next up previous
Next: ff_par_surf Up: ff_par_surf Previous: ff_par_surf

1. Inventing Surfaces

It is possible to invent surfaces of your own if you make them step-by-step. It helps especially to sweep them out by motions.



Example 1, revisited. To make a sphere of radius 1, first take an easy point, such as $ (1,0,0)$. Make it into a vertical semicircle by applying $ R _ t ^ {x\rightarrow z}$, where $ t$ varies from $ -\frac \pi 2$ to $ \frac \pi 2$; you get $ (1,0,0) R _ t ^
{x\rightarrow z}$. Then twist the semicircle around the $ z$-axis by applying $ R _ u ^ {x\rightarrow y}$, where $ u$ varies from 0 to $ 2 \pi$. You get

P$ (t,u) = (1,0,0) R _ t ^ {x\rightarrow z} R _ u ^
{x\rightarrow y}$, which reduces to

P$ (t,u) = (\cos t \cos u, \cos t \sin u, \sin t)$ for $ -\frac {\pi}2 \leq t \leq \frac {\pi}2, 0 \leq u \leq 2 \pi$.



Example 3, revisited. Pick an axis along which to twist the ribbon, say the $ y$-axis. Imagine the ribbon as swept out by a moving line segment, as the time $ t$ changes. Pick a line segment for time 0, say a segment of length 2 along the $ z$-axis: $ (0,0,u)$ for $ -1
\leq u \leq 1$. At time $ t$, the segment should be rotated by $ R _ t ^ {x\rightarrow z}$ and translated along the $ y$-axis a distance $ t$ (a distance proportional to $ t$ would also do), by adding $ (0,t,0)$. The time interval is arbitrary, but $ - \pi \leq t \leq \pi$ is nice, since it corresponds to one full twist. You get

P$ (t,u) = (0,0,u) R _ t ^ {x\rightarrow z} + (0,t,0)$, which reduces to

P$ (t,u) = ( -u \sin t, t, u \cos t)$, for $ -\pi
\leq t \leq \pi, -1 \leq u \leq 1$.



Example 4. Any graph $ (f(t,u),g(t,u),h(t,u))$. In fact, this is the most general parametric surface. You might think you would get an interesting surface by just choosing $ f,g,h$ to be unrelated messy functions, but such a method often gives disappointing results.



Example 5. A curve rotated about the $ x$-axis, with radius some function $ r(x)$, $ a \leq x \leq b$:

The curve drawn in the $ x,z$-plane would be $ z = r(x)$. To make this parametric in one variable, use $ x = t, y = 0, z =
r(t)$, so that the points are $ (t, 0, r(t))$. Now rotate about the $ x$-axis by multiplying by $ R _ u ^ {y\rightarrow
z}$, for $ 0 \leq u
\leq 2 \pi$. You get

P$ (t,u) = (t, 0, r(t)) R _ u ^ {y\rightarrow z}$ = $ (t,
- r(t) \sin u, r(t) \cos u)$ for $ a \leq t \leq b$, $ 0 \leq u
\leq 2 \pi$.



Example 6. A torus (doughnut): Pick a small radius $ r$ and a large radius $ s$, say 1 and 3. Start with one point, $ (r,0,0)$. Rotate it in the $ x,z$-plane by $ R _ t ^ {x\rightarrow z}$, to sweep out a circle. Then move the circle by a translation along the $ x$-axis so that its center is moved to $ (s,0,0)$. Then sweep out the surface by rotating around a vertical axis, using $ R _ u ^ {x\rightarrow y}$ for $ 0 \leq u
\leq 2 \pi$. You get

P$ (t,u) = \left ( (r,0,0) R _ t ^ {x\rightarrow z} +
(s,0,0) \right ) R _ u ^ {x\rightarrow y}$, which you can simplify, or program as is. (Be careful about parentheses in this formula. Which mathematical parentheses indicate a list of three coordinates and which are for grouping?)



Example 7. A Möbius strip: This is something like a combination of Examples 3 and 6. You need to move a line segment around a circle with time, turning it as you go. Start with a vertical line segment $ (0,0,u)$, $ -1
\leq u \leq 1$. Choose a radius $ R > 1$. At time $ t$, the segment should be rotated in the $ x,z$-plane, translated by $ (R,0,0)$, and then rotated by $ R _ t ^ {x\rightarrow y}$, where $ 0 \leq t
\leq 2 \pi$. But how fast should the segment be rotated in the $ x,z$-plane? It should get a half-twist during the time it sweeps around the circle of radius $ R$, so that the rotation should be $ R _ {t/2} ^ {x\rightarrow z}$. You get

P$ (t,u) = \left ( (0,0,u) R _ {t/2} ^ {x\rightarrow z} +
(R,0,0) \right ) R _ t ^ {x\rightarrow y}$ which you can simplify, or program as is. (Be careful about parentheses in this formula, lists versus grouping.)




next up previous
Next: ff_par_surf Up: ff_par_surf Previous: ff_par_surf
Kirby A. Baker 2002-03-01