Unit 6: Integration and Accumulation of Change
- Syllabus
- 2020
- Section
- —
- Level
- —

The signed area between a rate graph and the horizontal axis over an interval is the accumulated change in the original quantity. Area above the axis contributes positively; area below the axis contributes negatively.
\text{accumulated-change units}=(\text{rate units})(\text{input units})
Split the region wherever the graph crosses the axis or changes geometric shape. Find each rectangle, triangle, or other familiar area, attach a positive or negative sign according to its position relative to the axis, and add the signed contributions. If an initial amount is given, then final amount = initial amount + net accumulated change.
Suppose a tank's net flow-rate graph, in liters per minute, forms a triangle above the axis with base 4 minutes and height 6 liters per minute, followed by a triangle below the axis with base 2 minutes and height 3 liters per minute. The net change is 21(4)(6)−21(2)(3)=12−3=9 liters. If the tank began with 20 liters, it ends with 29 liters.
Net change is not the same as total geometric area or final amount. In the example, the total unsigned area is 12+3=15 liters, the net change is 9 liters, and the final amount is 29 liters. Preserve the sign before interpreting the context.
To approximate ∫abf(x)dx, partition [a,b] at a=x0<x1<⋯<xn=b. On each subinterval, multiply its own width Δxi=xi−xi−1 by a height chosen by the stated method, then add. This works for equal or unequal widths.
| Method | Height on [xi−1,xi] | Useful error clue |
|---|---|---|
| Left | f(xi−1) | increasing: under; decreasing: over |
| Right | f(xi) | increasing: over; decreasing: under |
| Midpoint | f((xi−1+xi)/2) | concave up: under; concave down: over |
| Trapezoidal | [f(xi−1)+f(xi)]/2 | concave up: over; concave down: under |
L=\sum_{i=1}^n f(x_{i-1})\Delta x_i,\quad R=\sum_{i=1}^n f(x_i)\Delta x_i,\quad T=\sum_{i=1}^n \frac{f(x_{i-1})+f(x_i)}{2}\Delta x_i
Suppose a table gives x=0,1,3,4 and f(x)=2,4,5,3. The widths are 1,2,1, not all equal. The left sum is 2(1)+4(2)+5(1)=15. The right sum is 4(1)+5(2)+3(1)=17. The trapezoidal sum is 22+4(1)+24+5(2)+25+3(1)=16.
Do not replace every Δxi by (b−a)/n unless the partition is uniform. An error direction also requires the relevant behavior across the interval: monotonicity supports left/right judgments, while concavity supports midpoint/trapezoidal judgments. Without that information, calculate the approximation but do not guess over or under.
A Riemann sum approximates accumulated change on [a,b] by adding contributions from a partition a=x0<x1<⋯<xn=b. In subinterval i, the product f(xi∗)Δxi uses a sampled function value as the rectangle height and Δxi=xi−xi−1 as its width.
\int_a^b f(x),dx=\lim_{\max \Delta x_i\to 0}\sum_{i=1}^{n} f(x_i^*)\Delta x_i
Refining the partition makes every rectangle narrower, so the sum follows the changing function more closely. The condition maxΔxi→0 matters: it prevents even one subinterval from remaining wide. For a continuous function, different valid choices of sample points approach the same definite integral.
For a positive rate r(t) on [0,4], a sum ∑r(ti∗)Δti adds approximate changes over short time intervals. As the largest time width approaches zero, the approximation approaches ∫04r(t)dt, the exact accumulated change over those four time units.
The integral is a limit of signed contributions, not automatically ordinary geometric area. Terms are negative where f(xi∗)<0, so regions below the axis reduce the accumulated value.
To translate notation, match four pieces: the interval [a,b], the width Δx, the sample point xi∗, and the function evaluated there. For n equal subintervals, Δx=(b−a)/n; a right-endpoint sample is xi=a+iΔx.
In limn→∞∑i=1n(1+3i/n)2(3/n), the width is 3/n, so b−a=3. The sample point is 1+3i/n, so a=1 and b=4. The squared factor is f(xi). Therefore the limit is ∫14x2dx.
For ∫021+x3dx, use Δx=2/n and right endpoints xi=2i/n. An equivalent limit is limn→∞∑i=1n1+(2i/n)3(2/n). The factor outside the function is the subinterval width.
Do not infer the upper endpoint from Δx alone; combine the width with the sample-point formula. Also distinguish i/n, which locates a sample point, from the separate width factor that multiplies every function value.
An accumulation function uses a fixed starting point and a moving endpoint: A(x)=∫axf(t)dt. For each input x, A(x) is the signed change accumulated from a to x. The letter t is a dummy integration variable, keeping it distinct from the endpoint x.
A(x)=\int_a^x f(t),dt \quad\Longrightarrow\quad A'(x)=f(x)
If f is continuous on an interval containing a and x, increasing the endpoint by a small amount Δx adds approximately f(x)Δx to the accumulation. Dividing by Δx and taking the limit leaves f(x). This is the Fundamental Theorem connection between integration and differentiation.
Let A(x)=∫0x(2t+1)dt. Then A(0)=0, because no interval has yet been accumulated. The Fundamental Theorem gives A′(x)=2x+1. Evaluating directly, A(x)=x2+x, so A(2)=6 and differentiating this expression confirms the same derivative.
A(x) is signed accumulation, not necessarily total geometric area: negative values of f subtract. Also, the stated derivative rule assumes continuity on the relevant interval and a variable upper limit exactly equal to x; do not silently drop those conditions.
For g(x)=∫axf(t)dt, the value g(x) is the signed area accumulated from a to x, while the Fundamental Theorem gives g′(x)=f(x). Thus a representation of f reveals both how much has accumulated and how g is changing.
| Information about f | Conclusion about g |
|---|---|
| f>0 / f<0 | g increases / decreases |
| f changes +→− / −→+ | g has a local maximum / minimum |
| f increases / decreases | g is concave up / concave down, since g′′=f′ |
| signed area from a to x | value of g(x) |
Suppose the signed area under f from a to c is 5, and the area from c to d lies below the axis with magnitude 2. Then g(c)=5 and g(d)=5−2=3. On (c,d), f<0, so g is decreasing, yet g remains positive. If f changes from positive to negative at c, then g has a local maximum there.
A zero of f means g′(x)=0; it is only a critical point of g, not automatically a zero of g. A zero of g occurs when the net signed area from a to x is zero. Likewise, f>0 tells whether g increases, not whether g itself is positive.
Use definite-integral properties when a requested integral can be built from known values or familiar geometric regions. Keep two sources of sign visible: regions below the axis contribute negatively, and reversing the integration limits changes the sign.
| Property | Rule |
|---|---|
| Constant multiple | ∫abcf(x)dx=c∫abf(x)dx |
| Sum | ∫ab(f+g)dx=∫abfdx+∫abgdx |
| Reversed limits | ∫baf(x)dx=−∫abf(x)dx |
| Adjacent intervals | ∫acfdx+∫cbfdx=∫abfdx |
If ∫02f(x)dx=5 and ∫27f(x)dx=−1, then ∫07f(x)dx=5+(−1)=4. Therefore ∫703f(x)dx=3[−∫07f(x)dx]=3(−4)=−12. The negative result comes from reversing the limits, not from discarding the given signed value.
If a graph on [−2,2] is the upper semicircle of radius 2, its integral is the semicircle's area, 21π(2)2=2π. The same semicircle below the axis would contribute −2π.
A removable discontinuity or a jump discontinuity can still allow a definite integral; changing one isolated function value does not change accumulated area because a point has zero width. This extension does not mean every discontinuous or unbounded function is automatically integrable.
An antiderivative of f is any function F satisfying F′(x)=f(x). If f is continuous on [a,b], the Fundamental Theorem of Calculus evaluates its definite integral by the net change in an antiderivative.
\int_a^b f(x),dx=F(b)-F(a)\qquad\text{when }F'=f
First confirm that the integrand is continuous on the interval. Find one antiderivative F, substitute the upper endpoint and lower endpoint separately, and compute upper minus lower. A quick derivative check of F protects against power-rule and coefficient errors.
Evaluate ∫13(2x2−4x+1)dx. A polynomial is continuous, and one antiderivative is F(x)=32x3−2x2+x. Then F(3)=3 and F(1)=−31, so ∫13(2x2−4x+1)dx=3−(−31)=310.
Use F(b)−F(a), not the reverse. You may write +C while describing the family of antiderivatives, but it is unnecessary in a definite-integral evaluation because (F(b)+C)−(F(a)+C) cancels the constant.
The indefinite integral ∫f(x)dx means the complete family F(x)+C whose derivative is f(x). The constant is necessary because all vertical shifts of F have the same derivative.
| Integrand | Antiderivative |
|---|---|
| xn, n=−1 | xn+1/(n+1)+C |
| 1/x, x=0 | ln∣x∣+C |
| ex | ex+C |
| cosx | sinx+C |
| sinx | −cosx+C |
Reverse the matching derivative rule for each term. Constants factor out and sums integrate term by term. After combining the terms, write one +C for the entire antiderivative family and differentiate the result to verify it.
For x=0, ∫(6x2−4/x+3ex)dx=2x3−4ln∣x∣+3ex+C. Differentiating gives 6x2−4/x+3ex, exactly the original integrand, so the coefficients, signs, and logarithm condition are consistent.
Not every function has a closed-form antiderivative; for example, e−x2 cannot be expressed using the usual finite collection of elementary functions. That does not mean a related definite integral is meaningless: it may still be represented by an accumulation function or approximated numerically.
Use substitution when an integrand contains a composite expression together with its derivative, possibly differing by a constant factor. Setting the inner expression equal to u reverses the chain rule and turns the integral into a familiar function of one variable.
Choose u as the inner expression; compute du; rewrite every factor and the differential in terms of u; integrate; then check by differentiating. For an indefinite integral, substitute the original expression back and add C. For a definite integral, convert both bounds using the same substitution before evaluating.
For ∫6x(3x2+1)4dx, let u=3x2+1, so du=6xdx. Then ∫u4du=u5/5+C, giving (3x2+1)5/5+C. Differentiation returns the original integrand.
For ∫012xex2dx, let u=x2 and du=2xdx. The original bounds become u(0)=0 and u(1)=1, so the integral is ∫01eudu=[eu]01=e−1.
Do not leave both x and u in the transformed integrand. For a definite integral, either change the bounds and finish entirely in u, or find an antiderivative, return to x, and use the original bounds. Mixing new bounds with an x-expression gives an invalid evaluation.
Before choosing a new integration rule, check whether algebra can expose a familiar one. Polynomial long division separates a polynomial part from a proper rational remainder; completing the square converts a quadratic into a shifted square plus a constant.
| Integrand feature | Rewrite | Structure revealed |
|---|---|---|
| numerator degree ≥ denominator degree | polynomial long division | polynomial terms plus a proper fraction |
| quadratic x2+bx+c | (x+b/2)2+c−b2/4 | shifted u2+a2 or related form |
For x=−1, x+1x2+1=x−1+x+12. Therefore ∫x+1x2+1dx=2x2−x+2ln∣x+1∣+C. Multiplying the quotient by the divisor and adding the remainder verifies the rewrite.
Since x2+4x+8=(x+2)2+4, ∫−20x2+4x+8dx=[21arctan(2x+2)]−20=21(π/4−0)=π/8. The completed square identifies the inverse-tangent form.
A rewrite must be algebraically equivalent on the original domain. Long division does not restore a denominator's excluded zeros, and completing the square does not change definite-integral bounds unless a separate variable substitution is introduced.
The product rule gives (uv)′=u′v+uv′. Integrating and rearranging produces integration by parts: choose one factor as u to differentiate and the remaining factor with dx as dv to integrate.
\int u,dv=uv-\int v,du
| Step | Decision | Check |
|---|---|---|
| 1 | choose u | differentiating it should simplify it |
| 2 | choose dv | it must have an antiderivative you can find |
| 3 | find du and v | include dx consistently |
| 4 | substitute into the rule | the new integral should be simpler than the original |
For ∫xexdx, take u=x and dv=exdx. Then du=dx and v=ex, so ∫xexdx=xex−∫exdx=xex−ex+C. Differentiating the result returns xex.
For a definite integral, keep the bounds: ∫01xexdx=[xex]01−∫01exdx=[xex−ex]01=0−(−1)=1. The result is a number, so no +C is added.
Do not choose u by a memorized ordering alone: the real test is whether u is easier after differentiation and dv is easy to integrate. For definite integrals, either retain the original bounds throughout or first find an antiderivative and then evaluate both endpoints—do not mix the two methods or drop a boundary term.
Use linear partial fractions when a rational function is proper (numerator degree is less than denominator degree) and the denominator factors into distinct linear factors. For (x−a)(x−b) with a=b, write (x−a)(x−b)P(x)=x−aA+x−bB, then solve for A and B.
| Step | Action | Purpose |
|---|---|---|
| 1 | if necessary, use long division | make the rational part proper |
| 2 | factor the denominator | identify each distinct linear factor |
| 3 | assign one constant numerator per factor | build the partial-fraction form |
| 4 | multiply through by the denominator and solve | determine the constants |
| 5 | integrate each simple fraction | obtain logarithmic terms |
Decompose (x+1)(x+2)3x+5=x+1A+x+2B. Multiplying through gives 3x+5=A(x+2)+B(x+1), so A=2 and B=1. Therefore ∫(x+1)(x+2)3x+5dx=2ln∣x+1∣+ln∣x+2∣+C.
Since (x+1)(x+2)1=x+11−x+21, ∫01(x+1)(x+2)dx=[ln(x+1)−ln(x+2)]01=ln(4/3). The interval contains no denominator zero, so the integrand is continuous there.
This card covers distinct, nonrepeating linear factors only. Do not use the same template unchanged for repeated factors or irreducible quadratic factors. Keep ln∣x−a∣ in an indefinite result, and remember that decomposition does not remove excluded values from the original denominator.
An integral is improper when an endpoint is infinite or the integrand becomes unbounded on the interval. Replace each improper feature by a limit of proper definite integrals; the improper integral converges only when every required limit exists and is finite.
| Improper feature | Required rewrite |
|---|---|
| upper endpoint ∞ | ∫a∞f(x)dx=limb→∞∫abf(x)dx |
| unbounded at endpoint a | ∫acf(x)dx=limt→a+∫tcf(x)dx |
| unbounded at interior point c | split at c and evaluate two one-sided limits separately |
For an infinite interval, ∫1∞x−2dx=limb→∞[−x−1]1b=limb→∞(1−1/b)=1. The finite limit means the integral converges to 1.
The integrand 1/x2 is unbounded at 0, so ∫−11x−2dx must be split. On the right, limt→0+∫t1x−2dx=limt→0+(1/t−1)=∞. Because one required one-sided integral diverges, the original integral diverges.
Never substitute ∞ into an antiderivative or integrate straight across an interior singularity. Opposite infinite contributions cannot be canceled: if any required one-sided limit fails to be finite, the improper integral diverges.
Before integrating, simplify and classify the integrand. Ask which recognizable structure is present and choose a method that converts the expression into a more familiar antiderivative; a good choice reduces complexity rather than merely changing notation.
| Visible structure | First method to test | Diagnostic question |
|---|---|---|
| direct sum, constant multiple, or known derivative pattern | basic antiderivative rule | can terms be integrated immediately? |
| composite function with an inner derivative factor | substitution | is f′(x) present with g(f(x))? |
| improper rational function or useful quadratic rewrite | algebra first | will division or completing the square reveal a known form? |
| product where one factor simplifies when differentiated | integration by parts | will the new integral be simpler? |
| proper rational function with distinct linear factors | linear partial fractions | can it split into logarithmic pieces? |
Examples of selection: ∫2xcos(x2)dx suggests u=x2 because du=2xdx is present. In contrast, ∫xexdx has no inner-function pattern, but differentiating x simplifies it, so integration by parts is appropriate.
For ∫x+1x2+1dx, the numerator degree signals long division before integrating. For ∫(x+1)(x+2)1dx, the proper rational form with distinct linear factors signals partial fractions. These decisions come from structure, not from the Topic number.
No single acronym replaces inspection, and the first attempted method is not automatically correct. Check that each transformation is equivalent on the original domain, retain definite-integral bounds correctly, and differentiate an indefinite result to verify it. If the new integral is harder, reconsider the choice.