ICB2 SPRING 2007 1

Exploration of Parametric Relationships in

Electrostatic Precipitators
Victoria Hsiao and Yifan Sun

Abstract— We examine the parameters involved in designing

the particle collection plates of dry electrostatic precipitators.
Through numerical simulation and analytical calculation we iden-
tify the correlations between individual parameters and the land-
ing distance of the charged particle. In the process, we also deter-
mine whether or not the analytical calculations accurately predict
numerical trajectory results. The major parameters for the par-
ticle are charge, entry position, initial velocity and mass; for the
“collection” plates the parameters are charge density and radius.
Changing one parameter at a time, we generated numerical and
analytical projections of landing distance. Despite large discrepan-
cies in actual values, the analytical calculations were determined
to be fairly accurate in predicting the general trends reflected in
simulation results.
Index Terms— electrostatic precipitators, electric field, ionized
particles, pollution

I. I NTRODUCTION TO E LECTROSTATIC P RECIPITATORS

As concerns over the global effects of air pollution and green-
house gases increase, electrostatic precipitators are a topic of
increasing relevance to both scientists and engineers alike in
the modern world. Electrostatic precipitators, with up to 99%
Fig. 1. (Left)Labeled Diagram of Electrostatic Precipitator. Those used in
efficiency, are particularly effective in removing silicon dioxide industrial smokestacks would be much larger but the general components are
(SiO2 ), sulfur dioxide (SO2 ) and nitrogen oxide (N Ox ), all the same. Large particles in the smoke are first filtered out by the aluminum
mesh filter; the remaining particles are ionized by the ionizing section. These
significant causes of acid rain, from fossil fuel emissions. For ionized particles are then passed through the collection cell, where a majority
the general public, air ionizers use electrostatic precipitators to are attracted and collected by the charged collection plates. (Right) The indus-
remove particles from the surrounding air, though the overall trial electrostatic precipitator in its natural habitat.
effectiveness of these devices has been questionable, because
of limited range and possible side effects such as the production
of ozone. Understanding the basics of electrostatic precipitators
is valuable for the interested student because it brings together
physics, chemistry and environmental studies into an easily un-
derstood modern application. solve analytically with accuracy, it is fairly reasonable to model
Electrostatic precipitators (Fig. 1) are particle filters which numerically. Thus we will solely model the particle collection
take advantage of an induced electric field to remove ionized plate system.
pollutant particles out of rising smoke or flowing air. These sys- In this work, we focus solely on the physics of a charged par-
tems consist of two main components: the ionizing section and ticle traveling through two oppositely charged electric plates
the collection plates. The ionizing section, as its name implies, generating a voltage gradient between parallel metal plates.
ionizes passing particles by applying several thousand volts into Through simulation and analysis, we seek to answer two ques-
sharpened spikes. A high enough voltage will cause a cer- tions: What sort of relationships did individual parameters have
tain type of electrical discharge, called a corona, which causes with the overall effectiveness of the collection plates, and how
the air around the wires to break down and become ionized. well can analytically derived equations predict these trends or
The charged particles now move into the collection cell, where relationships? To that end, we have defined ‘effectiveness’ as
they are attracted to charged metal collection plates and thus the shortest distance traveled by a charged particle before be-
removed from the air entering the atmosphere. The phenom- ing trapped by the collection plate. We realize such a definition
ena provided by the shorting coronas is governed by high-level lends itself to predicting the most energy consuming parame-
plasma physics, which, although fascinating, are very difficult ters, but have also kept the voltage put into the collection plates
to model. Conversely, the electric field surrounding the col- constant for all of our tests, with the exception of the test with
lection plates turns to be the opposite; though very difficult to variable voltage.
ICB2 SPRING 2007 2

with the edge of the two plates (R~o =< −Rplate , 0, 0 >) and
shot directly inwards with an initial velocity between 1,000 m/s
and 1,000,000 m/s, depending on restrictions of other param-
eters (V~o =< 104 , 0, 0 > m/s or < 106 , 0, 0 > m/s). These
seven parameters are needed to strictly characterize the motion
of any particle in this system.
In reality, the collection cell of an electrostatic precipitator
consists of multiple plates to catch ionized particles. For the
purposes of our simulation, however, we are testing parametric
effects on one particle at a time, and thus only need to model
a single pair of plates. We also assume the particle has already
been ionized by the corona immediately preceding its entry into
Fig. 2. Diagram of Defined Variables in Collection Plate System. We the collection cell with a charge of +1 coulomb, that the ef-
assume the pollutant particle of mass,m, has already been ionized by the corona
and enters the collection plate area with a charge, q, initial velocity, v~o , and fects of gravity are negligible in comparison to the electrostatic
initial point of entry, r~o . The metal collection plates have charge densities of forces on particles of this size; the proportionality constant for
2
+σ and −σ, the strength of which are determined by the amount of voltage, V ,
put into the plates. The distance between the two plates is d and the dependent
gravity is G = 6.67 · 10−12 Nkgm2 whereas the proportionality
2
Nm
variable being measured is the landing distance from entry point, x. In context, constant for electrostatics is k = 9.99·109 coulomb 2 . In addition,
this entire system would be turned 90◦ since the smoke particles would be
rising; however, since gravity has very little effect on particles of this size, we assume that the particles do not interact with each other in
having the plates horizontally oriented makes the system easier to visualize. the form of collisions, attractions etc. To be fair, molecules are
not infinitesimally small, and realistically undergo collisions
that cause their motions to become more random; however, cal-
II. D EFINING THE C OLLECTION P LATE S YSTEM culating these situations extends beyond the scope of our study
in electrostatics, and for this reason it will not be covered.
In our system, we eliminate randomness by strictly defining
all parameters that characterize the system and the trajectory
of any particles. It is also for this reason that we did not try III. C ALCULATIONAL A PPROACHES
testing the collecting plate for a multitude of charged particles The electric field over a plate can be characterized as the
under similar conditions; each trial is so strictly defined that all summation, or integral, of its composite field-causing points.
particles can only have one trajectory. Such parameters are de- When looking at the system analytically, we assume the parti-
lineated in Figure 2. cles are so small in comparison to the plates that we can cal-
As Figure 2 illustrates, we have distinguished seven param- culate the electric field for an infinite sheet of charge. This
eters to test: particle charge q, particle mass m, initial velocity assumption is useful because the integral describing the field
v~o , initial position r~o , voltage put into the collection plates V , caused by finite radius plates is currently unsolvable analyti-
distance between the collecting plates d, and radius of the col- cally. Solving the electric field integral over an infinite sheet of
lection plates Rplate . All of these independent variables are charge begets this equation:
gauged with the dependent variable landing distance, x.
In order to test each parameter independently for correlation ~ sheet = σ
E (1)
with landing distance, we must choose default values for the 2o
other parameters which will be held constant. First, though the For two sheets of charge:
problem can be analytically solved for collection plates of in-
finite radius, this is physically impossible; thus, a finite value ~ sheet = σ
E (2)
must be assigned for plate radius, Rplate . The default value for o
Rplate was determined to be 100 meters; although this value is
higher than realistic values, it more closely mirrors the analyt- Using the Lorentz Force Law, F~ = q E
~ + I~ × B
~ = m~a, we
ical assumption of infinitely sized plates. With the addition of solve for acceleration.
plate boundaries, the distance between the plates also becomes q σ
significant, and is assigned the variable d, given a default value ~a = k̂ (3)
m o
of 1 meter. The charge density, σ, is also significant in the effec-
tiveness of the plates. In industry, this value is usually encap- We used this acceleration in the kinematics equation, x(t) =
1 2
sulated in the voltage measurement between the two plates, as 2 at since force in only in the ĵ direction, solved for time t
V = Cq = σ·A σ·d
A =  . Our default voltage value was 10,000 and used this in x(t) = vo t. The resulting equation, where
o· d o
volts. σ = Vdo :
In addition, the particle itself also holds properties. In our s r
simulation, we modeled the particle as a silicon oxide (SiO2 ) dmo m
60.1 x = vox · = vox · d (4)
molecule. Its mass m therefore was 6.022·10 23 grams = 9.98 · qσ qV
−26
10 kilograms. It carries an ionized charge, q, of 1 coulomb.
The particle is placed initially at a height exactly between the where x is the horizontal distance traveled by the charged par-
plates, spanned radially from the origin such that it is aligned ticle before it hits the collection plate. Eq. 4 represents our
ICB2 SPRING 2007 3

estimation of an analytical solution to our problem; we later of which illustrate the advantages and disadvantages of our as-
compare numerical solutions to this equation to test its validity. sumptions. We will also briefly explain the remaining results,
This equation predicts some parameter trends for us. It shows of which are less significant.
that radial distance traveled is directly proportional to initial ve-
locity, directly proportional to the square roots of plate spacing
and particle mass, and indirectly proportional to the reciprocal General Trends
of the square roots of particle charge and plate density. How-
The parameters d,v0 ,and m all follow the trends predicted
ever, collection plates in reality do not have infinite radius, and
by the analytical equation: landing distance varies linearly with
are bounded by defined edges. These edges cause nonuniform
d, by the square root of m, and linearly with v0 . Conversely,
electric fields that point not only axially, but also radially, dis-
landing distance varies linearly with Rplate numerically but has
torting the validity of our analytical assumption. Figure 3 illus-
no impact analytically, and varies inversely with V numerically
trates this problem, showing the difference between field lines
but as the inverse of the square root of V analytically. Values
found near the center of the plate that act as if the plate is in-
for the numerical solution varies between 1000 times and 10
finitely large, and field lines near the boundary of the plate that
trillion times as large as analytical solution.
curve outward.

Fig. 3. Electric Field Lines of Collection Plates. The analytical calculation

of the field, calculated for an infinite sheet of charge does not take into account
boundary effects of the field, drawn out here. The numerical simulation of the
field does, and so this accounts for some of the discrepancy in results.

To solve for a situation with bounded plates (Fig. 3), a nu-

merical approach was taken. Numerical integral of the electric
~ = k ~r−r~0 dq , was done using MatLab’s trapezoidal
R
field, E r −r~0 |3
|~ Fig. 4. Effects of Charge on ESP Efficiency Varying particle charge has is
summation method, and derived at each point using the values related by an factor of an inverse square root, as shown by the log-log slope of
0.5. In this case, the analytical and numerical models agree.
provided as parameters. To find the effectiveness of parameter
values on total distance traveled, the simulation iterated through
possible values at a given range surrounding the variables, and
the distance values derived were plotted against the parameter
values given. Varying the Charge of the Particle: A Valid Analytical Repre-
sentation
IV. R ESULTS The analytical equation predicts particle charge as having an
inverse square root relationship with the landing distance, x.
In order to find the parameter/landing distance relationships,
The fitted line on the log-log graph does indeed have a slope of
as well as determine the validity of analytically derived predic-
approximately -0.5, indicating an exponent of − 12 . Alterations
tions, our test trials systematically worked through all seven
of charge seem to cause the same sorts of variance in landing
parameters affecting the charged particle collection system:
distance numerically and analytically.
q, m, Rplate , r~o , v~o , σandd. We held the other six parameters
at the default values described earlier and ran the variable pa-
rameter for a range of about 100 values. For each value, the
Varying Particle Starting Position: An Invalid Analytical Rep-
electric field was calculated, the charged particle trajectory was
resentation
calculated, and the final landing coordinates of the particle were
recorded. After the simulation, each parameter value was plot- On an infinite sheet of charge, the starting position of the par-
ted against its dependent landing distance. The predicted land- ticle (r~0 ) is inconsequential, and is not a factor in determining
ing distances were also analytically calculated using Eq. 4 for landing distance; the log-log slope of the analytical data is zero
the same range of parameter values, and both graphs were plot- as expected. For the numerical data, however, (?) plot indicates
ted side by side for trend comparison. A wide variety of re- that the farther away the particle starts from the charged collec-
sults were found for the different parameters, some more sig- tion plates, the farther it will have to travel before landing, in
nificant than others. In this section, we will further explore which the log-log relationship of the two factors relate with a
the effects of particle charge and starting position alterations slope of 0.8.
ICB2 SPRING 2007 4

analytical assumption drastically overestimates the amount of

downward field the particle experiences. Also, the symmetry
associated with plates of infinite radius are not present for one
of finite radius; this could also account for the drastic decrease
in distance traveled in analytical cases.
However, this magnification seems to have been accom-
plished in a somewhat uniform fashion, as each parameters
trend of increase still appears similar when derived analytically
and numerically. The exceptions to this assertion are r~o and
Rplate . Analytically, neither term appears to affect the distance
traveled, while numerically, they seem to both cause a linear in-
crease in distance traveled (though less precisely for r~o ). This
conclusion is not surprising: analytically, the radius of the plate
is infinity, and there is no logical starting position for the par-
ticle. Numerically, larger plates and different starting positions
form different geometries that affect the final outcome of the
Fig. 5. Effects of Initial Position on ESP Efficiency Varying entry position trajectory. For the other parameters, the trends seem fairly con-
of the charged particle. In this case, the analytical and numerical models do not
agree. sistent with analytical predictions.

VI. C ONCLUSION AND F URTHER S TUDY

V. D ISCUSSION In this study, we attempted to find the correlations between
individual parameters and the landing distance of a charged
For the property of charge is unrelated to the property of
particle moving between two oppositely charged metal plates;
changing radius, it seems evident that changing the two factors
in addition, we also sought to determine the validity of us-
would cause similar trends in landing distances. Furthermore,
ing an analytically derived equation to predict the behavior of
the decay of distance with increasing charge suggests that par-
the numerical model. Results of numerical simulation juxa-
ticles with higher charge are easier to capture in the collector.
posed with analytical predictions show that although the analyt-
This information suggests that corona effectiveness in ionizing
ically calculated solution was fairly accurate in predicting gen-
particles is an advantageous factor in precipitator design.
eral trends exhibited by numerical simulation, the actual values
The property of starting position, however, is significantly
varied drastically. Such discrepancies may be the function of
tied to plate radius, as in an analytically infinite plate, the par-
poorly chosen default values, which may or may not have re-
ticle cannot start anywhere but the middle of the plate. (On
flected actual conditions. Particle charge and starting particle
every side, it feels the effect of infinitely large sides, and thus
position were discussed in greater detail as example cases of
constantly feels radially symmetric fields.) By assuming finite
when the analytical model was valid and when it was not.
radius, however, our predictions become more realistic, as the
For the enthusiastic and time-rich student, we suggest taking
particles to be captured must start from somewhere outside the
our results to the next level: how could all of these parameters
plate. Our numerical trends show an increasing distance trav-
be combined into one massive optimization problem? Addi-
eled as the particle is carried further from the center of the
tional factors which should be considered in an optimization
plates. This trend indicates that the further away the starting
problem, which were not considered here and partially explain
position, the longer the distance must be traveled before the
our inconclusive results, are: energy restrictions, cost limita-
particle feels significant force from the plates causing it to land.
tions, size limitations, particle interactions and experimentation
This result suggests that the efficiency of a particle collector is
with actual particle data. For the purposes of simply determin-
increased for particles that wander closer to its boundary; how-
ing relationships and trends we used general default values, but
ever, as almost all of our particles landed in the plates eventu-
taking the problem a step further would involve optimization
ally, the effectiveness is not affected.
for these conditions. In cases where only certain pollutant par-
The superposition of our numerical results as a whole to that
ticles need to be filtered out, how could such a discriminating
predicted by analysis shows some similar trends, but do not
filtration system be designed?
match exactly because of value differences. The most proba-
ble explanation for this discrepancy is that the assumption that
plates of finite radius can be treated as sheets of infinite radius
is inadequate. Because our tests are run such that the particle
begins at the edge of the plate, it feels strongly the boundary
effects of electric fields. Figure 3 shows a qualitative represen-
tation of such boundary effects, of which act non-uniformly in
a significant way of altering data.
These boundary effects are not the only consequence of our
assumption. Unfortunately, a plate of infinite radius is quite a
bit larger than a plate of finite radius, and for this reason, our