Projectile motion: optimal launch angle for weak quadratic drag

In the absence of air resistance, a launch angle of 45° maximises range. When there is drag linear in speed, the equations of motion can be integrated analytically, and closed-form expressions in terms of the Lambert W-function can be obtained for the optimal launch angle; see Packel & Yuen (2004). However, a more realistic model of air resistance has drag proportional to the square of speed, for which the equations of motion are nonlinear and analytic solutions cannot be obtained.

In high school there was this rather horrible investigation where we had to (experimentally) compare the optimal launch angle of a golf ball and a ping-pong ball. At the time I believed that nothing much could be done in terms of modelling other than solving the equations of motion numerically, but back then I knew nothing of scaling and perturbation theory. Now armed with some basic knowledge of these very useful tools, I have been able to derive an asymptotic expansion for the optimal launch angle when air resistance is relatively weak.

Solution

Equations of motion

Suppose a projectile of mass $m$ is launched at speed $u$ and angle $\phi$ from the ground, which has gravitational field strength $g$, and let there be drag proportional to the square of the projectile's speed, with constant of proportionality $b$. Let dots denote time derivatives. The drag force on the projectile has magnitude $b \norm{\dot{\vec{x}}}^2$, and it acts in the direction opposite to the projectile's velocity, i.e. in the direction $-\dot{\vec{x}} / \norm{\dot{\vec{x}}}$. Therefore the drag force is $-b \dot{\vec{x}} \norm{\dot{\vec{x}}}$, and so in components the equations of motion are

Scaling

Since we shall be making a perturbation from the dragless $b = 0$ case, it is appropriate to choose the length scale and time scale thereof. In the dragless case the mass $m$ is irrelevant, and the only parameters are the initial speed $u$ and gravitational acceleration $g$, yielding the length scale

and the time scale

Therefore we put $\hat{x} = x / L$, $\hat{y} = y / L$, and $\hat{t} = t / T$ to obtain scaled (dimensionless) variables $\hat{x}$, $\hat{y}$, and $\hat{t}$. Dropping the hats, the equations of motion become

where

is the initial drag-to-weight ratio, the only dimensionless group in the problem. By definition, the projectile's terminal speed $c$ is given by

Dividing these, we see that

Now the optimal angle is dimensionless, so it can depend only on the sole dimensionless group $B$. Thus, the optimal angle depends only on $\sqrt{B} = u / c$, the ratio between the initial and terminal speeds. (I wish I knew this back in Year 12.)

Perturbed trajectory

By "weak drag" I mean that $B \ll 1$, i.e. $u^2 \ll c^2$. We make an asymptotic expansion in powers of $B$ about $B = 0$:

Substituting these into the equations of motion and equating coefficients, we obtain successive 2nd-order ordinary differential equations for $x_0$, $y_0$, $x_1$, $y_1$, etc. Thus $\dot{x}_0$, $x_0$, $\dot{y}_0$, $y_0$, $\dot{x}_1$, …, $y_2$ can be determined by straight integration, the details of which I leave to the manuscript (page 3 onwards).

You might be wondering why we have stopped at order $B^2$. Initially I thought that integration could be performed to arbitrary order (although the amount of algebra required grows very quickly). However I was wrong; it turns out that the following integrals appear at cubic order, which have no closed-form expression:

If you are able to evaluate either of the two integrals above, please contact me at yawnoc.outsell414@passmail.net. On the other hand, quartic terms, even if they could be found, would be of little practical use, since the expansion is asymptotic and as such probably does not converge.

Flight time

Having determined the trajectory, we then determine the flight time, given by $y (t) = 0$. To quadratic order, i.e. with

this becomes

The unperturbed (dragless) flight time is given by the positive solution to $y_0 (t_0) = t_0 (\sin\phi - t_0 / 2) = 0$, which is

From the linear and quadratic terms we then obtain $t_1$ and $t_2$ (see page 20 onwards of manuscript).

Range

Substituting the flight time into the horizontal component of the trajectory gives the range

where $R_0$, $R_1$, and $R_2$ are (horribly complicated) functions of $\phi$ (see page 24 onwards of manuscript).

Optimal launch angle

Let primes denote $\phi$ derivatives. Then the optimal launch angle is given by $R' (\phi) = 0$, or, to quadratic order,

From the constant term we have $R'_0 (\phi_0) = 2 \cos (2 \phi_0) = 0$, yielding the familiar

in the absence of air resistance. From the linear and quadratic terms we obtain after much differentiation and algebra (pages 27–35 of manuscript) the constants $\phi_1$ and $\phi_2$, and hence the result:

Result

For small initial drag-to-weight ratios (or small initial-to-terminal kinetic energy ratios)

the optimal launch angle has the asymptotic expansion

or

Numerical verification

In the following table we compare numerically computed optimal launch angles with those from the asymptotic result above:

Indeed the asymptotic expansion is very accurate for $B < 0.5$ (or equivalently $u / c < 0.7$).

The true optimal launch angle is a decreasing function of $B$. Thus, very crudely, the asymptotic expansion becomes useless when it stops decreasing, which occurs at

Finally, here is a plot of the optimal launch angle in terms of $\sqrt{B} = u / c$, the initial-to-terminal speed ratio:

See also

• Projectile motion: optimal launch angle from a platform

Cite this page

Conway (2023). Projectile motion: optimal launch angle for weak quadratic drag. <https://yawnoc.github.io/math/projectile-weak-drag> Accessed 2025-04-05.