A Triangle Ratio Theorem

and its Generalization to Higher Dimensions

Hari Rajesh
Ridgewood High School, Ridgewood, New Jersey

Abstract: In this paper, I discuss an interesting theorem involving a triangle and lines through a point parallel to
the sides of the triangle, including both a geometric and a coordinate-based proof of the theorem. Later, I discuss
and prove a generalization in higher dimensions. I conclude with a conjecture about a further generalization.

1 Introduction
I first noticed this 4 years ago while I was still in middle school. I made a little document to try and prove it, but
I quickly forgot about it. I decided to look through my old stuff one day and reopened the document. I began to
look into it and have since proven the theorem and generalized it.

2 A Statement of the Theorem

Take any triangle 4ABC and an internal point P . Construct three line segments parallel to the sides of the triangle
through P. This will create a diagram similar to the following:

Figure 1: Triangle ABC with point P

The sum of the ratios between a segment and the corresponding parallel side of the triangle is 2:

IF EH GD
+ + =2 (1)
AB BC CA
Given that this involves ratios in a triangle, I will call it the Triangle Ratio Theorem.

3 Proof
DB
First, notice that 4DBG ∼ 4ABC. This implies that AB
= GD
CA
. Then, notice that IP = AD because they are
IP AD AD DB
parallel sides in the parallelogram ADP I. Thus, AB
= AB
. Next, we shall use the fact that AB
+ AB
= 1. We

Mathematical Reflections 3 (2023) 1

 IP GD
can then use the previously found equalities to get that AB
+ CA
= 1.
AE EH
We may now repeat the same reasoning to get a second equation. 4AEH ∼ 4ABC, so AB
= BC
. Being
PF EB AE EB
parallel sides of parallelogram EBF P , P F = EB, so AB
= AB
. Since AB
+ AB
= 1, and given the previous
PF EH
equalities, AB
+ BC
= 1.
IP PF EH GD IP PF IF
Adding the two equations gives us AB
+ AB
+ BC
+ CA
= 2, and since AB
+ AB
= AB
, we get the desired
equation:

IF EH GD
+ + =2
AB BC CA

3.1 Related Results

There are a few related results of the Triangle Ratio Theorem apart from the main one. The following three
equations are all true:

DE FG HI
+ + =1
AB BC CA
EB GC IA
+ + =1
AB BC CA
AD BF CH
+ + =1
AB BC CA
Each of these three equations can be proven either directly from the Triangle Ratio Theorem or independently.
I shall leave the proofs to the reader, except for the first one, which I shall prove in Section 6.2.5 when discussing
generalizations to the theorem.

4 Generalization: What if P is Outside of the Triangle?

P does not have to lie in the triangle for this to work. However, we need to make a few modifications in order to
allow this to happen. First, we must extend the sides of the triangle by considering the whole line because otherwise,
the segments would not hit anything, so no distance calculation could be made. The other modification is to turn
the segments into vectors. Why? Return to Figure 1. Notice that point I is to the left of point F . However, if P is
IF
above point C, this will flip, so point F will be to the left of point I. This means that AB needs to be subtracted
from the sum. This can be done easily using vectors. This negative sign is quite important.
−→ −−→
In Figure 2, the full lines are not shown for simplicity. Let’s take vectors IF and AB. If they are pointing in
−→ −→
IF IF
the same direction, we add − → but if they point in the opposite direction, we subtract −
− −→ . This can be easily
AB AB
accomplished using the dot product.
Recall the following identity involving the dot product:

Given vectors u and v,

u·v
cos (θ) = kukkvk
where θ is the angle between u and v
−→ −−→
Thus, if IF and AB are parallel, cos (θ) will be 1 and if they are anti-parallel, cos (θ) will be -1. This gets us
the desired sign, so we can just multiply it by the ratio and add them all up.
Let un be one of the vectors through P and let vn be the parallel/anti-parallel side on the triangle. Using the
previous formula multiplied by the ratio, we get the following:
un ·vn kun k un ·vn
kun kkvn k kvn k = kun k2

We can then get the following equation which we will need to prove:
P un ·vn
kun k2 = 2

Mathematical Reflections 3 (2023) 2

 Figure 2: Triangle ABC with point P using vectors instead of segments

5 Proof of the Generalization

The first step in proving this generalization will be to find a way to remove vectors from the equation. This is not
that difficult. Since un and vn are either parallel or anti-parallel, they are scalar multiples of each other. If you
define the scalar kn as uvnn , then we are simply looking for
P
kn and showing that it equals 2.
The second step is to define the triangle in terms of points on the coordinate plane. You could define it as
(xa , ya ), (xb , yb ), and (xc , yc ), but this is computationally inefficient and unnecessary. Instead, we can assume
without loss of generality that A lies at (0, 0) and B lies at (1, 0). We can do this because any translation of the
triangle will maintain the ratios and since we are dealing with length, scaling the triangle does not change the
ratios. We thus get the following diagram:

Figure 3: Triangle ABC with point P using lines

Note that while Figure 3 appears the same as any of the other triangles, P can lie anywhere on the plane. In
IF FC
order to find the ratio of IF to AB, we can simply use the fact that AB = BC , following the logic from the original
proof. In this case, however, we can simply use a ratio of the y-coordinates of P and C because IF is parallel to
y −y
the x-axis. This ratio works out to cyc p . You can easily verify that this gives us the desired sign change when

Mathematical Reflections 3 (2023) 3

yp > yc , which is the main goal of this generalization.
To do the other ratios, we will need to find equations for the lines through EH and GD. This will allow us to
find the intersection point with the line through AB. From there, we can use the following two facts to find the
AE
ratios: AB = EHBC
and DB
AB
= GD
CA
.
For the line through EH, we’ll use the slope of the parallel BC, which is xcy−1
c
. Using that and the point-slope
yc
form through P yields the following equation: y = xc −1 (x − xp ) + yp . Doing the same for the line through GD
yields the following equation: y = xycc (x − xp ) + yp .
Now, we can solve for the x-value of the intersection between these lines and the line through AB, which will be
y x
the x-values of points D and E, by simply setting y = 0. Doing this yields the following two results: xd = xp − pyc c
yp (xc −1)
and xe = xp − yc . To find the ratios, we can first calculate AE, DB, and AB. By definition, AB = 1,
AE EH yp (xc −1)
which simplies the computations. AE simply equals xe − xa = xe , so AB
= BC
= xe = xp − yc . DB is
yp xc
xb − xd = 1 − xd , so DB
AB
= GD
CA
= 1 − xd = 1 − xp + yc .
You can verify that this also results in the necessary sign
changes when P is beyond any of the corners of the triangle.
IF
Now we can calculate AB + EH
BC
+ GD
CA
. First, we can expand and get a common denominator on all of the terms:
EH yc xp −yp xc +yp GD yc −yc xp +yp xc
BC
= yc and CA
= yc . Now we can simply add.

IF EH GD yc − yp yc xp − yp xc + yp yc − yc xp + yp xc
+ + = + +
AB BC CA yc yc yc
yc − yp
+yp +(yp(
xc (−(yp(x( yc x
c +( −(
(p( yc(x(p + yc
=
yc
2yc
= =2
yc
Thus, we get the desired result. One easily identifiable error is when yc = 0, but this makes sense as this would
create a triangle that is a straight line, so any parallel lines that aren’t collinear to the “triangle” would never
intersect. There is a second issue that must be dealt with, however, when xc is 0 or 1; this is because some of the
slopes put xc or xc − 1 in the denominator. Notice, however, that neither scaling nor rotating the triangle affect the
ratios. Thus, we can simply rotate/scale the triangle as necessary to create a similar triangle that has a different
side on the x-axis and xc 6= 0 or 1, as can be seen in Figure 4.

Figure 4: Transformation of a right triangle

6 Generalizing to Higher Dimensions

The seemingly obvious next step forward is to generalize the result to a tetrahedron. Take a tetrahedron ABCD
and an internal point P , as is shown in Figure 5.
Now, construct four planes through P , each parallel to a face of the tetrahedron. You might assume we can
take the ratio of the areas, but if we do this, we would need to take the root of the ratio. Instead, we can just find
the scaling factor. If you do this and add it up, you will find that the sum is 3. We will prove this later.
The next question is whether this applies to the higher dimensional cases. The fact that the 2- and 3-dimensional
cases both work in a similar manner suggests that it will work. In fact, it also works in the 1-dimensional case,
though trivially.
AP
Figure 6 clearly shows that AB +P B
AB
= 1. If P falls outside of the segment, then one can simply subtract as
necessary. In any case, the reasoning behind this should require no proof. In the 0-dimensional case, where there is
only a point, it is difficult to properly determine a notion of scale, so it would be difficult to conclude whether the
theorem applies or not. However, this leads me to a more generalized theorem for any dimension ≥ 1.

Mathematical Reflections 3 (2023) 4

 Figure 5: Tetrahedron ABCD with point P

Figure 6: Segment AB with point P

6.1 The Simplex Ratio Theorem

In an n-dimensional space, take any n-simplex whose vertices are non-cohyperplanar, and an interior point P . Let
ui be a hyperplane through P parallel to vi , an facet of the simplex. If the ratio of ui and vi is taken to be the
scaling factor between them, then the following is true:

X ui
=n (2)
i
vi

I will now briefly explain what the theorem is saying in a bit more detail. First, a simplex is simply a general-
ization of a triangle to higher dimensions. Here is a little chart:
n-simplex Common Name
0-simplex Point
1-simplex Line
2-simplex Triangle
3-simplex Tetrahedron
4-simplex 5-cell
A hyperplane is simply a generalization of a plane to higher dimensions. In 4-dimensions, for instance, a
hyperplane is a 3-dimensional object. The Triangle Ratio Theorem fails when the triangle is just a line because
the parallel lines through P never intersect the “triangle.” Thus, the contraint that the vertices of the simplex are
non-cohyperplanar must be made.
Facet refers to the generalization of the sides of a triangle. Triangles have sides, tetrahedrons have faces, 5-cells
have cells, and so on, which are all considered facets of the n-simplex.
The scaling factor means the ratio between the hyperplane and the parallel facet while the fact that it is signed
refers to the fact that it takes on a negative sign if the orientation of the hyperplane through P is opposite to that
of the parallel facet.

Mathematical Reflections 3 (2023) 5

 The theorem states that the sum of those signed scaling factors should equal the number of dimensions. So far,
I have shown it to be true in the 1- and 2-dimensional cases, but by the end of the paper, I will have proven it.

6.2 Lemmas
Before I prove the theorem, I must first provide a few lemmas. Most of these are necessary to proving the theorem
while the first one is an important observation.

6.2.1 Lemma 1: When P is at a Vertex

If P is at a vertex, then the proof of the theorem is easy because each hyperplane is identical in size to their parallel
facet (so the ratio for each one is 1), except for the facet opposite the vertex, whose ratio is 0. Since there are n + 1
facets in an n-simplex, the ratio would add up to n + 1 − 1 = n. Thus, we have proven it for every n-simplex when
P is at a vertex.

6.2.2 Lemma 2: When P is on a Facet, an (n-1)-case is Formed

Figure 7: Triangle ABC with point P on AB

AP
Notice in Figure 7 that the bottom side is a 1-dimensional case, where AB +P B
AB
= AB
AB
= 1. This also holds for
the 3-dimensional case, as can be seen in Figure 8. Notice that face ABD with point P forms the 2-dimensional
case with lines parallel to AB, AC, and BC.
A proof of this is quite simple. For an n-simplex, where there are n + 1 facets, there are n + 1 hyperplanes
through P . However, if P lies on a facet, then only n of those hyperplanes are non-cohyperplanar with the simplex.
These hyperplanes all go through P and are parallel to the facets of the hyperplane through P . This is equivalent to
having an (n−1)-simplex and hyperplanes through P parallel to the facets of that simplex, which is the (n−1)-case.

6.2.3 Lemma 3: A Hyperplane Through P Parallel to a Facet Forms an (n-1)-case

This is actually easiest to notice in the 3-dimensional case. Refer to Figure 9.
Notice that the intersection of plane KLM and the other planes forms a shadow that is identical to a 2-
dimensional case. This is also true of the 2-dimensional case, where a line through P forms a 1-dimensional case.
This naturally follows from Lemma 2. If you were to only consider the part of tetrahedron above plane KLM , you
would end up with P being on the face of the tetrahedron DKLM , which we know forms the (n − 1)-case from
Lemma 2.

Mathematical Reflections 3 (2023) 6

 Figure 8: Tetrahedron ABCD with point P on facet ABD

Figure 9: Tetrahedron ABCD with Point P and Plane KLM

6.2.4 Lemma 4: The Desired Sum is the Number of Facets Minus the Ratio of the Central Sections
In Figure 10, the central sections are segments DE, F G, and HI. In higher dimensions, it works similarly. In
Figure 11, the central sections are the outer faces of the small tetrahedra shown.
To prove the lemma, let’s look at Figure 10. Notice that 4ABC ∼ 4IF C ∼ 4DEP . Additionally, notice
the height of 4DEP + the height of 4IF C = the height of 4ABC. Thus, a linear quantity, such as the side
lengths, of 4DEP + the corresponding quantity of 4IF C yields the corresponding quantity of 4ABC, by similar
triangles. We can easily rearrange this and apply it to get the following: IF + DE = AB. This is true of the other
sides and this method can generalize to higher dimensions.
For example, in 3-dimensions, refer to Figures 9 and 11. In Figure 11, we shall use the bottom tetrahedron as
an example. Notice that it, tetrahedron KLM D in Figure 9, and tetrahedron ABCD are all similar, and that the
height of the bottom tetrahedron + the height of KLM D yields the height of ABCD. Thus, as they are similar
by a linear factor, all linear quantities such as side lengths of the bottom tetrahedron + tetrahedron KLM D yields
that of ABCD. Thus, the sum of the ratios of the central sections + the sum of the ratios of the hyperplanes,
which is what we want, = the sum of the ratios of the facets themselves. The ratio of a facet to itself is simply 1,

Mathematical Reflections 3 (2023) 7

 Figure 10: Triangle ABC with point P

so the sum of these ratios is the number of facets. Rearranging gives us that the desired sum of the ratios of the
hyperplanes to their respective facets = the number of facets − the sum of the ratios of the central sections.

Figure 11: Tetrahedron ABCD and Point P with Smaller Tetrahedra

6.2.5 Lemma 5: The Sum of the Ratios of the Central Sections to their Respective Sides is 1
In the 2-dimensional case, it follows naturally from the original theorem, as I mentioned in Section 3.1, when
discussing related results of the Triangle Ratio Theorem. However, it will be important to start by proving this
first, so here is a proof that doesn’t use the Triangle Ratio Theorem as a prior step. More importantly, it also
generalizes to higher dimensions extremely easily.
Return to Figure 10. First, notice the following:

Mathematical Reflections 3 (2023) 8

 IH IP
=
IC IF
GF PF
=
CF IF
This is due to similar triangles. We can then do the following:

IH GF IP PF
+ = + =1
IC CF IF IF
IF CF IC
Next, notice that AB
= BC
= AC
, again by similar triangles. Thus, we can do the following:

  
IF IH GF IF
= +
AB IC CF AB
IH IF GF IF
= +
IC AB CF AB
IH IC GF CF
= +
IC AC CF BC
IH GF
= +
AC BC
DE
Now if we add in AB
, we get the following:

IH GF DE IF DE
+ + = +
AC BC AB AB AB
IF DE
We already know AB
+ AB
= 1 from Lemma 4, so we get our desired result:

IH GF DE
+ + =1 (3)
AC BC AB
The general process is similar in 3-dimensions. Return to Figure 11. You can prove that the sum of the scaling
factor of the tetrahedrons is equal to 1 using similar logic. It revolves around the earlier fact that the plane through
P forms the 2-dimensional case. Return to Figure 8 to see this more clearly. The sum of the central sections in
plane KLM is 1. We can multiply both sides by, for instance, KL AB
. Then, adding in the last tetrahedra and using
Lemma 4, we can get that the sum of the ratios of the inner sections is 1.
Thus, the n-dimensional case relies on the (n − 1)-dimensional case. If we have an n-simplex and we look at
just one of the hyperplanes through P , we get the (n − 1)-case. Assuming that the (n − 1)-case is true, then we
can multiply by the scaling factor between the hyperplane and the parallel facet and add the remaining smaller
n-simplex. By Lemma 4, this gives us 1. This sets us up for a proof by induction, which I shall now complete.

6.3 The Proof of the Simplex Ratio Theorem

First, take an n-simplex and a point P and construct hyperplanes through P parallel to the facets of the n-simplex.
Assume that the (n − 1)-case holds.
By Lemma 5, if the (n − 1)-case holds, then the sum of the ratio of the center sections to their respective facets
is 1 in the n-case. By Lemma 4, the desired ratio is the number of facets minus the sum of the ratios of the center
sections. In an n-simplex, there are n + 1 facets, so the desired ratio is n + 1 − 1 = n.
Given that the n-case is true if the (n − 1)-case is true, and since it is true for n=2, then it is true for every
n-simplex by induction.

Mathematical Reflections 3 (2023) 9

7 Points Outside of the Simplex
The purely geometric proof applies to all internal points of the simplex. For external points, however, things are
not as straightforward. You could simply define the scaling factor to be signed, but the geometric proof doesn’t
definitively prove that it works, though I am confident it is provable. A proof analogous to the coordinate-based one
shown earlier is likely the best candidate, though it would be far more complex in higher dimensions, unfortunately.
Another thing I must mention is that while you can certainly just define the scaling factor to be signed, when
dealing with computations or demonstrations, such as through Geogebra, there must be a way to calculate the sign.
Using vectors still works in higher dimensions, though there are a few methods of using them. One of these is to
use a unit vector normal to the facet and another unit vector normal to the hyperplane, then use the dot product of
those vectors. If orientation is preserved by these vectors, then this will yield the correct sign. Another option is to
simply have one vector lie on edge of the simplex while another one lies on the hyperplane parallel or anti-parallel
to the first one, depending on the required sign. Then, the dot product of these vectors, divided by the product of
the magnitudes, yields the sign required, which is analogous to the 2-dimensional case.
Again, this has not been proven, but given the fact that the 2-dimensional case has been proven and the n-
dimensional case has been proven for internal points, I believe it can be done, so I will call it the General Simplex
Ratio Conjecture.

Mathematical Reflections 3 (2023) 10