There is a trigonometric identity:

$$\cos^2A+\cos^2B+\cos^2C+2\cos A\cos B\cos C\equiv 1\text{ when }A+B+C=\pi$$

It is easy to prove it in an algebraic way, just like that:

$\quad\cos^2A+\cos^2B+\cos^2C+2\cos A\cos B\cos C\\=\cos^2A+\cos^2B+\cos^2\left(\pi-A-B\right)+2\cos A\cos B\cos \left(\pi-A-B\right)\\=\cos^2A+\cos^2B+\cos^2\left(A+B\right)-2\cos A\cos B\cos \left(A+B\right)\\=\cos^2A+\cos^2B+\left(\cos A\cos B-\sin A\sin B\right)^2-2\cos A\cos B\left(\cos A\cos B-\sin A\sin B\right)\\=\cos^2A+\cos^2B+\cos^2A\cos^2B+\sin^2A\sin^2B-2\sin A\cos A\sin B\cos B-2\cos^2A\cos^2B+2\sin A\cos A\sin B\cos B\\=\cos^2A+\cos^2B-\cos^2A\cos^2B+\left(1-\cos^2A\right)\left(1-\cos^2B\right)\\=\cos^2A+\cos^2B-\cos^2A\cos^2B+1-\cos^2A-\cos^2B+\cos^2A\cos^2B\\=1$

Then, I want to find a geometric way to prove this identity, as $A+B+C=\pi$ and it makes me think of the angle sum of triangle. However, it is quite hard to prove it in a geometric way. Therefore, I hope there is someone who can help. Thank you!

  • 1
    $\begingroup$ For future reference, the algebraic proof can also be found on the site here. $\endgroup$ – Arnaud D. Sep 10 at 20:31
  • $\begingroup$ I've also found this question, for an identity of which yours is a particular case (though you'll still need a bit of algebra to see it). $\endgroup$ – Arnaud D. Sep 10 at 20:33
  • $\begingroup$ I'm still thinking about possibility that the equation will be derived when we divide a triangle into four pieces and consider areas, but I can't find a good partition for that. $\endgroup$ – Seewoo Lee Sep 11 at 8:47
  • $\begingroup$ @Isaac YIU Math Studio It means that $A$, $B$ and $C$ are positives? $\endgroup$ – Michael Rozenberg Sep 11 at 11:41
  • $\begingroup$ @MichaelRozenberg It doesn't mention that, but it's alright if you add this condition. $\endgroup$ – Isaac YIU Math Studio Sep 11 at 11:45

Since the accent in the OP is put on a purely geometric solution, i can not even consider the chance to write $\cos^2 =1-\sin^2$, and rephrase the wanted equality, thus having a trigonometric function which is better suited to geometrical interpretations.

So this answer has two steps, first we reformulate the given identity in a mot-a-mot geometric manner, the geometric framework is introduced, some strictly geometrically transposed equivalent relations are listed, then we give a proof:

In the triangle $\Delta ABC$ let $AA'$, $BB'$, $CC'$ be the heights, $A'\in BC$, $B'\in CA$, $C'\in AB$, intersecting in $H$, the orthocenter. We assume that the diameter $2R$ of the circumcircle is normed to be the unit. Then we have the following situation for the lengths of some segments in the picture:

Some relations in the triangle, values for the trigonometric functions of the angles

$$ \begin{aligned} AH &=\cos A\ , \qquad & HA'&=\cos B\cos C\ ,\\ BH &=\cos B\ , \qquad & HB'&=\cos A\cos C\ ,\\ CH &=\cos C\ , \qquad & HC'&=\cos A\cos B\ . \end{aligned} $$

Proof: We have: $$ \sin \hat B =\sin \widehat{C'HA} =\frac{C'A}{AH} =\frac{AC\;\cos A}{AH} =\frac{2R\sin B\; \cos A}{AH} =\frac{\sin B\; \cos A}{AH} \ , $$ which implies $AH=\cos A$, and the similar relations. Then we express twice the area of $\Delta HBC$ as $$ HA'\cdot BC =2[HBC]=HB\cdot HC\cdot \sin\widehat{BHC}\ ,$$ thus getting $HA'=\cos B\cos C$.

We are in position to give a geometric mask to the given equality:

We use the above notations in $\Delta ABC$. We denote by $a,b,c$ the lenghts of the sides. Let $M_A, M_B,M_C$ be the mid points of the sides $BC$, $CA$, respectively $AB$. Let $G=AM_A\cap BM_B\cap CM_C$ be the intersection of the medians, the centroid. Let $A^*, B^*, C^*$ be the mid points of $HA$, $HB$, $HC$. Let $N$ be the center of the Euler circle $(N)$ passing through the nine points $A',B',C'$; $M_A,M_B,N_C$; $A^*, B^*,C^*$. It is the mid point of $OH$, and $M_AA^*$, $M_BB^*$, $M_CC^*$ are diameters in $(N)$, having the lenght $R=OA=OB=OC$. (For $OM_AA^*A$ is a parallelogram.)

Nine points circle in a triangle

Then we have the following relations: $$ \begin{aligned} HA^2+HB^2+HC^2 + 2 HA\cdot HA' &= 4R^2\ ,\\ HA^*{}^2+HB^*{}^2+HC^*{}^2 + HA^*\cdot HA' &= R^2\ ,\\ 4OM_A^2+4OM_B^2+4OM_C^2 &= 3R^2+OH^2\ ,\\ 9R^2 &= a^2 +b^2 + c^2 +OH^2\\ 9R^2 &= a^2 +b^2 + c^2 +9OG^2\ . \end{aligned} $$

Proof: The relations above are equivalent:

  • $AH^2=4A^*H^2=4OM_A^2$, and $2 HA\cdot HA'=4 HA^*\cdot HA'$ is the power of $H$ in the circle $(N)$, so it can be rewritten using its radius $NA^*=\frac 12 R$ and the distance to its center, $NH=\frac 12 OH$ as $2 HA\cdot HA'=4 HA^*\cdot HA'=R^2-OH^2$.

  • From the triangle $OBM_A$, $4OM_A^2+BC^2 =4(OM_A^2+BM_A^2)=4OB^2=4R^2$.

  • Note that $G$ cuts the median $AM_A$ in the proportion $AG:GM_A=2:1$, so it projects on $BC$ in the same proportion. This also holds for the colinear points $H,G,O$, so $HG:GO=2:1$, so $HO=3GO$.

  • The last relation, $OG^2 = R^2-\frac 13(a^2+b^2+c^2)$, is a standard formula. We have in general the formula for an arbitrary point $P$: $$PA^2+PB^2+PC^2=GA^2+GB^2+GC^2+3GP^2\ .$$ We apply it for $P=O$, getting $3R^2=3OG^3+\sum AG^2=3OG^3+\frac 49\sum AM_A^2=3OG^3+\frac 49\sum \left(\frac 12b^2+\frac 12 c^2-\frac 14 a^2\right)=3OG^3+\frac 49\sum \frac 34a^2=3OG^3+\frac 13\sum a^2\ .$



For positives $A$, $B$ and $C$ there is the following way.

Let $A=\max\{A,B,C\},$ $\pi-A=\alpha,$ $\frac{\pi}{2}-B=\beta$ and $\frac{\pi}{2}-C=\gamma$.

Thus, $\alpha$, $\beta$ and $\gamma$ are measured angles of the triangle and let sides-lengths of the triangle be $a$, $b$ and $c$ respectively.

Thus, since by law of sines $$\frac{a}{\sin\alpha}=\frac{b}{\sin\beta}=\frac{c}{\sin\gamma},$$ by law of cosine we obtain: $$\sin^2\alpha=\sin^2\beta+\sin^2\gamma-2\sin\beta\sin\gamma\cos\alpha$$ or $$\sin^2(\pi-A)=\sin^2\left(\frac{\pi}{2}-B\right)+\sin^2\left(\frac{\pi}{2}-C\right)-2\sin\left(\frac{\pi}{2}-B\right)\sin\left(\frac{\pi}{2}-C\right)\cos(\pi-A)$$ or $$1-\cos^2A=\cos^2B+\cos^2C+2\cos B\cos C\cos A$$ and we are done!

  • $\begingroup$ Nice. This is as “geometric” a proof as any here so far. $\endgroup$ – John Bentin Sep 11 at 17:36

I don't know if this counts as a proof, but following your suggestion, I used the cosine laws to obtain your result.

Suppose you have a triangle ABC as in the figure: enter image description here

Since the angles $A+B+C=\pi$, these are the internal angles of a general triangle. Using the law of cosines, you can write:

$a^2=b^2+c^2-2bc\cos A\\b^2=a^2+c^2-2ac\cos B\\c^2=a^2+b^2-2ab\cos C$

It follows from here that:

$ \cos A=\frac{a^2-b^2-c^2}{-2bc}\\ \cos B=\frac{b^2-a^2-c^2}{-2ac}\\ \cos C=\frac{c^2-a^2-b^2}{-2ac} $

Now, to verify your formula, we have:

$\cos^2A+\cos^2B+\cos^2C+2\cos A\cos B\cos C=\\ \left(\frac{a^2-b^2-c^2}{-2bc}\right)^2+\left(\frac{b^2-a^2-c^2}{-2ac}\right)^2+\left(\frac{c^2-a^2-b^2}{-2ac}\right)^2+2\left( \frac{a^2-b^2-c^2}{-2bc} \right)\left( \frac{b^2-a^2-c^2}{-2ac} \right)\left( \frac{c^2-a^2-b^2}{-2ac} \right) $

It is now a matter of manipulation of the equation to show that this equals 1. Observe that the least common multiple of the first three terms is $4a^2b^2c^2$, which is equal to the product of the last term,

$ \frac{a^2(a^2-b^2-c^2)^2+b^2(b^2-a^2-c^2)^2+c^2(c^2-a^2-b^2)^2}{4a^2b^2c^2}-\frac{(a^2-b^2-c^2)(b^2-a^2-c^2)(c^2-a^2-b^2)}{4a^2b^2c^2} $

expanding the products in the numerator, you can verify that: $a^2(a^2-b^2-c^2)^2+b^2(b^2-a^2-c^2)^2+c^2(c^2-a^2-b^2)^2-(a^2-b^2-c^2)(b^2-a^2-c^2)(c^2-a^2-b^2)=4a^2b^2c^2$

so the fraction simplifies to

$ \frac{4a^2b^2c^2}{4a^2b^2c^2}=1 $

  • 4
    $\begingroup$ Right, but not a geometric way to prove the claim. $\endgroup$ – Michael Hoppe Sep 10 at 18:39
  • $\begingroup$ The critical first step, "Since the angles $A + B + C = \pi$, these are the internal angles of a general triangle. [Therefore] using the law of cosines..." is certainly itself a geometric argument, even if the rest of the solution is algebraic manipulation. $\endgroup$ – Travis Willse Sep 10 at 18:49
  • $\begingroup$ Though it consists of some geometry, but it is almost using algebra. $\endgroup$ – Isaac YIU Math Studio Sep 11 at 8:53

A purely geometric way doesn't look likely, because the degrees of the cosine terms (two and three respectively) don't match. For what it's worth, here is an alternative trigonometric derivation.

Writing $2\cos B\cos C$ in the second term as $\cos (B+C)+\cos(B-C)$ transforms our expression to$$\cos^2A+\cos^2B+\cos^2C+[\cos(B+C)+\cos(B-C)]\cos A.$$Notice that $\cos(B+C)=-\cos A$, and use this conversion forwards and backwards to give$$\cos^2B+\cos^2C-\cos(B-C)\cos(B+C).$$Now write the last term as $-\frac12(\cos2B+\cos2C)$ and express the first terms also in double-angle format. Then cancellation yields the required result.

  • $\begingroup$ I slightly disagree with the first line since $\cos$'s are dimensionless. $\endgroup$ – Seewoo Lee Sep 11 at 3:01
  • 1
    $\begingroup$ @SeewooLee : Sure. But I haven't seen any essentially geometric derivations of results like this before. If you try hard enough, you can always work any trigonometric identity back into a geometric diagram, but the proof doesn't generally gain in clarity. $\endgroup$ – John Bentin Sep 11 at 6:20
  • $\begingroup$ Sorry, I want a geometric way. Therefore, this is an unavailable answer. $\endgroup$ – Isaac YIU Math Studio Sep 11 at 8:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.