Suppose we have the differential equation M(x,y)+N(x,y)y′=0 and a region D. If there is a function Ψ(x,y) so that Ψx(x,y)=M(x,y),Ψy=N(x,y) for all (x,y)∈D, then we call the differential equation exact in D. In this case, the implicit solution is Ψ(x,y)=c for (x,y)∈D, and we call Ψ(x,y) the potential function.
Theorem (Test for exact differential equation) Suppose the region D is simply connected. The differential equation M(x,y)+N(x,y)y′=0 is exact in D if and only if My(x,y)=Nx(x,y) at each (x,y)∈D.
Problem 1 Determine whether each of the equations is exact. If it is exact, find the solution. (a) 3x2−2xy+2+(6y2−x2+3)y′=0; (b) (exsiny−2ysinx)+(excosy+2cosx)y′=0.
Solution 1(a) Since ∂y∂3x2−2xy+2=−2x and ∂x∂6y2−x2+3=−2x agree for all (x,y), the DE is exact. Let Ψ(x,y) be the potential function. Since Ψx=3x2−2xy+2, Ψ=x3−x2y+2x+C(y). Since Ψy=−x2+C′(y)=6y2−x2+3, C(y)=2y3+3y. The implicit solution to the DE is then x3−x2y+2x+2y3+3y=C.
Solution 1(b) Since ∂y∂exsiny−2ysinx=excosy−2sinx and ∂x∂excosy+2cosx=excosy−2sinx agree for all (x,y), the DE is exact. Let Ψ(x,y) be the potential function. Since Ψx=exsiny−2ysinx, Ψ=exsiny+2ycosx+C(y). Since Ψy=excosy+2cosx+C′(y)=excosy+2cosx, C(y)=0. The implicit solution to the DE is then exsiny+2ycosx=C.
Idea It is sometimes possible to convert a differential equation that is not exact into an exact equation by multiplying the equation by a suitable integrating factor.
Problem 2 Given that the integrating factor depends on y only, solve y+(2x−yey)y′=0.
Solution Let μ(y) be the integrating factor such that μ(y)y+μ(y)(2x−yey)y′=0 becomes exact. By the test for exactness, (μ(y)y)y=(μ(y)(2x−yey))x, equivalently, μ′(y)y+μ(y)=2μ(y). Thus μ(y) satisfies μ′=yμ(y) and so μ(y)=y. Therefore y2+(2xy−y2ey)y′=0 is exact. Let Ψ be the potential function. Since Ψx=y2, Ψ=xy2+C(y). Since Ψy=2xy+C′(y)=2xy−y2ey, C(y)=∫−y2eydy=−ey(y2−2y+2). The solution to the DE is then xy2−ey(y2−2y+2)=C.
Remark In general, M and N have to satisfy certain condition so that M(x,y)+N(x,y)y′=0 has an integrating factor μ that depends on y only. Let μ(y) be an integrating factor. This means μ(y)M(x,y)+μ(y)N(x,y)y′=0 is exact. Thus (μ(y)M(x,y))y=(μ(y)N(x,y))x. The left hand side equals μ′M+μMy, and the right hand side equals μNx. This shows μ′M+μMy=μNx, that is μ′=MNx−Myμ. The condition for the existence of such μ is that MNx−My depends only on y.
Definition Consider two families of curves F and G. We say that they are orthogonal whenever any curve from F intersects any curve from G, and the two curves are orthogonal at the point of intersection.
Example The families y=kx and x2+y2=r2 are orthogonal.
Question Given a family of curves F, is it possible to find a family of curves which is orthogonal to F?
Example Each line through the origin, that is, y=ax, we have y′=a, and so xy′=ax=y. In other words, every line in the given family satisfies the DE xy′=y. Rewrite the DE as y′=y/x. The slope diagram is as follows:
Imagine that each little segment in the slope diagram turns 90 degrees:
On one hand, the 2nd slope diagram satisfies the DE y′=−1/(y/x)=−x/y (why?). On the other hand, a solution (an integral curve) to the 2nd slope will intersect every line through the origin at a right angle (again, why?). This means the family of curves that is orthogonal to the family of lines through the origin satisfies the DE y′=−x/y. By separation of variables, x+yy′=0, and so x2+y2=C. This represents the family of circles centered at the origin.
We summarize the above discussion as follows.
Step 1 Find the differential equation associated to the given family of curves F and rewrite this differential equation in the explicit form dxdy=f(x,y).
Step 2 Write down the differential equation dxdy=f(x,y)−1 associated to the orthogonal family G, and solve the differential equation.
Optional step Give a geometric view of the two families.
Exercise 1 Find the orthogonal family of the family of circles x2+y2=2cx.
Exercise 2 (Hard) Find the orthogonal family of the family of cardiodsx2+y2+ax=ax2+y2.