Department of Mathematics Indian Institute of Technology Bombay Powai, Mumbai - 76
August 2, 2021 Method of solving first order ODE Separable-Variable: Homogeneous form: ODE Exact form: ODE Seperable-variable Let I be an interval in R containing t0 and J be an interval in R containing y0 . Let g ∈ C (I ) and h ∈ C (J) with h(z) 6= 0 for all z ∈ J (i.e., either h(z) > 0 or h(z) < 0 for all z ∈ J). Consider the ODE: y 0 (t) = h(y (t))g (t), y (t0 ) = y0 . y 0 (t) Step 1. Rewrite the ODE: h(y (t)) = g (t), y (t0 ) = y0 . Ru 1 Step 2. Consider the anti-derivatives: H(u) = y0 h(s) ds and Rt G (t) = t0 g (s) ds. Note that H(y0 ) = 0 and G (t0 ) = 0. Step 3. Note that if a solution of the ODE exists, then putting u = y (t), R y (t) 1 H(y (t)) = y0 h(s) ds. It gives
d y 0 (t) d (H(y (t)) = H 0 (y (t))y 0 (t) = = G (t). dt h(y (t)) dt Step 4 . Thus, the ODE is transformed to an algebraic equation:
H(y (t)) = G (t) + c.
The constant is determined by the initial value: H(y0 ) = G (t0 ) + C
and thus c = 0. The solution of the ODE is in implicit form satisfying the algebraic equation: H(y (t)) = G (t). If the inverse of H exists and can be obtained by any known function, then the solution can be written in an explicit form: y (t) = H −1 (G (t)).
t2 Ex. Solve: y 0 (t) = 1−y 2 (t) , y (0) = 1.
Ans: Rewriting the ODE: (1 − y 2 (t))y 0 (t) = t 2 , y (0) = 1. Denoting
Z u Z t H(u) = (1 − s 2 ) ds, G (t) = s 2 ds, 1 0
we have y3 t3 H(y ) = y − − 2/3, G (t) = . 3 3 The solution of ODE satisfies H(y (t)) = G (t), i.e.,
y 3 (t) 2 t3 y (t) − − = . 3 3 3 Homogeneous form
Let f : R2 → R be called a homogeneous function of order (or degree)
n ∈ N ∪ {0} if f can be written as
f (λt, λz) = λn f (t, z), forall λ ∈ R, λ 6= 0.
Ex. f (x, y ) = x + y , homogeneous of order 1 and g (x, y ) = x − y ,
x+y homogeneous of order 1. Now gf (x, y ) = x−y is scale invariant. (In fact, homogeneous of order 0.) Let us consider the ODE:
M(t, y (t)) + N(t, y (t))y 0 (t) = 0, y (t0 ) = y0 ,
where N 6= 0 and M and N are homogeneous function of same order n.
Qn. Can the ODE be transformed into a variable-separable form? Reduced to variable-separable form Rewrite the ODE: for t 6= 0,
M(t, y (t)) M(1, y (t)/t)
y 0 (t) = − =− . N(t, y (t)) N(1, y (t)/t)
I Substituting y (t) = tv (t) and thus y 0 (t) = v (t) + tv 0 (t) for t 6= 0.
I Putting this in the ODE: M(1, v (t)) y0 tv 0 (t) = − v (t) + , v (t0 ) = . N(1, v (t)) t0 I The ODE is now variable-separable form in variables t and v (t).
Ex. (y 2 − t 2 )dy + 2tydt = 0, y (1) = 2.
Ex. tan−1 (y /t)(y 2 − t 2 )dt − 2ty log(y /t)dy = 0, the ODE is in homogeneous form. Ex. (y 3 − 3t 2 y )dt + 4t 2 y 2 dy = 0, the ODE is not in homogeneous form. Exact form Let us consider the ODE: M(t, y (t)) + N(t, y (t))y 0 (t) = 0, y (t0 ) = y0 , where N 6= 0, and M and N are defined on a domain D ⊂ R2 with (t0 , y0 ) ∈ D. Definition The ODE is said to be in Exact form if there exists a differentiable function φ : D → R such that
∂φ(t, y ) ∂φ(t, y ) = M(t, y ), = N(t, y ), ∀ (t, y ) ∈ D. ∂t ∂y
Remark. Note that if such φ exists then the ODE is reduced to
d φ(t, y (t)) = 0, dt and thus the family of solutions is given by φ(t, y ) = constant. The constant can be determined by the initial condition. Qn. How to check if the ODE is in Exact form. Refer. Apostol/ Rudin’s book on advanced calculus. Theorem Suppose D is a convex/ simply connected (open, connected, without any hole...) subset of R2 . Let M, N be continuously differentiable functions. Then the ODE M(t, y (t)) + N(t, y (t))y 0 (t) = 0, is Exact if and only if ∂M ∂N = on D. ∂y ∂t
Remark. In R2 , this results is a version of the result that ‘Curl’ free
vector always can be written as gradient of a potential function, provided the domain satisfies certain conditions. Proof of the theorem: If the ODE is in exact form, i.e., there exists a C 2 function φ : D → R, such that ∂φ(t, y ) ∂φ(t, y ) = M(t, y ), = N(t, y ), ∀ (t, y ) ∈ D. ∂t ∂y Then due to the assumptions on M and N and φ, taking partial derivatives of M with respect to y and partial derivative of N with respect to t and equating φty = φyt on D, we get that the condition holds:
∂M ∂N = on D. ∂y ∂t
Conversely, let ∂M ∂y = ∂t ∂N on D hold. Then using Green’s theorem in R , we can show the existence of a C 2 function φ (since M and N are 2
continuously differentiable) such that M = φt and N = φy on D.
Ex. Check if the ODE in Exact form. Solve the ODE.
1. 3t(ty − 2)dt + (t 3 + 2y )dy = 0. 2. sin t + t 2 e y y 0 − y 0 = −y cos t − 2te y . Integrating factor: If the ODE is not in Exact form. It may be possible to make the ODE in an exact form multiplying the ODE by an integrating factor. Let µ(t, y ) (6= 0) be an integrating factor such that after multiplying the ODE with it, the ODE becomes exact:
µ(t, y )M(t, y ) + µ(t, y )N(t, y )y 0 (t) = 0.
Since the ODE is exact,
∂ ∂ µ(t, y )M(t, y ) = µ(t, y )N(t, y ) , ∂y ∂t i.e., it yields h ∂M ∂N i h ∂µ ∂µ i µ − = N −M , on D. ∂y ∂t ∂t ∂y
Few cases, for example assuming further, µ = µ(t) (independent of y ),
the above equation is reduced to a linear ODE of µ and hence easily solvable: ∂µ h ∂M ∂N i N = − . ∂t ∂y ∂t
