\[\frac{dx}{dt} = \frac{d}{dt}\left( -e^t \right) = -e^t = -y\] \[\frac{dy}{dt} = \frac{d}{dt}\left( e^t \right) = e^t = -x\]
The only point at which \(\frac{dx}{dt} = \frac{dy}{dt} = 0\) is \((2, 1)\).
Integrating the equations, \[x = \int \left(\frac{dx}{dt}\right)dt = -\int \left( y - 1 \right) dt = -t\left( y - 1 \right)\] As \(t \to \infty\), \(x(t) \to \infty\) if \(y_0 < 1\) or \(x(t) \to -\infty\) if \(y_0 > 1\).
\[y = \int \left(\frac{dy}{dt}\right)dt = \int \left( x - 2 \right) dt = t\left( x - 2 \right)\]
As \(t \to \infty\), \(y(t) \to \infty\) if \(x_0 > 2\) or \(y(t) \to -\infty\) if \(x_0 < 2\).
Since \((x,y) \to (\pm \infty, \pm \infty)\) for \((x_0, y_0) \notin (2,1)\), the point is unstable.
\[\frac{dt}{dx} = \left( \frac{dx}{dt} \right)^{-1} = \frac{1}{(a - by)x}\] \[\frac{dy}{dx} = \frac{dy}{dt} \times \frac{dt}{dx} = \frac{(m - nx)y}{(a - by)x}\]
\[\frac{1}{y} \frac{dy}{dx} = \frac{1}{x} \frac{m - nx}{a - by} \longrightarrow \frac{ady}{y} - bdy = \frac{mdx}{x} - ndx\]
\[\int \frac{ady}{y} - \int bdy = \int \frac{mdx}{x} - \int ndx \longrightarrow a\ln y - by + K_1 = m\ln x - nx + K_2\]
\[y^a e^{-by} = K x^m e^{-nx}\]
\[\frac{d}{dy} f(y) = ay^{a-1} \times e^{-by} + (-be^{-by}) \times y^a = y^{a-1}e^{-by} \left( a - by\right)\]
Setting \(f^\prime = 0\) to find the maximum: \[y^{a-1}e^{-by} \left( a - by\right) = 0 \longrightarrow a - by = 0 \longrightarrow y = \frac{a}{b}\] At \(y = \frac{a}{b}\), \[f(\frac{a}{b}) = \left( \frac{a}{b} \right)^a e^{-a} = \left( \frac{a}{eb} \right)^a\]
\[\frac{d}{dx} g(x) = mx^{m-1} \times e^{-nx} + (-ne^{-ny}) \times x^m = x^{m-1}e^{-nx} \left( m - nx\right)\]
Setting \(g^\prime = 0\) to find the maximum: \[x^{m-1}e^{-ny} \left( m - nx\right) = 0 \longrightarrow m - nx = 0 \longrightarrow x = \frac{m}{n}\] At \(x = \frac{m}{n}\), \[g(\frac{m}{n}) = \left( \frac{m}{n} \right)^m e^{-m} = \left( \frac{m}{en} \right)^m\]
The constant \(K\) from part b is equivalent to \[K = \frac{e^{K_1}}{e^{K_2}}\] Where \(K_1\) and \(K_2\) are the integration constants from the left and right sides, respectively. As \((x, y) \to (m/n, a/b)\), it is clear from the provided equations that \(dx/dt, dy/dt \to 0\).
Using this information, the constants can be solved for: \[a \ln y - by + K_1 = 0 \longrightarrow e^{K_1} = \frac{y^a}{e^{by}}\]
\[m \ln x - ny + K_2 = 0 \longrightarrow e^{K_2} = \frac{x^m}{e^{nx}}\]
Thus, as \((x, y) \to (m/n, a/b)\), \[\lim \left[ \left( \frac{y^a}{e^{by}}\right) \left( \frac{e^{nx}}{x^b} \right)\right] = \frac{e^{K_1}}{e^{K_2}} = K\]
For \(y_0 > a/b\), \(f(y_0) < M_y\). This implies \[\frac{M_y}{M_x} \left( \frac{x^m}{e^{nx}} \right) = \frac{y_0^a}{e^{by_0}} <M_y \longrightarrow \frac{x^m}{e^{nx}} < M_x\]
Per the plot of \(g(x)\), there is a single value \(x > m/n\) satisfying this condition. Thus, there exists a unique trajectory approaching \((m/n, a/b)\) from above.
From above, the derivative is given by \[f^\prime(y) = y^{a-1}e^{-by} \left( a - by\right)\]
Setting \(f^\prime = 0\) to find the maximum: \[y^{a-1}e^{-by} \left( a - by\right) = 0 \longrightarrow a - by = 0 \longrightarrow y = \frac{a}{b}\]
Since \(y = \frac{a}{b}\) is the only root of the equation \(f^\prime(y) = 0\), it is the only critical point of \(f(y)\).
The second derivative is given by \[f^{\prime\prime}(y) = a(a-1)y^{a-2} \times e^{-by} +ay^{a-1} \times -be^{-by} + b^2e^{-by} \times y^a + ay^{a - 1} \times -be^{-by}\] \[f^{\prime\prime}(y) = y^{a - 2} e^{-by} \left(a^2 - 2aby - a + b^2 y^2\right)\]
At the critical point, this is \[f^{\prime\prime}(\frac{a}{b}) = \left( \frac{a}{b} \right)^{a - 2} e^{-a} \left(a^2 - 2a^2 - a + a^2\right) = -\frac{a^{a-1}}{e^a\ b^{a-2}}\] This second derivative, indicating that the slope is decreasing. Thus, the critical point is a maximum.
The limit of the equation \(\lim_{y \to \infty} f(y)\) does not provide a useful limit. Using L’Hopital’s rule to take the derivative of the numerator and denominator, however, provides useful insight: \[\lim_{y \to \infty} \frac{y^a}{e^{by}} = \lim_{y \to \infty} \frac{\frac{d^{n}}{dy^n}y^a}{\frac{d^{n}}{dy^n}e^{by}} = \lim_{y \to \infty} \frac{\prod_{i}^n \{a - i + 1\}y^{a-n}}{b^ne^{by}}\]
From this, it is clear that the denominator will grow much larger than the numerator; thus \[\lim_{y \to \infty} f(y) = 0\]
# function to implement Euler's method
euler <- function(dxdt, dydt, t0, x0, y0, delta_t, n) {
# initial conditions
t <- t0
x <- x0
y <- y0
# run loop to get approximations
for (i in 1:n) {
t_i <- t[length(t)] + delta_t
x_i <- x[length(x)] + f(x[length(x)], y[length(y)]) * delta_t
y_i <- y[length(y)] + g(x[length(x)], y[length(y)]) * delta_t
t <- c(t, t_i)
x <- c(x, x_i)
y <- c(y, y_i)
}
# return estimate
return(data.frame(t, x, y))
}
# functions for dx/dt and dy/dt
f <- function(x, y) {return(2 * x + 3 * y)}
g <- function(x, y) {return(3 * x + 2 * y)}
# get results for delta_t and 0.5*delta_t
est_a <- euler(f, g, 0, 1, 1, 1/4, 3)
est_b <- euler(f, g, 0, 1, 1, 1/8, 6)
# calculate analytical solutions
soln <- data.frame(t = est_b$t,
x = 0.5 * exp(-est_b$t) + 0.5 * exp(5 * est_b$t),
y = -0.5 * exp(-est_b$t) + 0.5 * exp(5 * est_b$t))
The results of the above estimates (Euler’s method using \(\Delta t\), Euler’s method using \(\Delta t/2\), and the analytical solution) are presented below in tabular format and graphically on the following page.
| \(t\) | \(x_{E(\Delta t)}\) | \(y_{E(\Delta t)}\) | \(x_{E(\Delta t / 2)}\) | \(y_{E(\Delta t / 2)}\) | \(x_A\) | \(y_A\) |
|---|---|---|---|---|---|---|
| 0 | 1 | 1 | 1 | 1 | 1 | 0 |
| 0.125 | NA | NA | 1.625 | 1.625 | 1.375 | 0.4929 |
| 0.25 | 2.25 | 2.25 | 2.641 | 2.641 | 2.135 | 1.356 |
| 0.375 | NA | NA | 4.291 | 4.291 | 3.604 | 2.917 |
| 0.5 | 5.062 | 5.062 | 6.973 | 6.973 | 6.395 | 5.788 |
| 0.625 | NA | NA | 11.33 | 11.33 | 11.65 | 11.11 |
| 0.75 | 11.39 | 11.39 | 18.41 | 18.41 | 21.5 | 21.02 |