See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/230926875

Spectroscopic studies and rotational and vibrational temperature

measurements of atmospheric pressure normal glow plasma discharges in air

Article in Plasma Sources Science and Technology · September 2006

DOI: 10.1088/0963-0252/15/4/027

CITATIONS READS

151 1,004

4 authors, including:

David Staack Bakhtier Farouk

Texas A&M University Drexel University
134 PUBLICATIONS 2,177 CITATIONS 251 PUBLICATIONS 4,787 CITATIONS

SEE PROFILE SEE PROFILE

Alexander F. Gutsol
Actively searching for new opportunity
180 PUBLICATIONS 8,856 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Piezoelectric Actuators Design and Piezoelctric Micropump Design View project

Optimization View project

All content following this page was uploaded by Alexander F. Gutsol on 01 June 2014.

The user has requested enhancement of the downloaded file.

INSTITUTE OF PHYSICS PUBLISHING PLASMA SOURCES SCIENCE AND TECHNOLOGY
Plasma Sources Sci. Technol. 15 (2006) 818–827 doi:10.1088/0963-0252/15/4/027

Spectroscopic studies and rotational and

vibrational temperature measurements of
atmospheric pressure normal glow plasma
discharges in air
David Staack, Bakhtier Farouk, Alexander F Gutsol and
Alexander A Fridman
Department of Mechanical Engineering and Mechanics, Drexel University, Philadelphia, PA
19104, USA
E-mail: [email protected]

Received 2 May 2006, in final form 10 August 2006

Published 20 September 2006
Online at stacks.iop.org/PSST/15/818

Abstract
DC normal glow (NG) discharges were created in atmospheric pressure air
for a pin to plate type geometry. The rotational and vibrational temperatures
of the discharge were measured by comparing modelled optical emission
spectra with spectroscopic measurements from the discharge. The
temperatures were measured as a function of discharge current, ranging
from 50 µA to 30 mA, and discharge length, ranging from 50 µm to 1 mm.
Rotational temperatures from 400 to 2000 K were measured over this range.
Vibrational temperatures vary from 2000 K to as high as 5000 K indicating a
non-equilibrium plasma discharge. Spectroscopic measurements were
compared using several different vibrational bands of the 2nd positive
system of N2 , the 1st negative system of N+2 and the UV transitions of NO.
NO and N+2 transitions were also used to determine the electronic
temperature and N+2 density. The discharge temperature appears to be
controlled by two cooling mechanisms: (1) radial conductive cooling which
results in an increase in temperature with increasing discharge current and
(2) axial cooling to the electrodes which results in a temperature saturation
with increase in discharge current. The measured discharge temperature
initially increases rapidly with discharge current then becomes nearly
constant at a higher discharge current. Thus, radial cooling appears to
dominate at lower discharge currents and the axial cooling at higher
discharge currents. The vibrational temperature decreases with increasing
rotational temperature due to increased vibrational to translation relaxation
but the discharge remains non-thermal and stable over the range studied.
The discharge appears to have a maximum vibrational temperature at the
low current limit of the NG regime.
(Some figures in this article are in colour only in the electronic version)

1. Introduction conditions [2]. A requirement for many of these techniques

is the creation of a non-thermal plasma. Non-thermal plasmas
Atmospheric pressure plasma discharges [1] are of growing (also known as non-equilibrium plasmas or cold plasmas)
interest because of the possibility of lower facility and are typically characterized by their electron temperature (Te ),
process costs for a variety of plasma processing and micro- vibrational temperature (Tvib ), rotational temperature (Trot )
manufacturing techniques currently performed in vacuum and translational temperature (Ttrans ). In non-thermal plasmas

0963-0252/06/040818+10$30.00 © 2006 IOP Publishing Ltd Printed in the UK 818

 Spectroscopic studies and rotational and vibrational temperature measurements of atmospheric pressure normal glow plasma discharges

created by externally applied electric fields typically Te > Tvib >

Trot = Ttrans . In plasma assisted materials processing the non-
equilibrium nature allows for the creation of active species
to interact with the substrate without generating excessive
heat which could destroy the substrate. In plasma chemistry
applications the non-equilibrium nature allows for the creation
of chemically active species by lowering activation potentials
without generating excessive heat which leads to excessive
dissociation and other problems.
Large volume, spatially uniform, non-thermal plasmas
can readily be created at low pressures because of the
relative lack of interspecies collisions and rapid cooling
to the walls of the discharge chamber [3]. One reason
atmospheric pressure discharges are not widely used in micro-
fabrication is that the increase in power density and increased
collision frequency make the attainment of stable non-
thermal discharges more difficult [1]. Common atmospheric
pressure arcs are thermal equilibrium plasma discharges Figure 1. Photo of the experimental setup which shows the
with characteristic temperatures of about 10 000 K. Dielectric spectrometer, microscope and discharge chamber.
barrier discharges (DBDs) and corona discharges are non-
equilibrium atmospheric discharges near room temperature temperatures but was studied because of its simplicity. More
but have spatial and temporal non-uniformities as well as sophisticated electrode geometries and added gas flows [4,10]
low concentrations of active species which can limit their can probably be viewed as a variant of the dc NG discharge
applications. Atmospheric pressure plasmas have recently studied here. Air was studied because it is readily attainable,
been developed which lie between the corona and arc regimes used in practical applications and because air represents a
[4–6]. Due to ‘pd’ and ‘normal current density’ scaling working gas in which a stable discharge is difficult to attain due
these discharges are typically sub-millimetre in size. These to the presence of oxygen. The temperatures were measured
‘microplasmas’ [7] are non-equilibrium, but still warm, with using optical emission spectroscopy because the technique is
characteristic gas temperatures measured between 1000 and non-intrusive, accurate over a wide range of conditions and
2000 K [4, 6, 8]. We have shown [6] that many of these dc can differentiate between the temperatures of different internal
microplasmas are pressure and temperature scaled versions of energy modes.
the normal glow (NG) discharges observed in vacuum.
The exception to the similarity between low pressure 2. Experimental setup
normal glow (LPNG) and atmospheric pressure normal glow
(APNG) discharges is that at atmospheric pressure the Figure 1 is a photo of the experimental setup used. This
discharge that is often not confined by side walls and the gas experimental setup is similar to that used in our previous studies
temperature can no longer be considered constant and near [6] although several improvements have been made. Most of
room temperature. These thermal effects can significantly these improvements were made to support ongoing research
affect the physics of the discharges, the non-equilibrium nature into using the microplasmas for atmospheric pressure plasma
and their applicability for use in micro-fabrication and other enhance chemical vapor deposition (AP-PECVD) and were
processes [9]. For example, at constant pressure an increase not directly relevant to these experiments. The improvements
in the gas temperature and corresponding decrease in density, made are (1) a tightly sealed vessel is used for operating the
n, thus changes the reduced electric field, E/n,assuming discharge at lower pressures and in other gases at high purity,
constant electric field E. The reduced electric field in turn (2) a three axis computer controlled positioning system is used
determines Te and the rates of electron-initiated chemical to control the position of the electrode, (3) a thermocouple and
reactions and ionization rates. Also an increase in the gas heater capable of heating substrates to 700 ◦ C were included
temperature exponentially increases the rate of energy transfer in the planar electrode and (4) electrical connections are
between vibrational and translational modes, thus decreasing provided for RF excitation and current measurements. The
the non-equilibrium nature of the discharge and promoting discharge chamber is a 10 cm cube with two quartz windows
unstable transition to an arc. For some applications an for spectroscopy and visualizing the discharge. The chamber
excessive increase in gas temperature can be undesirable is sealable and there are gas inlets and outlets for testing in
due to overheating of the substrate and increased molecular a variety of pressures and discharge gases. In the studies for
dissociation. atmospheric pressure air no gas flow was used and the chamber
Detailed parametric temperature measurements are was unsealed as shown in figure 1.
reported here for atmospheric pressure glow discharges in In these experiments a thin wire anode is suspended over a
air. The goals were to find useful operating regimes, compile planar cathode. Although this may traditionally be considered
results for comparison to future modelling and understand the a pin to plate configuration the electrodes may actually be
nature of the temperature dependencies. NG discharges in considered to be two planar electrodes for several reasons:
non-flowing air between effectively parallel electrodes were the diameter of the discharge is several times smaller than the
studied. This geometry may not be ideal for attaining low diameter of the anode electrode; an increase in the diameter

819
D Staack et al

emission spectra (OES) of the discharge were taken in a range

1000:1 HV probe
200–450 nm averaging emission from the entire discharge.
Oscilloscope
A Roper Scientific model 7430 CCD camera was mounted onto
the exit slit to digitally acquire the spectra at approximately
0.01 nm resolutions. Background images with the discharge
Ballast off were subtracted and a low pressure mercury lamp was used
Anode to determine the ‘slit (apparatus) function’ and to calibrate the
Current
Power probe Plasma spectrometer.
Supply
Cathode 3. Spectroscopic model and best fit temperatures
Shunt Discharge Chamber SPECAIR [8, 11, 12] was used to model OES and for the
determination of the temperature by comparisons between the
Figure 2. Schematic of the circuit to generate and analyse the glow model and experimental spectra. SPECAIR is a FORTRAN
discharge. program which can simulate the emission spectra of air in
the ultra-violet (UV), visible and near infrared (NIR) ranges.
of the anode wire does not change the characteristics of the The emission spectra are functions of the temperature of the
discharge; the end of the wire is flat, and the edges of the wire air and the relative population of different atomic/molecular
are rounded. The cathode plate was 25 mm square and 9.4 mm species within the air mixture. SPECAIR is able to
thick. The anode was 0.5 mm in diameter attached to a 12 mm simulate air in non-equilibrium conditions. This is done
diameter stainless steel anode holder which acts as a heatsink. by allowing the temperatures of translational, rotational,
The anode is fixed in place, and the cathode can be moved vibrational and electronically excited states (Ttrans , Trot , Tvib ,
through a flexible bellows using a micrometer to change the Telex ) to be arbitrarily specified. Telex is the electronic
inter-electrode spacing. excitation temperature which corresponds to the Boltzmann
The circuit used to generate the glow discharge is similar distribution among excited electronic states and may differ
to that used before [6] and that used in low pressure glow from Te , the electron temperature which is the translational
discharges [3]. Figure 2 is a schematic of the circuit used temperature of free electrons. SPECAIR can also simulate
for creating and analysing the discharge. A dc power supply various air component mixtures since the relative population of
is connected in series to a ballast resistor, a shunt resistor the different species within air can also be arbitrarily specified.
and the discharge. Various values of shunt resistor, ballast SPECAIR was originally used and validated in determining
resistor, power supply polarity and applied voltage were used temperatures for thermal equilibrium and chemically non-
depending on the regime of study. For studies in the NG equilibrium discharges in excess of 2000 K. However, the
discharge regime in air the shunt resistance was 100
and the approach and the code have been used for thermally non-
ballast was 200 k
. The ballast resistor acts to electrically equilibrium discharges with temperatures approaching room
stabilize the discharge circuit ensuring that the discharge temperature [13–17].
current increases monotonically with an increase in power SPECAIR calculates the OES by determining the
supply voltage. This is necessary since a dc electric discharge population of the upper state of the transitions as determined
has a nonlinear voltage versus current characteristic. The by a Boltzmann distribution and the specified temperature of
ballast resistor is placed physically close to the discharge in the internal energy mode. A collisional–radiative model is not
order to reduce the stray capacitance of the circuit and to used to determine the upper state populations from the ground
improve the stability of the discharge. The shunt resistor is state populations. The modelled transition probabilities are
connected to an oscilloscope and is used to measure dc and low calculated based upon tabulated data for a given set of
frequency current. A Pearsons model 2877 current transformer transitions. In the present calculations the following transitions
probe was connected inline to measure high frequency current were used: N2 2nd positive system (N2 C 3 u → B 3 g ),
oscillations up to 50 MHz. A BK-precision high impedance, N+2 1st negative system (N+2 B 2 u+ → X 2 g+ ), and the NO
1000 : 1, high voltage (HV) probe connected to an oscilloscope β, γ , δ and ε systems (NO B 2  → X 2 , A2  → X 2 ,
is used to measure the discharge voltage and oscillations up to C 2  → X 2  and D 2  → X 2  transitions, respectively).
1 MHz. From the calculated probability and population, the light
For visualizing the discharge, a 640 × 480 resolution emission intensity is calculated at the wavelength of the
DIGITAL colour CCD video camera was mounted on a transition. A line shape is determined from the natural line
microscope focused on the discharge. The microscope–camera width, thermal Doppler broadening and collisional broadening.
setup provided a variable magnification allowing between a The intensities from all calculated transitions and species
2 mm × 1.5 mm and a 20 mm × 15 mm field of view. The are combined at each wavelength in a specified wavelength
corresponding minimum pixel size at maximum magnification range. This ‘narrow line’ spectra is then convolved with the
was thus about 3 µm. Still images from the CCD camera were spectrometer ‘slit function’ to model the observed spectra.
captured digitally by a PC. In thermally non-equilibrium plasmas, because of their
For spectroscopy, an Acton Research SpectraPro 500i very nature, there are at least five temperatures Ttrans ,
scanning monochromator was used. Light from the discharge Trot, Tvib , Telex , Te . These temperatures cannot be assumed
was focused onto the entrance slit of the spectrometer to be equal nor can they be assumed the same from species
using a UV-lens and mirror as shown in figure 1. Optical to species. Even the often assumed conditions such as

820
 Spectroscopic studies and rotational and vibrational temperature measurements of atmospheric pressure normal glow plasma discharges

Tvib = Telex = Te should not be assumed and in some cases to speak of an apparent Tvib and Telex . Although in practice
have been directly shown to be incorrect [12]. The modelled when comparing the experimental rotational and vibrational
spectra are sensitive to Trot , Tvib , Telex because the upper states line structure with Boltzmann computed distributions of the
of the emission transition have different electronic excitation upper state there is fairly good superposition as shown later in
energies, vibrational energy and/or rotational energies. The figure 4. In those cases the good superposition indicates that
population of the upper states is given by equation (1) [12] the Boltzmann assumption may be valid.
which assumes the internal energy levels follow Boltzmann In the present study, the temperature measurements were
distributions at Telex , Tvib and Trot , respectively. n is the made by comparing the modelled spectra with the experimental
population of the upper state (denoted by ‘) with a given spectra. The accuracy of the fit was determined by the root
electronic excitation (e), vibrational (v) and rotational (J ) mean square error (RMSE). The best fit corresponds to the
quantum level. ge and (2J +1) are the degeneracy (or statistical RMSE closest to zero. The conditions of the discharge were
weighting) of the electronic excitation and rotational internal determined to be those parameters which result in the best fit.
energy levels, respectively; the vibrational level degeneracy is The model fitting parameters were Ttrans , Trot , Tvib , Telex and
one. L is line alteration factors due to nuclear spin and σ is 2 for species concentrations. Depending on the spectral range the
homo-nuclear molecules and 1 for hetero-nuclear molecules. best fit might only be sensitive to a subset of these parameters.
ε is the energy of the respective level. The summation in the Additional parameters which change the spectra are due to
denominator represents the combined rotational, vibrational the experimental setup. Some of them are (1) wavelength offset
and electronic excitation partition function: due to spectrometer resolution and calibration, (2) wavelength
ne v J  dispersion due to spectrometer optics, (3) intensity offset

 due to background signal and noise, (4) changes in camera
εe  εv εJ 
ge (2J + 1) exp − − − sensitivity and light transmission with wavelength and (5)
L Telex Tvib Trot
= ntotal
. apparatus line broadening due to the dimensions of the slit
σ  εe εv εJ entrance aperture. Items 1 to 4 can be taken into account
g
e,v,J e (2J + 1) exp − − −
Telex Tvib Trot by correcting the experimental measured wavelength, λ, and
(1) intensity, I. Assuming linear variations we have λcorrected =
The term exp(−ε/T ) in equation (1) represents the λ + C1 + C2 × (λ − λ0 ) and Icorrected = I + C3 + C4 × (λ − λ0 ),
Boltzmann distribution. However, it is not always correct where C1 to C4 correspond, respectively, to the parameters
to assume that within a given internal energy mode the described and listed as (1) to (4) above and λo is some
energy distribution function is Boltzmann. For rotational reference wavelength (in our case the middle wavelength of
energy the assumption of a Boltzmann distribution with Trot the range). The scanning monochromator used can acquire the
is accurate because the energy levels are closely spaced, and spectrum over an approximately 10 nm range in a single camera
the characteristic temperature of the rotational energy mode is image. Over such short wavelength ranges it is reasonable to
significantly less than the typical rotational temperature. For assume C1, C2, C3 and C4 constant, with C4 = 0. Longer
most species it can be assumed that the Trot is close to Ttrans and wavelength ranges can be acquired by turning the grating and
is effectively considered Tgas of the mixture because rotational- stitching together with the images from different wavelength
to-translational relaxation is fast at atmospheric pressure [12]. sub-ranges. For longer wavelength ranges C1, C2, C3, C4
For vibrational and electronically excited states the are piecewise non-constant and can complicate correcting the
temperature of the mode is on the order of the characteristic experimental spectra. For simplicity and accuracy in most
energy of the mode and assumption of a Boltzmann distribution cases only normalized spectra were compared corresponding
can be inaccurate since the quantization of the states becomes to a single camera image. The apparatus line broadening
important and the rates of specific paths of energy transfer is taken into account by convolving the modelled natural
from state to state need to be considered. The deviation (narrow-line) spectra with the slit function. The slit function
from Boltzmann can occur in several ways. For example, was measured and then approximated using a pseudo-Voight
in certain vibrational systems Treanor distributions have been function of the same characteristic full width at half maximum
measured [9] due to preferential energy transfer to the higher (FWHM). The range of uncertainty in these experimental
vibrational states. A deviation from Boltzmann can also be parameters was reduced by using a mercury lamp to calibrate
due to the fact that OES only probes the states of electronically the spectrometer and measured the slit function. However,
excited species. It is possible that the distribution functions are better fits to experimental data could be achieved by using C1,
altered by the electronic excitation. For Trot it has been shown C2, C3 and the FWHM of the slit function as fitting parameters.
that the rotational distribution of the upper state is sufficiently This is because the slit function cannot be measured for all
similar to that of the lower state to have equal temperatures wavelengths and the calibration cannot be accurate over all
even with small changes due to the excitation [13]. For the wavelength ranges. The search for the best fit conditions was
Tvib and Telex collisional–radiative models attempt to take into automated using a MATLAB routine based on the simplex
account such processes but were not used in our modelling. search method [18]. This allowed for the rapid analysis of a
Another complication is that in atmospheric pressure plasma large number of experimental spectra.
the de-excitation time of the electronically excited states is
non-negligible in comparison with the time between collisions. 4. Experimental results and discussion
For N2 C–B transitions the de-excitation time is a few tens of
nanoseconds and at atmospheric pressure the time between Figure 3 is an image of the APNG in air for a current of
collisions is about 1 ns. For these reasons it is only possible 3.0 mA and an electrode spacing of 400 µm and is similar to

821
D Staack et al

illustrated by plotting the RMSE value as a function of the

fitting parameters. Figure 5 is a contour plot of the RMSE as
a function of Tvib and Trot for (a) the 334 nm and (b) 375 nm
spectra. The best fit temperatures correspond to the minimum
value of the RMSE, RMSEbestfit . Comparisons of the fitting
error, RMSE, with the statistical error, STD, where STD is the
standard deviation of multiple experimental spectra, can be
used to determine the sensitivity and precision of the spectral
model with regard to the fitting parameters. In figure 5 the
contour plots of RMSE as functions of modelled Trot and
Tvib are shown. The contour corresponding to a value of
RMSEbestfit + STD is the blue-filled contour in the centre and
represents the conditions where the change in error between
experiment and model is on the order of the statistical error
in the data. Any Tvib , Trot pair within this contour is within
Figure 3. APNG in air at 3.0 mA current and 400 µm electrode the errors of the measurements. Thus, the width and high of
spacing. the RMSEbestfit + STD contour can be used to determine an
error bar for Trot and Tvib , respectively. We calculate that at
those reported earlier [6]. The anode wire is visible in the top 334 nm the Tvib cannot be precisely determined (worse than
of the image and the cathode plane is visible at the bottom. +4000 K/−2000 K) and Trot can be determined to within ±
Distinct features which characterize this as a NG discharge are 39 K. For the 375 nm spectra the precision in determining
the bright region near the cathode, the negative glow, above that the temperatures were ± 130 K and ± 37 K for Tvib and Trot ,
the Farady dark space and lastly the luminous positive column respectively. The developers of SPECAIR use an accuracy of
continuing to the anode. Due to ‘pd’ and normal current ± 50 K [8] for rotational and vibrational temperature. The total
density scaling, the characteristic sizes of the NG discharge are error is due to both inaccuracies of the model and imprecision
hundreds of microns when operating at atmospheric pressure. of the experiments. For measurements using the 375 nm
We see from this figure that although the geometry is pin-to- spectra the error bars for vibrational temperature determined
plate because of the small size of the discharge as described are less than ± 180 K; the error bars for rotational temperature
earlier it is effectively plate-to-plate. are ± 87 K. Another matter illustrated by figure 5(b) is the
The experimental results are reported in three sections. ‘line of approach’ to the best fit rotational and vibrational
Section 1 focuses on calculating the error bars associated temperatures (as shown by the dashed lines and represent either
the best fit Trot for a given Tvib or the best fit Tvib for a given
with the spectroscopic measurements and determining the best
Trot ); when parallel to axes this indicates that the best fit Trot
transitions to use for measuring Trot and Tvib . Section 2 focuses
and Tvib are decoupled. The search routine to find the best fit
on the use of the NO and N+2 transitions for determining Telex
temperatures is expedited by taking this into account.
and the relative N+2 density. Section 3 reports the results of a
Figures 6 and 7 are plots of the rotational and vibrational
parametric study. Trot and Tvib are measured as a function of the
temperature as a function of discharge current comparing
discharge current and the electrode spacing. A discussion and
measurements using the four spectra ranges shown in figure 4.
explanation of the observed trends is also included in section 3.
The error bars were calculated for each spectra range as
described above. As can be seen the measured temperatures are
4.1. Spectral sensitivity of results within the error bars of measurements using other transitions.
For the majority of the temperature measurements made
Figure 4 shows the measured spectra and corresponding
later in this paper the transitions around 375 nm were used
best fit modelled spectra for a 2 mA discharge and 300µm
because they could most precisely determine both Trot and
electrode spacing for spectra centred at (a) 296 nm, (b) 313 nm,
Tvib . Compared with the transition at 334 nm, the transition at
(c) 334 nm and (d) 375 nm, respectively. These transitions
375 nm provides a significantly more precise Tvib since several
correspond to different vibrational bands of the N2 (C–B)
strong vibrational transitions are compared. Compared with
transitions. These transitions were chosen because they
the transitions at 296 and 313 nm, the transition at 375 nm
represent regions of the spectra where the N2 (C–B) transitions
is more intense and the spectrometer has higher resolution
can be seen independent of transitions for different species
resulting in slightly better determination of both Tvib and Trot .
and electronic states. These spectra are thus independent of
From these measurements we see that the discharge is non-
species fractions and Telex . The relative heights of two band equilibrium Tvib > Tgas over a range of discharge currents.
heads with different initial vibrational quantum number are
sensitive to Tvib . The vibrational band heads are labelled
in the graphs. The band tails are due to sub-transitions 4.2. Measurements of electronic temperature and ion density
between different rotational quantum numbers and are the NO transitions in the range from 218 to 228 nm were
small frequent overlapped peaks decaying to the left of the used to determine the vibrational, electronic and rotational
band head in figure 4. The relative height of the band head to temperatures. The strongest lines in this range are due to
the tail is sensitive to Trot . the γ transitions, also weaker but significant are the β, δ
The sensitivity of the fitting procedure and the precision and ε transitions. Each of these transitions has a different
to which the fitting parameters can be determined can be initial electronic state and so the OES is sensitive to Telex .

822
 Spectroscopic studies and rotational and vibrational temperature measurements of atmospheric pressure normal glow plasma discharges

Figure 4. Measured spectra and corresponding best fit modelled spectra for a 2 mA discharge for spectra centred at (a) 296 nm, (b) 313 nm,
(c) 334 nm and (d) 375 nm.

Within the γ transitions there are different vibrational levels excitation space and results in an overpopulation of the ground
and rotational levels and thus the spectra are sensitive to Trot state and low-level electronic states relative to the equilibrium
and Tvib . Figure 8 is a plot of the modelled and experimental population. This overpopulation would be perceived as a lower
spectra for a 3.8 mA discharge in air with electrode spacing Telex . A lower bound on Telex of Tvib might occur due to
of 300 µm. The best fit temperatures were Trot = 2200 K, collisions and vibrational–excitation energy transfer: however,
Tvib = 4077 K and Telex = 4074 K. The rotational temperatures such complexities in molecular systems such as air are poorly
measured are higher than for the same discharge using the understood.
N2 lines around 375 nm. This may be due to the fact that Spectra in the range from 351 to 359 nm can be used to
NO is produced as a result of chemical reactions and residual determine the relative densities of N+2 and N2 . The SPECAIR
chemical energy may manifest itself as elevated temperatures modelled OES from this range are due to both N2 and N+2
[19]. That Telex and Tvib are measured to be equal is very transitions. The relative concentration of nitrogen ions N+2 /N2
interesting. It should be restated that this Telex corresponds can thus be determined and used to approximate a lower bound
to the distribution of electronic energy among the excited on the ionization fraction and plasma density in the discharge.
states (A, B, C, D) and not the electronic temperature relative N+2 is not the only ion present in the air plasma; however, it is
to the ground (X) state. One might expect that Te = typically the most prevalent ion. Figure 9 shows a spectrum
Telex since the electronically excited states are attained by attained around 355 nm for a 3.8 mA discharge at an electrode
collision with electrons and should mirror the electron energy spacing of 300 µm. The two larger band heads at 357.5 and
distribution function. However, in atmospheric pressure 353.3 nm are due to the 1–0 and 2–1 vibrational transition of
plasmas lower excitation temperatures as in our case with Texec the N2 2nd positive system, the smaller peak at 358.3 nm is
near Tvib are measured [17,20]. Such non-equilibrium systems due to the N+2 1st negative system. The N2 transitions can be
may be classified as ionizing systems [21] where excitation used to determine both Trot and Tvib . The relative intensity
and ionization exceed recombination and radiative decay. of the two systems is due to both species concentration and
Excitation and ionization is balanced by particle diffusing Telex as shown in the inset equation of figure 9, where I is the
from the plasma region. There is thus a net upward flow in intensity of the line, n is the density of the species and ε is the

823
D Staack et al

9000
296nm

Vibrational Temp (K)

8000 313nm
334nm
7000 375nm
6000
5000
4000
3000
2000
1 2 3 4
Discharge Current (mA)

Figure 7. Plot of Tvib versus discharge current comparing results

using different vibrational transitions of the N2 2nd positive system.
(a)

(b)
Figure 5. Contour plots of RMSE as a function of modelled Trot and Figure 8. Experimental and modelled NO spectra used to determine
Tvib for spectra around (a) 334 nm and (b) 375 nm attained from a Trot = 2200 ± 100 K, Tvib = 4077 ± 250 K, Telex = 4074 ± 400 K,
2 mA discharge and 300 µm electrode spacing. for a 3.8 mA glow discharge in air and 300 µm electrode spacing.

1500 accurate within a factor of 2. One estimate of N+2 /N2 can

296nm be implemented by assuming Telex = Tvib . This assumption
313nm
Rotational Temp (K)

1400
334nm is somewhat justified by the measurements using the NO
1300 375nm spectra where Tvib = Telex was found. Progressing from the
measurement of Tvib (NO) = Telex (NO) to the assumption of
1200
Tvib (N2 ) = Telex (N2 ) = Telex (N+2 ), however, is not entirely
1100 justified but was implemented nontheless to determine N+2 /N2 .
Using the spectra in figure 9 we determined Trot = 1282 K, Tvib
1000
= 5015 K and N2+ /N2 = 1.25 × 10−8 . This corresponds to a
900 nitrogen ion density of about 5.5 × 1010 cm−3 . Similar values
1 2 3 4 of the ion density have been estimated for other atmospheric
Discharge Current (mA) pressure discharges in air [22]; however, measurements and
estimates of electron densities have a somewhat higher value
Figure 6. Plot of Trot versus discharge current comparing results
using different vibrational transitions of the N2 2nd positive system.
of about 1013 cm−3 [16, 23]. A possible explanation for the
low nitrogen ion density is that N+4 [24] or another ion is the
dominant ion in the air plasma. Another possible assumption
energy of the excited state. An increase in the height of the for interpreting the spectra is to assume Telex = Te . Te was
N+2 line could be due to either an increase in the ion density estimated from the measured electric field and density to be
or a decrease in Telex since the energy of the N+2 excited state about 1.2–1.4 eV [6]. Using this assumption we determine
is less than the energy of the N2 excited state (εN2+ < εN2 ). N2+ /N2 = 2 × 10−3 and nitrogen ion density of 9 × 1015 cm−3 ,
Telex and N+2 /N2 thus cannot be determined independently. which is at the upper limit of what might be expected for a glow
The multiple bestfit solutions can be approximated by the discharge [7]. The appropriate Telex to use in this situation is
equation N2+ /N2 = 1.383 × exp(−90500/Telex ) which is unclear.

824
 Spectroscopic studies and rotational and vibrational temperature measurements of atmospheric pressure normal glow plasma discharges

1900
100um
1700 300 um

Temperature (K)
1500

1300

1100

900

700

500

300
0 5 10 15 20 25 30 35

Current (mA)
Figure 10. Variation of rotational temperature with discharge
current for a glow discharge in atmospheric pressure air at 100 and
300 µm electrode spacing.
Figure 9. Experimental and modelled N2 and N+2 spectra used to
determine Trot = 1282 ± 100 K, Tvib = 5015 ± 250 K and
N2+ /N2 = 1.25 × 10−8 assuming Tvib = Telex for a 3.8 mA glow 1400 100um
discharge in atmospheric pressure air and 300 µm electrode spacing. 300 um
Discharge Voltage (V) 1200

4.3. Temperature characteristics and heat balance in 1000

discharge
800
Results of measurements of the discharge temperature as a
600
function of discharge current over a range from 10 µA to
40 mA are shown in figure 10. The low current conditions 400
are for an electrode spacing of approximately 10 µm; the
200
higher currents are at a spacing of 300 µm. Two different
electrode spacings were used because the smaller currents 0
could only be attained with the smaller electrode spacing. 0 5 10 15 20 25 30 35

The temperature increases rapidly at low currents then Current (mA)

appears to asymptotically approach a constant temperature for
Figure 11. Discharge voltage variation with discharge current for a
larger discharge currents. The voltage–current characteristic
glow discharge in atmospheric pressure air at 100 and 300 µm
corresponding to this temperature–current characteristic is electrode spacing.
shown in figure 11. The discharge voltage is relatively constant
over most of the range of discharge currents indicating NG 5000
operation. The sharp, approximately 600 V, increase in the 100um
300 um
discharge voltage at currents below 300 µA corresponds with 4500
Vibrational Temp (K)

operation in the sub-normal mode and was noted by high

frequency oscillations in current and voltage. Above 300 µA 4000
all the discharges had no oscillation in either voltage or current
and were in the so-called NG mode. The approximately 3500
50 V decrease in discharge voltage in going from currents of
1 to 35 mA can be shown to correspond with the observed 3000
increase in temperature. As the temperature increases there
is a corresponding decrease in density (at constant pressure = 2500

1 atm). The decrease in density corresponds with a decrease

2000
in electric field (at constant E/n and Te ) and thus a decrease 0 5 10 15 20 25 30 35
in discharge voltage (for constant electrode spacing) at higher Current (mA)
currents.
Figure 12 shows the change in measured vibrational Figure 12. Variation of vibrational temperature with discharge
temperature with discharge current for the same conditions as current for a glow discharge in atmospheric pressure air at 100 and
300 µm electrode spacing.
shown in figures 10 and 11. Going from subnormal to normal
discharge there is an increase in the vibrational temperature. In
the NG discharge regime there is a decrease in the vibrational rotational temperature with electrode spacing. The vibrational
temperature with increasing current which corresponds to temperature is seen to have a peak at around 400 µm. It
an increase in the rotational temperature. Figure 13 shows would appear from figures 12 and 13 that there is an optimal
the variation of discharge temperature with electrode spacing current and electrode spacing for achieving high vibrational
for a 3.5 mA discharge. There is a measured increase in temperatures.

825
D Staack et al

4800 is consistent with the experimental results shown in figures 10

4300 TRot and 13. The gas temperature increases with discharge current
TVib and then levels off. Also for smaller electrode spacing
3800
the electrodes provide a larger proportion of the cooling
Temperature (K)

3300 and the temperature decreases with electrode spacing. This

2800 simplified analysis neglects some important effects such as
2300 non-uniform heating and cooling along the discharge, property
variations with temperature, buoyancy induced convection and
1800
instabilities such as ionization overheating. But, it is sufficient
1300 to explain the major experimentally observed trends in the NG
800 discharge operating regime.
300 The vibrational temperature is determined by the balance
0 500 1000 1500 of vibrational excitation and vibrational de-excitation or
relaxation. By considering the excitation rate due to collision
Electrode spacing (um)
with electrons, the rate of vibrational to translational (V–T)
Figure 13. Rotational and vibrational temperature versus electrode relaxation and the rate of collisions with the electrode, the
spacing (in µm) for a 3.5 mA discharge using the transitions around observed experimental trends for Tvib can be explained. At very
375 nm. low currents there is a decrease in the vibrational temperature
with current. This is because in the sub-normal regime
The dependence of the gas temperature on discharge the electron density and rate of excitation decreases and the
current as shown by the shape of the curve in figure 10 can vibrational temperature can be expected to be lower. In the
be explained by considering the two major mechanisms of NG regime the current density is relatively constant and no
cooling the discharge during steady state. Consider the plasma significant change in the excitation rate is expected. The
as a cylindrical heat source of radius R and length d. The variations of the Tvib with discharge current while operating
power  generated can be expressed as wp = I (Vn + Ed) = in the NG regime can be understood by taking into account
π R 2 j (Vn + Ed) where I is the discharge current, Vn is the exponential increase in the rate of V–T relaxation with gas
the voltage drop in the cathode region and E is the electric temperature [26]. The vibrational temperature decreases at
field in the positive column. Since we are operating in a NG higher currents because of the increase in gas temperature and
discharge mode, the diameter of the discharge is related to the corresponding increase in the rate of V–T relaxation. Thus,
current by the normal current density j and I = πR2 j . This the decrease in Tvib correlates with an increase in Trot . For a
equation roughly takes into account that a longer discharge given electrode spacing Tvib thus experiences a maximum as a
does not create proportionally large power due to the lower function of discharge current as shown in figure 12. Typically
power density in the positive column compared with the this occurs slightly above the low current limit of the NG
cathode region. The first mode of cooling is radial cooling regime. This low current limit is determined approximately
through the gas. The effective cooling area is the surface by where the radius and length of the negative glow region are
area of the plasma 2πRd and the radial heat transfer can be equal [9]. Below that current the discharge extinguishes or has
expressed as q1 = (2πRd) h1 T , where T is the difference oscillations in the subnormal mode. There is also a decrease in
in temperature between the plasma and surroundings and h1 Tvib when the electrode spacing is decreased due to increased
is the effective heat transfer coefficient. h1 is determined by relaxation at the electrodes. There is also a decrease in Tvib
the diffusion of gas out of the discharge and will decrease with larger gap spacing due to the increase in gas temperature.
with an increase in the radius of the discharge.
 Treating the For a given current Tvib thus also experiences a maximum as a
surrounding gas as a cooling shell, h1 = k (R ln(R∞ /R)) function of electrode spacing as shown in figure 13. The gap
[25], where k is the conductivity of the gas and R∞ is an length corresponding to the maximum varies with discharge
effective outer boundary where the temperature is that of current.
the surroundings. The logarithm varies slowly and can be
approximated by a constant of order unity. If radial cooling 5. Conclusions
were the only method of cooling then T ≈ I (Vn + Ed) dk
and an increase in temperature with discharge current is Non-thermal glow discharges of the NG type were created
predicted.The second
 mode is axial cooling to the electrodes in atmospheric pressure air. Optical emission spectroscopy
and q1 = π R 2 h2 T , where h2 is the effective heat transfer provided measurements of the vibrational and rotational
coefficient to the electrodes. Assuming highly conducting temperature over a range of discharge conditions within the
electrodes, h2 is constant and is independent of the size of the NG regime. The method of determining the temperatures
discharge. If axial cooling were the only method of cooling was to fit modelled OES with spectroscopic measurements
then T ≈ j (Vn + Ed) h2 and the discharge temperature from the discharge. An analysis of the sensitivity of the
would be independent of the discharge current because j = RMSE to the fit parameters determined error bars on the
const. The ratio of the radial to axial cooling is Rd:R 2 and temperature measurements. Different transitions within the
axial cooling becomes more important for larger diameter, N2 2nd positive system measured similar rotational and
higher current discharges. Thus, at low discharge currents an vibrational temperatures. The transitions around 375 nm
increase in temperature with discharge current is expected and appeared best suited for measurements of both vibrational and
at higher discharge currents a near constant temperature. This rotational temperatures because of the higher spectrometry

826
 Spectroscopic studies and rotational and vibrational temperature measurements of atmospheric pressure normal glow plasma discharges

sensitivity and stronger emission strength of the line. The [5] Mohamed A, Block R and Schoenbach K 2002 IEEE Trans.
most intense of the transitions at 334 nm were inaccurate Plasma Sci. 30 182
[6] Staack D, Farouk B, Gutsol A and Fridman A 2005 Plasma
at determining the vibrational temperature because only one
Sources Sci. Technol. 14 700
strong vibrational transition occurred in the range. Spectra [7] Becker K, Schoenbach K and Eden J 2006 J. Phys. D: Appl.
of the UV NO transitions around 224 nm showed that for NO Phys. 39 R55
Telex and Tvib are equal. Measurements using the overlapped [8] Laux C O, Spence T G, Kruger C H and Zare R N 2003
spectra of N2+ and N2 were used to estimate an ion number Plasma Sources Sci Technol. 12 125
[9] Fridman A and Kennedy L A 2004 Plasma Physics and
density, assuming Telex = Tvib and gave results for nitrogen
Engineering (New York: Taylor and Francis)
ion density of about 5.5 × 1010 cm−3 . The gas temperature [10] Schoenbach K H, Moselhy M, Shi W and Bentley R 2003
measured in the NG regime show an increase in rotational J. Vac. Sci. Technol.A 21 1260
temperature with discharge current and electrode spacing. [11] Yu L, Pierrot L, Laux C and Kruger C H 2001 Plasma Chem.
The observed variations can be explained by a simplified Plasma Process. 21 483
[12] Laux C 2002 von Karman Institute Special Course on
heat balance analysis. Radial cooling appears dominant for
Physico-Chemical Modeling of High Enthalpy and Plasma
smaller diameter discharges and the axial cooling for larger Flows (Rhode-Saint-Genese, Belgium)
diameter discharges. The vibrational temperature decreases [13] Chelouah A, Marode E, Hartmann G and Achat S 1994
with increasing rotational temperature and also decreases at J. Phys. D: Appl. Phys. 27 940
low discharge currents and smaller electrode spacing. This [14] Chelouah A, Marode E and Hartmann G 1994 J. Phys. D:
Appl. Phys. 27 770
results in an optimal regime for creating vibrationally excited
[15] Donnelly V M and Malyshev M V 2000 Appl. Phys. Lett.
species. Overall the atmospheric pressure NG discharge in 77 2467
air was shown to be non-equilibrium. Depending on the [16] Rahman A, Yalin A P, Surla V, Stan O, Hoshimiya K, Yu Z,
application, regimes can be chosen for maximum vibrational Littlefield E and Collins G J 2004 Plasma Sources Sci.
temperature or low gas temperature and moderate vibrational Technol. 13 537
[17] Iza F and Hopwood J A 2004 IEEE Trans. Plasma Sci.
excitation.
32 498
[18] Langarias J C, Reeds J A, Wright M H and Wright P E 1998
Acknowledgment SIAM J. Optimization 9 112
[19] Slovetsky D I 1980 Mechanisms of Chemical Reactions in
The work reported is supported by the US National Science Non-Equilibrium Plasma (Moscow: Nauka (Science))
[20] Stoffels E, Flikweert A J, Stoffels W W and Kroesen G M W
Foundation under grant DMII-0423409.
2002 Plasma Sources Sci. Technol. 11 383
[21] van der Mullen J A M 1990 Phys. Rep. 191 190
References [22] Brok W J M, Bowden M D, van Dijk J, van der Muller J J A M
and Kroesen G M W 2005 J. Appl. Phys. 98 013302
[1] Kunhardt E E 2000 IEEE Trans. Plasma Sci. 28 189 [23] Leipold F, Stark R H, El-Habachi A and Schoenbach K H
[2] Schutze A, Jeong J Y, Babayan S E, Park J, Selwyn G S and 2000 J. Phys. D: Appl. Phys. 33 2268
Hicks R F 1998 IEEE Trans. Plasma Sci. 26 1685 [24] Varney R N 1953 Phys. Rev. 89 708
[3] Raizer Y P 1997 Gas Discharge Physics (Berlin: Springer) [25] Kaviany M 2002 Princples of Heat Transfer (New York:
[4] Akishev Y, Grushin M, Kochetov I, Karal’nik V, Napartovich Wiley)
A and Trushkin N 2005 Plasma Source Sci. Technol. 14 S18 [26] Lifshitz 1974 J. Chem. Phys. 61 2478

827

View publication stats