Sei sulla pagina 1di 8

Journal of the Korean Physical Society, Vol. 42, February 2003, pp.

S859S866

The Effect of Inertial Terms in the Momentum Equation in Fluid Simulation


of High Density Plasma Discharge
Hee-Hwan Choe
LCD R & D Team, Device Solution Network, Samsung Electronics Co., LTD., Suwon 449-711

N. S. Yoon
School of Electrical and Electronic Engineering, Chungbuk National University, Chungju 361-763
The validity of the frequently used drift-diffusion representation of the momentum equation is
investigated in fluid simulations of high-density plasma discharges. The drift-diffusion expressions
are good approximations for both the electrons and ions at high pressure. However, for low pressure,
the inertial term cannot be neglected in the ion momentum equation, although the drift-diffusion
representation
 of electrons is still valid. The available criterion for the valid pressure regime is
vB / i/e,N L  1; where vB is the Bohm velocity, i/e,N is the ion/electron-neutral collision
freuquency and L is the dimension of the system. In addition, a new method to treat the inertial
terms by introducing an effective ambipolar electric field, Eeff , is proposed. The results are compared
to the full ion momentum equation case.
PACS numbers: 52.50.Qt
Keywords:

ni
+ i = ion ne ,
t
ne
+ e = ion ne ,
t
E = 4e (ni ne ) ,
3 (ne Te )
= Q ee E + Pext Pcoll ,
2 t

I. INTRODUCTION

Plasma processes play important roles in the production of high-density, high-performance microelectronic
products. Modeling and simulations of plasma processes
are studied to understand and optimize the processes as
well as experiments [1,2]. As a part of modeling and simulations of the processes, the fluid simulation [36], the
kinetic simulation, and the hybrid model [79] are used
to describe the dynamics of plasma discharge. Among
those methods, the fluid simulation has been used widely
because it helps us to understand the physical structure of the discharge, and is less time consuming in numerical computation compared to the kinetic simulation,
and gives accurate results if the distribution function of
species is not highly deviated from Maxwellian [10].
In such fluid simulations, the following set of equations
is usually solved: the momentum, continuity, Poisson
and temperature equations.
eni
1
i
+ (vi i ) =
E iN i
(ni Ti ),
t
M
M
e
ene
1
+ (ve e ) =
E eN e (ne Te ),
t
m
m
E-mail:
E-mail:

(1)
(2)

(3)
(4)
(5)
(6)

where i is ion flux, e is electron flux, ni is ion density, ne is electron density, ion is the ionization collision
frequency, M is ion mass, m is electron mass, Ti is ion
temperature, Te is electron temperature, iN is ion neutral collision frequency, eN is electron neutral collision
frequency, E is ambipolar electric field, Pcoll is collisional
power loss per volume, Q is electron energy flux, and Pext
is externally applied power per volume.
In a numerical treatment of the above equations, the
inertial term of (v) in the momentum equations
requires much more computation time and a specific stabilization technique, such as upwind scheme [11]. Therefore, the replacement of the momentum equations by the
drift-diffusion equations is used frequently. The driftdiffusion forms are achieved by neglecting the left hand
side of Eqs. (1) or (2), and by assuming that the electron
or ion motion is collisionally dominated and in a steady
state. That is, this method replaces Eq. (1) with

h.choe@samsung.com
nsyoon@vod.chungbuk.ac.kr

i =
-S859-

eni
1
E
(ni Ti )
M iN
M iN

(7)

-S860-

Journal of the Korean Physical Society, Vol. 42, February 2003

or Eq. (2) with


e =

1
ene
E
(ne Te ).
meN
meN

(8)

For example, the substitutions of Eq. (7) for Eq. (1)


and Eq. (8) for Eq. (2) are used for high pressure capacitive RF (Radio Frequency) plasma discharges in Refs.
[12, 13], and for low pressure ICP (Inductively Coupled
Plasma) discharges in Ref. [14]. The replacement of
Eq. (2) by Eq. (8) is used for low pressure capacitive
RF plasma discharges in Ref. [3], and for low pressure
ICP discharges in Ref. [10]. The advantage of using this
method is that it saves a great deal of computation time
as well as having tractable form. However, the validity
of the drift-diffusion representation is not clearly verified
for a wide range of pressure. Therefore, in this work,
the effect of the inertial terms and the valid condition
for the application of the drift-diffusion equations are investigated and detailed discussions are given in Sec. II.
. Also, the drift-diffusion representation is sometimes
useful, even in a low-pressure condition in which this approximation is not good. In Sec. III. , we will introduce
an effective ambipolar electric field, Eeff , and test the
accuracy using a one-dimensional simulation.

II. VALIDITY CRITERION FOR THE


APPLICATION OF THE
DRIFT-DIFFUSION EXPRESSION

For the steady state, it becomes




2
vi
1+
vi i = dd

vi ni ,
i
iN
iN

(9)

Crude estimations of the ion inertial term to the drift


and diffusion terms are
2
M vB
M ni vi vi

O(1),
eni E
eVps

(10)

2
M ni vi vi
M vB

 1.
Ti ni
Ti

(11)

where Vps is the pre-sheath voltage drop and vB is the


Bohm velocity. These results show that the ion inertial
term can be comparable to the drift term and is larger
than the diffusion term, even at sufficiently high pressure. For example, a rough estimation for 50 mTorr Ar
discharge gives vB 2.1 105 cm/sec, Te 1.8 eV,
eVps 1.0 eV, and Ti 0.05 eV which results in
2
2
M vB
/(eVps ) 2 O(1) and M vB
/Ti 40  1. To
see the effect of the inertial term in the ion momentum

(12)
(13)

(14)

where dd
= (eni E Ti ni ) / (M iN ) is the drifti
diffusion representation of the ion flux.
In the
above equation, the two terms of 2 ( vi ) i /iN and
vi (vi ni ) /iN appear newly because of the inclusion
of the inertial term. As we can see from Eq. (14)
that the difference in the magnitude of i and the magnitude of dd
is caused by the following two terms:
i
2 ( vi ) i /iN and vi (vi ni ) /iN . This means that
the inertial term affects the flux considerably and thus
alters results greatly, especially in a low-pressure discharge condition. Therefore, the substitution of the driftdiffusion representation of the ion flux for the ion momentum equation should be carefully considered according to discharge conditions, with pressure being the most
important.
Similarly, the above discussions for the ion momentum
equation can be repeated for electrons also. Steady state
electron momentum equation is
eN mne ve = ene E (ne Te ) mne ve ve .

To start, we roughly compare the order of each term in


the momentum equations to check whether the inertial
term can be neglected. The ion momentum equation for
steady state is given by
iN M ni vi = eni E Ti ni M ni vi vi .

equation, we rearrange Eq. (1) as




2
1 i
+ 1+
vi i
iN t
iN
Ti
vi
eni
=
E
ni
vi ni .
M iN
M iN
iN

(15)

Rough estimations yield


2
mne ve ve
mvB

 1,
ene E
eVps

(16)

2
mne ve ve
mvB

 1.
(ne Te )
Te

(17)

We can easily check that the inertial term in the electron momentum equation can always be neglected in the
2
usual discharge conditions, because eVps Te  mvB
=
Te (m/M ); Equation (2) for steady state can be rewritten
as


2
ve
1+
ve e = dd

ve ne , (18)
e
eN
eN
where dd
= (ene E (ne Te )) / (meN ) is the drifte
diffusion representation of the electron flux. The two
terms of 2 ( ve ) e /eN and ve (ve ne ) /eN can be
neglected in nearly all cases, because vi ve , ni ne ,
and iN  eN .
A validity criterion of drift-diffusion representation
will be derived as follows. If we give attention to the
facts that the escaping velocity of particles decreases and
the ion/electron-neutral collision frequency increases as

The Effect of Inertial Terms in the Momentum Equation Hee-Hwan Choe and N. S. Yoon

-S861-

pressure increases, there must be the possible conditions


for using i,e dd
i,e for high pressure
2
vi,e  1 ,
iN,eN
vi,e
vi,e ni,e  dd
i,e .
iN,eN

(19)
(20)

For electrons, we can see clearly that


vB
2
ve
,
eN
eN L
ve ve ne
vB

,
dd
eN e
eN L

(21)
(22)

where L is the typical dimension of system. Similarly for


ions, we have
2

vB
,
iN
iN L
vi vi ni
vB

.
iN dd

iN L
i
vi

(23)
(24)

Therefore, the final form of the validity criterion for the


validity of the drift-diffusion equations becomes
vB
 1.
i/e,N L

(25)

For example, in an Ar discharge with L = 10 cm,


the drift-diffusion equations can be used, if the following
conditions are satisfied:
3

vB
10

 1,
eN L
P rs

vB
Te

 1,
iN L
P rs

(26)
(27)

where P rs is the gas pressure in mTorr. These rough


estimations allow us to use the drift-diffusion representation for electron flux for pressures higher than a few
mTorr, but for ion flux the pressure should be larger than
100s mTorr at a minimum, if we take 102  1. Furthermore, even though the pressure is sufficiently high so
that the effect of the inertial terms seems to be almost
negligible in the body of discharge, we have to consider
carefully the fact that the effect of inertial terms near
the boundaries become greater, because ion velocity and
the its gradient are large around the boundaries.
Some quantitative prediction of the effect of the ion
inertial term can be given as follows; A rearrangement
of Eq. (9) gives


e
Ti
vi
vi =
E
ni
vi . (28)
M iN
M ni iN
iN
The bracket in the right hand side is ion velocity due to
the drift and diffusion, and the last term is the inertial
term. Since |vi | is a monotonically increasing function

Fig. 1. Magnitude
which stem from the
of two terms

inertial term; (a) 2 v/iN for 10 mTorr(real line),
100 mTorr(dashed
line), and

 1000 mTorr(dash-dotted line),
for 10 mTorr(real line), 100
(b) vv ni / iN dd
i
mTorr(dashed line), and 1000 mTorr(dash-dotted line).

as the position goes outward, the term of vi vi /iN


always decreases |vi |. This reduction of the ion velocity
will decrease the particle losses and thus the inclusion of
the inertial term increases the ion density. The increased
ion density will decrease the electron temperature and
the plasma potential.
Figures 1(a) and 1(b) show the magnitude of two
terms in Eq. (14) due to the ion inetial term for various pressures. These results say that the ion inertial
terms for low-pressure discharge (< 100 mTorr) cannot be neglected. As previously mentioned, 2 v/iN
and vv ni /(dd
iN ) stem from the ion inertial term
i
and they diminish as pressure increases. The concavely
shaped graphs show that the effect of the ion inertial
term is very small in the body of discharge compared to
that in the boundary regions where the magnitude of v
and ni are relatively large.
For a more exact inspection, the usual one-dimensional
fluid simulation of Ar ICP discharge is used in this study.
In the simulation, we applied a numerical stabilization

-S862-

Journal of the Korean Physical Society, Vol. 42, February 2003

Fig. 2. Profiles of (a) ion density (ni ), (b) plasma potential (), (c) electron temperature (Te ), (d) ion flux (i ), and (e)
ion velocity for 5 mTorr and 100 mTorr Ar discharge at 500 Watt; the upwind scheme (real lines) and method of using Eeff
(dashed lines) are used to calculate the inertial term, and the dashed-dotted lines are results obtained from the drift-duffusion
representation of the momentum equation.

method of dielectric relaxation scheme [15] to overcome


the time step limitation, and the exponential scheme described in Ref. [13] in the finite difference expression of
drift-diffusion components of fluxes. The first method
enables us to use a larger time step compared with the
physical relaxation time and a larger grid size can be
used by adopting the second method.
In Figs. 2(a)(e), profiles for ni , Te , , i , and vi are
plotted for low (5 mTorr) and high (100 mTorr) pressures. The inclusion of the ion inertial term increases
the value of ni approximately about three times for 5
mTorr. For 100 mTorr, there is little change due to the
ion inertial term as expected. Also, the ion inertial term

decreases the values of Te and . Plasma potential is


lowered by approximately 9 Volts for 5 mTorr and by 2
Volts for 100 mTorr, and electron temperature is lowered
by 1 eV for 5 mTorr and by 0.3 eV for 100 mTorr. For
ion flux, the change of values due to the ion inertial term
increses near the boundaries where the inertial terms are
great because the values of ion velocity and the gradient
of ion density are large there. The profiles of ion velocity show the importance of the ion inertial term. For 100
mTorr, ion velocity obtained by the inclusion of the ion
inertial term is almost idential to that calculated by the
drift-diffusion representation. For 5 mTorr, ion velocity
obtained from the drift-diffusion equation is exaggerated.

The Effect of Inertial Terms in the Momentum Equation Hee-Hwan Choe and N. S. Yoon

-S863-

Fig. 3. Relative differences of peak values between the upwind scheme and the drift-diffusion representation for 500W
Ar discharge at various pressures; ion density (square), electron temperature (circle), poential (upward triangle), and ion
flux at z = L (downward triangle).

Ion velocity must be the same order of magnitude as the


Bohm velocity, or somewhat larger, and much smaller
than the electron thermal velocity. However, the velocities at z = L obtained using the drift-diffusion equation
are vi = 9.47 107 cm/s, vB = 2.70 105 cm/s, and
ve,th = 1.04 108 cm/s, which shows that ion velocity
is much larger than the Bohm velocity and comparable to the electron thermal velocity. This means that
the drift-diffusion equation cannot fairly represent the
particle transport phenomenon. This unphysical situation was also pointed out in Ref. [14] and was corrected
by modifying the diffusion coefficient and mobility. The
ion inertial term corrects this situation without artificial
modification of coefficients. The velocities at z = L obtained using the inertial terms are vi = 1.28 106 cm/s,
vB = 2.58 105 cm/s, and ve,th = 9.89 107 cm/s;
vi  ve,th and vi 5vB . The decreased ion velocity
reduces the particle loss, and results in the increased
ion density, which gives the lowered electron temperature and plasma potential compared to results obtained

Fig. 4. Peak values of ion density (square) and electron


temperature (circle) for 500W Ar discharge at low pressures.

Fig. 5. Profiles of E (dashed lines) and Eeff (real lines) for


(a) 5 mTorr and (b) 100 mTorr.

using the drif-diffusion equations.


Relative differences of peak plasma density, electron
temperature, plasma potential, and ion flux at the
boundary (z = L) versus pressure are given in Fig. 3.
The relative difference is defined by f in f dd /f in
100 (%), where f in is the value calculated with the inertial term and f dd is that calculated by the driftdiffusion representation. We can see that, as expected,
the differences decrease as pressure increases. It is apparent that Eq. (25) is a good criterion for classifying
the low and high pressures. If we know the pressure
and gas species, we can determine whether the inertial
terms should be included using the approximate Bohm
velocity and the ion-neutral collision frequency. For example, Ar discharge for 100 mTorr with L = 10 cm,
vB / (iN L) = 8.21103 , and vB / (eN L) = 3.60105 :
In this case, if we take 102  1, then the inertial
terms can be neglected for both the ion and electron.
And, for 10 mTorr, vB / (iN L) = 1.01 101 , and
vB / (eN L) = 3.60 104 : In this case, although we
take 102  1, the ion inertial term should be included.
These estimations are in good agreement with the results
given in Fig. 3.
Another important result is that available range of

-S864-

Journal of the Korean Physical Society, Vol. 42, February 2003

Fig. 6. Profiles of (a) ion density (ni ), (b) plasma potential (), (c) electron temperature (Te ), (d) ion flux (i ), and (e) ion
velocity for 5 mTorr and 100 mTorr Ar discharge at 500 Watt; the upwind scheme (dashed lines) and method of using Eeff (real
lines) are used to calculate the inertial term.

pressure is extented toward the low value by the inclusion of the inertial term. This result is given in Fig. 4.
The drift-diffusion equation cannot give steady state solutions for pressures lower than 2 mTorr, whereas with
the inertial term it is possible. The steady state plasma
density is determined by the power balance relation [16],
and the slope of the power loss is proportional to vi [17]
if the power loss and absorption are expressed as a function of plasma density. If the slope of the Ploss (np ) is so
large that Ploss (np ) > Pabs (np ), then the steady state solution cannot be obtained: Ploss and Pabs are the power
loss and absorption as a function of plasma density, np .
As pointed out previously, the inertial term decreases

the ion velocity, which means the slope of Ploss (np ) is


also reduced. Therefore the steady state solutions can
be obtained.

III. DRIFT-DIFFUSION REPRESENTATION


AT LOW PRESSURE
Sometimes, the form of the drift-diffusion representation is still useful even in a low-pressure situation. In
that case, it may be possible to introduce the concept of
an effective electric field. A modified form of the driftdiffusion equation introducing Eeff and considering the

The Effect of Inertial Terms in the Momentum Equation Hee-Hwan Choe and N. S. Yoon

-S865-

time derivative term in the momentum equation is proposed in Ref. [18]. We extend this method by adding
the vi vi term:
e
Ti
Eeff
ni ,
M iN
M ni iN
vi
e
Ti
+ vi vi =
E
ni iN vi .
t
M
M ni

vi =

(29)
(30)

Subsequently, we obtain
eni
Ti
Eeff
ni ,
(31)
M iN
M iN

1 Eeff
+ Eeff E =
iN t

 eff


e Eeff
T
E
T

ni

ni
M iN
eiN ni
iN
eiN ni
(32)
i =

The right hand side of Eq. (32) occurs due to the


ion inertial term, thus Eeff includes the effect of the ion
inertial term. Solving these two equations, we can easily
calculate the ion flux including the effect of the inertial
term.
Profiles of E and Eeff for 5 mTorr and 100 mTorr are
given in Fig. 5. For low pressure, Eeff and E give different values because of the inertial term, and they are
almost the same for 100 mTorr because the inertial term
becomes small. This shows that the effects of the inertial
term are included in Eeff , which reflects the inertial term
effect well.
Figure 6 shows profiles for ni , Te , , i , and vi for low
(5 mTorr) and high (100 mTorr) pressure. The profiles
were identical within 1% except for the ion density at 5
mTorr; the maximum difference of ni is 7.5%.
In Fig. 7, relative differences of peak plasma density,
electron temperature, plasma potential, and ion flux at
the boundary (z = L) versus pressure are given for methods using upwind scheme and Eeff . Results using the
effective electric field method are close to those obtained
by the ion momentum equation with the inertial term
compared with those developed using the drift-diffusion
equations. Results using Eeff are identical to those using
upwind scheme within 5% for pressures higher than 10
mTorr. For pressures lower than 5 mTorr, the results
differ within 15% except for the ion density at 2 mTorr
(about 30 %). The results indicate that the new method
describes the inertial term well.

IV. CONCLUSION
We investigate the valid condition of the driftdiffusion equations,
and a useful criterion is obtained;

vB / i/e,N L  1. According to the criterion, the driftdiffusion representation of electron flux can be used for
usual Ar ICP discharge conditions, but the ion inertial

Fig. 7. Relative differences of peak values of ion density


(square), electron temperature (circle), potential (upward triangle), and ion flux at z = L (downward triangle) for 500W
Ar discharge at various pressures; filled symbols are differences between the upwind scheme and the method using Eeff ,
hollow symbols are differences between the upwind scheme
and drift-difusion representation.

term has to be included in the momentum equation for


low-pressure. It turns out that the criterion is effective by comparing the results obtained from the driftdiffusion equations with those from the full momentum
equations.
Another result is that the drift-diffusion equation cannot give steady state solution at low-pressure, because
the exaggerated escaping velocity of particles yields a
large power loss, whereas, the inclusion of the inertial
term lowers the power loss by correcting the ion velocity.
In the case that the drift-diffusion form is useful even
at a low pressure, the effect of the inertial term may be
included in the momentum equation by introducing an
effective ambipolar electric field, Eeff . Results obtained
using this method are in good agreement with those calculated by a full treatment of the momentum equation.

REFERENCES
[1] J. Kim and K. Chung, J. Korean Phys. Soc. 40, 381
(2002)
[2] Hyeon-Soo Kim, Martin. D. Dawson and Geun-Young
Yeom, J. Korean Phys. Soc. 40, 567 (2002)
[3] T. E. Nitschke and D. B. Graves, J. Appl. Phys. 76, 5646
(1994)
[4] N. S. Yoon, N. -H. Choi, B. -H Park, and D. -I. Choi,
IEEE Trans. Plasma Sci. 23, 609 (1995)
[5] S. S. Kim, N. S. Yoon, and C. S. Chang, J. Korean Phys.
Soc. 29, 678 (1996)
[6] S. -K. Park and D. J. Economou, J. Appl. Phys. 68(8),
3904 (1990)
[7] T. J. Sommerer and M. J. Kushner, J. Appl. Phys. 71,
1654 (1992)
[8] J.-P. Boeuf and L. C. Pitchford, IEEE Trans. Plasma
Sci. 19, 286 (1991)
[9] N. H. Choi, W. H. Koh, N. S. Yoon, H. B. Park, and
Duk-In Choi, IEEE Trans. Plasma Sci. 23, 617 (1995)

-S866[10] J. D. Bukowski and D. B. Graves, J. Appl. Phys. 80,


2614 (1996)
[11] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P.
Flannery, Numerical Recipes in FORTRAN (Cambridge,
New York, 1992)
[12] D. B. Graves, J. Appl. Phys. 62, 88 (1987)
[13] J.-P. Boeuf, Phys. Rev. A, 36, 2782 (1987)
[14] P. L. G. Ventzek, R. J. Hoekstra, and M. J. Kushner, J.
Vac. Sci. Technol. B, 12, 461 (1994)
[15] H. -H. Choe, N. S. Yoon, S. S. Kim, and Duk-In Choi,

Journal of the Korean Physical Society, Vol. 42, February 2003


J. Comp. Phys. 170, 550 (2001)
[16] M. A. Lieberman and A. J. Litchenberg, Principles of
Plasma Discharges and Material Processing (Wiley, New
York, 1994)
[17] N. S. Yoon, B. C. Kim, J. G. Yang, and S. M. Hwang,
IEEE Trans. Plasma Sci. 26, 190 (1998)
[18] E. Gogolides and H. H. Sawin, J. Appl. Phys. 72, 3971
(1992)

Potrebbero piacerti anche