diff --git a/src/differential_equations_1.md b/src/differential_equations_1.md index 21545ba4f70056d2efe3ede6e7b7c545d5997987..261674a8229cb7e4836781754961d08606e37216 100644 --- a/src/differential_equations_1.md +++ b/src/differential_equations_1.md @@ -369,7 +369,6 @@ $$**\Phi**(t) \dot{**c**}(t) = **b**(t). $$ In order to cancel terms in the previous line we made use of the fact that $**\Phi**(t)$ solves the homogeneous equation $\dot{**\Phi**} = **A** **\Phi**$. - By way of inverting and integrating, we can write an equation for the coefficient vector $**c**(t)$ @@ -380,9 +379,94 @@ the particular solution, $$**\psi**(t)= **\Phi**(t) \cdot \int **\Phi**^{-1}(t) **b**(t) dt .$$ - - - +!!! check "Example: Inhomogeneous first order linear differential equation" + + The technique for solving a system of inhomogeneous equations also works for a + single inhomogeneous equation. Let us apply the technique to the equation + + $$ \dot{x} = \lambda x + a. $$ + + In this particular inhomogenous equaiton, the function $g(t)=a$. As discussed in + an earlier example, the solution to the homogenous equation is + $c e^{\lambda t}$. Hence we define $\phi(t)=e^{\lambda t}$ and make the ansatz + + $$\psi(t) = c(t) e^{\lambda t}. $$ + + Solving for $c(t)$ results in + + $$c(t) = \int e^{- \lambda t} a dt$$ + $$c(t) = \frac{- a }{\lambda} e^{- \lambda t} $$ + + Overall then, the solution (which can be easily verified by substitution) is + + $$\psi(t) = - \frac{a}{\lambda}. $$ + +## Homogeneous linear system with constant coefficients ## + +The type of equaiton under consideration in this section looks like + +$$ \dot{**x**}(t) = **A** **x**(t),$$ + +where throughout the section $**A**$ will be a constant matrix. It is possible +to define a formal solution using the *matrix exponential*, +$**x**(t) = e^{**A** t}$. + +!!! info "Definition: Matrix Exponential" + + Before defining the matrix exponential, recall the definition of the regular + exponential function in terms of Taylor series, + + $$e^{x} = \overset{\infty}{\underset{n=0}{\Sigma}} \frac{x^n}{n!},$$ + + in which it is agreed that $0!=1$. The matrix exponential is defined in + exactly the same way, only now instead of taking powers of a number or + function, powers of a matrix are calculated + + $$e^{**A**} = \overset{\infty}{\underset{n=0}{\Sigma}} \frac{{**A**}^n}{n!}.$$ + + It is important to use caution when translating the properties of the normal + exponential function over to the matrix exponential, because not all of the + regular properties hold generally. In particular, + + $$e^{**X** + **Y**} \neq e^{**X**} e^{**Y**},$$ + + unless it happens that + + $$**XY** = **YX**.$$ + + The necessary condition for this property to hold, stated on the previous + line, is called *commutativity*. Recall that in general, matrices are not + commutative so such a condition is only met for particular choises of + matrices. The property of *non-commutativity* (what happens when the + condition is not met) is of central importance in the mathematical + structure of quantum mechanics. For example, mathematically, + non-commutativity is responsible for the Heisenberg uncertainty relations. + + On the other hand, one property that does hold, is that $e^{- **A** t}$ is + the inverse of the matric exponential of $**A**$. + + Furthermore, it is possible to derive the derivative of the matrix + exponential by making use of the Taylor series formulation, + + $$\frac{d}{dt} e^{**A** t} = \frac{d}{dt} \overset{\infty}{\underset{n=0}{\Sigma}} \frac{(**A** t)^n}{n!} $$ + $$\frac{d}{dt} e^{**A** t} = \overset{\infty}{\underset{n=0}{\Sigma}} \frac{1}{n!} \frac{d}{dt} (**A** t)^n$$ + $$\frac{d}{dt} e^{**A** t} = \overset{\infty}{\underset{n=0}{\Sigma}} \frac{n **A**}{n!}(**A** t)^{n-1}$$ + $$\frac{d}{dt} e^{**A** t} = \overset{\infty}{\underset{n=1}{\Sigma}} \frac{**A**}{(n-1)!}(**A** t)^{n-1}$$ + $$\frac{d}{dt} e^{**A** t} = \overset{\infty}{\underset{n=0}{\Sigma}} \frac{**A**}{n!}(**A** t)^n$$ + $$\frac{d}{dt} e^{**A** t} = **A** e^{**A** t}.$$ + +Armed with the matrix exponential and it's derivative, +$\frac{d}{dt} e^{**A** t} = **A** e^{**A** t}$, it is simple to verify that +the matrix exponential solves the differential equation. Properties of this +solution are: + +1. The columns of $e^{**A** t}$ form a basis for the solution space. +2. Accounting for initial conditions, the full solution of the equation is + $\dot{**x**}(t) = e^{**A** t} {**x**}_{0}$, with initial condition + $**x**(0) = e^{**A** 0}{**x**}_0 = \mathbbm{1} {**x**}_{0} = {**x**}_{0}$. + +Next we will discuss how to determine a solution in general, going beyond the +formal solution just presented.