JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

The Integrating Factor, Step by Step

The previous guide explained why the integrating factor works; here we slow right down and actually drive it — four unhurried steps, two fully worked examples, and the small slips that quietly wreck the answer.

From idea to procedure

In the last guide you saw why the integrating factor exists: multiplying y' + p(x) y = q(x) by a cleverly chosen mu(x) collapses the whole left side into the single derivative (mu y)', after which one integration finishes the job. That was the why. This guide is the how — the same idea slowed to a walking pace so your hand learns the moves, not just your head. Knowing a method and executing it cleanly under pressure are two different skills, and the second only comes from watching every step land.

Here is the whole recipe as four steps you will repeat for the rest of this rung. They never change; only p and q change. Read them once now, then watch them run on real equations below.

  1. Standard form. Rewrite the equation as y' + p(x) y = q(x), with the coefficient of y' exactly 1. If you start from a(x) y' + b(x) y = c(x), divide everything by a(x) first — this single step is the most common place to go wrong.
  2. Build mu. Read off p(x), integrate it, and form the factor mu(x) = e^(integral of p(x) dx). You need only one antiderivative — drop the +C here, since any single mu does the job.
  3. Collapse and integrate. Multiply through by mu, recognize the left side as (mu y)', so the equation reads (mu y)' = mu(x) q(x); then integrate both sides — and now you DO keep the +C.
  4. Solve for y. Divide by mu to isolate y, giving the general solution; if an initial condition is given, plug it in to pin down C.

Worked example one: a variable coefficient

Solve x y' + 3 y = x^2 (taking x > 0). It does not start in standard form — the coefficient of y' is x, not 1 — so step one is to divide everything by x. That gives y' + (3/x) y = x, and now we can read off p(x) = 3/x and q(x) = x. Notice that p is a genuine function of x, a variable coefficient, not a constant; the recipe will not care in the slightest.

Step two: build mu. Integrate p, so integral of (3/x) dx = 3 ln x, and mu(x) = e^(3 ln x) = e^(ln x^3) = x^3. The exponential and the logarithm cancel, which is exactly why integrating factors so often come out as clean powers of x. Step three: multiply the standard-form equation through by x^3. The right side becomes x^3 times x = x^4, and the left side is engineered to be the single derivative (x^3 y)'. So the equation now reads (x^3 y)' = x^4 — the whole point of the factor, achieved.

Step four finishes it. Integrate both sides of (x^3 y)' = x^4 to get x^3 y = x^5/5 + C — keeping the constant C this time, because this is the integration that carries the freedom of the general solution. Divide by x^3 to isolate y: the answer is y = x^2/5 + C/x^3. You can read its two pieces straight off — x^2/5 is the particular response shaped by the forcing q = x, while C/x^3 is the homogeneous part carrying the free constant.

Worked example two: pinning the constant with an initial value

Most real problems come as a linear initial value problem: an equation plus a starting value, asking for the one curve that passes through a given point. Solve y' + 2 y = 6 with y(0) = 1. The coefficient of y' is already 1, so we are in standard form with the constant coefficient p = 2 and q = 6. Here integral of 2 dx = 2x, so the integrating factor is mu(x) = e^(2x). With a constant p the factor is always a plain exponential e^(p x) — the simplest case there is.

Multiply through by e^(2x): the left side collapses to (e^(2x) y)', and the right side becomes 6 e^(2x). So (e^(2x) y)' = 6 e^(2x). Integrate both sides — the integral of 6 e^(2x) is 3 e^(2x) — to get e^(2x) y = 3 e^(2x) + C. Dividing by e^(2x) gives the general solution y = 3 + C e^(-2x). Only now do we use the initial condition: at x = 0 the equation says 1 = 3 + C, so C = -2, and the one curve we want is y = 3 - 2 e^(-2x).

Look at the shape of y = 3 - 2 e^(-2x). As x grows the term -2 e^(-2x) shrinks toward zero, leaving the steady value 3, while at the start it pulls the curve down to y(0) = 1. That decaying piece is the transient and the surviving 3 is the steady state — the transient-plus-steady-state reading that the final guide of this rung unfolds in full. You have, in effect, already computed it.

Where step by step goes wrong

The recipe is reliable, but four slips trip almost everyone at first, and all four are mechanical rather than conceptual. The good news: once you know where they hide, they are easy to avoid. Keep this short list nearby while the moves are still new.

SLIP 1  Forgetting standard form.
        Reading p straight off a y' + b y = c gives the WRONG p.
        Fix: divide by a(x) FIRST so y' stands alone.

SLIP 2  Multiplying only the left side by mu.
        mu must hit EVERY term, the q(x) on the right included.

SLIP 3  Dropping +C, or adding it too early.
        No +C when building mu (step 2).
        DO add +C at the integration in step 3.

SLIP 4  Using the initial condition before C exists.
        Pin C only at the very end, on the general solution.
The four mechanical slips, paired with their one-line fixes.

One slip deserves a second look because it changes the answer's domain, not just its value. In example one we wrote mu = x^3 only because we assumed x > 0; the very act of dividing by x in step one excludes x = 0 from the picture. So y = x^2/5 + C/x^3 lives on an interval of validity that stops at the origin, where C/x^3 blows up. Whenever a step divides by something that can vanish, the solution is honest only away from those points — the integrating factor never frees you from watching the domain.

Why the linear walk is so much safer than separation

Recall the quiet trap from guides one and two of this rung: separating variables can silently throw away a constant equilibrium, because you divide by an expression in y and the case where that expression is zero slips through the crack. The integrating-factor walk has no such pitfall. The only division anywhere in the four steps is by mu(x), and mu depends on x alone, never on y — so it can never erase a y-solution. Nothing gets lost; the general solution you compute genuinely captures every solution.

That tidiness is a structural gift of linearity, and it is worth being honest about its reach. 'Always solvable' means the four steps always apply when p and q are continuous — it does not promise a tidy formula. The integral in step three may have no elementary antiderivative (think q = e^(-x^2)); then the recipe still wins, just handing you the answer as an integral to evaluate numerically. And the moment y enters nonlinearly — y' = y^2, say — this whole machine stops applying, and the gentler troubles of nonlinear equations, including finite-time blow-up, return.

You now own the integrating factor as a procedure, not just a story: standard form, build mu, collapse and integrate, solve for y. Run it three or four times by hand and the four steps fuse into one fluid motion. The last guide of this rung returns to the answers you have been producing — y = 3 - 2 e^(-2x) and its kin — to read them as a transient settling onto a steady state, the picture that makes first-order linear models feel alive.