20

A triangle is inscribed in a circle so that three congruent circles can be inscribed in the triangle and two of the segments. Each circle is the largest circle that can be inscribed in its region.

Triangle inscribed in circle, with three inscribed congruent circles

Keeping the shortest side (in red) fixed, move the vertex opposite the shortest side along the circle. As before, each inscribed circle is the largest circle that can be inscribed in its region. Do not let the triangle become degenerate.

Triangle inscribed in circle, with three inscribed non-congruent circles

Show that the radii of the inscribed circles are always in a geometric sequence.


I made a Geogebra page where you can move the vertex and observe the circles.

I will post my solution, which involves complicated algebra. I am looking for a more elegant solution.

Ellie K
  • 486
Dan
  • 35,053
  • 2
    You should say right from the beginning that this property is only valid for a unique ratio : "chord length"/"radius of circumcircle". – Jean Marie May 25 '25 at 06:57
  • 3
    @JeanMarie At the beginning, I state that there are three congruent inscribed circles. This defines a unique ratio of the length of the red chord to the radius of the circumcircle. – Dan May 25 '25 at 07:18
  • Thanks for your answer, but what do you mean by adjective "congruent" in this context ? – Jean Marie May 25 '25 at 07:25
  • @JeanMarie By "congruent", I mean they have the same radius. – Dan May 25 '25 at 07:31
  • Will you first amend the Title to include the triangle? Without the triangle, how should I understand the Question? – Robbie Goodwin May 25 '25 at 23:30
  • And please add an explanation of congruent in this particular context. – Trunk May 27 '25 at 15:26
  • 2
    Speaking of circles, what else could "congruent" mean? – lesath82 Jun 03 '25 at 19:15

5 Answers5

11

I have a proof using trigonometry. I label the points as the following figure.

incircle

Let's denote the inradius by $r$ and the radii of the three circles by $t_{A}, t_{B}, t_{C}$.

The key formulas in this proof are: $$r = 4R\sin\dfrac{A}{2}\sin\dfrac{B}{2}\sin\dfrac{C}{2}$$ and $$t_{A} = R{\left(\sin\dfrac{A}{2}\right)}^{2}, t_{B} = R{\left(\sin\dfrac{B}{2}\right)}^{2}, t_{C} = R{\left(\sin\dfrac{C}{2}\right)}^{2}$$

I guess I should provide a proof for these formulas.

From the incenter-excenter lemma, $D$ is the circumcenter of triangle $IBC$. $$IB\cdot IC\cdot\sin\angle BIC = 2\cdot \text{area}(IBC) = r\cdot BC$$

On the other hand

$$\sin\angle BIC = \sin\left(\dfrac{\pi}{2} + \dfrac{A}{2}\right) = \sin\left(\dfrac{\pi}{2} - \dfrac{A}{2}\right) = \cos\dfrac{A}{2}$$

and

$$BC = 2R\sin A = 4R \sin\dfrac{A}{2}\cos\dfrac{A}{2}, IB = \dfrac{r}{\sin\dfrac{B}{2}}, IC = \dfrac{r}{\sin\dfrac{C}{2}}$$

so we obtain that

$$r = 4R \sin\dfrac{A}{2}\sin\dfrac{B}{2}\sin\dfrac{C}{2}.$$

According to the law of sine

$$DX = DB\cos\dfrac{\pi - A}{2} = 2R\sin\dfrac{A}{2}\sin\dfrac{A}{2}$$

so $t_{A} = R{\left(\sin\dfrac{A}{2}\right)}^{2}$.

Back to the original problem.

$$ \begin{align*} r^{2} = t_{B}t_{C} & \iff 16R^{2}{\left(\sin\dfrac{A}{2}\right)}^{2}{\left(\sin\dfrac{B}{2}\right)}^{2}{\left(\sin\dfrac{C}{2}\right)}^{2} = R^{2} {\left(\sin\dfrac{B}{2}\right)}^{2}{\left(\sin\dfrac{C}{2}\right)}^{2} \\ & \iff {\left(\sin\dfrac{A}{2}\right)}^{2} = \dfrac{1}{16} \\ & \iff \sin\dfrac{A}{2} = \dfrac{1}{4} \end{align*} $$

This implies the desired result.

Duong Ngo
  • 2,159
  • Lovely perspective but you can't just assert that $r^2 = t_B;t_C$ as this is what we are are trying to prove ! You could use your reasoning above after you evaluate the value of angle $A$ from the geometry of the congruent circles case. To do this consider the congruent circles and use the expressions, $b = r \cot(A/2) + r\cot(B/2) ;$, $A/2 = 90-B$ and $;b = 2R\sin{B}$ to deduce a value for the angle $B/2$. This leads to a value for angle $A/2$. Now compare expressions for $r^2$ and $t_B;t_C$ and deduce a geometric sequence. Still a hindsight proof but thanks for the lemma. – Trunk Jun 02 '25 at 19:27
  • I didn't assert that. Instead, I showed that $r^{2} = t_{B}t_{C}$ is equivalent to something else. – Duong Ngo Jun 03 '25 at 00:35
  • Okay, so $r^2 = t_B;t_C \Leftrightarrow \sin(A/2) = 1/4$ . . . How does this imply the desired result unless you already know that $\sin(A/2) = 1/4$ is correct ? Please elaborate. – Trunk Jun 03 '25 at 10:54
  • 1
    If $\triangle ABC$ is $A$-isosceles and $\sin\dfrac{A}{2} = \dfrac{1}{4}$ then $r^{2} = t_{B}t_{C}$. Moreover, the equivalence implies that for any point $A^{\prime}$ such that $\angle BAC = \angle BA^{\prime}C$, one also has $r^{2} = t_{B}t_{C}$. – Duong Ngo Jun 03 '25 at 11:56
  • "If $\triangle ABC$ is A-isosceles and $\sin{\dfrac{A}{2}}= \dfrac{1}{4}$ . . ." But that is just the point. You had not found the value of A before moving on to the general (non-isosceles) situation. The value of A is readily deducible but is not obvious. Even if it were, it would merit noting that fact at the outset so that readers could follow your thinking. – Trunk Jun 03 '25 at 12:18
  • The point is a triangle $ABC$ such that $\sin\dfrac{A}{2} = \dfrac{1}{4}$ does exist and the proof doesn't rely on a fixed figure. – Duong Ngo Jun 03 '25 at 12:28
  • Yes, a triangle $\Delta ABC$ with $\sin{\dfrac{A}{2}} = \dfrac{1}{4} $ can exist. Yet the fact that it can exist does not mean that the value of $\sin{\dfrac{A}{2}}$ in this case is actually $\dfrac{1}{4}$. All you have deduced here is that the assertion $t_B;t_C = r^2$ is not disproved by an absurd implication - a different thing to proving it. Since this geometric sequence between $r_a$, $r$ and $r_c$ is only observed for a particular value of $\dfrac{b}{R}$ then the proof does indeed rely on a particular vale of $\sin{\dfrac{A}{2}}$. – Trunk Jun 03 '25 at 15:27
  • Alternatively, you could prove that a triangle within a circle such that its $A$ angle satisfies the expression $\sin{\dfrac{A}{2}} = \dfrac{1}{4}$ will have inscribed circles of equal radius. But you just can't say from your derivation that having found a triangle which produces a set of circles whose radii are in geometric sequence, this triangle's base side $b$ is the same as that in a triangle which will produce congruent circles. It might be the case that the two triangles have different base lengths . . . – Trunk Jun 04 '25 at 10:50
  • Apologies. I get it at last. The geometric sequence ratio between $t_A, r, t_C$ is maintained through the full range of allowable $A$ angles. This must therefore be true when $\triangle ABC$ is isosceles where $t_A = t_C$. So in the isosceles situation the radius of the circle in the middle of $\triangle ABC$, i.e. $r$, must satisfy the expression $ \dfrac{t_A}{r} = \dfrac{r}{t_C} $ and this means that $t_A = r = t_C$ here. hence your proof also covers the congruent circles case.. – Trunk Jun 04 '25 at 17:31
9

The result generalizes a bit. Since other answers provide detailed proofs of the original result, I'll just hit the highlights.

Let $P$ be an interior point of $\triangle ABC$. Specifically, we can write $$P = \frac{u A + v B + w C}{u+v+w}$$ for positive real constants $u$, $v$, $w$ (that we'll consider fixed throughout). Let lines $PB$ and $PC$ meet the sides $AC$ and $AB$ at $B'$ and $C'$. Let circles tangent to $AC$ and $AB$ at $B'$ and $C'$ also be tangent to the circumcircle (while not overlapping the triangle itself). Let $b$ and $c$ be the radii of these circles, and let $i$ be the triangle's inradius.

Then (barring degeneracies) $b$, $i$, $c$ are in geometric progression (that is, $i^2 = bc$) when $$\sin^2 \frac12A = \frac{u^2vw}{(u+v)^2(u+w)^2}\tag{$\star$}$$

Since $(\star)$ only involves angle $A$, if one finds any instance where the geometric progression property holds (eg, when $b=i=c$), then the property holds so long as $\angle BAC$ is an inscribed angle of fixed size.

enter image description here

With some Mathematica-assisted coordinate bashing, I find that the tangent circles have radii given by $$ b = \frac{4 r uw \sin^2B_2}{(u + w)^2} \qquad c = \frac{4 r uv \sin^2C_2}{(u + v)^2} \tag1$$ where $r$ is the circumradius of $\triangle ABC$, and $X_2 := X/2$ (to save typing and reduce visual clutter). Moreover, we "know" that the inradius is given by $$i = 4r\sin A_2\sin B_2\sin C_2 \tag2$$ Thus, radii $b$, $i$, $c$ are in geometric progression iff $$0 = b c-i^2 = 16r^2\sin^2B_2\sin^2C_2\left( \frac{u^2vw}{(u+v)^2(u + w)^2} - \sin^2A_2 \right) \tag3$$ from which $(\star)$ follows (since $\sin B_2$ and $\sin C_2$ are non-zero in a non-degenerate triangle). $\square$


In the question as stated, we note that the tangent circles touch the midpoints of the triangle's sides. Consequently, in the notation of this answer, $BB'$ and $CC'$ are medians, making $P$ the centroid, whence $u=v=w$. In this case, $(\star)$ reduces to $\csc A_2 = 4$, echoing the result in @Duong Ngo's answer.


There's actually no requirement that $P$ be interior to the triangle. The result holds for configurations as shown:

enter image description here


It's a bit intriguing that we can re-write $(1)$ as $$ \frac{1}{b} = \frac{1}{4r}(u + w)\left(\frac1u+\frac1w\right)\csc^2B_2 \qquad \frac{1}{c} = \frac{1}{4r}(u+v)\left(\frac1u+\frac1v\right)\csc^2C_2 \tag{1'}$$ and $(\star)$ as $$\csc^2 A_2 = (u+v)\left(\frac1u+\frac1v\right)\,(u+w)\left(\frac1u+\frac1w\right) \tag{$\star'$}$$ I wonder if there's some geometric significance to these forms.

Update. There's this: Values $u$, $v$, $w$ can be considered the barycentric coordinates of point $P$, typically expressed as $u:v:w$ (since the proportion is all that matters). Given $P$, the point with bary-coordinates $\frac1u:\frac1v:\frac1w$ is called the isotomic conjugate of $P$. It's the point whose corresponding $B'$ is the reflection of $B'$ in the midpoint of $AC$; likewise for $C'$ (and the appropriately-defined $A'$). "Obviously", the tangent circles touching these points must be congruent to the respective tangent circles for $P$. Consequently, we should expect the formulas for their radii to be unchanged by the substitutions $(u,v,w)\to(1/u,1/v,1/w)$.

So, that helps explain the forms of $(1')$ and $(\star')$, albeit not their underlying geometry. (Perhaps it would be instructive to think of the product in, for instance, the $b$ formula as $(u+w)\csc B_2 \cdot \left(\frac1u+\frac1w\right)\csc B_2$.)


Another tidbit: We can eliminate angle references in the formulas for $b$, $c$, $i$, and the corresponding radius $a$ for a tangent circle along side $BC$. This gives a general relation among the four radii:

$$2 \;+\;\widehat{a} \;+\;\widehat{b} \;+\;\widehat{c} \;-\;\widehat{a}\,\widehat{b}\,\widehat{c} \;=\; 0 \tag{5}$$

where, leaning into my previous observation, I define cyclically $$ \widehat{a}:=\frac{a}{i}(v+w)\left(\frac1v+\frac1w\right) \qquad \widehat{b}:=\cdots\qquad \widehat{c}:=\cdots\tag6$$ Hmmm ... That $2$ seems out of place. Perhaps this form of the equation is better: $$ \left(1+\widehat{a}\right) \;+\;\left(1+\widehat{b}\right) \;+\;\left(1+\widehat{c}\right) \;=\;1+\widehat{a}\,\widehat{b}\,\widehat{c} \tag{5'}$$

Blue
  • 83,939
6

Self-answering. This solution involves complicated algebra. I am looking for a more elegant solution.


Assume that the largest circle (in which the triangle is inscribed) is a unit circle.

The length of the fixed side of the triangle

Consider the diagram with the isosceles triangle.

Let:

  • $a=$ half the length of the fixed side of the triangle
  • $L=$ length of one of the longer sides of the isosceles triangle
  • $r=$ radius of the three inscribed circles

Isosceles triangle inscribed in circle

Pythagorus gives

$$L=\sqrt{a^2+\left(\sqrt{1-a^2}+1\right)^2}=\sqrt{2+2\sqrt{1-a^2}}$$

We have

$$r=\frac{2\times(\text{area of triangle})}{\text{perimeter of triangle}}$$

$$\frac12\left(1-\sqrt{1-\left(\frac{L}{2}\right)^2}\right)=\frac{2a\left(\sqrt{1-a^2}+1\right)}{2a+2L}$$

which is satisfied by $a=\frac{\sqrt{15}}{8}$. (So the length of the fixed side of the triangle is $2a=\frac{\sqrt{15}}{4}$, and $L=\frac{\sqrt{15}}{2}$.)

A relationship between the other two sides

Now consider the variable triangle.

Let the two variable side lengths be $2x$ and $2y$, where $x\le y$ (so $0<x\le\frac{L}{2}$).

Inscribed non-isosceles triangle

The angles at the centre of the circle subtended by the three sides are $2\arcsin x,\space 2(\pi-\arcsin y),\space 2\arcsin a$.

$$2\arcsin x+2(\pi-\arcsin y)+2\arcsin a=2\pi$$

$$y=\sin({\arcsin x+\arcsin a})=x\sqrt{1-a^2}+a\sqrt{1-x^2}$$

Expressions for the three radii

The radii of the circles in the segments are

$$R_1=\frac12\left(1-\sqrt{1-x^2}\right)$$

$$R_2 = \begin{cases} \frac12\left(1+\sqrt{1-y^2}\right) & \text{if }\space 0<x\le\frac78 \\ \frac12\left(1-\sqrt{1-y^2}\right) & \text{if }\space \frac78<x\le\frac{L}{2} \end{cases}$$

(There are two expressions for $R_2$ because when $x=\frac78$, the longest side of the triangle is a diameter of the circle, so when $x$ increases through $\frac78$, the circle in the segment bounded by the longest side changes from being in the major segment to being in the minor segment.)

The radius of the circle in the variable triangle is (again)

$$R_3=\frac{2\times(\text{area of triangle})}{\text{perimeter of triangle}}$$

Using Heron's formula,

$$R_3=\frac{\sqrt{(x+y+a)(-x+y+a)(x-y+a)(x+y-a)}}{x+y+a}=\sqrt{\frac{(-x+y+a)(x-y+a)(x+y-a)}{x+y+a}}$$

Showing that the radii are in a geometric sequence

Then we only need to show that $R_1R_2={R_3}^2$ for $0<x\le\frac{L}{2}$. This can be checked by a graph.

Dan
  • 35,053
1

Inscribed circles diagram, congruent and general

As Jean Marie noted, this geometric series relationship between $r_a, \; r$ and $r_c$ only exists for triangles standing on a unique chord of the circle, i.e. that chord upon which the congruent inscribed circles results. Refer to the diagram on the left above.

Determining the properties of the general case, we can easily deduce from the big circle geometry:

$$ r_a = \dfrac{R}{2} (1 - \cos{A}) = R \sin^2{\dfrac{A}{2}}$$

$$ r_b = \dfrac{R}{2} (1 - \cos{B}) = R \sin^2{\dfrac{B}{2}}$$

$$ r_c = \dfrac{R}{2} (1 - \cos{C}) = R \sin^2{\dfrac{C}{2}}$$

where $R$ is the outer circle's radius.

And $\;\;\; a = 2R\;\sin{A} \;\;\;\; b = 2R\; \sin{B} \;\;\;\; c = 2R \;\sin{C}$

Looking at the inscribed circle of radius $r$, we can write: $ \;\;a = r \cot{\dfrac{B}{2}} + r \cot{\dfrac{C}{2}} $

$$ \implies r = \frac{2R\sin{A}}{\cot{\dfrac{B}{2}} + \cot{\dfrac{C}{2}} } $$

$$ $$

The congruent situation first.

Clearly for $r_a = r_c = r\;$ we must have $ \;a = c \;$ and thus $ \; A = C$.

Thus $\; B = 180^{\text{o}} - 2A \;\;$ or $\;\; \dfrac{B}{2} = 90^{\text{o}} - A $.

$$ r_a = r \;\; \implies \;\; R \sin^2{\dfrac{A}{2}} = \frac{2R\sin{A}}{\cot{\dfrac{B}{2}} +\cot{\dfrac{C}{2}}} = \frac{4R\sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}}}{\tan{A} +\cot{\dfrac{A}{2}}} $$

$$ \implies \tan{\dfrac{A}{2}} = \frac{4}{\tan{A} + \cot{\dfrac{A}{2}} } $$

$$ \implies \tan{\dfrac{A}{2}} \; (\frac{2 \tan{\dfrac{A}{2}}}{1 - \tan^2{\dfrac{A}{2}}}) + 1 = 4 $$

$$ \implies 2\; \tan^2{\dfrac{A}{2}} = 3 - 3 \; \tan^2{\dfrac{A}{2}} \; \implies \tan{\dfrac{A}{2}} = \sqrt{\dfrac{3}{5}} $$

$$ \implies \tan{A} = \dfrac {2\; \sqrt{\dfrac{3}{5}} } { (1 - \dfrac{3}{5}) } = \sqrt{15} \;\; \implies \; \cos{A} = \dfrac{1}{4} = 75.5225 ^{\text{o}} $$

Hence since $\sin{\dfrac{A}{2}} = \sqrt{\dfrac{3}{8}} \;\;$ we have $\;\; r_a = r_c = r = R \; \sin^2{\dfrac{A}{2}} = \dfrac{3R}{8} $

The general situation when $b$ is fixed at its congruent length.

Keeping $b$ fixed is effectively keeping angle $B$ fixed at its congruent value of $B = 180^{\text{o}} - 2 \cos^{-1}{\dfrac{1}{4}} = 28.955^{\text{o}} $. This means that changes to angles $A$ and $C$ come at the expense of each other, i.e.

$ A + C = 2A_o \;\; \implies \;\; \dfrac{C}{2} = A_o - \dfrac{A}{2} \;\; $ where $\;\; A_o = \cos^{-1}{\dfrac{1}{4}} $

$$ r = \dfrac {2R \sin{A}} { \cot{\dfrac{B}{2}} + \cot{\dfrac{C}{2}} } = \dfrac { 4R \sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}} } { \sqrt{15} + \cot{(A_o - \dfrac{A}{2}}) } = \dfrac { 4R \sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}} } { \sqrt{15} + \dfrac{ 1 + \sqrt{15} \tan{\dfrac{A}{2}} } { \sqrt{15} - \tan{\dfrac{A}{2}} } } $$

$$ \require{cancel} = 4R \sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}} \; (\dfrac { \sqrt{15} - \tan{\dfrac{A}{2}} } { 15 - \cancel {\sqrt{15} \tan{\dfrac{A}{2}} } + 1 + \cancel {\sqrt{15} \tan{\dfrac{A}{2}} } } ) $$

$$ \implies \; r = \dfrac { 4R \sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}} (\sqrt{15} - \tan{\dfrac{A}{2}} ) } {16} = \dfrac{R}{4} \sin{\dfrac{A}{2}} \cos{\dfrac{A}{2}} (\sqrt{15} - \tan{\dfrac{A}{2}} ) $$

$$ \boldsymbol{ \require{cancel} \dfrac{r_a}{r} } = \dfrac{ \cancel{R} \sin^\cancel{2}{\dfrac{A}{2}} } { \dfrac{\cancel{R}}{4} \cancel{\sin{\dfrac{A}{2}} } \cos{\dfrac{A}{2}} (\sqrt{15} - \tan{\dfrac{A}{2}} ) } = \boldsymbol{ \dfrac{4 \tan{\dfrac{A}{2}}} {\sqrt{15} - \tan{\dfrac{A}{2}}} } $$

$$ $$

$$ r_c = R\;\sin^2{\dfrac{C}{2}} = R\;(\sin{A_o} \cos{\dfrac{A}{2}} - \cos{A_o} \sin{\dfrac{A}{2}})^2 $$

Recalling that $\;\; \sin{A_o} = \dfrac{\sqrt{15}}{4} $ and $ \cos{A_o} = \dfrac{1}{4} $ we have:

$$ r_c = R\;\cos^2{\dfrac{A}{2}} (\dfrac{\sqrt{15}}{4} - \dfrac{1}{4}\;\tan{\dfrac{A}{2}})^2 = \dfrac{R}{16}\; \cos^2{\dfrac{A}{2}}\;(\sqrt{15} - \tan{\dfrac{A}{2}})^2 $$

$$ \require{cancel} \boldsymbol{\dfrac{r}{r_c}} = \dfrac{\dfrac{\cancel{R}}{4} \sin{\dfrac{A}{2}} \cancel{\cos{\dfrac{A}{2}} } \cancel{(\sqrt{15} - \tan{\dfrac{A}{2}})}} {\dfrac{\cancel{R}}{16}\; \cos^\cancel{2}{\dfrac{A}{2}}\;(\sqrt{15} - \tan{\dfrac{A}{2}})^\cancel{2}} = \boldsymbol{ \dfrac{4 \tan{\dfrac{A}{2}}} {\sqrt{15} - \tan{\dfrac{A}{2}}} } $$

So we have:

$$ \boldsymbol{ r_a:r \; = \; r:r_c \; = \; \dfrac{ 4 \tan{\dfrac{A}{2}} } {\sqrt{15} - \tan{\dfrac{A}{2}} } } $$

which corresponds to a geometric sequence $a, ar, ar^2, \dots \; $ if we have them in the ascending order $ r_c, r, r_a$.

The sequence ratio of

$$\; \dfrac{ 4 \tan{\dfrac{A}{2}} } { \sqrt{15} - \tan{\dfrac{A}{2}} }\;$$

collapses to unity (though still mathematically a geometric sequence with ratio $1$ !) at the congruent situation where $A = C = A_o $ and it rises to $\; +\infty \;$ as $A \; \rightarrow 2A_o = 151.045^\text{o} \;$ and then increases from $ -\infty $ after this.

I know that this is far from the elegant solution sought by Dan. I have a feeling that I am missing a geometrical perspective that might simplify things. I am also curious about the neglected small circle of radius $r_b$. But we must solve a problem elaborately before we find the elegant solution . . .

Trunk
  • 442
0

Triangle Inscribed Circles

This may be long late answer, but one realizes that even if the question is about a geometric sequence, the real question would be the following: Given a circle of center $O$, radius $R$ and a chord $[AB]$; let $C$ be a point on some arc $\overparen{AB}$. If $M$ is the midpoint of $[AC]$ and $N$ the midpoint of $[BC]$ then $$\dfrac{(R-\overline{OM})(R-\overline{ON})}{r^2}$$ is a constant $k$ independent of the position of $C$, $r$ is the inradius of triangle $ABC$. The algebraic measure $\overline{ON}$ is negative if the angle at $A$ become obtuse and the same for $\overline{OM}$ at vertex $B$. Coordinates circle

Here for the particular case, one find some configuration for wich the constant $k$ is $4$.

A proof is presumed computational because we have the inradius and circumradius in one expression, taking $R=1$, in a coordinates system, the coordinates are as follows: $A(-a;-\sqrt{1-a^2})$, $B(a;-\sqrt{1-a^2})$, $C(-b;\pm\sqrt{1-b^2})$, $a>0$ and $b\ge 0$. We will take $C(-b;\sqrt{1-b^2})$ (the same goes for the other case). Line $(AC)$ has the equation: $$(a-b)y-(\sqrt{1-a^2}+\sqrt{1-b^2})x-b\sqrt{1-a^2}-a\sqrt{1-b^2}=0$$

Line $(BC)$ has the equation: $$(-a-b)y-(\sqrt{1-a^2}+\sqrt{1-b^2})x-b\sqrt{1-a^2}+a\sqrt{1-b^2}=0$$ Now the distance of $O$ to $(AC)$ equals: $$\dfrac{a\sqrt{1-b^2}+b\sqrt{1-a^2}}{\sqrt{2-2ab+2\sqrt{(1-a^2)(1-b^2)}}}=d$$

The distance of $O$ to $(BC)$ equals: $$\dfrac{|-a\sqrt{1-b^2}+b\sqrt{1-a^2}|}{\sqrt{2+2ab+2\sqrt{(1-a^2)(1-b^2)}}}$$ Set $$h=\dfrac{a\sqrt{1-b^2}-b\sqrt{1-a^2}}{\sqrt{2+2ab+2\sqrt{(1-a^2)(1-b^2)}}}$$ $$r^2=\dfrac{4a^2\left(\sqrt{1-a^2}+\sqrt{1-b^2}\right)^2}{\left(2a+\sqrt{2-2ab+2\sqrt{(1-b^2)(1-a^2)}}+\sqrt{2+2ab+2\sqrt{(1-b^2)(1-a^2)}}\right)^2}=\dfrac{4S^2}{P^2}$$ $S$ is the area of $ABC$ and $P$ its perimeter. Finally check that $$\dfrac{(1-h)(1-d)}{r^2}$$ is independent of $b$. Taking $b=0$ one needs to show that the previous fraction is equal to $\dfrac{1+\sqrt{1-a^2}}{2a^2}$.

Notice that $\sqrt{2+2ab+2\sqrt{(1-b^2)(1-a^2)}}=\sqrt{(1+a)(1+b)}+\sqrt{(1-a)(1-b)}$ Doing some algebra the identity to prove is $$4\left(\sqrt{1-a^2}+\sqrt{1-b^2}\right)^3(1+\sqrt{1-a^2})=\left(2a+(\sqrt{1+b}+\sqrt{1-b})(\sqrt{1-a}+\sqrt{1+a})\right)^2\left(2\sqrt{1-a^2}+2\sqrt{1-b^2}+a^2-b^2+(b-a)(\sqrt{(1+b)(1+a)}+\sqrt{(1-b)(1-a)})-(a+b)(\sqrt{(1-b)(1+a)}+\sqrt{(1+b)(1-a)})\right) $$ which is true.

Also, to verify check that $$a^2-b^2=(-\sqrt{1-a^2}+\sqrt{1-b^2})(\sqrt{1-a^2}+\sqrt{1-b^2})$$ and that $$(a+b)(\sqrt{(1-b)(1+a)}+\sqrt{(1+b)(1-a)})=(\sqrt{1-a^2}+\sqrt{1-b^2})(\sqrt{(1+a)(1+b)}-\sqrt{(1-a)(1-b)})$$ Simplify by $(\sqrt{1-a^2}+\sqrt{1-b^2})$ to get:

$$4\left(\sqrt{1-a^2}+\sqrt{1-b^2}\right)^2(1+\sqrt{1-a^2})=\left(2a+(\sqrt{1+b}+\sqrt{1-b})(\sqrt{1-a}+\sqrt{1+a})\right)^2\left(2-\sqrt{1-a^2}+\sqrt{1-b^2}-\sqrt{(1+b)(1+a)}+\sqrt{(1-b)(1-a)}-\sqrt{(1-b)(1+a)}+\sqrt{(1+b)(1-a)}\right) $$ The factors are squares as $2(1+\sqrt{1-a^2})=(\sqrt{1-a}+\sqrt{1+a})^2$ and $$2\left(2-\sqrt{1-a^2}+\sqrt{1-b^2}-\sqrt{(1+b)(1+a)}+\sqrt{(1-b)(1-a)}-\sqrt{(1-b)(1+a)}+\sqrt{(1+b)(1-a)}\right)=\left((\sqrt{1+b}+\sqrt{1-b})-(\sqrt{1+a}-\sqrt{1-a})\right)^2$$

Taking square root, we prove: $$2\left(\sqrt{1-a^2}+\sqrt{1-b^2}\right)(\sqrt{1-a}+\sqrt{1+a})=\left(2a+(\sqrt{1+b}+\sqrt{1-b})(\sqrt{1-a}+\sqrt{1+a})\right)\left((\sqrt{1+b}+\sqrt{1-b})-(\sqrt{1+a}-\sqrt{1-a})\right) $$

Computations are checked with GeoGebra.

Toni Mhax
  • 1,362
  • 1
    FYI: Your argument simplifies significantly when re-cast trigonometrically. In particular, one readily shows, without the need for coordinates, $|OM|=R\cos A$ and $|ON|=R\cos B$, so that your target expression immediately reduces thusly (defining $X_2:=X/2$): $$\frac{R^2(1-\cos A)(1-\cos B)}{r^2}=\frac{R^2\cdot 2\sin^2A_2\cdot 2\sin^2 B_2}{16R^2\sin^2A_2\sin^2B_2\sin^2C_2}=\frac14\csc^2C_2$$ which is independent of $A$ and $B$, hence remains fixed as $C$ varies along a given arc of a (circum)circle with chord $AB$. (I lose the thread of your discussion at "Notice that ...") – Blue Jun 04 '25 at 01:18
  • @Blue, yes ty I doubted that, I had a trig proof with tangents, but it needed some work, sure simpler than this. ( In notation $|OM|=R\cos(B)$). However I tempted this because I found the ratio of side $AB=2a$ to the circumradius $R=1$ explicitly. If you notice the formula $\dfrac{1+\sqrt{1-a^2}}{2a^2}=k$. – Toni Mhax Jun 04 '25 at 02:45
  • The argument is just heavy expansion, I tried to make it complete. And I am not very in the trig formulas for the triangle. Anyway thank you, that was on point@Blue – Toni Mhax Jun 04 '25 at 02:50
  • It's not the elegant solution sought by Dan. Even if it had a suitable diagram with it. Neither is mine, of course - just a long derivation lacking knowledge of the incenter-excenter lemma. I don't know why they call it a lemma. It stands on things that are theorems from Euclid's Elements . . . – Trunk Jun 04 '25 at 11:32
  • 1
    @Trunk, the comment of Blue is a good solution (elegant?). I just wanted to prove a different approach that seems to hold. The problem here is a particular case. (The rest I did not understand what you mean for the lemma). – Toni Mhax Jun 04 '25 at 12:55
  • 1
    @ToniMhax: I often begin investigations with brute-force coords, in hopes that resulting eqns will hint at cleaner approaches. My go-to setup for inscribed $\triangle ABC$ leverages the Inscribed Angle Thm to write $$C=(R,0)\qquad A=(R\cos2B,R\sin2B)\qquad B=(R\cos(-2A),R\sin(-2A))$$ Various aspects of the triangle reduce pretty nicely; in particular here, the eqns of lines through $C$: $$CA:;x\cos B+y\sin B= R\cos B\qquad CB:;x\cos A-y\sin A=R\cos A$$ Simplifying trig can get tricky (and rewarding!), but potentially avoiding rampant square roots and sign issues is quite a benefit. Cheers! – Blue Jun 04 '25 at 13:32
  • I've added a sketch for the situation as you set it out. I have made no attempt to put in coordinates - that's too much like work and besides Blue's use of trig algebra is much cooler. You say: $$\dfrac{(R-\overline{OM})(R-\overline{ON})}{r^2}$$ is a constant $k$ independent of the position of $C$ . . . Is this another lemma unknown to me or is it just an assertion à la Duong Ngo ? If you are multiplying the radii rather than diameters of the inscribed circles then surely it should be $\dfrac{(R-\overline{OM})}{2} ; \dfrac{(R-\overline{ON})}{2}$ ? – Trunk Jun 04 '25 at 17:06
  • @Trunk, ty I will add some figure if it is better, (should be simple). Yes I multiplied the diameters so the constant $k$ is multiplied by $4$. The proof gives $k=\dfrac{1+\sqrt{1-a^2}}{2a^2}$ independent of $-b$ (abscissa of $C$). – Toni Mhax Jun 04 '25 at 19:38
  • I thought you were going to dump the long algebra and use Blue's handy trig proof. I can't see anyone wading through as it stands. – Trunk Jun 04 '25 at 20:42