Production of space charge at the boundaries of layer clouds

Limin Zhou
1,2,3,4
and Brian A. Tinsley
2
Received 5 September 2006; revised 6 February 2007; accepted 8 February 2007; published 6 June 2007.
[1] The ionosphere-Earth current density J
z
creates space charge at the upper and
lower boundaries of layer clouds. This occurs because clouds have an order of magnitude
lower conductivity than the clear air at the same altitude, and as the current density
flows through the boundaries, it creates a gradient of electric field that must be satisfied by
the accumulation of space charge, according to Gausss law. We have modeled the
production of space charge and its partition between charges on droplets, aerosol particles,
and ions and performed sensitivity tests for the variation of a number of relevant
atmospheric parameters. We find typical droplet charges of 50100 elementary charges,
positive at cloud top and negative at cloud base, consistent with recent observations.
The charges are of sufficient magnitude to suggest measurable electrical effects on
scavenging of ice-forming nuclei and cloud condensation nuclei. The results are relevant
to the modeling of solar or internally forced changes of J
z
and space charge on cloud
microphysics as a possible cause of small effects on weather and climate.
Citation: Zhou, L., and B. A. Tinsley (2007), Production of space charge at the boundaries of layer clouds, J. Geophys. Res., 112,
D11203, doi:10.1029/2006JD007998.
1. Introduction
[2] The ionosphere is charged to a potential of about
250 kV by the upward flow of about 1000 A of current from
the total of highly electrified convective clouds around the
globe [Williams, 2005]. The return current is in the form of a
current density, J
z
, downward through the thickness of the
global atmosphere to the land and ocean surface. J
z
is in the
range 16 pA m
2
, depending on the elevation of that
surface, the output of the tropical thunderstorm generators,
the cloud cover and aerosol content in the atmospheric
column, the radioactive emanations from the land, the
cosmic ray flux, and other inputs modulated by space
weather such as energetic protons and electrons, and auroral
current systems that affect the ionospheric potential in the
polar caps [Tinsley and Zhou, 2006].
[3] The conductivity within clouds is reduced compared
to that of clear air at the same altitude because of the
attachment of atmospheric ions to cloud droplets. The
decrease in s within clouds is by a factor of between 3
and 30 compared to the clear air, according to the simu-
lations of Griffiths et al. [1974]. Thus we have
J
z
sE 1
where E is ambient electric field, which we take here to be
positive downward, and s is the conductivity. With constant
current density, as s is reduced, E must increase.
[4] As J
z
passes through the upper and lower boundaries
of clouds, it creates a gradient of E because of the gradient
in s between cloudy and clear air. In turn, the gradient in E
entails the accumulation of space charge r (difference
between the total positive and total negative charge per unit
volume) according to Gausss law:
r E r=e
0
2
Thus, assuming horizontal stratification,
r e
0
J
z
d
dz
1
s
_ _
3
where z is the distance coordinate, measured vertically
downward.
[5] A layer of positive space charge is produced in the
gradient of conductivity at the tops of clouds, (resistivity r =
1/s is increasing in the direction of positive z and J
z
),
accompanied by a layer of negative space charge in the
decreasing resistivity at cloud base. Most of the space
charge accumulates on droplets, because for the mixture
of air ions, aerosol particles and droplets in typical clouds,
the droplets have the greatest radii of curvature and surface
area. The remaining space charge is distributed between the
aerosol particles and ions, depending on their relative
concentrations. The total space charge is
r e n
1
n
2
g
1
SS
1
g
2
SS
2
pSN
A
4
where e is the elementary charge (1.6 10
19
C); n
1
and n
2
are the concentrations of positive and negative ions; SS
1
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, D11203, doi:10.1029/2006JD007998, 2007
Click
Here
for
Full
Article
1
State Key Laboratory of Environmental Geochemistry, Institute of
Geochemistry, Chinese Academy of Sciences, Guiyang, China.
2
W. B. Hanson Center for Space Sciences, University of Texas at
Dallas, Richardson, Texas, USA.
3
Graduate School of the Chinese Academy of Science, Beijing, China.
4
Now at Department of Geography, East China Normal University,
Shanghai, China.
Copyright 2007 by the American Geophysical Union.
0148-0227/07/2006JD007998$09.00
D11203 1 of 17
and SS
2
are the concentrations of positively and negatively
charged aerosol particles with mean positive and negative
charges of g
1
and g
2
in units of e; SN
A
is the total droplet
concentration, summed over all values of the droplet radius
A; and p is the (algebraic) mean of the number of
elementary charges on both positively and negatively
charged droplets.
[6] Under steady state conditions the total downward
current J
z
is constant, and is the sum of a downward current
density due to downward moving positive ions and a
downward current density due to the upward moving
negative ions. The charged aerosol particles are taken as
stationary (negligible mobility). Thus
J
z
J
1
J
2
5
where
J
1
s
1
E m
1
n
1
eE 6
and
J
2
s
2
E m
2
n
2
eE 7
and where m
1
and m
2
are the mobilities of the positive and
negative ions.
[7] The relative amounts of J
1
and J
2
vary with distance
because the relative amounts of n
1
and n
2
vary on account
of the so-called electrode effect, which creates the space
charge. Just below the top of the cloud more positive ions
can flow downward than negative ions can flow upward,
because of the decreased ion concentration in the cloud.
Similarly, more negative ions can flow upward just above
the bottom of the cloud than positive ions can flow down.
[8] We have modeled the flow of current density through
a set of representative layer clouds, with representative
aerosol concentrations, in order to evaluate the amounts of
charges on droplets and aerosol particles for equilibrium
electrostatic conditions. This is necessary in order to model
the electrical effects on the scavenging by droplets of ice-
forming nuclei (IFN) and cloud condensation nuclei (CCN).
This scavenging appears to have significant effects on cloud
cover and precipitation [Tinsley, 2000; Burns et al., 2007].
2. Construction of Model
[9] The continuity equation for the positive ions is
dn
1
dt
q an
1
n
2
b
1
Sn
1
g
1
S
2
n
1

1
e
r J
1
4pD
1
n
1

1
A0
AN
A
_ _
pc
e
pc
1
8
and for negative ions is
dn
2
dt
q an
1
n
2
b
2
Sn
2
g
2
S
1
n
2

1
e
r J
2
4pD
2
n
2

1
A0
AN
A
_ _
pc
e
pc
1
9
where q is the ion pair production rate per unit volume, a is
the ion-ion recombination rate coefficient; S the concentra-
tion of neutral aerosol particles; b
1
and b
2
are the
attachment rate coefficients for the positive and negative
ions, respectively, to the neutral particles; g
1
and g
2
are the
attachment rate coefficients for positive ions to negatively
charged (concentration S
2
) particles and negative ions to and
positively charged (concentration S
1
) aerosol particles,
respectively; D
1
and D
2
are the diffusion coefficients for
positive and negative ions at the altitude of the cloud; and
from work by Pruppacher and Klett [1997, section 18.3] the
value of p for the equilibrium condition of droplet charging
is given by
pc ln
D
1
n
1
D
2
n
2
_ _
10
where c depends inversely on aerosol radius and in SI units
is
c
e
2
4pe
0
AkT
11
where

A is the mean droplet radius; k is Boltzmanns
constant, and T is the absolute temperature. The opposite
sign for the divergence terms in (9) as compared to (8)
arises because the negative charges move in the opposite
direction to J
z
. The last terms in (8) and (9) are for the
diffusion charging process. The drift charging process
considered by Griffiths et al. [1974] can be shown to be
small compared to diffusion charging for the droplet sizes
and electric fields that we have modeled.
[10] Equations (8) and (9) state that the rate of change of
ion concentration (dn/dt) is the production rate, minus the
loss rates due to ion-ion recombination and attachment,
minus the net flow out of the volume, minus the attachment
rates to droplets. The latter loss rates are for droplets with
their equilibrium average number of charges p.
[11] Although a detailed treatment of ion attachment to
the aerosol would consider the polydispersive aerosol par-
ticle size distribution, with multiple charges on the larger
particles [Hoppel, 1985; Hoppel and Frick, 1986; Yair and
Levin, 1989], we represent the particles by a monodisper-
sive size distribution, representing the dominant size cate-
gory, with at most a single positive or negative charge. Thus
b and g in each case apply to the radius of the dominant
particles. We will consider consequences of these approx-
imations in the Discussion section.
[12] The continuity equations for the positively and
negatively charged aerosol particles are
dS
1
dt
b
1
Sn
1
g
2
S
1
n
2
12
and
dS
2
dt
b
2
Sn
2
g
1
S
2
n
1
13
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
2 of 17
D11203
and at equilibrium when dS
1
/dt = dS
2
/dt = 0, we have
S
1
S
b
1
n
1
g
2
n
2
_ _
14
S
2
S
b
2
n
2
g
1
n
1
_ _
15
and with the total aerosol concentration S
T
= S
1
+ S
2
+ S we
have
S
S
T
1
b
1
n
1
g
2
n
2

b
2
n
2
g
1
n
1
16
[13] The terms for divergence of J, with horizontal
stratification, using (6) and (7) can be expressed as
1
e
r J
1

1
e
@
@z
J
1
m
1
n
1
@
@z
E E
@
@z
m
1
n
1
17
and
1
e
r J
2

1
e
@
@z
J
2
m
2
n
2
@
@z
E E
@
@z
m
2
n
2
18
with
E
J
z
e
1
m
1
n
1
m
2
n
2
_ _
19
from (5), (6), and (7).
[14] On substituting (14) and (15) in (8) and (9) we see
that on the right side of these equations, for equilibrium
conditions, the sums of the third and fourth terms, (U), in
both cases are given by
U S b
1
n
1
b
2
n
2
20
The fifth terms are also equal, since at equilibrium
r J r J
1
r J
2
0 21
[15] If (10) is substituted into (8) and (9), we see that the
sixth terms, (W), are also equal and are given by
W 4pD
1
D
2
n
1
n
2

1
A0
AN
A
_ _
ln D
1
n
1
=D
2
n
2

D
1
n
1
D
2
n
2
22
[16] With the first terms already being equal, and also the
second terms, in (8) and (9), we see that these two equations
are effectively the same, and only one of them can be used
for evaluating n
1
and n
2
, and subsequently evaluating S
1
, S
2
,
S, and p. Either of the equations (8) or (9) can be trans-
formed using (17) or (18), and (19) and (20), with S given
by (16), so that the only unknowns on the right side are n
1
and n
2
. Thus (8) becomes
dn
1
dt
q an
1
n
2
S b
1
n
1
b
2
n
2
m
1
n
1
dE
dz
E
d
dz
m
1
n
1

4pD
1
D
2
n
1
n
2
ln D
1
n
1
=D
2
n
2

D
1
n
1
D
2
n
2

1
A0
AN
A
_ _
23
[17] The other equation that determines the unique values
of n
1
and n
2
(for specified values of J and S
T
, and SN
A
A as a
function of height z through the cloud) is Gausss law
(equation (2)). Making the assumption that only single
charges can exist on the monodispersive aerosol particles,
then g
1
= g
2
= 1, and SS
1
= S
1
, and SS
2
= S
2
. Then, using
(4) and (10) and (11), equation (2) becomes
@E
@z

e
e
0
_
n
1
n
2
S
1
S
2

4pkTe
0
e
2

1
A0
AN
A
_ _
ln D
1
n
1
=D
2
n
2

_
24
Substituting for S
1
and S
2
with (14) and (15), (24) becomes
@E
@z

e
e
0
n
1
n
2
S
b
1
n
1
g
2
n
2

b
2
n
2
g
1
n
1
_ _ _ _

4pkT
e

1
A0
AN
A
_ _
ln D
1
n
1
=D
2
n
2
25
[18] The electric field E and its derivative with distance
are determined in terms of n
1
and n
2
by (19), and S is given
by (16), thus leaving n
1
and n
2
the only unknowns in
equation (25), so that equations (23) and (25) can together
determine a unique solution. The approach that was used in
our modeling was to assume approximate equilibrium con-
ditions and make a first estimate of ion concentration as a
function of distance z by adding together (8) and (9) and
taking n
1
= n
2
= n
0
and dn
0
/dt = 0 and using (20) and (16).
The solutions to the quadratic equations yielded a value of
n
0
for each value of z in the modeled cloud region.
[19] The electric field and its derivative with distance in
this first estimate can be calculated using n
1
= n
2
= n
0
in
(19) with a specified current density J.
[20] Next, make estimates of n
1
and n
2
by putting
n
1
1 x n
0
26
n
2
1 x n
0
27
where 1 < x < 1.
[21] Then (26) and (27) are put into (25) so that it
becomes
@E
@z

e
e
0
_
2n
0
x S
b
1
g
2

1 x
1 x

b
2
g
1

1 x
1 x
_ _

4pkTe
0
e
2

1
A0
AN
A
_ _
ln
D
1
D
2

1 x
1 x
_ _
_
28
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
3 of 17
D11203
In this equation x is the only unknown variable and the
equation can be solved using an iterative numerical method.
Thus we get the first estimate of n
1
and n
2
, which represent
their initial values at the time of the assumed approximate
equilibrium, i.e., as a function of altitude at time t = 0.
Starting with these initial n
1
and n
2
values, the evolution of
n
1
and n
2
with time can be calculated step by step by
iterating (23) and (25) until convergence on a more precise
equilibrium is attained:
n
1
t n
1
t 1
dn
1
t 1
dt
Dt 29
n
2
t n
2
t 1
dn
2
t 1
dt
Dt 30
where n
1
(t 1) and n
2
(t 1) are obtained from the earlier
estimate. In evaluating the derivatives with respect to z
smoothing over adjacent values was used to ensure stability
in the iterations. It should be noted that this continued
iteration in time does not simulate a change in n
1
and n
2
from their values in cloud-free air. Rather, after the first
estimate of n
1
and n
2
we are already very close to
equilibrium, and further iterations are to ensure an even
more precise value.
[22] The values of m
1
, m
2
, D
1
, and D
2
depend on altitude,
i.e., on the pressure and temperature of the atmosphere, and
b
1
, b
2
, g
1
and g
2
depend on particle radius, (a), as well as
on pressure and temperature. We used a model atmosphere
derived from the US Standard Atmosphere (U.S. Standard
Atmosphere, 1976) as in work by Tinsley and Zhou [2006].
We use values of m
1
and m
2
at standard sea level atmo-
spheric pressure and temperature (1013.25 hPa and
288.15 K) of 1.4 10
4
and 1.9 10
4
m
2
V
1
s
1
.
These values originated with Bricard [1965] and were said
to be for normal pressure and temperature, but are
described as being for STP by Shreve [1970] and
Pruppacher and Klett [1997, section 18.1]. However, the
variation of mobility with altitude given by these authors
implies that the STP temperature is the standard sea level
temperature of 288.15 K. We regard Bricards experimental
values, which depend on the differing molecular weights of
the molecular clusters forming the positive and negative air
ions, as representative. They are dependent on atmospheric
chemical processes; however, for cloud environments in the
troposphere we follow Pruppacher and Klett [1997] and
consider them representative. The relationship between
mobility of ions and their diffusion coefficient D in air is
[e.g., McDaniel and Mason, 1973]
D mkT=e 31
D varies with temperature, in part because the cross section
for collision of the ions with air molecules varies with
collision energy. To include this effect and we took the
variation of mobility with altitude, i.e., of m
z
with pressure
p
z
and temperature T
z
at altitude z, referred to m
o
at pressure
p
o
and temperature T
o
as
m
z
m
0

p
0
p
z
T
z
T
0
_ _
1:5
T
0
120
T
z
120
32
where the expression is derived in Appendix A. Using
Bricards values at z = 0 km, the values of m
1
and m
2
for 2
km are then 1.63 10
4
and 2.21 10
4
m
2
V
1
s
1
,
respectively, and at 4 km 1.90 10
4
and 2.58 10
4
m
2
V
1
s
1
, respectively, and at 10 km 3.43 10
4
and 4.64
10
4
m
2
V
1
s
1
, respectively. The formula (32) differs
from an alternative formula for variation of mobility with
altitude: m
z
/m
0
= (p
0
/p
z
)(T
z
/T
0
) that has often been used in
atmospheric electricity literature. It evidently originated
with Bricard [1965], and does not take into account the
variation of D and of m with temperature at constant density.
This variation is implicit in the treatment by McDaniel and
Mason [1973] and in the present treatment. For further
discussion of this point see Appendix A.
[23] The values of b
1
, b
2
, g
1
and g
2
were obtained from
the theory of Keefe et al. [1968] and Hoppel and Frick
[1986]. These formalisms include the effects of image
charge as well as three-body trapping of the ions, with
diffusion bringing the ions close enough to the neutral or
charged aerosol particle to be captured. Table 1 gives the
values of b
1
, b
2
, g
1
and g
2
for 4 km and 10 km for particles
with a = 2 10
8
m; 4 10
8
m; and 10 10
8
m.
3. Results
3.1. General Features
[24] Figure 1 shows a set of input and output parameters
from one run of the model. The model consists of five
Table 1. Values of b
1
, b
2
, g
1
, and g
2
Used in the Simulations at 2 km, 4 km, and 10 km Altitude for the Aerosol Particle
Radii 0.02 mm, 0.04 mm, and 0.10 mm
a, mm b
1
, 10
12
m
3
s
1
b
2
, 10
12
m
3
s
1
g
1
, 10
12
m
3
s
1
g
2
, 10
12
m
3
s
1
z
c
= 2 km
0.02 0.61 0.86 2.36 3.13
0.04 1.53 2.16 3.32 4.55
0.10 4.55 6.38 6.23 8.65
z
c
= 4 km
0.02 0.63 0.88 2.56 3.37
0.04 1.61 2.27 3.64 4.98
0.10 4.90 6.88 6.83 9.49
z
c
= 10 km
0.02 0.70 0.96 3.06 3.91
0.04 1.93 2.70 4.84 6.49
0.10 6.62 9.34 9.72 13.47
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
4 of 17
D11203
regions, divided according to the droplet concentration. As
in Figure 1a, the first 2 m, with z increasing downward,
contain aerosol and ionization, but no cloud droplets. The
next region is the cloud upper boundary layer, of variable
thickness, where the droplet concentration increases to its
maximum value, after which there is a layer of arbitrary
thickness (for convenience set at 11 m) to represent a
uniform cloud interior, with constant droplet concentration
equal to the maximum. Then the lower boundary layer is of
variable thickness, with droplet concentration decreasing to
zero, and a lower droplet free layer that is set at 2 m
thickness. Since the interior is uniform, the results for the
boundary layers are applicable to clouds with interiors that
are uniform and of any thickness, not just the 11 m
thickness used here. To avoid discontinuities in the numer-
ical integration, the droplet concentration in the boundary
layers follows a sin
2
(p(z z
1
)/(2z
b
)) variation, where z
b
is
the thickness of the boundary layer, and z
1
is either edge of
the cloud. There is little difference in the results using other
smooth monotonic variations. The total aerosol concentra-
tion S
T
is kept constant throughout the cloud, and the mean
droplet radius

A = 10 mm in all runs.
[25] In Figure 1 the cloud is at the altitude of 4 km; with
ion production q = 8.0 10
6
m
3
s
1
. The current density
J
z
= 3.0 10
12
A m
2
; the maximum droplet concentra-
tion (SN
A
)
max
= 2.0 10
8
m
3
(and thus SAN
A
=

A SN
A
);
the aerosol concentration S
T
= 5.0 10
9
m
3
; the (mono-
dispersive) aerosol radius 4 10
8
m (0.04 mm), and each
boundary layer is 10 m thick. In subsequent runs sensitivity
tests are made by varying these parameters.
[26] We will examine the mean ion concentration (n
0
), the
strength of electric field (E), the concentrations of positive
ion and negative ions (n
1
, n
2
) and their difference (n
1
n
2
),
the concentration of positively charged and negatively
Figure 1. Profiles of parameters of a simulation of space charge production in a cloud, as a function of
distance z measured downward through it: (a) droplet concentration, (b) mean ion concentration,
(c) electric field, (d) divergence of field, (e) ion concentration difference (n
1
n
2
), (f) charged aerosol
particle concentration difference (S
1
S
2
), (g) mean droplet charge, (h) charged aerosol concentrations
(S
1
and S
2
), and (i) ion concentrations (n
1
and n
2
). For other parameters for this simulation, see text. In
Figures 1b1g, the solid line is for the initial estimate, and the dotted line is for 600 s simulation. In
Figures 1h and 1i, the solid and dotted lines are for the initial estimates, and the dashed and dash-dotted
lines are for 600 s. In Figure 1h, the solid and dashed lines are for positively charged aerosol particles,
and the dotted and dash-dotted lines are for the negatively charged particles. In Figure 1i, the solid and
dashed lines are for the positive ions, and the dotted and dash-dotted lines are for the negative ions. The
thin solid lines are the zero references.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
5 of 17
D11203
charged aerosol particles (S
1
, S
2
) and their difference (S
1

S
2
), the mean number of elementary charges on the droplets
(p) in the cloud layer as a function of layer thickness for up
to 600 s iteration time. We will also examine the effect of J
z
;
(SN
A
)
max
; S
T
; a; the thickness of the cloud boundary layers
z
b
; and height of the cloud z
c
on the various elements of
space charge that are generated by the flow of J
z
into the
cloud.
3.2. Charge Distribution in the Cloud Layer
[27] Figures 1b1g shows relevant parameters as a func-
tion of z; with initial values (t = 0) after the first estimate of
the parameters shown as solid lines, and values after 600 s
shown as dotted lines. Figure 1b gives n
o
; Figure 1c gives
E; Figure 1d gives the derivative of the electric field
@E
@z
;
Figure 1e gives (n
1
n
2
); Figure 1f gives (S
1
S
2
); and
Figure 1g gives p (as a net mean charge p can be positive
or negative). Figure 1h shows S
1
as solid and dashed lines
for t = 0 and t = 600 s and S
2
as dotted and dash-dotted lines
for the same times. Figure 1i shows n
1
and n
2
with the same
line designations.
[28] From Figures 1a and 1b it can be seen that with the
increase of droplet concentration with depth into the cloud
in the upper boundary layer the ion concentration n
0
decreases because of ions attaching onto the droplets. In
the middle part of the cloud the constant droplet concen-
tration ensures that n
0
remains constant until the decrease of
droplet concentration in the lower boundary layer causes n
0
to increase back to its original value. In Figure 1c the electric
field E rises in the upper boundary layer, is constant in the
uniform part of the cloud, and decreases in the lower boundary
layer, which causes the positive peak of
@E
@z
in the upper
boundary layer and negative peak of
@E
@z
in the lower boundary
layer (see Figure 1d). In accordance with equation (28),
@E
@z
determines the space charge distribution in the cloud.
Although the electric field derivatives reach essentially
equal and opposite amplitudes in the upper and lower
boundary layers, this is not true for all the components of
the space charge, because the mobility of the negative ions
is greater than that of the positive ones, and more negative
Figure 2. Profiles for variation of vertical current density (J
z
) for 600 s simulation. Dashed lines are for
J
z
= 4 10
12
A m
2
; solid lines for 2 10
12
A m
2
; and dotted lines are for 1 10
12
A m
2
. Thin
solid lines are the zero references. The parameters plotted are (a) mean ion concentration, (b) electric
field, (c) divergence of field, (d) ion concentration difference, (e) charged aerosol concentration
difference, and (f) mean droplet charge.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
6 of 17
D11203
aerosol particles are formed, leaving more positive ions
than negative. Thus in Figures 1i and 1e it can be seen that
n
1
> n
2
in the cloud-free areas and in the uniform central
section, with the differences being about 1.8 10
8
m
3
and 1 10
7
m
3
, respectively. (Note the logarithmic scale in
Figure 1i and the linear scale in Figure 1e). In the upper
positive space charge regions the magnitude of the (n
1
n
2
)
difference reaches 4 10
8
m
3
, but in the lower negative space
charge region the difference only reaches 1.1 10
8
m
3
.
[29] The related asymmetry in S
1
and S
2
can be seen
in Figures 1h and 1f. In the cloud-free and central regions
S
1
< S
2
, with the differences being about 1.8 10
8
m
3
and 5 10
7
m
3
, respectively. In the upper space charge
region the difference (S
1
S
2
) is positive and reaches a
maximum of 1.1 10
9
m
3
, then gradually reduces to the
constant small negative value in the central region, while in
lower space charge region it gradually approaches its great-
est negative value of 1.25 10
9
m
3
, and then rapidly
increases up to the cloud-free value.
[30] Figure 1g shows the mean number of charges on the
droplets in the cloud. In the upper space charge region the
mean charge is positive and its amount sharply rises to
maximum value of about 90e and then gradually decreases
to 0.17e in the central region (i.e., only slightly more
positive than negative charged droplets). In the lower space
charge region the mean droplet charge is negative, gradually
approaching the peak value of about 90e and then rapidly
decreasing in (absolute) value to near zero again.
[31] After 600 s iteration time only small differences in
the parameters are found as compared to the initial estimate
of equilibrium, as is apparent in all the panels of Figure 1.
These differences do not increase significantly after more
than 600 s of iteration. The differences are partly due to the
smoothing, and partly (as in the uniform middle section)
from the approximation m
1
= m
2
and n
1
= n
2
in the initial
estimate for n
0
. Simulations for t = 2 s are closer to that of
t = 600 s. We consider that compared to other uncertainties
in the model the t = 0, t = 2 s, and t = 600 s results are
equivalent.
3.3. Effect of the Current Density J
z
[32] From equation (19) reduction of J
z
will reduce E and
@E
@z
and affect the space charge distribution. Figures 2a2f
show the results for 600 s iteration, with three different
current density values. The dashed lines are for J
z
=
4 10
12
A m
2
; the solid lines for 3 10
12
A m
2
; and
the dotted lines for 2 10
12
A m
2
. The zero level in
Figures 2c2f is shown as a thin solid line.
Figure 3. Profiles for variation of droplet concentration. Dashed lines are for the maximum droplet
concentration of 4 10
8
m
3
; solid lines are for 2 10
8
m
3
; and dotted lines are for 1 10
8
m
3
. The
parameters that are plotted are the same as in Figure 2.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
7 of 17
D11203
[33] The other parameters are: z
c
= 4 km; q = 8 10
6
m
3
s
1
; (SN
A
)
max
= 2 10
8
m
3
; S
T
= 5 10
9
m
3
; and
z
b
= 10 m. In Figure 2a it can be seen that changing J
z
did
not significantly change n
0
in the cloud layer. In Figure 2b
an increase in J
z
causes a proportionate increase of electric
field (E) throughout the cloud. In Figures 2b and 2c the
electric field gradient
@E
@z
scales as E and J
z
, and causes the
space charge to scale proportionately. So in Figures 2d2f
there are increases in (n
1
n
2
), (S
1
S
2
), and p in the
boundary layer due to increases of
@E
@z
, especially in p and
(S
1
S
2
) that carry most of the space charge.
3.4. Effect of Cloud Droplet Concentration
[34] Enhancement of the concentration of cloud droplets
increases the attachment surface for ions, reduces the ion
concentration and conductivity, and increases the electric
field and its gradients and the space charge. Figures 3a3f
shows the results after 600 s iteration with three values of
droplet concentration; the dotted lines are for (SN
A
)
max
1 10
8
m
3
; solid lines for 2 10
8
m
3
; and dashed
lines for 4 10
8
m
3
. The other parameters are: z
c
= 4 km;
J
z
= 3 10
12
A m
2
; q = 8 10
6
m
3
; S
T
= 5 10
9
m
3
:
a = 4 10
8
m; z
b
= 10 m.
[35] In Figure 3a the increases in droplet concentration
causes decreases of n
0
in the cloud layer. In Figures 3b and
3c the decreased conductivity causes higher electric field E
in the cloud layer, and corresponding increases in
@E
@z
. In
Figure 3d with higher droplet concentration the curve for
(n
1
n
2
) peaks at higher values, and closer to the edge of
the cloud. This applies also to (S
1
S
2
) as shown in
Figure 3e and to p as shown in Figure 3f. For the uniform
central region of the cloud (n
1
n
2
) decreases because of
the increasing droplet concentration and decreasing n
0
.
3.5. Effect of Thickness of Boundary Layer
[36] The boundary layer thickness z
b
determines
@E
@z
,
which affects all the components of the space charge.
Figures 4a4f show the results for 600 s iteration; the
dashed lines are for z
b
= 5 m; dotted lines for 10 m; and
solid lines for 50 m. The other parameters are: z
c
= 4 km;
q = 8 10
6
m
3
; J
z
= 3 10
12
A m
2
; (SN
A
)
max
= 2
10
8
m
3
; S
T
= 5 10
9
m
3
; a = 4 10
8
m. In Figure 4a,
the increase of boundary layer thickness causes a less rapid
decrease of n
0
with distance into the upper boundary layer,
and a less rapid increase coming out of the lower one. So in
Figures 4b and 4c there are lower values of
@E
@z
for the
Figure 4. Profiles for variation of boundary layer thicknesses. Dashed lines are for thicknesses of 5 m;
dotted lines are for 10 m; and solid lines are for 50 m. The parameters that are plotted are the same as in
Figure 2.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
8 of 17
D11203
greater boundary thicknesses. In Figure 4d in the upper
boundary layer the value of (n
1
n
2
) decreases as the layer
widens, while in the lower boundary layer (n
1
n
2
) shows a
negative peak only for the 5 m and 10 m layers. For the 50 m
thick boundary layer the lower negative peak disappears.
This is because the weaker space charge is overcome by the
effect of the greater negative ion mobility. In Figure 4e the
wider boundary layer causes lower (S
1
S
2
) within it.
Similarly in Figure 4f the wider boundary layer results in a
reduction in p within it. However, theory requires, and the
figures indicate that the integrated space charge with depth
through either of the boundary regions remains constant.
The theoretical reason is that integrated space charge is
effectively a surface charge s
s
= e
o
(E
i
E
o
) to accommo-
date the change in electric field from E
o
outside the cloud to
E
i
in the uniform interior of the cloud.
[37] For the 50 m case in Figure 4f the small negative
excursion at the very top of the upper space charge layer
appears to be an anomalous feature. However, it is also due
to the preponderance of the effects of negative ion mobility
exceeding positive mobility. The relatively slow increase of
space charge and thus of the ratio n
1
/n
2
with distance into
the cloud (see (n
1
n
2
) for the 50 m case in Figure 4d)
creates a small region where the droplet concentration is
nonzero and n
1
/n
2
is less than the inverse of the ratio D
1
/D
2
=
m
1
/m
2
= 1.4/1.9. Under these circumstances, (D
1
n
1
)/(D
2
n
2
) < 1,
and from equation (10) p is negative.
3.6. Effect of Altitude of Cloud
[38] There are several parameters which increase with the
altitude of the cloud, z
c
, and affect the space charge
components: the ion mobility and the ion-aerosol attach-
ment rate coefficients as noted earlier, and the ion produc-
tion rate, q, which increases with altitude up to about 15 km
[Tinsley and Zhou, 2006]. For 10 km q is taken as 3.9 10
7
m
3
; for 4 km 8.0 10
6
m
3
and for 2 km 4.5 10
6
m
3
.
The other parameters are J
z
= 3 10
12
A m
2
; (SN
A
)
max
=
2 10
8
m
3
; S
T
= 5 10
9
m
3
; a = 4 10
8
m; z
b
= 10 m.
Figures 5a5f shows the results; the dashed lines are for
10 km; dotted lines for 4 km; and solid lines for 2 km.
[39] In Figure 5a the mean ion concentration at 10 km is
higher than that at 4 km which is higher than 2 km, due
mainly to the higher ion production rates at higher altitudes.
In Figure 5b the enhancement of the ion concentration and
mobility at the higher altitudes causes the decrease of the
electric field E. Then in Figure 5c in the space charge
region,
@E
@z
shows a marked decrease with the increase in
Figure 5. Profiles for variation of cloud altitude. Dashed lines are for 10 km altitude; dotted lines are for
4 km; and solid lines are for 2 km. The parameters that are plotted are the same as in Figure 2. The
waviness near 13 m and 23 m in Figure 5c for the 2 km case is due to an instability in the computing.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
9 of 17
D11203
cloud height. So in Figures 5e and 5d for the 10 km
simulation the tendency for the predominantly negative
ion mobility to produce negative (S
1
S
2
) and positive
(n
1
n
2
) has comparable effects to that of the space charge,
and so in Figure 5f p is negative, or only slightly positive.
Because of an instability in computing
@E
@z
for the 2 km case,
the iteration was stopped just beyond the first estimate of n
1
and n
2
, i.e., for just 2 s; however, for the 4 km and 10 km
cases, the results are indistinguishable from iterations for
600 s.
3.7. Effect of Aerosol Concentration
[40] The rate of attachment of ions to aerosols is directly
proportional to their concentration, affecting the distribution
of charges on aerosols, ions, and droplets. Figures 6a6f
show the results; the dotted lines are for S
T
= 5 10
7
m
3
;
the dash-dotted lines for 5 10
8
m
3
; the dashed lines for
5 10
9
m
3
; and the solid lines for 6 10
10
m
3
, from
simulations for 600 s. The other parameters are z
c
= 4 km;
J
z
= 3 10
12
A m
2
; (SN
A
)
max
= 1 10
8
m
3
(this is
smaller than in previous cases); a = 4 10
8
m, z
b
= 10 m.
[41] In Figure 6a, there exits a larger difference in n
0
between the curves for S
T
= 6 10
10
m
3
and 5 10
9
m
3
,
where attachment to aerosol particles dominates the
ion recombination, than between those of 5 10
9
m
3
and 5 10
8
m
3
, and between those of 5 10
8
m
3
and
5 10
7
m
3
,where attachment is becoming less important
compared to ion-ion recombination outside the cloudy
areas, and less important compared to attachment to droplets
inside them. The higher electric field (E) curve for the
higher aerosol concentration cases reflects the lower ion
concentrations, but there is little difference in
@E
@z
for the
four aerosol conditions, as illustrated in Figures 6b and 6c.
In Figure 6d the lower aerosol concentration can be seen to
allow a greater net positive charge do develop in the upper
boundary layer, and a greater net negative charge in the
lower boundary layer, while for the highest aerosol concen-
tration of 6 10
10
m
3
, the effects of space charge are very
small, and there is only a small net positive ion excess
throughout the cloud layer. In Figure 6e the differences in
the amounts of positively and negatively charged aerosol
varies approximately as the total aerosol concentration.
Figure 6. Profiles for variation of aerosol concentration. Dotted lines are for total concentration of
aerosol particles 5 10
7
m
3
; dash-dotted lines are for 5 10
8
m
3
; dashed lines are for 5 10
9
m
3
;
and solid lines are for 6 10
10
m
3
. The parameters that are plotted are the same as in Figure 2.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
10 of 17
D11203
Figure 6f shows the mean charges on the droplet for the four
values of aerosol concentration, and it can be seen that for
most of the boundary layer, the effect of aerosol concentra-
tion on the droplet charge is small except for very high
aerosol concentrations.
[42] The same small negative excursion in p at cloud top
is present, as in Figure 4f, for the lowest aerosol concen-
tration. In this case the effect of greater negative mobility
appears, not because of a wider boundary layer, but because
n
1
/n
2
is closer to unity for the lowest aerosol concentrations,
and the lower peak droplet concentration extends the region
of initial very low droplet concentration.
3.8. Effect of (Monodispersive) Aerosol Radius
[43] As noted in section 2, the aerosol size affects the ion-
aerosol attachment coefficients (b, g) according to the
complex formalism of Keefe et al. [1968], Hoppel [1985],
and Hoppel and Frick [1986], and these in turn affect the
partition of the space charge between ions, charged aerosol
particles and droplets. Figures 7a7f show model results;
the dashed lines are for a = 10 10
8
m; the solid line is
for 4 10
8
m; and the dotted line for 2 10
8
m.
The other parameters are: are z
c
= 4 km; q = 8 10
6
m
3
;
J
z
= 3 10
12
A m
2
; (SN
A
)
max
= 2 10
8
m
3
; S
T
= 5
10
9
m
3
; z
b
= 10 m.
[44] Generally, smaller particles have smaller attachment
coefficients, and in Figure 7a, the reduction of aerosol
radius causes the reduction of attachment coefficient and
increase in concentration of n
0
. As a result, in Figures 7b
and 7c, the reduction of aerosol radius causes the decrease
of the strength of the electric field, E. However, as in earlier
cases,
@E
@z
is essentially unchanged in the three cases. In
Figures 7d, 7e, and 7f the reduction of aerosol size and
attachment diverts the space charge into the ions as it
decreases it in the charged aerosol particles, and increases
it slightly in the charged droplets.
3.9. Effect of the Difference in Ion Mobilities
[45] As has been mentioned earlier, the effect of the
negative ion mobility being greater than the positive ion
mobility has caused the magnitude of the maximum positive
value of p in the positive space charge region to exceed the
magnitude of the maximum negative value of p in the
negative space charge region. Also, in the absence of space
charge, the value of (S
1
S
2
) is negative, while the value of
(n
1
n
2
) is positive. This can be seen for the droplet-free
regions above and below the cloud in Figures 1e, 1f, 1h, and
1i and in Figures 3d, 3e, 4d, 4e, 5d, 5e, 6d, 6e, 7d, and 7e.
In the uniform central region of the cloud the effects are not
Figure 7. Profiles for variation of (monodispersive) aerosol radius. Dashed lines are for radius 10 10
8
m;
solid lines are for 4 10
8
m; and dotted lines are for 2 10
8
m. The parameters that are plotted are
the same as in Figure 2.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
11 of 17
D11203
as readily apparent, on account of the greatly decreased ion
concentration, but in Figure 6f, with low droplet concen-
tration and a wide range of aerosol concentrations, p shows
a small positive value (2.3e) for the largest S
T
of 6 10
10
m
3
, and a barely detectable negative value (0.5e) for the
smallest S
T
of 5 10
7
m
3
.
[46] To illustrate these effects better, and minimize the
space charge, the model results in Figure 8 are for (SN
A
)
max
=
1 10
8
m
3
; q = 1 10
7
m
3
; J
z
= 1 10
12
A m
2
; with
the dotted lines for S
T
= 5 10
9
m
3
; the dashed line for
1 10
10
m
3
; and the solid line for 3 10
10
m
3
. The other
parameters were: z
c
= 4 km; a = 4 10
8
m, z
b
= 10 m. With
the expanded scale for p in Figure 8f it can be seen that for
the highest aerosol concentration (solid line) the mean
droplet charge in the central region rises to about 1.6e, and
at the extreme lower edge of the cloud, even in the presence
of some residual space charge, the average droplet charge
rises to about 2.4 elementary charges.
[47] The parameters that determine the droplet charge are
given in equations (10) and (11), so that with D
1
/D
2
= 1.4/
1.9, a positive value of p will be present when n
1
D
1
> n
2
D
2
,
i.e., when n
1
/n
2
> 1.357. This is the case in the central
region for all of the aerosol concentrations in Figure 8.
While (n
1
n
2
) is greater for the lower aerosol concen-
trations, the ratio n
1
/n
2
is greater for the highest, on account
of the lower values of n
1
and n
2
(thus of n
0
in Figure 8a). In
Figure 6f, n
1
D
1
> n
2
D
2
in the space charge free regions for
the largest aerosol concentration (6 10
10
m
3
) but not for
the smallest (5 10
7
m
3
).
[48] Also, there are small regions of near-zero droplet
concentration and values of n
1
/n
2
< 1.357, which give
negative p at the upper edges of the cloud. This occurs for
the S
T
= 5 10
9
m
3
case in Figure 8f and also for the 10 km
case in Figure 5f and the 5 10
7
m
3
and 5 10
8
m
3
cases
in Figure 6f. This requires low aerosol concentrations or high
ion production rates, so that n
1
/n
2
! 1, enhanced by the
absence of large amounts of positive space charge, which
otherwise bring n
1
/n
2
> 1.357 near the upper cloud edges.
4. Discussion
4.1. Dependence of Space Charge on J
z
and on Droplet
Concentration
[49] J
z
is controlled by the ionosphere potential V
i
and the
local atmosphere columnar resistance R, so that J
z
is given
by V
i
/R. There are variations in V
i
on many timescales, from
Figure 8. Profiles to illustrate the consequences of the difference between positive and negative ion
mobilities, with reduced ion production rate, reduced droplet concentration, and reduced current density,
as described in the text. The results are for three values of total aerosol particle concentration. Dotted
lines are for S
T
= 5 10
9
m
3
; dashed lines are for 1 10
10
m
3
; and solid lines are for 3 10
10
m
3
.
The parameters that are plotted are the same as in Figure 2.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
12 of 17
D11203
a diurnal variation in universal time to the Milankovich
timescales of orbital variation [Tinsley and Zhou, 2006].
These variations are in response to both external and
internal forcings (cosmic ray and other space weather
inputs, and tropical climate change). R can vary by tens
of percent in response to cosmic ray flux changes over the
11-year solar cycle in the high geomagnetic latitude regions.
Associated with these must be comparable variations in J
z
.
[50] In most cases considered here, most of the attach-
ment of ions within the clouds occurs on droplets rather
than on aerosol particles, and so the mean ion concentration
n
0
is determined mainly by the droplet concentration SN
A
as
in equations (8) and (9). So in equation (19) (m
1
n
1
+ m
2
n
2
)
depends approximately inversely on the droplet concentra-
tion, and either a higher droplet concentration or a larger J
z
produces a proportionately larger E and larger
@E
@z
in the
boundary layer, as is illustrated in Figures 2c and 3c. Then from
equation (25), high
@E
@z
will cause proportionately larger space
charge to appear, mainly on the droplets but with smaller
amounts on the charged aerosol particles and ions, as in
Figures 2e and 2f and 3e and 3f.
[51] Thus the enhancement of J
z
in clouds with moderate to
high droplet concentration can increase the mean charge on the
droplets and aerosol particles (positive in the upper boundary
layer and negative in the lower layer). A 50% variation of J
z
can cause up to 40% variation of charge on the droplets and
particles. These variations can affect the collision process
between the droplet and the particles that constitute ice-
forming nuclei and condensation nuclei [Tinsley et al., 2000,
2001, 2006], and may affect cloud cover and climate.
4.2. Dependence of Space Charge on Thickness of
Boundary Region
[52] The models presented here have been for stable
cloud layers. However, localized downward (upward) air
motion in the upper (lower) boundary layer causes a
narrowing of the layer, as seen, for example, in the sharp
upper surfaces of convective clouds. A downdraft or updraft
also inserts higher-conductivity air into the cloud layer, that
has lower conductivity, and thus acts to increase the electric
field, in the same way as a mountain peak concentrates
electric field around its summit [see Reiter, 1992, pp, 158
162]. This would increase J
z
and the droplet charges. The
narrowing of the boundary layer increases the gradient of
conductivity and of
@E
@z
as can be seen in Figure 4c and also
produces larger droplet and aerosol charges, as in Figures 4f
and 4e. The measurements of droplet charges in thick
altostratus and stratocumulus clouds by Beard et al.
[2004] showed largest positive average droplet charges
(p $ 85e) in downdrafts at cloud top and largest negative
average values (p $ 65e) in updrafts at cloud base.
[53] Harrison and Carslaw [2003, Figure 5] discussed
the inferred space charge from electric field measurements
made by Reiter [1992, p. 197] above and below a stable
stratus cloud in a valley, with the measurements being made
on instruments suspended below a cable car, ascending
through the cloud. Reiter showed electric field gradients
consistent with conductivity gradients above and below the
visible boundaries of the cloud, which he described as
very thin stratocumulus cloud that from his sketch had
a total thickness of about 50 m. The electric field and
conductivity gradients imply space charge up to 120 m
above the upper visible boundary of the cloud. There is little
difference between the average conductivity or electric field
within the designated cloud and a point 50 m above the
upper surface. Below the cloud the conductivity increased
and electric field decreased uniformly over a distance of
$190 m to the fair weather values. It appears that the space
charge near the boundaries of this very thin cloud was not
determined by gradients in conductivity caused by attach-
ment of ions to cloud water droplets, but caused by
gradients in the concentration of aerosol or haze or pre-
condensation particles below and above the cloud. A
summary of a large number of similar valley stratus clouds
from the instrumented cable car is given by Reiter [1992,
Figure 4.38] where he shows that a thickness of 500 m of
these clouds is needed to produce an increase in electric
field (or reduction in conductivity) by a factor of two, and
that a thickness of less than about 40 m has no effect.
[54] Thus this cloud is not representative of the clouds we
have modeled here, or those observed by Beard et al.
[2004]; clouds with relatively high liquid water content
and droplet concentrations 100150 cm
3
, where the
charge is present on droplets inside the cloud boundaries.
In particular, the width of the space charge regions near the
boundaries in the valley stratus is evidently much greater
than is the case in dense stratocumulus clouds, where
updrafts and downdrafts appear to produce the hard edges
characteristic of considerably narrower boundary regions.
4.3. Dependence of Space Charge on Altitude of Cloud
[55] In the troposphere (except for the lowest 1 or 2 km
over land) the ion production rate is determined by the
cosmic ray flux, which increases rapidly with altitude
[Tinsley and Zhou, 2006]. In the present model the ion
production rate at 10 km is 4.9 times larger than that at
4 km. The mobility also increases with altitude. Then the
conductivity in and around the cloud at high altitude is
considerably larger, and the electric field and its gradients
considerably smaller, than that at low altitude, as in
Figures 5a5c. In equation (25), smaller
@E
@z
in the boundary
layers for the high cloud causes smaller net charge on the
aerosol and droplets, compared to the low-altitude case, as in
Figures 5e and 5f. Thus a given value of J
z
causes much
greater accumulation of space charge on droplets and
aerosol particles for low clouds as compared to high clouds,
and greater amounts of electroscavenging, as suggested by
Tinsley et al. [2000, 2001, 2006]. This greater amount of
charging of low clouds compared to high clouds may be
a factor in explaining the observations of Marsh and
Svensmark [2000] that the greatest response of clouds over
the solar cycle is found in low clouds.
4.4. Dependence of Space Charge on Aerosol
Concentration and (Monodispersive) Radius
[56] The effect of changing aerosol concentration is
mainly on ion concentration outside the cloud and on its
fringes (where the droplet concentration tends to zero)
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
13 of 17
D11203
except when the aerosol concentration is so high that
attachment to aerosol dominates over attachment to droplets
in the space charge region. Since concentrations of 10
8
to
10
9
m
3
are typical of the 410 km level of the atmosphere
[Tinsley and Zhou, 2006], the result of Figure 6, show that
variations of aerosol concentration, except in heavily pol-
luted situations, have a relatively minor effect on droplet
charges in the space charge region.
[57] The effect of reducing the (monodispersive) aerosol
radius is, on the whole, similar to that of a reduction in
aerosol concentration, as can be seen by comparing Figures
6 and 7. This is on account of the reduced attachment rates
when either concentration or size is reduced. There are
departures from this similarity in the areas outside and on
the fringes of the cloud, where differences in the ratio of
attachment coefficients of positive and negative ions, de-
pendent on aerosol particle radius [Hoppel and Frick, 1986]
have significant effects.
4.5. Effect of a Polydispersive Aerosol Size Distribution
[58] To consider how the present results would be affected
by the presence of a polydispersive aerosol size distribution
instead of the monodispersive one used, we consider the
results of Hoppel and Frick [1986], whose formalism was
used in the present simulations, and the results of Yair and
Levin [1989]. Hoppel and Frick [1986] considered a two-
component aerosol dominated by a monodispersive
component with a = 0.02 mm, and used values of m
1
and
m
2
of 1.20 10
4
and 1.35 10
4
m
2
V
1
s
1
,
respectively.
[59] With the attachment to particles dominating over ion-
ion recombination, they obtained a ratio n
1
/n
2
= 1.310. With
this ion concentration ratio considered as fixed and deter-
mining the charging of a second monodispersive compo-
nent, with radii up to 1 mm, they showed that for particles
with radii greater than 0.02 mm there would be more
positively than negatively charged particles; that is, the
particles would have a mean charge p that was positive.
Also, the larger particles had larger mean charges. Thus the
present simulations, where the large particles are 10 mm
radius droplets, are consistent with their results.
[60] Yair and Levin [1989] took the simulations to a
greater level of complexity, by calculating the distribution
of charges as a function of particle size for several poly-
dispersive aerosols, where the ion asymmetry ratio, n
1
/n
2
, is
determined in a self consistent way by the varying attach-
ment coefficients as a function of particle size and the
values of m
1
and m
2
. As in the work of Hoppel and Frick
[1986], for a situation where the dominant particles in the
distribution have a radius comparable to the ion mean free
path, and the temperature is about that of the lower
atmosphere, then with D
2
larger than D
1
the dominant small
particles charge negatively on average. If the concentration
of aerosol particles is relatively high so that (S
1
+S
2
) )(n
1
+n
2
)
with negligible ion-ion recombination, then the formalism
leads to an ion asymmetry ratio n
1
/n
2
that is greater than
D
2
/D
1
, the inverse of the diffusion coefficient ratio, so that
D
1
n
1
> D
2
n
2.
[61] For the larger particles in a polydispersive distribu-
tion the mean number of charges on each tends (with
increasing radius compared to mean free path), to the
limiting value given by equations (10) and (11):
p 1:157 10
7

A ln D
1
n
1
= D
2
n
2
33
for the atmospheric temperature at 4 km altitude.
[62] Thus, with D
1
n
1
> D
2
n
2
, the mean number of charges
on the larger particles will be positive, although the space
charge remains zero. This also requires that the larger
particles have a small enough concentration so that even
when they take up multiple positive charges, their share of
the positive charge does not reduce the ion asymmetry ratio
below the inverse of the diffusivity ratio.
[63] The work of Hoppel and Frick [1986, Figure 9]
provides a guide as to the particle radius with which this
limit is approached. They show results for the equilibrium
ion ratio n
2
/n
1
, for a monodispersive aerosol for which
almost all of the charge is on the particles, and a negligible
amount on the ions, and where there is no space charge (so
that p ! 0). In this case the aerosol radius for which the
limit of equation (33) is approached is when n
2
/n
1
approaches the ratio D
1
/D
2
(0.89 in this case). This limit,
for charge on the particle as a function of radius given by
(33), is met within about 1% for particles of radius 1 mm or
greater.
[64] This means that for mixtures of droplets with a
polydispersive aerosol, provided that the total charge on
the larger aerosol particles is small compared to that on the
droplets, then the charges on the droplets and larger aerosol
particles are all given, to a degree of approximation, which
improves with aerosol size, by equation (33). The above
result, i.e., that the charges are approximately proportional
to the radii, applies whether or not the actual ion asymmetry
ratio is caused by space charge.
[65] For our simulations, when the 10 mm radius droplets
have charges of positive or negative magnitude of 50e, then
to a good approximation the charges on aerosol particles of
1, 2, and 5 mm would be 5e, 10e, and 25e, respectively. For
these particles, and for somewhat smaller ones, the charges
are sufficient to significantly increase the rate of them being
electroscavenged by droplets [Tinsley et al., 2001, 2006].
For the subset of these particles with properties of contact
ice nuclei, the resulting primary ice nucleation would tend
to increase rates of precipitation in cold clouds. Also, such
scavenging of the larger CCN would narrow the CCN size
distribution, and this would subsequently narrow the droplet
size distribution in subsequent cycles of cloud formation,
and tend to reduce rates of precipitation in both warm and
cold clouds.
4.6. Other Effects
[66] The simulations presented in Figures 18 were for
total aerosol concentration S
T
constant with height z, with
only SN
A
varying in the cloud boundary regions. These
results suggest that for isolated layers of aerosol, positive
and negative space charge, of magnitude proportional to J
z
,
would similarly develop at the upper and lower boundaries.
At least for the upper boundary of the tropospheric region
where upward mixing of aerosols from the surface takes
place (the top of the mixing, or tropospheric boundary
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
14 of 17
D11203
layer) observations of electric field variations show the
presence of positive space charge [Sagalyn and Faucher,
1954; Reiter, 1992].
[67] Multiple layers of aerosol distributed in altitude are
sometimes observed, especially after volcanic eruptions,
and generally multiple layers of clouds are common. J
z
,
will flow through such layers, producing space charge in
each. It is not necessary that the layers be distinct; vertical
gradients in ion production or in the concentration of
aerosol particles, or of droplets within the clouds, will also
result in space charge.
[68] The aerosol in the atmosphere in general is poly-
dispersive, and thus the effects described in section 4.5
would apply, with the larger aerosol particles tending to
charge positive, and the smaller ones negative, with the
distribution of charge between them depending on the
amount of space charge. Such differential charging, if large
enough, may affect the rates of coagulation of small aerosol
particles onto larger ones, with the effects depending on the
magnitude of J
z
.
[69] For ultrafine aerosol particles near the boundaries of
clouds, where there is a supply of volatiles from evaporating
droplets that can allow ultrafine particles to grow large
enough to act as cloud condensation nuclei, the presence of
space charge, depositing like charges on the ultrafine
particles and droplets, may act to preserve these particles
against scavenging on droplets. This protection and en-
hancement of the small CCN component, together with the
narrowing of the distribution due to the removal by electro-
scavenging of large CCN, would tend increase droplet
concentrations and reduce droplet radii in subsequent cycles
of cloud formation, reduce rates of coagulation and precip-
itation, and also increase cloud cover (the Twomey effect)
[see Tinsley, 2004].
[70] Where evaporation of charged cloud droplets is
occurring at the boundaries of clouds, the presence of space
charge not only generates the charge on the droplets before
evaporation, but serves to slow the decay of the charge on
the residual aerosol particle, beyond the initial decay time
constant of about 10 min. The final charge is determined
by the amount of space charge, and as noted earlier, could
be $5e for 1 mm particles. Thus the evaporation residue has
charges of tens of elementary charges for periods of 10 min,
and depending on radius, decaying down to several ele-
mentary charges for as long as the space charge persists.
These evaporation residues are considered to be efficient
ice-forming nuclei [Beard, 1992] and this characteristic,
along with the electrically enhanced scavenging, may con-
tribute to significant increases the rates of contact ice
nucleation [Tinsley et al., 2001].
5. Conclusions
[71] The variable flow of ionosphere-Earth current den-
sity J
z
generates variable amounts of space charge near the
upper and lower boundaries of clouds. We have modeled the
production of space charge and its partitioning between
droplets, aerosol particles, and atmospheric ions, for a range
of relevant atmospheric variables. The results give average
charges on droplets of 50 to 100 elementary charges;
positive at cloud top and negative at cloud base, which
supports the recent measurements of Beard et al. [2004].
The effects of moderate downdrafts at cloud top and
updrafts at cloud base may be necessary to provide a narrow
enough boundary layer together with a more concentrated
current density, to generate the observed charges.
[72] When taken with models of electrical effects on
cloud microphysics, the results are consistent with obser-
vations of meteorological responses to processes affecting
J
z
, and suggest that the accumulation of space charge in
layer clouds may affect cloud microphysics sufficiently to
cause measurable effects on weather and climate.
Appendix A: Variation in Ion Mobility With
Altitude
[73] The ions diffuse through the air and are accelerated
by the ambient electric field between collisions with air
molecules. We consider the case appropriate to the motion
of ions in the fair weather electric field, where the concen-
tration of ions is small compared to that of the air mole-
cules, and the energy gained from the electric field, between
collisions, is small compared to the thermal energy. The
ions gain an increment of energy between collisions, but
lose it as the collisions randomize and thermalize the ion
motion. The ion gains a drift velocity v small compared to
the mean thermal velocity. The mobility is m defined as
m vE A1
where E is the applied electric field. The increments of
velocity gained between collisions, which average to v,
depend both on the mean thermal velocity and on the mean
free path. In turn, the mean free path depends on the
collisional cross section for ion collisions with air
molecules, and thus on the effective size of ions and
molecules in the collisions, which vary with the relative
velocity between the ions and molecules, and thus on the
temperature. The mean free path depends inversely on the
concentration of air molecules. These factors are also
characteristic of diffusion of one species of gas through
another, and it was first shown by Townsend [1900], and
more recently by Chapman and Cowling [1970, section
19.12] and McDaniel and Mason [1973, section 5-1-A] that
m eD= kT A2
where D is the diffusion coefficient for ions through air, e is
the elementary charge, and k is Boltzmanns constant. It was
shown by Chapman and Cowling [1970, section 14.5] that
for diffusion of two gases through each other, where the
molecular cross sections vary similarly with collision
energy, the diffusion coefficient D and the coefficient of
viscosity h are related by
D const h=r A3
where r is the gas density, determined by pressure (P) and
temperature:
r P=T A4
[74] For molecules considered as rigid elastic spheres of
diameter s and mass m, Chapman and Cowling [1970,
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
15 of 17
D11203
section 12.1], using different symbols, give the first ap-
proximation to h as
h
5
16s
2
kmT
p
_ _1
2
A5
which gives
h
h
o

T
T
o
_ _1
2
A6
so that using (A2), (A3), and (A4), and for rigid elastic
spheres,
m
m
o

P
o
P
T
T
o
_ _1
2
A7
where the subscripts (o) refer to reference conditions, which
could be STP (T = 273.15 K, P = 1013.25 hPa) as used in
physics and chemistry, or T = 288.15 K, P = 1013.25 hPa,
as used in standard atmospheres. The diffusion coefficients
and viscosities are difficult to derive theoretically, as it is
necessary to allow for the variation of collision cross section
with collision energy. At a higher level of accuracy,
empirical determinations of D or h with temperature, fitted
by theoretical expressions, are used.
[75] An improvement on the rigid elastic spheres model
is Sutherlands model [Bircumshaw and Stott, 1929;
Montgomery, 1947; Chapman and Cowling, 1970, section
12.32], where
h
h
o

T
T
o
_ _3
2
T
o
S
T S
A8
where S is Sutherlands constant; an adjustable parameter to
ensure agreement with experiment. S is determined by
viscosity measurements, and depends on the species of gas
involved. Assuming that the cross section for ions in air
varies in a similar way with temperature as that of other
gases in air, we adopt the value for S for air, given by
Rogers and Yau [1989] of S = 120 K, consistent with the
value from Bircumshaw and Stott [1929]. Then, using (A2),
(A3), and (A4),
m
m
o

P
o
P
T
T
o
_ _3
2
T
o
S
T S
A9
[76] At this level of approximation, an alternative ap-
proach would be to use the empirically determined temper-
ature variation of the diffusion coefficient [Boynton and
Brattain, 1929]
D
D
o

P
o
P
T
T
o
_ _
n
A10
with the value of n varying with the species of gas. For
diffusion of water vapor in air the value of n = 1.81 has been
obtained [Montgomery, 1947; List, 1966]. The use of this
expression in (A2) gives
m
m
o

P
o
P
T
T
o
_ _
0:81
A11
which agrees to better than 1% with (A9) over the range
200320 K.
[77] Strictly speaking, at the highest level of accuracy a
more complex fit to empirical data for D should be more
appropriate than a fit to data for h, especially for very low
and very high temperatures [McDaniel, 1989, section 1.6].
However, for the temperature range considered here, ade-
quate accuracy is obtained by using (A9).
[78] A completely different approach to specifying the
variation of ion mobility with height is based on the concept
of reduced mobility, where a value for m is determined
experimentally, for specified conditions of pressure P and
temperature T, and reduced to STP, i.e., to a value of m
o
at STP, by means of the formula
m
o

P
P
o
T
o
T
m A12
[79] As discussed by McDaniel and Mason [1973, section
1.3], this corrects only for the variation of mobility with gas
concentration (number density), and is applicable for meas-
urements at a temperature close enough to STP that the
variation of mobility with temperature (at constant density)
can be ignored. The reversal of the use of this equation, to
give a value of m in terms of m
o
, P, and T; that is,
m
P
o
P
T
T
o
m
o
A13
would only be valid for a small range of temperature around
T
o
, as in work by Bricard [1965]. To use it for a theoretical
model of ion mobility throughout the troposphere, strato-
sphere, and lower mesosphere as was done by Shreve
[1970], is not appropriate.
[80] One could argue that the assumption implicit in
equation (A9), that the collision cross section of air ions
with air molecules behaves like that of other gases diffusing
through air, leads to uncertainties in the temperature varia-
tion such that there is little improvement in accuracy to be
gained by using (A9) rather than (A13). However, there are
additional problems with regard to the results of Shreve
[1970]. He made an exact calculation of mobility with
respect to altitude using (A13), together with the temper-
atures and pressures vs. altitude of the U.S. Standard
Atmosphere (1962), and values of positive and negative
mobility m
1
and m
2
at STP. These STP values were
from work by Bricard [1965] and were 1.4 10
4
m
2
V
1
s
1
and 1.9 10
4
m
2
V
1
s
1
, respectively. Shreve then
generated the fitted expressions, m
1
(z) = 1.4 10
4
exp
(0.14z) m
2
V
1
s
1
and m
2
(z) = 1.9 10
4
exp (0.14z) m
2
V
1
s
1
where z is in km, and is the altitude above the zero
at sea level in the standard atmosphere. These expressions
show that the values from Bricard were used by Shreve for
the temperature at zero altitude in the standard atmosphere
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
16 of 17
D11203
(288.15 K) rather than the STP temperature of 273.15 K,
and his designation as STP values is misleading.
[81] An additional concern is that Shreves expressions
for the altitude variation of mobility were fitted to the whole
altitude range from 0 to 60 or 70 km, whereas the exact
calculation departs significantly from the log linear fitted
curve near the 10 km region and the 50 km region.
[82] We have recalculated the height variations of the
mobilities m
1
(z) and m
2
(z) for the pressures and temperatures
of the U.S. Standard Atmosphere (1976) using the Bricard
[1965] values for m
1
and m
2
at zero altitude, and equation
(A9) for the variation. The linear fits to the curves are given
for three height ranges, to better fit the nonlinearity in the
curves of the calculated variation:
z < 10 km; m
1
z 1:40 10
4
exp 0:111z m
2
V
1
s
1
m
2
z 1:90 10
4
exp 0:111z m
2
V
1
s
1
10 < z < 50 km m
1
z 4:25 10
4
exp 0:154 z 10 m
2
V
1
s
1
m
2
z 5:76 10
4
exp 0:154 z 10 m
2
V
1
s
1
50 < z < 60 km m
1
z 2:01 10
4
exp 0:099 z 50 m
2
V
1
s
1
m
2
z 2:72 10
4
exp 0:099 z 50 m
2
V
1
s
1
[83] The above height variations are appropriate for use
with standard atmosphere models. They do not, however,
consider the changing composition and structure of the ions
as a function of absolute humidity [Fujioka et al., 1983] or
changes in the amounts of trace constituents such as HNO
3
,
NH
3
, and H
2
SO
4
, which can replace H
2
O molecules in the
cluster surrounding the ion. Such processes are involved in
the process of ion mediated nucleation [Yu and Turco,
2001], especially in the downward branch of the strato-
spheric Brewer-Dobson circulation, where the air can be-
come supersaturated in H
2
SO
4
vapor [Tinsley and Zhou,
2006].
[84] Acknowledgments. This work has been funded by grant ATM-
0242827 by the U.S. National Science Foundation. Limin Zhou has been
supported in part by a fellowship from the Chinese Academy of Sciences.
References
Beard, K. V. (1992), Ice initiation in warm based convective clouds: An
assessment of microphysical mechanisms, Atmos. Res., 28, 125152.
Beard, K. V., H. T. Ochs III, and C. H. Twohy (2004), Aircraft measure-
ments of high average charges on cloud drops in layer clouds, Geophys.
Res. Lett., 31, L14111, doi:10.1029/2004GL020465.
Bircumshaw, L. L., and V. H. Stott (1929), Viscosity of gases, in Interna-
tional Critical Tables, vol. 5, edited by E. W. Washburn, pp. 16,
McGraw-Hill, New York.
Boynton, W. P., and W. H. Brattain (1929), Interdiffusion of gases and va-
pors, in International Critical Tables, vol. 5, edited by E. W. Washburn, pp.
6263, McGraw-Hill, New York.
Bricard, J. (1965), Action of radioactivity and of pollution upon parameters
of atmospheric electricity, in Problems of Atmospheric and Space Elec-
tricity, edited by S. C. Coroniti, pp. 82117, Elsevier, Amsterdam.
Burns, G. B., B. A. Tinsley, A. V. Frank-Kamenetsky, and E. A. Bering
(2007), Interplanetary magnetic field and atmospheric electric circuit
influences on ground-level pressure at Vostok, J. Geophys. Res., 112,
D04103, doi:10.1029/2006JD007246.
Chapman, S., and T. G. Cowling (1970), The Mathematical Theory of Non-
Uniform Gases, 3rd ed., Cambridge Univ. Press, London.
Fujioka, N., Y. Tsunoda, A. Sugimura, and K. Arai (1983), Influence of
humidity on ion mobility with life time in atmospheric air, IEEE Trans.
Power Apparatus Syst., 102, 911917, doi:0018-9510/83/0002-0911.
Griffiths, R. F., J. Latham, and V. Myers (1974), The ionic conductivity of
electrified clouds, Q. J. R. Meteorol. Soc., 100, 181190.
Harrison, R. G., and K. S. Carslaw (2003), Ion-aerosol-cloud processes in
the lower atmosphere, Rev. Geophys., 41(3), 1012, doi:10.1029/
2002RG000114.
Hoppel, W. A. (1985), Ion-aerosol attachment coefficients, ion depletion,
and the charge distribution on aerosols, J. Geophys. Res., 90, 5917
5923.
Hoppel, W. A., and G. F. Frick (1986), Ion-aerosol attachment coefficients
and the steady-state charge distribution on aerosols in a bipolar ion en-
vironment, Aerosol Sci. Technol., 5, 121.
Keefe, D., P. J. Nolan, and J. A. Scott (1968), Influence of Coulomb and
image forces on combination in aerosols, Proc. R. Ir. Acad., Sect. A, 66,
1724.
List, R. J., (Ed.) (1966), Table 113, in Smithsonian Meteorological Tables,
6th ed., pp. 394395, Smithson. Inst., Washington, D. C.
Marsh, N., and H. Svensmark (2000), Low cloud properties influenced by
cosmic rays, Phys. Rev. Lett., 85, 50045007.
McDaniel, E. W. (1989), Atomic Collisions, John Wiley, New York.
McDaniel, E. W., and E. A. Mason (1973), The Mobility and Diffusion of
Ions in Gases, John Wiley, New York.
Montgomery, R. B. (1947), Viscosity and thermal conductivity of air and
diffusivity of water vapor in air, J. Meteorol., 4, 193196.
Pruppacher, H., and J. D. Klett (1997), Microphysics of Clouds and Pre-
cipitation, 2nd ed., Kluwer, Dordrecht, Netherlands.
Reiter, R. R. (1992), Phenomena in Atmospheric and Environmental Elec-
tricity, Elsevier, Amsterdam.
Rogers, R. R., and M. K. Yau (1989), A Short Course in Cloud Physics,
3rd ed., Pergamon, Oxford, U. K.
Sagalyn, R. C., and G. A. Faucher (1954), Aircraft investigation of the large
ion content and conductivity of the atmosphere, and their relation to
meteorological factors, J. Atmos. Terr. Phys., 5, 253272.
Shreve, E. L. (1970), Theoretical derivation of atmospheric ion concentra-
tions, conductivity, space charge density, electric field and generation rate
from 0 to 60 km, J. Atmos. Sci., 27, 11861194.
Tinsley, B. A. (2000), The influence of the solar wind on the global electric
circuit, and inferred effects on clouds and climate, Space Sci. Rev., 94,
231258.
Tinsley, B. A. (2004), Scavenging of condensation nuclei in clouds: De-
pendence of sign of electroscavenging effect on droplet and CCN sizes,
paper presented at 14th International Conference on Clouds and Precipi-
tation, Inst. of Atmos. Sci. and Clim., Bologna, Italy.
Tinsley, B. A., and L. Zhou (2006), Initial results of a global circuit model
with variable stratospheric and tropospheric aerosols, J. Geophys. Res.,
111, D16205, doi:10.1029/2005JD006988.
Tinsley, B. A., R. P. Rohrbaugh, M. Hei, and K. V. Beard (2000), Effects of
image charges on the scavenging of aerosol particles by cloud droplets
and on droplet charging and possible ice nucleation processes, J. Atmos.
Sci., 57, 21182134.
Tinsley, B. A., R. P. Rohrbaugh, and M. Hei (2001), Electroscavenging in
clouds with broad droplet size distributions and weak electrification,
Atmos. Res., 5960, 115135.
Tinsley, B. A., L. Zhou, and A. Plemmons (2006), Changes in scavenging
of particles by droplets due to weak electrification in clouds, Atmos. Res.,
79, 266295.
Townsend, J. S. (1900), The diffusion of ions into gases, Philos. Trans. R.
Soc. London, Ser. A, 193, 129158.
Yair, Y., and Z. Levin (1989), Charging of polydispersive aerosol particles
by attachment of atmospheric ions, J. Geophys. Res., 94, 13,08513,901.
Yu, F., and R. P. Turco (2001), From molecular clusters to nanoparticles:
The role of ambient ionization in tropospheric aerosol formation, J. Geo-
phys. Res., 106, 47974814.
Williams, E. R. (2005), Lightning and climate: A review, Atmos. Res., 76,
272287.

B. A. Tinsley, University of Texas at Dallas, Richardson, TX 75083-

0688, USA.
L. Zhou, Department of Geography, East China Normal University, 3663
North Zhongshan R. D., Shanghai, China.
D11203 ZHOU AND TINSLEY: SPACE CHARGE IN CLOUDS
17 of 17
D11203