A Quantitative Investigation on the Average Shadow Area of a Cube

Mathematics: Analysis and Approaches HL

Aim: To investigate the average shadow area of a cube with three degrees of

rotational freedom

Page Count: 19 Pages

 I INTRODUCTION

1 Rationale

"I like my shadow; it reminds me that I exist" (Ildan, n.d.). I find Mehmet Murat

Ildan's quote fascinating because of how it captures us identifying ourselves in

colourless, two-dimensional areas. Beyond our own, shadows are one of the most

interesting phenomena. They are natural two-dimensional images in a three-

dimensional world with no texture, character, or colour and can be cleverly

manipulated by their object's position relative to lights. Nevertheless, they always

appear to retain some semblance of their respective object. Noticing these

properties, I wondered whether shadows, whilst having an artistic simplicity, would

be mathematically complex. To further this exploration, I considered a cube; one of

the simplest shapes (excluding trivial shapes such as a sphere). While I initially

thought the shadow would be rather mundane, I was amazed by the various

shadows created when I informally experimented with my Rubix cube's shadow

created by my desk light. This investigation is a formalised mathematical extension

of this interest, which will explore the mathematical relationships governing the

shadows of cubes, bringing to light the intricacies of an everyday phenomenon.

2 Aim

This investigation aims to explore the average shadow area with three

degrees of rotational freedom. Intuitively, it is clear that the magnitude of a cube's

side length affects its average shadow area. Hence, this investigation will achieve its

aim by culminating with a mathematical relationship between the average shadow

area for a cube and the cube's side length. To further this investigation, crucial

assumptions should be made. The light source creating the shadow will be assumed

1
to be a point light source, and the displacement vector compromising the light source

and the centre of the cube will be perpendicular to the plane of the shadow. As a

result, the light source will always be directly above the cube and the shadow

throughout this investigation. Notably, we will also assume that the point light source

is infinitely distanced from the cube to facilitate the exploration.

II MATHEMATICS

Firstly, we must introduce the mathematical problem formally. The cube will

be positioned above a plane at height h (the distance from the plane to the centre of

the cube), with a point light source directly above it but infinitely far away, as

demonstrated by figure 1.

Figure 1: The Cube is Positioned Above a Plane with a Light Source Directly Above

As the light source approaches an infinite distance away from the plane, the

light rays will all approach a 90° angle to the plane.

We can prove this by considering a right-angled triangle ABC, where A is the

point light source, and B and C are points of intersection between the plane and two

consecutive rays of light from the point light source like the one presented in figure 2.

2
 Figure 2: Consecutive light rays from the light source intersecting the plane

As B and C are consecutive light rays, they are infinitesimally close to each

other. However, all that's important is that it is a finite distance. Furthermore, the

point light source is infinitely far away from the plane:

AB=∞ units

AC=∞ units

BC=x units , 0< x < ∞

We can use trigonometry to find angle a:

AB
tan a=
BC

a=arctan (BC
AB
)
Although the point light source is infinitely distanced from the plane, it is more

appropriate to evaluate this equation considering the length AB approaching infinity.

I will explain this decision retrospectively, proceeding with the angle's calculation.

a= lim arctan
AB → ∞
( BC
AB
)
π
∴ a=
2

3
 While I could have substituted the distance for infinity, I used limit notation to

consider the angle a as AB approaches infinity because a right-angled triangle

cannot have two right angles (as the third angle must be non-zero). Hence, this

conclusion proves that as AB approaches infinity, the angles between subsequent

π
light rays and the plane will approach . However, it will never be this exactly.
2

Moreover, the cube will have three degrees of rotational freedom as described

in the aim. Hence, the cube cannot be translated (moved), but it can swivel left and

right (yawing), tilt forward and backward (pitching), and pivot from side to side

(rolling), as seen in figure 3.

Figure 3: The three degrees of rotational freedom for a cube

Initially, I found that the three-dimensionality of the cube made it significantly

harder to mathematically model with three degrees of rotational freedom. As a result,

I have decided to compartmentalise the problem. To begin with, I will only consider

one face of the cube. However, for now, I have decided to compartmentalise this

further by investigating only the cross-section or a 'slice' of this single face,

represented in figure 4.

4
 Figure 4: The cross-section of a single face and its created shadow.

Interestingly, as the light rays are all perpendicular to the plane, the created

shadow is not dependent on the height of the face as seen in figure 4. However, to

maintain a more realistic model, we must ensure that the face (and cube, for that

matter) stays above the plane, otherwise, the parts below the plane will not have a

registered shadow.

As the cross-section of the face can rotate, I found that we could describe its

rotation mathematically by comparing its normal vector to the shadow's normal

vector (which will not change). We can achieve this by ascribing the angle θ as the

angle between the face's cross section's normal and the shadow's normal (identical

to the plane's normal as the shadow is on the plane), as represented in figure 5.

5
 Figure 5: The Angle Between the Shadow's Normal and the Face's Normal

Here, we can use basic geometric manipulation in figure 6 to find that the

angular relationship between the shadow and the face of the cube is also θ.

Figure 6: Graphical proof of the angle between the plane and shadow

As by definition, AC∨¿ GD, we can use transversal angle rules first to

^ B by using the corresponding angle rule:

determine the supplement angle of G A

^ B=θ+ 90°
Supplement angle of G A

^ B to determine G A
Subsequently, we can use the supplement angle of G A ^B

itself:

^ B=180 ° −(θ+ 90° )

GA

^ B=90 °−θ
GA

^ B and
As Δ ABF is a triangle where two angles are known in terms of θ (G A

^ ), we can use the rule for the sum of interior angles in a triangle to determine
A BF

A F^ B in terms of θ . This is important because A F^ B is the angle between the shadow

and the face of the cube, which will allow for a relationship between these two

variables to be developed.

6
 ^ B−A B
A F^ B=180 °−G A ^F

A F^ B=180 °−(90 °−θ)−90 °

A F^ B=θ

Consequently, we have found that the angle between the shadow and the face of the

cube is also θ.

Hence, using trigonometry, a relationship between the length of the face

(represented by AF in figure 6) and the length of the shadow (represented by BF as

BF=CD) can be deduced:

l shadow
cos θ=
l cross-section

l shadow =cos θ × l cross-section

Next, by considering the above equation for all single ‘slices’ on a cubic face,

we can expand this relationship to a singular face of the cube. We can achieve this

by finding the infinite sum of all these infinitesimally narrow cross-sectional shadows

and lengths:

l shadow1 +l shadow2 +l shadow3 +…=cos θ ( l cross-section1 +l cross-section2+ l cross-section3+ … )

` These sums actually comprise the areas of the entire face as it is taking the

sum of the lengths for each instance of width. As the area of the face would be a

square, we can rewrite and simplify where s is the side length:

Ashadow =cos θ × s2

7
 However, I noticed that this equation would produce a negative shadow for an

angle greater than 180°. As a shadow cannot have a negative area, we should take

the modulus of the trigonometric function to model the investigation more accurately:

Ashadow =¿ cos θ∨s2

To better visualise the problem, I decided to assume that the face's normal is

a unit vector. Remembering that the plane has three degrees of rotation, placing all

possible face rotations on a common centre would mean that the unit normal would

form a sphere with a unit radius around the faces. Each unit vector would be the

radius of this sphere. I had trouble conceptualising this at first. As a result, I

constructed figure 7 to aid in visualising this, with three possible random rotations of

the face shown. I constructed figure 8 to show the spherical nature, through a 2D

circular diagram, of the combination of all the normal vectors.

Figure 7: 3 Rotations of a Cube's Face with Their Unit Normal Vectors (Left)

Figure 8: All Rotations of Normal Vectors Would Form a Sphere (2D Shown) (Right)

Now that we have determined and visualised the relationship between the

shadow's area and a single cubic face's orientation (demonstrated by its unit normal

vector's angle), I will now try to determine the average shadow area caused by a

single face. This is aligned with the approach that was established earlier in this

8
report, where the relationship found for the shadow of a "deconstructed" model of a

cube (in this case, a single face) will be able to be used to determine a relationship

for the shadow caused by the entire cube.

Usually, the arithmetic mean can be found by considering the sum of the

values of a finite amount of data points divided by the number of data points.

Typically, the mean would be represented mathematically as:

∑ xi
μ= i=1
n

Here, I used μ to represent this as the arithmetic mean of a population of data

rather than a sample of data, as in this investigation, we are finding the mean

shadow area of all possible rotations of a face (and later, all possible rotations of a

cube) rather than a sample of rotations. However, I noticed that there are infinitely

many possible rotations as θ (the angle between the shadow (which is on the plane)

and the cubic face) is a continuous variable, which would lead to infinitely many

shadow area data points. Hence, as this definition of the arithmetic mean requires

0< n<∞ , we cannot use this approach in this investigation.

Clearly, it is necessary for us to finitise the infinite number of possible angles

θ. Instead of considering each possible individual angle, we should consider a range

of angles. Firstly, we should determine the probability that a random rotation of a

9
cubic face will be within a particular range of angles. We can visualise this in figure

8.

Figure 8: Graphical Representation of the Plane's Normal Vector in the Sphere

Interestingly, rotating the cubic face around the shadow's normal vector

results in an identical angle θ and hence an identical shadow. Here I used the

notation of dθ, which typically demonstrates a small 'chunk' (or range) of θ, as this

will be helpful later when we take the sum of shadows for all dθ using integration.

While Figure 8 demonstrates dθ as a relatively large range, in reality, dθ is taken to

approach an infinitesimally small angular range. We should notice that dθ only

approaches an infinitesimally small angular range because dθ must still be an actual

range greater than 0. Furthermore, I should clarify that whilst dθ is a range of angles.

We will approximate it as forming the height of a rectangular section that is a

continuous ‘band’ of the sphere’s surface area, as seen in figure 9:

Figure 9: Approximation of the Sphere Using Rectangular ‘Bands’

10
 Figure 9 demonstrates why it is beneficial for dθ to be taken as approaching

an infinitesimally small angular range. This is because the horizontal length clearly

changes depending on θ, so a smaller dθ increases the accuracy of the band’s area

as the horizontal length would vary less between the top and bottom of each band as

the band would be shorter.

As the horizontal length would change depending on the rotation of the cubic

face, we can use figure 8 to deduce this relationship using basic trigonometry

because the cubic face's normal vector, the horizontal length of each 'band' and the

shadow's normal vector form a right-angled triangle:

l horizontal
sin θ=
r

As we took the cubic face's normal vector (represented by r) to be a unit

vector, it has a magnitude of 1. Hence, we can simplify the equation:

l horizontal =sin θ

Hence, we should now consider the probability that a given cubic face's

normal vector will be within a particular band. As the probability of an event can be

expressed as:

n(Event)
P ( Event )=
n(U )

Clearly, the probability that a random cubic face's vector would be in a

particular band (a small range of angles) can be summarised as:

A band
P ( Vector in Band )=
SA sphere

11
 The area of the band will be the circumference of the sphere multiplied by the

height (dθ). Hence, we can write the probability of the vector is in the band as:

2 π l horizontal d θ
P ( Vector in Band )=
4 π r2

Earlier, we found the horizontal length (the horizontal distance from the

shadow's normal vector to the sphere) in terms of θ, so we shall substitute for this.

Once again, the sphere was defined as having a unit radius, r =1. Substituting:

2 π sin θ d θ
P ( Vector in Band )=
4π

We can simplify this relationship:

1
P ( Vector in Band )= sin θ d θ
2

Now that we have developed a relationship between the area of shadow

created and θ and a relationship explaining the probability of the cubic face having

an angle within a specific range, we should now determine the average shadow area

produced by a cubic face. We can do this by finding the sum of shadows created by

a range of θ (dθ) concerning the probability of randomly selecting that particular

range for a cubic face with three degrees of rotational freedom. Mathematically, this

can be expressed as:

π
Average Ashadow =∑ A shadow × P ( Vector in Band )
θ =0

Substituting the respecting relationships:

π
1
Average Ashadow =∑ ¿ cos θ∨s × sin θ d θ
2

θ =0 2

12
 Simplifying by moving constants outside the sum:

π
1
Average Ashadow = s2 ∑ ¿ cos θ∨sin θ d θ
2 θ=0

Notice that this sum is identical to an integral. It is beneficial for us to treat this

sum formally as an integral so we can use integration rules to evaluate this equation:

π
1
Average A shadow = s 2∫|cos θ|sin θ d θ
2 0

However, this raises a problem as we have a modulus in our integral. Luckily, as

sin θ is positive from 0 to π , and the graph of sin θ , θ ∈[0 , π ] has the line of symmetry

π
of x= , the following mathematical relationship can be used:
2

[ ]
sin θ , θ ∈ [ 0 , π ] =2sin θ , θ ∈ 0 ,
π
2

Whilst cosine would usually be negative in the second quadrant, ¿ cos θ∨¿ is

π
clearly positive here. ¿ cos θ∨¿ also has the line of symmetry x= . Hence, likewise:
2

[ ]
¿ cos θ∨, θ ∈ [ 0 , π ] =2 cos θ ,θ ∈ 0 ,
π
2

As a result, the integral can be rewritten without requiring a modulus by

finding the integral from 0 to

π
2
π
twice as cos θ> 0 ,θ ∈ 0 , :
2 [ ]
π
2
1 2
Average A shadow = s × 2∫ cos θ sin θ d θ
2 0

13
 Whilst integration by parts could be used here. I observed that this integral is

sin θ multiplied by its derivative cos θ . Hence, integration by substitution is an effective

method for evaluating this integral. Defining the substitution:

u=sin θ

Determining the necessary substitution for dθ by differentiating the

substitution equation in terms of θ:

du
=cos θ
dθ

Rearranging:

d u=cos θ d θ

We should also revaluate the boundaries of the integral in terms of the new

variable:

π
u Upper Boundary=sin
2

∴ uUpper Boundary =1

uLower Boundary =sin 0

∴ uLower Boundary =0

Now that we have revaluated the boundaries, we can rewrite the integral in

terms of u:

1
Average A shadow =s
2
∫ud u
0

We can now evaluate this integral using the reverse power rule:

14
 [ ]
1
u2 2
Average A shadow =s
2 0

Evaluating this equation in terms of the upper and lower boundary:

2
Average A shadow =s (
12 02
−
2 2 )
Simplifying:

1
∴ Average A shadow = s2
2

Now we have found the relationship between the average shadow area of a

singular cubic face and the cube's side length. We will now proceed to determine the

relationship between the average shadow area of a singular cubic face and the

average shadow area of the cube as an intermediate to find the relationship between

the average shadow area of the cube and the cube's side length.

Intuitively, not all the faces of the cube will contribute to the shadow area. To

formalise this intuition, I was guided toward the concept of convexity. Notice that a

convex shape is characterised by all internal angles strictly less than 180° (Lassak,

1993), and a cube satisfies this. Therefore, as the cube is convex, all possible rays

of light enter the cube in one face and go out through a different face for the same

rotation according to convexity. Figure 10 visualises this concept applied to a square;

however, this phenomenon is also readily applicable to a cube.

15
 Figure 10: Convexity of a Square

Hence, we can see that the shadow area is only affected by half of the cube's

faces. Specifically, the shadow area caused by a cube will be the sum of the average

cubic face's shadow for half of the faces. Mathematically:

6
1
ACube Shadow = ∑A
2 f =1 Face Shadow
f

Although this is not a very intuitive conclusion to draw, we must remember

that as we are taking the average shadow area for all rotations of the cube. Each

face will produce the same average shadow, which follows the probability

relationship we established earlier. Furthermore, I initially thought that perhaps the

positioning of each face in the cube (as they are constrained to be next to each

other) might decrease its resultant shadow. However, I remembered that the height

of a cubic face (h) does not affect its shadow which was found earlier.

Hence, accounting for all possible rotations of the cube, we can represent all

the possible shadow areas for the cube:

1
ACube Shadow (R 1) ¿ [A ( R1 )¿ + ¿ A Face (R1)¿ +…+ ¿ A Face ( R1 )]
2 Face f1 Shadow f2 Shadow fn Shadow

1 ¿ ⋮¿ ⋱ ¿ ⋮ ¿ A Cube Shadow (Rn ) ¿

ACube Shadow (R 2) ¿ [A ( R2) ¿ +¿ A Face (R 2)¿ +…+ ¿ A Face ( R2 )]
2 Facef 1
Shadow f2
Shadow fn
Shadow

⋮ ⋮ ¿

As mentioned above, the cubic face's average shadow for each face in a cube will be

identical. Hence, the average cube shadow area can be determined from this matrix to be:

6
1
Average ACube Shadow = ∑ Average A Square Shadow
2 f =1

As the average shadow area of one face was found earlier, this can be substituted

into the equation:

16
 6
1 1
Average ACube Shadow = ∑ s 2
2 f =1 2

Simplifying:

3 2
Average ACube Shadow = s
2

Here we have successfully determined the investigation's relationship between the

average area of a cube's shadow and its side length. This is graphically summarised in

figure 11.

Figure 11: Graph of Average Shadow Area of Cube vs Side Length of Cube

III CONCLUSION

1 Conclusion

Ultimately, this investigation aimed to investigate the average shadow area of

a cube with three degrees of rotational freedom. This was done by exploring the

effect of a cube's side length on the cube's average shadow due to a point light

source positioned directly above an infinite distance. As represented in figure 11, the

following relationship was determined:

17
 3
Average ACube Shadow = s 2
2

This relationship was ascertained by first considering the effect of a square's side

length on its average shadow. Then, the relationship between the average shadow

of a singular cubic face and the average shadow of a cube was determined,

culminating in producing a relationship between the cube's side length and its

average shadow, accounting for all possible rotations.

2 Reflection

While this investigation featured a valid use of mathematics to determine the

relationship between the average shadow cube and the cube's side length, several

assumptions were made to facilitate this exploration, which decreased the validity of

this relationship. Firstly, the light source was assumed to be an infinite distance

away. This allowed the modelling of this investigation to only feature rays of light that

were perpendicular to the shadow's plane, which meant that h (the distance between

the shadow's plane and the cube) did not affect the shadow area caused by the cube

if the whole cube was above the plane. However, in reality, light sources cannot be

infinitely far away. Secondly, the light source was assumed to be a point light source.

This assumption would have had a marginal effect in this investigation because of

the former assumption. A more accurate model would require modelling a light

permeating from the surface area of a sphere or other shape rather than from a

point.

An appropriate extension to this investigation would be to explore the average

shadow area of a cube for a finitely distanced point light source. However, this would

require a far less elegant and more sophisticated solution, as our methodology would

be invalid. Most notably, more variables would affect the shadow area, such as the

18
distance from the cube to the shadow's plane and the distance between the point

light source and the cube. Furthermore, the compartmentalised method we used to

find the average shadow caused by a singular cubic face before expanding this to

find the average shadow caused by the entire cube would not be valid. If we were to

find the average shadow caused by a plane from a finitely distanced light source, the

cube's shadow would not be related to this because of the geometric constraints on

the cube (the necessary heights and angles of all six faces so that a cube is formed).

The differing heights of the faces forming a cube would not allow similar relationships

with a finitely distanced point light source as was used in our exploration.

Furthermore, as a finitely distanced point light source would result in non-

perpendicular rays of light to the shadow’s plane, an umbra and penumbra would

exist due to the interference of light which cannot occur if all light rays are

perpendicular to the shadow’s plane.

The discussed mathematical difficulties that would be created without

assuming the light source is an infinite distance from the cube demonstrates the

elegance of the methodology used in this investigation. While the model used is not

an entirely accurate model of reality, this investigation's results provide a solid

approximation of the average shadow area of a cube in the real world, especially

when a light source is above the cube and when it is significantly distanced from the

cube and the shadow's plane.

IV BIBLIOGRAPHY

Ildan, M. M. (n.d.). Retrieved May 8, 2022, from AZQuotes.com:

https://www.azquotes.com/quote/1365580

19
Lassak, M. (1993). Approximation of convex bodies by rectangles. Geometriae

Dedicata, 111-117.

20