Journal of Physics D: Applied

Physics

PAPER You may also like

- A twisted generalization of linear Poisson
Simulation of the hydrodynamic expansion brackets of hydrodynamic type
Dongping Hou and Chengming Bai
following a nanosecond pulsed spark discharge in - Homogenization of the planar waveguide
with frequently alternating boundary
air at atmospheric pressure conditions
D Borisov and G Cardone

To cite this article: Fabien Tholin and Anne Bourdon 2013 J. Phys. D: Appl. Phys. 46 365205 - Two-component soliton systems and the
Painlevé equations
Mikio Murata

View the article online for updates and enhancements.

This content was downloaded from IP address 183.173.89.241 on 09/11/2024 at 03:31

IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 46 (2013) 365205 (18pp) doi:10.1088/0022-3727/46/36/365205

Simulation of the hydrodynamic

expansion following a nanosecond
pulsed spark discharge in air at
atmospheric pressure
Fabien Tholin1,2 and Anne Bourdon1,2
1
CNRS, UPR 288 ‘Laboratoire d’Energétique Moléculaire et Macroscopique, Combustion’, Grande voie
des vignes, 92295 Châtenay-Malabry, France
2
Ecole Centrale Paris, Grande voie des vignes, 92295 Châtenay-Malabry, France
E-mail: [email protected] and [email protected]

Received 19 April 2013, in final form 5 July 2013

Published 22 August 2013
Online at stacks.iop.org/JPhysD/46/365205

Abstract
This paper presents 2D simulations of the dynamics of formation of a nanosecond spark
discharge between two point electrodes in air at atmospheric pressure at 300 and 1000 K, the
induced air heating and the following hydrodynamic expansion. As a first step, we have
considered that 30% of the discharge energy instantaneously heats the ambient air. At the end
of the voltage pulse, it is shown that the energy density and the air temperature distributions
are non-uniform in the interelectrode gap. Rapidly after the nanosecond voltage pulse, a
cylindrical shock wave is formed and propagates with a velocity very close to the speed of
sound of the surrounding ambient air. Furthermore, the rapid dilatation of the hot channel
formed on the discharge path is observed for t
1 µs, as in experiments. Then we have
carried out a parametric study on the influence of the value of the fraction of discharge energy
going to fast heating on the hydrodynamic expansion at 1000 K, assuming an instantaneously
fast gas heating. For all values in the range of 15% to 60% studied in this work, we have
observed a very similar gas dynamics. Then, we have considered that the nanosecond spark
discharge heats the ambient air at 1000 K with a longer relaxation time of 1 µs, and in this case
we have observed the propagation of a weak pressure wave and no dilatation of the hot channel
on the discharge path. Finally, the comparison with experiments seems to validate the
hypothesis that the 10 ns spark discharges studied in this work at 300 and 1000 K, significantly
heat the ambient air on very short time-scales.
(Some figures may appear in colour only in the online journal)

1. Introduction ensure a low power consumption. Finally, the use of repetitive

voltage pulses results in the accumulation of metastable species
Nanosecond repetitively pulsed (NRP) discharges are being important for sustaining the discharge and active species
increasingly studied for various applications such as plasma- interesting for applications.
assisted combustion (Pilla et al 2006, Starikovskaia 2006), Pai et al (2010a) have carried out a detailed experimental
aerodynamic flow control (Starikovskii et al 2009) or study to identify the different regimes of NRP discharges in
nanomaterial synthesis (Pai 2011). In these discharges, the air at atmospheric pressure in the temperature range 300 to
application of high-voltage pulses generates a high electric 1000 K between two metallic point electrodes with a radius of
field that allows an efficient electron impact ionization and curvature of about 200 µm, for a gap size in the range 5–10 mm,
the production of highly reactive species. Furthermore, a for 10 ns voltage pulses with an amplitude in the range 5–10 kV
voltage pulse with a duration of a few nanoseconds is chosen to and a repetition frequency in the range 1–30 kHz, and with

0022-3727/13/365205+18$33.00 1 © 2013 IOP Publishing Ltd Printed in the UK & the USA
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

an external flow aligned with the axis of electrodes at 0.5– starting from 50 ns after the discharge pulse were recorded.
2.6 m s−1 . It is interesting to note that the cumulative effect of These images clearly show a shock wave propagation and the
repeated pulsing achieves a stable ‘quasi-periodic’ behaviour. expansion of the heated gas channel formed on the discharge
Thus, even though they are transient, the NRP discharges path. Then, the significant fast gas heating is believed to play
have a visual resemblance with dc discharges. Three different a key role in the application of NRP spark discharges for
discharge regimes have been observed as the applied voltage aerodynamic flow control (Bityurin et al 2008, Starikovskii
increases: corona, glow and spark regimes. It is interesting et al 2009) and plasma-assisted combustion (Pilla et al 2006,
to note that corona and spark discharge regimes were easily Starikovskaia 2006).
obtained under almost all experimental conditions, whereas It is interesting to note that the gas heating following a
the glow regime existed only over a limited range of conditions dc spark in air at atmospheric pressure has been extensively
between 750 and 1000 K. In Pai et al (2009), Rusterholtz studied experimentally (Maly 1984, Marode 1975a) and
(2012), Tholin and Bourdon (2011), Tholin and Bourdon numerically, mostly in 0D and 1D (Marode 1975b, Marode
(2013), experiments and simulations have been carried out to et al 1979, Bastien and Marode 1985, Naidis 1999, Naidis
extend the domain of existence of the glow regime between two 2005, Riousset 2010). However, so far, only few studies have
point electrodes in air at atmospheric pressure down to 300 K been devoted to the simulation of NRP discharges in the spark
in varying the gap size, the radius of curvature of electrodes, regime. In particular, Popov (2011b) has simulated in 1D the
the applied voltage and the external air flow. production of active species and the neutral gas temperature
Pai et al (2010b) have carried out an extensive evolution during and after an NRP spark discharge voltage
experimental study of the spark regime of NRP discharges pulse and Naidis (2008) has simulated in 0D the multi-pulse
in air at atmospheric pressure in the temperature range 300 to nanosecond spark dynamics and the neutral gas temperature
1000 K between two metallic point electrodes with a radius of evolution after several voltage pulses. In this work, we
curvature of about 200 µm, for a gap size of 5 mm, for 10 ns propose to simulate in 2D the formation of a nanosecond spark
voltage pulses with an amplitude up to 10 kV and a repetition between two point electrodes in air at atmospheric pressure,
frequency in the range 1–30 kHz, and with an external flow the fast gas heating during and after the voltage pulse and the
aligned with the axis of electrodes at 0.5–2.6 m s−1 . The following hydrodynamic expansion. We propose to compare
nanosecond spark discharge regime is characterized by an our simulation results with experimental results obtained in
intense emission, a high conduction current and a significant Xu et al (2011) on the shock wave propagation and on the
heating of the gas (by about 1000 K (Pai et al 2010b, expansion of the heated gas channel formed on the discharge
Rusterholtz et al 2012)) within a few tens of nanoseconds path.
following each voltage pulse. Indeed, during a nanosecond In section 2.1, we present the equations needed for the
discharge, a significant part of the discharge energy is used numerical simulation of air streamer discharges at atmospheric
to produce vibrationally excited molecules and electronically pressure in a point-to-point geometry during a nanosecond
excited atoms and molecules. In the literature, these last years, voltage pulse. In section 2.2, we set out the model for the
several works (Popov 2001, Naidis 2008, Aleksandrov 2010, spark phase. In section 2.3, we present the model for the
Popov 2011a, 2011b, Mintoussov et al 2011, Rusterholtz et al hydrodynamic expansion after the voltage pulse. In section 3.1
2012, Rusterholtz 2012) have been devoted to the study of the we simulate a nanosecond spark in air at 1000 K, the fast
fast relaxation of the energy stored in the different internal gas heating and the hydrodynamic expansion after the voltage
modes of molecules and atoms in air and the subsequent rapid pulse and simulation results are compared with experimental
increase of the translational temperature of the neutral gas. As results obtained in Xu et al (2011). In sections 3.2 and 3.3 we
initially proposed by Popov (2001), an important process for carry out a parametric study on the fast gas heating source
fast gas heating in air at atmospheric pressure by nanosecond term and its relaxation time to study its influence on the
pulsed discharges seems to be the fast quenching of electronic hydrodynamic expansion observed after the voltage pulse.
excited levels of nitrogen in the so-called two-step mechanism. Finally, in section 3.4, we go into the results obtained for the
First, during the nanosecond voltage pulse, electronically simulation of a nanosecond spark in air at 300 K with fast gas
excited nitrogen molecules are produced by electron impact heating and compare with experimental results obtained in Xu
due to the high electric field present in the discharge. In a et al (2011).
second step, these molecules are quenched by neutral particles
and mainly by O2 molecules. These quenching reactions
2. Model formulation
are very efficient and occur quickly after the discharge in
less than a few tens of nanoseconds (Rusterholtz et al 2012). 2.1. Model for the streamer phase of the discharge during the
As a consequence, O2 molecules are dissociated and part of voltage pulse
the energy stored in the electronic excited levels of nitrogen
molecules is transferred into kinetic energy of the oxygen As shown in our previous numerical studies (Celestin et al
atoms produced. In a few collisions, these oxygen atoms 2009, Bourdon et al 2010, Tholin and Bourdon 2011,
thermalize and then increase the neutral gas temperature. Tholin and Bourdon 2013) and validated by comparison with
Recently, in Xu et al (2011) single-shot schlieren images experiments in Tholin et al (2011), the discharge during
of NRP discharges between two point electrodes in air at a nanosecond voltage pulse between two point electrodes
atmospheric pressure in the temperature range 300 to 1000 K consists of a positive streamer propagating from the anode

2
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

to the cathode and a negative streamer propagating in the density: N1000 = 0.3 × N300 where N300 = 2.45 × 1019 cm−3
opposite direction until they impact each other to form a is the density of air at 300 K. The decrease of the total density
plasma channel. The most common and effective model to by a factor of 3.3 as the temperature increases from 300 to
study streamer dynamics in air at atmospheric pressure is 1000 K, increases by the same factor the local reduced electric
based on the following drift–diffusion equations for electrons, field E/N and then has a direct impact on transport parameters
positive and negative ions coupled with Poisson’s equation and reaction rates for air.
(e.g. Kulikovsky 1997): The charged species transport equations are solved using
a modified 1D Scharfetter–Gummel algorithm (Kulikovsky
∂t ne − ∇ · (ne µe E + De ∇ne ) = Se + Sph 1995) with the parameter of this scheme, ISG = 10−2 . To
∂t nn − ∇ · (nn µn E + Dn ∇nn ) = Sn simulate 2D discharges, we split the numerical resolution into
(1) two 1D problems in the x and r directions, respectively and
∂t np + ∇ · (np µp E − Dp ∇np ) = Sp + Sph , we alternate the order of the resolution between x and r axes
qe at each time-step. In this work, as in Celestin et al (2009)
∇ 2V = − (np − nn − ne ) we take into account simplified boundary conditions: near
0
the anode and cathode surfaces, gradients of electron density
where subscripts ‘e, ‘p’ and ‘n’ refer to electrons, positive and are assumed to be zero. The ghost fluid method is used to
negative ions, respectively, ni is the number density of species take into account the exact shapes of the electrodes in the
i, V is the potential, E = −∇V is the electric field, Di and resolution of Poisson’s equation on a rectilinear grid (Celestin
µi are the diffusion coefficient and the absolute value of the et al 2009). The finite difference form of Poisson’s equation
mobility of species i, respectively, qe is the absolute value of is solved using the SuperLU solver (Demmel et al 1999)
electron charge, and 0 is the permittivity of free space. Sph (http://crd.lbl.gov/ xiaoye/SuperLU/) with Dirichlet boundary
is the photoionization source term and the Si terms stand for conditions on electrodes and Neumann boundary conditions
the source terms of species i due to ionization, attachment, on the other axial and radial boundaries of the computational
recombination and detachment processes. domain. For the time integration during the voltage pulse, a
We study axisymmetric streamers propagating between simple first-order Euler time integration is used for fluxes and
two aligned point electrodes and thus cylindrical coordinates the photoionization source term, and an explicit fourth order
(x, r), are used with the x-axis as axis of the discharge. In Runge–Kutta method (Ferziger and Peric 2002) is used for
this work, simulation results are presented for point-to-point other source terms. For stability and accuracy, the time-step
configurations with hyperboloid electrodes with a radius of in our explicit simulations is limited by several considerations.
curvature of Rp = 300 µm. In Xu et al (2011), the point The time-scales of relevance are the Courant δtc time-scale,
electrodes are separated by an adjustable distance of 2–4 mm defined as in Vitello et al (1994), the chemistry time-
and in this work, we have considered a 2.5 mm gap. For all scale δtS due to ionization, attachment, recombination and
discharge simulations, a computational domain with a radius detachment processes,

and the dielectric relaxation time-scale
of Rmax = 1.0 cm and a length Lmax = 1.0 cm is used with δtD = 0 /(qe i |ni µi |). The model time-step is calculated
a fixed no-uniform rectilinear grid defined as in Tholin and as δt = min(Ac δtc , AS δtS , AD δtD , 10−10 ) with Ac = 0.5,
Bourdon (2011). AS = 0.05 and AD = 0.5. The upper limit value
In this work, we study the discharge dynamics at of 10−10 s is used as a characteristic time-step to follow
atmospheric pressure between two points in ambient air at accurately the time evolution of the nanosecond applied
Tg = 300 K and in preheated air at Tg = 1000 K. In the charged voltage.
species transport equations, all transport coefficients and rate For the voltage pulse, we model the whole shape of the
coefficients in source terms are assumed to be functions of the nanosecond applied voltage pulse with a steep nanosecond
local reduced electric field E/N , where E is the electric field voltage rise, voltage plateau and a steep nanosecond voltage
magnitude and N is the air neutral density. At Tg = 300 K, decrease, as detailed in section 3.1 at Tg = 1000 K and
all transport parameters for charged species are taken from section 3.4 at 300 K. In the following, the value of the voltage
Morrow and Lowke (1997) except diffusion coefficients for plateau is referred to as the applied voltage. As initial
ions which are derived from mobilities using Einstein relation. condition, we consider a low uniform preionization of ambient
Reaction rate coefficients for ionization, attachment and air with 104 cm−3 electrons and positive ions as discussed
recombination processes are taken from Morrow and Lowke in Pancheshnyi (2005). Finally, we compute the conductive
(1997). For detachment, as in Benilov and Naidis (2003), we current I and the total energy density eJ (t) of the discharge at
use the detachment rate coefficient proposed by Mnatsakanyan time t from: 
and Naidis (1985) which is a function of the air temperature I = jc · dS , (2)
and the electric field. For photoionization, the three-group SP3 S
model is used (Bourdon et al 2007) with boundary conditions and  t
given in Liu et al (2007). As discussed in our previous eJ (t) = jc · E dt, (3)
works (Celestin et al 2009, Bourdon et al 2010, Tholin and 0
Bourdon 2011, Tholin and Bourdon 2013), to take into account where jc is the conductive current density and S is a surface
the fact that the discharge occurs at atmospheric pressure at described by equation {x = Const}. The choice of the location
Tg = 1000 K, we have only scaled the value of the total air of the surface S will be discussed in section 3.1.

3
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

2.2. Model for the spark phase of the discharge during the section is carried out when two conditions are fulfilled: (1) the
voltage pulse axial electron density is ten times higher than in the positive
streamer channel before connection (i.e. at Tg = 1000 K, the
In Tholin and Bourdon (2011), it has been shown that after the transition is for an electron density higher than 1014 cm−3 ) and
connection of both streamer discharges, the electric field in (2) the axial electric field is uniform and equal within 10% to
the plasma channel between point electrodes becomes rapidly the average electric field (i.e. the applied voltage by the gap
rather uniform and equal to the average electric field (i.e. the length). After the transition to the spark model, this model
applied voltage divided by the gap length). If the value of is used until the end of the discharge pulse. The transition
the average electric field is higher than the breakdown field, between both models usually occurs 0.5 ns after the connection
the electron density goes on increasing while the electric field of both streamer discharges. During the spark phase, the model
remains constant and the discharge transits to a spark phase. time-step is calculated as δt = min(10−10 , AS δtS ) where AS
This results in a very fast increase of the conductive current and and δtS are defined in section 2.1. As in the streamer phase
of the energy released in the discharge. As the electron density in section 2.1, the upper limit value of 10−10 s is used as a
increases in the spark phase, the dielectric relaxation time of the characteristic time-step to follow accurately the time evolution
plasma δtD decreases to values of the order of 10−16 s, i.e. much of the nanosecond applied voltage.
less than the other characteristic time-scales of relevance for
the selection of the explicit simulation time-step as explained
in section 2.1. With such small time-steps, it is very time 2.3. Model for the hydrodynamic expansion after the voltage
consuming to simulate the spark phase of a nanosecond pulsed pulse
discharge with the fluid model and the explicit numerical To study the coupling of fast-heating nanosecond pulsed
method described in section 2.1. However, it is important discharges with ambient air, we have to question the necessity
to note that during the spark phase the electric field remains to solve the neutral gas flow equations during the voltage pulse.
rather constant and the chemistry of charged species is much If the temperature increase during the voltage pulse is very
slower than in the streamer phase. In fact, the severe time-step small (<10 K), the neutral gas density remains constant and
limitation is due to thin non-neutral regions close to electrodes then, there is no need to solve flow equations during the voltage
and on the wings of the discharge channel. pulse. If the temperature increase during the voltage pulse is
As mentioned in the introduction, so far, only few studies more significant, the gas density may vary during the voltage
have been devoted to the simulation of NRP discharges in the pulse and then change the reduced electric field. In this case,
spark regime (Popov 2011b, Naidis 2008). In both studies, it is the gas heating has a significant influence on the discharge
assumed that the electric field is uniform over the cross-section dynamics. We have then compared the duration of the pulse
of the discharge channel and is either constant or derived from Tpulse with the characteristic time for the change of neutral gas
an experimental current distribution. In this work, in order to density due to heating, defined as Rs /cmax where Rs is the
simulate the nanosecond spark discharge in 2D, we have used typical radius of the discharge, and cmax is the maximum of the
as a first step a simple approach, that we validate a posteriori speed of sound in the path of the discharge. The increase
by comparison of simulation results with experimental results of the neutral gas temperature leads to an increase of the

in section 3. We have assumed that after the connection, the speed of sound: cmax ∝ Tg,max . Then, the time necessary
structure of the nanosecond spark discharge remains the same for acoustic waves to propagate inside the heated region is a
and that the electric field in the spark phase is proportional good approximation of the time-scale of the neutral gas density
to the applied voltage. As a consequence, the space charge variation. It is interesting to note that if
density ρc = qe (np − nn − ne ) has to evolve proportionally to
the applied voltage during the spark phase: Rs
Tpulse  (5)
cmax
ρc (t) = AV (t) (4)
neutral species are immobile during the voltage pulse. Then
where V (t) is the applied voltage and A is a constant. It is even if the temperature increases during the voltage pulse, the
reasonable to assume that only electrons have a sufficiently neutral gas density remains the same (Naidis 2008). In this
high mobility to ensure the fast variation of the space charge case, the temperature increase has an influence on transport
proportionally to the applied voltage. Then, as a first step, coefficients and rate coefficients in source terms but its impact
we propose to add an electron source term in each point of on the discharge dynamics is small. It is important to note
the computational domain to modify the space charges when that the condition given by equation (5) is rather strict as the
the voltage varies to ensure the condition of equation (4) temperature field generated by the discharge is highly non-
without solving consistently all the drift–diffusion fluxes of uniform. For an intense spark discharge condition with a
charged species in equations (1). This simple approach is non- temperature increase up to 3000 K and a discharge diameter
conservative as we take into account an additional electron of 500 µm, the condition given by equation (5) becomes
source term, but it is important to note that the lack or excess Tpulse  500 ns. As in our case, the durations of voltage pulses
of electrons in the thin non-neutral regions during the spark are in the range of 5 to 10 ns, the condition given by equation (5)
regime is small compared to the total electron density which is fulfilled. Then in our conditions, there is no need to solve
increases by orders of magnitude during the spark phase. the flow equations during the voltage pulse and we have used
In this work, the transition between the streamer model the approximation of a constant neutral gas density during the
presented in section 2.1 and the spark model presented in this voltage pulse.

4
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

As mentioned in the introduction, in recent years, several specific internal energy u, ht is the specific total enthalpy, Rs
works (Popov 2001, Naidis 2008, Aleksandrov 2010, Popov is the specific gas constant, γ is the isentropic coefficient of the
2011a, 2011b, Mintoussov et al 2011, Rusterholtz et al 2012, neutral gas, assumed to be constant and equal to 1.4 and Sener
Rusterholtz 2012) have been devoted to the study of the fast is an energy source term which is equal to zero in the case of an
relaxation of the energy stored in the different internal modes instantaneous gas heating. The system given by equations (7)
of molecules and atoms in air and the subsequent rapid increase is closed by the ideal gas state equation:
of the translational temperature of the neutral gas. However,
there are still some uncertainties on the chemical processes p = ρu(γ − 1). (8)
involved and the reaction rate coefficients of some reactions
involving excited states. Therefore, as a first step, in this Second, we have assumed that the heating of the neutral gas by
work, we have carried out a parametric study and assumed the discharge occurs at a finite rate after the voltage pulse. More
that a fraction ηR of the discharge energy goes to fast heating. precisely, we have considered a negligible gas heating during
For the conditions of the nanosecond spark discharge studied the voltage pulse and we have assumed that a fraction of the
in this work, based on Popov (2011a) and Aleksandrov (2010), discharge energy at the end of the voltage pulse ηR eJ (Tpulse ) is
we have considered as a reference in sections 3.1 and 3.4 that put into the air heating with a characteristic time of τh  Tpulse .
ηR = 30% of the discharge energy instantaneously heats the Then as initial condition of the Euler simulations, we have
ambient air. Then, to discuss the influence of ηR on the results, considered ambient air at a given and uniform temperature Tg
in section 3.2 we consider two other values of ηR = 15% and and we have solved Euler equations (equations (7)) with the
60%. Finally, we have assumed in section 3.3 that the heating energy source term given by:

of the neutral gas by the discharge (with ηR = 30%) occurs S ηR eJ (Tpulse )
ener = (t − Tpulse ) for Tpulse
t
τh
during a characteristic relaxation time τh after the voltage τh (9)
pulse. 
Sener = 0 for t > τh .
First, if the discharge instantaneously heats the neutral gas
For both gas heating models, the governing 2D Euler equations
during the voltage pulse, at each point of the computational
domain during the integration time-step dt we have: (equations (7)) are solved for a cylindrical geometry using
a finite volume approach. The third-order classical MUSCL
 t+dt scheme proposed by van Leer (1979) is used for fluxes with
ηR
cv dTg (t) = (eJ (t + dt) − eJ (t)) (6) a Lax–Friedrichs Riemann solver (Shu 1997) and a Minmod
t ρ0
slope limiter. A second order Runge–Kutta method is used for
where cv is the specific thermal capacity of air at constant time integration. The time-step of integration for the explicit
volume and ρ0 is the density of the neutral gas which is Euler simulations is calculated as δt = 0.5δtCFL where δtCFL
assumed to be constant during the voltage pulse. Equation (6) is the classical Courant–Friedrich–Levy (CFL) time-step. We
is solved to derive Tg (t) assuming that cv is constant and using a have checked that with this choice, the temperature increase is
predictor–corrector method. As an instantaneous gas heating less than 10 K during one time-step. Validation test-cases of
is assumed, the maximal air temperature is obtained at the the 2D axisymmetric Euler code and discussion on the choice
end of the voltage pulse. This temperature field is used as of the numerical scheme can be found in Tholin (2012).
initial condition for the 2D neutral gas flow equations. In the The electrode surfaces are considered as reflecting
following, to avoid confusion, we use time tt which starts at surfaces for the flow. Conversely, to avoid any reflection on the
the beginning of the voltage pulse to present results on the outer boundary of the computational domain during the Euler
nanosecond discharge and time t = tt − Tpulse to present flow simulations, we have considered a large computational domain
results after the voltage pulse. In this work, as we study the of [−39 cm : 40 cm] in the axial direction and [0 cm : 40 cm] in
hydrodynamic expansion on short time scales (t
10 µs) after the radial direction. A refined and uniform mesh of 10 µm on
a nanosecond spark discharge to compare with experimental 1 cm × 1 cm centred on the discharge gap is used and then the
results (Xu et al 2011), as a first step, we have neglected viscous mesh is expanded in both directions following a geometric
and diffusive terms. Then we consider 2D axisymmetric Euler progression up to the outer boundaries. It is important to
equations: note that the computational domain and the mesh of the Euler


 ∂t ρ + ∇ · (ρ V ) = 0 simulations are larger and less refined, respectively, than those

∂t (ρVx ) + ∇ · (ρVx V ) = −∂x p used for the discharge simulations. Then at the end of the
(7)

 ∂t (ρVr ) + ∇ · (ρVr V ) = −∂r p voltage pulse, the gas temperature and the discharge energy

∂t (ρut ) + ∇ · (ρht V ) = Sener calculated from the discharge code are interpolated linearly on
1 R s Tg the mesh used to compute the Euler equations.
with ut = V 2 + u, u = cv Tg =
2 (γ − 1)
p 3. Results and discussion
and ht = ut + ,
ρ
3.1. Simulation of the flow expansion following a nanosecond
where ρ is the density of the neutral gas, Vx and Vr are, spark discharge at Tg = 1000 K: reference test-case
respectively, the axial and the radial components of the fluid
velocity V , ut is the specific total energy of the neutral gas, As a reference test-case for nanosecond spark discharge in air
defined as the sum of the specific kinetic energy and of the at Tg = 1000 K, we consider a point-to-point discharge with

5
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

a gap size of 2.5 mm, electrodes with a radius of curvature of 6000

current
45
300 µm and an applied voltage of 5100 V. This test-case is very voltage 40
close to the multi-pulse spark discharge studied experimentally 5000
35

Applied voltage [V]

in Rusterholtz et al (2012). Figure 1 shows the shape of
4000 30
the voltage pulse with rise and decrease times of 2 ns, and

current [A]
a pulse duration of 10 ns and the conductive current calculated 25
3000
using equation (2) for a surface S located close to the cathode 20
at x = 0.75 cm. Figure 2 shows the distributions of the
2000 15
absolute value of the electric field, the electron density and the
discharge energy at tt = 4, 6, 8 and 10 ns. The gas temperature 1000
10

distributions during the voltage pulse are also shown, assuming 5

that ηR = 30% of the discharge energy heats instantaneously 0 0
the ambient air at Tg = 1000 K. In this test-case, the connection 0 2 4 6 8 10

between positive and negative discharges occurs at tt = 4.5 ns. time [ns]

During the streamer phase for tt
4.5 ns, figure 1 shows Figure 1. Time evolution of the applied voltage and of the
that the conductive current calculated for a surface S located conductive current calculated using equation (2) for a surface S
close to the cathode at x = 0.75 cm is very small. During the located close to the cathode at x = 0.75 cm for the reference spark
streamer phase, the conductive current depends of the location discharge in air at Tg = 1000 K for a point-to-point configuration
with a gap size of 2.5 mm and electrodes with a radius of curvature
of the S surface in equation (2). However, whatever the of 300 µm, close to the conditions studied experimentally in
location of the S surface, we have checked that the calculated Rusterholtz et al (2012).
conductive current remains very small, in agreement with the
conductive current obtained in experiments (Rusterholtz et al density has significantly decreased in the discharge channel and
2012). For tt  4.5 ns, the discharge transits rapidly to a spark particularly close to the electrode tips. We also note that the
phase with a fast increase of the conductive current and of the velocity of the fluid has a cylindrical shape all around the heated
energy density. It is interesting to note that during the spark channel with a slightly higher value close to the hot spots. For
phase, the conductive current is independent of the location of t > 400 ns, a cylindrical shock wave characterized by jumps in
the S surface in equation (2). In figure 1, we observe a peak of pressure, density, temperature and velocity propagates radially.
conductive current at about tt = 8 ns of 40 A, in agreement with These jumps are more significant in the region close to the
the current obtained in experiments (Rusterholtz et al 2012) electrode tips due to a higher initial temperature. However,
and then the current decreases due to the decrease of the applied the shock wave propagation velocity is at every point of the
voltage. Figure 2 shows that the energy density distribution is wave front equal to the speed of sound in the unperturbed
non-uniform in the interelectrode gap during the voltage pulse air at 1000 K (650 m s−1 ) as observed in experiments (Xu
with peaks close to both electrodes. Indeed, as the discharge et al 2011). Behind the shock wave, a low density region
radius is smaller close to point electrodes than in the middle of is formed with a pressure less than the atmospheric pressure.
the gap, the conductive current and then the discharge energy This depletion is a consequence of the air displacement due to
are higher close to electrodes. Figure 2 also shows that the air the velocity induced by the shock wave. It can be interpreted
temperature increases rapidly during the spark phase. At the as the fast dilatation of the hot channel produced by the
end of the voltage pulse at tt = Tpulse = 10 ns, we observe a discharge. Between t = 800 ns and 1.2 µs two other shock
hot channel on the discharge path with two hot spots close to waves propagate from the electrode tips in the low pressure,
electrodes. As we have assumed an instantaneous heating of low density channel. These shock waves are reflections of
ambient air by the nanosecond discharge, the air temperature the first cylindrical shock wave on the electrode tips. These
Tg (tt = Tpulse ) obtained at the end of the voltage pulse is used successive shock waves increase the gas density in the hot
as initial condition for the gas-dynamic Euler equations (i.e. channel and the pressure slowly converges to the atmospheric
t = tt − Tpulse = 0). pressure for t > 1.2 µs. Figure 5 shows that at t = 10 µs,
The gas density is assumed to be constant during the a hot channel with a low density is formed on the discharge
voltage pulse, then the pressure at the end of the voltage pulse path with a uniform atmospheric pressure. We point out that
is computed using the ideal gas law considering a density the temperature in the hot channel at t = 10 µs is less than
of 0.35 kg cm−3 for air at 1000 K. Figure 3 shows the 2D the initial one in figure 3, as the shock wave propagation has
distributions of gas temperature, air density, pressure and removed part of the initial hot channel energy.
magnitude of the flow velocity at the end of the voltage pulse Figure 6 shows the time evolution of the air density
and then at the beginning of the Euler simulation. Figure 3 and temperature in the radial direction in the middle of the
shows that the initial condition consists of an homogeneous inter-electrode gap every 100 ns up to 900 ns. In the first
gas density in the whole computational domain, and a channel 600 ns, we note a fast depletion of the air density and of the gas
on the discharge path with higher pressure and temperature temperature inside the hot channel. On the density profiles, we
than ambient air with two hot spots close to electrode tips. note that the shock wave is formed on the wings of the heated
Figure 4 shows the 2D distributions of gas temperature, air channel before propagating radially towards the surrounding
density, pressure and magnitude of the flow velocity at t = 400, ambient air. Due to the cylindrical symmetry, the amplitude
800 ns and 1.2 µs. At t = 400 ns, figure 4 shows that the gas of the shock wave decreases as it propagates radially. The

6
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

tt = 4 ns

tt = 6 ns

tt = 8 ns
Figure 2. Cross-sectional views of the magnitude of the electric field, electron density, discharge energy density and neutral gas temperature
for the reference nanosecond spark of figure 1 in air at Tg = 1000 K. The neutral gas temperature is obtained assuming an instantaneous gas
heating with ηR = 30% (see section 2.3).

7
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

tt = 10 ns

Figure 2. (Continued).

t=0 µs
Figure 3. Cross-sectional views of the density, pressure, air temperature and magnitude of the flow velocity at the end of the voltage pulse
for the reference nanosecond spark discharge of figure 1 in air at Tg = 1000 K assuming an instantaneous gas heating with ηR = 30% (see
section 2.3). This corresponds to the initial condition for the 2D Euler simulations.

maximum density inside the shock wave is around 0.4 kg cm−3 jump decreases and becomes very small. As a consequence,
which corresponds to an increase of 15% in comparison of as the speed of sound is proportional to the square root of
the density of 0.35 kg cm−3 of the surrounding ambient air at the air temperature, the speed of sound of the steady air is
Tg = 1000 K. The maximum temperature jump behind the barely affected by the presence of the shock wave. Then,
shock wave in the radial direction in the middle of the gap in the generated shock waves propagate at velocities very close
figure 6 is around 40 K. Figure 4 shows that the temperature to the speed of sound at Tg = 1000 K. Figure 6 also shows
field is non-uniform axially and higher temperature jumps up that the hot channel increases in size as a function of time
to 200 K are obtained close to the electrodes. The maximum for t
1 µs. If we define the radius of the hot channel as the
temperature jump due to the shock wave is then of about distance from the axis of symmetry at which the density is equal
20%, but as the shock wave propagates this temperature to the density of the ambient air (in our case 0.35 kg cm−3 ), our

8
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

t=400 ns

t=800 ns

t=1.2 µs
Figure 4. Cross-sectional views of density, pressure, air temperature and magnitude of flow velocity for t = 400, 800 and 1.2 µs for the
initial condition of figure 3 in air at Tg = 1000 K.

9
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

t=10 µs
Figure 5. Cross-sectional views of the density, pressure, air temperature and magnitude of the flow velocity at t = 10 µs for the initial
condition of figure 3 in air at Tg = 1000 K.

0.4 1900

1800
0.35 1700

1600
Temperature [K]
Density [kg m ]
-3

0.3
1500
t=0 ns
t=100 ns 1400
0.25
t=200 ns
t=300 ns 1300
t=400 ns
t=500 ns 1200
0.2 t=600 ns
t=700 ns
t=800 ns 1100
t=900 ns
0.15 1000
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
r [cm] r [cm]

Figure 6. Time evolutions of the air density and the air temperature in the radial direction in the middle of the gap every 100 ns up to 900 ns
for the same condition as figure 4.

results show that this radius increases between 300 and 900 ns in the middle of the gap: in the first 400 ns, the temperature
by 60%. The expansion velocity of the heated channel for decrease is around 400 K in the middle of the gap and around
t
1 µs is around 660 m s−1 which is a little higher than 1000 K at the tip of the electrodes. This faster decrease of
the speed of sound at Tg = 1000 K. It is interesting to point temperature close to electrodes is due to the higher pressure
out that this fast expansion of the heated channel on short and the faster shock wave formation, which is able to remove
time-scales is also observed in experiments (Xu et al 2011). a significant amount of the deposited energy. However, for
Figure 7 shows the time evolution of the air density and the t > 600 ns, the temperature difference between the electrode
temperature in the axial direction on the symmetry axis every tips and the middle of the gap remains rather constant and
100 ns up to 900 ns. The initial temperature distribution is equal to about 600 K. As a consequence, the 2D temperature
non-uniform on the discharge axis with a 1000 K difference distribution in the hot channel obtained in figure 5 for t = 10 µs
between the temperature at electrode tips and in the middle of is more homogeneous than at the end of the voltage pulse
the gap. Figure 7 shows that the air density and temperature (figure 3). Finally, it is interesting to note that the different
decreases are more significant close to the electrode tips than succeeding shock waves have removed a significant amount

10
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

0.4 3200
t=0 ns
3000 t=100 ns
t=200 ns
0.35 t=300 ns
2800 t=400 ns
t=500 ns
0.3 2600 t=600 ns

Temperature [K]
Density [kg m-3]

t=700 ns
2400 t=800 ns
t=900 ns
0.25
2200

0.2 2000

1800
0.15
1600

0.1 1400
0.65 0.7 0.75 0.8 0.85 0.65 0.7 0.75 0.8 0.85
x [cm] x [cm]

Figure 7. Time evolutions of the air density and the air temperature in the axial direction on the symmetry axis every 100 ns up to 900 ns for
the same condition as figure 4.

of the discharge energy deposited in the gas: the temperature of its propagation for ηR = 60% than for ηR = 15%. On
decrease after the shock wave propagations is around 30% at the radial profiles of the air density, we observe that between
the tip of the electrodes and 20% in the middle of the gap. 100 and 900 ns, the radius of the hot channel increases by 50%
Then the good agreement obtained with experiments in for ηR = 15%. For ηR = 60%, the radius of the hot channel
Xu et al (2011) on the flow expansion after the nanosecond increases by 67% in the same time, corresponding to a much
pulsed spark discharge, validates a posteriori the simple spark faster and more significant dilatation of the hot channel. It is
model presented in section 2.2 and seems to confirm that the interesting to note that the rapid dilatation of the hot channel
gas heating by the nanosecond spark discharge in Xu et al for t
1 µs has been obtained for all values of ηR studied
(2011) is very fast. in this work in assuming an instantaneous heating of ambient
air. As this fast expansion of the hot channel is also observed
3.2. Influence of the fraction of discharge energy going to fast in experiments (Xu et al 2011) at Tg = 1000 K, the results
gas heating on the flow expansion following a nanosecond obtained in section 3.1 and in this section seem to confirm
spark discharge at Tg = 1000 K the very fast heating of the ambient air by the nanosecond
spark discharge in Xu et al (2011). However, additional more
In the previous section, we have assumed that ηR = 30% of the quantitative experimental studies are required to derive the
total discharge energy instantaneously heats the ambient air. In value of ηR from the comparison of experimental results in
this section, we have considered two other values of ηR = 15% Xu et al (2011) and numerical results.
and 60% to study its influence on the gas dynamics. Figure 8
shows the time evolutions of the air density and air temperature 3.3. Influence of the gas heating relaxation time on the flow
in the radial direction in the middle of the inter electrode gap
expansion following a nanosecond spark discharge at
every 100 ns up to 900 ns for ηR = 15% and ηR = 60%.
Tg = 1000 K
Figure 9 shows for the same conditions, the time evolutions
of the air density and the temperature in the axial direction on In previous sections, we have assumed that the fraction ηR of
the symmetry axis. First, it is interesting to note that for the the total discharge energy instantaneously heats the ambient
different studied values of ηR in the range 15% to 60%, we air. However, the air heating by the nanosecond discharge is
observe very similar shock wave structures and a very similar due to the relaxation of the energy stored in different internal
dynamics of the flow. In all cases, the propagation velocity modes of molecules and atoms, which may have very different
of shock waves is very close to the speed of sound in steady timescales. The hypothesis of an infinitely fast-heating rate is
air at 1000 K. As expected, as the value of ηR increases, more valid only if the fast-heating process occurs in a characteristic
energy is put into the ambient air and the generated shock wave time which is much less than the diameter of the discharge
induces higher density, pressure, and temperature gradients. divided by the speed of sound in the hot channel, which is
For example, at t = 0 ns the temperature increase is about around 500 ns for a strong spark discharge (see section 2.3).
four times higher on the axis in the middle of the gap for In experiments (Rusterholtz et al 2012), the characteristic time
ηR = 60% than for ηR = 15%. Correspondingly, the air for the fast gas heating by the nanosecond spark discharge
density decrease in the first 600 ns on the axis in the middle of is about 20 ns. As part of preparatory works for the results
the gap is two times higher for ηR = 60% than for ηR = 15%. presented in this work, we have carried out simulations with
In figure 8, we also note that the shock wave generated by the a 20 ns relaxation time for fast gas heating with ηR = 30%
low density channel propagates slightly faster at the beginning and no significant differences have been observed on the

11
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

0.38 1450

1400
0.36
1350
0.34
1300

Temperature [K]
Density [kg m-3]

0.32 1250
t=0 ns
0.3 t=100 ns 1200
t=200 ns
t=300 ns 1150
0.28 t=400 ns
t=500 ns 1100
t=600 ns
0.26 t=700 ns
t=800 ns 1050
t=900 ns
0.24 1000
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
r [cm] r [cm]

0.45 2800

2600
0.4
2400
0.35
2200
Temperature [K]
Density [kg m-3]

0.3 2000
t=0 ns
0.25 t=100 ns 1800
t=200 ns
t=300 ns 1600
0.2 t=400 ns
t=500 ns 1400
t=600 ns
0.15 t=700 ns
t=800 ns 1200
t=900 ns
0.1 1000
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
r [cm] r [cm]

Figure 8. Time evolutions of the air density and the air temperature in the radial direction in the middle of the gap every 100 ns after the
reference nanosecond spark discharge of figure 2 at Tg = 1000 K assuming an instantaneous gas heating with ηR = 15% (top figures) and
ηR = 60% (bottom figures).

gas dynamics with the reference case of an infinitely fast we note that the maximum air temperature on the axis in the
gas heating (section 3.1). In this section, we consider no middle of the interelectrode gap obtained at t = 1 µs is 100 K
gas heating during the voltage pulse and we assume that a less than the maximum initial temperature for the case of an
fraction of the discharge energy at the end of the voltage pulse infinitely fast heating (see figure 6). This is due to the fact
ηR eJ (Tpulse ) is put into the air heating with a characteristic that with a relaxation time of τh = 1 µs, the pressure wave
time of τh  Tpulse . As the upper limit value of τh for the is generated and starts to propagate during the temperature
nanosecond spark discharge studied in this work, in this section increase, removing some energy from the hot channel during
we consider τh = 1 µs as observed in experiments on air its formation. For t > 1 µs, the temperature on the axis in the
surface dielectric barrier discharges at atmospheric pressure middle of the gap starts to slowly decrease and the air density
(Aleksandrov 2010). Figure 10 shows the time evolutions of increases slowly up to t = 5 µs.
the air density and of the temperature in the radial direction It is interesting to note that there are two major differences
in the middle of the gap for ηR = 30% with τh = 1 µs up between the case of the air heating with a characteristic time
to t = 5 µs. Figure 11 shows the time evolutions of the air τh = 1 µs and an instantaneous heating: first, due to the
density and of the temperature in the axial direction on the slow increase of the air temperature when a characteristic
symmetry axis for the same conditions. On both figures, we time τh = 1 µs is considered, the amplitude of the generated
note that the air temperature at t = 0 s is equal to 1000 K pressure wave is smaller and smoother than the one obtained for
and increases progressively during 1 µs. Conversely, figure 10 an instantaneously fast heating. Second, with an instantaneous
shows that the air density decreases progressively and that a heating, a rapid expansion of the heated channel is observed
pressure wave is generated on the wings of the hot channel for t
1 µs in figure 6. With an air heating with a relaxation
and propagates radially towards the ambient air. In figure 10, time of τh = 1 µs, figure 10 shows only a small increase of the

12
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

0.36 2100
t=0 ns
0.34 2000 t=200 ns
t=400 ns
t=600 ns
0.32 1900 t=800 ns
t=1000 ns
0.3 1800 t=1200 ns

Temperature [K]
Density [kg m-3]

t=1400 ns
0.28 1700 t=1600 ns
t=1800 ns
0.26 1600

0.24 1500

0.22 1400

0.2 1300

0.18 1200
0.65 0.7 0.75 0.8 0.85 0.65 0.7 0.75 0.8 0.85
x [cm] x [cm]

0.4 5500
t=0 ns
t=200 ns
5000 t=400 ns
0.35 t=600 ns
4500 t=800 ns
t=1000 ns
0.3 t=1200 ns
Temperature [K]
Density [kg m-3]

4000 t=1400 ns
t=1600 ns
t=1800 ns
0.25 3500

3000
0.2
2500
0.15
2000

0.1 1500
0.65 0.7 0.75 0.8 0.85 0.65 0.7 0.75 0.8 0.85
x [cm] x [cm]

Figure 9. Time evolutions of the air density and the air temperature in the axial direction on the symmetry axis every 100 ns after the
reference nanosecond spark discharge of figure 2 at Tg = 1000 K assuming an instantaneous gas heating with ηR = 15% (top figures) and
ηR = 60% (bottom figures).

radius of the hot channel for t
1 µs. As the fast expansion Recently, Naidis (2009) has simulated in 1D the significant
of the hot channel and the shock wave propagation are clearly effect on the power-supply electric circuit on the limitation of
observed in the experimental results from Xu et al (2011), the current of pulsed discharges in air at atmospheric pressure.
our simulation results seem to validate the hypothesis of a However, the modelling of the electric circuit is beyond the
fast heating on short time-scales much less than 1 µs in the scope of this paper and will be addressed in a future work.
nanosecond spark discharge studied in Xu et al (2011). Therefore, as a first step, we have considered a configuration
in which the voltage pulse decreases only a few nanoseconds
3.4. Simulation of the flow expansion following a nanosecond after the connection time, to avoid a too intense spark discharge
spark discharge at Tg = 300 K at Tg = 300 K. In this section, we consider the same geometry
and initial conditions as for the reference nanosecond spark
In this section, we consider a nanosecond spark discharge in discharge defined in section 3.1 at Tg = 1000 K but with an
air at atmospheric pressure at Tg = 300 K. As discussed in applied voltage of 15 kV, scaled relatively to the air density at
Tholin and Bourdon (2011), if we consider a voltage pulse Tg = 300 K and with a pulse duration of Tpulse = 5 ns (with a
with the same duration and we scale the applied voltage rise and a decrease time of 2 ns), shorter than for the reference
with the air density, the air heating is more significant at case at Tg = 1000 K. Figure 12 shows the distributions of
Tg = 300 K than at 1000 K. Then after the connection of the absolute value of the electric field, the electron density
both streamer discharges, the obtained spark is more intense at and the discharge energy at tt = 3, 4 and 5 ns. The gas
Tg = 300 K than at 1000 K. In experiments, the pulsed power- temperature distributions during the voltage pulse are also
supply limits the current delivered during the spark phase. shown, assuming that ηR = 30% of the discharge energy heats

13
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

0.38 1800

0.36 1700
0.34 1600
0.32
1500

Temperature [K]
Density [kg m-3]

0.3
1400
0.28
t=0 ns
t=200 ns 1300
0.26
t=400 ns
t=600 ns 1200
0.24 t=800 ns
t=1000 ns 1100
0.22 t=2000 ns
t=3000 ns
0.2 t=4000 ns 1000
t=5000 ns
0.18 900
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
r [cm] r [cm]

Figure 10. Time evolutions of the air density and the air temperature in the radial direction in the middle of the gap up to t = 5 µs after the
reference nanosecond spark discharge of figure 2 at Tg = 1000 K assuming that ηR = 30% of the discharge energy is put into the air heating
with a characteristic time of τh = 1 µs.

0.4 2800
t=0 ns
2600 t=200 ns
t=400 ns
0.35 t=600 ns
2400 t=800 ns
t=1000 ns
0.3 2200 t=2000 ns
Temperature [K]
Density [kg m ]
-3

t=3000 ns
2000 t=4000 ns
t=5000 ns
0.25
1800

0.2 1600

1400
0.15
1200

0.1 1000
0.65 0.7 0.75 0.8 0.85 0.65 0.7 0.75 0.8 0.85
x [cm] x [cm]

Figure 11. Time evolutions of the air density and the air temperature in the axial direction on the symmetry axis up to t = 5 µs after the
reference nanosecond spark discharge of figure 2 at Tg = 1000 K assuming that ηR = 30% of the discharge energy is put into the air heating
with a characteristic time of τh = 1 µs.

instantaneously the ambient air at Tg = 300 K. In this test- in the reference case at Tg = 1000 K). This constriction of
case, the connection between positive and negative discharges the discharge increases the local current density and then the
occurs at tt = 3.5 ns and then the discharge transits to a spark deposited energy density. At tt = Tpulse = 5 ns, at the end of
phase with a fast increase of energy density and air temperature, the voltage pulse, figure 12 shows that a hot channel is obtained
but as the voltage pulse lasts only 5 ns, the temperature increase in the discharge path, with a non-uniform axial temperature
is less than the one obtained at Tg = 1000 K (figure 2). profile with a hot cylinder from the middle of the gap up to
Figure 12 shows that the energy density and the air temperature the anode tip and a hot spot close to the cathode tip. This
distributions are non-uniform in the interelectrode gap. We temperature field is used as initial condition for the 2D neutral
point out at tt = 4 ns, just after the connection, two hot gas flow equations at t = tt − Tpulse = 0. Figure 13 shows the
spots close to the electrode tips and a third one in the middle gas dynamics following the nanosecond spark at Tg = 300 K.
of the gap. At Tg = 1000 K (figure 2), only the two hot We observe at t = 600 ns a fast depletion of the air density close
spots close to the electrode tips are observed just after the to electrode tips and also in the hot region between the anode
connection. At Tg = 300 K, the third hot spot in the middle of and the middle of the inter-electrode gap. On the pressure
the gap is directly linked to the discharge structure: indeed, in and velocity distributions, we point out that three shock waves
figure 12, we observe at tt = 3 and 4 ns a constriction of the emerge at the same time from the three initial hot spots. At
discharge radius in the middle of the gap (which is not seen t = 1.2 µs the three spherical shock waves merge into a

14
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

tt = 3 ns

tt = 4 ns

tt = 5 ns

Figure 12. Cross-sectional views of the magnitude of the electric field, electron density, discharge energy density and air temperature for a
nanosecond spark in air at Tg = 300 K. The air temperature is obtained assuming an instantaneous gas-heating with ηR = 30% (see
section 2.3).

15
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

Figure 13. Cross-sectional views of density, pressure, air temperature and magnitude of flow velocity for t = 600, 1.2 and 1.8 µs after the
nanosecond spark discharge of figure 12 in air at Tg = 300 K.

16
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

cylindrical shock wave propagating radially at approximately observed the propagation of a weak pressure wave and no
360 m s−1 , i.e. a velocity very close to the speed of sound at dilatation of the hot channel on the discharge path. As
Tg = 300 K. These results are in good qualitative agreement the hot channel expansion and a shock wave propagation
with experiments in Xu et al (2011) at Tg = 300 K. Finally at are clearly observed in the experimental results from Xu
t = 1.8 µs, we observe a hot channel, with a lower temperature et al (2011), our simulation results seem to validate the
and slightly more homogeneous temperature distribution than hypothesis of a gas heating by the nanosecond spark
at t = 0 ns due to the shock wave propagation. For t > 1.8 µs, discharge on very short time-scales. However, additional
we have checked that the pressure converges to the atmospheric more quantitative experimental studies are required to
pressure and then that we obtain a hot, low density channel derive the value of ηR from the comparison of numerical
on the discharge path, as for the reference spark discharge at results and experimental results in Xu et al (2011).
Tg = 1000 K. These results are in good qualitative agreement Finally, in this work, we have simulated the hydrodynamic
with experiments in Xu et al (2011) at Tg = 300 K and seem expansion on short times (t
10 µs) after a single pulse
to confirm the very fast heating of the ambient air by the spark discharge at Tg = 300 and 1000 K to compare with
nanosecond spark discharge in Xu et al (2011). experimental results in Xu et al (2011) and then we have
neglected heat transfer in the gas and to the electrodes.
4. Conclusions Furthermore, in this work, we have assumed that the
applied voltage is independent of the discharge current.
The principal results and contributions, which follow from In a future work, we will model the power-supply electric
studies presented in this paper on air discharges in point-to- circuit used in experiments and heat transfer processes to
point geometry at atmospheric pressure, can be summarized study their influence on the multi-pulse spark discharges
as follows: and on the air hydrodynamics after several successive
voltage pulses.
(i) At Tg = 300 K and 1000 K, we have simulated in 2D
the formation of a nanosecond spark discharge in air at Acknowledgments
atmospheric pressure, the induced air heating and the
following hydrodynamic expansion. In a first step, we The authors thank the Agence Nationale de la Recherche for
have considered that a fixed fraction ηR of the discharge its support of the PREPA (Grant No ANR-09-BLAN-0043-03)
energy instantaneously heats the ambient air. As a and the PLASMAFLAME (Grant No ANR-11-BS09-025-01)
reference, we have used the value of ηR = 30% based projects and Mr Da Xu, Drs Diane Rusterholtz-Duval, David
on the works of Popov (2001) and Aleksandrov (2010). Pai, Deanna Lacoste and Professor Christophe Laux for helpful
At the end of the voltage pulse, we have shown that discussions on their experimental results on nanosecond spark
the energy density and the air temperature distribution discharges in air.
are non-uniform in the interelectrode gap with two hot
spots close to the electrode tips at Tg = 1000 K. At
References
Tg = 300 K, at the end of the voltage pulse, a hot cylinder
is formed from the middle of the gap up to the anode Aleksandrov N L, Kindysheva S V, Nudnova M M and
tip and a hot spot is also seen close to the cathode tip. Stariskovskiy A Yu 2010 Mechanism of ultra-fast heating
After the nanosecond voltage pulse, we have observed the in a non-equilibrium weakly ionized air discharge plasma
formation and propagation of a cylindrical shock wave in high electric fields J. Phys. D: Appl. Phys. 43 255201
Bastien F and Marode E 1985 Breakdown simulation of
and its reflection on the electrode tips for Tg = 300 K and
electronegative gases in non-uniform field J. Phys. D: Appl.
1000 K. The propagation velocity of the cylindrical shock Phys. 18 377
wave is very close to the speed of sound of air at the studied Benilov M S and Naidis G V 2003 Modelling of low-current
Tg . We have also observed the rapid dilatation of the hot discharges in atmospheric-pressure air taking account of
channel formed on the discharge path for t
1 µs. These non-equilibrium effects J. Phys. D: Appl. Phys. 36 1834–41
Bityurin V A, Bocharov A N and Popov N A 2008 Numerical
results are in good qualitative agreement with experiments
simulation of an electric discharge in supersonic flow Fluid
in Xu et al (2011). Dyn. 43 642–53
(ii) In a second step, we have carried out a parametric study Bourdon A, Pasko V P, Liu N Y, Célestin S, Ségur P and Marode E
on the influence of the value of ηR on the air heating 2007 Efficient models for photoionization produced by
and hydrodynamic expansion at Tg = 1000 K, assuming non-thermal gas discharges in air based on radiative transfer
an instantaneously fast gas heating. For all values of and the Helmholtz equations Plasma Sources Sci. Technol.
16 656–78
ηR studied in this work in the range of 15% to 60%, Bourdon A, Bonaventura Z and Celestin S 2010 Influence of the
we have observed the formation and propagation of a pre-ionization background and simulation of the optical
cylindrical shock wave with a propagation velocity close emission of a streamer discharge in preheated air at
to the speed of sound in air at Tg = 1000 K, and the atmospheric pressure between two point electrodes Plasma
rapid dilatation of the hot channel formed on the discharge Sources Sci. Technol. 19 034012
Celestin S, Bonaventura Z, Zeghondy B, Bourdon A and Ségur P
path for t
1 µs. Then we have considered that the 2009 The use of the ghost fluid method for Poisson’s equation
nanosecond spark discharge heats the ambient air with a to simulate streamer propagation in point-to-plane and
longer relaxation time of 1 µs, and in this case we have point-to-point geometries J. Phys. D: Appl. Phys. 42 065203

17
J. Phys. D: Appl. Phys. 46 (2013) 365205 F Tholin and A Bourdon

Demmel J W, Eisenstat S C, Gilbert J R, Li X S and Liu J W H 1999 Pancheshnyi S 2005 Role of electronegative gas admixtures in
A supernodal approach to sparse partial pivoting SIAM J. streamer start, propagation and branching phenomena Plasma
Matrix Anal. Appl. 20 720–55 Sources Sci. Technol. 14 645–53
Ferziger J and Peric M 2002 Computational Methods for Fluid Dyn. Pilla G, Galley D, Lacoste D A, Lacas F, Veynante D and Laux C O
3 edn (Berlin: Springer) 2006 Stabilization of a turbulent premixed flame using a
Kulikovsky A A 1995 A More Accurate Scharfetter–Gummel nanosecond repetitively pulsed plasma IEEE Trans. on Plasma
algorithm of electron transport for semiconductor and gas Sci. 34 2471–7
discharge simulation J. Comput. Phys. 119 149–55 Popov N A 2001 Investigation of the mechanism for rapid heating of
Kulikovsky A A 1997 Positive streamer between parallel plate nitrogen and air in gas discharges Plasma Phys. Rep.
electrodes in atmospheric pressure air J. Phys. D: Appl. Phys. 27 886–96
30 441–50 Popov N A 2011a Fast gas heating in a nitrogen-oxygen discharge
Liu N, Celestin S, Bourdon A, Pasko V P, Segur P and Marode E plasma: I. Kinetic mechanism J. Phys. D: Appl. Phys.
2007 Application of photoionization models based on radiative 44 285201
transfer and the Helmholtz equations to studies of streamers in Popov N A 2011b Kinetic processes initiated by a nanosecond
weak electric fields Appl. Phys. Lett. 91 211501 high-current discharge in hot air Plasma Phys. Rep. 37 807–15
Maly R 1984 Spark ignition: its physics and effect on the internal Riousset J A, Pasko V P and Bourdon A 2010 Air density dependent
combustion engine Fuel Economy in Road Vehicles Powered model for analysis of air heating associated with streamers,
by Spark Ignition Engines ed J C Hilliard and G S Springer leaders, and transient luminous events J. Geophys. Res.
(New York: Plenum) pp 91–148 115 A12321
Marode E 1975a The mechanism of spark breakdown in air at Rusterholtz D L 2012 Nanosecond repetitively pulsed discharges in
atmospheric pressure between a positive point and a plane: atmospheric pressure air PhD Thesis Ecole Centrale Paris
I. Experimental: nature of the streamer track J. Appl. Phys. France
46 2005–15 Rusterholtz D L, Pai D Z, Stancu G D, Lacoste D A and Laux C O
Marode E 1975b The mechanism of spark breakdown in air at 2012 Ultrafast heating in nanosecond discharges in
atmospheric pressure between a positive point and plane: atmospheric pressure air 50th AIAA Aerospace Sciences
II. Theoretical: computer simulation of the streamer track J. Meeting including the New Horizons Forum and Aerospace
Appl. Phys. 46 2016–20 Exposition (Nashville, TN January, 2012) AIAA-2012-0509
Marode E, Bastien F and Bakker M 1979 A model of the Shu C-W 1997 Essentially non-oscillatory and weighted essentially
streamer-induced spark formation based on neutral dynamics non-oscillatory schemes for hyperbolic conservation laws
J. Appl. Phys. 50 140–6 NASA/CR-97-206253, ICASE Report No 97-65
Mintoussov E I, Pendleton S J, Gerbault F G, Popov N A and Starikovskaia S M 2006 Plasma assisted ignition and combustion
Starikovskaia S M 2011 Fast gas heating in nitrogen-oxygen J. Phys. D: Appl. Phys. 39 R265
discharge plasma: II. Energy exchange in the afterglow of a Starikovskii A Y, Nikipelov A A, Nudnova M M and Roupassov D V
volume nanosecond discharge at moderate pressures J. Phys. 2009 SDBD plasma actuator with nanosecond pulse-periodic
D: Appl. Phys. 44 285202 discharge Plasma Sources Sci. Technol. 18 034015
Mnatsakanyan A K and Naidis G V 1985 The vibrational energy Tholin F 2012 Numerical simulation of nanosecond repetitively
balance in a discharge in air High Temp. 23 506–13 pulsed discharges in air at atmospheric pressure: application to
Morrow R and Lowke J J 1997 Streamer propagation in air J. Phys. plasma-assisted combustion PhD Thesis Ecole Centrale Paris,
D: Appl. Phys. 30 614–27 France
Naidis G 1999 Simulation of streamer-to-spark transition in short Tholin F and Bourdon A 2011 Influence of temperature on the glow
non-uniform air gaps J. Phys. D: Appl. Phys. 32 2649–54 regime of a discharge in air at atmospheric pressure between
Naidis G V 2005 Dynamics of streamer breakdown of short two point electrodes J. Phys. D: Appl. Phys. 44 385203
non-uniform air gaps J. Phys. D: Appl. Phys. 38 3889 Tholin F and Bourdon A 2013 Simulation of the stable
Naidis G V 2008 Simulation of spark discharges in high-pressure air ‘quasi-periodic’ glow regime of a nanosecond repetitively
sustained by repetitive high-voltage nanosecond pulses J. Phys. pulsed discharge in air at atmospheric pressure Plasma Sources
D: Appl. Phys. 41 234017 Sci. Technol. 22 045014
Naidis G V 2009 Simulation of streamer-induced pulsed discharges Tholin F, Rusterholtz D L, Lacoste D A, Pai D Z, Celestin S,
in atmospheric-pressure air Eur. Phys. J. Appl. Phys. 47 22803 Jarrige J, Stancu G D, Bourdon A and Laux C O 2011 Images
Pai D Z 2011 Nanomaterials synthesis at atmospheric pressure using of a nanosecond repetitively pulsed glow discharge between
nanosecond discharges J. Phys. D: Appl. Phys. 44 174024 two point electrodes in air at 300 K and at atmospheric pressure
Pai D Z, Stancu G D, Lacoste D A and Laux C O 2009 Nanosecond IEEE Trans. Plasma Sci. 39 2254–55
repetitively pulsed discharges in air at atmospheric van Leer B 1979 Towards the ultimate conservative difference
pressure—the glow regime Plasma Sources Sci. Technol. scheme: V. A second-order sequel to Godunov’s method
18 045030 J. Comput. Phys. 32 101–36
Pai D Z, Lacoste D A and Laux C O 2010a Transitions between Vitello P A, Penetrante B M and Bardsley J N 1994 Simulation of
corona, glow, and spark regimes of nanosecond repetitively negative-streamer dynamics in nitrogen Phys. Rev. E
pulsed discharges in air at atmospheric pressure J. Appl. Phys. 49 5574–98
107 093303 Xu D A, Lacoste D A, Rusterholtz D L, Elias P-Q, Stancu G D and
Pai D Z, Lacoste D A and Laux C O 2010b Nanosecond repetitively Laux C O 2011 Experimental study of the hydrodynamic
pulsed discharges in air at atmospheric pressure–the spark expansion following a nanosecond repetitively pulsed
regime Plasma Sources Sci. Technol. 19 065015 discharge in air Appl. Phys. Lett. 99 121502

18