This is the case-atlas for the parent topic . The parent showed you the machinery. Here we drive it through every kind of pothole : each sign pattern, definite integrals whose limits force us to think about quadrants, a degenerate limit, a real-world arc, and an exam twist where the pattern is hidden.
Before symbols, one promise: every letter below is earned. a is always a fixed positive number (a length — think "radius"). x is our variable (a "position"). θ (theta) is a Greek letter we use for an angle , measured in radians. A radian is just an angle size; π radians = a straight line = 18 0 ∘ .
Intuition How we "machine-check" every answer
After each worked problem you'll see a Verify line. Every numeric answer and every antiderivative on this page is also re-checked by a computer in the ===VERIFY=== block: for an indefinite integral the machine differentiates our answer and confirms it equals the original integrand (differentiation is easy and unambiguous, so it's the perfect referee); for a definite integral it re-evaluates the number symbolically and compares. So "Machine-checked" means "a symbolic-algebra engine independently reproduced this."
Every trig-sub problem lands in exactly one of these cells. Our examples below cover all of them.
#
Cell (the scenario)
Which sub
Covered by
A
a 2 − x 2 , indefinite , root in denominator
x = a sin θ
Ex 1
B
a 2 − x 2 , definite integral (limits → quadrant care)
x = a sin θ
Ex 2
C
a 2 + x 2 , root on top , needs ∫ sec θ d θ
x = a tan θ
Ex 3
D
( a 2 + x 2 ) 2 power denominator, no visible root
x = a tan θ
Ex 4
E
x 2 − a 2 with x > 0 (branch [ 0 , 2 π ) )
x = a sec θ
Ex 5
F
Negative x region x < − a (branch [ π , 2 3 π ) ) — sign trap
x = a sec θ
Ex 6
G
Completing the square first (offset centre, hidden pattern)
x − h = a sin θ
Ex 7
H
Word problem / geometry — area of an ellipse cap (real-world)
x = a sin θ
Ex 8
I
Degenerate / limiting case: a → 0 , does the formula survive?
—
Ex 9
Worked example Ex 1 · Cell A —
∫ 16 − x 2 x 2 d x
Forecast: the radical is "number minus x 2 ", 16 = 4 2 . Guess the sub before reading on.
Step 1. Recognise a 2 − x 2 with a = 4 . Set x = 4 sin θ , so d x = 4 cos θ d θ , and restrict θ ∈ [ − 2 π , 2 π ] .
Why this step? Only 1 − sin 2 θ = cos 2 θ (a Pythagorean identity ) turns a difference of squares into a single square, killing the root.
Step 2. Simplify the root: 16 − 16 sin 2 θ = 16 cos 2 θ = 4 cos θ .
Why no ± ? On [ − 2 π , 2 π ] we have cos θ ≥ 0 , so ∣4 cos θ ∣ = 4 cos θ .
Step 3. Assemble. Since x 2 = 16 sin 2 θ :
∫ 4 c o s θ 16 s i n 2 θ ⋅ 4 cos θ d θ = 16 ∫ sin 2 θ d θ .
Why this step? The cos θ on top (from d x ) and bottom (from the root) cancel — that cancellation is the entire payoff of the sub.
Step 4. Power-reduce: sin 2 θ = 2 1 − c o s 2 θ (from Power-reduction & double-angle formulas ):
16 ⋅ 2 1 ( θ − 2 s i n 2 θ ) + C = 8 θ − 4 sin 2 θ + C = 8 θ − 8 sin θ cos θ + C ,
using sin 2 θ = 2 sin θ cos θ .
Why this step? sin 2 has no elementary antiderivative on sight; the double-angle identity linearises it into things we can integrate term by term.
Step 5. Back-substitute via the reference triangle : sin θ = 4 x , adjacent = 16 − x 2 , so cos θ = 4 16 − x 2 and θ = arcsin 4 x .
8 arcsin 4 x − 8 ⋅ 4 x ⋅ 4 16 − x 2 + C = 8 arcsin 4 x − 2 x 16 − x 2 + C .
Verify: the VERIFY block differentiates 8 arcsin 4 x − 2 x 16 − x 2 and confirms it simplifies to exactly 16 − x 2 x 2 . ✓
Worked example Ex 2 · Cell B —
∫ 0 3 9 − x 2 d x
Forecast: this is the area under a quarter of a circle of radius 3 . Predict the number before computing.
Step 1. a = 3 , x = 3 sin θ , d x = 3 cos θ d θ , root → 3 cos θ .
Why this step? Same a 2 − x 2 pattern.
Step 2 — convert the LIMITS, don't back-substitute. For a definite integral it's cleaner to push the bounds into θ :
x = 0 ⇒ 3 sin θ = 0 ⇒ θ = 0 .
x = 3 ⇒ 3 sin θ = 3 ⇒ sin θ = 1 ⇒ θ = 2 π .
Why this step? If we change the bounds now, we never have to undo the substitution — the answer pops out as a number.
Step 3. The integrand becomes from the root 9 − x 2 3 cos θ ⋅ from d x 3 cos θ d θ — two cosines multiply, one from replacing the root and one from d x :
∫ 0 π /2 3 cos θ ⋅ 3 cos θ d θ = 9 ∫ 0 π /2 cos 2 θ d θ .
Why do the two cosines appear? The substitution touches the integral in two places at once: the root 9 − x 2 becomes 3 cos θ , and the differential d x becomes 3 cos θ d θ . Their product is 9 cos 2 θ — that is why a cos 2 (not a bare cos ) shows up.
Step 4. cos 2 θ = 2 1 + c o s 2 θ :
9 [ 2 θ + 4 s i n 2 θ ] 0 π /2 = 9 ( 4 π + 0 − 0 ) = 4 9 π .
Why the sin 2 θ term vanishes: sin ( π ) = 0 and sin 0 = 0 — both endpoints kill it.
Verify: the VERIFY block re-integrates ∫ 0 3 9 − x 2 d x symbolically and gets 4 9 π ; this also matches the geometry, a quarter-disc of radius 3 : 4 1 π ( 3 ) 2 = 4 9 π . ✓
The figure below shows exactly what number we computed — the shaded quarter-disc.
Intuition Reading figure s01
The cyan curve is y = 9 − x 2 , the top half of a circle of radius 3 . The amber region is the area we integrated, from x = 0 to x = 3 . Because it is exactly one quarter of a full disc, its area must be 4 1 π r 2 = 4 9 π — the picture is the sanity check for the algebra.
Worked example Ex 3 · Cell C —
∫ x x 2 + 25 d x (take x > 0 )
Forecast: "number plus x 2 ", 25 = 5 2 . Guess the sub.
Step 1. a 2 + x 2 with a = 5 : set x = 5 tan θ , d x = 5 sec 2 θ d θ , θ ∈ ( − 2 π , 2 π ) .
Why this step? 1 + tan 2 θ = sec 2 θ turns a sum of squares into one square.
Step 2. Root: 25 + 25 tan 2 θ = 5 sec θ (with sec θ > 0 on that range). Substitute:
∫ 5 t a n θ 5 s e c θ ⋅ 5 sec 2 θ d θ = 5 ∫ t a n θ s e c 3 θ d θ .
Step 3. Convert to sin/cos to expose structure: tan θ sec 3 θ = sin θ / cos θ 1/ cos 3 θ = cos 2 θ sin θ 1 .
Why this step? Everything in sin/cos lets us hunt for a clean substitution or a known integral.
Step 4. Split cos 2 θ sin θ 1 = cos 2 θ sin θ sin 2 θ + cos 2 θ = cos 2 θ sin θ + sin θ 1 (used sin 2 + cos 2 = 1 ).
Why this step? The first piece is a perfect u = cos θ sub; the second is the standard ∫ csc θ d θ .
5 ∫ c o s 2 θ s i n θ d θ + 5 ∫ csc θ d θ = 5 ⋅ c o s θ 1 + 5 ln ∣ csc θ − cot θ ∣ + C = 5 sec θ + 5 ln ∣ csc θ − cot θ ∣ + C .
Step 5. Reference triangle: tan θ = 5 x → opposite x , adjacent 5 , hyp x 2 + 25 . So sec θ = 5 x 2 + 25 , csc θ = x x 2 + 25 , cot θ = x 5 .
= x 2 + 25 + 5 ln x x 2 + 25 − 5 + C .
Step 6 — the x < 0 branch (why we said "take x > 0 "). For x < 0 the substitution x = 5 tan θ needs a θ with tan θ < 0 , i.e. θ ∈ ( − 2 π , 0 ) , where sec θ is still > 0 — so the root is still 5 sec θ (roots are never negative), but now x = 5 tan θ < 0 while x 2 + 25 = 5 sec θ > 0 . Re-reading the triangle, csc θ = x x 2 + 25 becomes negative , and the absolute-value bars in 5 ln ∣ csc θ − cot θ ∣ absorb exactly that sign flip.
Why this matters: the antiderivative x 2 + 25 + 5 ln x x 2 + 25 − 5 + C is written with ∣ ⋅ ∣ precisely so it stays valid for both signs of x — the ∣ ⋅ ∣ is not decoration, it is what stitches the x > 0 and x < 0 branches into one formula.
Verify: the VERIFY block differentiates this answer (for x > 0 ) and confirms it collapses to x x 2 + 25 . ✓
Worked example Ex 4 · Cell D —
∫ ( x 2 + 4 ) 2 d x
Forecast: there's no ! But ( x 2 + 4 ) 2 = ( x 2 + 4 ) 4 . Guess the sub anyway.
Step 1. x = 2 tan θ , d x = 2 sec 2 θ d θ . Then x 2 + 4 = 4 sec 2 θ , so ( x 2 + 4 ) 2 = 16 sec 4 θ .
Why this step? The sign structure a 2 + x 2 dictates tan whether or not a root is written; tan makes x 2 + 4 collapse to a single sec 2 .
Step 2. ∫ 16 s e c 4 θ 2 s e c 2 θ d θ = 8 1 ∫ s e c 2 θ d θ = 8 1 ∫ cos 2 θ d θ .
Why this step? 1/ sec 2 θ = cos 2 θ — the powers of sec cancel to a friendly cos 2 .
Step 3. Power-reduce: 8 1 ⋅ 2 1 ( θ + 2 s i n 2 θ ) + C = 16 θ + 16 s i n θ c o s θ + C .
Step 4. Triangle: tan θ = 2 x , hyp x 2 + 4 , so sin θ = x 2 + 4 x , cos θ = x 2 + 4 2 , θ = arctan 2 x .
= 16 1 arctan 2 x + 16 1 ⋅ x 2 + 4 2 x + C = 16 1 arctan 2 x + 8 ( x 2 + 4 ) x + C .
Verify: the VERIFY block differentiates this answer and confirms it equals ( x 2 + 4 ) 2 1 . ✓
Worked example Ex 5 · Cell E —
∫ x 2 x 2 − 9 d x , take x > 3
Forecast: "x 2 minus number". Which sub?
Step 1. x 2 − a 2 , a = 3 : x = 3 sec θ , d x = 3 sec θ tan θ d θ . For x > 3 we're on the branch θ ∈ [ 0 , 2 π ) , where tan θ ≥ 0 .
Why this step? sec 2 θ − 1 = tan 2 θ removes the root; the branch choice keeps tan θ (hence the root) non-negative.
Step 2. Root: 9 sec 2 θ − 9 = 3 tan θ . Also x 2 = 9 sec 2 θ . Substitute:
∫ 9 s e c 2 θ ⋅ 3 t a n θ 3 s e c θ t a n θ d θ = 9 1 ∫ s e c θ d θ = 9 1 ∫ cos θ d θ .
Why this step? Everything cancels down to cos θ — the reward.
Step 3. = 9 1 sin θ + C .
Step 4. Triangle: sec θ = 3 x → hyp x , adjacent 3 , opposite x 2 − 9 , so sin θ = x x 2 − 9 .
= 9 x x 2 − 9 + C .
Verify: the VERIFY block differentiates 9 x x 2 − 9 (for x > 3 ) and confirms it equals x 2 x 2 − 9 1 . ✓
The reference triangle we used in Step 4 is drawn below.
Intuition Reading figure s02
This is the reference triangle for x = 3 sec θ . We chose the adjacent side = a = 3 and the hypotenuse = x , because sec θ = adj hyp = 3 x is exactly our substitution. Pythagoras then forces the opposite side to be x 2 − 9 — the very root in the problem. Reading sin θ = hyp opp = x x 2 − 9 straight off this picture is how Step 4 turns the angle back into x .
Worked example Ex 6 · Cell F —
∫ − 4 − 3 4 x x 2 − 4 d x (all x < − 2 )
Forecast: same shape as Ex 3 in the parent, but now x is negative . Will x 2 − 4 = 2 tan θ still be true? Predict.
Step 1. a = 2 , x = 2 sec θ . For x < − 2 , sec θ < − 1 , so we must use the branch θ ∈ [ π , 2 3 π ) . There tan θ ≥ 0 , so x 2 − 4 = 4 tan 2 θ = 2 tan θ (still + — the branch was chosen precisely so this holds).
Why this step? u 2 = ∣ u ∣ . Choosing the second branch guarantees tan θ ≥ 0 , so the absolute value drops cleanly — this is the whole reason for the [ π , 2 3 π ) restriction.
Step 2 — convert the limits, staying on the branch [ π , 2 3 π ) . On this branch each angle is the principal arccos value plus π (so it lands in quadrant III):
x = − 4 ⇒ sec θ = 2 x = − 2 ⇒ cos θ = − 2 1 . The quadrant-III angle with this cosine is θ 1 = π + 3 π = 3 4 π .
x = − 3 4 ⇒ sec θ = 2 x = − 3 2 ⇒ cos θ = − 2 3 . The quadrant-III angle with this cosine is θ 2 = π + 6 π = 6 7 π .
Why add π ? The principal arccos lands an angle in quadrant II (between 2 π and π ), where tan < 0 and our root identity would fail . Reflecting through by + π moves it to quadrant III on our chosen branch, where tan ≥ 0 and x 2 − 4 = 2 tan θ holds. The negative x is what forces these quadrant-III angles.
Step 3. With d x = 2 sec θ tan θ d θ and root 2 tan θ :
∫ 2 s e c θ 2 t a n θ ⋅ 2 sec θ tan θ d θ = ∫ 2 tan 2 θ d θ = 2 ∫ ( sec 2 θ − 1 ) d θ = 2 ( tan θ − θ ) + C .
Why sec 2 − 1 : tan 2 isn't directly integrable; tan 2 = sec 2 − 1 is.
Step 4 — evaluate between the branch angles. The endpoint x = − 4 is the lower x -limit and maps to θ 1 = 3 4 π ; the endpoint x = − 3 4 is the upper x -limit and maps to θ 2 = 6 7 π . So
∫ − 4 − 4/ 3 x x 2 − 4 d x = [ 2 ( tan θ − θ ) ] θ = 4 π /3 θ = 7 π /6 .
On this branch, tan 3 4 π = 3 and tan 6 7 π = 3 1 . Therefore
=2\left(\tfrac{1}{\sqrt3}-\sqrt3+\tfrac{\pi}{6}\right).$$
Numerically $\tfrac{1}{\sqrt3}-\sqrt3=-\tfrac{2}{\sqrt3}\approx-1.1547$, so the answer is $2(-1.1547+0.5236)\approx -1.262$.
**Step 5 — back-substitution sanity (express in $x$).** Since $\tan\theta=\tfrac{\sqrt{x^2-4}}{|x|}$ and here $x<0$ so $|x|=-x$, the antiderivative $2(\tan\theta-\theta)$ reads $-\dfrac{2\sqrt{x^2-4}}{x}-2\,\theta(x)$ in $x$-language; evaluating this $x$-form at $x=-4$ and $x=-\tfrac4{\sqrt3}$ reproduces the same $2\left(\tfrac1{\sqrt3}-\sqrt3+\tfrac\pi6\right)$. The integrand $\frac{\sqrt{x^2-4}}{x}$ is **negative** on $x<0$ (numerator $\ge0$, denominator $<0$), and we integrate left-to-right over positive width, so a **negative** answer is exactly expected. ✓
**Verify:** the VERIFY block re-computes $\int_{-4}^{-4/\sqrt3}\frac{\sqrt{x^2-4}}{x}\,dx$ symbolically; it equals $2\left(\tfrac{1}{\sqrt3}-\sqrt3+\tfrac{\pi}{6}\right)\approx -1.262$, and the sign is confirmed negative. ✓
> [!mistake] The branch trap
> For $x=-4$ it is tempting to write $\theta=\tfrac{2\pi}{3}$ (also has $\cos=-\tfrac12$). But $\tfrac{2\pi}{3}$ lies in quadrant II where $\tan\theta<0$, so $\sqrt{x^2-4}=2\tan\theta$ would be **false**. For $x<-a$ you must land in $[\pi,\tfrac{3\pi}{2})$ — hence $\tfrac{4\pi}{3}$, not $\tfrac{2\pi}{3}$.
The figure below shows why quadrant III is the correct branch: it is the only place where the substitution's tangent stays non-negative so the root identity survives.
Intuition Reading figure s03
The cyan unit circle carries two shaded wedges. The amber wedge in quadrant III (π ≤ θ < 2 3 π ) is our chosen branch for x < − 2 : there both sin and cos are negative, so tan θ = c o s s i n > 0 — exactly what makes x 2 − 4 = 2 tan θ (a non-negative root) valid. The white wedge in quadrant II is the trap: cos < 0 but sin > 0 , so tan θ < 0 there and the root identity would break. Our two limit-angles 6 7 π and 3 4 π both sit inside the amber wedge.
Worked example Ex 7 · Cell G —
∫ 5 + 4 x − x 2 d x
Forecast: there's a linear term 4 x , so no bare a 2 − x 2 . What must we do before substituting?
Step 1. Complete the square inside the root:
5 + 4 x − x 2 = − ( x 2 − 4 x ) + 5 = − ( ( x − 2 ) 2 − 4 ) + 5 = 9 − ( x − 2 ) 2 .
Why this step? Trig sub only recognises the three canonical shapes. Completing the square converts the shifted quadratic into a 2 − u 2 with u = x − 2 , a = 3 .
Step 2. Let u = x − 2 (so d u = d x ). Now ∫ 9 − u 2 d u , a standard a 2 − u 2 with a = 3 .
Why this step? This is now identically the parent's memorised form.
Step 3. u = 3 sin θ ⇒ ∫ 3 c o s θ 3 c o s θ d θ = θ = arcsin 3 u .
Step 4. Restore u = x − 2 :
∫ 5 + 4 x − x 2 d x = arcsin 3 x − 2 + C .
Verify: the VERIFY block differentiates arcsin 3 x − 2 and confirms d x d = 9 − ( x − 2 ) 2 1 = 5 + 4 x − x 2 1 . ✓
Worked example Ex 8 · Cell H — Area of an elliptical window cap
A stained-glass window is the ellipse 16 x 2 + 9 y 2 = 1 (units: metres). A horizontal glazing bar runs at y = 1.5 . Find the area of the cap above the bar (the region y ≥ 1.5 inside the ellipse).
Forecast: solving for x will produce number − y 2 . Guess the sub in y .
Step 1 — set up. From the ellipse, x = ± 4 1 − 9 y 2 = ± 3 4 9 − y 2 . At height y the cap stretches from x = − 3 4 9 − y 2 to x = + 3 4 9 − y 2 , a width of 2 ⋅ 3 4 9 − y 2 . So
A = ∫ 1.5 3 3 8 9 − y 2 d y .
Why this step? Area = sum of horizontal strip widths from the bar y = 1.5 up to the ellipse top y = 3 .
Step 2. Recognise 9 − y 2 , a = 3 : y = 3 sin θ , d y = 3 cos θ d θ , root → 3 cos θ .
Limits: y = 1.5 ⇒ sin θ = 2 1 ⇒ θ = 6 π ; y = 3 ⇒ sin θ = 1 ⇒ θ = 2 π .
Why this step? 9 − y 2 is the a 2 − y 2 pattern; converting the limits avoids back-substitution.
Step 3. Root 3 cos θ times d y = 3 cos θ d θ gives two cosines → a cos 2 :
A = 3 8 ∫ π /6 π /2 3 cos θ ⋅ 3 cos θ d θ = 24 ∫ π /6 π /2 cos 2 θ d θ .
Step 4. cos 2 θ = 2 1 + c o s 2 θ , and ∫ cos 2 θ d θ = 2 θ + 4 s i n 2 θ :
=24\left[\left(\frac\pi4+\frac{\sin\pi}{4}\right)-\left(\frac{\pi}{12}+\frac{\sin(\pi/3)}{4}\right)\right].$$
With $\sin\pi=0$ and $\sin\tfrac\pi3=\tfrac{\sqrt3}{2}$:
$$A=24\left(\frac\pi4-\frac\pi{12}-\frac{\sqrt3}{8}\right)=24\cdot\frac\pi6-24\cdot\frac{\sqrt3}{8}=4\pi-3\sqrt3\approx 7.371\ \text{m}^2.$$
*Why this step?* Power-reduction is the only way to integrate $\cos^2$; the endpoints then plug straight in.
**Verify (units + magnitude):** the whole ellipse area is $\pi(4)(3)=12\pi\approx 37.7\text{ m}^2$; our cap is a top slice, so a few m² is sensible. Units: $\text{m}\times\text{m}=\text{m}^2$. The VERIFY block re-integrates and confirms the exact value $4\pi-3\sqrt3$ and the numeric $7.371$. ✓
Intuition Reading figure s04
The cyan ellipse is the window; the dashed white line is the glazing bar at y = 1.5 . The amber region above it is the cap whose area we found, 4 π − 3 3 ≈ 7.37 m 2 . Compare its size to the whole ellipse (12 π ≈ 37.7 m 2 ): the shaded slice is clearly a modest fraction, matching our number.
Worked example Ex 9 · Cell I — does
∫ 0 a a 2 − x 2 d x behave sensibly as a → 0 ?
Forecast: geometrically this integral is a quarter-disc of radius a . As a shrinks to 0 , the disc shrinks to a point — the area should go to 0 . Predict how fast .
Step 1 — do the integral generally. Using x = a sin θ exactly as in Ex 2 (limits 0 → 2 π , root → a cos θ , d x → a cos θ d θ ):
∫ 0 a a 2 − x 2 d x = a 2 ∫ 0 π /2 cos 2 θ d θ = a 2 ⋅ 4 π = 4 π a 2 .
Why this step? Same a 2 − x 2 machinery; the a 2 factors out front, isolating the a -dependence.
Step 2 — the degenerate check a → 0 + . 4 π a 2 → 0 , and it does so quadratically (like a 2 ) — halving a quarters the area, exactly as a shrinking disc should. No blow-up, no undefined arcsin : its argument x / a stays in [ − 1 , 1 ] throughout the interval.
Why this step? Degenerate inputs are where formulas secretly divide by zero. Here the answer is a clean polynomial in a , so it survives the limit.
Step 3 — the other degeneracy: x → a from below. In the indefinite formula 2 a 2 arcsin a x + 2 x a 2 − x 2 , as x → a the term 2 x a 2 − x 2 → 0 and arcsin a x → arcsin 1 = 2 π — both finite . So even though the integrand's slope d x d a 2 − x 2 blows up (a vertical tangent) at x = a , the area is finite.
Why this matters: a vertical tangent of the integrand does not make the area infinite — only a vertical asymptote (integrand → ∞ ) would. Here the integrand itself stays bounded (0 ≤ a 2 − x 2 ≤ a ), so we're safe.
Verify: the VERIFY block confirms ∫ 0 a a 2 − x 2 d x = 4 π a 2 symbolically, that it → 0 as a → 0 , and that at a = 2 it equals π (a quarter of π ⋅ 2 2 ). ✓
Recall One-line reflex per scenario
Difference a 2 − x 2 ::: x = a sin θ ; limits → arcsin , quarter-disc geometry.
Sum a 2 + x 2 (root on top) ::: x = a tan θ ; watch for a sec 3 / csc integral.
Power ( a 2 + x 2 ) n no root ::: still x = a tan θ ; powers of sec cancel to cos powers.
Difference x 2 − a 2 , x > a ::: x = a sec θ , branch [ 0 , 2 π ) .
Difference x 2 − a 2 , x < − a ::: x = a sec θ , branch [ π , 2 3 π ) — sign trap!
Linear term present ::: complete the square first, then sub u = x − h .
Definite integral ::: convert the LIMITS to θ ; skip back-substitution.
Degenerate a → 0 or endpoint ::: check the formula stays finite; vertical tangent ≠ infinite area.
Parent topic (Hinglish) — the core machinery.
Pythagorean identities — kills each root.
Integration by substitution (u-sub) — the inner u -subs (Ex 3, Ex 7).
Power-reduction & double-angle formulas — every cos 2 / sin 2 step.
Reference right triangle method — all back-substitutions.
Partial fractions — the alternative when there is no root.
Arc length and surface area — where these a 2 ± x 2 integrals arise naturally.
Hyperbolic substitution — a parallel tool for x 2 ± a 2 .