Differential Equations › Section C · Second order differential equations › Topic C.7
More on the Wronskian
The Wronskian does more than certify a fundamental set. It tests whether two functions are linearly independent, and Abel's theorem lets you find it for an equation's solutions without solving the equation at all.
Two questions about one determinant
The previous topic, "Fundamental sets and the Wronskian", used \(W = y_1y_2' - y_1'y_2\) to decide whether two solutions can match every initial condition. Two natural questions follow. First, how does the Wronskian relate to the everyday idea that two functions are "really different", meaning neither is just a rescaled copy of the other? Second, is there a way to know the Wronskian of an equation's solutions before finding them? This topic answers both: the first with linear independence, the second with Abel's theorem.
Linear dependence and independence
Two functions \(f\) and \(g\) are linearly dependent on an interval \(I\) if there are constants \(c\) and \(k\), not both zero, with
Otherwise, if only \(c = k = 0\) works, they are linearly independent on \(I\). For two functions, dependence means exactly that one is a constant multiple of the other. For instance, \(e^{t + 3} = e^3 \cdot e^t\), so \(e^{t + 3}\) and \(e^t\) are dependent. Sometimes an identity hides the dependence: \(4\cos^2 t - 2\) and \(\cos 2t\) are dependent, because the double angle formula shows the first is \(2\cos 2t\).
The Wronskian test
Checking the definition directly can be awkward, so the Wronskian offers a shortcut. If \(f\) and \(g\) are differentiable on \(I\) and \(W(f, g)(t_0) \ne 0\) for some \(t_0\) in \(I\), then \(f\) and \(g\) are linearly independent on \(I\). The reason: if \(cf + kg = 0\) for all \(t\), then differentiating gives \(cf' + kg' = 0\) too. At \(t_0\) these two equations form a linear system for \(c\) and \(k\) whose determinant is \(W(t_0) \ne 0\), so its only solution is \(c = k = 0\). Turned around, the same fact says that dependent functions have \(W = 0\) at every point of \(I\).
Worked example: three pairs
Decide whether each pair is independent. 1. \(f = t\) and \(g = t^3\): \(W = t \cdot 3t^2 - 1 \cdot t^3 = 2t^3\), which equals 2 at \(t = 1\), so the pair is independent on any interval containing \(t = 1\); in fact, on any interval at all, since \(2t^3\) vanishes only at \(t = 0\). 2. \(f = 4\cos^2 t - 2\) and \(g = \cos 2t\): \(f' = -8\cos t\sin t = -4\sin 2t\) and \(g' = -2\sin 2t\). Writing \(f = 2\cos 2t\),
The zero matches the dependence found earlier. 3. \(e^{t + 3}\) and \(e^t\): \(W = e^{t + 3}e^t - e^{t + 3}e^t = 0\), again matching.
A zero Wronskian is not proof
The test only works in one direction. A Wronskian that is zero everywhere does not, for general functions, prove dependence. The classic example is \(f = t^2\) and \(g = t|t|\) on \((-1, 1)\). Here \(g' = 2|t|\), so \(W = t^2 \cdot 2|t| - 2t \cdot t|t| = 0\) for every \(t\). Yet the functions are independent: if \(ct^2 + kt|t| = 0\) for all \(t\), then \(t = 1\) gives \(c + k = 0\) and \(t = -1\) gives \(c - k = 0\), so \(c = k = 0\). On \((0, 1)\) alone, though, \(g = f\), and they are dependent there. Independence always depends on the interval.
Abel's theorem
For solutions of an equation, the picture is much cleaner. Let \(y_1\) and \(y_2\) solve \(y'' + p(t)y' + q(t)y = 0\) with \(p\) and \(q\) continuous on \(I\). Differentiate the Wronskian: \(W' = y_1y_2'' - y_1''y_2\), since the \(y_1'y_2'\) terms cancel. Replacing each \(y''\) by \(-py' - qy\), the \(q\) terms cancel too, leaving \(W' = -p(t)W\). This first order equation solves to
or, more briefly, \(W = Ce^{-\int p\,dt}\). That is Abel's theorem, after the Norwegian mathematician Niels Henrik Abel. It requires the equation in standard form, with leading coefficient 1.
Zero once means zero always
The exponential in Abel's formula is never zero. So for two solutions on \(I\), there are only two possibilities: if \(W(t_0) = 0\) at one point, then \(W\) is zero everywhere on \(I\), and if \(W(t_0) \ne 0\), then \(W\) is never zero there. Moreover, for solutions, a zero Wronskian does mean dependence: if \(W(t_0) = 0\), some \(c\) and \(k\), not both zero, make \(cy_1 + ky_2\) have value 0 and slope 0 at \(t_0\), and by uniqueness that combination is the zero solution. The upshot is a clean equivalence: two solutions form a fundamental set exactly when they are linearly independent on \(I\). It also shows that \(t^2\) and \(t|t|\) cannot both solve such an equation on an interval containing 0.
Worked example: a Wronskian without solving
Find the Wronskian of two solutions of \(t^2y'' + ty' + (t^2 - 9)y = 0\) on \(t > 0\). 1. Divide by \(t^2\) to reach standard form: \(p(t) = \frac{1}{t}\). 2. Integrate: \(\int p\,dt = \ln t\). 3. Abel's theorem: \(W = Ce^{-\ln t} = \frac{C}{t}\). 4. So any two solutions have a Wronskian of the form \(C/t\), even though their formulas involve Bessel functions, which come later in "The Frobenius method". 5. Check the method on a known pair: \(t^2y'' - 3ty' + 4y = 0\) has \(p = -\frac{3}{t}\), so Abel gives \(W = Ce^{3\ln t} = Ct^3\), matching the value \(t^3\) found directly for \(t^2\) and \(t^2\ln t\). For constant coefficients, \(ay'' + by' + cy = 0\) gives \(W = Ce^{-bt/a}\).
Common mistakes to avoid
First, concluding that two arbitrary functions are dependent because their Wronskian is zero; that inference is valid only for solutions of a linear equation like the one in Abel's theorem. Second, applying Abel's formula before dividing by the leading coefficient: for \(t^2y'' + 3ty' + y = 0\), \(p\) is \(\frac{3}{t}\), not \(3t\). Third, dropping the minus sign in \(e^{-\int p\,dt}\). Fourth, forgetting that independence is a statement about an interval. Fifth, thinking a fundamental set must have a constant Wronskian; Abel shows it usually varies.
Key takeaways
Two functions are linearly dependent on \(I\) when one is a constant multiple of the other there. If their Wronskian is nonzero at even one point of \(I\), they are independent; a zero Wronskian proves nothing for general functions, as \(t^2\) and \(t|t|\) show. For solutions of \(y'' + py' + qy = 0\), Abel's theorem gives \(W = Ce^{-\int p\,dt}\), so the Wronskian is either never zero or always zero, and fundamental sets are exactly the independent pairs of solutions.
Flashcards
Flashcard
cf + kg = 0 on I with c, k not both zero
Test your understanding
Question 1 of 5
Which pair of functions is linearly dependent on all real numbers?
Question 2 of 5
Two differentiable functions, not known to solve any equation, have W(f, g) = 0 at every point of I. What follows?
Question 3 of 5
Without solving, what form does the Wronskian of two solutions of y″ + 4y′ − 5y = 0 take?
Question 4 of 5
What form does the Wronskian of two solutions of t²y″ + 3ty′ + y = 0 take on t > 0?
Question 5 of 5
Two solutions of y″ + p(t)y′ + q(t)y = 0, with p and q continuous on I, have W(t₀) = 0 at one point t₀ in I. What follows?
