2

I'm trying to construct the trigonometric functions in a analytic-geometrically way without using the concepts of integral and derivative. So I came up with the idea of using the arclength measure, but without mentioning it.

I started introducing the equation of the unit circumference:

$$x^2+y^2=1,$$

and the function $\gamma:[-1,1]\rightarrow\mathbb{R}$ defined as $\gamma(t)=\sqrt{1-t^2}$. Then, I introduce the concept of a partition of an interval and defining the length of $\left.\gamma\right|_{[a,b]}$ where $[a,b]\subset[-1,1]$ is a non-degenerate interval with respect to $P\in\mathcal{P}([a,b])$.

After that, I proved that for any $x\in[-1,1)$, the following set is upper-bounded:

$$\left\{L_a^b(\gamma,P)\middle|P\in\mathcal{P}([a,b])\right\}.$$

Let denote $L_a^b(\gamma)=\sup\left\{L_a^b(\gamma,P)\middle|P\in\mathcal{P}([a,b])\right\}$ and $\pi=L_{-1}^{1}(\gamma)$.

Therefore, let define $\arccos:[-1,1]\rightarrow[0,\pi]$ such that:

$$\arccos(x)=\left\{\begin{array}{ll} L_x^1(\gamma) & \text{if $-1\leq x<1$.} \\ 0 & \text{if $x=1$.} \end{array}\right.$$

Notice that the function is well-defined as I've proved before that $L_a^b(\gamma)=L_a^c(\gamma)+L_c^b(\gamma)$ for any $c\in(a,b)$ which allowed us to demonstrate that the function $L_{(\cdot)}^1(\gamma)$ is decreasing. Because of that, I could say that the arccosine function is injective and therefore, I can invert it. However, I cannot define the inverse of the arccosine as the cosine because I don't know yet that the image of the arccosine function is the entire interval $[0,\pi]$ for sure. How could I do it?


Solution. We'll prove that the arccosine function is continuous. Given $a\in[0,1)$ and $b\in[-1,0)$, let be $x\in(a,1)$ and $y\in(b,0)$. Let consider $P=\left\{t_i\right\}_{i=0}^{n}\in\mathcal{P}\left(\left[a,x\right]\right)$ and $Q=\left\{s_j\right\}_{j=0}^{m}\in\mathcal{P}\left(\left[b,y\right]\right)$. As $\gamma$ in continuous on $[-1,1]$, increasing on $[-1,0]$ and decreasing on $[0,1]$, then:

$$\begin{array}{l} L_a^x(\gamma,P)\leq\sum_{i=1}^{n}{(|t_{i-1}-t_i|+|\gamma(t_{i-1})-\gamma(t_i)|)}=\sum_{i=1}^{n}{(t_{i}-t_{i-1})}+\sum_{i=1}^{n}{[\gamma(t_{i-1})-\gamma(t_i)]}=x-a+\gamma(a)-\gamma(x)=|x-a|+|\gamma(x)-\gamma(a)|. \end{array}$$

$$\begin{array}{l} L_b^y(\gamma,Q)\leq\sum_{j=1}^{m}{(|s_{j-1}-s_j|+|\gamma(s_{j-1})-\gamma(s_j)|)}=\sum_{j=1}^{m}{(s_{j}-s_{j-1})}+\sum_{j=1}^{m}{[\gamma(s_{j})-\gamma(s_{j-1})]}=y-b+\gamma(y)-\gamma(b)=|y-b|+|\gamma(y)-\gamma(b)|. \end{array}$$

Therefore:

$$|\arccos(x)-\arccos(a)|=L_a^x(\gamma)\leq |x-a|+|\gamma(x)-\gamma(a)|\stackrel{x\to a^+}{\rightarrow} 0+0=0.$$ $$|\arccos(y)-\arccos(b)|=L_b^y(\gamma)\leq |y-b|+|\gamma(y)-\gamma(b)|\stackrel{y\to b^+}{\rightarrow} 0+0=0.$$

By Sandwich's theorem, the arccosine function is continuous to the right on $[-1,1)$. Analogously, it can be proved that it's continuous to the left on $(-1,1]$. In conclusion, the arccosine function is continuous on $[-1,1]$.

  • Can't you just define $\pi$ to be $L_{-1}^1(\gamma)$? – Hans Lundmark Aug 26 '21 at 13:41
  • That's what I did and that doesn't have anything to do with the question. – Raúl Filigrana Villalba Aug 26 '21 at 13:44
  • Show continuity of arccos and hence you can use IVT. – user10354138 Aug 26 '21 at 14:07
  • Yeah, I know is the other option. But I think we have the same difficulties. What would you do with $|\arccos(x)-\arccos(a)|$? We don't have any formula here. We only know that if $x<a$ (resp. $x>a$) then $\arccos(x)-\arccos(a)=L_x^a(\gamma)$ (resp. $-L_a^x(\gamma)$). – Raúl Filigrana Villalba Aug 26 '21 at 14:14
  • To expand user10354138's comment, you know $\arccos(-1) = \pi$ and $\arccos(1) = 0$. If you know/show that every definite integral is continuous in the lower limit (in the spirit of your definition of $\arccos$), the intermediate value theorem guarantees surjectivity. – Andrew D. Hwang Aug 26 '21 at 14:55
  • I would like to do without using integrals. If I wanted to use integrals and derivatives, I just use the formula of arclength for $\mathcal{C}^1$-curves and derive in order to know that the function is invertible, continuous and anything I want. It's just I want to explain the concept of trigonometric function without using too many tools. As I were a new student of maths. – Raúl Filigrana Villalba Aug 26 '21 at 14:58
  • Of course, using the intermediate value theorem is the natural thing to do here, but my impression is that you want to avoid the notion of continuity. Let $f(t)$ be the length of the arc between $x = 1$ and $x = 1 - t$. Then you've proved that $f$ is strictly increasing from $[0,2]$ into $[0,f(2)]$. The only part missing is to prove that the image of $f$ is dense in $[0,f(2)]$, and then with only general arguments (using the completeness of $[0,2]$), you can prove that $f$ is surjective. But you can prove this as follows. Show first that there are arbitrarily small positive values of $f(t)$. – Anonymous Aug 26 '21 at 20:55
  • So you have an arc of small length $\varepsilon$. Now by rotating it you can show that there are arcs of length $\varepsilon, 2\varepsilon, 3\varepsilon, \dots.$ So the image of $f$ is dense. Probably, in terms of the entire development of the theory, it would be most economical to prove invariance of arc length under rotation first. – Anonymous Aug 26 '21 at 20:57
  • In any case, if you don't want to use an argument based on rotations, you can give an argument similar to the IVT based on the inequality $|f(t) - f(s)| \leq |t - s| + |\sqrt{1 - t^2} - \sqrt{1-s^2}|$, valid if $s$ and $t$ are on the same side of $1$. – Anonymous Aug 26 '21 at 21:19
  • I like that last solution. I don't mind if we use continuity (I suppose you want to use uniform continuity). It's just that with integrals and derivatives it makes the problem easier, but we need too many tools and I want a way to construct formally the trigonometric functions and it's understandable for a student who sees Calculus for the first time. – Raúl Filigrana Villalba Aug 26 '21 at 22:54
  • I'll try it and paste the solution if I find it. – Raúl Filigrana Villalba Aug 26 '21 at 22:56
  • Without continuity: assume that some $p \in (0,f(2))$ is not in the range of $f$. Since $f$ is increasing, by a simple argument using completeness, the set of $t \in [0,2]$ such that $f(t) < p$ is of the form $[0,a]$ or $[0,a)$ for some $a \in [0,2)$. Now you can derive a contradiction from the inequality above. – Anonymous Aug 27 '21 at 00:26

0 Answers0