< Home

Chapter 4

4.1

nnth order differential equations

an(x)y(n)(x)+an−1(x)y(n−1)(x)+⋯+a2(x)y′′(x)+a1(x)y′(x)+a0(x)y(x)=f(x) a _n (x) y ^{(n)} (x) + a _{n-1} (x) y ^{(n - 1)} (x) + \dots + a _2 (x) y\prime\prime(x) + a _1 (x) y\prime(x) + a_0(x)y(x) = f(x)

if f(x)=0f(x) = 0 then "homogeneous"
otherwise "non-homogeneous"

Example: y′′−4y=0 y\prime \prime - 4y = 0

linear, 2nd order, homogeneous, constant coefficient

Example: y′′+ry′−3y=2t y \prime \prime + ry\prime - 3y = 2t

linear, 2nd order, non-homogeneous, constant coefficient

Example: 3x3y′′′−6xy+y=10ex 3x^3y\prime\prime\prime-6xy+y=10e^x

linear, 3rd order, non-homogeneous

Example: y′′′+49y=0 y\prime\prime\prime + 49y = 0

Guessed Solution:

y(x)={sin⁡(7x),cos⁡(7x)} y(x) = \left\{ \sin(7x), \cos(7x) \right\}

Suppose we have an nnth-order linear, homogeneous DE:

an(x)y(n)(x)+an−1(x)y(n−1)(x)+⋯+a2(x)y′′(x)+a1(x)y′(x)+a0(x)y(x)=0 a _n (x) y ^{(n)} (x) + a _{n-1} (x) y ^{(n - 1)} (x) + \dots + a _2 (x) y\prime\prime(x) + a _1 (x) y\prime(x) + a_0(x)y(x) = 0

if we find nn solutions to it, [unclear] consider

y=C1y1+C2y2+⋯+Cnyn(*) y = C_1y_1 + C_2y_2 + \dots + C_ny_n \tag{*}

Initial conditions: y(x0)=b0y′(x0)=b1⋮y(n−1)(x0)=bn−1 \begin{align} y(x_0) =& b_0 \\ y\prime(x_0) =& b_1 \\ & \vdots \\ y^{(n-1)}(x_0) =& b_{n-1} \\ \end{align}

Plug in ICs for (*)\text{(*)}:

(**){C1y1(x0)+C2y2(x0)+…+Cnyn(x0)=b0C1y1′(x0)+C2y2′(x0)+…+Cnyn′(x0)=b1C1y1′′(x0)+C2y2′′(x0)+…+Cnyn′′(x0)=b2⋮C1y1(n−1)∗(x0)+C2y2(n−1)(x0)+…+Cnyn(n−1)(x0)=bn}⇔[y1(x0)y2(x0)…yn(x0)y1′(x0)y2′(x0)…yn′(x0)⋮y1(n−1)(x0)y2(n−1)(x0)…yn(n−1)(x0)][C1C2⋮Cn]=[b1b2⋮bn] \begin{align} \text{(**)} \left\{ \begin{align} C_1y_1(x_0) & + C_2y_2(x_0) & + \dots & + C_ny_n(x_0) &= b_0& \\ C_1y_1\prime(x_0) & + C_2y_2\prime(x_0) & + \dots & + C_ny_n\prime(x_0) &= b_1& \\ C_1y_1\prime\prime(x_0) & + C_2y_2\prime\prime(x_0) & + \dots & + C_ny_n\prime\prime(x_0) &= b_2& \\ & \vdots \\ C_1y_1^{(n-1)}*(x_0) & + C_2y_2^{(n-1)}(x_0) & + \dots & + C_ny_n^{(n-1)}(x_0) &= b_n& \\ \end{align} \right\} \Leftrightarrow &\left[ \begin{array}{cccc} y_1(x_0) & y_2(x_0) & \dots & y_n(x_0) \\ y_1\prime(x_0) & y_2\prime(x_0) & \dots & y_n\prime(x_0) \\ \vdots & & & \\ y_1^{(n-1)}(x_0) & y_2^{(n-1)}(x_0) & \dots & y_n^{(n-1)}(x_0) \\ \end{array} \right] \left[ \begin{array}{c} C_1 \\ C_2 \\ \vdots \\ C_n \\ \end{array} \right] = \left[ \begin{array}{c} b_1 \\ b_2 \\ \vdots \\ b_n \\ \end{array} \right] \end{align}

Only valid if W=det⁡((**))≠0W = \det(\text{(**)}) \neq 0

(WW: "Wronskian")

For the previous Example:

y1(x)=sin⁡(7x)y2(x)=cos⁡(7x) \begin{align} y_1(x) &= \sin(7x) \\ y_2(x) &= \cos(7x) \\ \end{align}

∣sin⁡(7x)cos⁡(7x)7cos⁡(7x)−7sin⁡(7x)∣=−7sin⁡2(7x)−7cos⁡2(7x)=−7(sin⁡2(7x)cos⁡2(7x))=−7≠0 \begin{align} \left| \begin{array}{cc} \sin(7x) & \cos(7x) \\ 7\cos(7x) & -7\sin(7x) \\ \end{array} \right| &= -7\sin^2(7x)-7\cos^2(7x) \\ &= -7\left(\sin^2(7x)\cos^2(7x)\right) \\ &= -7 \\ &\neq 0 \end{align}

Linear independence of a function

Definition

We say y1(x),y2(x),…,yn(x)y_1(x), y_2(x), \dots, y_n(x) is linearly independent on the interval II if C1y1(x)+C2y2(x)+⋯+CnYn(x)C_1y_1(x) + C_2y_2(x) + \dots + C_nY_n(x) on II only has the solution C1=C2=⋯=Cn=0C_1 = C_2 = \dots = C_n = 0

Example: y1(x)=cos⁡2(x)y2(x)=8sin⁡2(x)y3(x)=4 \begin{align} y_1(x) &= \cos^2(x) \\ y_2(x) &= 8\sin^2(x) \\ y_3(x) &= 4 \\ \end{align} Is {y1(x),y2(x),y3(x)}\left\{y_1(x), y_2(x), y_3(x)\right\} linearly independent over R\R? Observe: 8y1+y2=8cos⁡2(x)+8sin⁡2(x)=8=2y3(x) \begin{align} 8y_1 + y_2 &= 8\cos^2(x)+8\sin^2(x) \\ &= 8 \\ &= 2y_3(x) \\ \end{align}

⇒8y1(x)+y2(x)−2y3(x)=0For all x in R⇒{y1(x),y2(x),y3(x)}is not linearly independent over R \begin{align} \Rightarrow 8y_1(x) + y_2(x) - 2y_3(x) = 0 &&& \text{For all x in $\R$} \\ \Rightarrow \left\{y_1(x), y_2(x), y_3(x)\right\} &&& \text{is not linearly independent over $\R$} \end{align}

If {y1(x),y2(x),…,yn(x)}\left\{ y_1(x), y_2(x), \dots, y_n(x) \right\} is a collection of n−1n-1 differentiable functions over II, and W(x0)≠0W(x_0) \neq 0 at any pt x∈Ix \in I, then {y1(x),y2(x),…,yn(x)}\left\{ y_1(x), y_2(x), \dots, y_n(x) \right\} is linearly independent.

Conversely, if yi(x)y_i(x)s are infinitely differentiable, then W(x)=0W(x)= 0 implies linear dependence.

Example: y1(x)=xy2(x)=∣x∣ \begin{align} y_1(x) &= x \\ y_2(x) &= |x| \\ \end{align} Is {y1(x),y2(x)}\left\{ y_1(x), y_2(x) \right\} linearly independent on (−1,1)⏟I\underbrace{(-1, 1)}_{I}?

Observe: for any x≥0,y1(x)=x=y2(x)⇒y1(x)−y2(x)=0for any x<0,y1(x)+y2(x)=0 \begin{align} \text{for any } & x \ge 0, & y_1(x) = x = y_2(x) \Rightarrow y_1(x) - y_2(x) = 0 \\ \text{for any } & x \lt 0, & y_1(x) + y_2(x) = 0 \end{align} Since constants differ, {y1,y2}\left\{y_1, y_2\right\} is independent.

4.3 - Homogeneous linear DEs with constant coefficients

any(n)+an−1y(n−1)+⋯+a2y′′+a1y′+a0y=0 a _n y ^{(n)} + a _{n-1} y ^{(n - 1)} + \dots + a _2 y\prime\prime + a _1 y\prime + a_0y = 0

aia_i constant

Educated guess that eλxe^{\lambda x} is the solution

Plug it in:

an(eλx)(n)+an−1(eλx)(n−1)+⋯+a1(eλx)′+a0eλx=0anλneλx+an−1λn−1eλx+⋯+a1λeλx+a0eλx=0eλx[anλn+an−1λn−1+⋯+a1λ+a0]=0⇔anλn+an−1λn−1+⋯+a1λ+a0=0"Characteristic Equation" \begin{align} a_n(e^{\lambda x})^{(n)} + a_{n-1}(e^{\lambda x})^{(n - 1)} + \dots + a_{1}(e^{\lambda x})\prime + a_0e^{\lambda x} &= 0 \\ a_n \lambda^n e^{\lambda x} + a_{n-1} \lambda^{n-1} e^{\lambda x} + \dots + a_1 \lambda e^{\lambda x} + a_0 e^{\lambda x} &= 0 \\ e^{\lambda x}\left[a_n \lambda^n + a_{n-1} \lambda^{n-1} + \dots + a_1 \lambda + a_0 \right] &= 0 \\ \Leftrightarrow a_n \lambda^n + a_{n-1} \lambda^{n-1} + \dots + a_1 \lambda + a_0 &= 0 &&& \text{"Characteristic Equation"} \\ \end{align}


Fundamental Theorem of Algebra: The characteristic equation in principle always has nn roots, λ1,λ2,…,λn\lambda_1, \lambda_2, \dots, \lambda_n, some may repeat, some may be complex.

Simple case when n=2n = 2:

y′′+py′+2y=0 y\prime\prime + py\prime + 2y = 0

Form characteristic equation:

λ2+pλ+q=0 \lambda^2 + p\lambda + q = 0

Cases:

−p±p2−4q2⇒{(1) p2−4q>0⇒{λ1=−p+p2−4q2λ2=−p−p2−4q2(2) p2−4q=0⇒λ=−p2(3) p2−4q<0⇒{λ1=−p+ip2−4q2λ2=−p−ip2−4q2 \begin{align} \frac{-p \pm \sqrt{p^2-4q}}{2} \Rightarrow \begin{cases} \text{(1) } p^2-4q \gt 0 \Rightarrow \begin{cases} \lambda_1 = \frac{-p + \sqrt{p^2-4q}}{2} \\ \lambda_2 = \frac{-p - \sqrt{p^2-4q}}{2} \\ \end{cases} \\ \text{(2) } p^2-4q = 0 \Rightarrow \lambda = \frac{-p}{2} \\ \text{(3) } p^2-4q \lt 0 \Rightarrow \begin{cases} \lambda_1 = \frac{-p + i\sqrt{p^2-4q}}{2} \\ \lambda_2 = \frac{-p - i\sqrt{p^2-4q}}{2} \\ \end{cases} \end{cases} \end{align}

Example of case 1 (real roots):

y′′+4y′+3y=0{y(0)=1y′(0)=2 \begin{align} y\prime\prime + 4y\prime + 3y &= 0 \\ \begin{cases} y(0) &= 1 \\ y\prime(0) &= 2 \\ \end{cases} \end{align}

λ2+4λ+3=0(λ+1)(λ+3)=0⇒λ=−1,3⇒y1(x)=e−1x,y2(x)=e−3x \begin{align} \lambda^2 + 4\lambda + 3 &= 0 \\ \left(\lambda + 1 \right)\left(\lambda + 3\right) = 0 \\ \Rightarrow \lambda &= -1, 3 \\ \Rightarrow y_1(x) = e^{-1x}, y_2(x) = e^{-3x} \end{align}

General Solution: y(x)=C1e−x+C2e−3xy′(x)=−C1e−x+−3C2e−3x \begin{align} y(x) &= C_1e^{-x} + C_2e^{-3x} \\ y\prime(x) &= -C_1e^{-x} + -3C_2e^{-3x} \end{align} Specific Solution: {C1+C2=1−C1−3C2=2⇒−2C2=−1⇒C2=12⇒C1=12 \begin{align} &\begin{cases} C_1 + C_2 &= 1\\ -C_1 - 3C_2 &= 2 \\ \end{cases} \\ & \Rightarrow -2C_2 = -1 \\ & \Rightarrow C_2 = \frac{1}{2} \\ & \Rightarrow C_1 = \frac{1}{2} \\ \end{align}

Example for case 2 (repeated roots): y1(t)=eλty2(t)=?→Guess:y2(t)=teλt \begin{align} y_1(t) = e^{\lambda t} \\ y_2(t) = ? \\ \rightarrow \text{Guess:} y_2(t) = te^{^\lambda t} \end{align}

Check 0=(teλt)′′+p(teλt)′+qteλt=(λeλt+tλeλt)′+pλteλt+peλt+qteλt=λeλt+λeλt+tλ2eλt+pλteλt+peλt+qteλt=2λeλt+(λ2+pλ+q)(teλt)=???=(2λ+p)(eλt)=2(−p2)+p=0  ✓ \begin{align} 0 &= \left(te^{\lambda t}\right)\prime\prime + p\left(te^{\lambda t}\right)\prime + qte^{\lambda t} \\ &= \left(\lambda e^{\lambda t} + t\lambda e^{\lambda t}\right)\prime + p\lambda t e^{\lambda t} + pe^{\lambda t} + qte^{\lambda t} \\ &= \lambda e^{\lambda t} + \lambda e^{\lambda t} + t\lambda^2e^{\lambda t} + p\lambda t e^{\lambda t} + pe^{\lambda t} + qte^{\lambda t} \\ &= 2\lambda e^{\lambda t} + \left(\lambda^2 + p\lambda + q\right)\left(te^{\lambda t}\right) \\ &= ??? \\ &= \left(2\lambda + p\right)\left(e^{\lambda t}\right) \\ &= 2\left(\frac{-p}{2}\right) + p \\ &= 0 \, \, \checkmark \end{align} Now Check ∣y1y2y1′y2′∣=∣eλtteλtλeλteλt+λteλt∣=e2λt+λte2λt−λte2λt=e2λt≠0  ✓ \begin{align} &\left| \begin{array}{cc} y_1 & y_2 \\ y_1\prime & y_2\prime \\ \end{array} \right| \\ &= \left| \begin{array}{cc} e^{\lambda t} & te^{\lambda t} \\ \lambda e^{\lambda t} & e^{\lambda t} + \lambda te^{\lambda t} \\ \end{array} \right| \\ &= e^{2\lambda t} + \cancel{\lambda te^{2\lambda t} - \lambda te^{2\lambda t}} \\ &= e^{2\lambda t} \\ &\neq 0 \, \, \checkmark \end{align}

Example: y′′−6y′+9y=0IC={y(0)=1y′(0)=−1 \begin{align} y\prime\prime - 6y\prime + 9y &= 0 \\ \text{IC} &= \begin{cases} y(0) &= 1 \\ y\prime(0) &= -1 \end{cases} \\ \end{align} Step 1: Find general solution: 0=λ2−6λ+9=(λ−3)2⇒λ=3⇒y(t)=C1e3t+C2te3t \begin{align} 0 &= \lambda^2 - 6\lambda + 9 \\ &= \left(\lambda - 3\right)^2 \\ &\Rightarrow \lambda = 3 \\ &\Rightarrow y(t) = C_1e^{3t} + C_2te^{3t} \\ \end{align} Step 2: Plug in ICs y′(t)=3C1e3t+C2(e3t+3te3t){1=y(0)=C1−1=y′(0)=3C1+C2=3+C2⇒C2=−4 \begin{align} y\prime(t) = 3C_1e^{3t} + C_2\left(e^{3t}+3te^{3t}\right) \\ \begin{cases} 1 &= y(0) = C_1 \\ -1 &= y\prime(0) = 3C_1+C_2 = 3 + C_2 \Rightarrow C_2 = -4 \\ \end{cases} \end{align}

Case 3 (complex roots): λ=−p±i4q−p22=−p2±i4q−p22⇓eλt=e−p2⋅t±i⋅4q−p22⋅t=e−p2⋅t±ei⋅4q−p22⋅t=e−p2⋅t⋅±(cos⁡(4q−p22⋅t)±isin⁡(4q−p22⋅t))=???⇓y1(t)=e−p2⋅tcos⁡(4q−p22⋅t)y2(t)=e−p2⋅tsin⁡(4q−p22⋅t)⇒y(t)=C1e−p2⋅tcos⁡(4q−p22⋅t)+C2e−p2⋅tsin⁡(4q−p22⋅t) \begin{align} \lambda &= \frac{-p \pm i\sqrt{4q - p^2}}{2} \\ &=\frac{-p}{2} \pm i\frac{\sqrt{4q - p^2}}{2} \\ &\Big\Downarrow \\ e^{\lambda t} &= e^{\frac{-p}{2}\cdot t \pm i\cdot\frac{\sqrt{4q - p^2}}{2}\cdot t} \\ &= e^{\frac{-p}{2}\cdot t} \pm e^{i\cdot\frac{\sqrt{4q - p^2}}{2}\cdot t} \\ &= e^{\frac{-p}{2}\cdot t} \cdot \pm \left(\cos\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right) \pm i\sin\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right)\right) \\ &= ??? \\ &\Big\Downarrow \\ y_1(t) &= e^{\frac{-p}{2}\cdot t} \cos\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right) \\ y_2(t) &= e^{\frac{-p}{2}\cdot t} \sin\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right) \\ \Rightarrow y(t) &= C_1e^{\frac{-p}{2}\cdot t} \cos\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right) + C_2e^{\frac{-p}{2}\cdot t} \sin\left({\frac{\sqrt{4q - p^2}}{2}\cdot t}\right) \end{align}

Example: {y′′−4y′+5y=0y(0)=2y′(0)=−3λ2−4λ+5=0⇒λ=−4±16−202=−4±−42=−4±2i2=−2±i \begin{align} &\begin{cases} y\prime\prime - 4y\prime + 5y = 0 \\ y(0) = 2 \\ y\prime(0) = -3 \end{cases} \\ &\lambda^2 - 4\lambda + 5 = 0 \\ &\Rightarrow \lambda = \frac{-4 \pm \sqrt{16 - 20}}{2} = \frac{-4 \pm \sqrt{-4}}{2} = \frac{-4 \pm 2i}{2} = -2 \pm i \end{align} General Solution: y(t)=C1e2tcos⁡(t)+C2e2tsin⁡(t)y′(t)=C1(2e2tcos⁡(t)−e2tsin⁡(t))+C1(2e2tsin⁡(t)+e2tcos⁡(t)) \begin{align} y(t) &= C_1e^{2t}\cos(t) + C_2e^{2t}\sin(t) \\ y\prime(t) &= C_1\left(2e^{2t}\cos(t) - e^{2t}\sin(t)\right) + C_1\left(2e^{2t}\sin(t) + e^{2t}\cos(t)\right) \end{align} Solve IVP: 2=y(0)=C1−3=y′(0)=C1(2)+C2(2)=2C1+C2=4+C2=−3⇒C2=−7y(t)=2e2tcos⁡(t)−7e2tsin⁡(t) \begin{align} 2 &= y(0) = C_1 \\ -3 &= y\prime(0) = C_1(2)+C_2(2) = 2C_1 + C_2 = 4 + C_2 = -3 \Rightarrow C_2 = -7 \\ \\ y(t) &= 2e^{2t}\cos(t) - 7e^{2t}\sin(t) \\ \end{align} .

$$

Higher order (n>3n \gt 3) linear, homogeneous, constant coefficient eqns

any(n)+an−1y(n−1)+⋯+a2y′′+a1y′a0y=0 a_ny^{(n)} + a_{n - 1}y^{(n - 1)} + \dots + a_2y\prime\prime + a_1y\prime a_0y = 0

Characteristic Equation:

λnany(n)+λn−1an−1ny(n−1)+⋯+a1λ+a0=0 \lambda ^n a_n y^{(n)} + \lambda ^{n-1} a_{n-1}n y^{(n - 1)} + \dots + a_1\lambda + a_0 = 0

Roots: λ1,…,λn\lambda_1, \dots, \lambda_n

General Idea:

Each distinct real root λi⇔yi(t)eλit\lambda_i \Leftrightarrow y_i(t)e^{\lambda_i t}
Each pair of distinct complex roots:

λ=a±bi⇔{y+(t)=eatcos⁡(bt)y−(t)=eatsin⁡(bt) \begin{align} \lambda = a \pm bi \Leftrightarrow \begin{cases} y_+(t) &= e^{at}\cos(bt) \\ y_-(t) &= e^{at}\sin(bt) \\ \end{cases} \end{align}

If any repetition happens (kk times):

{y1=eλt,teλt,…,tk−1eλty±=eatcos⁡(bt),eatsin⁡(bt),…,y~±=tk−1eatcos⁡(bt),tk−1eatsin⁡(bt) \begin{align} \begin{cases} y_1 &= e^{\lambda t}, te^{\lambda t}, \dots, t^{k-1}e^{\lambda t} \\ y_\pm &= e^{at}\cos(bt), e^{at}\sin(bt), \dots, \widetilde{y}_\pm = t^{k-1}e^{at}\cos(bt), t^{k-1}e^{at}\sin(bt) \end{cases} \end{align}

This page is not yet finished!