Rigorous Elementary Functions

  • Thread starter Thread starter wrobel
  • Start date Start date
wrobel
Science Advisor
Insights Author
Messages
1,283
Reaction score
1,100
The goal of this post is to formally construct the theory of elementary functions without gaps, from the perspective of a university-level analysis course. It is intended as a unifying overview for undergraduate students who have previously encountered these functions in school but perhaps without a rigorous foundation.

A key point I want to emphasize is that, contrary to popular belief, it is quite straightforward to bridge the gap between the definition of trigonometric functions via the complex exponential and their classical geometric definition via the unit circle.



1. The Exponential Function​

Definition: The exponential function is defined by the power series:

$$\exp(z) = 1+\sum_{k=1}^\infty \frac{z^k}{k!}, \quad z = t + i\tau \in \mathbb{C}$$
The exponential is an entire function; its radius of convergence is infinite, and it satisfies the property ##\exp(\mathbb{R})\subset\mathbb{R}##.

By direct verification, we can see that the exponential satisfies the following Cauchy problem (initial value problem):

$$\frac{dw}{dz} = w, \quad w(0) = 1, \quad w(z) = \exp(z) \qquad (1)$$
Due to the group property of solutions to ordinary differential equations, we have:

$$\exp(z_1 + z_2) = \exp(z_1)\exp(z_2) \qquad (2)$$
From this identity, it is clear that the exponential never vanishes in ##\mathbb{C}##. If it were zero at any point ##z_1##, it would be identically zero everywhere, which contradicts ##w(0)=1##.

Consequently, the function ##\exp(t)## is positive and increasing for ##t \in \mathbb{R}##. This follows directly from (1). From the definition of the series, it also follows that ##e := \exp(1) > 1##.

Using formula (2), we can establish that for any rational number:

$$\exp(m/n) = \sqrt[n]{e^m}, \quad m \in \mathbb{Z}, n \in \mathbb{N}$$
As a result, we observe the asymptotic behavior: ##\lim_{t \to \infty} \exp(t) = \infty## and ##\lim_{t \to -\infty} \exp(t) = 0##.

2. Logarithms and Powers​

Definition: The natural logarithm is defined as the inverse of the exponential function:

$$\ln: (0, \infty) \to \mathbb{R}, \quad \ln(\exp t) = t, \quad \exp(\ln \xi) = \xi, \quad \xi > 0$$
Furthermore, we can define general logarithms and powers for ##a > 0, a \neq 1##:

$$\log_a \xi := \frac{\ln \xi}{\ln a}, \quad a^t := \exp(t \ln a)$$

3. Trigonometric Functions​

Definition: We define the sine and cosine functions using Euler's formulas:

$$\cos z := \frac{\exp(iz) + \exp(-iz)}{2}, \quad \sin z := \frac{\exp(iz) - \exp(-iz)}{2i}$$
From these definitions, the following identities follow immediately:

$$\cos^2 z + \sin^2 z = 1, \quad \frac{d}{dz}\cos z = -\sin z, \quad \frac{d}{dz}\sin z = \cos z$$

4. Geometric Interpretation​

Consider a curve in the Cartesian plane ##(x, y)## defined by:

$$x(t) = \cos t, \quad y(t) = \sin t \qquad (3)$$
From the identities above, we find:

$$x^2(t) + y^2(t) = 1, \quad \dot{x}^2 + \dot{y}^2 = 1, \quad \begin{vmatrix} x(t) & y(t) \\ \dot{x}(t) & \dot{y}(t) \end{vmatrix} = 1$$
Theorem: The functions in (3) represent the parametric equation of a unit circle centered at the origin. Here, ##t## acts as the natural parameter (arc length). As ##t## increases, the circle is traversed in a counter-clockwise direction.

By defining the length of the unit circle as ##2\pi## (which serves as the definition of the number ##\pi##), we obtain the periodicity of trigonometric functions:

$$\cos(t + 2\pi) = \cos t, \quad \sin(t + 2\pi) = \sin t$$
 
Last edited:
Physics news on Phys.org
Have you seen the recent ArXiv preprint paper by Andrzej Odrzyvolek

https://arxiv.org/html/2603.21852v2

It's related to your discussion, and you might find it interesting. Basically, he has reduced all intrinsic functions to a single function: and

eml(x, y) = exp(x) - ln(y)

He came up with this while wondering why, in math, we don't have an equivalent of the NAND gate in Boolean algebra, which is a fundamental building block for other logic gates.

Through an evolutionary search, he uncovered the eml() function, named exp-minus-ln, while trying to minimize the keys on his scientific calculator.

Although quite impractical, it does show interesting mathematical ideas that could be used to great effect in evolutionary equation discovery from data.
 
No, I haven't seen this paper. Thank you. I’m not an expert in algorithm theory and math. logic, but it seems he is addressing a scientific question and has obtained a new result. My goal is much more modest: it is purely educational and methodological
 
Yeah, my take was he was trying to find a programming building block for symbolic regression and evolutionary equation discovery. He mentions that the lowly NAND gate can simulate all other primitive gates and seeks something similar for more general math.

It's mathematically elegant. It plays a growth function exp() against its inverse. I think it's for some theoretical framework he is building. The paper is a preprint, but he has 40-50 published papers.

The paper also caught the attention of Sabine Hossenfelder. She did a video on it.



My research area is symbolic regression, and I was investigating similar schemes to eliminate nonsense and trivial expressions from the evolutionary part while providing scale factors and offsets to match an expression to measured data and to componentize it so that mutations are easy to implement ie cutting a subtree from an expression tree and pasting it into another expression tree.
 

Similar threads

  • · Replies 25 ·
Replies
25
Views
2K
  • · Replies 3 ·
Replies
3
Views
943
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 0 ·
Replies
0
Views
2K
Replies
0
Views
2K
  • · Replies 0 ·
Replies
0
Views
4K
  • · Replies 27 ·
Replies
27
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
Replies
7
Views
3K