1.

1 Aim
Verify Coulombs
deducing
how

Law by investigating the force between 2 charged spheres and

electric field strength relates to spark discharges.

1.2

Introduction

Tasks 1 & 2 of

this experiment revolves around Coulombs Law, relating

Fig 1. Setup Outline

electrostatic force

and 2 charges Q1 and

Q2:

4 0 r
QQ
|F|= 1 2

, assume we are using permittivity of air,

0 =8.8542 1012 F m1

since we are not using any other dielectric, |

r| being the distance between 2 point charges (in this case, they are spheres,
but they have to be approximated as point charges in other to utilise this
equation).
Task 1 (F against V (potential difference)): Assume electric potential on
each sphere is of magnitude |V/2| (Fig 1). Given that we are using steel
conducting spheres, we can approximate each sphere being a point charge
in the centre of sphere. The potential on the surface of each sphere would be

V
Q
Fig 2. Change in Charge Distribution
=
2
4 0 R , R being radius of each sphere (given to be 20mm).

We obtain the equation of the right hand side by:

a=

d
dr , being electric potential,

acceleration (similar electrostatic force and electric potential energy relationship). We then get

being electrostatic

Q= 2 0 R V

as the charge of each spheres. Substitute this into Coulombs equation:

distances constant, we need not be concerned with

r 2
for this task.

r
2
2
0 R V . Since we keep the
|F|=

|F| V 2

Task 2 (F against distance): Distance is less than 5mm for our experiment, which is small relative to diameter of
sphere, 40mm. Since these are conducting spheres, charges can move freely about the sphere, thus, by having such
small separation between the spheres, opposite charges in the spheres will approach each other (Fig 2), though they
will still maintain at the surface as those are their equilibrium positions. Approximating them as point charges, these
points will move away from their respective centres and towards each other as they approach. This results in the point
charges being around the edge of the spheres that is the closest to the other sphere. This means the distance between
them, |r|, will be about r in Fig 1.
2

We then get:

|F|=

0 R V
r

(Note:

Q= 2 0 R V

change.).

|F|

still holds because the amount of charge didnt

1
2
r .

Task 3 (r for sparking): Similar to work done in a circuit, work done by electric field on a charged particle is the
distance the particle is being moved by the electric field. Here, we assume the electric field () between the spheres is
constant throughout r, since r is significantly smaller than the radius of spheres.
charge, E is the work done by E-field (being converted to kinetic energy, KE) and

E=q e l ,
l

qe

is electron

is distance electron travels

before colliding with another electron. We simplify sparking as a chain reaction of 1 electron to another. A free
electron on the conductor (metal has free electrons) accelerates continuously due to the E-field, gains enough KE in

the process, and collide with another atom (gas atom), ionising it by knocking out 1 of its electrons, continuing the
chain reaction. The ionisation potential (U I) of air, majority being nitrogen gas (70%), is 15eV. E> U I must be
satisfied for the chain reaction/sparking to occur. So,

electron =

UI
l , we estimate the space occupied of an
q e l . To find

( a2 ) l 1

a being the Bohr radius ( 5 1011 m , 1/n being number of gas molecules per unit
n ,

volume (N/V). Ideal gas equation: PV=Nk BT, so V/N= kBT/P=n. We then get

k T
1
= B 2
2
n ( a ) P ( a ) , substitute this

U I P ( a2 ) U I

=
into
.
qe l
k B T qe

1.3 Apparatus

2 Ruler (15cm)
1 Micrometre drive
2 Metal balls
Plastic tighteners
1 High Voltage source
2 Banana plugs
Spirit level
Track
(0-6kV)

2 Stages (with holders, one

also with Micrometre drive)
Force sensor, connected to PC

1.4
Measured
Variables

Constant
Independent
Variables:
Variables:

( 40 ) kV
1 ) mm
Diameter
Voltage, V:
of ball:( 0.05

1.5
& Precaution

20
Distance
Turns

Dependent Variables:
Electrostatic force, F: (

Force Sensor

between
of
micrometre
balls,

Methodology

Metal Balls
Connected High voltage Source

Connect to computer
Track

Fig 3. Experimental Setup

1. Set up experiment as shown in Fig 3 using the given components.
2. Use banana plugs to connect both stages to High Voltage source, ensure Force sensor is connected to PC (by
checking if the PC registers any change in values as you change the potential difference between the balls,
controlled by the Voltage source).
3. Ensure track is level using a spirit level (important for ensuring balls vertical alignment in the next step).
4. After tightening the screw supporting the balls, place spirit level on the balls to ensure they are at the same height
(always ensure balls alignment after tightening the screw).
5. Ensure balls are horizontally aligned by palming 2 rulers flat on the side of the balls, then tighten the rulers with
tighteners placed at the centre of rulers. Once it is tight enough, the balls will align horizontally. Remove the
combination by pulling it upwards slowly.
6. Task 1: Vary V (from 0-6kV) by intervals of 0.2kV, wait for the Force registered on the PC to stabilise around a
small range of values, record. Plot a graph of F against V using the data obtained.
7. Task 2: i) Set a particular value of V first. Vary r by turning the micrometre drive. Wait for the Force registered
on the PC to stabilise around a small range of values, record. Plot F against r. ii) Vary V value (1-6kV, with
intervals of 1kV), and repeat the previous step. There should be a total of 6 graphs.
8. Task 3: Set V to 0kV, and r to around 1mm. Gradually increase V until sparking first occurs. Repeat this for
increasing r in intervals of 0.1mm, until r is too great for sparking to occur at maximum V. Graph V vs r.

2.1 Results & Calculation

Calculations of uncertainties (electric field strength, , will be calculated in Discussion, as it depends on data):

drive:
r:

V uncertainty = 0.05kV, due to rounding error of the value displayed on Voltage source, which is in 2 d.p.
Length of 20 micrometre turns = (101) mm since we are reading from 2 points on the scale beside the track.
Length of 1 micrometre turn = (0.50.05) mm. However, we are not using this uncertainty (0.05mm) because
when we measure, we measure directly from the micrometre reading instead of the scale beside the track.
r = 0.0025mm because we are reading 1 point, so it is one half the smallest increment (0.005mm, every turn
has 100 increments 0.5/100 = 0.005mm per increment) of the micrometre.
F depends on the fluctuation displayed on computer, hence it has no guidelines and is seemingly random.

Graphs:

Fig 6. Task 2, F vs r at 2kV

Fig 4. Task 1, F vs V

Fig 8. Task 2, F vs r at 4kV

Fig 10. Task 2, F vs r at 6kV

Fig 7. Task 2, F vs r at 3kV

Fig 5. Task 2, F vs r at 1kV

Fig 9. Task 2, F vs r at 5kV

Fig 11. Task 3, V vs r for sparkle

(For clearer graphs and tabulated data, please refer to Appendix 2.1.)

2.2 Discussion
Obtaining F values
F is random, depending on the fluctuations of the F values displayed on computer. We obtained F by observing the
highest and lowest values displayed, average the values, find the difference of this value from both sides, it should be
the same on either side. This difference will be F. F will be the y-error bars for Fig 4-10.
In fact, all error bars will be the uncertainties of the values in the respective axes.
Coefficient of Determination
Notice all graphs (Fig 4-11) are being fitted according to
their appropriate function curves. In it, there is a table
that has a value labelled Adj R-Square, which is useful
for determining how closely our data points follow the
theoretical relationship.
This value is called Coefficient of Determination, R 2.
Refer to Fig 12, let f be the function curve used for
fitting of the given data points.

of all y-values of the data points,

th

value of i particle. R is defined by:

being the average

yi

Fig 12. Coefficient of Determination

being the y-

R2 1

SS res
SStot .
2

SSres is residual sum of squares,

SS res = ( y if i)
i

, basically the deviation of every points y-value from

functions y-value at for the same x-value.

2

SStot is total sum of squares,

SS tot = ( y i y )
i

, basically gives the spread of y-values of the points across the

graph. Since we record a range of values, SS tot should be non-zero.

This suggests the closer the R2 value is to 1, the less deviation there is of the data points from function curve.
I will be using R2 values as part of the discussion element.
Task 1:
2

As mentioned in Introduction, the relation between F and V is

|F| V

r
2
2
R V ). Therefore, we
(
|F|= 0

did a parabolic curve fitting of the data points, Fig 4.

The R2 value is 0.99, which suggests the relation that we obtained theoretically closely resembles what we got
empirically. This is expected and relatively straightforward because we need not consider

r , the distance

between 2 approximated point charges, yet. All that we are concerned about is how V contributes to F value.
Task 2:
The graphs we plotted here (Fig 5-10) are F vs r graphs. From Introduction, |F|=

Logically, the fitting we used is reciprocal squared function

y=

1
0 R 2 V 2
|F| 2
2
,
r .
r

k
+A
2
.
x

The R2 value is 0.99 on average, which suggests the relation we obtained theoretically closely resembles the data we
got. This holds because the r is relatively small compared to ball diameter as well as the value of potential difference.
Due to this, excess charges are concentrated at a region nearer to the other ball due to electrostatic attraction forces,
hence I approximated the point charge to be at the point nearest to the adjacent ball.

If r is significant compared to ball diameter and V, the approximated point charge will be more centred, and

|F|

1
2
r

will not be as accurate an approximation as the present case, since

now cannot be

approximated as r only. One would then see greater deviation from the function curve, hence lower R 2.
Task 3:
From Introduction, the dielectric breakdown field

P ( a2 ) U I
=
. This will be our theoretical expectation value. Assume room temperature and
k B T qe

sparkling to occur,
pressure,

U I P ( a2 ) U I

=
, if we obtain a distance just enough for
qe l
k B T qe

P=101325
6

Pa,
1

Theor . 2.90 10 V m

UI=15eV,

a=0.5*10-10m,

qe=1.6022*10-19C,

kB=1.3806*10-23JK-1,

Since r is small relative to ball diameter, we approximate the system as parallel charged plates,
uncertainty of ,

T=298K,

V
r . Then,

( V ) ( r )
( V ) (r)
=
+
= (
+
) . Since ( V ) ( r ) are constants, the

V
r
V
r

greater the V and r, the smaller the , since we assume would not change much.
We can now calculate the percentage error of our data:

V
r
(0.05)kV (0.0025)mm
3.10
1.0000
3.30
1.1000
3.70
1.2000
4.40
1.3000
4.70
1.4000
5.00
1.5000
5.60
1.6000
5.70 Support1.7000

-1

/MVm

3.100.06
3.000.05
3.080.05
3.380.04
3.360.04
3.330.04
3.500.04
3.350.03

Fig 13. Improved mechanism

Percentage
Error
6.89
3.45
6.32
16.71
15.76
14.94
20.69
15.62

| Expt Theor|

Percentage Error=

Theor

100 .

From Table 1, the average is 3.26 MVm-1, with percentage

error 12.5%. This is higher than 10%, suggesting that there
could be a disparity in the room conditions being assumed,
and also errors which will be discussed under Errors &
Improvements.
In Fig 11, we used linear fitting, because we expect to be
same, then, V= (r). Ideally, the line should intercept the
Origin, with a gradient of (theoretically, 2.90 MVm-1). But
we got an intercept of y=-1.01kV and gradient of 4.04 MVm1

Table 1. Calculation of and % error

instead, which again suggests that there are errors (random and systematic) that significantly affect the readings. As
expected, decreases as V and r increase.

3.1 Errors & Improvements

Random errors in general (precision):
Balls orientation when not touching Even though we have ensured the balls are horizontally and vertically aligned
(steps 4-5 under Methodology & Precaution), we realised external vibrations were able to cause the balls to change
orientation and misalign. This results in random errors because sometimes we were unable to spot any changes done to
the alignment due to vibrations (they may be too minor to us, but may cause significant orientation changes). This
results in different r values, which affects the results of the data for all 3 Tasks.
One way to minimise this error is instead of using screws to secure the balls, we can have them mechanically attached
with extra supports to the stages instead. With the same mechanism used for securing both balls, they will be

horizontally aligned since they should be symmetrical. Also, there is no need for tightening (less inconsistency, the
more you tighten, the more the ball tilts) and the support will ensure the balls are consistently vertically aligned.
However, as time goes by, deterioration will occur, and the balls will not be as secured, external vibrations would still
affect the orientation. The difference is that there is no need for securing the screw, making it more consistent.
Balls alignment when touching To make sure r values (observed on micrometre drive) directly reflects the
distance between the balls, we need to turn the knob to 0, then bring the balls together such that touch each other. At
the same time, ensuring that they align (steps 4-5 under Methodology & Precaution). Thereafter, turn the knob, and the
distance displayed on the micrometre will be r. 2 problems arise:
i)
ii)

As we shift the stages, the vibration will knock the balls out of alignment, and as they approach, we are
not sure if they would just touch nicely when we were to align them.
Even if they are seemingly aligned, as we move them closer, we cannot judge with our naked eyes if they
touch or not. If they were to be slightly misaligned, as they touch, they would slide against each other
(because they are smooth and spherically shaped, so sliding is possible) even more misaligned.

2 ways to improve this: i) lubricate the track so that shifting the stages will not be so jerky. ii) Shift the stage such that
the balls almost touch (leaving a tiny space in between), then turn the knob so that they just touch. This way, it is
much easier to control the system and will reduce the random errors caused by the 2 issues
above.
Pivot
Large fluctuations of F values The y-axis for Fig 4-10 are the electrostatic force (F)
r
recorded by the force sensor. Notice that the y-error bars are large relatively to the scale of the
graph. This is because we observed large fluctuations in the values displayed on the computer. Fig 14. Torque on ball
We can attribute this the high sensitivity of the Force sensor. But also, notice that the ball
which the sensor measures the force it experiences is quite far (about 13cm) from the pivot (relative to the width of
force sensor, about 4cm) that is holding it upright. We know that in such a system, the torque on ball

=fr , r being

distance from balls centre of mass to pivot (Fig 14.), f being any force exerted on the ball (usually vibrations) that is
perpendicular to r, for there to be a torque. In order to counter this torque that would otherwise cause this ball to
move around, there must be a force exerted somewhere within the force sensor. The perpendicular distance of this
force from the pivot would be a fraction of r, and thus this force must a few times greater than f. Although this is
hypothetical, it would make sense that the greater the r, the more the force measured would be magnified, since the
force is sensed somewhere within the sensor.
Some ways to minimise this error is shorten the arm, that way, force will be magnified less. Also, isolate the
experiment in a secluded room, so that there will be less external vibrations, such as people walking around and
touching the table the setup is on. All these improvements are meant to reduce the fluctuations of F values, for higher
precision.
Backlash Given that the micrometre uses a gear mechanism, backlash would contribute to the random error that we
got. This is apparent when we turn the knob in opposite directions as before, and if there are gaps between the gears,
the knob would turn but the ball will not move (actual r unchanged). This causes the r measured to be greater than
what it actually is. This will affect the results of Tasks 2 & 3 as they involve changing r.
Backlash arises due to deterioration of gears. Hence, an improvement would be to switch the micrometre drive to a
newer one. But this is only temporary, because in time, deterioration would happen and backlash would happen again.
F value takes a long time to register We noticed that as we changed the independent variables, it took a long time
(at least 30 seconds) for the Force to stabilise around a range of values. Coupled with the aforementioned error of F
having large fluctuations, even if we were to wait (steps 6&7 under Methodology & Precautions), we were still unsure
if F is already stable. Besides, recording the highest and lowest F seen on screen and averaging them is highly
subjective because we cannot be sure when will F be highest/lowest. This cause random errors in Tasks 1&2 (Task 3
has no F).
To minimise fluctuations, the suggestions are aforementioned. To obtain F values properly, we should use a program
that can average the F values for us instead. That way, it is less subjective, F values being more certain, reducing
random errors.
Systematic errors (accuracy):
Faulty voltage source The voltage source does not provide voltage that is being displayed. Using a multimeter to
measure the actual voltage output when 1kV is displayed, only 820V was recorded. Further measurements of varying
voltages confirmed that the actual voltage is consistently lower than the displayed value. This contributes to

systematic errors and is a contributing factor to the high percentage errors obtained for Task 3, since the experimental
is always greater than the theoretical , which is because V is actually smaller than displayed (= V/r).
We can improve this by switching the source to a newer, less faulty one. That way, it is possible that the Task 3
percentage errors would then be smaller because the V registered would be lower, and closer to the theoretical ,
leading to greater accuracy.
Insufficient time to recharge after sparkle Upon sparkling, excess charges on each ball are reduced significantly.
They then require time to recharge for sparkle to happen again. Although at close enough distances, continuous
sparkle happens, we measure the distance where sparkle first occurs, which is not continuous as excess charges in
each sphere are arguably at maximum for the given voltage across the spheres, and after the first sparkle, they reduce,
between them is not high enough to cause another long distant sparkle. By right, we are supposed to give them
enough time to recharge, but we are unsure how long they need. So we might not have given them enough time to
recharge to maximum charge and proceeded with the rest of the readings. This will cause higher V measured,
because we increased voltage if they were not sparkling. Thus, this plays a part in the large percentage error we got for
Task 3.
If we waited long enough (30 seconds after every reading), a lower V would be required to sparkle for a given r.
This would reduce the high percentage errors we obtained.
Errors in approximations made:
approximation error Under Discussion, we used the approximation

V
r , suggesting we treated the balls

as parallel conducting plates. This, however, is not an accurate approximation because they are spherical, there is only
1 point that is the closest to the other ball (unlike parallel plates, infinitely many). The point charge we approximated
for each sphere when they are close are at that point (closest to neighbouring sphere). If charges are concentrated at 1
point, approximating the spheres as parallel plates would be inaccurate. Besides, we observed that the sparkling bolts
arc do not necessarily connect both closest points. This suggests that electrons gain less KE (

E=q e l ) than

expected because is weaker than expected. Thus, we need to be higher in order to sparkle, which is why
experimental is higher than theoretical, because higher V is required.
To improve on this, instead of using metal spheres, we should use parallel conducting plates, which are the basis for

the approximation of
r . This would be more accurate and hence percentage error would decrease.

4.1 Conclusion
From the results obtained in Tasks 1& 2,

|F| V 2 is an accurate expression ( R2 1 , not much scatter) for F

& V relationship. At relatively close distances,

|F|

1
r2

is also accurate ( R 1 , not much scatter). But at

further distances, it will be less accurate. Hence, we are able to verify Coulombs Law since the equation is valid for
this experiment.
From Task 3, it does not seem that

V
r

is an accurate way to finding dielectric breakdown field of air, since

the percentage error is on average, above 10%.

4.3 Questions and Answers

1) Does the position of the analyser affect the results of the experiment?
Based on the results of Task 1, we were able to verify that

|F| V 2 , which in turn verifies the estimation of Eq.

(3). However, that equation does not estimate the relationship between F and r, which we found in Task 2 to be

|F|

1
r2 . Both relationship estimates are relatively accurate, with coefficients of determination close to 1.

2) If the coils and light source are turned 90o, how will the results change? Describe the changes, if any.

V varies linearly with r (Fig 11). Our experimental breakdown field is deviated from the approximate theoretical
value (with average percentage error of 12.5%). The approximated value is around 2.90MVm -1, but the experimental
values we got are on average, 3.26 MVm-1. The reasons why percentage errors are so high are mentioned previously
under Errors & Improvements.
3) What are the safety hazards associated with this experiment, and what are the safety precautions that one should
take in relation to these hazerds?
No. The charge and voltage in both cases are very different due to the scale of the system. The voltage source
produces high potential difference relative to r, hence continuous sparkling is possible. But, distance between clouds
and land is much greater than r. Clouds need way more time to gather enough charges before causing dielectric
breakdown of air due to the immense potential difference between clouds and ground for the electric field strength
between them to equate the dielectric breakdown of air. Thus, thundercloud cannot continuously release lightning
bolts.
However, the electric field strength for both cases are about the same because they are just the dielectric breakdown
field of air.

References
http://www.phys.ufl.edu/~acosta/phy2061/lectures/MagneticDipoles.pdf

Appendix
2.1

Task 1, F vs V

Task 2, F vs r at 1kV

Task 2, F vs r at 2kV

Task 2, F vs r at 3kV

Task 2, F vs r at 4kV

Task 2, F vs r at 5kV

Task 2, F vs r at 6kV

Fig 11. Task 3, V vs r for sparkle