Calculus.

OpeningWhat calculus is.

Calculus is the mathematics of change — of how a quantity behaves as another quantity shifts continuously. It turned the static algebra of equation-solving into a dynamic study of motion, growth, and accumulation.

Two operations sit at the centre. Differentiation extracts the instantaneous rate of change of one quantity with respect to another. Integration accumulates a varying quantity over an interval. The fundamental theorem of calculus says these operations are inverses of each other.

The discipline has the strangest origin story in mathematics: invented twice, almost simultaneously, by two giants who then spent decades arguing about who got there first.

Chapter IGreeks and exhaustion.

Eudoxus of Cnidus (c. 408 – c. 355 BCE) developed the method of exhaustion: bracket a curved region by inscribed and circumscribed polygons and squeeze the difference toward zero. Archimedes (c. 287 – 212 BCE) deployed it brilliantly — computing the area of a parabolic segment, the volume and surface of a sphere, and bounds for π between 3 + 10/71 and 3 + 1/7.

The 1906 rediscovery of the Archimedes Palimpsest revealed his lost text The Method, in which he describes his heuristic discovery procedure: imagine a parabolic segment as a sum of indivisible line segments, balanced against the lines of a known triangle on a mechanical lever.

It was, in essence, integration. Archimedes did not have the limit concept that would make the procedure rigorous, so he reproved every result by exhaustion. But the spirit of integration — adding up infinitely many infinitely small pieces — is unambiguously his.

Chapter IIKepler's wine barrels.

In 1613, having married for the second time, Johannes Kepler noticed that wine merchants gauged barrel volumes with a single dipstick measurement and standardised dimensions. Was the standard barrel shape efficient?

His Nova Stereometria Doliorum Vinariorum (1615) — "New stereometry of wine barrels" — answered yes by computing barrel volumes as sums of infinitely many infinitely thin disks. The technique is recognisably a limit-of-Riemann-sums calculation, sixty years before the formal subject existed.

Kepler's work, like Archimedes's Method two thousand years before, treated areas and volumes as sums of "indivisibles" without a rigorous limiting framework. He computed the volumes of 92 different solids of revolution. The book deserves to be more famous than it is — calculus before calculus, in print.

Chapter IIICavalieri's principle.

Bonaventura Cavalieri (1598–1647), a student of Galileo, published Geometria indivisibilibus continuorum (1635), which argued that two solids with equal cross-sectional areas at every height have equal volumes.

The principle works without metaphysics. Imagine the solids as stacks of paper sheets; if every corresponding pair of sheets has the same area, the total volume must agree. Cavalieri used it to compute volumes that had been intractable by exhaustion.

His "indivisibles" troubled critics: an area cannot literally be a sum of lines (lines have no thickness). But the answers were correct. Cavalieri's principle is now standard in introductory calculus and is one of the most reliable tools in elementary volume computation. The formal justification, by way of integration, came two centuries later.

Chapter IVTangent lines.

Pierre de Fermat's 1636 method for finding the maximum of a function — circulated in manuscript, not published until 1679 — set f(x + e) = f(x), treated e as small, divided through by e, then set e = 0.

The procedure is the derivative-equals-zero technique we still teach. Fermat called it adequation; it works because at a critical point the function changes by a quantity of order e², not e. The same method also produced tangent lines to curves and tangent slopes — the prototype of differentiation.

Newton later acknowledged that Fermat's method "gave him the hint" for the derivative. Fermat is the closest pre-Newton-and-Leibniz figure to actually inventing the calculus; what he lacked was the integral-derivative connection that the next generation would call the fundamental theorem.

Chapter VNewton's Principia.

Isaac Newton (1643–1727) developed his "method of fluxions" during the annus mirabilis of 1665–66, while plague closed Cambridge and he retreated to Woolsthorpe. Newton conceived of quantities as flowing in time; their rates were "fluxions" (our derivatives), and the inverse operation produced "fluents" (our integrals).

The Philosophiae Naturalis Principia Mathematica (1687), commissioned and paid for by Edmund Halley, applied the new methods to derive Kepler's laws of planetary motion from inverse-square gravitation. The result was the most important scientific book ever written.

The Principia presents its proofs in classical Euclidean style — Newton expected geometric arguments to be more persuasive to contemporaries — but the underlying machinery is calculus. De Methodis Serierum et Fluxionum, Newton's explicit calculus text, was written in 1671 and not published until 1736.

Chapter VILeibniz and the dx.

Gottfried Wilhelm Leibniz (1646–1716), philosopher and diplomat, developed an independent calculus during a 1675–76 stay in Paris. His first published account appeared in Acta Eruditorum in 1684 — three years before the Principia.

Leibniz's notation is the one that survived. dy/dx for the derivative; ∫ y dx for the integral; d²y/dx² for second derivatives. The notation makes operations look like algebra: the chain rule reads dy/dx = (dy/du)(du/dx), as if differentials canceled. They don't, formally; but the notation makes the machinery flow.

Leibniz's calculus spread quickly across the Continent. The Bernoulli brothers — Jacob and Johann — taught it to a generation, including Johann's student Leonhard Euler, who would become its greatest expositor.

Chapter VIIThe priority dispute.

The most famous priority dispute in mathematics. Newton had developed his method first (1665–66) but published last; Leibniz had published first (1684) but developed it independently in the 1670s.

The Royal Society — of which Newton was president — formed an investigating committee in 1712. The committee's report, Commercium Epistolicum, ruled for Newton. Newton wrote it himself, anonymously.

The modern verdict: independent invention. Newton was first chronologically; Leibniz was first to publish; both made original contributions; Leibniz's notation was better and is what survived. The dispute was nationally damaging — British mathematicians, defending Newton's notation, fell behind the Continental school by half a century. By 1820 they had to import the Leibnizian notation back via the Cambridge Analytical Society to catch up.

Chapter VIIILimits made rigorous.

Newton and Leibniz's "infinitesimals" — quantities smaller than any positive number but not zero — drew justified philosophical objections. Bishop Berkeley's 1734 The Analyst mocked them as "ghosts of departed quantities" and asked how a divisor that becomes zero at the end of a calculation can have been used as a divisor in the middle.

The answer took a century. Augustin-Louis Cauchy's Cours d'analyse (1821) defined the limit as the value a sequence approaches without necessarily reaching it; Karl Weierstrass (1860s) gave the modern epsilon-delta formulation: lim f(x) = L means for every ε > 0 there is δ > 0 such that |x − a| < δ implies |f(x) − L| < ε.

The infinitesimal was redeemed in the 1960s by Abraham Robinson's non-standard analysis, which constructed a coherent number system extending the reals with genuine infinitesimals. Newton and Leibniz had been right; they just lacked the model theory.

Chapter IXThe derivative.

The instantaneous rate of change of f at a:

Geometrically: the slope of the tangent line at (a, f(a)). Physically: if f is position, f' is velocity; if f is velocity, f' is acceleration.

The basic rules — sum, product, quotient, chain — let one differentiate any expression built from polynomials, exponentials, logarithms, and trigonometric functions. The exponential e^x is the unique function that is its own derivative; this defines the number e ≈ 2.71828.

Derivatives turn up everywhere: optimisation (find where f' = 0), physics (the Hamilton–Lagrange formulation), economics (marginal utility), machine learning (gradient descent on loss functions). Half of applied mathematics is derivative-finding in disguise.

Chapter XThe integral.

The accumulation of a varying quantity. The Riemann definition (Bernhard Riemann, 1854) partitions an interval, takes a representative value of the function on each piece, and sums the products:

Geometrically: the signed area between the curve y = f(x) and the x-axis. Physically: if f is velocity, the integral over time gives displacement; if f is power, the integral gives energy.

Computing integrals is harder than differentiation. Antiderivatives — functions whose derivative is the integrand — exist for elementary functions but cannot always be expressed in closed form. The Gaussian e^(−x²) has no elementary antiderivative; computing its integral required the development of the error function and special-function theory.

Chapter XIThe fundamental theorem.

Differentiation and integration are inverse operations. In one form:

In another form (the evaluation theorem):

The result was glimpsed by James Gregory (1668) and Isaac Barrow (1670 — Newton's teacher); Newton and Leibniz made the connection central. It is the bridge between the two halves of calculus and what makes the subject computationally tractable. Most integrals are evaluated by finding antiderivatives, not by Riemann-summing.

Higher-dimensional generalisations — the divergence theorem, Green's theorem, Stokes's theorem — are all forms of the same statement: integrating the derivative over a region equals integrating the original function over the boundary.

Chapter XIIDifferential equations.

An equation involving a function and its derivatives. F(x, y, y', y'', …) = 0. Solve it: find every function y(x) that satisfies the equation.

Differential equations are the language of physics and engineering. Newton's second law F = ma is a second-order ODE. Maxwell's equations are first-order PDEs in the electromagnetic field. The heat equation, the wave equation, the Schrödinger equation, the Navier–Stokes equations — all are differential equations, and most of physics consists of solving them.

For most real-world DEs, no closed-form solution exists. Numerical methods (Euler's method, Runge–Kutta, finite-element methods) give approximate solutions. Qualitative analysis — phase portraits, bifurcation theory, dynamical systems — extracts structure even when exact solutions are unavailable. The chaotic behaviour of nonlinear DEs (Lorenz 1963) was one of the great surprises of 20th-century mathematics.

Chapter XIIIOrdinary DEs.

One independent variable, usually time. The simplest first-order ODE: dy/dx = ky, whose solution y(x) = y₀ e^{kx} describes exponential growth (k > 0) or decay (k < 0). Radioactive decay, compound interest, and unconstrained population growth all follow this equation.

Second-order linear ODEs with constant coefficients describe oscillations: m ÿ + c ẏ + k y = 0 is the damped harmonic oscillator, with characteristic equation m λ² + c λ + k = 0. The roots determine whether the system oscillates, decays exponentially, or critically damps.

The major qualitative theorems: Picard–Lindelöf (existence and uniqueness for Lipschitz ODEs), Peano (existence under continuity alone), the Hartman–Grobman theorem on stability of equilibria. The mathematical theory is well-developed; computational solution is largely a solved problem in the linear regime and an open art in the nonlinear.

Chapter XIVPartial DEs.

Multiple independent variables. The three classical second-order linear PDEs:

The wave equation ∂²u/∂t² = c² ∇²u — vibrations on a string, sound, light. d'Alembert solved it in 1747.

The heat equation ∂u/∂t = α ∇²u — diffusion of heat, of probability, of any conserved quantity. Joseph Fourier's 1822 Théorie analytique de la chaleur introduced Fourier series to solve it.

Laplace's equation ∇²u = 0 — equilibrium states; harmonic functions; potential theory. The mother of complex analysis (every holomorphic function's real and imaginary parts are harmonic).

Nonlinear PDEs are vastly harder. The Navier–Stokes existence and smoothness problem — does the equation governing fluid flow have smooth solutions for all time? — is one of the seven Clay Millennium Prize Problems and remains open.

Chapter XVMultivariable calculus.

Calculus of functions of several variables. The derivative becomes the gradient ∇f = (∂f/∂x, ∂f/∂y, ∂f/∂z) — a vector pointing in the direction of steepest increase. The integral becomes a multiple integral — over areas, volumes, and higher-dimensional regions.

The chain rule, the change-of-variables formula (with its Jacobian determinant), and the multivariable Taylor expansion all generalise their single-variable analogues, with notation that becomes increasingly tensorial as dimensions grow.

Multivariable calculus is the everyday computational language of physics, statistics, optimisation, and machine learning. Gradient descent — start at a point, move in the direction of −∇f, repeat — is the workhorse of modern optimisation, including the backpropagation that trains every neural network.

Chapter XVIVector calculus.

The differential geometry of vector fields in three-dimensional space. Three operators: gradient (scalar to vector), divergence (vector to scalar), curl (vector to vector).

Maxwell's equations (1865) — the four equations governing electromagnetism — are the canonical use case: divergence of E equals charge density divided by ε₀; divergence of B equals zero; curl of E equals minus the time derivative of B; curl of B equals μ₀ J plus a time-derivative term. Four lines that describe every classical electromagnetic phenomenon.

The classical operators generalise to differential forms on manifolds, where they become a single operation — the exterior derivative d — and the classical theorems (Green, Stokes, Gauss) become a single generalised Stokes theorem: ∫_M dω = ∫_{∂M} ω. Élie Cartan's framework, now standard.

Chapter XVIIStokes's theorem.

For a smooth oriented surface S with boundary curve ∂S, and a smooth vector field F:

The flux of the curl through the surface equals the line integral of the field around the boundary. The theorem unifies several classical results: in two dimensions it reduces to Green's theorem; over a closed surface it implies Gauss's divergence theorem.

The result was first stated in a 1850 letter from Lord Kelvin to George Stokes; Stokes set it as a Smith's Prize examination question at Cambridge in 1854, where it was solved by James Clerk Maxwell.

The general Stokes theorem on differential forms — ∫_M dω = ∫_{∂M} ω — is one of the most beautiful identities in mathematics. The fundamental theorem of calculus, Green, Stokes, Gauss, and the higher-dimensional analogues are all instances of one statement.

Chapter XVIIIGreen's theorem.

The two-dimensional case of Stokes. For a positively-oriented simple closed curve C bounding a region D in the plane, and smooth functions P, Q:

George Green (1793–1841) was a self-taught miller's son in Nottinghamshire who published his foundational Essay on the Application of Mathematical Analysis to the Theories of Electricity and Magnetism in 1828 by subscription, sold to 51 subscribers locally.

The essay was forgotten until William Thomson (Kelvin) rediscovered it in 1845. By then Green's theorem had been independently rediscovered. The Green's function method — for solving inhomogeneous linear differential equations — is named for him and is foundational in physics, especially in quantum field theory and partial differential equations generally.

Chapter XIXEpsilon-delta.

The 19th-century rigour project. Bolzano, Cauchy, Weierstrass, Dedekind, and Cantor rebuilt calculus on the foundation of an explicitly defined real-number system and explicitly quantified limit definitions.

The motivating worries: pathological functions. Weierstrass's 1872 example of a continuous everywhere-but-differentiable-nowhere function shocked the field — many had assumed continuity implied differentiability except at isolated points. Dirichlet's function (1 on rationals, 0 on irrationals, 1829) is everywhere discontinuous yet representable in a Fourier-series-like fashion.

Real analysis as we now teach it — sequences, series, continuity, differentiability, Riemann integrability, uniform convergence, metric spaces — was largely codified between 1870 and 1900. Rudin's Principles of Mathematical Analysis (1953) and Real and Complex Analysis (1966) became the standard graduate texts; both remain in print.

Chapter XXLebesgue integration.

The Riemann integral handles continuous functions and most well-behaved discontinuous ones, but breaks down on more pathological functions — Dirichlet's, for instance, is not Riemann integrable.

Henri Lebesgue's 1902 thesis built a more powerful integral. Where Riemann partitions the domain and approximates by horizontal rectangles, Lebesgue partitions the range and asks how big the preimage of each interval is. The Lebesgue integral handles all bounded measurable functions and gives the right answer to many integrals where Riemann's fails.

The Lebesgue integral is the framework for modern probability theory, functional analysis, harmonic analysis, and partial differential equations. The dominated convergence theorem and Fatou's lemma — the central convergence results — are powerful in ways no Riemann-era theorem can match. Most working analysts default to Lebesgue without thinking about it.

Chapter XXIMeasure theory.

The infrastructure beneath Lebesgue integration. A measure on a set is a function that assigns to certain subsets a non-negative size — finitely or countably additive. Lebesgue measure on R^n generalises length, area, and volume.

The Banach–Tarski paradox (1924) shows that, assuming the axiom of choice, a solid ball in three-dimensional space can be decomposed into finitely many pieces and reassembled into two balls of the same size — a violation of measure that proves not every subset of R³ can be measure-assigned.

The standard remedy: restrict measures to a sigma-algebra of "measurable" sets. Lebesgue measure is defined on the Lebesgue-measurable subsets of R^n, which include all open and closed sets and pretty much anything you can construct without explicit recourse to choice. Borel sets are the smallest sigma-algebra containing the open sets and form the standard probability-theoretic playground.

Chapter XXIIStochastic calculus.

Calculus where the integrator is a random process. The motivating object is Brownian motion — a continuous random walk with stationary, independent, normally distributed increments.

Brownian paths are continuous everywhere but differentiable nowhere; they have unbounded variation on any interval. Standard Riemann–Stieltjes integration with respect to Brownian motion is undefined. Kiyosi Itô's 1942–51 work built a coherent calculus that interprets stochastic integrals as limits of left-endpoint Riemann sums, leading to Itô's lemma — the chain rule for stochastic calculus, which has an extra second-order term arising from the non-zero quadratic variation of Brownian paths.

Black, Scholes, and Merton's 1973 option pricing formula is the most famous application: the price of a European option is the solution of a parabolic PDE that arises from Itô calculus on a geometric Brownian motion. Modern quantitative finance is built on this foundation.

Chapter XXIIICalculus of variations.

Find the function — not the number — that minimises (or extremises) a functional. The brachistochrone problem: of all curves from point A to point B, which one minimises the time a sliding bead takes under gravity? Johann Bernoulli posed it in 1696; Newton, Leibniz, l'Hôpital, and Jacob Bernoulli all solved it (the answer: a cycloid).

The general framework: the Euler–Lagrange equation, derived by Euler and Lagrange in the 1750s. To extremise ∫ L(x, y, y') dx, the function y(x) must satisfy ∂L/∂y − d/dx (∂L/∂y') = 0.

Hamilton's principle (1834) reformulated all of classical mechanics as a calculus-of-variations problem: physical trajectories are critical points of an action functional. The same principle organises general relativity, quantum field theory, and string theory. The deepest physical laws are extremum principles in disguise.

Chapter XXIVCalculus on manifolds.

Calculus extended to spaces that locally look like Euclidean but globally curve. A smooth manifold has charts — coordinate maps to R^n — that overlap smoothly. Vector fields, differential forms, integration, and curvature all generalise to this setting.

The connection: how to compare vectors at different points on a curved manifold. The Levi-Civita connection (compatible with the metric, torsion-free) is uniquely determined by the metric and gives the parallel transport.

The key geometric quantities — curvature tensor, Ricci tensor, scalar curvature — measure how the manifold deviates from flatness. Gauss's Theorema Egregium (1827): the curvature of a surface is intrinsic — it can be computed from measurements made entirely within the surface, without reference to ambient space. This is why a flat map of the Earth must distort distances or angles.

Chapter XXVComplex analysis.

Calculus of functions of a complex variable. The differentiability condition (Cauchy–Riemann equations) is far stronger than in the real case: a complex-differentiable function is automatically infinitely differentiable, equal to its Taylor series, and rigid in ways real-analytic functions are not.

Three landmark results. Cauchy's integral theorem: the integral of a holomorphic function around any closed loop in a simply-connected domain is zero. Cauchy's integral formula: the value of a holomorphic function inside a contour is determined by its values on the contour. The residue theorem: contour integrals can be evaluated by summing the residues at enclosed singularities.

Complex analysis gives the cleanest proofs of many real-analytic facts (the fundamental theorem of algebra, the prime number theorem) and is indispensable in physics — quantum scattering, signal processing, fluid dynamics. Bernhard Riemann's 1851 thesis on Riemann surfaces opened the door to algebraic geometry by way of complex analysis.

Chapter XXVINumerical methods.

Most calculus problems in practice cannot be solved in closed form. Numerical methods compute approximations, with controlled error.

For derivatives: finite differences. For integrals: Simpson's rule, Gaussian quadrature, Monte Carlo for high-dimensional cases. For ODEs: Runge–Kutta methods (Carl Runge 1895; Wilhelm Kutta 1901), with the fourth-order RK4 still the workhorse for non-stiff problems. For PDEs: finite-difference, finite-element, finite-volume, and spectral methods.

Numerical analysis as a discipline emerged with the digital computer. John von Neumann's 1947 Numerical Inverting of Matrices of High Order founded the modern subject. Today's computational science — climate models, drug design, fluid simulation, the rendering pipeline of every animated film — is numerical calculus at scale.

Chapter XXVIICalculus in physics.

Physics is calculus, applied to nature. Newton's second law: F = m a = m d²x/dt². Maxwell's equations: a system of first-order PDEs in vector calculus. The Schrödinger equation: a complex-valued PDE. Einstein's field equations: a tensor equation in differential geometry.

The variational reformulations are even more striking. The principle of stationary action says physical trajectories extremise an action functional. Lagrangian mechanics derives the Euler–Lagrange equations from L = T − V. Hamiltonian mechanics formulates the same content as a flow on phase space governed by a single function.

Modern physics — quantum field theory, the standard model, general relativity, string theory — extends these ideas to fields and infinitely many degrees of freedom. The path integral (Feynman 1948) computes quantum amplitudes by integrating over function spaces. Calculus, taken to its limit, is what physical theory looks like.

Chapter XXVIIICalculus in economics.

Marginal analysis — the central tool of microeconomics — is differential calculus. Marginal utility, marginal cost, marginal revenue are derivatives. Optimisation under constraints (Lagrange multipliers) gives consumer demand, firm supply, and equilibrium.

The discipline turned mathematical in the late 19th century. William Stanley Jevons, Léon Walras, and Alfred Marshall brought calculus into economics; Paul Samuelson's Foundations of Economic Analysis (1947) remade graduate economics as applied mathematics.

Modern economics deploys advanced calculus throughout: dynamic optimisation (Bellman equations, optimal control), stochastic calculus (Black–Scholes, Merton's continuous-time consumption-investment models), partial differential equations (option pricing). Macroeconomic models are systems of ODEs. The boundary between mathematical economics and applied calculus has dissolved.

Chapter XXIXTwenty essentials.

• 1687Principia MathematicaNewton
• 1748Introductio in analysinEuler
• 1797Théorie des fonctions analytiquesLagrange
• 1821Cours d'analyseCauchy
• 1822Théorie analytique de la chaleurFourier
• 1902Lebesgue's ThesisLebesgue
• 1953Principles of Mathematical AnalysisRudin
• 1961CalculusApostol
• 1967CalculusSpivak
• 1965Calculus on ManifoldsSpivak
• 1966Real and Complex AnalysisRudin
• 1976Differential Equations, Dynamical SystemsHirsch & Smale
• 1978Calculus, Vol. I & IIApostol
• 1991Real AnalysisRoyden
• 1994The Cauchy–Schwarz Master ClassSteele
• 1996Calculus Made Easy (centennial)Thompson
• 1998The Geometry of PhysicsFrankel
• 2007Stochastic Calculus for FinanceShreve
• 2010CalculusStewart
• 2017The Calculus LifesaverBanner

Chapter XXXWatch & read.

↑ 3Blue1Brown · The Essence of Calculus · series opening

More on YouTube

Watch · Crash Course · Newton and Leibniz
Watch · The Fundamental Theorem of Calculus, Part 1