WorksPhysics and Astronomy
Exact turning invariants for projectile motion under quadratic drag and a vertical-axis Magnus force
Abstract
A ball spinning about a vertical axis is pushed sideways by the Magnus force, so its ground track curves. We show that when the lift and drag coefficients are constant, the horizontal part of this problem is exactly solvable in elementary functions. The step that makes this work is to write the horizontal velocity as a complex number and to use path length rather than time as the independent variable. Both aerodynamic forces are quadratic in speed, so the speed cancels and what remains is a linear equation with constant coefficients in which gravity does not appear. Three exact results follow from it. The heading turns at a fixed rate per unit path length, so a half turn always costs the same distance whatever the launch conditions and whatever the drag. The horizontal speed decays exponentially in the turn angle at a rate fixed by the drag-to-lift ratio, which makes the horizontal hodograph a logarithmic spiral. The radius of curvature of the ground track equals the lift length times the cosine of the flight-path angle. The whole flight problem also collapses to one scalar differential equation controlled by two dimensionless numbers. We use this to settle a concrete question. Two identical balls thrown in opposite horizontal directions with opposite spin can meet head-on when drag is absent, on a one-parameter family we give in closed form, and they fail to meet once drag acts. Every analytic claim is checked against machine-precision numerics, and the code and data are openly available.
Introduction
The sideways deflection of a spinning ball is one of the most familiar everyday demonstrations of fluid dynamics, and one of the oldest to be studied quantitatively. Briggs measured the deflection of a baseball spinning about a vertical axis in a wind tunnel and found how it depends on spin rate and speed[1]. Mehta’s review is still the standard entry point to sports-ball aerodynamics[2], and measurements of the lift and drag coefficients of spinning balls are now routine[3, 4].
The equations of motion for a spinning ball under gravity, quadratic drag and the Magnus force are easy to write down, and they are almost always integrated numerically. Bray and Kerwin, in the standard treatment of a soccer ball struck with sidespin, put the position plainly: “These equations have no closed form solutions but can be solved numerically using a Runge–Kutta routine”[5]. Nathan’s reference study of baseball flight also uses fourth-order Runge–Kutta[3], as does the recent pedagogical treatment by Mencke et al.[6]. In the projectile-dynamics literature, Turkyilmazoglu and Altundag obtain closed forms when either the quadratic drag or the Magnus term can be dropped, and they describe the case in which both act together as highly nonlinear, with only perturbative solutions available[7]. Chudinov’s asymptotic analysis of the same problem gives the velocity hodograph as an approximate implicit formula[8]. Reviews of projectile motion with drag[9, 10] trace the analytic tradition back to Johann Bernoulli’s 1719 idea of parametrising planar motion by the tangent angle rather than by time, which decouples the equations but leaves the solution implicit.
This paper points out that one physically natural special case of the system is exactly integrable in elementary functions. When the spin axis is vertical and and are constant, the horizontal motion separates completely from gravity and can be solved in closed form. The invariants that come out are simple, and as far as we can tell they have not been written down before.
Section II fixes the model. Section III derives the central identity and reduces the full flight problem to a single scalar equation. Section IV collects the invariants. § 0.5 poses and answers a concrete geometric question that puts the machinery to work, namely whether two counter-spinning balls thrown in opposite directions can be made to collide head-on. Section VI describes the numerical verification, Section VII discusses scope and an open problem, and Section VIII concludes.
Model
A ball of mass and radius , with cross-sectional area , moves through still air of density under gravity, quadratic drag and the Magnus force,
where is the unit spin vector. Throughout we take the spin axis vertical, , which is the pure-sidespin configuration studied in Refs.[1] and[5]. It is convenient to absorb the coefficients into two inverse lengths,
and to give a name to the lift length,
Equation (3) is the exact analogue of the radius of a constant- turn in aircraft performance, where the centripetal requirement and the available lift both scale as and the speed cancels.
We write for the azimuth of the horizontal velocity, measured as an unwrapped angle rather than through an inverse tangent, and for path length, so that . The flight-path angle is , with , where is defined in Sec. III. Sections III and IV assume and constant, and Sec. VI A reports what happens when that assumption is relaxed.
Exact reduction
The central identity
Combine the two horizontal components of the velocity into a single complex number,
Three observations make the horizontal problem linear. Gravity is vertical and contributes nothing to . Drag is antiparallel to and contributes . With the Magnus term involves , which is horizontal and which in complex notation is exactly , a rotation by . Putting these together,
The right-hand side is proportional to , and this is what makes the change of variable from time to path length pay off. Using , the speed cancels identically and Eq. (5) becomes linear with constant coefficients,
Equation (6) is exact. It holds with gravity acting throughout, at any launch angle and for any ball, and it does not involve the vertical motion at all. Everything that follows comes from it.
Two features are worth emphasising. First, the quadratic velocity dependence of both aerodynamic forces is essential, because it is the common factor that cancels under . A Magnus force linear in speed, of the kind that arises for vortices in superfluids where the Magnus–Lorentz correspondence is standard[11], would not have this property, and the associated turning radius would then depend on speed. Second, the reparametrisation is in the spirit of Bernoulli’s tangent-angle method[9], with the difference that here the angle variable is an exact affine function of path length. That is what upgrades the solution from implicit to explicit.
Reduction to one scalar equation
Because is now known in closed form, the whole flight problem reduces to a single scalar equation. Take the turn angle as the independent variable, which by Eq. (6) means , and let
The Magnus force has no vertical component, so . Dividing this by gives . Multiplying by the integrating factor and using together with yields
with the single dimensionless group
where is the launch elevation and the launch speed. Equation (8) carries the whole problem. Given , the heading is , the horizontal speed is , the total speed is , and the position follows by quadrature. The three-dimensional velocity equation (1) has been reduced exactly to one scalar equation in the two parameters and .
Two functionals of the solution encode the geometry we shall need. Since , the ball returns to its release height after a turn if and only if
Similarly , so the launch-to-landing chord is
Equation (11) is exact and is the key formula of § 0.5.
Invariants
Constant turning per unit path length
Taking the argument of Eq. (6),
so the heading advances by a fixed angle per unit distance travelled. Gravity, launch angle, launch speed and are all absent from Eq. (12). Drag is antiparallel to and so contributes nothing to . It slows the ball along the path without bending it. A heading change of therefore always costs
and in particular a 180° turn costs exactly of path, whatever the launch conditions and whatever the drag. For a baseball with this is m.
Logarithmic-spiral hodograph
Taking the modulus of Eq. (6) and eliminating using Eq. (12),
so the horizontal velocity traces a logarithmic spiral whose pitch is set entirely by the lift-to-drag ratio. At this becomes the compact statement .
We stress that Eq. (14) governs the horizontal speed. The total speed obeys no such law, because gravity re-accelerates the ball on descent and terminal velocity places a floor under . Using for underestimates the true ratio by more than an order of magnitude at small , as Table IV shows.
| simulated | |||
| 0.5 | 0.0019 | 0.0435 | |
| 1.0 | 0.0432 | 0.2110 | |
| 2.0 | 0.2079 | 0.4622 | |
| 3.0 | 0.3509 | 0.5938 | |
| 5.0 | 0.5335 | 0.7263 | |
| 10.0 | 0.7304 | 0.8506 |
Table IV. Residual speed at closure, comparing the exact horizontal law (14) with the same expression applied, incorrectly, to the total speed.
Equation (14) inverts to
which suggests a measurement. The lift-to-drag ratio of a real ball can be read off a single tracked trajectory by comparing the decay of its horizontal speed with the swing of its heading. Gravity drops out of Eq. (15) identically, so no correction for the vertical motion is needed.
Curvature of the ground track
The radius of curvature of the horizontal projection is , and , so
The track is tightest where the ball is climbing or diving most steeply, and it is flattest at the apex, where and exactly. This gives a second measurement, since the radius of curvature of the ground track at the apex is directly, without separate knowledge of , or . In the drag-free case runs from to , so
which says exactly how far the track departs from a circle.
| ball | drag | fitted radius (m) | (m) | rel. error | speed change (m/s) |
| baseball | off | 281.267705304 | 281.267705306 | ||
| baseball | on | 281.267705317 | 281.267705306 | ||
| frisbee | off | 8.075877046 | 8.075877046 | ||
| frisbee | on | 8.075877047 | 8.075877046 |
Table 1. Zero-gravity ground track, showing the fitted circle radius against . Drag changes the speed by more than an order of magnitude and leaves the radius unchanged.
Figure 1. Numerical verification of the turning invariants with drag off. Left: the turning rate normalised by , plotted along the path for three launch conditions, with a deviation from unity that never exceeds . Right: the total path length required for a 180° turn, normalised by , swept over m/s at five launch angles, with a deviation that never exceeds .
Gravity is the sole source of non-circularity
Setting in Eq. (8) gives and hence , so the ground track is a perfect circle of radius whether or not drag is present. This is a sharper statement than it may look, because drag can reduce the speed by any factor without altering the geometry. Table I confirms it numerically. A baseball that loses 94% of its speed still traverses a circle whose fitted radius agrees with to eleven significant figures.
The counter-spinning rendezvous
Statement of the problem
We now put the machinery to work on a concrete question. A thrower at the origin releases two identical balls at the same moment in opposite horizontal directions, one spinning about and the other about . Both curve in the same sense in the horizontal plane. Can the launch conditions be chosen so that each ball turns through exactly and the two collide head-on at the height of release?
Label the balls (launched along bearing with spin ) and (bearing with spin ), and give them the same launch speed and the same elevation angle. The configuration is invariant under reflection in the vertical plane at bearing combined with the exchange , since a mirror reflection reverses the horizontal velocity and negates the axial spin vector. Ball ’s endpoint is therefore ball ’s endpoint reflected in , and the two coincide if and only if
that is, if and only if the launch-to-meeting chord is perpendicular to the launch direction. By Eq. (11) with , condition (18) reads
The whole question is the sign of a single integral of the positive weight against . The rendezvous succeeds precisely when that weight is balanced about .
Without drag: an exact one-parameter family
For Eq. (8) is separable, and integrating gives
The map preserves the equation, so if the solution satisfies then by uniqueness
identically. Two things follow at once. First, is odd about , so the closure condition (10) holds automatically. Second, is even about while is odd, so exactly. Without drag the chord is always exactly perpendicular to the launch direction, at every launch angle. The rendezvous is therefore possible, and imposing fixes
with . Combining Eq. (22) with Eq. (9) gives the launch speed in closed form,
for any , which is a one-parameter family of solutions. Note that is the classical arc length of a drag-free parabola, so Eq. (23) is just a restatement of Eq. (13). For a baseball, Eq. (23) gives 130.87 m/s at and falls to a shallow minimum of 84.99 m/s near before rising again. At the required speed is 86.89 m/s and the meeting point lies 515.4 m away.
A detail worth recording, because it is easy to get wrong, is that at the meeting point both balls are descending. Their vertical velocities are common-mode and cancel in the relative velocity. The closing speed is therefore
rather than . We confirm this numerically to relative over . At the closing speed equals the speed of a single ball.
With drag: the chord tips below perpendicular
Once the antisymmetry (21) is broken and need not vanish. Because the closure condition (10) determines once and are given, the chord bearing at closure is a function of and alone. It cannot depend on and separately, nor on the mass, the radius or the air density. This follows from the structure of Eq. (8) rather than from inspection of the output, and it is confirmed numerically to six decimal places for coefficient pairs that share a common .
We evaluate the leading behaviour for small drag and small launch angle. Writing for the chord bearing, a perturbative solution of Eq. (8), given in the Appendix, yields
in radians. The coefficient is , and Richardson extrapolation of the numerically computed deficit reproduces it to a relative accuracy of , as shown in the right panel of Figure 2.
Figure 2. The obstruction to the rendezvous, computed from the reduced equation (8). Left: the chord-bearing deficit over and , strictly positive everywhere, with real sports balls marked at . Middle: the folded integrand of Eq. (27), which changes sign once, showing that the positivity of is not pointwise. Right: convergence of the numerically computed deficit to the analytic coefficient of Eq. (25) as .
What matters here is the sign. The coefficient in Eq. (25) is positive if and only if , which is to say if and only if
Since , the deficit is positive, the chord tips below perpendicular, and the rendezvous fails. The obstruction reduces to an elementary inequality.
A tempting argument that does not work
It is natural to try to prove pointwise. Folding Eq. (19) about gives
and one might argue that with drag the descent is always steeper than the ascent at matched turn angle, which would make the bracket positive everywhere. That argument is false. Because drag makes the ascending arc longer than the descending arc, the apex occurs at a turn angle greater than . We measure , and for the three cases in Figure 2. For slightly below both and then lie on the ascent, where is decreasing, so the ordering reverses. The bracket in Eq. (27) changes sign exactly once, as the middle panel of Figure 2 shows. The positivity of is genuinely an integral statement, in which a large positive contribution near launch outweighs a smaller negative one near the midpoint. We record this explicitly because the pointwise argument looks convincing at first.
Numerical survey and consequences
Beyond the asymptotic regime we evaluate the deficit directly from Eq. (8) over a grid spanning and , which brackets every ball in Table III. The deficit is strictly positive throughout, with a minimum of degrees at the corner , , where both small parameters vanish together. This is the left panel of Figure 2. We therefore state the following.
| object | closes | (m/s) | (s) | |||
| frisbee | 3.00 | yes | 64° | 19.85 | 2.813 | 0.590 |
| soccer | 1.00 | yes | 76° | 124.20 | 9.583 | 0.211 |
| table tennis | 0.667 | yes | 80° | 90.95 | 3.966 | 0.096 |
| golf | 1.00 | no | — | — | — | |
| baseball | 0.571 | no | — | — | — | |
| tennis | 0.455 | no | — | — | — |
Table III. Closure of a single 180° turn with drag active and capped at 150 m/s. The minimum- member of the -family is shown. Coefficients are representative constants rather than measurements.
Result. Take constant and with and a vertical spin axis. Two balls released in exactly opposite horizontal directions with opposite spin, at equal launch speeds and equal elevation angles, do not collide, at any launch speed or launch angle. This is proved for small by Eq. (25) and established numerically over the physically relevant domain. A proof valid for all and remains open, as does the corresponding question for unequal launch conditions, which we discuss in Sec. VII B.
The practical size of the failure is not small. The two balls miss by , which for the closure solutions of Table III is 1.0–1.9 m for a frisbee, 7.6–9.4 m for a table tennis ball and 30–46 m for a soccer ball.
We note in passing that this changes the count of conditions. The two residuals “return to launch height” and “” are two ways of writing the single scalar statement that two events coincide, over the three unknowns , and . The corresponding Jacobian is numerically rank one, with a singular-value ratio of . The physical problem still becomes genuinely two-dimensional once drag is present, through the independent condition (18), which without drag is satisfied automatically.
Numerics and validation
The full system is integrated as an eleven-component state , carrying and as state variables so that the heading unwraps correctly past without any inverse-tangent differencing. Integration uses explicit Runge–Kutta 4(5) with relative tolerance and event detection for the two closure events, never fixed steps[12, 13]. Parameter sweeps use a batched fixed-step integrator. Figures are produced with Matplotlib[14].
Table II summarises the verification. The reduced equation (8) reproduces from the full three-dimensional integration to better than across four balls and three launch angles, and the reduced chord bearing agrees with the full simulation to better than degrees. The invariants (12) and (13) hold to machine precision. Equation (12) holds equally well with drag active, which confirms that drag cannot torque the azimuth.
Beyond constant coefficients
The results of Secs. III and IV require and constant. Two remarks bound their scope. First, under constant the Magnus term depends on the spin direction only, so the spin magnitude is exactly inert. Trajectories for , 200 and 800 rad/s coincide to twelve digits. Any question about sensitivity to spin rate is therefore empty until is allowed to depend on the spin parameter . With the saturating form , a 10% spin mismatch between the two balls produces a 1084 m miss at the drag-free solution, which is larger than the chord itself. Second, we have implemented a smoothed drag-crisis model in which falls from 0.47 to 0.15 across –. With varying, no longer follows Eq. (14) exactly, although Eq. (12) survives so long as is constant, since drag never contributes to the turning rate whatever its magnitude.
Real balls
Table III collects the closure solutions with drag active, using representative coefficients[2, 3, 4]. Only three of the six objects can complete a turn at all below 150 m/s, because the required path must be covered before drag exhausts the ball. A baseball manages of turn at best, falling short by , and a solution exists mathematically only near m/s, which is Mach 1.8 and outside the validity of an incompressible quadratic-drag model. In view of § 0.5, we emphasise that the entries in Table III are single-ball closures and are not rendezvous solutions.
Discussion
Relation to known results
The individual ingredients used here are old. The complex substitution is standard for rotating-frame problems. The constant- turn radius is the aircraft-performance result that a wing at fixed lift coefficient turns with a speed-independent radius. The reparametrisation of projectile motion by a geometric angle is Bernoulli’s, dating to 1719[9]. What appears not to have been noticed is that these combine, for the vertical-axis case specifically, into an exactly solvable system. The turning angle is an affine function of path length, so Bernoulli’s implicit solution becomes explicit, and the turn radius survives the addition of drag untouched.
It is worth being precise about why the vertical-axis case is special while the general one is not. For a general spin orientation the Magnus term is not a pure rotation of , so the horizontal equation does not close and Eq. (6) fails. This is consistent with the perturbative status of the general problem in Ref.[7]. In exterior ballistics the situation is different again. For spin-stabilised projectiles the Magnus force is usually small enough to neglect, and the relevant lateral deflection arises instead from the Magnus moment through the yaw of repose[15], a gyroscopic mechanism unrelated to the one considered here.
Figure 3. Ground track of a drag-free closure solution at . Left: the track against its best-fit circle, with an RMS residual of 1.585 m on a fitted radius of 264.06 m, which is 43 times the ball radius. Right: the radius of curvature along the path, following with .
The Magnus–Lorentz analogy familiar from vortex dynamics[11] is instructive precisely where it breaks. There the Magnus force is linear in velocity, exactly like the Lorentz force, and the resulting circular orbits have a radius proportional to speed. The aerodynamic Magnus force is quadratic, and the radius is regardless of speed. That speed-independence is what makes Eq. (13) possible.
An open problem
If one drops the requirement that each ball turn through exactly and asks only that two oppositely thrown balls collide head-on at release height, then the head-on condition becomes
so the sum of the turns is fixed while the split need not be equal. The equal split is a drag-free accident. Counting gives four unknowns , , and against four conditions, which suggests isolated solutions with one ball turning past . We searched for these over a grid with pairing restricted to , and we found candidates with genuinely unequal turns of and . Every refinement drained back to the degenerate corner , and stalled at a residual miss of 1.5–2.0 mm, consistent with the law of Eq. (25). We were unable to establish either existence or non-existence of a finite-angle asymmetric solution, and we record it as open.
Figure 4. The drag-free rendezvous, showing the two counter-spinning trajectories and their common endpoint at for a baseball. The mirror symmetry is exact, with arrival times agreeing to .
Classroom use
Several of these results are accessible as exercises at intermediate undergraduate level, in a first course on classical mechanics that has already introduced quadratic drag. Deriving Eq. (6) from Eq. (1) requires only the observation that is a rotation, and it makes a clean illustration of why choosing the right independent variable matters. The statement that a 180° turn costs a fixed distance regardless of how hard the ball is thrown is counterintuitive enough to be worth setting as a prediction before the derivation. The circle of Table I is a good computational exercise, as is verifying that drag does not appear in Eq. (12). Finally, § 0.5 D is a useful cautionary example of a plausible pointwise argument that fails, with the failure diagnosable by plotting one function.
Conclusion
For a ball spinning about a vertical axis with constant lift and drag coefficients, the horizontal equation of motion becomes linear with constant coefficients when path length replaces time as the independent variable, and gravity drops out of it entirely. The heading then turns at the fixed rate per unit path length, so a given change of heading always costs a fixed distance. The horizontal speed decays exponentially in the turn angle at a rate set only by , and the ground-track radius of curvature is , which reduces to at the apex. Each of these gives a quantity that can in principle be measured from a single tracked trajectory without knowing the mass, the radius or the air density separately. The remaining vertical motion reduces exactly to the one scalar equation (8) in the two parameters and , which makes the whole flight problem cheap to survey.
Applying this to the symmetric two-ball rendezvous, we find that the meeting is possible without drag on a one-parameter family given in closed form, and that drag destroys it. The chord tips below perpendicular by at leading order, and the sign of that coefficient comes down to . The asymmetric version of the problem and a proof of the deficit’s positivity over the whole parameter range are both left open.
Author declarations
Conflict of interest. The author has no conflicts of interest to disclose.
Funding. This work received no external funding.
Data availability. The complete source code, regression test suite and figure-generating scripts are openly available[16]. All numbers quoted in this paper are reproduced by a single driver script, and the analytic claims are encoded as regression tests with the tolerances stated in Table II.
Table II. Verification of the analytic results against direct integration. Deviations are relative unless stated otherwise.
Leading-order deficit
We sketch the calculation leading to Eq. (25), working to first order in and second order in . For small , Eq. (8) linearises to , so
Imposing closure (10), which at this order reads , gives
With and ,
Using , and , and expanding to first order in , the terms independent of cancel identically, as they must by Eq. (21). What is left is
The imaginary part of Eq. (11) is at leading order, so
which is Eq. (25). The coefficient is positive because .
References
- [1]L. J. Briggs, “Effect of spin and speed on the lateral deflection (curve) of a baseball; and the Magnus effect for smooth spheres,” American Journal of Physics 27, 589–596 (1959).DOI
- [2]R. D. Mehta, “Aerodynamics of sports balls,” Annual Review of Fluid Mechanics 17, 151–189 (1985).
- [3]A. M. Nathan, “The effect of spin on the flight of a baseball,” American Journal of Physics 76, 119–124 (2008).
- [4]J. R. Kensrud and L. V. Smith, “Drag and lift measurements of solid sports balls in still air,” Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and Technology 232, 236–243 (2018).DOI
- [5]K. Bray and D. G. Kerwin, “Modelling the flight of a soccer ball in a direct free kick,” Journal of Sports Sciences 21, 75–85 (2003).pubmed.ncbi.nlm.nih.gov/12630787
- [6]J. E. Mencke, M. Salewski, O. L. Trinhhammer, and A. T. Adler, “Flight and bounce of spinning sports balls,” American Journal of Physics 88, 934–947 (2020).DOI
- [7]M. Turkyilmazoglu and T. Altundag, “Exact and approximate solutions to projectile motion in air incorporating Magnus effect,” The European Physical Journal Plus 135, 566 (2020).DOI
- [8]P. Chudinov, “Study of the asymptotic motion of a sporting projectile taking into account the Magnus force,” arXiv preprint arXiv:2409.15110 [physics.class-ph] (2024), 10.48550/arXiv.2409.15110.arxiv.org/abs/2409.15110
- [9]M. V. Lubarda and V. A. Lubarda, “A review of the analysis of wind-influenced projectile motion in the presence of linear and nonlinear drag force,” Archive of Applied Mechanics 92, 1997–2017 (2022).DOI
- [10]P. Chudinov, “On the problem of projectile motion in a medium with quadratic resistance,” arXiv preprint arXiv:2205.01564 [physics.class-ph] (2022), 10.48550/arXiv.2205.01564.arxiv.org/abs/2205.01564
- [11]E. B. Sonin, “Magnus force in superfluids and superconductors,” Physical Review B 55, 485–501 (1997).DOI
- [12]P. Virtanen et al., “SciPy 1.0: fundamental algorithms for scientific computing in Python,” Nature Methods 17, 261–272 (2020).
- [13]C. R. Harris et al., “Array programming with NumPy,” Nature 585, 357–362 (2020).
- [14]J. D. Hunter, “Matplotlib: A 2D graphics environment,” Computing in Science & Engineering 9, 90–95 (2007).DOI
- [15]R. L. McCoy, “Modern Exterior Ballistics: The Launch and Flight Dynamics of Symmetric Projectiles,” 2nd ed. (Schiffer Publishing, Atglen, PA) (2012).
- [16]P. Gaur, “Exact Arc-Length Reduction for Projectile Motion under Quadratic Drag and a Vertical-Axis Magnus Force, with a Non-Existence Result for the Symmetric Two-Ball Rendezvous,” https://doi.org/10.5281/zenodo.21979028 (2026), software and data.



