Logarithms: Taming Huge Numbers

The question a logarithm answers#

Exponentiation is repeated multiplication: $2^3 = 8$ means "multiply three 2s together." A logarithm runs this backwards. It asks the opposite question:

$$\log_b x = y \iff b^y = x$$

In words: $\log_b x$ is the power you must raise the base $b$ to in order to get $x$. So $\log_2 8 = 3$, because $2^3 = 8$. And $\log_{10} 1000 = 3$, because $10^3 = 1000$. The logarithm is nothing more mysterious than the exponent, pulled back out into the open.

This is the reframe worth holding onto. A logarithm is not "just a button on the calculator." It is the answer to a concrete question — what power of the base produces this number? — and that question turns out to be one of the most useful in all of applied mathematics.

Multiplication becomes addition#

Here is the property that made logarithms indispensable for three centuries before electronic calculators existed. Suppose $x = b^p$ and $y = b^q$. Then $xy = b^p \cdot b^q = b^{p+q}$. Take the logarithm of both sides and you read off the exponents:

$$\log_b(xy) = \log_b x + \log_b y$$

Multiplying two numbers becomes adding their logarithms. There is a companion rule for powers, which follows the same way from $(b^p)^k = b^{pk}$:

$$\log_b(x^k) = k \log_b x$$

Before the 1970s, engineers multiplied large numbers by looking up their logarithms in a table, adding, and looking up the reverse. A slide rule is exactly this trick made physical: two rulers marked not in equal distances but in equal ratios, so sliding one against the other adds logarithms and multiplies the underlying numbers. Every time you say two quantities differ by "three orders of magnitude," you are adding logarithms in your head.

Any base you like: change of base#

You will meet three bases constantly: base 10 (scientific notation, decibels), base 2 (computer science, information), and base $e \approx 2.718$ (calculus, natural growth, written $\ln$). They are all the same idea, and one formula converts between them:

$$\log_b x = \frac{\log_c x}{\log_c b}$$

So $\log_2 x = \ln x / \ln 2$. Changing base only multiplies by a constant, which is why computer scientists write $O(\log n)$ without specifying a base — the base changes the constant, not the shape of the growth.

The scale that fits atoms and galaxies#

If each step along an axis represents a fixed ratio rather than a fixed amount, you get a logarithmic scale. On a linear axis marked 0 to a billion, the numbers 1, 10, 100, and 1000 are all crushed indistinguishably against the origin, and the billion mark is a mile away. On a log axis, each factor of ten is an equal step, so all of them spread out and become readable at once.

Toggle between the two axes above. The payoff of the log scale is that equal spacing means equal ratio. This is why so many real-world scales are logarithmic: the Richter scale (each unit is ×10 in ground amplitude), decibels (a fixed factor of loudness per step), pH (each unit is ×10 in hydrogen-ion concentration), and stellar magnitudes (each step a fixed brightness ratio, a convention inherited from the ancient Greeks). None of these could show their full range on a linear axis without either losing the small end or running off the page.

The name is a trap: logarithmic growth is slow#

Now the misconception worth dismantling. "Logarithmic growth" sounds dramatic, and people routinely assume it means something is taking off. It means the exact opposite.

Because a logarithm is the inverse of an exponential, it is the slowest growth you commonly meet. Watch what it costs to make $\log_2 n$ increase:

| $n$ | 2 | 4 | 8 | 16 | 32 | 1,048,576 | |---|---|---|---|---|---|---| | $\log_2 n$ | 1 | 2 | 3 | 4 | 5 | 20 |

To push the logarithm up by a single unit, you have to double the input. Go from a thousand to a million to a billion, and $\log_2 n$ only climbs from about 10 to 20 to 30. Compare that to its inverse, the exponential $2^n$, which explodes past a billion by the time $n$ reaches 30.

The three curves make the contrast physical. The exponential shoots off the top of the plot almost immediately; the linear line climbs steadily; the logarithm barely lifts off the floor, gaining just one unit each time the input doubles. Exponential explodes, logarithmic crawls — they are mirror images, not synonyms.

That crawl is not a weakness; it is the whole point. It is exactly why binary search is fast: halving the search space each step means only $\log_2 n$ steps are ever needed, so a billion-item sorted list is searched in about thirty comparisons. It is the same reason a log scale can hold atoms and galaxies on one axis — the vast range gets compressed precisely because the logarithm grows so grudgingly.

Where logarithms show up#

Once you recognize the inverse-of-an-exponential shape, you see logarithms everywhere. The natural log $\ln$ is the antiderivative of $1/x$, which is why it appears whenever you integrate a rate that falls off inversely — a fact that connects directly to the derivative and the calculus of growth. In information theory, the number of bits needed to identify one option out of $n$ equally likely ones is $\log_2 n$, and Shannon's entropy $-\sum p_i \log p_i$ is a weighted average of exactly such logarithms. Even Euler's formula hides a logarithm: extending $\ln$ to the complex plane is what makes sense of raising numbers to imaginary powers.

The thread through all of it is the same simple definition. A logarithm answers "what power?", it converts multiplying into adding, and — despite its reputation — it grows about as slowly as anything ever does.