Comments about Chapter 3 of the Math 5335 (Geometry I) text

Joel Roberts
November 5, 2003; revised October 18, 2004

Contents:
Heron's formula (Theorem 8 in 3.5).
3.4: Another proof of Theorem 6.
3.7: The barycentric coordinates of the incenter.
3.7: The existence of angle bisectors (An alternative approach).
3.10: The inner product and the cosine.
3.11: The cosine function, the inner product, and right angle trigonometry
3.11: About the law of cosines

1. Heron's formula (Theorem 8 in 3.5). Here is a proof that doesn't depend on the formulas from
Chapter 2. The starting point is the observation that there are two different ways to use the
Pythagorean Theorem to write a formula for h2 in the following figure.

By definition, F is the foot of the altitude (perpendicular) drawn from B to AC . As usual, a, b

and c denote the lengths of the sides opposite the vertices A, B, and C respectively. We define x
to be the distance from C to F: positive in case F is in the ray CA and negative if not. (We
would get a negative value if we had an obtuse angle at C.) Similarly, b -x is the distance
(with sign) from A to F.
With this setup, h2 = a2 - x2, and also h2 = c2 - (b - x)2 = c2 - b2 + 2bx - x2. Setting these two
expressions equal, we obtain:
a2 - x2 = c2 - b2 + 2bx - x2,
so that:
2 2 2
2bx = a2 + b2 - c2, and therefore x = (a + b - c )/2b .
Substituting this into the equation h2 = a2 - x2, we obtain:
2 2 2 2 2 2 2 2 2 2 4 4 4
h2 = a2 ((a + b - c ) /4b2) = (2a b + 2a c + 2b c - a - b - c )/4b2 .
This leads to a formula which is equivalent to formula (3.3) of the text:
2 2 2 2 2 2 4 4 4
||ABC||2 = 1/4 b2h2 = (2a b + 2a c + 2b c - a - b - c )/16 .

And exactly as in the text, the numerator can be factored. One way to see that is to note that the
numerator, namely -(c4 - 2(a2 + b2)c2 + (a4 - 2a2b2 + b4)), is of 4th degree in c with only even
exponents and is thus a quadratic expression in c2. If we set it equal to zero and regard it as a
quadratic equation with c2 as the unknown, then the quadratic formula produces the following roots:
c 2 = (a 2 + b 2 ) 4a2 b 2 = (a2 2ab + b2)
Therefore, we have the following factorization:
-(c4 - 2(a2 + b2)c2 + (a4 - 2a2b2 + b4)) = -(c2 - (a + b)2)(c2 - (a - b)2).
Math 5335 Chapter 3 comments
Page 2
Since the last terms are differences of squares, this leads to:
-(c4 - 2(a2 + b2)c2 + (a4 - 2a2b2 + b4)) = -(c + (a + b))(c - (a + b))(c + (a - b))(c - (a - b)).
If the minus sign is absorbed into the second factor on the right, then we obtain a nearly final result:
||ABC||2 = (a + b + c)(a + b - c)(a + c - b)(b + c - a))/16
To get a more traditional version of the formula, we set s = (a + b + c)/2 (sometimes called the
semi-perimeter) and then observe that s - a = (a + b + c)/2, and so forth, thus leading to:
||ABC||2 = s(s - a)(s - b)(s - c), or !ABC = s(s " a)(s " b)(s " c)

2. 3.4: Another proof of Theorem 6. Here, we'll use a figure very similar to the one used in the proof
of Heron's formula presented above. Indeed, the only change is that we've labeled the distance from

A to F as y instead of b - x, since we'll actually want to find its value.

As in the previous proof, we have two ways to calculate h namely h2 = a2 - x2, and also
h2 = c2 - y2. Setting them equal to each other, we obtain the following equation:
a2 - x2 = c2 - y2.
And here is our other equation:
x + y = b.
Substituting y = b - x into the first equation, we have a2 - x2 = c2 - b2 + 2bx - x2. Thus:
2 2 2
2bx = a2 + b2 - c2, so that x = (a + b - c )/2b .
2 2 2
By doing a similar calculation, or by setting y = b - (a + b - c )/2b , we obtain:
2 2 2
y = (b + c - a )/2b .
Now, what does this tell us about the barycentric coordinates of F? A preliminary guess might be
2 2 2 2 2 2
that F = (y/b,0,x/b) = ((b + c - a )/2b2,0, (a + b - c )/2b2), but this guess would be wrong!!
Well, at a minimum, it would be the "opposite" of what's predicted in Theorem 6. To see why, and to
determine which choice really is correct, consider the actual barycentric coordinates F = (r,0,t) .
In rectangular coordinates, this is F = rA + tC. So, to get the distance from A, we calculate:
F - A = (rA + tC) - A = (r - 1)A + tC = -tA + tC = t(C - A).
Thus, the signed() distance from A is tb. In other words, the distance from A to F is associated
with the 3rd barycentric coordinate. And in a similar way the distance from B to F is associated
2 2 2 2 2 2
with the 1st barycentric coordinate. Accordingly: F = ((a + b - c )/2b2,0, (b + c - a )/2b2),
The results can summarized as in the following table. (The above calculation above gives the middle
row.)

Vertex opposite side Foot of altitude

2 2 2 2 2 2
A BC (0, (a + b - c )/2a2, (a + c - b )/2a2)
2 2 2 2 2 2
B AC ((a + b - c )/2b2,0, (b + c - a )/2b2)
2 2 2 2 2 2
C AB ((a + c - b )/2c2, (b + c - a )/2c2, 0)
Math 5335 Chapter 3 comments
Page 3
3. 3.7: The barycentric coordinates of the incenter. Well give a version of the proof of this formula
(Theorem 16 on page 78) based on a slightly different viewpoint from what is presented in the text.
Another significant feature is that the formula can be derived without using the existence of angle
bisectors (Proposition 15 of 3.7). And in fact, the proof in the text also works from a similar frame
of reference if you study it carefully.

The discussion that follows is based on the proof that was presented in class. Thus, we are
looking for an incenter (whose existence we need to prove): a point in the interior of the triangle that
is given in barycentric coordinates as I = (r,s,t). Thus, we need to find formulas for the barycentric
coordinates, under the assumption that r, s, and t all are positive. We denote the radius of the
inscribed circle as (rather than r) to avoid confusion with the first barycentric coordinate.
Accordingly, is the distance of I from each side of the triangle.

By Theorem 16 of Chapter 2, the distance of I = (r,s,t) from AC is hBs, where hB is the

distance from B to AC , i.e., the length of the altitude from B to AC . Thus, we have the equation
= hBs. {Literally applied, the theorem requires multiplying by |s| rather than s, but s > 0
because we are looking for a point in the interior of the triangle.} In a similar way, we obtain the
equations = hAr and = hCt. If we eliminate from these three equations, we are left with the
following two equations:
hAr = hBs and hBs = hCt.

To get an equation that relates the altitudes to other known quantities, we observe (for instance) that
we have the relation ||ABC|| = hBb/2, which gives hB = 2||ABC||/b. Using this, along with similar
equations that involve the other altitudes, we can transform our equations into:
2r||ABC||/a = 2s||ABC||/b and 2s||ABC||/b = 2t||ABC||/c,
or after canceling common factors:
r/a = s/b and s/b = t/c.
Along with these two equations, we have the standard equation r + s + t = 1. One way to solve the
resulting system involves setting equal to the common value of the ratios r/a, s/b and t/c. Thus,
= r/a = s/b = t/c so that r = a, s = b, and t = c. Substituting these values into the equation
1
r + s + t = 1, we find that (a + b + c) = 1, so that " = . If we substitute this into our
a+b+c
previous equations, then we obtain the following formula:
" a b c %(
I =$ , , ' .
#a +b+c a +b+c a +b+c& !
This certainly identifies a point in the interior of ABC which is equidistant from the 3 sides of the
triangle.
!
 Math 5335 Chapter 3 comments
Page 4
Incidentally, we also can combine this with the equations = hBs and hB = 2||ABC||/b to derive
2 "ABC
the formula ! = , also given in the text as part of Theorem 16.
a+ b+c

4. 3.7: The existence of angle bisectors (An alternative approach). This result corresponds to
Proposition 15 on page 75 of the text. By definition, an angle bisector of BAC is a ray AQ in
the interior of BAC such that BAQ CAQ. {And because Q is in the interior of the angle, it
follows that |BAQ| and |CAQ| both are equal to |BAC|/2.} Proposition 15 asserts that an angle
bisector exists and is unique. Proving that result (or at least mentioning it) before!discussing
Theorem 16 certainly is the more traditional approach. On the other hand, showing that a ray with
the required angle congruence properties lies in the interior of the angle is somewhat intricate. This
(somewhat subtle) question seems generally to be ignored in high school geometry courses, probably
to avoid confusing students who are learning geometry for the first time.

On the other hand, having already proven our result about the incenter, we actually have a
somewhat less intricate means available to prove the existence of the angle bisector:

Proposition. If I is the incenter of ABC, then AI is the angle bisector of BAC

PROOF. To get started, lets recall that weve shown that I is in the interior of BAC, so it follows
that the entire ray AI is contained in the
! interior of this angle. Next, let B be the point of AC
closest to I, and let C be the point of AB closest to I.

! !
!

C!

B!
Then the triangles ABI and ACI have right angles at B and C respectively. Hence, we can
2 2 2 2 2
apply the Pythagorean theorem to show that | AB" | = | AI | # B"I | = | AI | # $ and similarly that
| AC " |2 = | AI |2 # $ 2 . In particular, we have | AC " |=| AB" |. Therefore, we can apply the SSS criterion
to conclude that ABI ACI. It follows that there is an isometry that maps A to itself, I to
itself, and B to C. We conclude that ! BAI CAI, so that we have an angle bisector, as claimed.
! !
Remark. The isometry constructed in the previous proof maps the ray AI to itself and maps the
ray AB" to the ray AC " . We can use this fact as a starting point of proving that every point of AI
is equidistant from the lines AB and AC . Alternatively, see Part 2 of the proof of Proposition 15 of
3.7, which doesnt really use very much of the machinery from! Part 1 of that proof.
! ! !
5. 3.10: The inner product and the cosine. In Chapter 1, we defined angular measure of an angle,
! ! 1
ds
whose sides are rays with direction indicators U and V, to be equal to the integral " ,
U,V 1 ! s2
Math 5335 Chapter 3 comments
Page 5
We also observed that this expression defines the integral as a strictly decreasing function of its lower
endpoint, and we decided to call this function the arccosine. Thus, the arccosine turns out to be a
strictly decreasing function that maps the closed interval [-1,1] to the closed interval [0,]. Since
a strictly increasing function or a strictly decreasing function is a bijective mapping from its domain
to its range, it has an inverse function. The inverse function of the arccosine function is the cosine.
Thus, the cosine is a strictly decreasing function that maps the interval [0,] to the interval [-1,1].
1
ds
So, if = " = arccos(U,V) is the measure of our angle, then we have U,V = cos() in
U,V 1 ! s2
the case where U and V are unit vectors.
More generally, if U and V are direction indicators of the sides of an angle (but not necessarily
unit vectors), and if is the angular measure, then we have the following important identity:
U,V = ||U|| ||V|| cos,
which may be familiar from vector calculus courses. To check it in our situation, we observe that
U = aU0 and V = bV0, where U0 and V0 are unit vectors, and a and b are positive real numbers.
(We want U0 and V0 to point in the same direction as U and V respectively.) So, ||U|| = a, and
||V|| = b, while U,V = abU0 ,V0. Therefore, the identity U,V = ||U|| ||V|| cos follows from the
previously known formula U0 ,V0 = cos().

6. 3.11: The cosine function, the inner product, and right angle trigonometry. Historically, the
most basic definition of the cosine of an angle was the quotient of adjacent side over hypotenuse in a
right triangle. In our formulation, this is fairly immediate in the case where the unit vector (1,0) is
one of the sides of an angle.

To check this algebraically, we set U = (1,0) and V = (v1,v2) and then calculate:
U,V = (1,0),(v1,v2 = v1. Thus, (v1,0) = v1U = U,VU turns out to be at the base of the
perpendicular from the point V to the line OU . A negative value of U,V is interpreted as
meaning that the base of this perpendicular lies on the ray opposite to OU .

7. 3.11: About the law of cosines. As noted in the text, the proof of the "first version" of the Law of
Cosines uses Lemma 1 of Chapter 2. Since that lemma isn't proved in the text, we'll first state and
prove a variant of that auxiliary result.

Lemma. Let A, B, and C be points in |Rn. Then:

||A-B||2 = ||A-C||2 + ||B-C||2 - 2(A-C),(B-C)
Math 5335 Chapter 3 comments
Page 6
Proof: We recall that ||A-B||2 = (A-B),(A-B) and then insert the "missing term", namely C, to
make A-B appear as a difference of differences. More plainly, the idea is to write:
A-B = (A-C) - (B-C).
Using this, we calculate the inner product:
(A-B),(A-B) = (A-C) - (B-C), (A-C) - (B-C)
= (A-C), (A-C) - (A-C),(B-C) - (B-C), (A-C) + (B-C),(B-C).
{Formally, we used the fact that the inner product is linear in each of the variables. More informally,
we can view it as similar to expanding the binomial expression (X-Y)2.} Next, we can use the
symmetry of the inner product: U,V = V,U to obtain:
(A-B),(A-B) = (A-C), (A-C) - 2(A-C),(B-C) + (B-C),(B-C).
Finally, we replace each inner product U,U with the square of the corresponding norm to obtain:
||A-B||2 = ||A-C||2 + ||B-C||2 - 2(A-C),(B-C),
thus proving the lemma.

To apply this when we're thinking of A, B, and C as the vertices of a triangle in |R2, we use the
letters a, b, and c to denote the lengths of BC , AC , and AB respectively. Thus:
c = AB = ||A-B||,
and so forth. If we make these substitutions we obtain the identity:
c2 = a2 + b2 - 2(A-C),(B-C).
As a final step, we apply the identity U,V = ||U|| ||V|| cos from the previous section, with
U = A-C, V = B-C, and = |CB| = |C| to obtain the following identity:
c2 = a2 + b2 - 2||A-C||||B-C|| cos(C),
from which we deduce the law of cosines.

Law of Cosines (1st version). Given ABC, let a = BC , b = AC , and c = AB . Then:

c2 = a2 + b2 - 2ab cos(C),
Just to check that our answer makes sense at least in a special case, note that if |C| = /2, then
cos(C) = 0, and we recover the usual Pythagorean identity c2 = a2 + b2 .

Finally, we can transform our main identity algebraically to obtain the other version of the law of
cosines.

Law of Cosines (2nd version). Given ABC, let a = BC , b = AC , and c = AB . Then:

a 2 + b 2 ! c2
cos(C) = ,
2ab

In particular, if c2 < a2 + b2, this gives a positive value of cos(C), corresponding to an acute angle
at C. On the other hand, if c2 > a2 + b2, the formula gives a negative value of cos(C),
corresponding to an obtuse (i.e., non-acute) angle at C.

Go back:
To the class homepage

To the homework page

To the course description