The effect of the dynamic friction coefficient of a keycap’s material

on its required pull force

Introduction

As an avid computer hobbyist, I take pride in exploring the

niche sub-communities of this vast hobby, such as the
mechanical keyboard community. This community takes
keyboards up one level with their unique customizability and
experience. A major customizability factor in a mechanical
keyboard is its keycaps, the plastic keys sitting on top of the
keyboard switch that your fingers press down to make an
input. On a mechanical keyboard, these are meant to be
removed so they can be replaced with a new design or to
clean the insides of the keyboard. While I often remove these
keycaps to clean my keyboard, I notice how it takes quite a
bit of force to remove them, even using a keycap puller tool. Figure 1. Keycap being removed
from a keyboard using a keycap
It is a tedious task, and so I wondered if I owned a different puller.
set of keycaps made from completely different materials,
would they be easier to remove? Keycaps are made of different materials, such as PBT plastic,
which has a dynamic friction coefficient, the force between two objects when one object is
moving. In this experiment, I will be focusing on the amount of force required to remove a
keycap from a keyboard, with keycaps of different dynamic friction coefficients.

My research question is as follows: How does the dynamic coefficient of friction of the material
of a keyboard keycap affect the pull force (in N) required to remove the keycap from a keyboard
switch?

To predict our result, I formulated an equation using the formula for finding the force of friction.
To begin, I need to find the force acting on the keycap due to gravity, which is mg. However, due
to m being incredibly small, this value can be made negligible. Next, I need to find the force of
friction between the keycap and the keyboard switch, which is going to be our pull force. This
can be derived from the formula for frictional force, F = μ ⋅ N . Starting with the dynamic
friction coefficient, our independent variable, multiplied by the normal force applied from the
switch stem inside the keycap stem, over the area of contact of the switch stem, which is going to

1 of 9
be a partial surface area of the switch stem. The base of the switch stem is shaped in the shape of
a plus sign, and so the sides could be theoretically unfolded to form a rectangular prism. As the
top and the bottom of the switch stem do not make any sliding contact with keycap stem, the top
and bottom sides of the prism can be omitted. This creates the formula of the surface area of a
rectangular prism without two of its sides.

Figure 2. Image of a keyboard switch, Figure 3. Image of the underside of a

showing the switch stem. keycap, showing the keycap stem.

Putting everything together, this all equals our dependent variable, the pull force. The derived
equation is as follows:

Fpull = μk ⋅ FN ⋅ 2(hl + hw)

To find the values of our controlled variables, I must measure them using measuring tools,
specifically a Vernier Go Direct Force Gauge (specifications specified in Materials section) to
find the normal force and a 15cm ± 0.05cm ruler to measure the dimensions of the contact area
between the switch and the keycap. Starting with finding the normal force, I turned on the force
gauge and connected it to my computer, followed a simple measuring procedure shown in Figure
4, and returned a value of 6.89N.

2 of 9
Figure 4. Diagram of the measurement of the FN Figure 5. Measurements of the area of
control variable contact of the switch stem

The value given by the force gauge is the force applied by the keycap stem onto the switch stem
per cm. If I want to find the force per m2, I need to divide the given value by 0.01 to convert to
m, and square it to get the force over an area in m. To find the values of h, l and w, I use my ruler
and measure the length, width and height of the switch stem. These values were converted to
meters as well, dividing the values by 0.01. With the values obtained, they can be substituted into
my equation:

6.89 2 0.40 0.40 0.40 0.40

Fpull = μk ⋅ ( ) ⋅ 2( ⋅ + ⋅ )
0.01 0.01 0.01 0.01 0.01

Fpull = μk ⋅ 474721 ⋅ (6.40 ⋅ 10−5)

Fpull = μk ⋅ 30.38

Materials

The following materials are needed for the experiment:

• Up arrow keycaps of materials ABS, PBT and Wood

• Mechanical keyboard with Gateron Blue switches installed
• Vernier Go Direct Force Gauge (50N ± 0.01N) [1]
• C-clamp to prevent the keyboard from moving during the keycap pulling process
• Computer with Vernier Graphical Analysis installed

3 of 9
Risks

I must address the following risks during the duration of the experiment to avoid unexpected
events from occurring:

• The keyboard should be clamped tightly to avoid sudden movements that may cause
injuries.
• Tests should be performed in an uncrowded environment as keycaps may fly off the
keyboard once removed, with the chance of possibly hurting someone.
• As keycap removals end with a sudden burst of force, there is a chance I could accidentally
let go of the force gauge, thus possibly damaging it or hurting an individual.

Experiment Procedure

The experiment was conducted using the following methodology:

1. Turn on force gauge, connect it to the computer via Bluetooth.

2. Clamp the keyboard down to a stable surface using a C-clamp.
3. Place the keycap on the up arrow switch.
4. Use the force gauge to hook the keycap from the bottom, begin recording values.
5. Pull the keycap until the keycap has been removed from the switch, record the peak force
value attained.
6. Repeat steps 3-5 a total of five times for each keycap.

Figure 6. Image of the experiment being done

4 of 9
Raw Data

Dynamic Coefficient Pull Force (±0.01N)

of Friction
Trial 1 Trial 2 Trial 3 Trial 4 Trial 5
0.24 [2] 5.54 5.62 5.72 5.29 5.70
0.32 [3] 9.52 9.86 9.74 9.70 9.94
0.35 [4] 10.95 10.74 11.04 10.64 11.68

Table 1. Raw data table (2 decimal places)

Analysis & Sample Calculations

With multiple trials done, the average value was found using the formula for the mean value,
which is given as:

ΣForce of each Trial

X=
Number of Trials

Sample calculation for coefficient of 0.24:

5.54 + 5.62 + 5.72 + 5.29 + 5.70

5.57 =
5

To find the uncertainty of the pull force, I took the largest value, subtracted it with the smallest
value, and halved it. This is done with the following formula [5]:

Largest Value − Smallest Value

Absolute Uncertainty =
2

5 of 9
Sample calculations for slope and y-intercept:

68.18 − 36.53
15.83 =
2

−2.64 − (−11.66)
4.51 =
2

To justify the uncertainty for the independent variable, as they are obtained from reputable
sources online, I cannot procure an accurate uncertainty. Thus, I used the smallest decimal value
of the coefficients, which is 0.01, as the uncertainty.

For the graph, the coefficient of determination, or R 2 , was found by squaring the coefficient of
correlation given by Logger Pro. The slopes and y-intercepts were also similarly found using the
lines on Logger Pro.

Processed Data

Dynamic Coefficient Uncertainty of Average Pull Force Uncertainty of Pull

of Friction Dynamic Coefficient (N) Force
of Friction
0.24 0.01 5.57 0.22
0.32 0.01 9.75 0.21
0.35 0.01 11.01 0.52

Table 2. Processed data table including average values and uncertainties (2 decimal places)

6 of 9
Figure 7. Dynamic Coefficient of Friction vs Pull Force (N)
Line of Best Fit: y = (50.03x ± 15.83x) − (6.39 ± 4.51) (R 2 = 0.998)
Maximum Line: y = 68.18x − 11.66
Minimum Line: y = 36.53x − 2.64

7 of 9
Conclusion

To conclude, as the dynamic coefficient of friction of the keycap’s material increases, the pull
force required to remove the keycap from the keyboard increases, as hypothesized. The
correlation in the data has an accuracy of 99.8%, meaning the correlation matches the averaged
points almost perfectly.

Comparing the values from my original equation and from the equation of the line of best fit, we
get these values:

10.63 = 0.35 ⋅ 30.38

11.12 = (50.03 ⋅ 0.35) − 6.39

Considering the uncertainties of the line of best fit, these values are relatively similar and both
follow a linear trend. These results solidify the relationship between the independent variable and
dependent variable.

There were a few issues throughout this experiment. As the independent variable was found
through sources from the internet, I cannot confirm which tool they used or the method they used
to find these coefficients. I could not come up with a valid uncertainty that would work, due to
these coefficients being calculated from multiple different sources, with their method unknown.
There is no way for sure to know that these values are exactly the coefficients of the keycaps, as
there may be variations in the exact material used, such as the difference between fiber-filled
PBT and unreinforced PBT.

Wear and tear can also be a limitation during the procedure, as the number of times I have placed
and removed a keycap can affect the wear of both the keycap stem and the switch stem, reducing
the amount of normal force and creating errors in the relationship.

It would have been best to buy brand-new material for peak performance. Additionally, this
would also allow me to have more independent variables to test with, as I only had three keycaps
at my disposal for the experiment.

Overall, this investigation allowed me to well utilize my knowledge in the hobby of keyboards
with an experiment that provides me with valuable information for something I do very often.

8 of 9
Bibliography

[1] Go direct® force and acceleration sensor. Vernier. (n.d.). https://www.vernier.com/

product/go-direct-force-and-acceleration-sensor/

[2] Slavkina, Viktoria & Goncharova, Yuliya. (2020). Study of tribotechnical

characteristics of 3D printed abs plastic samples. IOP Conference Series: Materials Science and
Engineering. 963. 012032. 10.1088/1757-899X/963/1/012032.

[3] The Engineering ToolBox Friction - Friction Coefficients and Calculator. (2004).
https://www.engineeringtoolbox.com/friction-coefficients-d_778.html

[4] The Online Materials Information Resource. MatWeb. (n.d.). https://

www.matweb.com/search/datasheettext.aspx?matguid=781bd0e9d0854fd5a919866c39ea3065

[5] Irfansyahril. (2019, July 6). PHY C2: Errors & uncertainties. ProDuckThieves.
https://produckthieves.home.blog/2019/01/27/phy-c2-errors-uncertainties/

9 of 9
10 of 9