Singular Solutions and Envelopes

date2026-07-16document_iddoc_3bd2011ad7a7156260d720ba7169fb7cdescription古典的な特異解を一般解族の定数選択では得られず、典型的に包絡線として現れる解として整理し、除算で失う定数解や非 Lipschitz な非一意枝と区別する。prerequisites微分方程式とは何か / 初期値問題の存在・一意性と最大存在区間 / 一階微分方程式の分類と最初の判定type講義content_typelecturestatusactive

mathdifferential-equationssingular-solutionenvelopelecture

1Introduction

This lecture distinguishes a singular solution from a particular solution.

A particular solution is one specific function that satisfies the equation. A solution obtained by selecting the arbitrary constant in a general solution and a single solution constructed for a nonhomogeneous equation are both particular solutions.

By contrast, a singular solution may occur as a solution that cannot be obtained by selecting a constant from the general one-parameter family.

The classical situation treated in this lecture can be summarized as

[PARSE ERROR: Undefined("Command(\"boxed\")")] Aclassicalsingularsolutionisnotamemberofthegeneralfamilyandtypicallyoccursasitsenvelope [PARSE ERROR: Undefined("RBrace")]

Constant solutions discarded during division in separation of variables and nonunique branches that arise in a non-Lipschitz setting share the warning that a standard one-parameter family may not exhaust all solutions. They are not, however, treated here as synonyms for singular solutions.

2Diagnostic Procedure

When a singular solution is suspected, use the following procedure.

Determine whether the equation is nonlinear.
Determine whether a general one-parameter family of solution curves is available.
Examine whether the family has an envelope.
Substitute every candidate into the original differential equation.
Verify that a constant solution lost through division or a nonunique branch is not being conflated with a singular solution.

A geometric envelope is not automatically a solution of the differential equation. Direct substitution into the original equation is essential.

3Classification of Candidates

The principal mechanisms that may produce an additional candidate can be organized as follows.

Classification	Condition to examine	Required conclusion
Classical singular solution	$F (x, y, C) = 0, F_{C} (x, y, C) = 0$	Derive an envelope candidate for the general family and substitute it into the original equation
Degeneracy of an implicit equation	$F (x, y, p) = 0, F_{p} (x, y, p) = 0$	Examine a possible additional branch in the slope $p$ , without concluding from this condition alone that it is singular
Solution discarded by division	$h (y) = 0$	Check constant solutions separately by substitution; they need not be singular
Nonunique branch	A non-Lipschitz setting	Construct distinct solutions to verify nonuniqueness; this is conceptually distinct from a classical singular solution

Here $F_{C} = \partial F / \partial C$ and $F_{p} = \partial F / \partial p$ . The condition $F_{p} = 0$ is not sufficient for existence of a singular solution. It marks a degeneracy in the determination of the slope and is therefore an indicator for investigating an envelope or an additional branch.

4Caution: Division by $h (y)$ May Discard Constant Solutions

For the separable equation

y^{'} = f (x) h (y),

one may write

\frac{1}{h (y)} d y = f (x) d x

where $h (y) \neq 0$ . This operation excludes the values for which $h (y) = 0$ .

If $h (a) = 0$ , the constant function $y (x) = a$ satisfies

y^{'} = 0, f (x) h (a) = 0,

and is therefore a solution. Every constant root of $h$ must be checked before division.

Such a constant solution need not be singular, but its loss is a typical source of incomplete solution sets for nonlinear equations.

data/lecture/math/differential-equations/separable-equations-and-autonomous-systems.lecture.n.md

5Contrast with Linear Equations

For a linear differential equation, all solutions are constructed from the superposition of homogeneous solutions and one particular solution. That structure is developed in later lectures on linear equations. In the present context, envelope-type singular solutions are investigated primarily for nonlinear first-order equations.

6Example 1: A Clairaut-Type Equation and Its Envelope

Consider the nonlinear differential equation

y = x y^{'} - (y^{'})^{2.}

Writing $p = y^{'}$ gives

y = x p - p^{2.}

First suppose that $p$ is a constant $C$ . Then

y = C x - C^{2},

which is a family of straight lines parametrized by $C$ .

To find its envelope, write the family implicitly as

F (x, y, C) = y - C x + C^{2} = 0.

An envelope candidate is obtained from the simultaneous equations

F (x, y, C) = 0, \frac{\partial F}{\partial C} (x, y, C) = 0.

Here

\frac{\partial F}{\partial C} = - x + 2 C,

so $C = x / 2$ . Substitution into $y = C x - C^{2}$ gives

y = \frac{x}{2} x - {(\frac{x}{2})}^{2} = \frac{x^{2}}{4} .

6.1Verification

If $y = x^{2} / 4$ , then

y^{'} = \frac{x}{2} .

The right-hand side of the original equation is

x y^{'} - (y^{'})^{2} = x \cdot \frac{x}{2} - {(\frac{x}{2})}^{2} = \frac{x^{2}}{4},

which equals $y$ . Thus $y = x^{2} / 4$ is indeed a solution.

No fixed constant $C$ makes the parabola $y = x^{2} / 4$ equal to one line $y = C x - C^{2}$ . It is therefore a singular solution occurring as the envelope of the general family.

7Degeneracy in the Determination of the Slope

In the preceding example, differentiation of $y = x p - p^{2}$ with respect to $x$ gives

y^{'} = p + (x - 2 p) p^{'} .

Since $p = y^{'}$ ,

(x - 2 p) p^{'} = 0.

This equation has the branches

p^{'} = 0

and

x - 2 p = 0.

The first gives $p = C$ and the line family $y = C x - C^{2}$ . The second gives $p = x / 2$ and the envelope $y = x^{2} / 4$ .

Thus the singular branch appears where determination of the slope degenerates. Selecting only the branch used to construct the general family would omit the additional solution branch.

8Related Example: Nonunique Branches in a Non-Lipschitz Setting

Consider the initial value problem

y^{'} = \sqrt{| y |}, y (0) = 0.

The function $y = 0$ is a solution. For every $a [PARSE ERROR: Undefined("Command(\"ge\")")] 0$ , the function

y(x)= \begin{cases} 0 & (x\le a),\\ \dfrac{(x-a)^2}{4} & (x\ge a) \end{cases}y(x)= \begin{cases} 0 & (x\le a),\\ \dfrac{(x-a)^2}{4} & (x\ge a) \end{cases}

is also a solution. For $x < a$ , $y^{'} = 0 = \sqrt{| y |}$ . For $x > a$ ,

y^{'} = \frac{x - a}{2}, \sqrt{y} = \sqrt{\frac{(x - a)^{2}}{4}} = \frac{x - a}{2} .

At $x = a$ , both one-sided derivatives are zero. The displayed function is therefore a $C^{1}$ solution on all of $R$ . Because $a [PARSE ERROR: Undefined("Command(\"ge\")")] 0$ places $x = 0$ in the waiting interval, it satisfies the initial condition $y (0) = 0$ .

The same initial condition thus produces multiple solutions. This is consistent with the failure of $\sqrt{| y |}$ to satisfy a Lipschitz condition near $y = 0$ , which prevents application of the uniqueness hypothesis in the Picard--Lindelof theorem. These waiting-time solutions illustrate nonunique branches; they do not define envelope-type singular solutions.

data/lecture/math/differential-equations/lipschitz-condition-and-continuity.lecture.n.md

9Cautions in Applying the Terminology

A singular solution is not another name for a particular solution.
An envelope is only a candidate and need not automatically solve the differential equation. It must be substituted into the original equation.
A constant solution discarded by division need not be singular, even if it is absent from the derived one-parameter family.
A nonunique branch arising in a non-Lipschitz setting is conceptually distinct from an envelope-type singular solution.

10Summary

In the classical setting treated here, a singular solution is a solution that cannot be obtained by selecting the constant in the general family and typically appears as its envelope. Every candidate must be verified in the original differential equation, and the terminology should remain distinct from that for constant solutions and nonunique branches.

11Previous Lecture

data/lecture/math/differential-equations/existence-uniqueness-and-lipschitz-condition.lecture.n.md

12Next Lecture

data/lecture/math/differential-equations/direction-fields-euler-method-error-and-stability.lecture.n.md