Differential Equations

Complete Study Notes

Higher Order ODEs · Variable Coefficients · Variation of Parameters

Cauchy-Euler · Power Series · Legendre · Bessel Functions

# Topic

1 Higher Order ODEs with Constant Coefficients

2 Variable Coefficient ODEs & Reduction of Order

3 Variation of Parameters

4 Cauchy-Euler Equation

5 Ordinary vs Singular Points

6 Power Series Method

7 Legendre Polynomials

8 Bessel Functions of the First Kind

9 Hard Practice Problems

1. Higher Order ODEs with Constant Coefficients
A linear ODE with constant coefficients has the form:

a_n y^(n) + a_(n-1) y^(n-1) + ... + a_1 y' + a_0 y = f(x)

where a_0, a_1, ..., a_n are constants. The general solution is:

General Solution = CF + PI

■ CF = Complementary Function (solves equation with RHS = 0). PI = Particular Integral (accounts for
f(x)).

The Characteristic (Auxiliary) Equation

Assume y = e^(mx). Substituting into the homogeneous ODE gives the characteristic equation:

a_n m^n + a_(n-1) m^(n-1) + ... + a_1 m + a_0 = 0

Three Cases of Roots

Case Roots CF Contribution

Distinct Real m1, m2, ... C1·e^(m1·x) + C2·e^(m2·x) + ...

Repeated (k times) m, m, ... (k) (C1 + C2·x + ... + Ck·x^(k-1))·e^(mx)

Complex α ± iβ e^(αx) [C1·cos(βx) + C2·sin(βx)]

Complex Repeated (k) α ± iβ (×k) e^(αx)[(C1+C2x+...)cos(βx) + (D1+D2x+...)sin(βx)]

■ For repeated roots in Cauchy-Euler, multiply by ln(x) instead of x.

Method of Undetermined Coefficients (for PI)

f(x) Assume y_p

e^(ax) A·e^(ax)

sin(bx) or cos(bx) A·sin(bx) + B·cos(bx)

x^n A_n·x^n + ... + A_0

x^n · e^(ax) (A_n·x^n + ...)·e^(ax)

Examples

Example 1 — Distinct Real Roots

Solve: y'' - 5y' + 6y = 0

m^2 - 5m + 6 = 0 → (m-2)(m-3) = 0 → m = 2, 3

y = C1·e^(2x) + C2·e^(3x)

Example 2 — Repeated Root

Solve: y'' - 4y' + 4y = 0
 m^2 - 4m + 4 = 0 → (m-2)^2 = 0 → m = 2, 2

y = (C1 + C2·x)·e^(2x)

Example 3 — Complex Roots

Solve: y'' + 4y = 0

m^2 + 4 = 0 → m = ±2i (α=0, β=2)

y = C1·cos(2x) + C2·sin(2x)
2. Variable Coefficient ODEs & Reduction of Order
When coefficients are functions of x, the characteristic equation trick fails. The standard form is:

y'' + p(x)·y' + q(x)·y = f(x)

Reduction of Order
If one solution y1(x) is already known, find y2 by assuming:

y2 = v(x) · y1(x)

Steps:

• Assume y = v(x)·y1(x) and differentiate

• Substitute into the ODE — the v terms cancel (since y1 is a solution)
• Let w = v' — reduces to a 1st order ODE in w
• Solve for w, integrate to get v, then y2 = v·y1

■ v(x) is never assumed to be any specific form — it is an unknown function that the substitution
resolves naturally.

The Wronskian
Test whether y1 and y2 are linearly independent:

W(y1, y2) = |y1 y2 | = y1·y2' - y2·y1'

|y1' y2'|

If W ≠ 0, the solutions are linearly independent.

Example
Given y1 = x is a solution of x²y'' - xy' + y = 0, find y2.

Let y = vx, then y' = v'x + v, y'' = v''x + 2v'

Substituting and simplifying: x³v'' + x²v' = 0

Let w = v': xw' + w = 0 → w = 1/x → v = ln(x)

y2 = x·ln(x)

General Solution: y = C1·x + C2·x·ln(x)

3. Variation of Parameters
A universal method to find the PI for any f(x). It uses y1 and y2 from the CF as building blocks to construct
the PI.

PI = u1(x)·y1(x) + u2(x)·y2(x)

■ Always write the ODE in standard form first — the coefficient of y'' MUST be 1 before applying the
formulas.

The Formulas
Step 1: Compute the Wronskian:

W = y1·y2' - y2·y1'

Step 2: Find u1 and u2:

u1 = ∫ (-y2·f(x)) / W dx

u2 = ∫ (y1·f(x)) / W dx

Step 3: PI = u1·y1 + u2·y2

Step-by-Step Summary

Step Action

1 Write ODE in standard form (y coefficient = 1)

2 Solve homogeneous → get y1, y2 (the CF)

3 Compute W = y1·y2' - y2·y1'

4 u1' = -y2·f/W, u2' = y1·f/W

5 Integrate to get u1 and u2

6 PI = u1·y1 + u2·y2

7 GS = CF + PI

Example — y'' + y = sec(x)

This cannot be done by undetermined coefficients since sec(x) is not in the standard table.

CF: m^2 + 1 = 0 → m = ±i → y1 = cos(x), y2 = sin(x)

W = cos²(x) + sin²(x) = 1

u1' = -sin(x)·sec(x) = -tan(x) → u1 = ln|cos(x)|

u2' = cos(x)·sec(x) = 1 → u2 = x

PI = cos(x)·ln|cos(x)| + x·sin(x)

GS = C1·cos(x) + C2·sin(x) + cos(x)·ln|cos(x)| + x·sin(x)

4. Cauchy-Euler Equation
The Cauchy-Euler equation has a specific recognizable form:

ax²y'' + bxy' + cy = f(x)

■ Recognition trick: the power of x matches the order of the derivative. x² with y'', x¹ with y', x■ with y.

The Substitution
Assume y = x^m. Then:

y' = m·x^(m-1), y'' = m(m-1)·x^(m-2)

Substituting gives the auxiliary equation:

am(m-1) + bm + c = 0

Three Cases

Case Roots General Solution

Distinct Real m1 ≠ m2 C1·x^m1 + C2·x^m2

Repeated m1 = m2 = m (C1 + C2·ln(x))·x^m

Complex α ± iβ x^α [C1·cos(β·ln x) + C2·sin(β·ln x)]

Comparison: Constant Coefficient vs Cauchy-Euler

Feature Constant Coefficient Cauchy-Euler

Trial solution y = e^(mx) y = x^m

Repeated root multiplier x ln(x)

Complex roots argument β·x β·ln(x)

Auxiliary equation am² + bm + c = 0 am(m-1) + bm + c = 0

Examples

Example 1 — Distinct Real Roots

Solve: x²y'' - 2xy' - 4y = 0

m(m-1) - 2m - 4 = 0 → m² - 3m - 4 = 0 → m = 4, -1

y = C1·x^4 + C2·x^(-1)

Example 2 — Repeated Root

Solve: x²y'' - 3xy' + 4y = 0

m(m-1) - 3m + 4 = 0 → m² - 4m + 4 = 0 → m = 2, 2

y = (C1 + C2·ln x)·x²

Example 3 — Complex Roots

Solve: x²y'' + xy' + 4y = 0

m(m-1) + m + 4 = 0 → m² + 4 = 0 → m = ±2i

y = C1·cos(2·ln x) + C2·sin(2·ln x)
5. Ordinary vs Singular Points
Before solving a variable coefficient ODE with series methods, classify the point x = x0.

Write in standard form: y'' + p(x)·y' + q(x)·y = 0, then check p(x) and q(x) at x = x0.

Classification Rules

Point Type Condition Method to Use

Ordinary Point p(x0) and q(x0) are both finite Power Series: y = Σ an·x^n

Regular Singular (x-x0)·p(x) and (x-x0)²·q(x) → finite limits Frobenius: y = Σ an·x^(n+r)

Irregular Singular Limits above blow up Out of scope

Decision Flow
Step 1: Plug x0 into p(x) and q(x) → Both finite? → ORDINARY
POINT → Power Series Step 2 (if not finite): Compute limits
lim(x→x0) (x - x0)·p(x) and lim(x→x0) (x - x0)²·q(x) → Both
finite? → REGULAR SINGULAR → Frobenius → Either infinite? →
IRREGULAR SINGULAR

Examples

Legendre's Equation: (1-x²)y'' - 2xy' + n(n+1)y = 0

Standard form: p(x) = -2x/(1-x²), q(x) = n(n+1)/(1-x²)

Point p(x) at point Limits finite? Type

x=0 p(0) = 0 ✓ finite — Ordinary Point

x=1 p(1) → ∞ Yes: lim = 1, 0 Regular Singular

x = -1 p(-1) → ∞ Yes: finite Regular Singular

■ Since x=0 is an ordinary point for Legendre's equation, plain power series works there!
6. Power Series Method
Used when x0 is an ordinary point. Assume the solution is an infinite polynomial:

y = Σ an·x^n = a0 + a1·x + a2·x² + a3·x³ + ...

The goal is to find the coefficients a0, a1, a2, ...

The Method — Step by Step

• Step 1: Assume y = Σ an·x^n and differentiate term by term
• Step 2: y' = Σ n·an·x^(n-1), y'' = Σ n(n-1)·an·x^(n-2)
• Step 3: Substitute into the ODE
• Step 4: Shift indices so ALL sums have x^n
• Step 5: Pull out early terms manually until all sums start at same n
• Step 6: Combine under one sigma — set coefficient of x^n to zero
• Step 7: This gives the recurrence relation
• Step 8: a0 and a1 are free constants (= C1 and C2)

Index Shifting Rule

If you see x^(n-k), replace n with n+k everywhere (including the lower limit):

You see Replace n with Result

x^(n-2) n+2 x^n (lower limit: n=2 → n=0)

x^(n-1) n+1 x^n (lower limit: n=1 → n=0)

x^(n+2) n-2 x^n (lower limit: n=0 → n=2)

Full Example: y'' - xy' - y = 0

Substitute and shift index in y'' (n → n+2):

Σ(n+2)(n+1)·a(n+2)·x^n - Σ n·an·x^n - Σ an·x^n = 0

At n=0: 2a2 - a0 = 0 → a2 = a0/2

For n≥1: (n+2)(n+1)·a(n+2) - n·an - an = 0

Recurrence Relation: a(n+2) = an / (n+2)

Even terms (from a0): a2 = a0/2, a4 = a0/8, a6 = a0/48, ...

Odd terms (from a1): a3 = a1/3, a5 = a1/15, a7 = a1/105, ...

y = a0(1 + x²/2 + x■/8 + ...) + a1(x + x³/3 + x■/15 + ...)

■ Even powers always come from a0, odd powers always come from a1.
7. Legendre Polynomials
Legendre polynomials arise from solving Legendre's equation:

(1 - x²)y'' - 2xy' + n(n+1)y = 0

where n is a non-negative integer. x=0 is an ordinary point, so we use power series. The recurrence
relation is:

a(k+2) = [k(k+1) - n(n+1)] / [(k+1)(k+2)] · ak

■ When n is an integer, the series terminates — becoming a finite polynomial. After normalizing so that
Pn(1) = 1, these are the Legendre Polynomials.

First Six Legendre Polynomials

n Pn(x)

0 P0(x) = 1

1 P1(x) = x

2 P2(x) = (1/2)(3x² - 1)

3 P3(x) = (1/2)(5x³ - 3x)

4 P4(x) = (1/8)(35x■ - 30x² + 3)

5 P5(x) = (1/8)(63x■ - 70x³ + 15x)

Key Properties

Rodrigues' Formula
Pn(x) = (1 / 2^n · n!) · d^n/dx^n (x² - 1)^n

Orthogonality (Most Important)

∫[-1 to 1] Pm(x)·Pn(x) dx = 0 if m ≠ n

∫[-1 to 1] [Pn(x)]² dx = 2/(2n+1) if m = n

Recurrence Relation
(n+1)·P(n+1)(x) = (2n+1)·x·Pn(x) - n·P(n-1)(x)

Special Values & Symmetry

Pn(1) = 1, Pn(-1) = (-1)^n

Pn(-x) = (-1)^n · Pn(x) [even n → even function, odd n → odd

function]

Properties Summary Table

Property Formula

Legendre ODE (1-x²)y'' - 2xy' + n(n+1)y = 0

Normalization Pn(1) = 1

Rodrigues Pn = (1/2^n·n!) · d^n/dx^n (x²-1)^n

Orthogonality ∫Pm·Pn dx = 2/(2n+1) · δmn

Recurrence (n+1)P(n+1) = (2n+1)x·Pn - n·P(n-1)

Parity Pn(-x) = (-1)^n · Pn(x)

8. Bessel Functions of the First Kind
Bessel functions arise from Bessel's equation:

x²y'' + xy' + (x² - n²)y = 0

where n is the order. This appears in problems with circular/cylindrical symmetry.

x=0 is a REGULAR SINGULAR POINT → use Frobenius Method.

Frobenius Method — Full Derivation

Step 1: Assume
y = Σ ak · x^(k+r) (k = 0, 1, 2, ...)

y' = Σ (k+r)·ak · x^(k+r-1)

y'' = Σ (k+r)(k+r-1)·ak · x^(k+r-2)

Step 2: Substitute and simplify

Multiplying through by x² and x respectively, all terms get power x^(k+r) except the x²·y term which gives
x^(k+r+2). Shift that sum (k → k-2):

Σ [(k+r)² - n²]·ak·x^(k+r) + Σ a(k-2)·x^(k+r) = 0

Step 3: Indicial Equation (k=0 term)

At k=0, only the first sum contributes:

(r² - n²)·a0 = 0 → r² - n² = 0 → r = ±n

■ This is where the indicial equation comes from naturally — the k=0 term! We take r = +n for Jn(x).

Step 4: k=1 term

[(1+n)² - n²]·a1 = 0 → (1+2n)·a1 = 0 → a1 = 0

Since a1 = 0, ALL odd coefficients vanish.

Step 5: Recurrence (k ≥ 2) with r = n

ak = -a(k-2) / [k(k + 2n)]

Let k = 2m. Generating coefficients from a0:

a(2m) = (-1)^m · a0 / [4^m · m! · (1+n)(2+n)···(m+n)]

Step 6: Choose a0 = 1/(2^n · n!) → Final Formula

Jn(x) = Σ [(-1)^m / (m! · (m+n)!)] · (x/2)^(2m+n) (m=0,1,2,...)

First Two Bessel Functions

J0(x) = 1 - x²/4 + x■/64 - x■/2304 + ...

J1(x) = x/2 - x³/16 + x■/384 - ...

Key Properties

Recurrence Relation
J(n-1)(x) + J(n+1)(x) = (2n/x) · Jn(x)

Derivative Relations
d/dx [x^n · Jn(x)] = x^n · J(n-1)(x)

d/dx [x^-n · Jn(x)] = -x^-n · J(n+1)(x)

Special case: J0'(x) = -J1(x)

Symmetry
J(-n)(x) = (-1)^n · Jn(x) for integer n

Orthogonality
∫[0 to 1] x · Jn(λi·x) · Jn(λj·x) dx = 0 if i ≠ j

where λi, λj are different zeros of Jn(x).

Properties Summary

Property Formula

Bessel ODE x²y'' + xy' + (x²-n²)y = 0

Series Jn(x) = Σ (-1)^m / (m!(m+n)!) · (x/2)^(2m+n)

Recurrence J(n-1) + J(n+1) = (2n/x)·Jn

Derivative 1 d/dx [x^n·Jn] = x^n·J(n-1)

Derivative 2 d/dx [x^-n·Jn] = -x^-n·J(n+1)

Special J0'(x) = -J1(x)

Symmetry J(-n)(x) = (-1)^n·Jn(x)

9. Hard Practice Problems

Section A — Higher Order ODEs with Constant Coefficients

1 Solve: y^(4) - 8y'' + 16y = 0

2 Solve: y''' - 3y'' + 3y' - y = e^x

3 Solve: y^(4) + 4y = 0 (Hint: factor m■ + 4 carefully)

4 Solve: y''' - 6y'' + 11y' - 6y = e^(2x)·sin(x)

5 Solve: y^(4) - 2y''' + 5y'' - 8y' + 4y = 0 (Hint: m=1 repeated, m=±2i complex)

Section B — Variation of Parameters

1 Solve: y'' + 4y = sec(2x)

2 Solve: y'' - 2y' + y = e^x / x²

3 Solve: y'' + 2y' + y = e^(-x)·ln(x)

4 Solve: x²y'' - 2xy' + 2y = x■·sin(x) [given y1=x, y2=x²]

5 Solve: y'' - 3y' + 2y = e^x / (1 + e^(-x))

Section C — Cauchy-Euler Equation

1 Solve: x³y''' + 2x²y'' - xy' + y = 0

2 Solve: x²y'' - 3xy' + 4y = x²·ln(x)

3 Solve: x²y'' + xy' + 4y = cos(ln x)

4 Solve: x²y'' - xy' + 5y = x²·sin(ln x)

5 Solve: x■y^(4) + 6x³y''' + 9x²y'' + 3xy' + y = 0

Section D — Power Series & Recurrence

1 Find power series solution about x=0: y'' + xy' + x²y = 0

2 Find recurrence relation and first 4 nonzero terms: (1+x²)y'' - 2xy' + 2y = 0

3 Solve: y'' - 2xy' + 4y = 0 (notice something special about the even series!)

4 Find two independent series solutions: x·y'' + y' + xy = 0

5 Find power series solution and identify the polynomial solution: (1-x²)y'' - 2xy' + 6y = 0

Section E — Legendre Polynomials

1 Using Rodrigues' formula, derive P4(x) from scratch

2 Prove the orthogonality: ∫[-1 to 1] P2(x)·P4(x) dx = 0

3 Express f(x) = x■ in terms of P0, P2, P4

 4 Using recurrence relation only, find P5(x) from P3 and P4

5 Evaluate: ∫[-1 to 1] x·Pn(x)·P(n+1)(x) dx (Hint: use recurrence to replace x·Pn)

Section F — Bessel Functions

1 Prove using the derivative formula that J0'(x) = -J1(x)

2 Using recurrence, express J3(x) purely in terms of J0 and J1

3 Prove: J(1/2)(x) = sqrt(2/πx)·sin(x) (Hint: use series with n=1/2)

4 Show that y = sqrt(x)·J1(2·sqrt(x)) satisfies: x·y'' + y' + y = 0

5 Evaluate: ∫ x²·J1(x) dx (Hint: use d/dx[x^n·Jn] = x^n·J(n-1))

■ Attempt order for best results: C → A → B → D → E → F. Cauchy-Euler first builds confidence fastest,
Bessel last is most demanding.

End of Notes — Good luck on your exam! ■