Differential Equations › Section G · Higher order differential equations › Topic G.1

Basic concepts for nth order equations

Bending beams and many control systems lead to equations of third, fourth or higher order. The good news is that the theory built for second order carries over almost word for word; this topic sets out exactly what changes when the order is n.

Beyond second order

Every linear equation in the second order section involved at most \(y''\). Some models need more: a loaded beam's deflection satisfies a fourth order equation, and control-system models often reach third or fourth order. Rather than invent new theory, this section extends the old: superposition, fundamental sets, the Wronskian and the split into complementary and particular solutions all have nth order versions. The main new burden is size, with \(n\) constants, \(n\) conditions and \(n \times n\) determinants. Throughout this section the independent variable is \(t\).

The general linear equation of order n

A linear equation of order \(n\) has the form

P_n(t)y^{(n)} + P_{n-1}(t)y^{(n-1)} + \cdots + P_1(t)y' + P_0(t)y = G(t)

where \(y^{(k)}\) is the \(k\)th derivative. Dividing by \(P_n\) gives the standard form

y^{(n)} + p_{n-1}(t)y^{(n-1)} + \cdots + p_1(t)y' + p_0(t)y = g(t)

which the theorems use. The equation is homogeneous when \(g = 0\). An initial value problem fixes \(n\) values at one point \(t_0\):

y(t_0) = y_0, \ \ y'(t_0) = y_1, \ \ \ldots, \ \ y^{(n-1)}(t_0) = y_{n-1}

So a third order problem needs \(y\), \(y'\) and \(y''\) at \(t_0\), but not \(y'''\).

Existence and uniqueness

The guarantee is the same as before. If \(p_0, p_1, \ldots, p_{n-1}\) and \(g\) are all continuous on an open interval \(I\) containing \(t_0\), then the initial value problem has exactly one solution, and that solution exists on the whole of \(I\). In practice, put the equation in standard form, list the points where any coefficient or \(g\) fails to be continuous, and take the largest open interval around \(t_0\) that avoids them all. The theorem promises nothing beyond those trouble points.

Worked example: the interval of validity

Find the largest interval on which

(t - 3)y''' + (\ln t)\,y' + y = \cos t

with conditions given at \(t_0 = 1\), is guaranteed a unique solution. 1. Divide by \(t - 3\): \(p_1 = \frac{\ln t}{t - 3}\), \(p_0 = \frac{1}{t - 3}\), \(g = \frac{\cos t}{t - 3}\) and \(p_2 = 0\). 2. Trouble points: \(\ln t\) needs \(t > 0\), and every coefficient breaks at \(t = 3\). 3. So everything is continuous on \((0, 3)\) and on \((3, \infty)\). 4. The interval containing \(t_0 = 1\) is \((0, 3)\), so a unique solution exists there.

Superposition and independence

If \(y_1, \ldots, y_n\) solve the homogeneous equation, so does every combination \(c_1y_1 + \cdots + c_ny_n\), because the equation is linear. To be a general solution, the combination must be able to meet any \(n\) initial conditions, which requires the functions to be linearly independent. Functions are linearly independent on \(I\) if \(c_1y_1 + \cdots + c_ny_n = 0\) for every \(t\) in \(I\) forces every \(c_k\) to be 0. Otherwise one of them is a combination of the others, adds nothing new, and the set is dependent. For example, \(1\), \(\cos^2 t\) and \(\sin^2 t\) are dependent, because \(\cos^2 t + \sin^2 t - 1 = 0\).

The Wronskian of n functions

Independence is tested with the \(n \times n\) Wronskian: each column holds one function and its derivatives up to order \(n - 1\),

W(y_1, \ldots, y_n)(t) = \det\begin{pmatrix} y_1 & y_2 & \cdots & y_n \\ y_1' & y_2' & \cdots & y_n' \\ \vdots & \vdots & & \vdots \\ y_1^{(n-1)} & y_2^{(n-1)} & \cdots & y_n^{(n-1)} \end{pmatrix}

It is the determinant of the system you solve to fit initial conditions. The theorem: if \(y_1, \ldots, y_n\) solve the homogeneous equation and \(W \ne 0\) on \(I\), they form a fundamental set, and every solution is \(c_1y_1 + \cdots + c_ny_n\). For solutions of one such equation, \(W\) is either zero everywhere on \(I\) or nowhere, so one point is enough to check.

Worked example: a fundamental set

Show that \(e^t\), \(e^{-t}\) and \(e^{2t}\) form a fundamental set for \(y''' - 2y'' - y' + 2y = 0\). 1. Substituting \(e^{rt}\) gives \((r^3 - 2r^2 - r + 2)e^{rt}\). For \(r = 1\) the bracket is \(1 - 2 - 1 + 2 = 0\); for \(r = -1\) it is \(-1 - 2 + 1 + 2 = 0\); for \(r = 2\) it is \(8 - 8 - 2 + 2 = 0\). All three are solutions. 2. The Wronskian has rows \((e^t, e^{-t}, e^{2t})\), \((e^t, -e^{-t}, 2e^{2t})\) and \((e^t, e^{-t}, 4e^{2t})\). 3. Factor \(e^t\), \(e^{-t}\) and \(e^{2t}\) out of the columns, leaving the rows \((1, 1, 1)\), \((1, -1, 2)\), \((1, 1, 4)\), whose determinant is \(-6\). 4. So \(W = -6e^{2t}\), never zero: a fundamental set, with general solution \(c_1e^t + c_2e^{-t} + c_3e^{2t}\).

Worked example: fitting three conditions

Solve the same equation with \(y(0) = 0\), \(y'(0) = 0\) and \(y''(0) = 6\). 1. Differentiate the general solution twice and set \(t = 0\):

c_1 + c_2 + c_3 = 0, \quad c_1 - c_2 + 2c_3 = 0, \quad c_1 + c_2 + 4c_3 = 6

2. Subtract the first equation from the third: \(3c_3 = 6\), so \(c_3 = 2\). 3. The first two become \(c_1 + c_2 = -2\) and \(c_1 - c_2 = -4\), so \(c_1 = -3\) and \(c_2 = 1\). 4. Answer: \(y = -3e^t + e^{-t} + 2e^{2t}\). Check: \(y(0) = 0\), \(y'(0) = -3 - 1 + 4 = 0\), \(y''(0) = -3 + 1 + 8 = 6\). The coefficient matrix here is the Wronskian matrix at 0, with determinant \(-6\), so a unique answer was guaranteed.

Nonhomogeneous equations

If \(Y_1\) and \(Y_2\) both solve the nonhomogeneous equation, their difference solves the homogeneous one, because the \(g(t)\) terms cancel. So every solution has the form

y = y_c + y_p = c_1y_1 + \cdots + c_ny_n + y_p

where \(y_c\) is the complementary solution, built from a fundamental set, and \(y_p\) is any one particular solution. For example, \(y''' - 2y'' - y' + 2y = 4\) is solved by the constant \(y_p = 2\), since every derivative of a constant is 0 and \(2 \cdot 2 = 4\). Its general solution is \(y_c + 2\), with the three-exponential \(y_c\) found above. Finding \(y_p\) in harder cases is the subject of the topic Undetermined coefficients and variation of parameters.

Common mistakes to avoid

First, applying the existence theorem before dividing by the leading coefficient: its zeros are trouble points too, like \(t = 3\) above. Second, giving the wrong number of initial conditions; order \(n\) needs \(y\) through \(y^{(n-1)}\), all at the same point. Third, building the Wronskian with \(n\) rows of derivatives instead of rows from order 0 up to \(n - 1\). Fourth, using fewer than \(n\) functions: two solutions can never give the general solution of a third order equation. Fifth, applying initial conditions to \(y_c\) alone; add \(y_p\) first.

Key takeaways

A linear equation of order \(n\), in standard form, has a unique solution on any open interval around \(t_0\) where the coefficients and \(g\) are continuous, once \(y, y', \ldots, y^{(n-1)}\) are fixed at \(t_0\). Combinations of homogeneous solutions are solutions. Solutions \(y_1, \ldots, y_n\) with non-zero \(n \times n\) Wronskian form a fundamental set, and then \(c_1y_1 + \cdots + c_ny_n\) is the general homogeneous solution. Nonhomogeneous solutions are \(y_c + y_p\), with \(n\) constants fitted from \(n\) conditions.

Flashcards

Flashcard

Standard form

y⁽ⁿ⁾ + pₙ₋₁(t)y⁽ⁿ⁻¹⁾ + … + p₀(t)y = g(t)

Divide by the leading coefficient first.

Test your understanding

Question 1 of 5

Which initial conditions does a fourth order linear initial value problem need?

Question 2 of 5

On what interval is (t + 2)y‴ + y′/(t - 5) + y = 0, with conditions at t₀ = 0, guaranteed a unique solution?

Question 3 of 5

What is the Wronskian of 1, t and t²?

Question 4 of 5

Which set of functions is linearly dependent?

Question 5 of 5

y_p = 2 solves y‴ - 2y″ - y′ + 2y = 4. What is the general solution?

← Previous topicNext topic →