This chapter's toolkit reduces to a short list of building blocks, best remembered in groups.
Power, log, and exponential forms.∫xndx=n+1xn+1+c (n=−1, the power rule) and its edge case ∫x1dx=log∣x∣+c; ∫exdx=ex+c and ∫axdx=logaax+c; plus the two "do-nothing/scale" rules ∫kdx=kx+c and ∫0dx=c.
Trigonometric forms. The six basic derivative pairs run in reverse to give ∫sinxdx=−cosx+c, ∫cosxdx=sinx+c, ∫sec2xdx=tanx+c, ∫csc2xdx=−cotx+c, ∫secxtanxdx=secx+c, and ∫cscxcotxdx=−cscx+c. Alongside these sit four "integrated-by-manipulation" log-form results: ∫tanxdx=log∣secx∣+c, ∫cotxdx=log∣sinx∣+c, ∫secxdx=log∣secx+tanx∣+c, and ∫cscxdx=log∣cscx−cotx∣+c.
Inverse-trigonometric forms.∫1−x2dx=sin−1x+c and ∫1+x2dx=tan−1x+c are the two seeds from which every Type-I inverse-trig/log result of \S11.7.9 is built, simply by rescaling x→x/a.
The linearity and shift rules that make everything else work.∫kf(x)dx=k∫f(x)dx and ∫(f1(x)±f2(x))dx=∫f1(x)dx±∫f2(x)dx let a complicated integrand be split into pieces and handled one at a time; and if ∫f(x)dx=g(x)+c, then ∫f(ax+b)dx=a1g(ax+b)+c, so any already-known antiderivative can be reused whenever the variable is replaced by a linear expression ax+b.
The three general integration methods.Substitution replaces x by a new variable to turn an unfamiliar integrand into a recognisable one (or, in the ∫f′(x)[f(x)]ndx / ∫f(x)f′(x)dx shapes, is spotted directly without writing out a formal substitution). Decomposition rewrites an integrand -- most often a rational function -- as a sum of simpler pieces, via partial fractions or an algebraic/trigonometric identity, before integrating term by term. Integration by parts, ∫udv=uv−∫vdu, handles a product of two dissimilar functions (polynomial-times-trig, polynomial-times-exponential, or a lone log/inverse-trig factor); repeated application is systematised by Bernoulli's rule ∫udv=uv−u′v1+u′′v2−⋯, whose two signature results are the self-referencing pair