Low-NO x thermal plasma torches: A renewable heat source for the electrified process industry

Industrial thermal plasma torches can heat a gas up to 5000 – 20,000 K, i.e., well above the temperature needed to replace the heat generated from the combustion of traditional fossil fuels (e.g., coal, oil, and natural gas) in large-scale process industry furnaces producing construction materials (e.g., iron, steel, lime, and cement). However, there is a risk for significant NO x emissions when air or N 2 are used as plasma-forming gas since the temperature somewhere in the furnace always will be higher compared to the threshold NO x formation temperature of ~1800 K. Torch NO x forms inside the high temperature region of the plasma torch ( > 5000 K) when air is used as gas. Process NO x forms instead when the hot gas (when air or nitrogen is used as plasma forming gas) from the plasma torch mixes with process air downstream the torch. By analysing the complex chemistry of both the torch-and process NO x formation with thermodynamic equilibrium and one-dimensional chemical kinetic calculations it was shown that adding H 2 to the plasma-forming N 2 gas significantly reduces the NO x emissions with more than 90 %. Verifying experiments with air, pure N 2 , and mixtures of H 2 and N 2 as plasma-forming gas were performed in a laboratory scale insulated laboratory furnace with different pre-heating temperatures of process air (293, 673, and 1073 K) which the plasma gas mixes with downstream the torch. Depending on the pre-heating temperature the NO x emissions were between 12,000 – 14,000 mg NO 2 /MJ fuel when air was used as plasma forming gas. Substantial NO x emission reduction occurs both when N 2 replaces air, where the NO x emissions was in the span of 8000 – 11,500 mg NO 2 /MJ fuel and furthermore when H 2 was mixed into the N 2 gas stream. For the highest degree of H 2 mixing (28.6 vol%), the NO x emissions were between 450 – 1700 mg NO 2 / MJ fuel depending on the pre-heat temperature of the process air, i.e., a reduction of 88 – 96 % and 85 – 94 %, respectively when air or N 2 was used as plasma forming gas. The measured NO x emissions are then of the same order of magnitude as would be expected from the combustion of traditional fuels (coal, oil, biomass and pure H 2 ). Finally, by analysing the aerodynamics in an axisymmetric furnace with an experimentally validated computational fluid dynamics (CFD) model using reduced chemistry for the NO x formation (19 species and 70 reactions), further guidelines into the process of NO x reduction from thermal plasma torches are given.


Introduction
Industrial thermal plasma torches [1,2] are currently used as an energy source for waste treatment [3,4], solid-fuel and waste gasification [5][6][7], cutting [8], welding [9], metallurgical applications [10], and advanced nanomaterial production [11].In the near future, plasma torches may play a key role in the necessary electrification and decarbonization [12] of the construction material (e.g., iron, steel, and cement) production industry, providing that they are powered by renewable electricity [13], as illustrated in Fig. 1a.The energy demand in this industrial sector is high, and the CO 2 emissions from fuels and raw material (e.g., limestone) accounted for 16.0 % of global anthropogenic CO 2 emissions in 2020 [14].Furthermore, these processes are especially hard to decarbonize [15], due to the extremely high furnace temperatures needed for various phase-conversion processes (e.g., the induration of iron-ore pellets [16], the heat treatment [17] and hot working [18] of steel, and the calcination [19] and sintering [20] of limestone), now mainly created by the combustion of fossil fuels (see Fig. 1b).Typical thermal plasma torches can heat the plasma-forming gas to 5000-20,000 K [3] with an average plasma enthalpy of 1-100 MJ/kg [21], significantly higher than the enthalpy of the gas formed during fossil fuel combustion, which is ~ 2-3 MJ/kg in stoichiometric combustion.In the non-transfer arc configuration [22], the plasma jet leaves the torch like a free flame generated during the combustion of traditional fossil fuels (e.g., coal powder, fuel oil, and natural gas), and is therefore of particular interest as an alternative heating technology.Furthermore, since the plasma-heated gas does not contain any significant amount of particulate matter, the harmful particle emissions [23] and ash-related operational problems in industrial furnaces [24][25][26] powered by fossil fuels could also potentially be avoided in the future.
The main challenge is the presumed significant production of the harmful [27] and heavily regulated [28,29] nitrogen oxides (NO x ) formed inside the torch, if air is used as the plasma-forming gas (torch NO x ), or downstream of the torch where the hot gas mixes with the process gas (i.e., process NO x ) (see Fig. 1c).The threshold temperature for significant NO x formation from colliding N and O atoms and/or molecules is ~ 1800 K [30], i.e., substantially below the high temperature in a thermal plasma torch.Thermal plasma torches operate under local thermodynamic equilibrium [31], so the gas composition can be estimated using thermodynamic equilibrium calculations (TECs) (see Fig. 1d for the contributions of N and O radicals and NO and Fig. S1 for all significant gas components).Fig. 1d shows that the plasma NO x emissions are typically several orders of magnitude greater than those from the combustion of traditional fuels, which is unacceptable from an environmental perspective and would necessitate extensive and costly NO x flue gas cleaning [32] to fulfill environmental requirements [28,29].High NO x emissions from thermal plasma torches using air as the plasma-forming gas have recently been verified experimentally [33] (see Fig. 1d).Direct NO x reduction strategies (i.e., combustion optimization) during the combustion of traditional fuels have been thoroughly investigated [34,35].
To our knowledge, the concept of NO x reduction for industrial thermal plasma torches has not previously been investigated, and other studies of thermal plasma torches have had different focuses (e.g., see ref [3,4] and references therein).This may be because there is no significant NO x formation under the reduced/inert furnace conditions most often used in waste treatment, gasification, and nanomaterial production or because the scale of these units (e.g., used for cutting and welding) is too small in terms of yearly NO x production.Furthermore, the electrification of the construction material production industry has just started, for example, in Sweden with recent efforts addressing renewable iron and steel (the hydrogen breakthrough ironmaking initiative (Hybrit) [36]) and cement production (the CemZero initiative [37]).The relevant industrial production units are very large (i.e., MW scale) and operate continuously (i.e., 24/7).Replacement of current hydrocarbon burners with thermal plasma torches would therefore lead to significant yearly NO x production from the plasma torches unless the NO x issue is addressed.Using inert gas (e.g., N 2 ) as the plasma-forming gas might reduce the formation of torch NO x .However, industrial facilities (e.g., rotary kilns and induration machines) operate at high temperatures and contain moving parts, i.e., they are not gas tight, so the formation of process NO x may always be a problem.The question is whether it is possible to develop a universal strategy for operating thermal plasma torches with NO x emission levels similar or lower to those from burning traditional fuels.This development may open the door for the large-scale deployment of environmentally friendly low-NO x thermal plasma torches as energy sources in industry.
With this contribution, this question was answered positively.By first analysing the complex chemistry of both the torch and process NO x formation by employing thermodynamic equilibrium (i.e., torch NO x formation) and chemical kinetic (i.e., process NO x formation) calculations incorporating the full reaction mechanism, a NO x reduction strategy was developed based on adding H 2 to the plasma forming N 2 gas.Significant NO x reductions were then obtained under industrially relevant experimental conditions (i.e., combination of a plasma torch and air pre-heated up to 1073 K) in a well-insulated laboratory scale furnace.Finally, using an experimentally validated computational fluid dynamics (CFD) model incorporating reduced chemistry, further guidelines for NO x reduction during the operation of thermal plasma torches was suggested.The NO emissions are significantly higher than those of traditional fuels and must be reduced significantly before thermal plasma torches can be introduced as a heating source in the process industry.

Preheated air co-flow reactor
The preheated reactor consisted of a plasma torch installed vertically at the bottom of an L-shaped chamber into which the preheated co-flow air enters [38].A high-temperature plasma jet was generated using an 18-kW direct current (DC) plasma torch (model PNIX-100; Plasnix, Incheon, ROK).The reactor consists of a square-shaped chamber with an internal cross-section of 25 × 80 cm, made of steel and internally lined with ceramic fibre.The heated co-flow air was injected from the side (perpendicular to the plasma jet) via a horizontal channel with the same cross-section and lining as the main vertical chamber.The electrical heater (model 5000 HT; Leister Technologies, Kaegiswil, Switzerland) used to preheat 600 normal liter per minute (NLPM) of co-flow air to 1073 K was mounted centrally at the far end of the side channel.The flow rate of the plasma-forming gas for each gas mixture was set to deliver 9 kW of thermal power to the chamber.The experimental conditions are summarized in Table S1.Note that the electrical power is the net value, which differs from the set value of the power source due to heat losses from the burner elements to the cooling water.The heat losses to the cooling water were calculated to ~ 20 %.This calculation is based on the known flow rate of water and the temperature difference between the inlet and outlet water flows measured with K-type thermocouples.
The temperature and composition of the gas leaving the chamber were measured 5 cm upstream of the chamber exit using a fine wire Stype thermocouple and Fourier transform infrared spectroscopy (FTIR) (2000 MultiGas analyser; MKS Instruments, Andover, MA, USA), respectively.

Axis-symmetric reactor
The main aim of building up the axis-symmetric reactor was to create an experimental data set that could be used to validate the developed CFD model.An axis-symmetric geometry was used since it saves computational time in the CFD simulations.The 18-kW DC plasma generator (model PNIX-100; Plasnix) was installed at the bottom of the reactor, which consists of an inner cylinder separating the co-flow (process) air and the plasma jet as well as an outer chamber.See Fig. S1 for schematics and the dimensions of the reactor.The outer chamber is a cylindrical 2-mm-thick steel shell with an inner diameter of 125 mm and a length of 290 mm before contracting to an inner diameter of 70 mm through a 30 • contraction towards the end.The inner cylinder is made of 4-mm-thick steel with an inner diameter of 70 mm.Experiments were performed with two lengths of the inner cylinder, i.e., 100 mm (short inner cylinder [SIC]) and 200 mm (long inner cylinder [LIC]).To achieve an axisymmetric and uniform co-flow profile, the volume between the cylinder and chamber was filled with glass balls (2 mm diameter) to a height of 60 mm.The composition of the gas leaving the chamber was analysed using FTIR (2000 MultiGas analyser; MKS Instruments).The temperature was measured at the reactor outlet and at two points in the glass bed, i.e., at the uniform co-flow inlet and on the outer surface of the cylinder.
In all experiments, the volume flow rate of the co-flow air was 340 NLPM.The flow rate of the plasma-forming gas for each gas mixture was set to deliver 9 kW of thermal power at an average plasma enthalpy of ~ 7.5 MJ/kg.The operating conditions of each measured case are summarized in Table S2.

Thermodynamic equilibrium calculations (TECs)
Assuming that the plasma flow exiting the torch is (i) thermally and chemically homogeneous, (ii) steady state, and (iii) in thermal and chemical equilibrium, its state, i.e., composition and temperature, can be estimated using equilibrium calculations.This is a valid assumption in the middle of a plasma jet where no walls disturb the flow [39].This narrows the factors affecting torch NO x to the average plasma enthalpy, h p , determined as the ratio of the thermal (net) power of the plasma torch, p_{net}, to the flow rate of the plasma-forming gas, G, and the composition of the plasma gas.Here the open-source Cantera code [40], which uses an element potential method to find the chemical equilibrium state was used for the calculations.By setting h p and the plasma gas composition and holding the specific enthalpy and pressure of the torch constant, the equilibrium state was calculated.

One-dimensional chemical kinetic calculations
Although specific CFD simulations are required to calculate the process NO x emissions of a plasma torch, simple 1D calculations can be used to compare the NO x levels of a torch when using different plasmaforming gases.The counterflow diffusion flame (D-CFF) model in Cantera code [40], in which equations representing a steady axisymmetric stagnation reacting flow are solved for a 1D domain consisting of opposed plasma and oxidizing inlets was used for the calculations.The D-CFF model has often been used to study the evolution of the mixing field between jets in 2D geometry, such as in a 2D coflow diffusion flame [41,42], which has a structure similar to a plasma jet.The reaction mechanism used in this study is Gas Research Institute (GRI) 3.0 which is a detailed chemical kinetic model, with 53 species and 325 elementary reactions, designed for accurate modeling of natural gas combustion [43].It was chosen as it is optimized against a wide range of experimental data, and it includes nitric oxide (NO) formation and reburn chemistry.Using this model and assuming that (i) the plasma jet is in equilibrium with a volume flow rate of 60 NLPM and a total thermal power of 9 kW, (ii) the oxidizer inlet air is at 800 K with a volume flow rate six times that of the plasma one, and (iii) the validity of GRI 3.0 [43] reduce the factors affecting NO x formation in the D-CFF domain to the mass flux of two inlets, which determines the mean strain rate of the flame.

CFD modeling
Velocity components, temperature, and the composition vector inside the domain are obtained by solving a steady-state Reynolds-averaged formulation in which the Favre (i.e., mass-weighted)-averaged Navier-Stokes equations for the conservation of mass, momentum, and energy, with the evaluation of Favre-averaged quantities denoted by f = fρ/ρ, where the over-bar denotes the mean component of a quantity, are described as follows: In this system of partial equations, u is the velocity, ρ density, p pressure, Y k mass fraction of species k, h s sensible enthalpy, λ thermal conductivity, T temperature, D k species diffusivity, R k species net production rate, and Δh 0 f,k specific formation enthalpies of species k.The viscous heating and pressure-velocity correlation terms are neglected in Eq. ( 2).The species mass flux is described by Fick's law in the first term on the right-hand side of Eq. ( 3).The turbulent diffusivity may therefore be expressed as the ratio of the turbulent viscosity, µ t , and Schmidt number, Sc t , which is assumed here to have a constant value of 0.7.The dynamic mixture viscosity, µ, and thermal conductivity are calculated using Sutherland's law, and the mixture is considered a Newtonian fluid with a density following the incompressible ideal gas law.
To close this system of equations, the Reynolds stresses, ρ ũʹ i u ʹ j , are modeled using the standard k-ε model [44].Assuming the absorption, emission, and scattering of radiation by the medium to be negligible, the surface-to-surface approach [45] is used to model the radiation source term, S R , in Eq. 7. The Favre-averaged species reaction rates, R k , are modeled using the eddy-dissipation concept (EDC) [46], according to which detailed Arrhenius chemical kinetics may be incorporated in turbulent flows.A two skeletal mechanisms were used to model NO x formation as well as H 2 oxidation.A mechanism with seven species and eight reactions was used for the air and nitrogen cases, while N 2 :H 2 simulations were performed using a mechanism with 19 species and 70 reactions.The former was developed by excluding GRI 3.0 reactions that contain C and H species, while in the latter only reactions with C species were excluded.
The axis-symmetric reactor, see Fig. S1, was modelled as a 2D axissymmetric domain.The domain consists of (i) a fluid zone including plasma torch exit, co-flow inlet (located immediately after glass balls) and the chamber outlet and wall and (ii) a solid zone for modeling conjugate heat transfer in the inner cylinder.The computational domain discretized by approximately 8.8 × 10 4 quadrilateral dominant cells.A mesh sensitivity study, performed by refining the grid with a cell count of 3.52 × 10 5 , showed negligible sensitivity to grid refinement for the chosen mesh (8.8 × 10 4 cells).Plasma inlet temperature and composition were imposed using equilibrium calculations and fully developed turbulent velocity profiles.Co-flow inlet was set by uniform profile representing mass flow controllers and measured temperatures.Turbulence boundary conditions were imposed based on the known hydraulic diameters and estimated turbulence intensities.No wall functions were used, proper y + values were ensured using boundary layers.For the chamber wall a combined convection and external radiation was set with a convective heat transfer coefficient of 5 W/m 2 K.The temperature of the wall of the inner cylinder boundary was set according to the thermocouple measurements.

Reducing NO x production by thermal plasma torches: theory
Using pure N 2 instead of air as the plasma-forming gas in plasma applications might initially be seen as an attractive way to reduce the total formation of NO x .However, the TECs of the plasma gas leaving the torch indicate gas temperatures higher than those with the use of air (Figs.S2 and S3) and significant N radical production (Fig. 2).Furthermore, detailed 1D chemical kinetic calculations indicate significant process NO x formation when the hot plasma gas mixes with the surrounding air (Fig. 3), due to the reaction of N radicals with O 2 (N+O 2 → NO+O).
Introducing H 2 to the N 2 plasma-forming gas has positive effects.TECs made using a mixture of 25 vol% H 2 and 75 vol% N 2 indicate (Fig. 2) that the H 2 not only significantly decreases the plasma temperature, especially for h p > 4 MJ/kg, but also produces fewer N atoms than does the pure N 2 .The lower temperature of the N 2 :H 2 plasma (Fig. 2) is associated with the higher specific heat capacity and weaker atomic bonds of H 2 compared with N 2 .A large fraction of the arc energy is used for converting H 2 molecules to H radicals rather than for heating the working gas inside the torch.The remaining energy is insufficient for the formation of N atoms, even in high-enthalpy operations (Fig. 2).Note that small amounts of NH, NH 2 , and NH 3 radicals are formed by the arc at high temperatures, T>3500 K (Fig. S4).
Another benefit of adding H 2 to the N 2 plasma is that the gas (consisting partly of H 2 and H, Fig. 2) can be further oxidized with air downstream of the torch.This enables torch operation at lower enthalpy but with the same thermal power of the whole system, in turn permitting even lower plasma jet temperatures and the further reduction of N radical production.This is illustrated by gray vertical lines in Fig. 2. By using combustible gas, the torch can produce the same thermal power (9 kW) with only 5.45 kW of electric energy, resulting in a plasma jet with an approximately 2500 K lower average temperature (h p = 4.6 MJ/kg).The rest of the thermal power (3.55 kW) is supplied by the oxidization of H 2 and H downstream of the torch.
The formation of process NO was investigated using 1D chemical kinetic calculations.Fig. 3a shows a sketch of the computational domain in the counterflow geometry made to compare the NO levels of a torch using different plasma-forming gases: air, N 2 , and 25:75 vol% H 2 :N 2 .Calculations were performed using the counterflow diffusion flame (D-CFF) model in Cantera code [40] and the GRI 3.0 chemical mechanism [43].It was assumed that the plasma jet is in equilibrium with a volume flow rate of 60 normal liter per minute (NLPM) and a total thermal power of 9 kW.The oxidizer inlet (Fig. 3a) air was 800 K at a volume flow rate six times greater than that of the plasma gas.Calculations were made for a mean strain rate of 300 1/s.The average electric enthalpy was 7.7 MJ/kg for the air or nitrogen plasma but only 4.6 MJ/kg for the H 2 :N 2 mixture, as the remainder of the latter was compensated for by the heating value of H 2 .
The axial velocities and mixture fraction profiles shown in Fig. S5a  and b indicate that, at this strain rate and mass flux ratio, the stagnation planes for all jets are located near the plasma inlet where the mixing and, consequently, reactions are significant at 2.5 mm < x < 7 mm.As shown in Fig. 3b, when nitrogen or the H 2 :N 2 mixture replaces air as the plasma-forming gas, the plasma stream enters the domain at a much higher or lower temperature.
The profiles of the net production rate of NO (R NO ) due to the Zeldovich mechanism reactions (all other NO reduction and formation pathways are negligible and will be briefly discussed below) compared in Fig. 3c show that the NO formation for the air plasma starts at around 3 mm and continues up to approximately 5.8 mm (due to the reactions N 2 + O ⇌ NO+O (R1) and N+O 2 ⇌ NO+O (R2), see Fig. 3c and S6a), although the direction of these reactions changes at greater distances.The reaction path diagram of N flux at maximum net production rate for air plasma is presented in Fig. S7, the figure shows that an additional route to form NO through the N 2 O reactions is about 2 % of the NO produced through the Zeldovich reactions.The NO formation for the N 2 plasma displays a distinct concave-convex shape between 3.8 and 6 mm.consumed through N 2 O reactions for the N 2 plasma is below 3 % of total NO produced via the Zeldovich reactions.The R NO profile for the H 2 :N 2 mixture has a pronounced convex shape over the reaction zone between 4 and 6 mm (see Fig. 3c) and is mainly governed by the forward progress of R1 and R3 (N + OH⇌NO + H) in the extended Zeldovich mechanism, as shown in Fig. S6c.A reaction path diagram of N flux at maximum net production rate of NO for H 2 :N 2 plasma can be found in Fig. S9.Apart from the 1.6 % contribution from N 2 O reactions, NO formation is slightly increased by about 5.5 % due to oxidation of NH, HNO, and NNH.The total process NO formation rate, calculated by the integral of R NO over the whole domain, is 7.5, 17.4, and 3.0 kmole/m 3 s for the air, N 2 , and H 2 :N 2 plasmas, respectively.Nonetheless, the total NO formation for each plasma gasthat is, torch NO x plus process NO xcompared in Fig. 3d shows that replacing air with N 2 reduces the total NO emissions of the torch by a factor of two, while replacing N 2 with the H 2 :N 2 mixture results in a reduction of one order of magnitude.

Reducing NO x from thermal plasma torches: Experiment
Industrial processes often use highly preheated air [47], so verifying experiments were performed in an insulated laboratory furnace with different temperatures of preheated air (600 NLPM at 293, 673, or 1073 K) as the gas and an 18-kW plasma torch (see Fig. 4a).Experiments were performed with air, pure N 2 , and the H 2 :N 2 mixture at different mixing ratios as the plasma-forming gas.The thermal power of the plasma torch was reduced when H 2 was mixed with N 2 , so the total thermal power was ~ 9 kW for all experiments.The experimental conditions were chosen to mimic the conditions inside the furnace of an iron-ore-pellet induration machine.This is probably one of the most challenging industrial processes to electrify using low-NO x thermal plasma torches due to the high temperature of the preheated air and the large amount of excess air in the furnace [47], needed for both the oxidation and sintering of the iron-ore pellets.Depending on the initial temperature of the co-flow air, the outlet temperature of the gas after mixing with the gas from the plasma torch increased to ~ 830, ~1030, and ~ 1200 K, respectively.The experimental protocol and experimental results, i.e., outlet-gas temperature and NO x concentration, can be found in Table S1.
Fig. 4b presents the gas emissions of NO x normalized to mg/NO 2 MJ/ fuel.Similar to the results of conventional burners [48], the NO x emissions increase when using preheated air, and for the H 2 :N 2 gas mixture are ~ 2000 mg/MJ higher at an air temperature of 1073 K than at 293 K. Substantial NO x emission reduction first occurs when pure N 2 replaces air, regardless of the initial preheated air temperature.The gradual addition of H 2 to the N 2 plasma gas leads to further NO x reduction.For the highest degree of H 2 mixing (i.e., 28.6 vol%), the reductions are 88-96 % compared with air and 95-85 % compared with pure N 2 , depending on the preheated air temperature.The highest reduction was achieved at a preheated temperature of 20 • C.
Wiinikka et al [49].investigated NO x emissions during the combustion of traditional fossil fuels (i.e., fuel oil and coal), biomass (i.e., bio-oil, stem wood powder from pine and spruce, and steam-explosion-treated wood powder), and electrofuel (i.e., hydrogen gas) in a ~ 150-kW pilot-scale kiln with highly preheated air (i.e., 1073 K for gaseous and liquid fuels and 1173 K for solid fuels), simulating the hot sintering zone of an iron-ore induration machine.The results of that investigation are shown in Fig. 4b for comparison.Furthermore, the NO x emissions for coal (i.e., 870 mg/MJ) are comparable to those from other similar pilot-scale investigations performed at higher fuel loads (400-580 kW) with coal as fuel [50], showing consistencies between different pilot-scale experiments.Compared with pure H 2 combustion, which has the highest NO x emissions (~4200 mg/MJ), ~18 vol% of H 2 was needed to be injected in the N 2 stream to reach the same NO x emissions for the thermal plasma torch (see Fig. 4b).Compared with the other fuels, the NO x emissions are still about two times higher for the thermal plasma torch, even at the highest mixing rate of H 2 investigated here.Unfortunately, higher H 2 :N 2 ratios lead to lower required electrical power.For the used plasma systems, at H 2 :N 2 ratios above 30 %, the plasma generation is not supported due to insufficient electrical power.However, extrapolating the data towards higher mixing ratios of H 2 in N 2 indicates that similar NO x emissions could be reached for ~ 35 vol% H 2 in N 2 for the thermal plasma torch.

Further analysis with computational fluid dynamics
To gain further insights into the process NO x reduction potential, an axisymmetric lab-scale reactor consisting of a plasma torch, cylinder, and chamber was built and analysed both experimentally as well as numerically with CFD.Fig. 5a and 5b show the 3D sketch of the reactor and the axisymmetric 2D section used for the CFD simulations.A detailed sketch of the reactor including dimensions and experimental equipment is shown in Fig. S1.Fig. 5c shows the numerical domain and the boundary conditions used in the CFD simulations.The influence of mixing between the process air and the plasma gas was investigated by introducing an inner cylinder of two different lengths (i.e., 100 and 200 mm) into the reactor, hereafter referred to as the short inner cylinder (SIC) or long inner cylinder (LIC).
As with the preheated reactor, the experimental results showed that replacing air with N 2 or a mixture of 25:75 vol% H 2 :N 2 as the working gas reduces the NO x emissions of the system by over 35 and 90 %, respectively, independent of the inner cylinder configuration (Fig. S9).Increasing the H 2 content of the H 2 :N 2 mixture also reduces NO x formation in both reactor configurations.The consistency of the results with those measured in the preheated setup indicates that the concept is robust to the modest geometrical and operating condition changes in the system.
The developed CFD model was validated against the measured NO concentration in the outlet of the axis-symmetric reactor (Fig S10).The results show that the CFD model can predict the NO x emissions with relatively good accuracy (Fig. S10).Therefore, one can assume that it can be used for the detailed analysis of the reasons for NO x reduction in the experimental axis-symmetric reactor.Fig. 5 presents the reactive flow pattern in the reactor for the two inner cylinder configurations and for an H 2 :N 2 ratio of 15:85 vol%.The CFD results for air and different mixing ratios of H 2 and N 2 as the plasma working gas are presented in the supplementary material (Figs.S11-S13).The pure N 2 plasma jet has the highest axial momentum of all the air and H 2 :N 2 cases; this is because it reaches the highest bulk temperature of 5350 K (Fig. 2, hp = 7.7 MJ/kg), followed by air at 4021 K (Fig. S3).The lowest axial momentum occurs for the 25:75 vol% H 2 :N 2 case due to the lower temperature (i.e., 2884 K) of the gas leaving the torch (Fig. 2; h p = 4.6 MJ/ kg).Reducing the H 2 addition to the plasma working gas leads to a higher bulk temperature in the torch, which in turn causes greater density and axial momentum in the jet leaving the torch (Figs.S11c-f).The narrow central high-velocity plasma jet (Figs. 5 and S11) creates the first recirculation zone inside the inner cylinder.For the LIC, the plasma jet expands freely inside the inner cylinder, while for the SIC, the co-flow gas stream entering the inner cylinder hinders its expansion and extends its axial momentum farther downstream.This results in a second recirculation zone outside the inner cylinder as well as an entirely different mixing regime inside the inner cylinder for SIC, as shown by the streamlines in Fig. 5e and S12.The co-flow air entering into the inner cylinder results in O 2 molecules around the plasma jet for the SIC, as seen in Fig. S12a.This is critical for NO formation since the plasma jet can react with O 2 at very high temperatures.By increasing H 2 addition to the torch, the mass fraction of O 2 is reduced in the inner cylinder due to the oxidation of H 2 (Figs.S12c-f).For the LIC, the reaction of H 2 with the surrounding air is instead delayed until the end of the inner cylinder.
The temperature distribution and heat release are shown in Fig. 5f  and 5 g, respectively, for the 15:85 vol% H 2 :N 2 case and in Fig. S13 for the other cases.As discussed previously, the temperature of the plasma jet leaving the torch is strongly dependent on the properties of the plasma working gas (i.e., air and pure N 2 ) and the amount of H 2 added to the N 2 stream, and the average bulk temperature decreases from 5350 K for pure N 2 to only 2884 K for the 25:75 vol% H 2 :N 2 gas mixture.For both the LIC and SIC, the plasma jet loses heat to the inner cylinder wall mostly through convection heat transfer accelerated by the recirculation zone inside the cylinder.The inner cylinder loses heat to the cold co-flow air stream by convection and, more importantly, by radiation to the outer wall of the chamber.Since there is no co-flow mixing for the LIC, the temperature distribution inside the inner cylinder is uniform up to the mixing region at the end of the cylinder (x = 0.215 m).For the SIC, however, part of the much colder co-flow stream enters the recirculation zone and cools down the plasma jet faster, resulting in steeper radial and axial temperature gradients inside the inner cylinder.For the LIC, a narrow region of slightly higher temperatures is observed downstream of the end of the cylinder (0.25 < x < 0.28 m) for cases with H 2 addition due to the oxidation of H 2 with co-flowing air (Figs.S13c-f).The temperature plot does not clearly show this oxidation region for the SIC, because it is located inside the inner cylinder in a region with high temperature gradients.However, these regions as well as other regions with high reaction heat (i.e., at the end of the long cylinder) can be clearly visualized in the heat release contour plot shown in Fig. 5e and  S13.
Fig. 6 shows the mass fraction and net production rate of NO for all cases and reactor configurations.Regardless of configuration, when air is used as the plasma working gas (Fig. 6a), the NO formed inside the torch is transported to the chamber, promptly increasing the NO level to over 1000 ppm.For other plasma working gases (Fig. 6b-6f), the mass fraction of NO is determined by the production rates of the different mixing regimes inside the reactor, as there is no torch NO in these cases.Increasing the H 2 content of the H 2 :N 2 mixture significantly reduces the net production of NO, which consequently decreases the NO mass fraction in the reactor and the emission levels throughout the system (Fig. 6c-6f) This is mainly associated with the lower temperature of the H 2 plasma jets mixing with process air, as shown in Fig. S13, hindering the progress of the thermal NO reactions.
Altering the mixing regime of the plasma and the co-flow air by halving the length of the inner cylinder (i.e., the SIC configuration) has a two-fold effect on NO formation.First, it reduces the plasma temperature faster due to mixing the much colder air stream with plasma within the inner cylinder.Second, it supplies the O 2 essential for NO formation to the non-air plasma jet when it is high in temperature.The combined effect is stronger NO formation at the shear layer of the non-air plasma jet before x < 0.1 m in the SIC configuration, versus delayed NO formation after mixing with the co-flow air within 0.2 < x < 0.25 m in the LIC configuration.Therefore, to achieve an optimal design for the reduction of process NO in a non-air plasma system, the plasma temperature should be lowered as much as possible, preferably below 1800 K, using a mechanism other than mixing with cold air.Here the heat loss through the inner cylinder was an alternative mechanism for the LIC configuration.

Conclusions and outlook
The presumed high NO x emissions of industrial thermal plasma torches hamper their implementation as a furnace heat source in the construction material (e.g., iron, steel, and cement) production industry.This contribution demonstrates both theoretically and experimentally that it is possible to reduce the NO x emissions by replacing air with N 2 as the plasma-forming gas.Further reductions in NO x emissions are also possible by adding H 2 to the N 2 gas stream: with H 2 reaching 30 vol% of the stream, the NO x emissions are of the same order as would be expected from burning traditional fuels (e.g., coal, fuel oil, natural gas, and biomass).The results presented here indicate that environmentally friendly low-NO x thermal plasma torches powered by renewable electricity could play a role in the necessary electrification and decarbonization of the construction material production industry in the future.The main technical challenge is likely to upscale the proposed concept from the laboratory scale of ~ 10 kW to the ~ 10 MW scale of the burners today used to fire fossil fuels in the industry.However, the developed CFD model can probably be used in this development in the future.

Fig. 1 .
Fig. 1.Industrial thermal plasma torches.(a) Schematics of renewable-energy-powered plasma torches as a heat source in the construction material (e.g.iron, steel, and cement) process industry.(b) Photo of fuel oil flame used as a heat source in an iron-ore induration machine.(c) Schematics of the generation of torch NO x and process NO x .(d) Thermodynamic equilibrium calculations (TECs) showing the concentrations of NO,and O, and N radicals as a function of the enthalpy of an air plasma torch.The NO emissions are significantly higher than those of traditional fuels and must be reduced significantly before thermal plasma torches can be introduced as a heating source in the process industry.

Fig. 2 .Fig. 3 .
Fig. 2. Thermodynamic analysis of a plasma torch.Comparison of the equilibrium state of N 2 (red lines) and H 2 :N 2 (blue lines) plasma torches as a function of average plasma enthalpy.The H 2 :N 2 mixture has an H 2 content of 25 vol% and an N 2 content of 75 vol%.The grey vertical lines show the shift in torch enthalpy when N 2 is replaced with a mixture of 25 vol% H 2 and 75 vol% N 2 in a 9-kW torch with 60 normal liter per minute (NLPM) N 2 as the working gas.

Fig. 4 .
Fig. 4. Experimentally verified reduction of NO x from a thermal plasma torch.(a) Schematics of the experimental setup including important boundary conditions and approximate outlet gas temperature.Air, pure N 2 , or a mixture of N 2 and H 2 was used as the plasma-forming gas.The N 2 flow was constant (60 NLPM) and the H 2 flow increased sequentially to 20 NLPM.(b) Concentration of NO x expressed in mg NO 2 /MJ fuel as a function of the concentration of H 2 in the gas and at different preheated air temperatures.Results of using air as the plasma-forming gas are presented inside the ellipse and results [50] of burning traditional fossil fuels (fuel oil [FO] and coal [C]), biomass (bio-oil [BO], stem wood powder from pine and spruce [WP], steam explosion treaded wood powder sale under the name black pellets [BP]), and electrofuels (hydrogen gas [H 2 ]) in a pilot-scale kiln with highly preheated air are presented inside the rectangle.

Fig. 5 .
Fig. 5. CFD model and simulation results for the axisymmetric reactor.(a) 3D sketch of the reactor.(b) Axisymmetric domain used in the CFD simulations.(c) Schematics of the numerical domain and boundary conditions.(d-f) CFD results for a case with an H 2 :N 2 mixture of 15 and 75 vol%, respectively.(d) Axial velocity distribution bounded within the absolute 10 m/s range.Values below -10 m/s and above 10 m/s are colored the darkest blue and darkest red, respectively, shown with triangles in the color bar.(e) Streamlines of the flow plotted on top of O 2 mass fraction contours.(f) Temperature contours.(g) Heat release due to chemical reactions.Results for air, pure N 2 , and other mixtures of H 2 :N 2 can be found in Figs.S11-S13.