Atlas / Mathematics
Domain · 9 threads · 49 fields
Mathematics
How space split into many geometries and whole numbers into theories of their own, how calculus was made rigorous and mathematics met its own limits, and how solving equations became the study of symmetry. How counting puzzles became the mathematics of networks and equations of motion led to chaos, how reasoning from data and calculating by machine became mathematics of their own, and what forced each split.
Thread 01 · 9 fields
The Geometry Thread
From Euclid's axioms to the shape of three-dimensional space. One awkward postulate split geometry in two. The two halves met again in Riemann's lecture of 1854. What grew from that meeting eventually settled Poincaré's question about the shape of space, and it still runs into fog in dimension four. A second branch runs through the painter's perspective to the geometry of polynomial equations, and on to Fermat's Last Theorem.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- c. 300 BCEEuclidean GeometryWhat can be deduced about space from a handful of assumptions no one thinks to doubt?Root of the thread
- 1630s – 1820sProjective GeometryWhich properties of a figure survive projection, the way a painter projects a scene onto a canvas?Branched from Euclidean Geometry
- 18th century – 1820sDifferential Geometry of SurfacesHow do you measure the bending of a curve or surface with calculus, and how much of that bending can be detected from inside the surface?Branched from Euclidean Geometry + Calculus (The Analysis Thread)
- 1820s – 1830sNon-Euclidean GeometryWhat does geometry look like if the parallel postulate is simply false?Branched from Euclidean Geometry
- 1895 (roots from 1750)Algebraic TopologyWhich features of a shape survive any amount of stretching, and how can algebra detect them?Branched from Euclidean Geometry
- 1854Riemannian GeometryWhat is geometry when a space is known only from the inside, in any number of dimensions?Branched from Differential Geometry of Surfaces + Non-Euclidean Geometry
- 1850s – 1950sAlgebraic GeometryWhat shapes are defined by polynomial equations, and what does their geometry reveal about the equations' solutions?Branched from Projective Geometry + Algebraic Topology
- 1904 – 1980sGeometric TopologyCan every three-dimensional space be classified, and does geometry decide its shape?Branched from Algebraic Topology + Riemannian Geometry + Non-Euclidean Geometry
- 1930s – 1970sGeometric AnalysisWhat can the solutions of differential equations reveal about the shape of a curved space?Branched from Riemannian Geometry + Differential Geometry of Surfaces
Thread 02 · 5 fields
The Number Theory Thread
From Euclid's proof that the primes never end to the arithmetic that secures the internet. The whole numbers look like the simplest objects in mathematics, yet questions a child could ask about them have taken centuries. The effort to prove Fermat's Last Theorem split number theory in two: an analytic branch that counts primes with calculus, and an algebraic branch that builds new number systems when unique factorisation fails. Both rejoined geometry in the proof of Fermat's theorem, and their oldest problems now guard every encrypted connection. The fog here is some of the densest in mathematics: the Riemann hypothesis, the twin primes, abc.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- c. 300 BCE – 1801Elementary Number TheoryWhat can be proved about the whole numbers and how they divide one another?Branched from Euclidean Geometry (The Geometry Thread)
- 1832 – 1871Algebraic Number TheoryWhat happens to primes and factorisation when the whole numbers are extended to larger number systems?Branched from Elementary Number Theory
- 1737 – 1896Analytic Number TheoryHow are the primes distributed, and why can calculus answer questions about whole numbers?Branched from Elementary Number Theory + Complex Analysis (The Analysis Thread)
- 1922 – 1983Arithmetic GeometryWhat does the shape of an equation's solution set reveal about its solutions in whole or rational numbers?Branched from Algebraic Number Theory + Analytic Number Theory + Algebraic Geometry (The Geometry Thread)
- 1976 – 1985Public-Key CryptographyHow can two strangers communicate in secret without ever having shared a secret key?Branched from Elementary Number Theory + Arithmetic Geometry + Computational Complexity (The Foundations Thread)
Thread 03 · 5 fields
The Analysis Thread
From Archimedes' curved areas to the laws of chance. Calculus worked brilliantly from the day Newton and Leibniz invented it, but nobody could say what its infinitely small quantities were. Fourier's claim that any function is a sum of waves pushed intuition past breaking point, and repairing the foundations produced rigorous limits, a precise definition of the real numbers, and a theory of measure strong enough to put probability on firm ground. The fog here sits where waves concentrate: problems like Kakeya's needle, and the critical point of random networks.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- 1665 – 1700CalculusHow can quantities that change continuously be measured, and infinitely many infinitely small pieces be added up?Branched from Euclidean Geometry (The Geometry Thread)
- 1807 – 1829Fourier AnalysisCan every signal be built from simple waves, and what does that decomposition reveal?Branched from Calculus
- 1799 – 1851Complex AnalysisWhat happens to calculus when numbers are allowed to be complex?Branched from Calculus + Theory of Equations (The Algebra Thread)
- 1821 – 1902Real AnalysisWhat exactly are limits, continuity and the real numbers, and why did calculus need them defined?Branched from Calculus + Fourier Analysis
- 1654 – 1933Probability TheoryHow can chance be measured, and what laws does randomness obey in the long run?Branched from Calculus + Real Analysis
Thread 04 · 5 fields
The Foundations Thread
From Aristotle's syllogisms to the limits of computation. Around 1900 mathematicians tried to rest all of mathematics on one secure foundation: logic made exact, and sets as the universal building material. The attempt produced paradoxes, a bitter feud, and in 1931 Gödel's proof that no such foundation can ever be complete. Out of that wreckage came the theory of computation, which is the blueprint of every computer, and the deepest open question in computer science. The fog here is P versus NP, and the continuum hypothesis, a question the standard axioms cannot answer at all.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- 1847 – 1930Mathematical LogicWhat makes an argument valid, and can reasoning itself be turned into calculation?Root of the thread
- 1874 – 1908Set TheoryWhat is infinity, and can all of mathematics be built out of collections?Branched from Mathematical Logic + Real Analysis (The Analysis Thread)
- 1900 – 1936MetamathematicsWhat can mathematics prove about itself, about its own consistency, completeness and limits?Branched from Mathematical Logic + Set Theory
- 1931 – 1936Computability TheoryWhat can be computed at all, by any mechanical procedure?Branched from Metamathematics
- 1965 – 1972Computational ComplexityWhich problems can be solved efficiently, and why do some seem to need astronomical time?Branched from Computability Theory
Thread 05 · 5 fields
The Algebra Thread
From Babylonian recipes for finding an unknown to the mathematics of symmetry. For three thousand years algebra meant solving equations, and each advance, from the cubic formula to the proof that the quintic has none, came from a harder question about the roots. Galois answered the last of those questions with symmetry, and symmetry became the subject: groups, rings and fields, studied for their own sake after Emmy Noether, and represented as matrices that turned out to describe atoms and particles. The fog here includes the inverse Galois problem and the Jacobian conjecture.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- c. 820 – 1799Theory of EquationsHow can an unknown quantity be found from an equation, and is there always a formula for it?Root of the thread
- 1770 – 1846Galois TheoryWhy can some polynomial equations be solved by a formula and others not?Branched from Theory of Equations
- 1832 – 1882Group TheoryWhat is symmetry, and what are all the ways a structure can be transformed into itself?Branched from Galois Theory + Non-Euclidean Geometry (The Geometry Thread)
- 1843 – 1931Abstract AlgebraWhat do all number-like systems have in common, and can algebra be done without numbers at all?Branched from Group Theory + Algebraic Number Theory (The Number Theory Thread)
- 1873 – 1931Representation TheoryHow can abstract symmetries be made concrete as matrices, and what does that reveal?Branched from Group Theory + Abstract Algebra
Thread 06 · 5 fields
The Combinatorics Thread
From counting arrangements to the structure of networks. For most of its history combinatorics was a collection of puzzles: how many ways to choose, whether a walk can cross every bridge once, how many colours a map needs. In the twentieth century the puzzles became a subject. Ramsey showed that complete disorder is impossible, Erdős showed that randomness proves what explicit construction cannot, and the need to route, schedule and match at scale turned graphs into the mathematics of computing. The fog here is close to the surface: nobody knows the smallest party of guests guaranteed to contain five mutual friends or five mutual strangers.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- c. 200 BCE – 1964Enumerative CombinatoricsHow many ways can something be arranged, and can the answer be found without listing them all?One of the thread's roots
- 1736 – 1936Graph TheoryWhat can be said about a network from the pattern of its connections alone?One of the thread's roots
- 1916 – 1975Extremal CombinatoricsHow large can a structure grow before some pattern is forced to appear inside it?Branched from Graph Theory + Enumerative Combinatorics
- 1939 – 1979Combinatorial OptimisationAmong astronomically many possible arrangements, how can the best one be found without trying them all?Branched from Graph Theory + Computational Complexity (The Foundations Thread)
- 1947 – 1975Probabilistic CombinatoricsCan chance prove that something exists, and what does a typical large network look like?Branched from Extremal Combinatorics + Probability Theory (The Analysis Thread)
Thread 07 · 5 fields
The Dynamics Thread
From Newton's laws of motion to the limits of prediction. Differential equations promised that the future follows from the present, and for two centuries mathematicians tried to solve them. Most cannot be solved by formula, so Poincaré learned to describe their solutions without solving them, and in doing so found the first hint of chaos. Computers later showed chaos everywhere: deterministic systems whose long-term behaviour is unpredictable in practice, yet obeys laws of its own, from the statistics of ergodic theory to the infinite detail of the Mandelbrot set. The fog here includes Hilbert's sixteenth problem and whether the Mandelbrot set is locally connected.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- 1671 – 1890Differential EquationsIf we know how something is changing at every instant, can we work out where it will be?Branched from Calculus (The Analysis Thread)
- 1881 – 1967Dynamical SystemsWhat does a system do in the long run, when its equations can't be solved?Branched from Differential Equations
- 1961 – 1978Chaos TheoryHow can a system that follows exact rules be impossible to predict, and what order is hidden in its disorder?Branched from Dynamical Systems
- 1871 – 1959Ergodic TheoryWhen does the long-run behaviour of one trajectory match the average over all possible states?Branched from Dynamical Systems + Real Analysis (The Analysis Thread)
- 1918 – 1985Complex DynamicsWhat happens when a simple formula on the complex numbers is applied over and over again?Branched from Dynamical Systems + Complex Analysis (The Analysis Thread)
Thread 08 · 5 fields
The Statistics Thread
From combining the observations of astronomers to machines that learn from examples. Probability predicts data from a known chance mechanism. Statistics runs the argument backwards, from data to the mechanism, and for two centuries it was argued over as much as it was used. Least squares began in a priority dispute, Fisher and Neyman feuded over what a test means, and Bayesian reasoning was nearly banished before computers brought it back. Along the way Shannon measured information itself, Markov and Wiener gave laws to quantities that wander at random, and Vapnik and Valiant asked when a rule learned from examples can be trusted. The fog here is close to daily life: how to make published findings reliable, and why giant neural networks generalise when the theory says they should not.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- 1805 – 1935Statistical InferenceHow can conclusions about a whole population be drawn from a limited sample, with a known risk of being wrong?Branched from Probability Theory (The Analysis Thread)
- 1924 – 1952Information TheoryHow much information does a message contain, and how fast can it be sent reliably through a noisy channel?Branched from Probability Theory (The Analysis Thread)
- 1900 – 1944Stochastic ProcessesWhat laws govern quantities that change randomly over time?Branched from Probability Theory (The Analysis Thread) + Real Analysis (The Analysis Thread)
- 1774 – 1990Bayesian StatisticsHow should a degree of belief be updated as evidence arrives, and can probability measure belief at all?Branched from Probability Theory (The Analysis Thread) + Statistical Inference
- 1958 – 1995Statistical Learning TheoryWhen can a rule learned from examples be trusted on cases it has never seen?Branched from Statistical Inference + Information Theory + Computational Complexity (The Foundations Thread)
Thread 09 · 5 fields
The Computation Thread
From Newton's method to the training of neural networks. For centuries, numbers were computed by hand, by people who followed rules and made mistakes, and the question was only how to get an answer at all. Electronic computers answered that and raised a harder question: can an answer produced by billions of rounded operations be trusted? Turing and Wilkinson showed how to tell a bad method from a bad problem. Richardson's failed weather forecast became a daily routine once the grid was made to keep up with the physics, games of chance on the ENIAC became Monte Carlo methods, and Cauchy's idea of walking downhill now trains artificial intelligence. The fog here is how fast two matrices can be multiplied, and why gradient descent trains deep networks as well as it does.
Scroll the map sideways →
- Charted: settled results
- Contested: sources disagree on priority or causation
- Unmapped: open problems
Fields in this thread
- 1669 – 1963Numerical AnalysisHow can a finite machine, doing finitely many rounded operations, give answers we can trust?Branched from Calculus (The Analysis Thread)
- 1922 – 1977Numerical Methods for PDEsHow do you turn the continuous equations of physics into a finite computation whose answer converges to the truth?Branched from Numerical Analysis + Differential Equations (The Dynamics Thread)
- 1847 – 1970Continuous OptimisationHow do you find the lowest point of a function of thousands or billions of variables?Branched from Numerical Analysis + Real Analysis (The Analysis Thread)
- 1946 – 1953Monte Carlo MethodsHow can randomness compute answers that careful deterministic methods cannot reach?Branched from Numerical Analysis + Probability Theory (The Analysis Thread)
- 1947 – 1969Numerical Linear AlgebraHow do you solve a million linear equations on a machine that rounds every operation, and know the answer is right?Branched from Numerical Analysis