MHB Linear differential equation of first order in EXP(C)

mathmari
Gold Member
MHB
Messages
4,984
Reaction score
7
Hey! :o

I want to check if there is a solution of a linear differential equation of first order in the ring of exponential sums $\text{EXP}(\mathbb{C})$. I have done the following: The general linear differential equation of first order is $$ax'(z)+bx(z)=y(z) \tag{*}$$
where $x,y \in \text{EXP}(\mathbb{C})$.

We have that $$y(z)=\sum_{i=1}^N \alpha_i e^{\mu_i z}, \ \ x(z)=\sum_{i=1}^N \beta_i e^{k_i z} \ \ \text{ so } \ \ x'(z)=\sum_{i=1}^N \beta_i k_i e^{k_i z}$$

$$(*) \Rightarrow a\sum_{i=1}^N \beta_i k_i e^{k_i z}+b\sum_{i=1}^N \beta_i e^{k_i z}= \sum_{i=1}^N \alpha_i e^{\mu_i z} \Rightarrow \sum_{i=1}^N \left [a\beta_i k_i +b\beta_i \right ]e^{k_i z}=\sum_{i=1}^N \alpha_i e^{\mu_i z}$$ For $k_i=\mu_i$ we have the following:

$$\sum_{i=1}^N \left [\beta_i (a\mu_i +b) \right ]e^{\mu_i z}=\sum_{i=1}^N \alpha_i e^{\mu_i z} \Rightarrow \sum_{i=1}^N \left [(a \mu_i +b)\beta_i -\alpha_i\right ]e^{\mu_i z}=0$$ 1. If for at least one $i$ with $\alpha_i\ne0$ we have $a\mu_i+b=0$, there is no solution.

2. If for at least one $i$ with $\alpha_i = 0$ we have $a\mu_i+b \neq 0$, we have that $\beta_i=0$.

3. If for all $i$ it holds that $\alpha_i = 0 \land a\mu_i+b = 0$, then thee are infinitely many solutions.

4. If for all $i$ it holds that $\alpha_i\ne0 \land a\mu_i+b\ne0$, then there is exactly one solution, with $\beta_i=\frac{\alpha_i}{a\mu_i+b}$.
Is everything correct? Have I forgotten a case? (Wondering)
 
Physics news on Phys.org
I did it also as follows by taking cases:
  • $\textbf{Case 1.}$
    $a=0, b \neq 0$

    Then we have $bx(z)=y(z)$.
    So, the solution is $$x(z)=\frac{1}{b}y(z) \in \text{EXP}(\mathbb{C})$$
  • $\textbf{Case 2.}$
    $a\neq 0, b=0$

    Then we have $ax'(z)=y(z)$.

    We have that $\displaystyle{x'(z)=\frac{1}{a}y(z)}$, where $\displaystyle{y(z)=\sum_{i=1}^N \alpha_i e^{\mu_i z}}$, so $\displaystyle{x'(z)=\frac{1}{a}\sum_{i=1}^N \alpha_i e^{\mu_i z}}$.
    • If $\exists j$ such that $\mu_j =0 \land \alpha_j \neq 0$ then $\displaystyle{x'(z)=\frac{1}{a}\sum_{i=1}^{j-1} \alpha_i e^{\mu_i z}+\frac{\alpha_j}{a}+\frac{1}{a}\sum_{i=j+1}^N \alpha_i e^{\mu_i z}}$ then $\displaystyle{x(z)=\frac{1}{a}\sum_{i=1}^{j-1} \frac{\alpha_i}{\mu_i} e^{\mu_i z}+\frac{\alpha_j}{a}z+\frac{1}{a}\sum_{i=j+1}^N \frac{\alpha_i}{\mu_i} e^{\mu_i z}} \notin \textbf{EXP}{(\mathbb{C})}$. So, there is no solution.
    • If $\mu_i \neq 0$ for $i$ for which $\alpha_i \neq 0$ then the solution is $\displaystyle{x(z)=\frac{1}{a}\sum_{i=1}^N \frac{\alpha_i}{\mu_i} e^{\mu_i z}}\in \text{EXP}(\mathbb{C})$.
  • $\textbf{Case 3.}$
    $a=0, b=0$

    Then we have $0=y(z)$.
    • If $y(z)=0$ then we have infinitely many solutions.
    • If $y(z) \neq 0$ then we have no solution.
  • $\textbf{Case 4.}$
    $a \neq 0, b \neq 0$

    Then we have $$ax'(z)+bx(z)=y(z) \Rightarrow x'(z)+\frac{b}{a} x(z)=\frac{1}{a}y(z) \ \ \ \ \ (*)$$

    We have that $$y(z)=\sum_{i=0}^N \alpha_i e^{\mu_i z}, x(z)=\sum_{i=0}^N \beta_i e^{k_i z}=\beta_0 +\sum_{i=1}^N \beta_i e^{k_i z}, also x'(z)=\sum_{i=1}^N \beta_i k_i e^{k_i z} $$

    $$(*) \Rightarrow \sum_{i=1}^N \beta_i k_i e^{k_i z}+\frac{b}{a} \left [\beta_0 +\sum_{i=1}^N \beta_i e^{k_i z}\right ]=\frac{1}{a} \left [\sum_{i=0}^N \alpha_i e^{\mu_i z}\right ] \\
    \Rightarrow \frac{b}{a}\beta_0 +\sum_{i=1}^N \left [\beta_i k_i +\frac{b}{a}\beta_i \right ]e^{k_i z}=\frac{1}{a}\alpha_0+\sum_{i=1}^N \alpha_i e^{\mu_i z} $$

    So, the following must hold:

    $$\left\{\begin{matrix}
    \frac{b}{a}\beta_0=\frac{1}{a}\alpha_0\\
    \beta_i k_i+\frac{b}{a}\beta_i=\alpha_i, \ \ i=1, \dots , N\\
    k_i=\mu_i, \ \ i=1, \dots , N
    \end{matrix}\right. \Rightarrow \left\{\begin{matrix}
    \beta_0=\frac{\alpha_0}{b}\\
    \beta_i \mu_i+\frac{b}{a}\beta_i=\alpha_i, \ \ i=1, \dots , N\\
    k_i=\mu_i, \ \ i=1, \dots , N
    \end{matrix}\right. \Rightarrow \left\{\begin{matrix}
    \beta_0=\frac{\alpha_0}{b}\\
    \beta_i \left (\mu_i+\frac{b}{a}\right )=\alpha_i, \ \ i=1, \dots , N\\
    k_i=\mu_i, \ \ i=1, \dots , N
    \end{matrix}\right. $$
    • If $\displaystyle{\mu_i \neq -\frac{b}{a}}$, then the solution is $\displaystyle{x(z)=\frac{\alpha_0}{b}+\sum_{i=1}^N \frac{\alpha_i}{\mu_i+\frac{b}{a}}e^{\mu_i z}}$.
    • If $\displaystyle{\mu_i=-\frac{b}{a}}$ then if $\alpha_i=0$ then there are infinitely many solutions, but if $\alpha_i \neq 0$ then there is no solution.
Is this correct? Have I taken into consideration all the possible cases? (Wondering)
 
Hey mathmari! (Wave)

It seems you assume that $\{ e^{\mu_i z} \}$ forms an independent basis as long as the $\mu_i$ are distinct.
However, that is not the case.
For instance $1\cdot e^{i\pi/2 z} = -1 \cdot e^{-i\pi/2 z}$.
(Worried)
 
Last edited:
I like Serena said:
It seems you assume that $\{ e^{\mu_i z} \}$ forms an independent basis as long as the $\mu_i$ are distinct.
However, that is not the case.
For instance $1\cdot e^{i\pi/2} = -1 \cdot e^{-i\pi/2}$.
And $2e^{1+i} + 3e^{2+i} = (2e+3e^2)e^i$.
(Worried)

Ahaa... Ok... (Thinking)
And what could I change? (Wondering)
 
I like Serena said:
Hey mathmari! (Wave)

It seems you assume that $\{ e^{\mu_i z} \}$ forms an independent basis as long as the $\mu_i$ are distinct.
However, that is not the case.
For instance $1\cdot e^{i\pi/2 z} = -1 \cdot e^{-i\pi/2 z}$.
(Worried)

mathmari said:
Ahaa... Ok... (Thinking)
And what could I change? (Wondering)

Nevermind, I think $\{ e^{\mu_i z} :i=0..N, \text{ distinct } \mu_i\in \mathbb C \}$ does form a basis.
So I think your reasoning is correct, although you should probably mention that. (Thinking)
 
I like Serena said:
Nevermind, I think $\{ e^{\mu_i z} :i=0..N, \text{ distinct } \mu_i\in \mathbb C \}$ does form a basis.
So I think your reasoning is correct, although you should probably mention that. (Thinking)

Do you mean that I should mention that $\{ e^{\mu_i z} :i=0..N, \text{ distinct } \mu_i\in \mathbb C \}$ forms a basis? So, in the ring $\text{EXP}(\mathbb{C})$ the linear differential equation of first order has always a solution except at the following cases:
  • $a \neq 0 \land b =0 \land \mu_j =0 \land \alpha_j \neq 0$
  • $a=0 \land b=0 \land y \neq 0$
  • $a \neq 0 \land b \neq 0 \land \mu_i =-\frac{b}{a} \land \alpha_i \neq 0$

Is this correct? (Wondering)
 
Thread 'Direction Fields and Isoclines'
I sketched the isoclines for $$ m=-1,0,1,2 $$. Since both $$ \frac{dy}{dx} $$ and $$ D_{y} \frac{dy}{dx} $$ are continuous on the square region R defined by $$ -4\leq x \leq 4, -4 \leq y \leq 4 $$ the existence and uniqueness theorem guarantees that if we pick a point in the interior that lies on an isocline there will be a unique differentiable function (solution) passing through that point. I understand that a solution exists but I unsure how to actually sketch it. For example, consider a...
Back
Top