Initial value problems (IVP) vs boundary value problems (BVP)

WHAT are we even solving?

A differential equation alone has infinitely many solutions (a family with arbitrary constants). To pin down one member, we attach extra data called side conditions (also called auxiliary conditions).

The type of side condition decides whether the problem is an IVP or a BVP.

WHY does the location matter so much?

Intuition first (Feynman): Imagine a thrown ball.

IVP version: "It starts at height $0$ with upward speed $20$ m/s." You know position and velocity at $t=0$ . Physics then forces the whole future trajectory. This is causal / time-marching.
BVP version: "It is on the ground at $t=0$ AND back on the ground at $t=4$ ." You don't know the initial speed — you must find the trajectory that connects two endpoints. This is global / it's-a-puzzle.

HOW to tell them apart and solve — second-order template

Take the canonical second-order linear ODE: $y'' + p(x)\,y' + q(x)\,y = f(x).$ Its general solution has two arbitrary constants: $y(x) = c_1 y_1(x) + c_2 y_2(x) + y_p(x).$ We need two equations to fix $c_1,c_2$ .

IVP conditions (same point $x_0$ ): $y(x_0)=A,\qquad y'(x_0)=B.$

BVP conditions (two points $a\neq b$ ): $y(a)=\alpha,\qquad y(b)=\beta.$

Derivation of why a BVP can be singular

Plug the boundary data into the general solution. You get a linear system in $c_1,c_2$ :

\begin{pmatrix} y_1(a) & y_2(a)\\ y_1(b) & y_2(b)\end{pmatrix} \begin{pmatrix} c_1\\ c_2\end{pmatrix} = \begin{pmatrix} \alpha - y_p(a)\\ \beta - y_p(b)\end{pmatrix}.

Call the matrix $M$ .

If $\det M \neq 0$ → unique $(c_1,c_2)$ → unique solution.
If $\det M = 0$ → either no solution or infinitely many (depends on the right side).

For an IVP the analogous matrix is the Wronskian $W(x_0)=y_1 y_2' - y_2 y_1'$ evaluated at the same point, which is never zero for independent solutions — that's why IVPs are reliably uniquely solvable.

Figure — Initial value problems (IVP) vs boundary value problems (BVP)

Worked Examples

Worked example Example 1 — Same ODE, two different problems

ODE: $y'' + y = 0$ , general solution $y = c_1\cos x + c_2 \sin x$ .

(a) As an IVP: $y(0)=1,\ y'(0)=0$ .

$y(0)=c_1=1$ . Why this step? Plug $x=0$ into $y$ .
$y'(x) = -c_1\sin x + c_2\cos x$ , so $y'(0)=c_2=0$ . Why? Differentiate, then evaluate at the same point.
Unique: $y=\cos x$ . ✔

(b) As a BVP: $y(0)=0,\ y(\pi)=0$ .

$y(0)=c_1=0$ .
$y(\pi)=c_1\cos\pi + c_2\sin\pi = -c_1 + 0 = 0$ . This gives $c_1=0$ but says nothing about $c_2$ .
Infinitely many solutions: $y = c_2\sin x$ for any $c_2$ . 😮 Why this happened: $\det M = y_1(0)y_2(\pi)-y_2(0)y_1(\pi) = \cos0\cdot\sin\pi - \sin0\cdot\cos\pi = 0.$

Worked example Example 3 — A solvable BVP (steady heat in a rod)

ODE: $y''=0$ on $[0,L]$ with $y(0)=T_0,\ y(L)=T_1$ (temperature fixed at both ends).

General solution: $y=c_1 x + c_2$ . Why linear? $y''=0$ integrates twice to a straight line.
$y(0)=c_2=T_0$ .
$y(L)=c_1 L + T_0 = T_1 \Rightarrow c_1 = \dfrac{T_1-T_0}{L}$ .
Unique: $y(x)=T_0 + \dfrac{T_1-T_0}{L}x$ — a linear temperature profile. Why unique? $\det M = y_1(0)y_2(L)-y_2(0)y_1(L)=0\cdot1-1\cdot L=-L\neq0$ .

Common Mistakes (Steel-manned)

Flashcards

What single feature distinguishes an IVP from a BVP?

Whether all side conditions are at one point (IVP) or split across distinct points (BVP).

How many conditions does an

n

-th order ODE need to be well-posed?

n

conditions.

Why is an IVP (under mild conditions) always uniquely solvable?

The Wronskian at the initial point is nonzero, giving a unique constant determination; Picard–Lindelöf guarantees existence & uniqueness.

What matrix determines BVP solvability and what does its determinant being zero mean?

M=\begin{smallmatrix}y_1(a)&y_2(a)\\y_1(b)&y_2(b)\end{smallmatrix}

;

\det M=0

→ no solution OR infinitely many.

Give a BVP with infinitely many solutions.

y''+y=0

y(0)=0

y(\pi)=0

→

y=c\sin x

for any

c

Give a BVP with no solution.

y''+y=0

y(0)=0

y(\pi)=1

→ contradiction

0=1

Are

y(a)

and

y'(a)

together an IVP or BVP?

IVP (both at the same point

a

What classical eigenvalue theory arises from BVP singularity?

Sturm–Liouville theory;

\det M=0

gives eigenvalues.

For

y''=0

y(0)=T_0,y(L)=T_1

, what is the solution?

y=T_0+\frac{T_1-T_0}{L}x

, unique.

Conceptually, IVP = ? marching; BVP = ? puzzle.

IVP = forward (causal) marching; BVP = global two-point matching.

Recall Feynman: explain to a 12-year-old

Suppose you're drawing a path for a toy car. IVP: "Start HERE, pointing THIS way, going THIS fast." Now the car has to go one definite way — easy, just let it roll. BVP: "The car must start at the door AND end at the kitchen." You're not told how fast to push it; you have to figure out the push that lands it exactly in the kitchen. Sometimes there's one perfect push, sometimes no push works, sometimes lots of pushes work — that's why BVPs are trickier than IVPs even when both give "two facts."

Connections

Sturm-Liouville Theory — BVP eigenvalues come from $\det M=0$ .
Wronskian and Linear Independence — guarantees IVP uniqueness.
Picard-Lindelöf Existence and Uniqueness Theorem — the IVP safety net.
Separation of Variables (PDE) — produces BVPs in space + IVPs in time.
Heat Equation / Wave Equation — typically IVP in $t$ , BVP in $x$ .
Green's Functions — built to solve linear BVPs.

Concept Map

Hinglish (regional understanding)

Intuition Hinglish mein samjho

Dekho, ek differential equation ke paas infinite solutions hote hain — ek poora family with arbitrary constants. Us family me se ek specific solution choose karne ke liye humein extra conditions deni padti hain. Ab asli baat ye hai ki ye conditions kahan di gayi hain. Agar saari conditions ek hi point par hain (jaise $y(0)$ aur $y'(0)$ ), to wo IVP hai — Initial Value Problem. Agar conditions do alag-alag points par bati hui hain (jaise $y(a)$ aur $y(b)$ ), to wo BVP hai — Boundary Value Problem.

IVP ko samajhna easy: socho ek ball ko start position aur speed pata hai. Nature aage ka pura path khud decide kar leti hai — isliye IVP almost hamesha unique solution deta hai (Wronskian zero nahi hota, Picard-Lindelof guarantee deta hai). BVP me situation alag hai: tumhe sirf do ends pata hain, beech ka rasta khud match karna padta hai. Isliye BVP me kabhi ek hi solution, kabhi bilkul nahi, aur kabhi infinite solutions ho sakte hain.

Yaad rakhne ka trick: BVP me general solution daal ke ek matrix $M$ banti hai. Agar $\det M \neq 0$ → unique. Agar $\det M = 0$ → ya to no solution ya infinite. Yahi $\det M=0$ wali baat aage Sturm-Liouville eigenvalues banati hai — bahut important physics me (heat equation, wave equation, quantum). Example: $y''+y=0$ ke saath $y(0)=0,y(\pi)=0$ → infinite solutions $c\sin x$ , lekin $y(\pi)=1$ kar do → koi solution hi nahi! Same equation, sirf condition badli, aur duniya badal gayi.

Bas itna rakho dimaag me: I = Instant (ek jagah), B = Both ends (do jagah). Conditions ginna kaafi nahi hai — unki location dekho.

Initial value problems (IVP) vs boundary value problems (BVP)

WHAT are we even solving?

WHY does the location matter so much?

HOW to tell them apart and solve — second-order template

Derivation of why a BVP can be singular

Worked Examples

Common Mistakes (Steel-manned)

Flashcards

Connections

Concept Map

Hinglish (regional understanding)

Go deeper — visual, from zero

Test yourself — Partial Differential Equations

Connections