← Hub FP2 · 13 · Diff. Equations 0%
Paper 2 · Further Pure 2 ✦

Differential Equations

An equation whose unknown is a whole function. Multiply by the right factor, guess the right shape, and the answer assembles itself out of two pieces — the complementary function and a particular integral.

Syllabus 2.6Nothing given by MF19General solution = CF + PI

What the examiner expects you to do

§1 First order linear: the integrating factor

◆ The method

Put the equation into the standard form — coefficient of dydx equal to 1:

dydx+P(x)y=Q(x).

Multiply through by the integrating factor I=ePdx. The left-hand side then collapses into a single derivative:

ddx(Iy)=IQIy=IQdx.

No constant of integration is needed when finding I itself; one arbitrary constant appears at the final integration, which is exactly the one the general solution should have.

Why ePdx is the magic multiplier

We want a function I making Idydx+IPy the derivative of Iy. By the product rule,

ddx(Iy)=Idydx+dIdxy,

so we need dIdx=IP. That is separable: dII=Pdx, giving lnI=Pdx and I=ePdx.

✎ Worked example — a polynomial right-hand side

Solve dydx2y=x2.

Already in standard form with P=2, so I=e2x and

ddx(ye2x)=x2e2x.

Integrate the right-hand side by parts twice:

x2e2xdx=12x2e2x12xe2x14e2x+A=14e2x(2x2+2x+1)+A.

Hence ye2x=14e2x(2x2+2x+1)+A, and multiplying by e2x:

y=14(2x2+2x+1)+Ae2x.

Check: y=x12+2Ae2x, so y2y=x12+2Ae2x+12(2x2+2x+1)2Ae2x=x2

✎ Worked example — divide through first

Solve xdydxy=x4.

The coefficient of dydx is x, not 1, so divide first: dydxyx=x3. Then P=1x, Pdx=lnx, and

I=elnx=1x.

So ddx(yx)=x2, giving yx=x33+A and

y=x43+Ax.

Check: xyy=x(43x3+A)13x4Ax=43x413x4=x4Forgetting to divide through is the most common error in this whole section — the integrating factor would have been wrong from the first line.

✎ Worked example — a hyperbolic integrating factor

Solve dydx+ycothx=coshx.

P=cothx=coshxsinhx, whose integral is ln|sinhx|, so I=sinhx. Then

ddx(ysinhx)=sinhxcoshxysinhx=12sinh2x+A,

because sinhxcoshx is exactly 12ddx(sinh2x). Dividing,

y=12sinhx+Acosechx.

Spotting that Pdx is a logarithm — so that I comes out as a clean function rather than an exponential — is what makes these tractable.

§2 Complementary function and particular integral

◇ The two words, and why the sum works

For a linear equation L(y)=f(x):

  • the complementary function (CF) is the general solution of the homogeneous equation L(y)=0 — it carries the arbitrary constants;
  • a particular integral (PI) is any one solution of the full equation L(y)=f(x) — no arbitrary constants.
 general solution=CF+PI 

Why: if yp is a particular integral and y is any solution, then L(yyp)=L(y)L(yp)=ff=0, so yyp is a complementary function. Every solution is therefore a CF plus that one PI — and none is missed.

⚠ Count your arbitrary constants

A first order equation has one arbitrary constant; a second order equation has two. If your final answer has the wrong number, something has been dropped — most often the CF itself.

§3 Finding the complementary function

◆ The auxiliary equation

For ad2ydx2+bdydx+cy=0, try y=emx. Substituting gives (am2+bm+c)emx=0, and since emx0 we need the auxiliary equation

am2+bm+c=0.
RootsComplementary functionBehaviour
real and distinct, m1m2y=Aem1x+Bem2xpure growth / decay
real and repeated, my=(A+Bx)emxthe extra x is essential — two constants are needed
complex, p±qiy=epx(Acosqx+Bsinqx)oscillation inside an envelope epx

For a first order equation adydx+by=0 the auxiliary equation is just am+b=0, giving y=Aebx/a.

✎ Worked examples — one of each
  • Distinct real: y3y+2y=0 has m23m+2=(m1)(m2)=0, so y=Aex+Be2x.
  • Repeated: y4y+4y=0 has (m2)2=0, so y=(A+Bx)e2x. Check the Bx term: if y=xe2x then y=(1+2x)e2x and y=(4+4x)e2x, so y4y+4y=(4+4x48x+4x)e2x=0
  • Complex: y+2y+5y=0 has m=2±4202=1±2i, so y=ex(Acos2x+Bsin2x) — an oscillation of period π damped by ex.
★ A* tip — read the physics off the roots

The real part of the roots controls growth or decay, and the imaginary part controls oscillation. Negative real part means the solution dies away, whatever the constants — which is exactly the answer to "describe the long-term behaviour" questions.

§4 Finding a particular integral

◆ Trial forms — match the shape of the right-hand side
Right-hand side f(x)Try
a polynomial of degree na general polynomial of degree n
aebxCebx
acospx+bsinpx (or just one of them)Ccospx+Dsinpxalways both

Substitute the trial form into the equation and compare coefficients to pin down the constants.

⚠ Resonance: when the trial form is already in the CF

If your trial PI is a solution of the homogeneous equation, substituting it gives 0=f(x) — nonsense. The remedy is to multiply the trial form by x (and by x2 if it is still in the CF, which happens with a repeated root).

✎ Worked example — the ordinary case

Solve y3y+2y=e3x.

CF: m23m+2=0 gives m=1,2, so CF =Aex+Be2x.

PI: e3x is not in the CF, so try y=Ce3x. Then y=3Ce3x and y=9Ce3x, so

9C9C+2C=1  C=12.

General solution: y=Aex+Be2x+12e3x.

✎ Worked example — the resonant case

Solve y3y+2y=e2x. Same CF, but now e2x is in it, so try y=Cxe2x:

y=C(1+2x)e2x,y=C(4+4x)e2x, y3y+2y=C[(4+4x)3(1+2x)+2x]e2x=Ce2x.

Every x term cancels — as it must, since xe2x was chosen precisely to survive. So C=1 and

y=Aex+Be2x+xe2x.
✎ Worked example — "find k given the form"

Evaluate k given that kxcos2x is a particular integral of d2ydx2+4y=sin2x.

(The form has an x in it because the CF is Acos2x+Bsin2x — this is the resonant case.) Differentiating twice by the product rule:

y=kxcos2x,y=kcos2x2kxsin2x,y=4ksin2x4kxcos2x.

Then the x terms cancel:

y+4y=4ksin2x4kxcos2x+4kxcos2x=4ksin2x.

Comparing with sin2x gives 4k=1, so k=14, and a full general solution would be y=Acos2x+Bsin2x14xcos2x.

§5 Using a given substitution

x=et turns variable coefficients into constant ones

For equations of the form ax2d2ydx2+bxdydx+cy=f(x), put x=et. Then

xdydx=dydt,x2d2ydx2=d2ydt2dydt.

Substituting leaves a constant-coefficient equation in t, which you solve as in §3 and §4 — then convert back with t=lnx.

Deriving those two results

With x=et we have dxdt=et=x, so by the chain rule dydt=dydxx, which is the first result. Differentiating that with respect to t:

d2ydt2=ddt(xdydx)=xddx(xdydx)=x(dydx+xd2ydx2)=dydt+x2d2ydx2,

which rearranges to the second.

✎ Worked example — x=et

Use x=et to solve x2d2ydx2+xdydx4y=0 for x>0.

(d2ydt2dydt)+dydt4y=0  d2ydt24y=0.

The first-derivative terms cancel. The auxiliary equation m24=0 gives m=±2, so y=Ae2t+Be2t. Converting back with et=x:

y=Ax2+Bx2.

Check y=x2: x2(2)+x(2x)4x2=2x2+2x24x2=0 ✓  Check y=x2: x2(6x4)+x(2x3)4x2=(624)x2=0

✎ Worked example — y=ux makes it separable

Use y=ux to solve dydx=x+yxy.

With y=ux, the product rule gives dydx=u+xdudx, while the right-hand side becomes x+uxxux=1+u1uevery x cancels, which is the whole point of the substitution. So

xdudx=1+u1uu=(1+u)u(1u)1u=1+u21u.

Now separate and split the left-hand integrand into two standard pieces:

1u1+u2du=dxx  tan1u12ln(1+u2)=ln|x|+A.

Finally put u=yx. Leaving the answer in this implicit form is perfectly acceptable — do not waste time trying to make y the subject.

§6 Initial conditions and interpreting the model

◆ Find the general solution first — always

Apply the conditions only after you have the complete general solution including the PI. Applying them to the CF alone is a guaranteed loss of every remaining mark, because the constants come out wrong.

For a second order equation you will be given two conditions — typically y and dydx at one value of x — giving two simultaneous equations in A and B.

✎ Worked example — with interpretation

Solve y+2y+5y=0 given y=1 and y=1 when x=0, and describe the behaviour for large x.

From §3 the general solution is y=ex(Acos2x+Bsin2x). At x=0: y=A=1. Differentiating,

y=ex(Acos2x+Bsin2x)+ex(2Asin2x+2Bcos2x),

so at x=0, y=A+2B=1. With A=1 this gives B=1, and

y=ex(cos2x+sin2x).

Interpretation. The bracket oscillates between ±2 forever, but the factor ex0. So the solution is a damped oscillation: it crosses zero repeatedly, with the size of each swing shrinking, and y0 as x. In a modelling context that is a system returning to equilibrium while overshooting on the way.

§7 Interactive: the solution family, and a live check

Move the arbitrary constants and watch the whole family of solutions sweep out — every one of them satisfies the equation. The panel substitutes the stated solution back into the differential equation numerically and reports the largest residual over the plotted range: if that is essentially zero, the solution really is a solution.

your solution other members of the family the particular integral alone

every curve solves it ✎

§8 Examiner traps & A* checklist

⚠ The nine most common mark-losers
  • Not dividing through so the coefficient of dydx is 1 before finding the integrating factor.
  • Giving only the PI, or only the CF, as "the solution". The general solution is both.
  • Trying Cebx when ebx is already in the CF — multiply by x.
  • Using only Ccospx when the right-hand side is asinpx. You need both cos and sin in the trial form.
  • Applying the initial conditions to the CF before adding the PI.
  • Writing the repeated-root CF as Aemx+Bemx — that collapses to one constant. It must be (A+Bx)emx.
  • Mishandling complex roots: the answer is epx(Acosqx+Bsinqx), with p the real part and q the modulus of the imaginary part.
  • Forgetting the constant at the final integration of Iy=IQdx.
  • Not answering the "interpret" part. If the question mentions a model, say in words what happens for large x.
★ Before you turn the page
  • Does my answer have exactly the right number of arbitrary constants?
  • Have I substituted my PI back into the equation to confirm it works?
  • Does the solution satisfy the initial conditions when I put the numbers in?

§9 Video explainers (curated)

§10 Check yourself

Score 0 / 5
Q1 · integrating factor

The integrating factor for dydx+2xy=x is:

Solution

I=ePdx=e(2/x)dx=e2lnx=elnx2=x2. The trap is stopping at Pdx=2lnx — you must still exponentiate. Check: x2y+2xy=ddx(x2y)

Q2 · repeated root

The complementary function of d2ydx26dydx+9y=0 is:

Solution

m26m+9=(m3)2=0 gives the repeated root m=3, so the CF is (A+Bx)e3x. The third option is really Ae3x+Be3x=(A+B)e3x — one constant in disguise, so it cannot be a general solution of a second order equation.

Q3 · resonance

For y4y+3y=e3x, a suitable form for the particular integral is:

Solution

The auxiliary equation is (m1)(m3)=0, so the CF is Aex+Be3x and e3x is already in it. Trying Ce3x would give 0=e3x; multiplying by x once is enough, because the root 3 is not repeated. (You would need Cx2e3x only for a repeated root.)

Q4 · finding a given coefficient

Given that kxcos2x is a particular integral of d2ydx2+4y=sin2x, the value of k is:

Solution

With y=kxcos2x: y=kcos2x2kxsin2x and y=4ksin2x4kxcos2x. The xcos2x terms cancel in y+4y, leaving 4ksin2x. Comparing with sin2x gives 4k=1, so k=14. The sign is the whole question here.

Q5 · full method — try it on paper first

Solve d2ydx23dydx+2y=e3x, given that y=1 and dydx=0 when x=0. Describe the behaviour of y for large x.

Solution

Complementary function. The auxiliary equation is m23m+2=0, i.e. (m1)(m2)=0, so m=1 or m=2 and

CF=Aex+Be2x.

Particular integral. The right-hand side is e3x, and 3 is not a root of the auxiliary equation, so try y=Ce3x. Then y=3Ce3x, y=9Ce3x, and

9Ce3x9Ce3x+2Ce3x=e3x  2C=1  C=12.

General solution. y=Aex+Be2x+12e3x.

Apply the conditions — to the full general solution, not just the CF. At x=0:

y=A+B+12=1  A+B=12.

Differentiating, y=Aex+2Be2x+32e3x, so at x=0:

A+2B+32=0  A+2B=32.

Subtracting the first from the second gives B=2, and then A=12(2)=52. Hence

y=52ex2e2x+12e3x.

Check: at x=0, y=2.52+0.5=1 ✓ and y=2.54+1.5=0

Behaviour for large x. All three terms grow, but e3x grows fastest, so it dominates: y12e3x+. The solution grows without bound, at the rate set by the particular integral rather than by the complementary function.

FP2 · Topic 13 · Differential Equations