Calculus and Beyond: Change, Structure and Uncertainty

After algebra, geometry and trigonometry, mathematics opens into several large fields, each answering a different kind of question. Calculus studies change, linear algebra studies structure, differential equations study change that depends on itself, and statistics studies what we can honestly say from incomplete data. This page describes what each accomplishes and offers "gates", concrete tasks a learner can use to check real understanding.

Calculus 1: limits, derivatives, integrals

Established Calculus starts with the limit, a precise way to say what a function approaches. The derivative is a limit of difference quotients, f'(a) = lim of [f(a + h) - f(a)]/h as h approaches 0, and measures instantaneous rate of change. The integral accumulates quantity, and the Fundamental Theorem of Calculus links the two: if f is continuous on [a, b] and F' = f, then the integral of f from a to b equals F(b) - F(a). Note the hypothesis. Applying it carelessly to a function with a break in the interval can give wrong answers; for example, 1/x^2 has an antiderivative -1/x, but the integral of 1/x^2 across 0 does not exist.

Gate tasks:

• Derive the derivative of x^2 directly from the limit definition.
• Solve an optimisation problem with units, for example the dimensions of the open box of greatest volume made by cutting squares from a sheet.
• State the Fundamental Theorem with its hypotheses and use it on one example where it applies and one where it does not.

Calculus builds on the function language in Geometry, Euclid and Trigonometry, where trigonometric and inverse functions were first handled.

Calculus 2: techniques and series

The second course covers integration techniques (substitution, parts, partial fractions), improper integrals, and infinite series. The key distinction is between a sequence (a list of terms) and a series (their sum). Terms going to zero is necessary for convergence but not sufficient: the harmonic series 1 + 1/2 + 1/3 + ... diverges although its terms shrink to zero. Learners should be able to choose a convergence test (ratio, comparison, integral, alternating series) and justify the choice, and find where a power series converges: 1/(1 - x) = 1 + x + x^2 + ... holds for |x| < 1, while the series for e^x converges for every x. Approximating a function by a partial sum requires an error bound, not just a plausible-looking answer.

Linear algebra: structure

Linear algebra studies vectors, matrices and linear maps. A learner should be able to describe the complete solution set of a system (none, one, or infinitely many), find a basis and the rank, interpret what a matrix does geometrically, and compute eigenpairs: nonzero vectors v with Av = lambda v. A least-squares fit is a projection of the data vector onto the column space of a matrix. Hefferon's free text is a good first course; Axler's Linear Algebra Done Right is a proof-based second pass. Both are listed in the sources.

Multivariable calculus

This extends calculus to several variables: gradients point toward steepest increase, constrained extrema use Lagrange multipliers, multiple integrals measure volume and mass, and the theorems of Green, Stokes and divergence relate an integral over a region to one over its boundary. Each has hypotheses about orientation and smoothness that a gate should require you to state.

Differential equations

A differential equation relates a function to its own rates. Typical order: slope fields and checking that a proposed solution satisfies both the equation and the initial condition; separable and first-order linear equations; second-order equations with forcing and resonance; Laplace transforms; linear systems and the stability read from eigenvalues.

A model worth building: Newton's law of cooling, dT/dt = -k(T - Tenv), with solution T(t) = Tenv + (T0 - Tenv)e^(-kt). Write the units of k, state the assumptions (uniform temperature, constant surroundings), and compare with measured data. Where it disagrees, you have learned something about the assumptions. This is the beginning of the thinking in Systems Thinking: Feedback, Queues, Information and Dynamics.

Statistics and probability

Established Core ideas: conditional probability, independence, expectation and variance, sampling distributions, confidence intervals and tests. Cautions every learner should be able to explain:

• A p-value is the probability of data at least as extreme as observed, assuming the null hypothesis and the model; it is not the probability that the null hypothesis is true (ASA statement, 2016).
• Statistical significance is not practical importance.
• Correlation is not causation; random assignment of treatments is what supports the strongest causal claims.
• Time-ordered measurements are often autocorrelated, so formulas that assume independent samples can mislead.

The practical use is developed in Mathematics for Better Decisions and Evidence-Based Troubleshooting: Separating Explanations.

Advanced branches

After the standard sequence, a learner can choose routes: real analysis (rigorous limits and continuity; Abbott's Understanding Analysis is a gentle start), calculus-based probability (Blitzstein and Hwang), mathematics for machine learning (Deisenroth, Faisal and Ong; least squares as projection, gradients, matrix decompositions), convex optimisation (Boyd and Vandenberghe) and feedback systems (Astrom and Murray). Each assumes the earlier gates. Graph and counting structures run alongside these in Discrete Mathematics and Graphs: Counting, Relations and Networks, and the whole landscape is charted in A Map of Mathematics: Twenty-Five Areas and What Depends on What.

Models versus proofs

Provisional A simulation of a differential equation or a probability model is evidence about the model, not a proof about the world and not a proof of the theorem. Compare a calculation with a simulation, but keep the assumptions explicit. Proof habits are in Proof and Precise Reasoning: From Arguments to Theorems.

Try this

1. Build the cooling model above for a cup of liquid, predict the temperature after a set time, then compare it with a thermometer reading and write down one assumption that failed.
2. Pick one convergent and one divergent series, and name the test that decides each.
3. For a small 2 by 2 matrix, find the eigenvalues by hand and verify Av = lambda v.

Further reading

• Gilbert Strang, Calculus (MIT OpenCourseWare edition).
• Jim Hefferon, Linear Algebra; Sheldon Axler, Linear Algebra Done Right.
• William Trench, Elementary Differential Equations.
• Wasserstein and Lazar, "The ASA statement on p-values," The American Statistician 70(2), 2016.
• Blitzstein and Hwang, Introduction to Probability; Stephen Abbott, Understanding Analysis.
• Deisenroth, Faisal and Ong, Mathematics for Machine Learning (Cambridge, 2020).
• Boyd and Vandenberghe, Convex Optimization; Astrom and Murray, Feedback Systems.