We'll imagine that we care only about non-negative values of the state variables $x$ and $y$. Perhaps $x$ and $y$ represent concentrations or populations sizes for which negative values don't make sense. Hence, in the below phase plane, we'll plot just the first quadrant.
-
We'll begin by determining where the variables are not moving, i.e., the nullclines and the equilibria.
The $x$-nullcline is the set of points $(x,y)$ where $x$ is not changing, i.e., where $\diff{x}{t}=0$. (The $x$-nullcline will typically be composed of one or more curves in the phase plane.) Write the equation for the $x$-nullcline in terms of $x$ and $y$.
This curve is a
with slope
and $y$-intercept
. Plot this curve on the above phase plane and label it as either “$x$-nullcline” or “$\diff{x}{t}=0$”. (Pencil is recommended unless you draw quite accurately. The curve needs to be sufficiently accurate so that it intersects the $y$-nullcline in the correct places, as discussed below.)
-
The $y$-nullcline is the set of points $(x,y)$ where $y$ is not changing, i.e., where $\diff{y}{t}=0$. Write the equation for the $y$-nullcline in terms of $x$ and $y$.
This curve is a
with vertex at the point
that opens
.
Plot this curve on the above phase plane and label it as either “$y$-nullcline” or “$\diff{y}{t}=0$”. (Pencil is recommended, or wait until the next step so that you know exactly where the curve should intersect the $x$-nullcline.)
-
The equilibria are the points where both $x$ and $y$ are unchanging, i.e., we need both $\diff{x}{t}=0$ and $\diff{y}{t}=0$. In other words, the equilibria are the points where the $x$-nullcline and the $y$-nullcline
To determine for the equilibria, solve the $x$-nullcline for $y$ (if you haven't already) and substitute this value for $y$ into the $y$-nullcline. The result is a quadratic equation for $x$:
Solve this equation (e.g., by factoring or plugging into the quadratic formula) to determine the values of $x$ at the two different equilibria: $x_e =$
and $x_e =$
.
Plug each of these values of $x$ into either nullcline equation (probably the $x$-nullcline is easier) to determine the values of $y$ at the equilibria. The final result is that the equilibria are $(x_e,y_e) =$
and $(x_e,y_e) =$
.
Plot the equilibria on the above phase plane. At this point, you can finalize your sketches of the nullclines, if you haven't already, making sure they intersect at the equilibria.
-
At each point $(x,y)$, the rate of change of $x$, is determined by the first equation $\diff{x}{t} = - 2 x + 3 y$. To summarize how this rate of change depends on $x$ and $y$, let's define the function $f(x,y) = - 2 x + 3 y$ so that the dynamics of $x$ are given by $\diff{x}{t}=f(x,y)$. Similarly defined $g(x,y)=4 x - y^{2}$ so that the dynamics of $y$ are given by $\diff{y}{t}=g(x,y)$. Rewrite the equations for the $x$-nullcline and $y$-nullcline in terms of $f(x,y)$ and $g(x,y)$. The $x$-nullcline is the equation
and the $y$-nullcline is the equation
. The equilibria are the simultaneous solution to both equations
\begin{gather*}
_\\
_.
\end{gather*}
With those definitions of $f(x,y)$ and $g(x,y)$, we can compactly express the direction at which the solution must move at each point $(x,y)$. The solution must move through the point $(x,y)$ in the direction $(dx/dt, dy/dy) =$
. (Your answer should be a vector in terms of $f(x,y)$ and/or $g(x,y)$.)
Let's calculate the direction of movement at a few points in the phase plane. For each point, calculate $(dx/dt, dy/dt)=_$. Then sketch the point on the phase plane along with a vector coming out of that point and whose direction is given by the direction of motion of the solution through that point. The length of the vector is arbitrary so make it a convenient length -- long enough to see, but not too long to cross a nullcline into another region. The exact direction of the vector is less important than making sure you have the signs of $dx/dt$ and $dy/dt$ correct. (In other words, your vector should point in the correct compass direction, such as southwest, north, or northeast.)
- $(x,y)=\left ( 8, \quad 4\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 4, \quad 1\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 7, \quad 10\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 11, \quad 7\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 10.5, \quad 7\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 10.24, \quad 6.4\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 0, \quad 8\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 6, \quad 4\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 4, \quad 4\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 3, \quad 3\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 0, \quad 0\right )$, $(dx/dt,dy/dt)=$
- $(x,y)=\left ( 9, \quad 6\right )$, $(dx/dt,dy/dt)=$
-
Just considering the first quadrant (where $x \ge 0$ and $y \ge 0$), the nullclines cut the phase plane into
pieces. Away from a nullcline $\diff{x}{t}$
be zero and $\diff{y}{t}$
be zero. Since we $f(x,y)$ and $g(x,y)$ are continuous functions, they
change signs without passing through
. Therefore, within each of the those four pieces, the basic compass direction (northwest, east, southeast, etc.)
change. To get the basic direction everywhere in the phase plane, it suffices to draw just one direction arrow in each of those pieces, which you
already done.
-
The $y$-nullcline cuts the $x$-nullcline into
pieces. Along each piece of the $x$-nullcline, $\diff{x}{t}=$
and $\diff{y}{t}$
be zero. Any direction arrow crossing through one of those pieces of the $x$-nullcline must point
. Along one piece of the $x$-nullcline, the sign of $\diff{y}{t}$
change so the direction arrows
change direction on a single piece of the $x$-nullcline. To determine in which direction the solution crosses the $x$-nullcline, it suffices to draw just one direction arrow in each of those pieces, which you
already done.
-
Similarly, the $x$-nullcline cuts the $y$-nullcline into
pieces. Along each piece of the $y$-nullcline, $\diff{y}{t}=$
and $\diff{x}{t}$
be zero. Any direction arrow crossing through one of those pieces of the $y$-nullcline must point
. Along one piece of the $y$-nullcline, the sign of $\diff{x}{t}$
change so the direction arrows
change direction on a single piece of the $y$-nullcline. To determine in which direction the solution crosses the $y$-nullcline, it suffices to draw just one direction arrow in each of those pieces, which you
already done.
-
What must the solution look like when starting at the initial condition $(x(0),y(0)) = \left ( 8, \quad 1\right )$? To determine the shape of the solution, plot the point $\left ( 8, \quad 1\right )$ on the above phase plane. Starting at that point, in which direction must the solution move?
(You don't actually have to compute $(dx/dt,dy/dt)$ for $(x,y) = \left ( 8, \quad 1\right )$; you can just determine the general direction from the arrow(s) drawn in its piece of the phase plane. Draw the point in $(x,y) = \left ( 8, \quad 1\right )$ in the above phase plane and draw a curve moving from that point in the correct general direction.
As the curve keeps moving in that direction, what must it hit first?
When it hits _, in which direction must it move?
Once the solution trajectory crosses _, in which direction must it move?
As the trajectory continues moving _, can it cross _ again?
Can it cross the $y$-nullcline?
Since the trajectory is hemmed in and most continue moving _, where must the trajectory end up as time increases to infinity? It must end up at the
$(x,y)=$
.
-
Sketch the solution you determined on the below axes. Plot $x(t)$ versus time $t$ on the top axes and $y(t)$ versus time on the bottom axes. Since time is not represented on the phase plane, the values of $t$ are arbitrary (hence the $t$-axes have no scale). Although the shape will be approximate, certain values, such as the initial condition and the long term behavior must correctly match your solution. For intermediate values of time, just make sure that $x(t)$ and $y(t)$ move in the proper directions to match the sequence of directions that you determined in the phase plane.
Since the graphs must be graded by hand, you can achieve at most 70% correct via computer grading.