DR Apps
DR.LuftkissenbahnDR Apps

DR.Luftkissenbahn · Model & numerical methods

Physics background

DR.Luftkissenbahn makes momentum transfer and energy conversion visible using two gliders. Their motion follows a coupled physical model that responds to interventions during an experiment.

Model version: 6 November 2026 · Dr. Daniel Roth · For teaching and understanding experiments

1. Two gliders, one direction of motion

The horizontal track is one-dimensional. The coordinate x denotes the centre of each glider, and positive velocities point to the right. The mass sliders in the workspace take effect immediately, during motion or while paused. A free glider retains its momentum p at the instant of the change; its velocity adjusts according to v = p/m. This intervention does not change positions or simulated time. Gliders that are already joined remain joined: their combined momentum is conserved, and their common velocity follows from the new total mass. Air-cushion strength, air resistance and the spring constants in Expert mode can also be changed during a run. A spring adjustment preserves positions, momenta and simulated time; its energy change is recorded as external work. Changing initial positions, initial velocities or buffers, switching an end spring on or off, or applying a configuration through “Setup”, instead starts a new measurement. The chosen spring constants are retained through these actions and Reset.

Track length
2.40 m
Mass of each glider
0.10–1.00 kg
Glider width
0.18 m
Spring constants D (default)
100 N/m each
Uncompressed contact distance
0.34 m
End-spring travel
0.16 m

Internally, all quantities use SI units. The momentum display uses the name Huygens, following the accompanying teaching material: 1 Hy = 1 kg·m/s = 1 N·s. A momentum current of 1 Hy/s corresponds to a force of 1 N.

2. Forces in the model

The spring bumper becomes active when the distance between the glider centres falls below 0.34 m. Here, Δx denotes the non-negative magnitude of compression. Compression produces equal and opposite forces and stores energy. Expert mode lets you set the contact spring between the gliders and the left and right end springs independently from 1 to 1,000 N/m. Each spring defaults to D = 100 N/m. Turning Expert mode off immediately returns all three spring constants to 100 N/m; switching it on again therefore shows these normal values in the fields. The contact value is the effective spring constant of the complete spring contact between the gliders; the two drawn bumpers do not contribute two additional forces. The optional end springs follow the same linear law. They act only when enabled and compressed.

Fspring = D · Δx
Espring = · D · (Δx)²Ideal springs store energy reversibly. Their compression is also shown visually.

In this equation, Fspring is the force magnitude. The contact spring pushes glider A to the left and glider B to the right. The left end spring pushes to the right, and the right end spring pushes to the left. Each force uses that spring’s own stiffness. At the same compression, a higher spring constant produces a greater force and stores more energy.

During elastic contact, the spring expands again. In a sticking collision, the gliders become joined at maximum compression, normally when their velocities are equal. The remaining contact energy becomes internal energy. For an isolated sticking collision, v = (mA · vA + mB · vB)/(mA + mB). In an isolated elastic collision between equal masses, the gliders exchange velocities.

Air cushion and friction

The air-cushion strength s lies between 0 and 1. The remaining normal force is modelled as N = m · g · (1 − s). Kinetic friction acts against the motion with magnitude μk · N. Static friction holds a stationary glider as long as the required opposing force does not exceed μs · N. Both friction coefficients default to 0.12. Surface friction disappears only at 100% air-cushion strength.

Air resistance can act independently: Fair = −b · v − c · |v| · v. The interface switches the quadratic term on and off, using c = 0.02 kg/m per glider. The linear coefficient is zero in the interface. These forces transfer momentum to the environment and convert mechanical energy into internal energy.

The hand and end stops

During dragging, a controller guides the glider with a force of at most 3 N and a target speed of at most 0.75 m/s. Its position is not changed directly. On release, an additional idealised impulse from the hand sets the velocity estimated from the last 100 ms of pointer motion, limited to ±2 m/s. This instantaneous impulse is not subject to the 3 N force limit. Its energy and momentum transfers are accounted for separately.

A fixed end stop exerts only the necessary compressive reaction and permits motion away from it. Without an end spring, an arriving glider comes to rest at the stop; joined gliders are slowed together. The simulation and the motion of another, separate glider continue. The end spring itself is lossless, although friction can remove energy at the same time. At full compression, a hard stop prevents geometrical interpenetration. Elastic bumpers reflect the motion without energy loss, whereas sticking gliders become joined and convert relative kinetic energy into internal energy.

3. Runge–Kutta integration with contact conditions

Continuous time integration includes the two positions and momenta, the opposing momentum transfers to the environment and the hand, the work done by the hand, and the internal energy. The classical fourth-order Runge–Kutta method (RK4) integrates them together. Velocities and forces are recalculated at every intermediate stage. Discrete changes to mass or spring stiffness and their associated energy changes are accounted for separately at the instant of the intervention.

dxi/dt = pi/mi
dpi/dt = ΣFiPosition, momentum and the quantities used in the balances evolve within the same time step.

An internal step lasts at most 0.25 ms. Comparing one full step with two half-steps estimates the local error. The step size is halved if necessary. The more accurate state obtained from the two half-steps is accepted. This error control does not guarantee an equally small global error everywhere.

The beginning and end of contact, stopping, static-friction limits and end stops are treated as events. A root-finding procedure locates their times. Until an event, the current contact and friction regime remains fixed; afterwards, it is determined again. A hard stop is therefore not simply approximated by a large force acting over a full step. Hard collisions and release impulses are discrete state changes with their own momentum and energy accounting. Changes to mass or spring stiffness also take effect immediately and have their own work accounting.

4. Tracking momentum and energy

The complete momentum balance includes the environment and the hand as well as the gliders. Air and Earth are treated together as the environment. Displaying the opposing momentum transfer makes it clear when the subsystem consisting of the two gliders no longer has constant momentum.

pA + pB + penvironment + phand = pinitial

The energy accounting separates kinetic energy, the energy of all active springs, and the internal energy produced. The hand can add or remove energy. The power Fhand · v and the rate of dissipation −Ffriction · v are integrated using the same RK4 stages as the motion.

Ekin + Espring + Einternal = Einitial + Whand + Wmass + Wspring

The mass slider is also an external, idealised intervention. At constant momentum, it changes kinetic energy: for a moving glider, increasing mass reduces kinetic energy, while decreasing mass increases it. The total change Ekin, after − Ekin, before is added to the signed work Wmass for each intervention. For joined gliders, the change in their combined kinetic energy is used. This produces no additional internal energy, and the initial energy is not reset. The rule represents a controlled change to a model parameter; it does not explicitly model a real glider being loaded or unloaded.

Absorption and sticking record the corresponding loss of mechanical energy as internal energy. Ideal springs do not produce heat. However, RK4 is neither exactly energy-conserving nor symplectic. A numerical balance residual remains visible. It is not hidden as heat or removed by rescaling the energy afterwards.

Changing spring constants during a run

Changing stiffness is an external, idealised intervention. The current compression stays unchanged. Stored spring energy therefore changes without relocating a glider or applying an instantaneous impulse. The force adjusts immediately and affects the subsequent motion. The energy change, including its sign, is added to Wspring.

ΔWspring = · (Dnew − Dold) · (Δx)²Stiffer at fixed compression: add energy. Softer: remove energy.

Wspring accumulates changes to all three springs. An uncompressed or disabled spring stores no energy, so changing its stiffness does no work. After a sticking collision, the contact spring energy has already become internal energy. The gliders are then held together by a constraint: changing the contact value creates neither new spring energy nor adjustment work. Initial energy and the internal energy already produced remain unchanged by every spring adjustment.

Returning to the normal values when Expert mode is switched off uses the same accounting. Positions, momenta and measurement time stay unchanged. For a compressed spring, restoring the normal value can add or remove energy; the corresponding work remains visible in the balance.

5. Computing, measuring and plotting

The interface requests advances of the physics model in intervals of 1/240 s. The numerical core subdivides these further. A regular measurement point is recorded after every two intervals, giving a rate of 120 Hz. Single steps, interventions and final states can produce additional points. The plots and CSV export use these measurements. The screen display has its own update rate.

Choosing slow motion changes the ratio of real time to simulated time. The force laws remain the same. While dragging, time advances at 1× speed. The force plot shows the resultant force on each glider. Instantaneous hard collisions and the final release impulse do not have a resolved force peak. Their effects on momentum and energy are still included. A mass change is also an instantaneous intervention without a resolved force-time curve; its energy change is accounted for separately.

Run duration and measurement window

The run duration can be set to 30 s, 60 s or unlimited. By default, the experiment runs without a time limit. Simulated time continues even when the gliders are at rest or one reaches an absorbing end stop; the gliders can still be moved. With a finite duration, the experiment pauses at the selected time. Changing the duration neither resets the setup nor starts it automatically. After increasing a limit that has been reached, the experiment can be resumed with “Continue”.

The app retains the complete measurement points from the most recent 60 simulated seconds. Older points are removed from the measurement buffer. The plot and CSV contain this current time window, not the entire history of an unlimited run. At high measurement densities, the plot draws a selection of actual measurement points while preserving the first point, minimum, maximum and last point within each time interval. No intermediate values are invented for this purpose; the CSV export contains the original stored measurement points from the time window, including the masses, all three spring constants and the work from mass and spring changes at each sample.

6. What the model describes

The simulation is a teaching model whose assumptions can be understood. It does not include rotation, tilt, deformation of the glider bodies, spring hysteresis or a measured blower characteristic. Earth is idealised as infinitely massive, so its kinetic energy is neglected. Sound and graphical effects do not remove energy from the model. Air-cushion strength describes a simplified supporting effect, not an electrical power rating.

The model helps explain forces, motion and conservation laws. A small energy-balance residual alone does not establish that every detail of real motion is described correctly. Real tracks are subject to additional influences and must be calibrated separately for quantitative measurements.

Calculations and measurement recording run locally in the browser. The CSV export is generated on the user's own device.