Unit 4, lesson 20

Second-Order Constant-Coefficient Equations

How the characteristic equation solves a real constant-coefficient homogeneous equation of second order, with distinct real roots, a repeated root or complex conjugate roots, and how to keep the solutions real-valued.

  • Video 20
  • Linear Theory and Constant Coefficients

A mass on a spring with friction and the current in a simple circuit both lead to equations such as y′′+2y′+5y=0y'' + 2y' + 5y = 0, where tt is time and y′=dy/dty' = dy/dt. Because the coefficients are fixed numbers, an exponential guess turns the differential equation into a quadratic equation. A real quadratic has two distinct real roots, one repeated root or a pair of complex roots, and each case gives its own pair of basic solutions. This lesson works through all three and shows how complex roots still lead to real solutions and real initial-value answers.

Key terms

Constant-coefficient homogeneous equation

Definition A second-order constant-coefficient homogeneous equation is

y′′+a y′+b y=0,y'' + a\,y' + b\,y = 0,

where yy is an unknown function of tt, y′=dy/dty' = dy/dt, and aa and bb are real constants. An equation a2y′′+a1y′+a0y=0a_2y'' + a_1y' + a_0y = 0 with real constants and a2≠0a_2 \ne 0 takes this form after division by a2a_2, with a=a1/a2a = a_1/a_2 and b=a0/a2b = a_0/a_2. Constant coefficients are continuous everywhere, so all of R\mathbb{R} is a regular interval. Every initial-value problem y(t0)=k0y(t_0) = k_0, y′(t0)=k1y'(t_0) = k_1 therefore has exactly one solution, defined for all real tt, and once a fundamental set y1y_1, y2y_2 is known, every solution is C1y1+C2y2C_1y_1 + C_2y_2. The characteristic equation below finds such a set.

In plain words The coefficients are fixed numbers rather than functions of time, and nothing is left over on the right side. Every solution then lives for all time, and two independent solutions describe them all.

Example The equation 2y′′+4y′+10y=02y'' + 4y' + 10y = 0 becomes y′′+2y′+5y=0y'' + 2y' + 5y = 0 after division by 22, so a=2a = 2 and b=5b = 5. By contrast, y′′+ty=0y'' + ty = 0 has a coefficient that changes with tt, and the method of this lesson does not apply to it. An exponential trial erte^{rt} there gives (r2+t) ert(r^2 + t)\,e^{rt}, which is not zero for all tt for any constant rr. The equation y′′+2y′+5y=cos⁡ty'' + 2y' + 5y = \cos t is not homogeneous. Its solutions are the homogeneous family found here plus one particular solution, which still has to be found, for example by undetermined coefficients.

Characteristic equation

Definition Write L[y]=y′′+ay′+byL[y] = y'' + ay' + by. For every constant rr, real or complex,

L[ert]=(r2+ar+b) ert.L[e^{rt}] = (r^2 + ar + b)\,e^{rt}.

The polynomial r2+ar+br^2 + ar + b is the characteristic polynomial, and

r2+ar+b=0r^2 + ar + b = 0

is the characteristic equation. Since erte^{rt} is never zero, the exponential erte^{rt} solves L[y]=0L[y] = 0 exactly when rr is a root. The discriminant

Δ=a2−4b\Delta = a^2 - 4b

decides the case. If Δ>0\Delta > 0 there are two distinct real roots, if Δ=0\Delta = 0 there is one repeated real root, and if Δ<0\Delta < 0 there is a pair of complex conjugate roots.

In plain words An exponential keeps its shape under differentiation, and each derivative only multiplies it by rr. So the differential equation reduces to a quadratic equation for the rate rr.

Example For y′′−y′−2y=0y'' - y' - 2y = 0 the characteristic equation r2−r−2=0r^2 - r - 2 = 0 factors as (r−2)(r+1)=0(r - 2)(r + 1) = 0, and the discriminant is 1+8=9>01 + 8 = 9 > 0. For y′′+4y′+4y=0y'' + 4y' + 4y = 0 the discriminant is 16−16=016 - 16 = 0, and for y′′+2y′+5y=0y'' + 2y' + 5y = 0 it is 4−20=−16<04 - 20 = -16 < 0. These three equations illustrate the three cases below.

Distinct real roots

Definition If Δ>0\Delta > 0, the characteristic equation has two real roots r1≠r2r_1 \ne r_2, namely

r1,2=−a±Δ2.r_{1,2} = \frac{-a \pm \sqrt{\Delta}}{2}.

The solutions er1te^{r_1t} and er2te^{r_2t} form a fundamental set on R\mathbb{R}, because their Wronskian

W=(r2−r1) e(r1+r2)tW = (r_2 - r_1)\,e^{(r_1 + r_2)t}

is never zero. Every solution is

y=C1er1t+C2er2t.y = C_1e^{r_1t} + C_2e^{r_2t}.

In plain words Two different rates give two different exponentials, and together they build every solution. A positive root contributes a growing term and a negative root a decaying one, and the initial data decide how much of each is present.

Example For y′′−y′−2y=0y'' - y' - 2y = 0 the roots are 22 and −1-1, so every solution is y=C1e2t+C2e−ty = C_1e^{2t} + C_2e^{-t}, and the Wronskian is −3et-3e^{t}. The data y(0)=1y(0) = 1 and y′(0)=−1y'(0) = -1 give C1+C2=1C_1 + C_2 = 1 and 2C1−C2=−12C_1 - C_2 = -1, so C1=0C_1 = 0, C2=1C_2 = 1 and y=e−ty = e^{-t}, which decays to 00. The data y(0)=1y(0) = 1 and y′(0)=0y'(0) = 0 give 3C1=13C_1 = 1 instead, and

y=13e2t+23e−t,y = \tfrac{1}{3}e^{2t} + \tfrac{2}{3}e^{-t},

which grows without bound as t→∞t \to \infty because its coefficient of e2te^{2t} is not zero.

Repeated root

Definition If Δ=0\Delta = 0, the only root is r0=−a/2r_0 = -a/2. The characteristic polynomial is then (r−r0)2(r - r_0)^2, so a=−2r0a = -2r_0 and b=r02b = r_0^2. The exponential er0te^{r_0t} gives only one solution, and a second one is ter0tte^{r_0t}. Its derivatives are

(ter0t)′=(1+r0t) er0t,(ter0t)′′=(2r0+r02t) er0t.\begin{aligned} (te^{r_0t})' &= (1 + r_0t)\,e^{r_0t}, \\[0.4em] (te^{r_0t})'' &= (2r_0 + r_0^2t)\,e^{r_0t}. \end{aligned}

Substituting them into the equation with a=−2r0a = -2r_0 and b=r02b = r_0^2, the constant terms give 2r0−2r0=02r_0 - 2r_0 = 0 and the terms with tt give r02−2r02+r02=0r_0^2 - 2r_0^2 + r_0^2 = 0. The Wronskian W(er0t,ter0t)=e2r0tW(e^{r_0t}, te^{r_0t}) = e^{2r_0t} is never zero, so every solution is

y=(C1+C2t) er0t.y = (C_1 + C_2t)\,e^{r_0t}.

In plain words When the two rates merge into one, the exponential alone supplies only one solution. Multiplying it by tt gives the missing second one. For a negative root the factor tt changes the early shape of the curve, but the exponential still drives every solution to zero as t→∞t \to \infty.

Example For y′′+4y′+4y=0y'' + 4y' + 4y = 0 the characteristic polynomial is (r+2)2(r + 2)^2, so r0=−2r_0 = -2 and y=(C1+C2t) e−2ty = (C_1 + C_2t)\,e^{-2t}. Writing e−2te^{-2t} twice would not be enough. The pair e−2te^{-2t}, e−2te^{-2t} has Wronskian 00, and its combinations Ce−2tCe^{-2t} only reach data with y′(0)=−2y(0)y'(0) = -2y(0). For the data y(0)=2y(0) = 2 and y′(0)=−1y'(0) = -1, the derivative y′=(C2−2C1−2C2t) e−2ty' = (C_2 - 2C_1 - 2C_2t)\,e^{-2t} gives C1=2C_1 = 2 and C2−2C1=−1C_2 - 2C_1 = -1, so C2=3C_2 = 3 and

y=(2+3t) e−2t.y = (2 + 3t)\,e^{-2t}.

Complex conjugate roots

Definition If Δ<0\Delta < 0, the roots are r=α±iβr = \alpha \pm i\beta with

α=−a2,β=4b−a22>0.\begin{gathered} \alpha = -\frac{a}{2}, \\[0.4em] \beta = \frac{\sqrt{4b - a^2}}{2} > 0. \end{gathered}

They are complex conjugates of each other because aa and bb are real. By Euler’s formula the complex solution e(α+iβ)te^{(\alpha + i\beta)t} equals eαt(cos⁡βt+isin⁡βt)e^{\alpha t}(\cos \beta t + i\sin \beta t), and its real and imaginary parts

y1=eαtcos⁡βt,y2=eαtsin⁡βt\begin{gathered} y_1 = e^{\alpha t}\cos \beta t, \\[0.4em] y_2 = e^{\alpha t}\sin \beta t \end{gathered}

are real solutions with Wronskian βe2αt≠0\beta e^{2\alpha t} \ne 0. Every real solution is therefore

y=C1eαtcos⁡βt+C2eαtsin⁡βt\begin{aligned} y &= C_1e^{\alpha t}\cos \beta t \\[0.2em] &\quad + C_2e^{\alpha t}\sin \beta t \end{aligned}

with real constants C1C_1 and C2C_2, which is often written y=eαt(C1cos⁡βt+C2sin⁡βt)y = e^{\alpha t}(C_1\cos \beta t + C_2\sin \beta t).

In plain words Complex roots mean oscillation. The number β\beta is the angular frequency of the swings, and the real part α\alpha sets the envelope eαte^{\alpha t}. The swings die out when α<0\alpha < 0, keep a constant size when α=0\alpha = 0 and grow when α>0\alpha > 0.

Example For y′′−2y′+10y=0y'' - 2y' + 10y = 0 the roots of r2−2r+10=0r^2 - 2r + 10 = 0 are 1±3i1 \pm 3i, so every real solution is et(C1cos⁡3t+C2sin⁡3t)e^{t}(C_1\cos 3t + C_2\sin 3t). It oscillates with angular frequency 33 inside a growing envelope. For y′′+9y=0y'' + 9y = 0 the roots are ±3i\pm 3i, so α=0\alpha = 0 and the solutions C1cos⁡3t+C2sin⁡3tC_1\cos 3t + C_2\sin 3t keep a constant amplitude.

Real-valued solutions

Definition Let aa and bb be real, and let z=u+ivz = u + iv, with uu and vv real-valued, be a complex-valued solution of L[y]=y′′+ay′+by=0L[y] = y'' + ay' + by = 0. Then uu and vv are real solutions, because

0=L[z]=L[u]+i L[v]0 = L[z] = L[u] + i\,L[v]

and both L[u]L[u] and L[v]L[v] are real-valued, so both are zero. With complex roots the complex solutions are c1e(α+iβ)t+c2e(α−iβ)tc_1e^{(\alpha + i\beta)t} + c_2e^{(\alpha - i\beta)t} with complex constants c1c_1 and c2c_2, and such a combination is real-valued exactly when c2=c1‾c_2 = \overline{c_1}. Choosing c1=(C1−iC2)/2c_1 = (C_1 - iC_2)/2 and c2=c1‾c_2 = \overline{c_1} with real C1C_1 and C2C_2 gives the real form eαt(C1cos⁡βt+C2sin⁡βt)e^{\alpha t}(C_1\cos \beta t + C_2\sin \beta t), and real initial data always lead to real C1C_1 and C2C_2. When R=C12+C22>0R = \sqrt{C_1^2 + C_2^2} > 0, the same solution can be written

y=R eαtcos⁡(βt−φ),y = R\,e^{\alpha t}\cos(\beta t - \varphi),

where the phase φ\varphi satisfies Rcos⁡φ=C1R\cos\varphi = C_1 and Rsin⁡φ=C2R\sin\varphi = C_2.

In plain words Complex numbers are a tool for finding the solutions, not part of the final answer. The real and imaginary parts of one complex solution already give two real solutions, and real starting values select a real curve.

Example Solve y′′+2y′+5y=0y'' + 2y' + 5y = 0 with y(0)=3y(0) = 3 and y′(0)=1y'(0) = 1. The roots of r2+2r+5=0r^2 + 2r + 5 = 0 are −1±2i-1 \pm 2i, so every real solution has the form y=e−t(C1cos⁡2t+C2sin⁡2t)y = e^{-t}(C_1\cos 2t + C_2\sin 2t). The height gives C1=3C_1 = 3. The product rule applied to the whole product gives

y′=e−t(2C2−C1)cos⁡2t−e−t(2C1+C2)sin⁡2t,\begin{aligned} y' &= e^{-t}(2C_2 - C_1)\cos 2t \\[0.2em] &- e^{-t}(2C_1 + C_2)\sin 2t, \end{aligned}

so y′(0)=2C2−C1=1y'(0) = 2C_2 - C_1 = 1 and C2=2C_2 = 2. The solution is

y=e−t(3cos⁡2t+2sin⁡2t),y = e^{-t}(3\cos 2t + 2\sin 2t),

an oscillation with angular frequency 22 whose amplitude 13 e−t\sqrt{13}\,e^{-t} decays. Dropping the derivative of e−te^{-t} would give 2C2=12C_2 = 1 and the wrong coefficient C2=12C_2 = \tfrac{1}{2}.

Check yourself

  1. Solve y′′+y′−6y=0y'' + y' - 6y = 0 with y(0)=0y(0) = 0 and y′(0)=5y'(0) = 5.
Show the answer

The characteristic equation r2+r−6=0r^2 + r - 6 = 0 factors as (r−2)(r+3)=0(r - 2)(r + 3) = 0, so y=C1e2t+C2e−3ty = C_1e^{2t} + C_2e^{-3t}. The data give C1+C2=0C_1 + C_2 = 0 and 2C1−3C2=52C_1 - 3C_2 = 5. Substituting C2=−C1C_2 = -C_1 gives 5C1=55C_1 = 5, so C1=1C_1 = 1, C2=−1C_2 = -1 and

y=e2t−e−3t.y = e^{2t} - e^{-3t}.
  1. For y′′−6y′+9y=0y'' - 6y' + 9y = 0 a classmate writes y=C1e3t+C2e3ty = C_1e^{3t} + C_2e^{3t} and tries to meet y(0)=1y(0) = 1 and y′(0)=2y'(0) = 2. Explain what goes wrong and find the correct solution.
Show the answer

The characteristic polynomial is (r−3)2(r - 3)^2, so 33 is a repeated root. The classmate’s family is only Ce3tCe^{3t} with C=C1+C2C = C_1 + C_2, and every member has y′(0)=3y(0)y'(0) = 3y(0). With y(0)=1y(0) = 1 that forces y′(0)=3y'(0) = 3 rather than 22, so no member fits. The correct family is y=(C1+C2t) e3ty = (C_1 + C_2t)\,e^{3t}, with y(0)=C1y(0) = C_1 and y′(0)=3C1+C2y'(0) = 3C_1 + C_2. So C1=1C_1 = 1, C2=−1C_2 = -1 and

y=(1−t) e3t.y = (1 - t)\,e^{3t}.

Its derivative y′=(2−3t) e3ty' = (2 - 3t)\,e^{3t} gives y′(0)=2y'(0) = 2 as required.

  1. Solve y′′+4y′+13y=0y'' + 4y' + 13y = 0 with y(0)=1y(0) = 1 and y′(0)=0y'(0) = 0, and describe the behavior of the solution as t→∞t \to \infty.
Show the answer

The roots of r2+4r+13=0r^2 + 4r + 13 = 0 are −2±3i-2 \pm 3i, so y=e−2t(C1cos⁡3t+C2sin⁡3t)y = e^{-2t}(C_1\cos 3t + C_2\sin 3t). The height gives C1=1C_1 = 1. By the product rule y′(0)=−2C1+3C2y'(0) = -2C_1 + 3C_2, so −2+3C2=0-2 + 3C_2 = 0 and C2=23C_2 = \tfrac{2}{3}. The solution is

y=e−2t(cos⁡3t+23sin⁡3t).y = e^{-2t}\left(\cos 3t + \tfrac{2}{3}\sin 3t\right).

It oscillates with angular frequency 33 inside the envelope ±133e−2t\pm\tfrac{\sqrt{13}}{3}e^{-2t}, so it tends to 00. Forgetting the derivative of e−2te^{-2t} would give 3C2=03C_2 = 0 and the wrong answer C2=0C_2 = 0.

How this lesson connects