Logarithms: The Arithmetic of Scale

Some changes are too large for ordinary intuition. A million-to-one ratio and a ten-to-one ratio both become ungainly when they must share the same chart, calculation, or conversation. A logarithm does not make those differences small in a physical sense. It changes the bookkeeping: multiplication becomes addition, and equal steps represent equal ratios rather than equal increments.

That is why logarithms belong in sound levels, chemistry, earthquake magnitude, information, and any calculation that repeatedly multiplies. They are also easy to misuse. The base matters, the argument must be valid, and a physical quantity needs a reference before it may go inside a logarithm.

A person climbs equal staircase steps beside dramatically growing colored blocks and a branching pattern — On a logarithmic scale, equal vertical steps mean equal multiplication factors—not equal amounts added.

A logarithm asks for an exponentlog_b(x) = y if and only if b^y = x

For real logarithms: b > 0, b ≠ 1, and x > 0.

Read the definition before using a rule

bbase

The repeated multiplication factor. Base 10 and base e are common, but neither is the default in every field.

xargument

The positive quantity or, in physical work, a positive ratio to a stated reference.

ylogarithm

The exponent required to turn b into x. It is dimensionless.

enatural base

ln x means log_ex, useful whenever exponential rates and calculus appear.

The definition immediately gives the two inverse checks:

b^log_bx = xundo the logarithmlog_b(b^y) = yundo the exponent

If the base lies between zero and one, it is still allowed: log_1/10100 = −2 because (1/10)⁻² = 100. The curve simply runs downward. What is not allowed in the real-number setting is base zero, base one, a negative base, zero as an argument, or a negative argument. NIST’s DLMF definitions and OpenStax’s logarithmic-functions chapter set out these conditions and the inverse relationship.

Why multiplication becomes addition

The familiar laws are not shortcuts invented for a calculator. They follow from exponent arithmetic. If M = b^p and N = b^q, then MN = b^p+q; taking log_b on both sides gives the product rule.

Productlog_b(MN) = log_bM + log_bN

Quotientlog_b(M/N) = log_bM − log_bN

Powerlog_b(M^p) = p log_bM

Change baselog_ax = ln x / ln a

For real-valued work, the arguments M and N must be positive. The most expensive false rule is log(M + N) = log M + log N. It fails because addition is not exponent addition. As a quick counterexample, log₁₀(10 + 10) ≈ 1.3010, while log₁₀10 + log₁₀10 = 2. OpenStax collects the real-domain conditions and the product, quotient, and power laws in its logarithmic-properties section.

A person joins two ruler-like strips into one path leading toward objects at radically different scales — A logarithm turns a product into a sum. The conversion is exact, provided the base and positive-domain conditions are respected.

A logarithm needs a unitless argument

Writing log(3 m) has no physical meaning until the metre quantity is compared with a same-kind reference. The meaningful form is log₁₀(L/L₀), where both L and L₀ are lengths measured in the same unit. Their ratio is dimensionless before the logarithm is evaluated.

Good physical inputlog₁₀(P/P₀)

Power divided by a stated reference power.

Incomplete inputlog₁₀(3 watts)

A raw dimensional quantity has no reference and cannot be logged as written.

Required contextbase + ratio + reference

Changing any of these can change the reported number.

NIST’s Guide for the Use of the SI makes this reference-ratio requirement explicit. It is not fussy notation: it prevents a number from changing merely because someone switches from watts to milliwatts or metres to centimetres.

Worked example 1: six base-ten steps mean one million times

Let R = Q/Q₀ = 1,000,000. The base-ten logarithm is:

Write as a power1,000,000 = 10⁶

Take the loglog₁₀R = log₁₀(10⁶) = 6

Reverse check10⁶ = 1,000,000

Meaningsix equal ratio steps, not six times larger

The distinction matters. A difference of six on a base-ten scale represents a ratio of one million in the defined quantity. It says nothing, by itself, about energy, danger, or human impact.

Worked example 2: solve an exponential by taking logs

Solve 2^x = 300. There is no integer answer, but the logarithm can bring the exponent down where ordinary division can reach it.

Take ln of both sidesln(2^x) = ln 300

Use the power rulex ln 2 = ln 300

Isolate xx = ln 300 / ln 2 = 8.22881869049588

Magnitude and reverse check2⁸ = 256 < 300 < 512 = 2⁹; 2^{8.22881869049588} ≈ 300

The answer being between 8 and 9 is a free sanity check before trusting the displayed decimals. The same calculation can use any valid log base because a change-of-base factor cancels in the quotient.

One word, two decibel formulas

Decibels are a sharp reminder that a logarithm needs a model, not just a button press. A power ratio uses 10 log₁₀(P/P₀) dB. A field or amplitude ratio uses 20 log₁₀(F/F₀) dB only when the corresponding power is proportional to the square of the field. The factor is not a preference.

Power quantityL_P = 10 log₁₀(P/P₀) dB

A factor of 10 in power is +10 dB.

Field quantityL_F = 20 log₁₀(F/F₀) dB

A factor of 10 in amplitude is +20 dB when power is proportional to amplitude squared.

Always statequantity and reference

A bare dB value is incomplete without them.

NIST SP 330 gives the decibel relation and its status as a non-SI unit accepted for use with SI. The point is broader than acoustics: a logarithmic scale preserves ratios, but every field chooses its own reference, correction, and valid range.

An acoustic engineer turns an analog dial while wave ripples grow dramatically around equally spaced markers — The same logarithm supports several scales, but the reference and physical model decide what a reported step actually means.

Mistakes, boundaries, and when to stop

Shortcut	Why it fails	Better move
Leave the base unstated	`log` can mean base 10, base `e`, or a field convention.	Write `log₁₀`, `ln`, or name the base.
Apply laws to a sum	Only multiplication turns into addition under a log.	Keep `log(M + N)` intact.
Log a raw measurement	Changing measurement units changes the number.	Form a same-kind reference ratio first.
Interpret one log unit as one added unit	The coordinate represents a factor, not a difference.	Exponentiate the difference to recover the ratio.
Use a real log at zero or below	The real domain has ended.	Revisit the model or explicitly enter complex analysis with branch conventions.

Do not use a simple log conversion as a universal “severity scale.” The original local (Richter) earthquake magnitude, for example, uses measured wave amplitude under specified instrument and distance conventions, and it has limits for large events; USGS explains why it is not a standalone damage or energy dial in its magnitude overview. Near a zero or noise-floor measurement, relative uncertainty can also become enormous before a logarithm is even taken.

The durable idea is modest: logarithms translate multiplicative structure into additive structure. They make ratios easier to compare and exponentials easier to solve. They do not remove the need to know what was measured, which reference was chosen, or whether the model is valid.