Energy considerations regarding pulsed arc production of nanomaterials

Nanomaterials consisting of 0D, 1D, or 2D objects show many applications due to their extreme physical and chemical properties.^1,2 Until today, their main synthesis techniques include mechanical or chemical exfoliation, decomposition of molecular precursors in gas phase, and material ablation by arc discharge.^3–8 The most developed synthesis method of quality nanomaterials in industry is chemical vapor deposition (CVD), either thermal- or plasma-activated.^9,10 The deposition of high-purity samples and process scalability constitute the main strengths of this technique. However, typical temperatures of the gas precursors do not overcome several hundred Kelvin (<0.1 eV), thereby relying sample quality on thermal treatments applied to dedicated substrates in order to achieve an optimal growth of nanoparticles. Moreover, a successful CVD growth process often requires previous conditioning of surface properties via deposition of a catalyst layer and/or creation of a template.^5,11 CVD is, therefore, considered as a surface production technique because of the critical role played by the substrate conditions.

Physical vapor deposition (PVD) techniques constitute alternative means to generate nanoparticles. In contrast to CVD methods, anodic arc discharge is a PVD volume production technique based on material ablation by high-temperature plasma.^3,12 In the plasma column, the ejected species can surpass typical CVD energies by one order of magnitude (approximately 1 eV). Such conditions are favorable for the synthesis of a number of nanomaterials.^13–16 Thus, arc discharge on specific anode materials constitutes a very flexible synthesis tool, where nanoparticle growth is tuned via plasma parameters as well as substrate temperature. In fact, nanosynthesis by arc discharge is basically a multi-step growth process of atom agglomeration from the very anode ablation up to particle incidence onto the substrate.¹² Gas-phase processes are then followed by surface diffusion of the deposited elements on the substrate, ending up in material solidification. The arc experimental conditions are appropriate for mass production of nanomaterials showing quality crystalline properties. Besides DC arc discharges, AC arcs and pulsed arcs within the kHz range have been successfully applied to deposit carbon nanostructures and ceramic films with the aim to increase the control over the discharge.^17–20 In these cases, middle-to-high frequencies were adopted so that an important fraction of energetic electrons could be confined within the oscillating plasma volume. Pulsed arcs held at high temperature have enabled the growth of diverse nanocarbons with well-defined properties.^20,21 Recent research pointed out pulsed anodic arc discharge in the Hz range and without intentional heating as a novel arc discharge method.²² Hz-modulated pulsed arcs provide clearly distinguished active (ablating) and inactive (non-ablating) phases in virtue of the slower variation of thermal flux on the anode. Such a technique is promising for nanosynthesis thanks to a number of advantages, namely, more flexibility in the power waveform, process stability, and low build-up of powder macroparticles.²³ 

An important milestone in pulsed arc discharge nanosynthesis is to minimize energy investment per nanoparticle generated. Cost-effective synthesis aims at mass production of ultra-pure nanomaterials with lowest power consumption. The approach in the present article consists of comparing the power management issues between standard DC anodic arc discharge and low-frequency pulsed arc discharge. Such a topic illustrates one of the advantages shown by pulsed arcs for nanomaterial growth over conventional DC arc discharge. In this study, the considered materials are carbon nanostructures (graphene, carbon nanotubes) and few layers of MoS₂. The supplied power density in each arc experiment has been evaluated and compared with the average energy delivered to the gas species. To this end, the plasma volume from arc discharges has been correlated with the flux of ablated material and the consumed power. The energy invested per grown nanoparticle has been confronted with typical CVD yields. Finally, the propagation of carbon and MoS₂ species within the arc chamber was analyzed in situ. For this, the generated nanoparticles were extracted using a high-speed probe aimed to spatially resolved measurements. The chemical bonding and morphology of the samples were characterized by Raman spectroscopy and scanning electron microscopy (SEM), respectively.

The arc discharge experiments were performed in a plasma arc setup described thoroughly elsewhere [Fig. 1(a)].^22,24 In brief, the discharges were held between two vertically aligned electrodes, which were located at the axis of a cylindrical vessel of stainless steel (4500 cm³). The upper electrode acted as a cathode and consisted of a graphite bar of 10 mm in diameter. The lower electrode, which acted as an anode, was either a graphite solid bar (3 mm in diameter) for carbon discharge or a 1 mm-thick graphite tube (3 mm of inner diameter) filled with densely packed MoS₂ powder (purity: 99%) with grains of 1 μm in average size for MoS₂ discharge.¹⁵ The vertical position of the anode was regulated by means of a linear drive. In this way, arc discharges were ignited by separating the electrodes upon energizing the system with electrical power. All experiments were carried out in a helium atmosphere (purity: 99.995%) at 300 Torr. The arc discharges were fed by a Miller Gold Star SS300 DC power supply, which was operated remotely using an external signal generator. Hence, rectangular waveforms of arc current were programmed with pulse frequencies of 1, 2, and 5 Hz, and a duty cycle of 10% (respective ON times of 100, 50, and 20 ms). The temporal evolutions of discharge voltage and current were measured with an oscilloscope. Images of the arc discharges were captured using a digital video camera working at 15–25 frames per second. A strongly absorbing filter was installed at the chamber view port to reduce the excessively bright arc optical emission.

A horizontally oriented cylindrical probe was placed next to the electrode system. The probe was formed by a tungsten wire with 0.5 mm in diameter and 3 cm in length, which was mounted on a high-speed actuator [Fig. 1(b)]. The angular motion of the fast probe was driven by a second signal generator in order to extract the material from the discharge in synchrony with the pulse phase. A dwell time of 30 ms was set to minimize the material collecting time during the probe experiments and thus minimize the exposition to the thermal flux from the hot plasma. Additionally, a vertical plate with a slit for the probe was used as a divider to reduce undesired deposition onto the wire while it was retracted. The distance x between the probe tip and the electrode axis was 5 mm in the forward position. Further details regarding the probe operation are provided in a recent article by Corbella et al.²⁴ 

The chemical structure of the carbon and MoS₂ layers deposited on the probe was characterized using a Raman spectrometer Horiba LabRAM HR operated at a wavelength of 532 nm. The spot of the incident laser beam, which determined the spatial resolution of the probing area, was a few micrometers in size. Morphology of the samples was studied with a scanning electron microscope (SEM) Tescan XEIA-FEG SEM operated at accelerating voltages of 5 and 10 kV.

Figure 2(a) shows the arc current and arc voltage waveforms measured during pulsed MoS₂ discharge carried out at 2 Hz and 10% duty cycle (50 ms of ON time). Typically, the peak current reached values of around 250–300 A, whereas the arc voltage showed peak values between 30 and 60 V in carbon and MoS₂ discharges.^15,22 During the inactive period, out of the ON time, the arc was not extinguished but behaved as a constricted arc with minimal current (approximately 10 A) and voltage (15–20 V).^23,25 The curve of instantaneous power has been included in the figure. The associated peak power reached normally about 10 kW. However, the average power per pulse over one period did not exceed 1 kW in any of the experiments. In general, DC arc discharges are characterized by power values that double the pulsed arc power.²² The current and voltage waveforms are accompanied by an image of the light emitted by a MoS₂ discharge captured during the ON phase [Fig. 2(b)]. In that particular moment, the discharge region was expanded to the largest volume in an ellipsoidal shape. Similar emission patterns were observed in pulsed carbon discharges (see the supplementary material). The largest discharge volume, V_dis, was estimated by measuring the ellipsoid axis lengths a and b, defined in Fig. 2(b), and applying the expression V_dis = 4/3 × π × (a/2) × (b/2)².

The measured peak power values and discharge volumes in arc experiments conducted at different discharge currents are listed in Table I. Arc discharges were held in DC and pulsed modes. However, MoS₂ discharges were only run in pulsed mode to enhance powder evaporation and, simultaneously, minimize erosion of the graphite container (weak arc attachment on the anode sidewall).¹⁵ The peak power for each experiment was calculated by multiplying the corresponding peak values of arc voltage and current: P_peak = V_peak × I_peak. The volumetric power density is the ratio of P_peak over V_dis. On the other hand, P_peak divided by the anode top surface area provides the surface power density. Power densities span between 10 and 20 kW/cm³ for DC arcs, while it is only approximately 1 kW/cm³ in pulsed arcs, indicating a reduction of power density by one order of magnitude.

The average ablation rates were calculated by considering the anode mass, measured with a microbalance, before and after each arc process. Rates between 2 and 20 mg/s were evaluated in carbon DC arcs (non-constricted). In the case of pulsed discharges (approximately 1 mg/s), the average values were divided by the duty cycle, 0.1, to obtain the peak rates of evaporation during the pulse.²² The average evaporation rate in MoS₂ arcs, 75 mg/s, matches with reported rates from other compound powder-graphite anodes.²⁶ The temperature at the carbon arc core, T_arc, was measured by optical emission spectroscopy (OES) of carbon arc discharges,²² and it has been assumed to be similar in the case of MoS₂ discharge. The precursor flux during the pulse, J_prec, is obtained from the peak ablation rate values and considering 12 u and 160 u as elementary masses for C atoms and MoS₂ molecules, respectively. The helium diffuse flux, J_He, is obtained from the He gas pressure, p_He= 300 Torr, and T_arc by assuming a kinetic regime that follows the ideal gas law:

Here, M_He = 4 u is the mass of the He atoms and κ = 1.38 × 10⁻²³ J/K is the Boltzmann constant. Finally, the average energy density or specific energy (eV/particle) is computed by dividing the power density on the anode top surface (kW/cm²) by the total flux of precursor species and He atoms, J_prec + J_He (cm⁻² s⁻¹).

Figure 3 compares the volumetric power density with the specific energy per particle (surface power density over total flux) measured in each arc experiment. Pulsed arc discharges take more advantage of supplied power (1 Hz: 17 eV/particle and 0.7 kW/cm³) than DC discharges do (60 A: 3.1 eV/particle and 8 kW/cm³; 150 A: 13 eV/particle and 20 kW/cm³). In consequence, the electrical parts submitted to pulsed power undergo a lower thermal flux than in the DC counterpart. Also, the punctual power bursts enable significant material evaporation at relatively low energy cost. The process at 5 Hz constitutes an exception because it shows a lower specific energy (12 eV/particle) in parallel with a higher power density (2.3 kW/cm³). This energy situation could be ascribed to working in conditions comparable to DC arcs, probably because temperature variations underwent by the anode at higher frequencies are smaller than those experienced at 1 or 2 Hz. In this sense, the ablation performance might approach steady state, DC operation if the thermal inertia of the electrode materials masks the effects of rapidly changing arc currents. The constricted DC arc (10 A: 0.25 eV/particle and 14 kW/cm³) constitutes a particular case since the ablation is almost negligible (0.01 mg/s) and occurs in a very localized discharge (0.015 cm³).

As explained below, pulsed arcs promote higher ionization rates than DC arcs, and thermal load in pulsed processes is reduced because of a power distribution occurring within a larger discharge volume than in DC arc cases. The first ionization energies of C and Mo atoms are 11.3 eV and 7.1 eV, respectively. Since pulsed carbon arc experiments show specific energies that surpass the C ionization threshold, total ionization or at least relevant ionization rates are expected if a uniform energy distribution is assumed. Indeed, ionization degrees close to unity were estimated in previous measurements of ion and electron currents in pulsed carbon arcs (electron density higher than 3 × 10¹⁶cm⁻³).²⁵ The molybdenum atoms released from MoS₂ pulsed discharges are expected to show lower ionization rates because the Mo ionization threshold is not achieved. However, the delivered specific energy, 3.3 eV/particle, is high enough to dissociate an important fraction of MoS₂ molecules (Mo–S bond energy: 2.56 eV).²⁷ 

SEM analysis on the carbon nanotubes deposited onto the cathode surface provided a surface density of around 10⁹cm⁻² in pulsed carbon arc discharges. It can be demonstrated that such a density is equivalent to a production yield of 10¹¹cm⁻²kWh⁻¹ in our experimental conditions (see the supplementary material). On the other hand, average yields of 10¹²–10¹³cm⁻²kWh⁻¹ have been achieved for vertically aligned carbon nanotube forests prepared by CVD.^9,10 Nevertheless, in contrast with the PVD arc processes, the reported CVD experiments required using heated-up substrates (between 400 °C and 700 °C) with especially tuned surfaces (deposition of metal catalyst film) for an optimal nanoparticle growth. These are compelling arguments to state that arc discharge processes could be competitive with CVD yields upon improvement of substrate conditions and selective guiding of the generated nanoparticles, by using, for example, magnetic fields for electron and ion manipulation or by designing special arc waveforms to control the discharge region and gradients of plasma temperature and plasma density.

The nanoparticle distribution within arc discharge volume was captured by means of a fast probe experiment. In a pulsed carbon arc trial at 1 Hz, the probe was facing the discharge for 30 ms during the arc current plateau region of the ON phase. A total of 20 consecutive exposures at the same time interval were periodically performed within the experiment to improve statistics. Also, complementary probe experiments were carried out by setting the exposure time outside of the ON phase, i.e., during the inactive phase of the pulse. The deposition on the probe in such experiments was negligible, proving thereby that nanoparticles were exclusively generated during the ON time. In MoS₂ experiments, the probe was left at rest at the vicinity of the arc column to avoid electrostatic shielding of the arc (the probe was covered with the semiconducting material quickly). The middle point of the probe was located 7 mm away from the electrodes axis.

Figure 4(a) shows a representative Raman spectrum of the carbon layer deposited at x = 6 mm on the probe after the 20-pulses collection experiment. The parameter x denotes distance to the arc core. The characteristic Raman bands of carbon are identified. The G peak (approximately 1550 cm⁻¹) is ascribed to graphite growth, whereas the D peak (approximately 1350 cm⁻¹) accounts for the presence of defects in the sample.²⁸ The G* band (approximately 2450 cm⁻¹) is an overtone of the LO phonon and can be observed as an index of defect-free crystallinity. Finally, the G′ or 2D band (approximately 2700 cm⁻¹) reflects the presence of graphene flakes.²⁹ The G′/G intensity ratio, which can be considered as a marker of few-layer graphene platelets,³⁰ is plotted against distance to arc center in Fig. 4(b). Initial G′/G ratios around 0.6 start to decline at around x = 8 mm down to 0.2. The lower abundance of graphene at x > 8 mm is consistent with the expected steep gradient of plasma temperature, which limits the arc region of effective nanoparticle generation. The spatial evolution of G′/G ratio in pulsed arc is compared with data extracted from previous DC arc experiments conducted by Fang et al.³¹ The fact that the G′/G ratio remained at the interval 0.1–0.2 in DC experiments suggests that pulsed arc discharges enhance the propagation of nanomaterials compared to traditional DC arcs. According to a fluid simulation study of carbon arc discharges, the dominant transport mechanism of gas species in radial direction is expected to be convection associated to the natural expansions of helium gas and carbon vapor driven by the arc pulses.³² Recently, time- and spatially resolved captures using the fast probe confirmed the relevance of convective transport in the propagation of carbon nanoparticles in pulsed arcs.²⁴ The SEM image from x = 6 mm in Fig. 4(c) shows the aggregate morphology of graphene flakes stacked as platelets. The structural disorder, suggested by the D/G peak ratio close to unity, is probably due to defects introduced by the boundary edges of the graphene platelets. Such defects might be attributed to the repeated exposures in the plasma arc, which could lead to a disordered stacking of graphene flakes.

In summary, the expansion of the graphene deposition region in pulsed arc processes is correlated to the increase of the discharge volume, thereby suggesting an enhanced transport of nanoparticles from the arc column toward the chamber walls. Such a propagation might be explained by particle drag driven by the sudden expansion of helium gas and carbon vapor during the pulse onset.

Figure 4(d) shows a Raman spectrum corresponding to MoS₂ deposited on the probe at rest. The examined distance to arc core is x = 8 mm in a pulsed arc trial performed at 2 Hz and 10% duty cycle for 8 s (16 pulses). The characteristic Raman bands, assigned to E¹_2g (approximately 385 cm⁻¹) and A_1g (approximately 410 cm⁻¹) modes, appear separated by a frequency shift of Δω_layer = 23–24 cm⁻¹. Such a separation is an index to quantify MoS₂ elementary layers (atomic trilayers).³³ In the present case, stacks of MoS₂ layers with between 3 and 5 units have been detected. Complementary characterization of the arc-deposited MoS₂ samples has been reported elsewhere: the inter-layer thickness (approximately 0.65 nm) was confirmed by atomic force microscopy (AFM) measurements, while the chemical composition and crystallographic properties were assessed by electron microscopy analysis.¹⁵ The spectrum corresponding to the MoS₂ powder precursor is added in Fig. 4(d) as an example of bulk material Raman profile (Δω_bulk≈ 25 cm⁻¹). In conclusion, few-layer flakes of MoS₂ can also be produced in arc volume and have enough mobility to be efficiently transported toward a substrate in the arc vicinity.

The power distributions in DC and pulsed anodic arc discharges have been studied by comparing the patterns of emitted light in each experiment. Volumetric power densities are relatively large in DC arc processes (10–20 kW/cm³), whereas values characteristic of pulsed arc discharges are lowered by one order of magnitude. The specific energies per atom or molecule, though, are higher in pulsed arcs (approximately 15 eV/particle) than in DC discharges (approximately 10 eV/particle or less). Concerning the evaporated atoms, the aggregates or clusters that were transformed into nanomaterials could be transported through the plasma phase more efficiently in pulsed arc discharges thanks to the periodical events of hot gas expansion. The improved transport properties, supported by gas convection, increase the reach of the generated nanoparticles so that substrates may be placed farther from the arc column and still receive a significant particle flux. In conclusion, nanosynthesis by means of anodic arc discharges in low-frequency pulsed mode offers more rational energy consumption compared with the DC analog, which is translated into higher ionization rates, more efficient energy transfer from electrical power to particle mobility, and the possibility to coat delicate substrates by enabling a remote arc plasma source.

The present study provides approximate but significant values (within 15% of uncertainty) of power and energy densities, which were obtained by applying simple energy balance considerations on the evaluated electrical parameters, optical emissions, and ablation rates. In order to obtain energy values showing higher accuracy, time-resolved measurements of discharge volumes and erosion rates within an arc pulse are planned.

See the supplementary material for (1) images of the light emitted by DC and pulsed carbon arc discharges and (2) comparison of nanoparticle production yields of pulsed anodic arc discharge and standard CVD technique.

This work was supported by the U.S. Department of Energy (DOE), Office of Science, Fusion Energy Sciences program (Award No. DESC0015767) and by the National Science Foundation (NSF) (Grant No. 1747760). The authors acknowledge the assistance in SEM analysis by Dr. Jiancun Rao and Dr. Sz-Chian Liou from AIMLab at the Maryland NanoCenter.

The data that support the findings of this study are available from the corresponding author upon reasonable request.