4

I have a simple question : What are the condition on $a,b,c$ to have : $$\tan(a+b)\tan(b+c)\tan(a+c)=1$$ I think we can use the Huilier formula but I am not sure how to solve it. Thanks

Edit:

With : $\tan(a)\tan(b)<1$, $\tan(a)\tan(c)<1$, and $\tan(b)\tan(c)<1$.

My second idea would be to build a cyclic quadrilateral (convex in your case) with this formulas :

$\tan(\frac{\alpha}{2})\tan(\frac{\beta}{2})=\tan(\frac{\gamma}{2})\tan(\frac{\delta}{2})=1$

With $\alpha+\beta+\gamma+\delta=\pi$

On the other hand an other idea would be to use the semiperimeter :

For a cyclic quadrilateral with successive sides $a, b, c, d$ semiperimeter $s$, and angle A between sides $a$ and $d$,we have

$\tan(\frac{A}{2})=\sqrt{\frac{(s-a)(s-d)}{(s-b)(s-c)}}$

3 Answers3

3

You are looking for the solutions to $$ \bbox[lightyellow] { \left\{ \matrix{ \tan \left( {a + b} \right)\tan \left( {a + c} \right)\tan \left( {b + c} \right) = 1 \hfill \cr \tan \left( a \right)\tan \left( b \right),\;\tan \left( a \right)\tan \left( c \right),\;\tan \left( b \right)\tan \left( c \right) < 1 \hfill \cr} \right. } \tag{1}$$

The equation is symmetric in $a,b,c$, and the solutions will be centered around the symmetric solution $$ \bbox[lightyellow] { a = b = c = \pi /8 } $$

This hints to try and make the substitution $$ \bbox[lightyellow] { \left\{ \matrix{ a + b = \pi /4 + \gamma \hfill \cr b + c = \pi /4 + \alpha \hfill \cr a + c = \pi /4 + \beta \hfill \cr} \right.\quad \Leftrightarrow \quad \left\{ \matrix{ a = \pi /8 + \left( { - \alpha + \beta + \gamma } \right)/2 \hfill \cr b = \pi /8 + \left( {\alpha - \beta + \gamma } \right)/2 \hfill \cr c = \pi /8 + \left( {\alpha + \beta - \gamma } \right)/2 \hfill \cr} \right. } \tag{2}$$

Making this change of variables we get, for $tan(a+b)$ $$ \bbox[lightyellow] { \tan \left( {a + b} \right) = \tan \left( {\pi /4 + \gamma } \right) = {{1 + \tan \gamma } \over {1 - \tan \gamma }} } \tag{3.a}$$ while for $tan(a)tan(b)$ we get $$ \bbox[lightyellow] { \eqalign{ & \tan a\tan b = \tan \left( {\pi /8 + \gamma /2 - \left( {\alpha - \beta } \right)/2} \right)\tan \left( {\pi /8 + \gamma /2 + \left( {\alpha - \beta } \right)/2} \right) = \cr & = {{\tan \left( {\pi /8 + \gamma /2} \right) - \tan \left( {\left( {\alpha - \beta } \right)/2} \right)} \over {1 + \tan \left( {\pi /8 + \gamma /2} \right)\tan \left( {\left( {\alpha - \beta } \right)/2} \right)}}{{\tan \left( {\pi /8 + \gamma /2} \right) + \tan \left( {\left( {\alpha - \beta } \right)/2} \right)} \over {1 - \tan \left( {\pi /8 + \gamma /2} \right)\tan \left( {\left( {\alpha - \beta } \right)/2} \right)}} = \cr & = {{\tan ^{\,2} \left( {\pi /8 + \gamma /2} \right) - \tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right)} \over {1 - \tan ^{\,2} \left( {\pi /8 + \gamma /2} \right)\tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right)}} \cr} } $$

and imposing the condition that $tan(a)tan(b)<1$ we get $$ \bbox[lightyellow] { \eqalign{ & \tan a\tan b < 1\quad \to \; \cr & \to \;\;\tan ^{\,2} \left( {\pi /8 + \gamma /2} \right) - \tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right) < 1 - \tan ^{\,2} \left( {\pi /8 + \gamma /2} \right)\tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right)\;\; \to \cr & \to \;\;\tan ^{\,2} \left( {\pi /8 + \gamma /2} \right)\left( {1 + \tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right)} \right) < 1 + \tan ^{\,2} \left( {\left( {\alpha - \beta } \right)/2} \right)\;\; \to \cr & \to \;\;\tan ^{\,2} \left( {\pi /8 + \gamma /2} \right) < 1\;\; \to \cr & \to \;\; - 3\pi /4 < \gamma < \pi /4\quad \to \quad \tan \gamma < 1 \cr} } \tag{3.b}$$ and analogously for the other combinations.

So, the original equation is transformed into $$ \bbox[lightyellow] { \eqalign{ & \left\{ \matrix{ \tan \left( {a + b} \right)\tan \left( {a + c} \right)\tan \left( {b + c} \right) = 1 \hfill \cr \tan a\tan b,\;\tan a\tan c,\;\tan b\tan c < 1 \hfill \cr} \right.\quad \to \cr & \to \quad \left\{ \matrix{ a + b = \pi /4 + \gamma \hfill \cr b + c = \pi /4 + \alpha \hfill \cr a + c = \pi /4 + \beta \hfill \cr \left( {{{1 + \tan \alpha } \over {1 - \tan \alpha }}} \right)\left( {{{1 + \tan \beta } \over {1 - \tan \beta }}} \right)\left( {{{1 + \tan \gamma } \over {1 - \tan \gamma }}} \right) = 1 \hfill \cr - 3\pi /4 < \alpha ,\;\beta ,\;\gamma < \pi /4 \hfill \cr} \right.\quad \to \cr} } $$ $$ \bbox[lightyellow] { \to \quad \left\{ \matrix{ a = \pi /8 + \left( { - \alpha + \beta + \gamma } \right)/2 \hfill \cr b = \pi /8 + \left( {\alpha - \beta + \gamma } \right)/2 \hfill \cr c = \pi /8 + \left( {\alpha + \beta - \gamma } \right)/2 \hfill \cr 0 = \tan \alpha + \tan \beta + \tan \gamma + \tan \alpha \tan \beta \tan \gamma \hfill \cr \tan \alpha ,\;\tan \beta ,\;\tan \gamma < 1 \hfill \cr} \right. } \tag{4}$$

Conclusion

Equation (4) tells us that the solutions to (1) are reconducible to the roots of the symmetric polynomial $$ \bbox[lightyellow] { P(x,y,z) = x + y + z + xyz = x\left( {1 + yz} \right) + y + z } $$ which can be easily found by expressing one variable in terms of the other two, and which are rendered graphically by this plot.

Prod_3_Tan_1

Note the similarity of the solution to the above with the addition formula for $\tanh$ $$ \bbox[lightyellow] { x = - {{y + z} \over {1 + yz}}\quad \leftrightarrow \quad - \tanh \left( {c + d} \right) = - {{\tanh c + \tanh d} \over {1 + \tanh c\tanh d}} } $$ which would deserve to be further expanded.

Otherwise, dividing by one of the variables taken to be non-null (e.g., $z$), the solutions are reconducible to the points on the hyperbola $$ \bbox[lightyellow] { h(\xi ,\eta ;z) = z^{\,2} \xi \eta + \xi + \eta + 1 = 0 } $$ and we can use the well developed apparatus of conic sections to analyze the behaviour of the solutions (existence, axes, canonical form, etc.) and to find appropriated parametric representations.

example :

In equation (4), choosing $\tan \alpha = 0,\,\;\tan \beta = 1/2$ we get $\tan \gamma = - 1/2 $, and then

$$ \bbox[lightyellow] { \eqalign{ & \tan \alpha = 0,\,\;\tan \beta = 1/2,\;\tan \gamma = - 1/2 \cr & \quad \quad \Downarrow \cr & \alpha = 0,\,\;\beta = \arctan \left( {1/2} \right),\;\gamma = - \arctan \left( {1/2} \right) \cr & \quad \quad \Downarrow \cr & \left\{ \matrix{ a = \pi /8 \hfill \cr b = \pi /8 - \arctan \left( {1/2} \right) \hfill \cr c = \pi /8 + \arctan \left( {1/2} \right) \hfill \cr} \right. \cr & \quad \quad \Downarrow \cr & \tan \left( {a + b} \right)\tan \left( {a + c} \right)\tan \left( {b + c} \right) = \cr & = \tan \left( {\pi /4 - \arctan \left( {1/2} \right)} \right)\tan \left( {\pi /4 + \arctan \left( {1/2} \right)} \right)\tan \left( {\pi /4} \right) = \cr & = {{1 - 1/2} \over {1 + 1/2}}{{1 + 1/2} \over {1 - 1/2}} = 1 \cr} } $$

addendum

The similarity with the addition formula for $\tanh$ provides in fact an interesting parametric equation for $a,b,c$ in terms of $u,v$, which
for $-\infty<u,v<\infty$ always satisfies the original system (1) $$ \bbox[lightyellow] { \left\{ \matrix{ a = \pi /8 + \left( { - \alpha + \beta + \gamma } \right)/2 \hfill \cr b = \pi /8 + \left( {\alpha - \beta + \gamma } \right)/2 \hfill \cr c = \pi /8 + \left( {\alpha + \beta - \gamma } \right)/2 \hfill \cr \alpha = \arctan \left( {\tanh u} \right) \hfill \cr \beta = \arctan \left( {\tanh v} \right) \hfill \cr \gamma = \arctan \left( {\tanh \left( { - u - v} \right)} \right) = - \arctan \left( {\tanh \left( {u + v} \right)} \right) \hfill \cr} \right. } \tag{5}$$

In fact, we have $$ \bbox[lightyellow] { \left. {\left\{ \matrix{ 0 \equiv \tan \alpha + \tan \beta + \tan \gamma + \tan \alpha \tan \beta \tan \gamma \; \hfill \cr \tan \alpha ,\;\tan \beta ,\;\tan \gamma < 1\; \to \;\left\{ \matrix{ \tanh u,\;\tanh v < 1 \hfill \cr - 1 < \tanh \left( {u + v} \right) \hfill \cr} \right. \hfill \cr} \right.\quad } \right|\; - \infty < u,v < \infty } $$

G Cab
  • 35,272
  • Your answer seems the most complete and clear. Also the simplicity of the solution is elegant and I think there is nothing wrong. Perhaps I would have a few questions. –  Jun 14 '17 at 17:44
  • @FatsWallers: glad for the appreciation. I cleaned up a bit the exposition to make it more fluid. If you have any doubt/ question I'll do my best to answer. Meanwhile I have a question on my side: where does the limitation $tan(a)tan(b)<1$ come from ? (apart from the case in which it is $=1$) – G Cab Jun 14 '17 at 21:17
  • I saw: much interesting – G Cab Jun 15 '17 at 14:32
  • @FatsWallers: have a look at the addendum: I suppose it will be much of help for your further application. – G Cab Jun 19 '17 at 15:03
0

By expansion, we get

$$\frac{(\tan(a)+\tan(b))(\tan(b)+\tan(c))(\tan(c)+\tan(a))}{(1-\tan(a)\tan(b))(1-\tan(b)\tan(c))(1-\tan(c)\tan(a))}$$

And as you wrote the multiplication of the terms $\tan(a)\tan(b)<1$ and so on therefore, the denominator is always less than $1$ and hence the whole terms are greater than one. So all the terms $\tan(a+b)\tan(b+c)\tan(c+a)$ have to be $1$. So, $$\tan(a+b)=\tan(b+c)=\tan(c+a)=1$$.

MAN-MADE
  • 5,381
0

$$\tan(a+b)\tan(b+c)\tan(a+c)=1 \tag 1$$ $$\tan(b+c)\tan(a+c)=\frac{\cos(b-a)-\cos(b+a+2c)}{\cos(b-a)+\cos(b+a+2c)}$$ $$\tan(a+b)\frac{\cos(b-a)-\cos(b+a+2c)}{\cos(b-a)+\cos(b+a+2c)}=1$$ This can be solved for $\cos(b+a+2c)$ $$\cos(b+a+2c)=\frac{\tan(a+b)-1}{\tan(a+b)+1}\cos(a-b)$$ $$c=\frac{1}{2}\left(-a-b+\cos^{-1}\left(\frac{\tan(a+b)-1}{\tan(a+b)+1}\cos(a-b) \right) \right) \tag 2$$ If we chose arbitrary values of $a$ and $b$, Eq.(2) gives $c$ so that Eq.(1) is satisfied. But $c$ can be complex. To obtain $c$ real, the condition is : $$-1\leq \frac{\tan(a+b)-1}{\tan(a+b)+1}\cos(a-b) \leq 1$$

On the graph, the areas allowed for $(a,b)$ to achieve $c$ real are drawn in green. The whole is periodic as expected.

enter image description here

NUMERIAL EXAMPLE :

For example, $a=0.2$ and $b=0.75$

$\frac{\tan(a+b)-1}{\tan(a+b)+1}\cos(a-b)\simeq 0.141608$

$c=\frac{1}{2}\left(-a-b+\cos^{-1}\left(\frac{\tan(a+b)-1}{\tan(a+b)+1}\cos(a-b) \right) \right)\simeq 0.239355$

$\tan(a+b)\tan(b+c)\tan(a+c)\simeq 1.000000$

$\cos^{-1}$ is multi-valuated. Other convenient values of $c$ are obtained, for example $c\simeq 1.952237$

Of course, a numerical example isn't a proof. The analytic proof is above. This shows that one can obtain an infinity of solutions with the above method.

JJacquelin
  • 66,221
  • 3
  • 37
  • 87