U-substitution — technique, change of limits for definite integrals: Visual walkthrough

Step 1 — What the integral sign is even asking

WHAT. We are stacking thin rectangles under a curve and adding their areas.

WHY. Every symbol in U-substitution is a manipulation of these slivers. If we do not know that $d x$ is a width — a real geometric length — the phrase " $d u = g^{'} (x) d x$ " is meaningless. So we anchor it here first.

PICTURE. Look at the single amber strip: its top touches the curve at height $h (x)$ , its base has width $d x$ . The whole shaded region is the sum of thousands of such strips.

Figure — U-substitution — technique, change of limits for definite integrals

Step 2 — Two rulers for the same road: introducing $u = g (x)$

We now put a second measuring tape along the bottom axis.

WHAT. We laid a curved second ruler ( $u$ ) underneath the straight ruler ( $x$ ). Because $g$ is usually not a straight line, equal steps in $x$ become unequal steps in $u$ .

WHY this tool and not another? We rename because the integrand secretly contains $g (x)$ inside it. The whole point (from the parent) is that integration reverses the Chain Rule: differentiating $F (g (x))$ produced an inner $g (x)$ and an outer $F^{'}$ . Renaming the inner function to a single letter $u$ collapses the mess.

PICTURE. Notice how three evenly spaced ticks on the $x$ -ruler (bottom) map to three unevenly spaced ticks on the $u$ -ruler (the curved arrows). That unevenness is the entire story of Step 3.

Step 3 — WHY $d u = g^{'} (x) d x$ : the stretch factor of the ruler

Here is the heart of the picture. When you move one tiny step $d x$ on the old ruler, how big a step $d u$ do you take on the new ruler?

WHAT. We measured how much the new ruler is stretched or squeezed at each point.

WHY. A definite integral is a sum of areas. If we switch to the $u$ -ruler but the slivers keep their old widths, we would be adding the wrong widths and get the wrong area. The factor $g^{'} (x)$ is the correction that keeps every sliver's area honest. It converts old-width $d x$ into new-width $d u$ .

PICTURE. The two triangles show the same rise measured against different runs. Where $g$ is steep (big $g^{'}$ ), a small $d x$ fans out into a big $d u$ — watch the wide amber gap on the $u$ -axis.

Step 4 — The chain-rule reversal, drawn

Now we connect the picture to the actual theorem. Differentiation of a composed function produces two factors; we run the machine backwards.

WHAT. We recognised that the messy factor $g^{'} (x) d x$ sitting in the integrand is nothing but $d u$ in disguise, and $f (g (x))$ is just the height $f (u)$ .

WHY it must be $g^{'} (x)$ present, not some other factor. U-substitution only "clicks" when the integrand already carries its own stretch factor $g^{'} (x)$ . That is the fingerprint the chain rule leaves behind. If it is absent (or off by a variable), Step 4 fails — see Step 6.

PICTURE. Same shaded region as Step 1, now with both rulers drawn. The strip is labelled twice: $f (g (x)) d x$ on the bottom ruler and $f (u) d u$ on the top ruler. One shape, two names.

Step 5 — WHY the limits must move (definite integrals)

The two walls $x = a$ and $x = b$ were positions on the old ruler. Once we throw the old ruler away and read only $u$ , those walls need their new names.

WHAT. We looked up where the two boundary walls land on the $u$ -ruler.

WHY. If we lazily kept $\int_{a}^{b}$ after switching to $u$ , we would be integrating between the positions labelled $a$ and $b$ on the new ruler — a completely different stretch of road. The area would be wrong.

PICTURE. The left wall and right wall are drawn as vertical amber lines. Follow each down to the bottom ruler ( $a, b$ ) and across to the top ruler ( $g (a), g (b)$ ). Same walls, relabelled.

Step 6 — The degenerate case: when the stretch factor vanishes or reverses

We must cover every scenario, including the ones that break the naive picture.

WHAT. We checked what happens when the stretch factor is zero or negative.

WHY. A reader who only ever saw an increasing $g$ would panic when the new limits come out "backwards" ( $5 \to 1$ instead of $1 \to 5$ ). The picture shows this is not an error — it is orientation bookkeeping.

PICTURE. Left panel: $g^{'} = 0$ , three $x$ -ticks collapse onto one $u$ -tick (crush). Right panel: $g$ decreasing, arrows cross — $x = a$ lands to the right of $x = b$ on the $u$ -ruler.

Step 7 — Worked example, seen as area

PICTURE. Left: the original curve $x x^{2} + 1$ over $x \in [0, 2]$ , shaded. Right: the transformed curve $\frac{1}{2} u$ over $u \in [1, 5]$ , shaded. The two shaded areas are equal — that equality is U-substitution. Both are annotated with the same number $3.39$ .

Recall Predict before you peek

Because $u = x^{2} + 1$ grows fast, will the new limits be wider apart or narrower than $0 \to 2$ ? ::: Wider — $1 \to 5$ (span 4 vs span 2), because the steep stretch factor $g^{'} = 2 x$ pulls positions apart.

The one-picture summary

Everything above in a single diagram: the same amber region, measured under the $x$ -ruler (left) and the $u$ -ruler (right), with the walls relabelled and the stretch factor $g^{'} (x)$ connecting the two widths. Read it left-to-right and you have re-derived the whole theorem.

Recall Feynman retelling (explain the whole walkthrough to a friend)

Imagine a road with a normal tape measure ( $x$ ) glued along it, and the area under a fence is what we want. Now glue a stretchy tape measure ( $u = g (x)$ ) on top: where the road is "busy," the stretchy tape spreads its numbers far apart; where it is calm, they bunch up. The amount the tape is stretched at each spot is its slope $g^{'} (x)$ — that is why a derivative shows up. When we measure the fence's area with the stretchy tape, each thin strip keeps the same real area (a taller-looking strip is just narrower to compensate, and vice-versa) — that is the equation $f (g (x)) g^{'} (x) d x = f (u) d u$ . Finally, the two end-posts of the fence didn't move an inch, but on the stretchy tape they now read different numbers $g (a)$ and $g (b)$ — that is why we change the limits. If the stretchy tape ever runs backwards ( $g^{'} < 0$ ), the two posts swap order, and the built-in minus sign fixes the answer for free.

Connections

Chain Rule — the fingerprint $g^{'} (x)$ inside the integrand comes from here; U-sub reverses it.
Definite Integral & Fundamental Theorem of Calculus — the "area = sum of slivers" picture in Step 1.
Antiderivatives & Indefinite Integrals — what we compute once the integral is in $u$ .
Integration by Parts — the product-rule inverse; contrast the geometry.
Trigonometric Substitution — a stretchy ruler shaped like $sin$ or $tan$ .
Hinglish version