Euler's formula e^(ix) = cos x + i sin x places the exponential function on the unit circle. As x increases, e^(ix) moves around the circle at unit speed, and its real and imaginary parts are the cosine and sine of the angle.
Read in reverse, the formula expresses trigonometry through exponentials: cos x = (e^(ix) + e^(−ix))/2 and sin x = (e^(ix) − e^(−ix))/(2i). Angle-addition formulas then follow from the rule e^(i(a + b)) = e^(ia)e^(ib), with no geometry needed. Setting x = π gives e^(iπ) = −1.
The formula is the working language of oscillations. Engineers write AC voltages as phasors, Fourier analysis splits signals into e^(ikx) components, and quantum wavefunctions carry complex phases of exactly this form.
Cartesian, cylindrical and spherical coordinates describe the same space, and simple formulas move between them.
From Cartesian to cylindrical, ρ = √(x² + y²) and φ = arctan(y/x), with z kept. From Cartesian to spherical, r = √(x² + y² + z²), θ = arctan(√(x² + y²)/z) and φ = arctan(y/x). Between the two curved systems, ρ = r sin θ and z = r cos θ, or in reverse r = √(ρ² + z²). In practice the angle φ is computed with arctan2(y, x), which picks the correct quadrant.
Cylindrical coordinates fit problems with an axis of symmetry, and spherical coordinates fit problems with a centre. Picking the system that matches the geometry often turns a hard integral into an easy one.
Many common probability distributions are special cases, transformations or limits of one another. A map of these links shows how a few distributions generate most of the rest.
A binomial with large n and small p approaches a Poisson with λ = np, and for large n it approaches a normal with mean np and variance np(1 − p). Standardising a normal gives Z ~ N(0, 1), squaring and summing standard normals gives a chi-squared, and a ratio of two scaled chi-squared variables gives an F distribution. A gamma with shape 1 is an exponential, and exponentiating a normal gives a lognormal.
These links let statisticians approximate one distribution by another, derive tests from simpler pieces, and choose a model by asking how the data were generated.
On a sphere, the shortest surface route between two points follows a great circle. Its plane passes through the sphere's centre, so the circle has the sphere's own radius and divides the surface into two hemispheres.
These routes are geodesics, the curved-surface equivalent of straight lines. Their length is s = Rθ, where R is the sphere's radius and θ is the central angle in radians. Between two points on the same latitude, following that latitude is always longer, unless it is the equator.
This geometry explains why long-distance flights appear curved on most flat maps. For antipodal points, the shortest route is not unique: infinitely many great semicircles join them, each of length πR.
The double integral of 1/((1 − xy)(1 + x)(1 + y)) over the unit square equals ln 2. Two standard techniques are enough to evaluate it.
For fixed x, partial fractions split the y-part as 1/((1 − xy)(1 + y)) = (1/(1 + x))(x/(1 − xy) + 1/(1 + y)). Integrating over y from 0 to 1 gives (ln 2 − ln(1 − x))/(1 + x), which leaves the single integral I = ∫₀¹ (ln 2 − ln(1 − x))/(1 + x)² dx. The ln 2 term contributes (ln 2)/2, and integration by parts shows the other term also equals (ln 2)/2.
The integrand blows up at the corner (1, 1), yet the integral converges. Integrals of this family over the unit square are closely related to logarithms and values of the zeta function.
In a right-handed coordinate system, the thumb points along the positive axis and the curled fingers give the direction of a positive rotation. The same rule fixes the order of the cross product: x̂ × ŷ = ẑ.
A positive rotation about +x turns ŷ toward ẑ, about +y turns ẑ toward x̂, and about +z turns x̂ toward ŷ. The pattern is cyclic, x to y to z and back to x, and reversing the order of any cross product flips its sign.
Robotics, aerospace and 3D graphics all depend on this convention. A sign error in one axis turns a pitch-up command into a pitch-down one, so flight controllers and graphics engines state their handedness explicitly.
Lagrange multipliers find the maximum or minimum of f subject to a constraint g = 0. The method replaces the constrained problem by the Lagrangian L(x, λ) = f(x) − λg(x).
At the optimum, the level curve of f touches the constraint curve without crossing it. Two curves that touch share a tangent line, so their normals line up, and that means ∇f = λ∇g. Together with g(x) = 0 this gives n + 1 equations for the n coordinates and λ. To maximise x + y on the circle x² + y² = 1, ∇f = (1, 1) and ∇g = (2x, 2y). Parallel gradients force x = y, and the constraint gives x = y = 1/√2, where x + y = √2.
The multiplier λ has a meaning of its own. It measures how fast the optimal value changes when the constraint is relaxed, which is why economists call it a shadow price.
One point in space, three addresses. Cartesian coordinates give (x, y, z), cylindrical give (ρ, φ, z), and spherical give (r, θ, φ), with θ measured down from the z axis and φ around it.
The conversions come from right triangles. In cylindrical form x = ρ cos φ and y = ρ sin φ, with z unchanged. In spherical form x = r sin θ cos φ, y = r sin θ sin φ and z = r cos θ, so ρ = r sin θ links the two curved systems. The inverse relations use r = √(x² + y² + z²), θ = arccos(z/r) and φ = arctan2(y, x).
The right choice follows the symmetry of the problem. A long wire or a pipe is simplest in cylindrical coordinates, a point charge or a planet in spherical ones, and a box in Cartesian ones.
The brachistochrone is the curve of fastest descent between two points under gravity, and it is an inverted cycloid: x = r(θ − sin θ), y = r(1 − cos θ).
A bead released from rest has speed √(2gy) at depth y, so its time is T = ∫ ds/√(2gy). The cycloid drops steeply first, gains speed early and wins in π√(r/g) ≈ 3.142√(r/g); the circular arc takes 3.173, the parabola 3.269 and the straight line 3.716.
Johann Bernoulli posed the problem in 1696, and Newton, Leibniz, l'Hôpital and Jakob Bernoulli all solved it. The same curve is the tautochrone: Christiaan Huygens found in 1659 that a bead released anywhere on it reaches the bottom in exactly π√(r/g).
The wave, heat and Schrödinger equations all share one spatial operator, the Laplacian ∇². What changes is how it is tied to time.
In the wave equation ∂²u/∂t² = v²∇²u, the Laplacian sets the acceleration, so disturbances travel at speed v without spreading. It covers strings, sound and light, where v becomes c. In the heat equation ∂u/∂t = α∇²u, it sets a first rate of change, so peaks flatten and spread. The Schrödinger equation iħ ∂ψ/∂t = −(ħ²/2m)∇²ψ is also first order in time, but the factor i makes the solution oscillate and interfere instead of diffusing.
Einstein's field equations use a curved-spacetime version of the same second-order structure. Learn the Laplacian once and it reappears across physics.
1. Six Math Essentials, Terence Tao (2026): built around six ideas, numbers, algebra, geometry, probability, analysis and dynamics.
2. Flatland, Edwin A. Abbott (1884): a novella narrated by a square living in two dimensions, who is visited by a sphere.
3. A Mathematician's Lament, Paul Lockhart (2009): a research mathematician turned school teacher on how school mathematics hides the art of the subject.
4. Fermat's Last Theorem, Simon Singh (1997): the 358-year hunt for a proof of Fermat's margin note, ended by Andrew Wiles. Published in the US as Fermat's Enigma.
5. A Mathematician's Apology, G. H. Hardy (1940): Hardy's essay, written late in his career, defending a life spent on pure mathematics.
6. Infinite Powers, Steven Strogatz (2019): the story of calculus, from Archimedes to its uses in modern science.
7. Journey Through Genius, William Dunham (1990): twelve great theorems, from Hippocrates to Cantor, each with its proof and its history.
8. How to Solve It, George Pólya (1945): the four-step method; understand the problem, devise a plan, carry it out, look back.
9. Prime Obsession, John Derbyshire (2003): the Riemann hypothesis, with chapters alternating between the mathematics and the history.
10. Love and Math, Edward Frenkel (2013): the Langlands program explained for general readers, alongside Frenkel's own story.
11. Gödel, Escher, Bach, Douglas R. Hofstadter (1979): self-reference in Gödel's logic, Escher's art and Bach's music. Winner of the 1980 Pulitzer Prize for General Nonfiction.
12. What Is Mathematics?, Richard Courant and Herbert Robbins (1941): a survey of the central ideas of mathematics, from numbers and geometry to calculus.
13. Elements, Euclid (c. 300 BC): thirteen books of geometry and number theory built from definitions, postulates and proofs.
14. Proofs from THE BOOK, Martin Aigner and Günter M. Ziegler (1998): named after Paul Erdős's idea of a book in which God keeps the most elegant proof of each theorem.
15. The Princeton Companion to Mathematics, edited by Timothy Gowers (2008): a guide to modern mathematics written by working mathematicians.
16. Concrete Mathematics, Ronald L. Graham, Donald E. Knuth and Oren Patashnik (1989): the mathematics behind computer science, grown from a Stanford course. The name joins CONtinuous and disCRETE.
17. Calculus, Michael Spivak (1967): calculus built up from the axioms of the real numbers, with proofs throughout.
18. Visual Complex Analysis, Tristan Needham (1997): complex analysis taught through geometric pictures.
19. Linear Algebra Done Right, Sheldon Axler (1995): linear algebra that develops eigenvalues without determinants, which arrive near the end.
20. Principles of Mathematical Analysis, Walter Rudin (1953): "Baby Rudin", the standard terse text for a first course in real analysis.
Twenty books for a mathematics shelf, starting with Terence Tao's new Six Math Essentials, his first book for a general audience. How many of the twenty have you read?
Gradient, divergence and curl are the three first-order derivatives of vector calculus. Each is built from the same operator ∇ = (∂/∂x, ∂/∂y, ∂/∂z).
The gradient ∇φ turns a scalar field into a vector field pointing towards steepest increase. For φ = x² + y², ∇φ = (2x, 2y), pointing outward and growing with distance. The divergence ∇ · F turns a vector field into a scalar measuring net outflow. For F = (x, y, z), ∇ · F = 3 everywhere, a uniform source. The curl ∇ × F measures local rotation. For F = (−y, x, 0), ∇ × F = (0, 0, 2), a steady spin about the z axis.
These operators appear in Maxwell's equations, fluid flow and heat transfer, and they are used to model flow in blood vessels.
The volume of a sphere of radius R is (4/3)πR³, and spherical coordinates give it in three short integrals. The coordinates are a radius ρ, a polar angle φ measured from the z-axis and an azimuth θ.
The volume element in these coordinates is dV = ρ² sin φ dρ dφ dθ. Over 0 ≤ ρ ≤ R, 0 ≤ φ ≤ π and 0 ≤ θ ≤ 2π, the three integrals separate: the ρ integral gives R³/3, the φ integral gives 2 and the θ integral gives 2π. Their product is (4/3)πR³.
The same result follows from stacking thin shells of area 4πr² and thickness dr. Spherical coordinates are the standard tool wherever a problem has spherical symmetry, from gravitational fields to the hydrogen atom.
The del operator ∇ = x̂ ∂/∂x + ŷ ∂/∂y + ẑ ∂/∂z is a vector whose components are derivatives. How it combines with a field decides what it measures.
Applied to a scalar field f, it gives the gradient ∇f, which points towards the fastest increase and is perpendicular to the level curves. Dotted with a vector field F, it gives the divergence ∇ · F, the net outflow per unit volume, positive at a source and negative at a sink. Crossed with F, it gives the curl ∇ × F, computed as a determinant, which measures local rotation.
The symbol was introduced by William Rowan Hamilton, and the name nabla comes from a harp of the same shape. Maxwell's equations, the heat equation and the equations of fluid flow are all written with it.
An aurora is light from charged particles of the solar wind striking the upper atmosphere, gathered in two rings around the magnetic poles.
The force F = qv × B is always perpendicular to the motion, so a particle circles a field line with radius r = mv⊥/(qB) while sliding along it. Earth's dipole field has lines r = L sin²θ and a strength growing as 1/r³, so the spiral tightens downward and each line lands at colatitude arcsin(1/√L).
Kristian Birkeland argued around 1900 that auroras were solar particles steered by Earth's field, and lit a magnetised sphere in a vacuum to prove it. Most particles never land: the converging field reflects them, trapping them in the Van Allen belts found in 1958.