Paralleling Section 5, the square-root forms ∫dx/a2−x2 and
∫dx/x2±a2 are each derived once by a trigonometric substitution and then reused
for every quadratic radicand via completing the square.
Derivation of ∫dx/a2−x2 by trigonometric substitution. Let x=asinθ,
θ∈(−π/2,π/2), so dx=acosθdθ and
a2−x2=a2−a2sin2θ=acosθ (positive, since cosθ≥0 on this
interval). Then
∫a2−x2dx=∫acosθacosθdθ=∫dθ=θ+C=sin−1(ax)+C,
verified by dxdsin−1(x/a)=1−x2/a21/a=a2−x21.
The other two square-root log forms (by the substitutions x=atanθ and
x=asecθ respectively, omitted here but directly checkable by differentiation):
∫x2+a2dx=lnx+x2+a2+C,∫x2−a2dx=lnx+x2−a2+C.
A quadratic radicand ax2+bx+c is handled exactly as in Section 5 -- factor out a,
complete the square to a[u2±k2] or a[k2−u2] with u=x+b/(2a) -- reducing to one of
the three forms above. Exercise: Standard Forms -- Square Roots applies this to
∫dx/x2+6x+13 and ∫(x+3)dx/x2+6x+13 (numerator split exactly as in
Section 5, since the derivative of the radicand again equals a multiple of the numerator).
Integrating the square root itself: ∫a2−x2dx and ∫x2−a2dx.
The same substitutions, now applied to the radical rather than its reciprocal, give (after using
the double-angle identity cos2θ=(1+cos2θ)/2 -- carried out in full in Example 11):
∫a2−x2dx=2xa2−x2+2a2sin−1(ax)+C, …