2
$\begingroup$

Consider the vector differential equations \begin{equation} \mathbf{x}^{\prime}=\mathbf{A}(t)\cdot\mathbf{x}\tag{1} \end{equation} and \begin{equation} \mathbf{y}^{\prime}=-\mathbf{A}^{\ast}(t)\cdot\mathbf{y},\tag{2} \end{equation} where $\mathbf{A}^{\ast}$ is the complex conjugate transpose of $\mathbf{A}$ and $\mathbf{x},\mathbf{y}$ are column vectors. It is well-known that (1) and (2) are said to be adjoint to one another. Further, we know that if $\mathbf{x}$ and $\mathbf{y}$ are solutions of (1) and (2), respectively, then \begin{equation} \mathbf{y}^{\ast}\cdot\mathbf{x}=\text{constant}.\notag \end{equation}

Now, consider the higher-order (scalar) differential equations \begin{equation} \sum_{i=1}^{n}p_{i}(t)x^{(i)}(t)=0,\tag{3} \end{equation} where $p_{n}(t)\neq0$, and \begin{equation} \sum_{i=1}^{n}(-1)^{(i)}[p_{i}y]^{(i)}(t)=0.\tag{4} \end{equation} Also, (3) and (4) are said to be adjoint to one another. Further, if $x$ and $y$ are solutions of (3) and (4), respectively, then (see [1, (8.17) on pp. 67]) \begin{equation} \sum_{i=0}^{n}\sum_{j=0}^{i-1}(-1)^{j}x^{(i-j-1)}(t)[p_{i}z]^{(j)}(t)=\text{constant}.\label{hmfeq}\tag{*} \end{equation}

The inner sum in \eqref{hmfeq} resembles the matrix multiplication formula. So, recognizing the similarities between systems and scalar equations, is it possible to obtain the result for higher-order equations by transforming them into vector equations? I could not establish any bridge here. I am experiencing problems in transforming (4) into a useful matrix representation.

References

[1]. P. Hartman, Ordinary Differential Equations, SIAM, 2002.

1 Answers 1

0

Okay, I believe, I have constructed the bridge between the adjoint of a vector equation and the adjoint of a scalar equation. I will explain it step by step here for those who might be interested.

Higher-Order Two-Term Scalar Equations

Now, consider the following higher-order two-term scalar differential equation \begin{equation} x^{(n)}(t)+p(t)x(t)=0,\label{hottceeq1}\tag{1} \end{equation} where $p$ is a complex-valued continuous function. Then, the matrix representation for \eqref{hottceeq1} is \begin{equation} \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right)^{\prime} = \left( \begin{array}{cccc} &1&&\\ &&\ddots&\\ &&&1\\ -p(t)&&& \end{array} \right) \cdot \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right).\label{hottceeq2}\tag{2} \end{equation} Thus, we obtain \begin{equation} (-1) \left( \begin{array}{cccc} &1&&\\ &&\ddots&\\ &&&1\\ -p(t)&&& \end{array} \right)^{\ast} = \left( \begin{array}{cccc} &&&\overline{p}(t)\\ (-1)&&&\\ &\ddots&&\\ &&(-1)& \end{array} \right)\notag \end{equation} and \begin{equation} \left( \begin{array}{c} -y^{(n-1)}\\ y^{(n-2)}\\ \vdots\\ (-1)^{n}y \end{array} \right)^{\prime} = \left( \begin{array}{cccc} &&&\overline{p}(t)\\ (-1)&&&\\ &\ddots&&\\ &&(-1)& \end{array} \right) \cdot \left( \begin{array}{c} -y^{(n-1)}\\ y^{(n-2)}\\ \vdots\\ (-1)^{n}y \end{array} \right),\label{hottceeq3}\tag{3} \end{equation} where we have constructed the unknown matrix from bottom to the top. This system gives us the adjoint equation \begin{equation} (-1)^{n}y^{(n)}(t)+\overline{p}(t)y(t)=0.\notag \end{equation} Note that, if $n=\text{even}$, then both \eqref{hottceeq2} and \eqref{hottceeq3} represent \eqref{hottceeq1}, i.e., \eqref{hottceeq1} is self-adjoint. Further, using the definition of the inner product in the first post, we get \begin{equation} \left\langle\left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right), \left( \begin{array}{c} -y^{(n-1)}\\ y^{(n-2)}\\ \vdots\\ (-1)^{n}y \end{array} \right)\right\rangle =\sum_{j=0}^{n-1}(-1)^{n-j}x^{(n-1-j)}\overline{y}^{(j)}=\text{constant}.\notag \end{equation}

Higher-Order Autonomous Scalar Equations

Next, consider the \begin{equation} x^{(n)}(t)+p_{1}x^{(n-1)}(t)+\cdots+p_{n}x(t)=0,\label{hoaseeq1}\tag{4} \end{equation} where $p_{1},p_{2},\cdots,p_{n}$ are complex numbers. Then, the matrix representation for \eqref{hoaseeq1} is \begin{equation} \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right)^{\prime} = \left( \begin{array}{cccc} &1&&\\ &&\ddots&\\ &&&1\\ -p_{n}&-p_{n-1}&\cdots&-p_{1} \end{array} \right) \cdot \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right).\notag \end{equation} On the other hand, by using the adjoint coefficient matrix, we form the differential system \begin{equation} \begin{aligned}[] &\left( \begin{array}{c} (-1)^{n-1}y^{(n-1)}+(-1)^{n-2}\overline{p}_{1}y^{(n-2)}+\cdots+\overline{p}_{n-1}y\\ (-1)^{n-2}y^{(n-2)}+(-1)^{n-3}\overline{p}_{2}y^{(n-3)}+\cdots+\overline{p}_{n-2}y\\ \vdots\\ y \end{array} \right)^{\prime}\\ =& \left( \begin{array}{cccc} &&&\overline{p}_{n}\\ (-1)&&&\overline{p}_{n-1}\\ &\ddots&&\vdots\\ &&(-1)&\overline{p}_{1} \end{array} \right) \cdot \left( \begin{array}{c} (-1)^{n-1}y^{(n-1)}+(-1)^{n-2}\overline{p}_{1}y^{(n-2)}+\cdots+\overline{p}_{n-1}y\\ (-1)^{n-2}y^{(n-2)}+(-1)^{n-3}\overline{p}_{2}y^{(n-3)}+\cdots+\overline{p}_{n-2}y\\ \vdots\\ y \end{array} \right), \end{aligned} \notag \end{equation} which gives the scalar equation \begin{equation} (-1)^{n}y^{(n)}(t)+(-1)^{n-1}\overline{p}_{1}y^{(n-1)}(t)+\cdots+\overline{p}_{n}y(t)=0.\notag \end{equation} Further, putting $p_{0}:=1$ for simplicity, we have \begin{equation} \left\langle\left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right), \left( \begin{array}{c} \begin{array}{c} \sum_{j=0}^{n-1}(-1)^{j}\overline{p}_{n-1-j}y^{(j)}\\ \sum_{j=0}^{n-2}(-1)^{j}\overline{p}_{n-2-j}y^{(j)}\\ \vdots\\ y \end{array} \end{array} \right)\right\rangle =\sum_{k=0}^{n-1}\sum_{j=0}^{k}(-1)^{j}p_{k-j}x^{(n-1-k)}\overline{y}^{(j)}=\text{constant}.\notag \end{equation}

Higher-Order Scalar Equations with Variable Coefficients

Finally, we consider \begin{equation} x^{(n)}(t)+p_{1}(t)x^{(n-1)}(t)+\cdots+p_{n}(t)x(t)=0,\notag \end{equation} where $p_{i}(t)$ ($i=1,2,\cdots,n$) is complex-valued and is $i$ times continuously differentiable function. Thus, the matrix representation is \begin{equation} \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right)^{\prime} = \left( \begin{array}{cccc} &1&&\\ &&\ddots&\\ &&&1\\ -p_{n}(t)&-p_{n-1}(t)&\cdots&-p_{1}(t) \end{array} \right) \cdot \left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right).\notag \end{equation} Then, the associated adjoint matrix is \begin{equation} - \left( \begin{array}{cccc} &1&&\\ &&\ddots&\\ &&&1\\ -p_{n}(t)&-p_{n-1}(t)&\cdots&-p_{1}(t) \end{array} \right)^{\ast} = \left( \begin{array}{cccc} &&&\overline{p}_{n}(t)\\ (-1)&&&\overline{p}_{n-1}(t)\\ &\ddots&&\vdots\\ &&(-1)&\overline{p}_{1}(t) \end{array} \right),\notag \end{equation} which yields the system \begin{equation} \begin{aligned}[] &\left( \begin{array}{c} \sum_{j=0}^{n-1}(-1)^{j}[p_{n-1-j}\overline{y}]^{(j)}\\ \sum_{j=0}^{n-2}(-1)^{j}[p_{n-2-j}\overline{y}]^{(j)}\\ \vdots\\ \overline{y} \end{array} \right)^{\prime}\\ &= \left( \begin{array}{cccc} &&&\overline{p}_{n}(t)\\ (-1)&&&\overline{p}_{n-1}(t)\\ &\ddots&&\vdots\\ &&(-1)&\overline{p}_{1}(t) \end{array} \right) \cdot \left( \begin{array}{c} \sum_{j=0}^{n-1}(-1)^{j}[p_{n-1-j}\overline{y}]^{(j)}\\ \sum_{j=0}^{n-2}(-1)^{j}[p_{n-2-j}\overline{y}]^{(j)}\\ \vdots\\ \overline{y} \end{array} \right), \end{aligned}\notag \end{equation} where we put $p_{0}(t):\equiv1$ for simplicity. Transforming this into the differential equation, we get \begin{equation} (-1)^{n}y^{(n)}(t)+(-1)^{n-1}[\overline{p}_{1}y]^{(n-1)}(t)+\cdots+[\overline{p}_{n-1}y]^{\prime}(t)+\overline{p}_{n}(t)y(t)=0.\notag \end{equation} Using the inner product in the first post gives us \begin{equation} \left\langle\left( \begin{array}{c} x\\ x^{\prime}\\ \vdots\\ x^{(n-1)} \end{array} \right), \left( \begin{array}{c} \begin{array}{c} \sum_{j=0}^{n-1}(-1)^{j}[p_{n-1-j}\overline{y}]^{(j)}\\ \sum_{j=0}^{n-2}(-1)^{j}[p_{n-2-j}\overline{y}]^{(j)}\\ \vdots\\ \overline{y} \end{array} \end{array} \right)\right\rangle =\sum_{k=0}^{n-1}\sum_{j=0}^{k}(-1)^{j}x^{(n-1-k)}[p_{k-j}\overline{y}]^{(j)}=\text{constant},\label{finaleq}\tag{#} \end{equation} which is the desired identity.

I believe that \eqref{finaleq} and (*) in the first post are equivalent.