Flat rotation curves. Stars orbit on near-circular paths with vc2 = R ∂Φ/∂R. Here Φ is a bulge (Hernquist, 5×109 M☉), a Miyamoto-Nagai disc (6×1010 M☉) and a logarithmic dark halo, tuned so vc(8 kpc) = 220 km/s, an orbital period of 224 Myr and about 9×1010 M☉ inside the Sun's orbit. The visible matter alone (dashed line) would make the curve fall beyond ~10 kpc; observed spirals stay flat, which is the classic argument for a dark halo. Set the halo slider to 0 and watch the outer stars fly apart.
The winding problem. The angular speed Ω(R) = vc/R falls with radius, so any pattern made of the same stars is sheared. For a flat curve the pitch angle of a material arm obeys cot i = cot i0 + Ωt; after a few turns the arm is wound shut. Press "Material arms" and compare.
Density waves. Real arms are not made of fixed stars. In the kinematic (Kalnajs) picture stars follow slightly elliptical orbits, each described by the epicycle frequency κ2 = R dΩ2/dR + 4Ω2; if the ellipse orientation rotates smoothly with radius, orbits crowd together along a spiral. The ellipses precess at Ω − κ/2, so the pattern turns more slowly than the stars and lasts much longer than a material arm. Stars flow through the arms. The dust lanes here rotate at the pattern speed at 8 kpc.
Simplifications: test stars do not feel each other (no self-gravity, so the wave is not self-sustaining and shear still smears it slowly); no gas, star formation or bar; the bulge stars are on random near-circular orbits, not an equilibrium distribution function; the disc thickness is fixed; stars are drawn thousands of times brighter than real. Young blue stars are 18 % of the sample and are only a tracer. Numerical method: kick-drift-kick leapfrog, step 1.5 Myr.
Running numerical checks...