Differential Equations › Section C · Second order differential equations › Topic C.6
Fundamental sets and the Wronskian
So far we have trusted that two different-looking solutions are enough to build every solution. This topic makes that precise: a single determinant, the Wronskian, decides whether a pair of solutions can meet every possible initial condition.
The question left open
The earlier topics of this section kept calling a pair of solutions "suitable" without saying exactly what that means. The question matters. A general solution must be able to match any starting position and any starting velocity, otherwise some motions of the system would be missing from the formula. This topic works with the standard form
and, for the homogeneous case, \(g = 0\). Nothing here needs constant coefficients, so the results also apply to equations like those in "Reduction of order".
Existence and uniqueness
The foundation is a theorem. If \(p\), \(q\) and \(g\) are continuous on an open interval \(I\) containing \(t_0\), then for any numbers \(y_0\) and \(y_0'\) the initial value problem with \(y(t_0) = y_0\) and \(y'(t_0) = y_0'\) has exactly one solution, and it exists on the whole of \(I\). Compare the first order version in "Existence, uniqueness, and intervals of validity": here, too, the guaranteed interval is the largest open interval around \(t_0\) on which the coefficients, in standard form, are continuous. For \((t - 4)y'' + y' + ty = 0\) with conditions at \(t_0 = 1\), dividing by \(t - 4\) shows trouble only at \(t = 4\), so a unique solution exists on \((-\infty, 4)\).
Solving for the constants
Suppose \(y_1\) and \(y_2\) solve the homogeneous equation, and try to match \(y(t_0) = y_0\) and \(y'(t_0) = y_0'\) with \(y = c_1y_1 + c_2y_2\). That gives a linear system: \(c_1y_1(t_0) + c_2y_2(t_0) = y_0\) and \(c_1y_1'(t_0) + c_2y_2'(t_0) = y_0'\). Cramer's rule solves it:
Both share the same denominator \(W\). If \(W \ne 0\), every choice of initial values can be met. If \(W = 0\), some choices cannot.
The Wronskian
That denominator is the Wronskian of the two functions, a 2 by 2 determinant:
It is named after the Polish mathematician Józef Hoene-Wroński. Two solutions \(y_1\) and \(y_2\) of the homogeneous equation on \(I\) form a fundamental set of solutions if \(W(y_1, y_2)(t_0) \ne 0\) at a point \(t_0\) of \(I\). Then every solution on \(I\) can be written as \(c_1y_1 + c_2y_2\), because any solution is fixed by its value and slope at \(t_0\), and the uniqueness theorem says that the combination matching them is that solution. The next topic shows that for two solutions, nonzero at one point means nonzero everywhere on \(I\).
Worked example: the three cases confirmed
1. Real, distinct roots:
which is never zero because \(r_1 \ne r_2\). 2. Complex roots: for \(e^{\alpha t}\cos\beta t\) and \(e^{\alpha t}\sin\beta t\), the product rule and \(\cos^2 + \sin^2 = 1\) give
which is never zero because \(\beta \ne 0\). 3. Repeated root:
which is never zero. So all three forms of general solution used in this section are genuinely general.
Worked example: a variable coefficient pair
The topic "Reduction of order" found \(y_1 = t^2\) and \(y_2 = t^2\ln t\) for \(t^2y'' - 3ty' + 4y = 0\) on \(t > 0\). Confirm they form a fundamental set. 1. Derivatives: \(y_1' = 2t\) and \(y_2' = 2t\ln t + t\). 2. Wronskian: \(W = t^2(2t\ln t + t) - 2t \cdot t^2\ln t\). 3. The \(\ln t\) terms cancel, leaving \(W = t^3\). 4. For \(t > 0\), \(t^3 > 0\), so the pair is fundamental and \(y = c_1t^2 + c_2t^2\ln t\) is the general solution there. The point \(t = 0\) is excluded anyway: in standard form, \(p = -\frac{3}{t}\) and \(q = \frac{4}{t^2}\) are not continuous there.
A pair that fails
For \(y'' - 4y = 0\), both \(e^{2t}\) and \(3e^{2t}\) are solutions, but \(W = e^{2t}(6e^{2t}) - 2e^{2t}(3e^{2t}) = 0\). Their combinations are all of the form \(ke^{2t}\), whose slope at 0 is always twice its value. So the condition \(y(0) = 1\), \(y'(0) = 0\) can never be met, even though the solution \(\cosh 2t\) satisfies it. A zero Wronskian signals exactly this: the pair covers only part of the solutions. In contrast, \(\cosh 2t\) and \(\sinh 2t\), with \(W = 2\), form a fundamental set, just as \(e^{2t}\) and \(e^{-2t}\) do. A fundamental set is never unique.
Fundamental sets always exist
A theorem guarantees a fundamental set whenever \(p\) and \(q\) are continuous on \(I\). Pick \(t_0\) in \(I\). Let \(y_1\) be the solution with \(y_1(t_0) = 1\), \(y_1'(t_0) = 0\), and \(y_2\) the solution with \(y_2(t_0) = 0\), \(y_2'(t_0) = 1\); both exist by the existence and uniqueness theorem. Then \(W(t_0) = 1 \cdot 1 - 0 \cdot 0 = 1\). For \(y'' - y' - 6y = 0\) at \(t_0 = 0\), the roots 3 and \(-2\) give \(y_1 = \frac{2e^{3t} + 3e^{-2t}}{5}\) and \(y_2 = \frac{e^{3t} - e^{-2t}}{5}\). That pair is correct but clumsy; the simpler \(e^{3t}\) and \(e^{-2t}\) work equally well.
Common mistakes to avoid
First, getting the sign order wrong: \(W = y_1y_2' - y_1'y_2\), and swapping it changes the sign, which does not affect whether \(W\) is zero but does give a wrong value. Second, forgetting the product rule for functions like \(te^{rt}\). Third, applying the existence theorem without first dividing by the leading coefficient, and so missing the points where it vanishes. Fourth, thinking that two solutions must look completely different to be fundamental: \(e^{2t}\) and \(e^{2t} - e^{-2t}\) form a fundamental set, with \(W = 4\).
Key takeaways
If \(p\), \(q\) and \(g\) are continuous on an open interval containing \(t_0\), the initial value problem has exactly one solution on that interval. Fitting \(c_1y_1 + c_2y_2\) to initial conditions is possible for every choice exactly when the Wronskian \(W = y_1y_2' - y_1'y_2\) is nonzero at \(t_0\). Two solutions with nonzero Wronskian form a fundamental set, and then every solution is \(c_1y_1 + c_2y_2\). The exponential, oscillating and repeated root pairs all pass this test, and a fundamental set always exists.
Flashcards
Flashcard
y″ + p(t)y′ + q(t)y = g(t)
Test your understanding
Question 1 of 5
What is the Wronskian W(e^(2t), e^(−3t))?
Question 2 of 5
What is the Wronskian W(cos 3t, sin 3t)?
Question 3 of 5
Which pair is NOT a fundamental set of solutions for y″ − 4y = 0?
Question 4 of 5
For (t − 4)y″ + y′ + ty = 0 with y(1) = 2 and y′(1) = 0, what is the largest interval on which a unique solution is guaranteed?
Question 5 of 5
For y″ + y = 0 and t₀ = 0, the theorem builds a fundamental set from solutions with (y(0), y′(0)) = (1, 0) and (0, 1). What are they?
