the discharge is ignited and maintained. Additionally, gas-phase methane molecules are introduced into the plasma. 24,28 Argon flow rates vary from 600 to 2000 sccm, while methane flow rates vary from 2 to 8 sccm. The nanostructures are captured by a membrane filter system coupled to an Edwards BS2212 two-stage vacuum pump. The collection system usually accumulates up to around 1 2 mg of material until the filter permeability for the gas begins to decrease. After removing the material, the filter can be reused. The light emitted from the plasma was collected by a quartz optical fiber coupled to a Jobin-Yvon Spex 1250 spectrometer (1200 g mm 1 grating) equipped with a cryogenic, back illuminated UV sensitive CCD camera. Optical emission in the visible range ( nm) was detected. Furthermore, the 2D distribution of the wall temperature was recorded by a FLIR camera E60 while the discharge was on. The output gas stream from the plasma reactor was probed by a Stanford Research Systems RGA 200 mass spectrometer. The morphology of the created structures was analyzed by scanning electron microscopy (SEM). A JSM-7001F (FEG of JEOL), operating in secondary electron imaging mode (SEI) with an applied voltage in the range of kv, was used. The phonon properties of the structures were analyzed by using a LabRAM HR Visible (Horiba Jobin-Yvon) Raman spectrometer with 1 cm 1 spectral resolution and 633 nm He Ne laser excitation with a laser spot size of 2 mm. Measurements were performed with a laser power P l = mw to avoid overheating Plasma source characterization The emission spectrum in the range nm of argon/ methane plasma detected at an axial distance of 3 cm from the launcher is shown in Fig. 2. The spectrum is dominated by argon atomic lines and several sequences of vibrational bands corresponding to C 2 Swan bands in the nm spectral range. The typical C 2 emission is generated by the radiative decay of the C 2 *(A 3 P g ) state. Due to the low energy threshold (E ext = 2.4 ev), ground state C 2 molecules can easily be excited to this level either by electron impact, 36 C 2 (X) + e - C 2 *(A 3 P g ) + e (1) or by three body recombination processes involving C and Ar atoms C+C+Ar- C 2 *(A 3 P g ) + Ar (2) Additionally, reactions of C 2 molecules with hydrogen radicals result in C 2 dissociation and the formation of atomic carbon. The collision rates in the microwave plasma at atmospheric pressure are high enough to ensure local thermodynamic equilibrium. Therefore, the rotational temperature may be considered nearly equal to the gas kinetic temperature. Since the plasma setup operates at atmospheric pressure conditions, small amounts of nitrogen are present, resulting in creation of CN molecules in the plasma environment. The emission band from the CN species in the spectral range ( nm) corresponding to violet CN(B X 2 +) transitions was used to estimate the rotational temperature. By comparison of the rotational lines in the frame of the CN(1 1) vibrational band with simulated ones using Lifbase software 37 (Fig. 3A), a temperature of about 3000 K was obtained. The comparison Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018

4 Fig. 2 Plasma emission spectrum at P = 1 kw, p = 1 barr, QAr = 600 sccm, QM = 2 sccm, detected at 3 cm from the launcher. was made employing two different wavenumber ranges of the spectrum to be considered in the fitting procedure. The inset in Fig. 3A demonstrates the fitting of a narrow part of the detected spectrum. The measured temperatures along the axis and the theoretical profile are shown in Fig. 3B. The reference point z = 0 corresponds to the launcher position. As can be seen from Fig. 3B, the rotational temperature measured at 4 cm from the launcher was 2900 K and remains nearly the same (around 3000 K) from 3 to 6.5 cm away from the launcher, at the middle of the hot plasma zone. The latter extends up to about 10 cm distance from the launcher. Due to the presence of hydrogen in the precursor, the emission of hydrogen Balmer lines can also be detected. To this end, the Hb line intensity profile was measured at several positions along the discharge axis in the hot plasma zone. All measured profiles were fitted with a Voigt function, which is a convolution of Gaussian and Lorentz profiles (Fig. 4A). The instrumental broadening caused by the spectrometer was estimated to be Å. A deconvolution procedure was used to determine the full width at half-maximum of the Gaussian (DlG) and Lorentzian (DlL) components. Stark and van der Waals effects contribute to the Lorentzian broadening under the present conditions. To estimate the van der Waals broadening, the expression given in ref. 38 was used. The electron density has been estimated using the expression:38 Dlstark = a1/2ne2/3, where a1/2 = Å.39 The estimated electron density varies within the limits (7 9) 1013 cm 3 at QAr = 600 sccm, QM = 2 sccm, P = 1 kw and ( ) 1014 cm 3 at QAr = 1000 sccm, QM = 7.5 sccm, P = 1 kw. The data are related to the hot plasma zone, which in the axial direction extends up to approximately 7 cm away from the launcher (Fig. 4B). As shown in the previous studies28 31 the plasma reactor wall temperature is one of the key parameters influencing the This journal is the Owner Societies 2018 Fig. 3 (A) Emission spectra of argon/methane plasma at QAr = 600 sccm, QM = 2 sccm, P = 1 kw; (B) gas temperature profile experimental points and the theoretical profiles. process of synthesis of high quality graphene sheets. To this end, infrared sensitive measurements of the wall temperature Phys. Chem. Chem. Phys., 2018, 20,

5 Fig. 4 (A) Emission profile of the H b line; (B) estimated electron density along the plasma column at Q Ar = 600 sccm, Q M = 2 sccm, P = 1 kw and Q Ar = 1000 sccm, Q M = 7.5 sccm, P = 1 kw. were also performed. The 2D map of the wall temperature as detected by a thermal imager (FLIR camera) is shown in Fig. 5A. The corresponding variation of the wall temperature along the tube axis is shown in Fig. 5B. As seen, the wall temperature increases, reaching a maximum at the middle of the hot zone, and then decreases with increasing distance from the launcher. The measured wall temperature axial profile (Fig. 5B) as a boundary value of the gas temperature was fitted by an analytical expression and used as an external parameter in the numerical calculations Theoretical model The microwave plasma reactor was modeled considering the formation of different zones (Fig. 6). The first one is the surface wave sustained discharge zone, including the zone inside the launcher and the extended hot plasma zone outside the launcher. Here, the surface wave power is absorbed by plasma electrons, which transfer the power to heavy particles via elastic and inelastic collisions, resulting in high gas temperatures (up to 3000 K). The gas temperature remains nearly constant in the discharge zone when moving away from the launcher (up to B10 cm) and then drops sharply in the near (10 13 cm) afterglow plasma zone (Fig. 3B). These two regions form the hot plasma zone, where the equilibrium chemical reactions take place. The next zone in the axial direction that includes the late plasma afterglow (up to B20 cm) is the assembly zone (Fig. 3B and 6), where the kinetic processes of growth and assembly of carbon nanostructures take place. The model is based on a self-consistent treatment of plasma thermodynamics and chemical kinetics and is similar to that Fig. 5 (A) Thermal image of the discharge at P =1kW,Q Ar =600sccm,Q M = 2 sccm; (B) axial wall temperature profile along the discharge at Q Ar =600sccm, Q M =2sccm,P =1kWandQ Ar =1000sccm,Q M =7.5sccm,P =1kW Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018

6 Fig. 6 Scheme of the main processes in the microwave plasma. previously developed. 31,32 The model input parameters are externally controlled, i.e., gas flows, total power delivered to the launcher, pressure and background gas mixture composition, i.e., the Ar/CH 4 gas mixture injected into the discharge. Local thermal equilibrium and linear wave power absorption along the generated plasma column are assumed. The developed kinetic scheme is similar to the one presented in ref. 41 and allows the study of ethanol, methanol, and methane decomposition. The chemical kinetic model accounts for the formation of the solid carbon, both in the effluent plasma gas flows and deposited on the tube wall, the diffusion mechanism, and the detailed kinetics of C 2 radicals and gas phase carbon atoms. The kinetic scheme includes 57 components (Ar, CH 3 OH, H 2,H,HO 2,H 2 O 2,H 2 O, O, OH, O 2,CH 4,CH 3,CH 2,CH, CH 2 O, CH 2 (s), CH 2 CO, HCO, CH 3 O, CO 2, CO, HCOH, CH 2 OH, HCCO, C 2 O, HCCOH, C 2 H, C 2 H 2,C 2 H 4,C 2 H 6,C 2 H 5,C 2 H 3,H 2 CCCH, HCOOH, CH 2 CHCO, CH 2 CHCHO,CH 3 CHCO, C 3 H 8,C 3 H 2,C 3 H 6, C 3 H 5 (a), C 3 H 5 (p), C 3 H 5 (s), C 3 H 7 (i), C 3 H 7 (n), C 3 H 4 (p), C 3 H 4 (a), CHOCHO, CH 3 HCO, C 2 H 5 OH, C 2 H 4 OH, CH 3 CHOH, CH 3 CH 2 O, CH 2 HCO, CH 3 CO, C 2, C) and about 390 chemical reactions. The full list of the chemical reactions and the corresponding rate constants can be found in ref. 31. The rate constants expressions and corresponding activation energies for the required temperature range were taken from those available in literature data The rate coefficients of the most important chemical reactions are of the order of 10 6 s 1 or more in the temperature range K, which justifies the assumption of local thermal equilibrium. The rate constants were extrapolated to the experimentally determined upper limit of the temperature range (B3000 K) using a non-linear extrapolation method presented in ref. 42. It should be noted that the existing kinetic model for methane oxidation presented by GRI-3 43 was also applied in our study. The results from the GRI-3 model are consistent with the results obtained with the model proposed in this study for methane decomposition. Thermodynamic information on the processes involved was obtained from the available thermodynamic databases Under the conditions considered, the main role of the electrons is to absorb the surface wave power and to transfer it via collisions to the heavy particles. This is expressed by the assumption that nearly all the absorbed microwave power is transferred to the gas thermal energy. Therefore, the gas thermal balance equation reads, p 0 1 dt a V 0 C p k Bol T 0 dz ¼ 4wðTÞ R 2 ðt a T w Þþ d dp S dz where V 0 is the initial gas velocity (m s 1 ), T 0 is the initial gas temperature (K), T a is the temperature at the axis, T w is the wall temperature, p 0 is the gas pressure, k Bol is Boltzmann s constant, C p ¼ 5 2 k Bol is the heat capacity at constant pressure, S is the plasma cross-section, P is the absorbed microwave power, w(t) represents the argon thermal conductivity, and d is a coefficient expressing the fraction of absorbed power from the wave which is transferred to thermal energy of the gas. We further assumed that d = 0.9, which is a typical value for atmospheric pressure conditions, and B10% energy losses due to radiation and losses in the dielectric. The approach used has been validated in a series of extensive experimental and theoretical studies. 33,40,52 56 It should be stressed that the gas temperature is also radially inhomogeneous, with the radial profile assumed to be similar to the radial velocity distribution profile, i.e. close to a parabolic shape. The process of carbon sublimation corresponds to the diffusion mechanism of the gas phase carbon atoms and molecules towards the colder zone through the so-called boundary of vaporization, an isothermal surface with a temperature equal to approximately 2500 K (see below). The boundary of vaporization is schematically illustrated in Fig. 6. This zone spans the central part of the plasma (3) This journal is the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20,

7 column around the axis. The next zone after the hot plasma zone in the radial direction is the active chemical zone, where the processes of nucleation, crystallization and re-crystallization occur. According to the theory, the diffusion coefficient dependence T 3=2 on the temperature is D ¼ D 0,whereD 0 is the diffusion T 0 coefficient at temperature T 0. The transformation rate is found: k ¼ 6D 0 T s 1 ; (4) r 2 T 0 assuming that diffusion is the transport mechanism of gasphase carbon atoms from the hot to the active chemical pffiffiffiffi zone and keeping in mind that the average distance r 2 traveled by a particle within the time interval t is proportional to the square root of this time interval. 57 The diffusion rate was T 3=2 estimated as k est ¼ a C=C2 k 0 a C=C T 3=2. 32, Here the factor a C/C2 is 1 and for C and C 2, respectively. The set of rate balance equations for the species considered and the thermal balance equation were solved in a self-consistent manner yielding an integral description of the axial structure of the discharge and post discharge regions as well as of the outlet gas stream. The sets of rate balance equations for different species can be found in ref. 31. The algorithm used for the calculations is similar to that of the well-known computer code Reaction Design, Chemkin 58 or Kintech Lab, Chemical Workbench. 59 Achemical mechanism is a collection of reactions that transforms reactants into products, including all important intermediates. All reactions in the mechanism are treated as if they are elementary ones. Reactions describe interactions and conversions between gas-phase species. Each species in a reaction must be associated with corresponding thermodynamic data. The thermodynamic data are used to calculate equilibrium constants and reverse-rate coefficients for the reaction. The system of ordinary differential equations is solved by implicit techniques. 60 The gas temperature, which is radially inhomogeneous, influences the spatial dependence of the rate coefficients k ¼ AT n exp E a for reactions with high activation R gas T energy, E a. At the same time, the present model considers only axial temperature variations, assuming local thermal equilibrium. Therefore, effective radially averaged rate constants k eff ¼ 2 Ð R R 2 0 kt ð 2AÐ R Þrdr ¼ R 2 0 Tr ðþn exp E a rdr, were used Tr ðþ instead. When calculating k eff, the temperature radial profile was assumed to be parabolic with a maximum at the center of the tube: Tr ðþ¼ðt a T w Þ 1 r 2 þ T w. In our previous R work, 40,53 56 the microwave power transfer to the plasma is studied in detail both experimentally and theoretically. The main parameters of the microwave plasma at atmospheric pressure were determined, i.e. the electron density r10 14 cm 3,electron temperature r6000 K, and carrier gas concentration r10 18 cm 3. Therefore, the local thermal equilibrium assumption is well justified. 61 The cross section for the electron impact dissociation of methane is about 2 10 Å 2.UsingBolsig+ 55 we can estimate the rate constant for electron impact dissociation k elec.impact as less than cm 3 s 1, while the dissociation rate k Ar impact due to the collisions with all Ar atoms under the conditions considered is about cm 3 s 1. The difference in the number densities of the electrons and the argon atoms is about four orders of magnitude. Therefore, the contribution of electron impact dissociation is at least three orders of magnitude less than that of argon impact dissociation and can be neglected. Note that for low pressure conditions electron impact dissociation is predominant. It is to be noted that only ground state atoms and molecules have been considered in the present model. Additionally, the scheme of methane dissociation is presented in Fig. 6. The scheme was deduced from analysis of both modeling results and the rate coefficients and a detailed discussion of the methane dissociation mechanics is given below. 3. Results and discussion According to the numerous investigations dedicated to methane decomposition in microwave plasma, including the current study, among the stable output products are hydrogen, acetylene, solid carbon, ethylene, and unconverted methane The equilibrium values i.e. those obtained under assumption of a thermal equilibrium, of the stable substances and intermediate complexes along the hot and active chemical plasma zones are useful to understand the general workings of methane decomposition and the formation of carbon precursors. Two approaches are presented: homogeneous and heterogeneous approximations. The calculated equilibrium values assuming a homogeneous approximation are shown in Fig. 7. Three temperature regions can be distinguished, depending on the formed stable substances and intermediate complexes. The first region is approximately up to 1000 K where methane is present; the second region ranges from K where methane is decomposed, and acetylene and hydrogen are formed; and the third region is the region of complete Fig. 7 Simplified equilibrium diagram assuming a homogeneous approximation for the main methane decomposition products for Q Ar = 600 sccm and Q M = 2 sccm Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018

8 Fig. 8 Simplified equilibrium diagram assuming a heterogeneous approximation for the main methane decomposition products for Q Ar = 600 sccm and Q M = 2 sccm. Fig. 9 Evolution of the main species concentrations in the discharge and the afterglow plasma for Q Ar = 600 sccm, Q M = 2 sccm, P = 1 kw. decomposition of C 2 and H 2 to gas-phase atomic carbon and atomic hydrogen for temperatures more than 3000 K. The decomposition evolution assuming a heterogeneous approximation (Fig. 8) can be divided into two regions. The first region (up to approximately 2500 K) is of complete decomposition of methane into molecular hydrogen and carbon; and the second region is above 2500 K, where complete decomposition of C 2 and H 2 to gas-phase atomic carbon and atomic hydrogen occurs. Therefore, the evaporation boundary is at approximately K, which is B1.4 times higher than the evaporation boundary value estimated for the case of ethanol and methanol decomposition (B1800 K). 31 Apparently, the discrepancy in the boundary values of vaporization is due to the content of oxygen atoms in the ethanol and p methanol ffiffiffiffi molecules. The average distance r 2 traveled by a particle within the time interval t (4) corresponds to the radius determined by the vaporization boundary and is defined as shown in ref. 31 by the ratio: 8 0; T a T >< w r vapor rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi T a T >: vapor R ; T a T w T a T w where T vapor corresponds to the phase transformation temperature. Considering the measured gas temperature at the axis is B3000 K, the radius determined by the vaporization boundary is B0.5 mm (see eqn (5)). This value of 0.5 mm is significantly less than the radius determined for the vaporization boundary of ethanol and methanol decomposition in argon plasma. Therefore, unlike the case of ethanol/methanol decomposition, the solid fraction formation of carbon species in methane/ argon plasma begins in the hot plasma zone (B2200 K), where the active chemical reactions take place. The calculated relative concentrations of the main products, which result from the decomposition process of methane in the Ar plasma, are shown in Fig. 9 as a function of the distance from the launcher. The mechanism of methane dissociation begins at temperature B1000 K (z B 0.5 cm), where the major (5) initial reaction is C H bond breaking followed by the formation of radical CH 3 (cf. Fig. 6). Further, polymerization of CH 3 occurs to form CQC in the ethylene molecule C 2 H 4, followed by the formation of a triple CRC bond in acetylene C 2 H 2. Acetylene is a stable molecule that can be decomposed into simple C and H atoms at sufficiently high temperature. Based on the analysis of the kinetic scheme, 24 the formation of the acetylene can be represented by the following chemical mechanism: + CH 3 þ CH 3! C 2 H 6!! C 2 H 4! CH 2 þ CH 2! + CH þ CH!! C 2 H 2 (6) CH 2 þ CH! CH 3 þ CH! Additionally, the limit of H 2 formation according to the two methane dissociation reaction mechanisms, i.e. CH 4 2 C(solid)+ 2H 2 (heterogeneous mechanism) and CH 4 2 1/2C 2 H 2 +3/2H 2 (homogeneous mechanism) is also shown in the Fig. 9. As can be seen, the experimental result for the outlet hydrogen concentration (asdeterminedbymassspectroscopy)fromonemoleofmethane is between 3/2 and 2 hydrogen molecules. The accuracy of H 2 determination is evaluated in the range of B10%. Increasing the microwave power from 1 kw to 1.5 kw or higher, a faster than linear increase of the gas temperature occurs due to the temperature dependence of the argon thermal conductivity (cf. eqn (3)). In this case, the properties of the plasma approach those of the arc discharges, where the temperature difference between the electrons and the heavy particles decreases. The transformation rate of C and C 2 is proportional to T 3/2 (cf. eqn (4)). The evolution of the main species concentrations along the plasma zones calculated for P = 1.5 kw, Q Ar =600 sccm, Q M = 2 sccm is shown in Fig. 10. With the temperature increase, the volume of the hot plasma zone increases at the expense of the active chemical plasma zone (cf. eqn (5)). Reducing the volume of the latter actually results in the withdrawing of C and C 2 from the active chemical plasma zone This journal is the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20,

9 Fig. 10 Evolution of the main species concentrations in the discharge and the afterglow plasma for QAr = 600 sccm, QM = 2 sccm, P = 1.5 kw. The black solid line indicates the gas temperature profile. before the formation of carbon clusters occurs. Thus, for a discharge length of B6.5 cm, it can be assumed that active C and C2 particles are absent at the end of the microwave discharge (Fig. 10). These species are being deposited on the tube wall in the middle of the discharge zone instead. With the departure of C and C2, the majority of radicals also disappear from the plasma. Therefore, a B2 cm long chemically inactive zone with a low concentration of active species appears at the end of the discharge. This calculation brings us to the conclusion that increasing the microwave power would lead to an increase of the chemically inactive zone. The C and C2 deposited on the tube wall are expected to form disordered, amorphous carbon structures. These structures will be subsequently sputtered by the argon ions and withdrawn with the main flux and captured by the collecting filter. Thus, the solid carbon produced at higher power P = 1.5 kw is an order of magnitude more than the one obtained at P = 1 kw, which is further confirmed by sensitivity analysis (see below). Although the solid carbon production increases in the case of higher powers, the collected structures differ from graphene, due to the lower residence time that C and C2 spend in the active chemical zone. Thus, the time for carbon cluster formation is lower, which leads to a decrease in the graphene production rate. The evolution of the main species concentrations for the two considered conditions, i.e. QAr = 600 sccm, QM = 2 sccm and QAr = 1000 sccm, QM = 7.5 sccm is shown in Fig. 11. It should be noted that the mechanism of acetylene formation from methane is well studied The methane decomposition scheme obtained in the present study is fully consistent with the results and conclusions of these reports, which demonstrates the adequacy of the proposed model. Differentiating the relative C (solid) concentration curves (Fig. 11) with respect to time gives an estimate for the rate Ws of nanoparticle formation. At different positions along the discharge the relative rate of nanoparticle formation varies from 0 s 1 at the beginning of the stream, reaches the maximum value E0.2 s 1 in the discharge zone and decreases again to approximately zero in the middle post discharge zone. The total rate of formation of nanoparticles can be evaluated by integrating the relative rate of Phys. Chem. Chem. Phys., 2018, 20, Fig. 11 Evolution of the main species concentrations comparison of the two conditions: QAr = 600 sccm, QM = 2 sccm, P = 1 kw and QAr = 1000 sccm, QM = 7.5 sccm, P = 1 kw. nanoparticle formation along the entire length of the tube multiplied be the amount of methane supplied and Z (%) the percentage of the nanoparticles deposited on the tube wall. Thus, the mass rate of nanopartical formation is defined as follows: W ¼ 7: MC 1 ðl Z dz QCH4 Ws kg s 1 ; 100% V ð z Þ 0 where MC is the molar mass of C 12 g mol 1. For the typical conditions shown in Fig. 11, the mass rate of nanoparticle formation is B(2 4) 10 6 kg s 1. The obtained values are in good agreement with the measured weight of the nanostructures deposited on the filter. In order to assess the relative contribution of the different chemical reactions for solid carbon, acetylene and H2 production, a sensitivity analysis was performed. The relative sensitivity to the rate coefficients has been calculated and is shown in Fig. 12. The relative sensitivity is defined as:46 rel Si;r ln ci ln kr (7) Fig. 12 Sensitivity analysis for H2, C2H2 and C (solid) concentrations in the plasma at QAr = 600 sccm, QM = 2 sccm, P = 1 kw. This journal is the Owner Societies 2018

10 Fig. 13 Sensitivity analysis of the main decomposition products dependence on (1) the microwave power P, (2) additional hydrogen gas in the background gas mixture C H2 and (3) the argon flow rate Q Ar. where k r are the rate coefficients of the elementary reactions and c i are the species concentrations. The influence of the operational parameters such as microwave power (P) and Ar flow rate (Q Ar ) on the main outlet decomposition products is shown in Fig. 13. The relative sensitivities of different stable species as functions of the externally controlled parameters, such as the microwave power P delivered to the launcher, the hydrogen percentage in the background gas mixture C H2 and the argon flow rate Q Ar, were evaluated as follows: SðPÞ rel i rel SC H2 ¼ C H 2 i c i SQ ð Ar Þ rel i ¼ P c ln c ln P i ln c ln C H2 ; ¼ Q i ln c i c ln Q Ar where c i is the species concentration. As shown in ref. 31, additional H 2 in the background Ar gas significantly influences the ratio of sp 2 /sp 3 carbons. Therefore, here the influence of additional H 2 has also been analyzed. Generally, the impact of these parameters is similar to that of ethanol decomposition, 31 except that the products of methane decomposition do not include CO and water. Increasing the microwave power leads to an increase of solid carbon production rate, however it does not affect the generation of H 2 and C 2 H 2. This is due to the fact that the concentration of the solid carbon is several orders of magnitude lower in the final products of pyrolysis, therefore it cannot influence the absolute change of H 2 and C 2 H 2 concentrations. If additional hydrogen gas is introduced into the plasma the solid carbon production rate decreases, due to the shift of the equilibrium towards acetylene and ethylene formation from C 2. The increase of the background argon gas also decreases the total yield of solid carbon (Fig. 13) due to the decrease of the residence time of the decomposition products in the plasma environment. (8) Fig. 14 Sensitivity analysis on the influence of microwave power, additional hydrogen gas in the background gas mixture and the background argon flow rate on the methane decomposition (shaded) and ethanol decomposition in the hot plasma region at axial distance z =4cm(P =1kW,Q Ar =600sccm, Q M = 2 sccm) in the case of argon/methane and P =600W,Q Ar =600sccm, Q Et = 0.6 sccm in the case of argon/ethanol. In order to reveal the main trends concerning C and C 2 generation and correspondingly the range of operational parameters prone to predominant synthesis of sheets (sp 2 carbons) or particles (sp 3 carbons), sensitivity analysis has been performed. As demonstrated in a previous analysis 31 the ratio of sp 2 /sp 3 carbons depends on the number densities of carbon atoms and C 2 molecules. The influence of the microwave power, the hydrogen percentage in the background gas mixture and the argon flow rate on the formation of C and C 2 in the active hot plasma zone is shown on Fig. 14. The increase of the microwave power leads to an increase of both C and C 2, while the increase of the background gas flow and hydrogen decreases their formation. These results are compared with the ones obtained in argon/ethanol plasma and clear similarities can be observed. 31 However, it should be noted that additional hydrogen gas supply in the argon/methane plasma affects the decrease of C and C 2 equally, unlike the case of argon/ethanol. In the argon/ethanol plasma environment the additional hydrogen has greater effect on the destruction of C 2 than on C. More hydrogen leads to the formation of more H and OH radicals in the plasma via the reaction CO 2 +H 2 - CO + OH + H. Since the OH radical is more reactive than atomic hydrogen, OH has a stronger influence on the dissociation of C 2 (C 2 +OH- CO + CH) than atomic hydrogen (C 2 +H- C + CH). In the argon/methane plasma however, H and OH have approximately the same effect on the carbon via reaction mechanisms: C + OH - CO + H and C+H+M- CH + M. As has already been shown previously, 29,31 the main transport mechanism of the carbon species is diffusion into colder zones, both in the radial and axial directions, where the nucleation process occurs. The formed solid carbon is partially deposited on the tube wall, but the main part is gradually withdrawn from the hot plasma region (Fig. 6). In this way, the predominant part of the solid carbon nuclei is transferred with This journal is the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20,

11 Fig. 15 (A) SEM image of graphene sheets synthesized at Q Ar = 600 sccm, Q M = 2 sccm, P = 1 kw, p = 1 atm; (B) the corresponding Raman spectrum. the gas flux into the post discharge zone ( assembly zone ), where the kinetic processes of growth and assembly of carbon nanostructures occur. It should be emphasized that the carbon particles nucleate and grow in the plasma and gas phases, therefore the assembly zone is the volume in which this process occurs in contrast to the assembly mechanism at surfaces. 71 These nanostructures are subsequently captured from the gas stream by a membrane filter (cf. Fig. 1). The captured structures are easily peeled off the membrane and analyzed by SEM and Raman spectroscopy. A typical SEM image of the synthesized carbon nanostructures is shown in Fig. 15A. The image shows the intrinsic curved morphology of graphene sheets. As seen, wrinkled paper-like tiny structures are observed. This result is further confirmed by Raman spectroscopy measurements (Fig. 15B). The presented Raman spectrum corresponds to an average of the spectra collected at different regions of the sample. It reveals a G-peak related to the Brillouin zone center mode at 1585 cm 1 and a sharp two-phonon (2D) peak at about 2667 cm 1 due to the second order process, providing evidence that the obtained carbon nanosheets are indeed graphene sheets. The D peak (at 1333 cm 1 ) and the small shoulder of the G peak are related to structural disorders, the presence of sp 3 carbon and/or edge effects. The 2D peak is the most prominent feature in the Raman spectrum of graphene. Its position, shape and intensity are frequently used to distinguish between single and multi-layer graphene. In this case, the ratio between the 2D and G peak intensities is I 2D /I G B 0.8, the ratio of the D to G peak intensities is I D /I G B 0.62 and the FWHM of the 2D peak is B46 cm 1. These results are an indication that multi-layer graphene was synthesized at the conditions considered. Note that selective synthesis (one carbon allotrope present in the samples) of free-standing graphene sheets was obtained. Nearly a threefold change of the flux of carbon species for growing nanostructures in the assembly zone along with an increase of the background Ar flux up to 1000 sccm, results in the formation of a different type of nanostructure as seen in Fig. 16A and B. At conditions Q Ar = 1000 sccm, Q M = 7.5 sccm, P = 1 kw, p = 1 atm, particle-like structures are the predominantly synthesized structures shown in Fig. 14A, which is representative of the whole sample. However, albeit at a small amount, structures resembling graphene sheets were observed in Fig. 16B. The carbon flux for growing structures increases up to B0.2 s 1.Theresidence time in the assembly zone also changes from 0.1 s for Q Ar = 1000 sccm, Q M =7.5sccmto0.2sforQ Ar = 600 sccm, Q M =2sccm. Moreover, increasing the background flux leads to a change of the thermal map in the active chemical and the assembly plasma zones (Fig. 3B), and different nanostructures were synthesized (Fig. 16). It should be recalled that the reduction of the Gibbs free energy of a saturated environment (gas + solid phase) is a driving force for both nucleation and growth. This reduction is proportional to the gas temperature. Lower gas temperatures as well as large amounts of precursors foster supersaturation conditions in the environment (gas + solid phase) and thus promote the synthesis of plenty of multiple nuclei, while higher temperatures (and lower precursor amounts) favour the appearance of fewer nuclei and, consequently, a deviation from supersaturation conditions. At high enough temperatures only the most stable nucleation centres will survive. Moreover, the increase of the background flux means a decrease of the residence time of the precursorinboththe activechemical and the assembly zones. For higher partial methane fluxes and lower temperatures in the assembly zone (7.5 sccm) carbon nanoparticles with a size ranging from 70 to 800 nm are produced. The particle size distribution (as determined using ImageJ 72 ) is shown in Fig. 17. As expected, the density of the carbon species increases as the partial flux increases, and, as a result, the conditions in the assembly zone favor the creation of spherical particle-like nanostructures. Additionally, there is interplay between the density of the nucleation centers and the mean free path (l C C ) for carbon carbon atomic collisions. Once introduced in a volume already containing nucleation centers, carbon atoms can (a) collide with other C atoms (or small carbon clusters) and eventually form new nucleation centers or, (b) interact with previously formed nucleation centers and contribute to their further growth. The probability of the two cases will depend on the density of nucleation centers and the mean free path l C C.Ifthe Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018

12 Fig. 16 SEM image of carbon nanoparticles synthesized at QAr = 1000 sccm, QM = 7.5 sccm, P = 1 kw, p = 1 atm (A) typical view and (B) structures resembling graphene sheets; (C) the corresponding Raman spectrum; (D) fitted with four Lorentzians D and G-peaks. Fig. 17 Particle size distribution. lc C is very small, it will be more probable for an atom to collide with another atom forming a new nucleation centre, than to reach already existing nucleation centers. The density of nucleation centers increases until the average proximity between two This journal is the Owner Societies 2018 neighboring nucleation centers becomes smaller than lc C. Consequently, a decrease of lc C will increase the density of the nucleation centers. In our case, lc C for the conditions with larger (QAr = 1000 sccm, QM = 7.5 sccm, P = 1 kw, p = 1 atm) and smaller (QAr = 600 sccm, QM = 2 sccm, P = 1 kw, p = 1 atm) amounts of the precursor is estimated as 0.3 mm and 1 mm, respectively.73 Therefore, a higher density of nucleation centers is expected under the conditions with larger amounts of the precursor, forcing supersaturation conditions. However, the existence of a small amount of nuclei surpassing the critical size for nucleation and growth and conditions in which no supersaturation of species is possible given the limited flux promote the growth of existing nucleation centers rather than the creation of new seeds. Thus, the conditions in the assembly zone foster the creation of planar nanostructures as observed in Fig. 15. Contrary, when the methane partial flux increases and thus there exists a large amount of nuclei with size greater than the critical one, supersaturation conditions occur, which promotes the creation of new seeds and multiple nucleation centers. The Raman spectra corresponding to the structures shown in Fig. 16A and B are shown in Fig. 16C. Several significant Phys. Chem. Chem. Phys., 2018, 20,

13 changes in the spectrum occur as compared to the Raman spectrum for the conditions in Fig. 15B. There is a significant decrease of the graphene related 2D peak along with an increase of the D-peak with respect to the intensity of the G peak. The position of the G peak is also shifted to 1596 cm 1 compared to 1585 cm 1 for the graphene samples. The SEM observation demonstrates that small quantities of graphene sheets are still present in the sample as seen in Fig. 16B. Part of the Raman spectrum between 1000 and 1800 cm 1 is displayed in Fig. 16D. The complex shape of the Raman intensity can be fitted with 5 Lorentzians. Their frequencies are 1168, 1251, 1337, 1473, and 1598 cm 1. It is seen that as well as the dominating D and G bands observed for all sp 2 containing carbon materials, the weak so-called n 1 (1168) and n 3 (1473) bands are observed. 10 These two bands are always observed in the spectra of nanocrystalline diamond (NCD). Although it is doubtful whether they are really intrinsic for the NCD or rather they originate from a satellite co-phase (trans-polyacetylene), their observation in our spectra can be a sign for the presence of NCD in the sample. 4. Conclusions In an effort to apply a more deterministic approach, a complex theoretical and experimental study on the synthesis of carbon nanostructures such as multilayer graphene sheets and nanoparticles, using microwave plasma at 2.45 GHz at atmospheric pressure has been conducted. The procedure involves the introduction of methane into a microwave argon plasma environment, where decomposition of methane molecules takes place and solid carbon is created. The solid carbon nuclei, guided by the plasma gas stream, pass through the assembly zone, where free-standing structures are created. The freestanding graphene structures are captured by a membrane filter. The predominant synthesis of graphene sheets occurs at a very narrow range of operational conditions. Free-standing sheets are selectively synthesized at a microwave power of 1 kw, and Ar and methane flow rates of 600 sccm and 2 sccm, respectively. For the same power but at higher flow rates, i.e and 7.5 sccm of Ar and CH 4 respectively, the formation of diamond-like nanoparticles was achieved. The set of key parameters that determine which type of structures is going to be predominantly synthesized is the temperature gradient, the residence time, and the concentration of the carbon building units in the assembly zone of the plasma reactor. Moving closer to supersaturation conditions by increasing the concentration of the carbon species in the plasma environment, i.e. increasing the flux of the building units, and reducing the residence time in the assembly zone, results in the formation of multiple nuclei, thus favoring the predominant synthesis of carbon nanoparticles. A theoretical model previously developed was further updated to identify the main decomposition products of methane in argon plasma, i.e. hydrogen, acetylene, solid carbon, ethylene, and unconverted methane, to describe in detail the formation of carbon precursor species and the formation of solid carbon nuclei. The methane decomposition begins at 1000 K with the breaking of the C H bond, followed by the formation of the CH 3 radical, the polymerization of CH 3 to form CQC in the ethylene molecule C 2 H 4, followed by the formation of a triple CRC bond in acetylene C 2 H 2. At higher temperatures, acetylene is decomposed into gas-phase carbon atoms and molecules, which diffuse into colder nucleation zones, both in the radial and axial directions, resulting in gas-phase carbon transformation into solid carbon nuclei. The model predictions for H 2 were validated with the experimental results obtained by mass spectrometry. Sensitivity analysis of the influence of the main decomposition products dependence on the microwave power, additional hydrogen percentage in the background gas mixture and the argon flow rate was performed. The increase of the microwave power leads to an increase of both C and C 2, while the increase of the background gas flow and hydrogen decreases their formation. To accomplish a complex description of the processes of carbon nanostructure formation, the model should be further updated incorporating a mechanistic approach to describe in detail solid nanostructure formation. Conflicts of interest There are no conflicts to declare. Acknowledgements The work was performed under auspices of PEGASUS (Plasma Enabled and Graphene Allowed Synthesis of Unique nano Structures) project. The project has received funding from European Union s Horizon research and innovation programme under grant agreement no This work was also funded by Portuguese FCT Fundação para a Ciência e a Tecnologia, under Project UID/FIS/50010/2013, Project INCENTIVO/FIS/LA0010/2014, and grant SFRH/BD/52413/2013 (PD-F APPLAuSE). References 1 A. K. Geim and K. S. Novoselov, Nat. Mater., 2007, 6, K. Ostrikov, E. C. Neyts and M. Meyyappan, Adv. Phys., 2013, 62, K. S. Novoselov, A. K. Geim, S. V. Morosov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva and A. A. Firsov, Science,2004, 306, K. Ostrikov, U. Cvelbar and A. B. Murphy, J. Phys. D: Appl. Phys., 2011, 44, Y. Yurum, A. Taralp and N. Veziroglu, Int. J. Hydrogen Energy, 2009, 34, E. Tatarova, N. Bundaleska, J. Ph. Sarrette and C. M. Ferreira, Plasma Sources Sci. Technol., 2014, 23, L. M. Malard, M. A. Pimenta, G. Dresselhaus and M. S. Dresselhaus, Phys. Rep., 2009, 473, A. C. Ferrari, Solid State Commun., 2007, 143, Phys. Chem. Chem. Phys., 2018, 20, This journal is the Owner Societies 2018