Diff — Scientific notation

{{infobox

| title = Scientific notation

| image = sci-notation-anatomy.svg

| caption = The anatomy of a scientific-notation number: coefficient, base, exponent.

| also called = standard form · index notation · scientific form

| type = Mathematical notation

| form = a × 10ⁿ, where 1 ≤ |a| < 10 and n is an integer

| precursors = Archimedes' myriad orders · Stevin's decimals · Descartes' exponents · Napier's logarithms

| used in = Astronomy, physics, chemistry, engineering, computing

| related = [[Floating point]] · [[Engineering notation]] · [[Order of magnitude]]

}}

# Scientific notation

[[File:sci-notation-anatomy.svg|center|720px|The parts of a scientific-notation number: a coefficient between 1 and 10, multiplied by a power of ten.]]

**Scientific notation** is the way of writing numbers as a product of a coefficient and a power of ten — *a* × 10ⁿ — and it exists because the universe refuses to stay at a human scale. A hydrogen atom is 0.000000000106 meters across. A mole of anything is 602,200,000,000,000,000,000,000 things. Both of those sentences are unreadable. Rewrite them as 1.06 × 10⁻¹⁰ m and 6.022 × 10²³, and suddenly they are not only legible — they are *comparable*. The exponent is the size; the coefficient is the detail; the notation is the lens that lets a human look directly at scales we cannot touch.

Almost every discipline that measures the physical world adopted it because it does three jobs at once: it compresses absurd digit strings, it makes magnitude visible at a glance, and — quietly, importantly — it encodes how *precise* a number actually is. This is the story of how the notation came to be, how it works, and how it escaped paper entirely to become the way computers store the universe.

---

## The problem of inconvenient numbers

The trouble starts as soon as you leave the human scale in either direction. The mass of an electron is about 0.000000000000000000000000000000911 kilograms — thirty zeros past the decimal point before the actual information begins. The mass of the Sun is roughly 1,989,000,000,000,000,000,000,000,000,000 kilograms — the same problem in the other direction. These numbers are not exotic edge cases; they are *routine* working values in physics and chemistry.

Written longhand, such numbers are worse than inconvenient — they are dangerous. Miscount a zero in a drug dosage or an orbital parameter and the error is not small; it is a factor of ten, which in the physical world is often the difference between working and catastrophe. Engineers of the slide-rule era had a name for this occupational hazard: the "decimal point error." The fix was not more care — it was better notation.

## Archimedes counts the sand

The first person known to grapple with this problem systematically was **Archimedes** of Syracuse, around 216 BC, in a short work addressed to King Gelon called *The Sand Reckoner* (*Psammites*).[1] The premise was a provocation: a common belief held that the grains of sand on all the world's beaches were effectively infinite — beyond numbering. Archimedes set out to prove otherwise, and to do it he first had to solve the harder problem of his age: Greek numerals topped out at the **myriad** (10,000), and there was no notation for anything larger.

His solution was the germ of every scaled notation that followed. He grouped numbers into **orders**: the first order ran up to a myriad of myriads (10⁸); the second order ran up to 10⁸ squared (10¹⁶); and so on through **periods** of orders — effectively indexing numbers by the power of a large base rather than writing each one out. With this machinery he built a geometric model of the cosmos (on the heliocentric-scale estimates of Aristarchus, no less) and counted the sand needed to fill it. His answer: no more than about **10⁶³ grains** — an absurdly large number, but a *finite, writable* one. The point was never the count. It was that the count could be written at all.

## Decimals, exponents, logarithms

Three more inventions had to arrive before the modern form could exist. The first was the **decimal fraction**: **Simon Stevin**'s 1585 booklet *De Thiende* ("The Tenth") taught Europe to write fractional quantities in decimal form and — crucially for our story — marked each decimal position with a small circled index, treating the place value itself as a numbered thing.[2]

The second was **exponent notation**. **René Descartes**' *La Géométrie* (1637) established the now-universal convention of raised superscripts for powers — x², x³, xⁿ — replacing the clumsy verbal and letter-hybrid notations of earlier algebraists.[3] Once powers had a compact written form, "ten to the power of" was one small step from becoming the notation's engine.

The third was the **logarithm**. **John Napier**'s *Mirifici Logarithmorum Canonis Descriptio* (1614) gave science a way to turn multiplication into addition — and as a side effect made the exponent, rather than the digits, the operative part of a number. The slide rule (William Oughtred's refinement, 1620s) put that idea in every engineer's pocket for three centuries: a physical device that computed by *adding exponents*. Slide-rule discipline — carry the mantissa, track the order of magnitude separately — is precisely the mental model scientific notation encodes.[4]

## The notation settles

There was no single invention date for a × 10ⁿ. Through the nineteenth century, astronomers and physicists increasingly wrote unwieldy values as a decimal part times an explicit power of ten, and textbooks gradually converged on the convention: a coefficient with exactly one nonzero digit before the decimal point, and an integer exponent. By the early twentieth century the form was standard in scientific literature, and by mid-century it was entrenched in school curricula as **standard form** (the British name) or **scientific notation** (the American one). Florian Cajori's *A History of Mathematical Notations* (1928) traces the long coalescence of the superscript-power idea that made it possible.[3]

```fact

There is no single inventor of scientific notation. It is a sedimentary invention: Archimedes' scaling orders, Stevin's decimal places, Descartes' superscripts, and Napier's logarithms each deposited a layer before the modern form surfaced in nineteenth-century practice.

```

## Anatomy of the notation

The convention is exact:

- **Coefficient** *a*: a decimal number with **1 ≤ |a| < 10** — exactly one nonzero digit left of the point.

- **Base**: always **10** — the ladder is our decimal system itself.

- **Exponent** *n*: an integer. Positive counts leftward shifts (10²³); negative counts rightward (10⁻¹⁰).

So 6.022 × 10²³ says "take 6.022 and move the point twenty-three places right," and 1.06 × 10⁻¹⁰ says "move ten places left." The form where the coefficient obeys the range rule is called **normalized scientific notation**; allowing the coefficient to drift produces useful variants (engineering notation, below).

The payoff is instant comparison. Two numbers in normalized form can be ordered without reading digits at all: whichever has the bigger exponent is bigger, and only ties need the coefficient consulted. 9.11 × 10⁻³¹ kg and 1.67 × 10⁻²⁷ kg are separated by four orders of magnitude — roughly ten thousand to one — before you have read a single significant digit.

## A ladder of everything

The full span of the notation's territory fits on one ladder. Each rung is ten times the one below: from the proton at 10⁻¹⁵ m through atoms (10⁻¹⁰), cells (10⁻⁵), people (10⁰), planets (10⁷), star systems (10¹³), galaxies (10²¹), up to the observable universe at roughly 8.8 × 10²⁶ m. The same ladder reads in bytes on the computing side — kilo-, mega-, giga-, tera-, peta-, exa-, zetta-, yotta- — and the two columns meet at the top, where the observable universe and humanity's accumulated data live at eerily similar exponents.

Charles and Ray **Eames** immortalized this view in their 1977 film *Powers of Ten*, which zooms by a factor of ten every ten seconds from a Chicago picnic blanket to the edge of the observable cosmos and back — arguably the most effective piece of mathematics communication ever filmed.[5]

## Arithmetic in the notation

Working *in* scientific notation — rather than converting in and out — is where it earns its keep:

| Operation | Rule | Example |

|---|---|---|

| Multiply | multiply coefficients, add exponents | (3×10⁴)·(2×10³) = 6×10⁷ |

| Divide | divide coefficients, subtract exponents | (8×10⁶)÷(2×10²) = 4×10⁴ |

| Add / subtract | equalize exponents first | 5×10⁴ + 3×10³ = 5×10⁴ + 0.3×10⁴ = 5.3×10⁴ |

| Raise to a power | power the coefficient, multiply exponent | (2×10³)² = 4×10⁶ |

| Root | root the coefficient, divide exponent | √(9×10⁶) = 3×10³ |

| Renormalize | nudge coefficient back to [1,10), adjust exponent | 62×10⁷ = 6.2×10⁸ |

The renormalization step is the only subtlety: arithmetic can push the coefficient out of range (62 × 10⁷), and the fix is to slide the decimal point and compensate in the exponent. Every floating-point unit in every computer performs exactly this dance, billions of times per second.

## Honest precision: significant figures

Scientific notation carries a second, quieter payload: it encodes *precision*. In longhand, "0.0030" and "0.003" are easy to confuse and the trailing zero's meaning is ambiguous. In scientific form the ambiguity evaporates: **3.0 × 10⁻³** declares two significant figures, **3 × 10⁻³** declares one, and **3.00 × 10⁻³** declares three. The coefficient's digits are, by convention, exactly the digits the measurement earned.[6]

This is why the notation is not merely cosmetic in science. Avogadro's number written 6.022 × 10²³ asserts four significant digits; writing it 6.02214076 × 10²³ (the exact SI definition since 2019) asserts ten. The exponent carries the scale; the coefficient carries the confidence. Measurement science — uncertainty, error bars, laboratory reporting — is impossible to write cleanly without this separation.

## Engineering notation and the SI prefixes

A close cousin relaxes the coefficient rule: **engineering notation** pins the exponent to multiples of three — so every value lands on an SI prefix boundary. 47 × 10³ is not normalized scientific form, but it *is* instantly "47 kilo," which is how engineers actually talk about it. Modern chip datasheets, for instance, quote supply currents in mA and gate delays in ns — engineering notation wearing its prefix directly.[7]

| Exponent | Prefix | Symbol | A familiar example |

|---|---|---|---|

| 10⁻¹² | pico | p | picosecond laser pulses |

| 10⁻⁹ | nano | n | nanometer-scale transistors |

| 10⁻⁶ | micro | µ | microamp sensor currents |

| 10⁻³ | milli | m | millimeter wavelengths |

| 10⁰ | — | — | the human scale |

| 10³ | kilo | k | kilograms, kilometers |

| 10⁶ | mega | M | megapixels |

| 10⁹ | giga | G | gigabytes, gigahertz |

| 10¹² | tera | T | terabyte drives, teraflops |

| 10¹⁵ | peta | P | petascale supercomputing |

| 10¹⁸ | exa | E | exascale: 10¹⁸ FLOPS |

| 10²¹ | zetta | Z | the global datasphere |

| 10²⁴ | yotta | Y | yottabyte-scale totals |

In November 2022 the SI added four more — **ronna** (10²⁷) and **quetta** (10³⁰) above, **ronto** (10⁻²⁷) and **quecto** (10⁻³⁰) below — partly because data volumes and atomic-scale measurement had finally outgrown yotta.[8] The ladder, it turns out, still needed more rungs.

## E-notation: when keyboards can't superscript

Computers inherited the notation but not the typography: there is no ×10 key, and no superscript shift. The fix was the letter. **FORTRAN** (1957) established the conventions still printed today: `1.06E-10` for scientific notation in single precision, and `D` — `1.5D-08` — for double precision, a living fossil in Fortran codebases to this day. **ALGOL** tried something typographically nobler: a dedicated subscript-ten symbol `₁₀`, so 6.02486₁₀23 meant ×10²³ — beautiful in print, but a character that scarcely any real input device could type, and one of the era's standing character-encoding headaches.[9]

C's `printf` cemented `%e` as the format specifier (and `%g` for "pick the tidier form"), and virtually every language since accepts `1e-10`, `6.02e23`, or `4.7E3` as literals — ALB among them. The E is pure convention: it stands for *exponent*, and the number it writes is identical to its superscripted ancestor.

## Floating point: scientific notation in binary

The deepest inheritance runs one level lower, into the hardware itself. A floating-point number **is** scientific notation — just with base 2 instead of base 10. Sign bit, exponent field, mantissa: exactly our *s*, *n*, and *a*, packed into a register.

And the machine that did it first is one of computing history's great buried facts: **Konrad Zuse**'s [[Z1]] (1938) and [[Z3]] (1941) — mechanical and relay machines respectively — computed in **22-bit binary floating point**: 1 sign bit, a 7-bit two's-complement exponent, and a 14-bit mantissa with a *hidden* leading 1, the same normalization trick IEEE 754 standardized four decades later. Zuse was solving ballistic and structural calculations; decimal fixed-point simply could not span the scales involved, so his machines normalized every value into sign-plus-exponent-plus-fraction form — scientific notation, implemented in telephone relays.[10]

IEEE Standard 754 (1985, revised 2019) generalized the scheme: a binary32 packs sign, 8-bit biased exponent, and 23-bit mantissa (24 effective bits ≈ 7 decimal digits) into a word; binary64 widens that to 15–16 digits.[11] Every normalization rule from the table above — the sliding point, the exponent bookkeeping, the coefficient that must live in range — runs in silicon, roughly 10¹⁸ times per second across the world's computers at any instant. Even the failure modes inherit faithfully: overflow (`1e309` → `Infinity`), underflow to zero, and the notorious `0.1 + 0.2` are all base-2 shadows of the same notation's limits.

```fact

The commonest scientific-notation bug in history wasn't human at all: floating-point cannot represent most decimal fractions exactly, so 0.1 + 0.2 evaluates to 0.30000000000000004 — a rounding artifact that has confused nearly every programmer who has ever typed it.

```

## Variants in the wild

[[File:sci-notation-ways.svg|center|720px|One value — a hydrogen atom's diameter — written four ways.]]

The same number can legally wear several costumes: longhand decimal, normalized scientific, engineering notation, or E-notation — each a different answer to a different constraint. A few more conventions worth knowing by name:

- **Order of magnitude**: the exponent alone, used as a rounded claim — "within an order of magnitude" means within a factor of about ten.

- **Scientific E notation**: the textual form (`1.06e-10`), normalized coefficient assumed.

- **Hexadecimal float** (`0x1.8p+3` in C): scientific notation with base 2 mantissa and explicit binary exponent — the hardware's dialect surfaced in source.

- **Magnitude notation in physics**: astronomers additionally repurpose "magnitude" for logarithmic brightness scales — related spirit, different beast.

## See also

- [[Floating point]] · [[Engineering notation]] · [[Order of magnitude]] · [[Significant figures]]

- [[Z1]] · [[Z3]] · [[Konrad Zuse]] · [[Plankalkül]] — the first floating-point machines

- [[IEEE 754]] · [[FORTRAN]] · [[ALGOL]] · [[C_Programming_Language|C]]

- [[History of Mathematics]] · [[Powers of Ten]] · [[UNIVAC I]]

## References

[1] Archimedes, *The Sand Reckoner* (Psammites), c. 216 BC — Thomas Heath's translation, *The Works of Archimedes* (1897).

[2] Simon Stevin, *De Thiende* (The Art of Tenths), Leiden, 1585 — the first systematic European treatment of decimal fractions.

[3] Florian Cajori, *A History of Mathematical Notations*, Vol. I, Open Court, 1928 — on Descartes' superscripts and the evolution of power notation.

[4] John Napier, *Mirifici Logarithmorum Canonis Descriptio*, Edinburgh, 1614 — the logarithm, and the exponent-as-quantity idea.

[5] Charles and Ray Eames, *Powers of Ten*, film, Eames Office, 1977 (expanded from the 1968 prototype; book by Philip and Phylis Morrison, 1982).

[6] BIPM, *Guide to the Expression of Uncertainty in Measurement* (JCGM 100:2008) — significant figures and reported precision.

[7] Bureau International des Poids et Mesures, *The International System of Units (SI Brochure)*, 9th edition, 2019.

[8] 27th General Conference on Weights and Measures (CGPM), November 2022 — adoption of ronna, quetta, ronto, and quecto.

[9] P. Naur (ed.), *Revised Report on the Algorithmic Language ALGOL 60*, 1963 — including the subscript-ten `₁₀` basic symbol; FORTRAN E/D exponent conventions trace to the 1956–58 IBM 704 manuals.

[10] Raúl Rojas, "Konrad Zuse's Legacy: The Architecture of the Z1 and Z3," *IEEE Annals of the History of Computing* 19(2), 1997 — the 22-bit floating-point word format.

[11] IEEE Standard 754-2019, *IEEE Standard for Floating-Point Arithmetic*.

## Quiz

- [ ] Write the mass of an electron (0.000000000000000000000000000000911 kg) in scientific notation.

- [ ] In a × 10ⁿ normalized form, what is the allowed range of the coefficient?

- [ ] Who estimated the grains of sand needed to fill the universe, and in what work?

- [ ] Which 1585 text by Simon Stevin established decimal fractions in Europe?

- [ ] How do you add 5 × 10⁴ and 3 × 10³ — and what's the result?

- [ ] What do the trailing zeros in 3.00 × 10⁻³ tell you that 3 × 10⁻³ does not?

- [ ] What is engineering notation's rule for the exponent, and why?

- [ ] What did FORTRAN use the letters E and D for?

- [ ] Which 1938 machine computed in binary floating point — and how many bits was its word?

- [ ] What do the 2022 SI prefixes ronna and quetta denote?

```hide Show answers

1 — **9.11 × 10⁻³¹ kg** · 2 — **1 ≤ |a| < 10** — exactly one nonzero digit before the point · 3 — **Archimedes**, in *The Sand Reckoner* (~216 BC); his bound was ≈10⁶³ grains · 4 — ***De Thiende*** ("The Tenth") · 5 — **equalize exponents first**: 5×10⁴ + 0.3×10⁴ = **5.3 × 10⁴** · 6 — **three significant figures** — a claim about measurement precision · 7 — **exponents must be multiples of 3**, so values map onto SI prefixes (kilo, mega, …) · 8 — **E** marks the exponent in single precision, **D** in double precision (`1.5D-08`) · 9 — **Zuse's Z1** — a **22-bit** floating-point word (sign, 7-bit exponent, 14-bit mantissa + hidden 1) · 10 — **10²⁷** and **10³⁰** respectively — adopted November 2022, extending the ladder beyond yotta.

```

[[Category:Mathematics]]

[[Category:History]]

[[Category:Science]]