The exponential thread - The puzzle

This is a story about how a single line of 3rd-grade arithmetic, taken seriously for long enough, leads to one of the most extraordinary unifications in modern mathematics.

The line is this:

$$3\times 4=4+4+4=12$$

Multiplication is repeated addition. Most people meet this idea once, accept it, and never look at it again. But every interesting question in this series is hiding inside it, waiting for someone to keep asking "what if".

By the end of the climb, the same chain of questions will have walked us to

$$e^{i\pi }+1=0$$

and through it to a structural picture in which $e$, $\ln$, $\sin$, $\cos$, complex numbers, matrices, and rotations are not five separate ideas that happen to be related. They are projections of one object: the exponential map of a one-dimensional Lie group (more about both later). Right now that sentence is a mouthful with no meaning attached to it. By Part 9, every word in it will have a concrete picture behind it.

The cast

Before we climb, here is the gallery of objects this series is about. You have probably met all of them in school.

• $e\approx 2.71828\dots$, a special constant that turns up everywhere
• $\ln (x)$, the natural logarithm (where "natural" gets a real definition in Part 2)
• $\sin (x)$ and $\cos (x)$, the functions of the unit circle
• $e^{i\pi }+1=0$, Euler's identity, often called the most beautiful equation in mathematics

These look like residents of different chapters of a textbook. One from algebra, one from calculus, two from geometry, the last from a chapter of complex analysis nobody assigned. But, they are not. The whole point of this series is to see, slowly and with the right pictures, that they are the same idea presented in different languages.

The first crack

Look again at the opening line.

$$3\times 4=4+4+4$$

Multiplication is the shorthand for "add the same number to itself N times". Once you have a notation for "do an operation N times", it becomes natural to ask what happens at the edges of the notation. What if N is zero? Negative? A fraction? An irrational number? Imaginary?

For multiplication those edge cases push the number system to widen. "Multiply by zero" forces zero to exist as a thing you multiply with, with its own rule ($n\times 0=0$). "Multiply by negative two" forces the negative integers to make sense as multipliers, with their own sign rules. Each shorthand we adopt opens a new question, and each new question demands the number system grow to accommodate the answer.

Now apply the trick one rung up the ladder. If multiplication is repeated addition, then exponentiation is repeated multiplication.

$$2^3=2\times 2\times 2=8$$

Same story, one level higher. We have a shorthand for "multiply N copies of a number together". The same family of edge cases shows up at the door, now louder.

Expression	What it means	What it forces into existence
$2^0$	zero copies multiplied	the convention that $2^0=1$
$2^{-1}$	a negative exponent	the rationals (with $2^{-1}=1/2$)
$2^{1/2}$	a fractional exponent	the irrationals (and a Pythagorean crisis)
$2^{\pi }$	a real-number power	continuity and limits
$2^i$	an imaginary power	the complex numbers

The simple game of shorthand has already opened the door to most of analysis. We have not even touched the last row, which is where the rest of this series lives.

Two inverses, not one

Subtraction is the inverse of addition. Division is the inverse of multiplication. Most school math leaves it at that, one inverse per operation.

Exponentiation breaks the pattern, because it has two inverses, not one. Both deserve a name.

Take the equation

$$b^x=y$$

with three letters in it. Fix any two and try to recover the third, and you get a different inverse for each choice.

Given	Solve for	Operation	Notation
$b$ and $x$	$y$	exponentiation	$y=b^x$
$x$ and $y$	$b$	the $x$-th root	$b=\sqrt[x]{y}$
$b$ and $y$	$x$	the logarithm in base $b$	$x={\log }_{b}y$

Roots and logarithms are siblings, two distinct inverses of the same operation. Most people's intuition for the logarithm stays fuzzy through their entire mathematical life because they never noticed it lived in a separate family from the root.

Once you see it cleanly, the logarithm becomes one of the central characters of this series. The "natural log" $\ln$ in particular will turn out to be the inverse of one specific exponential, not just ${\log }_e$ written with a different notation.

Where this is going

A nine-part climb. Each part hangs on one intuitive picture, and each step revisits a fact you "already knew" from the previous part and reveals it to have been a special case.

1. The puzzle (this post). The cast and the climb.
2. Why $e$ is the natural base. A unique number with the property that the exponential function $e^x$ equals its own derivative. Compounding interest, learning-rate decay, and the softmax function in machine learning all reach for $e$ for the same structural reason.
3. Sin and cos as a moving point. Geometry first. All the trig identities fall out of a single point moving on the unit circle. A pure tone in audio is the projection of that point.
4. Taylor series. A function reconstructed from its derivatives at one point. This is also how libm actually computes $\sin$ and $\exp$ on your CPU.
5. The reveal. Plug $ix$ into the Taylor series for $e^x$. Out comes $\cos x+i\sin x$. Multiplication by $i$ is a 90° rotation, and that fact makes Euler's identity feel inevitable. The Fourier transform sits one paragraph downstream.
6. Same equation, different costume. $y^{\prime }=ay$ describes growth and decay. $y^{\prime \prime }=-y$ describes oscillation. They are the same equation in disguise, and numerical ODE solvers exploit this.
7. Matrix exponentials. Once $e^x$ accepts a matrix as input, $\sin$ and $\cos$ appear as entries of a rotation matrix. RoPE (Rotary Position Embedding), the position encoding inside every modern transformer, is exactly this idea applied to attention vectors.
8. Groups and homomorphisms. Identities like $e^{a+b}=e^a\cdot e^b$ and the $\sin (a+b)$ formula are the same fact stated in two presentations. RSA's modular exponentiation and the Diffie-Hellman key exchange live in this house too.
9. One map to rule them all. The 1-dimensional Lie groups classified, and the exponential map identified. Looking back at the climb, we will see we were studying one function the entire time.

What's next

Part 2 picks up the thread at the number $e$ itself. Where does it come from, why does this one number out of all real numbers get the title "the natural" base, and why does every probabilistic and machine-learning context that needs an exponential reach for this one and not $2$ or $10$?