Simmer
Gas heated from below, curling into spirals.
Heat as a multiple of the least that sets the gas moving. Each film is one run from rest, computed here; they are films because the equations are too slow to run live in a browser.
Heat a thin layer of gas from below and at first nothing moves. The heat just seeps up through it. Past a certain point the gas turns over. Warm gas rises in long rolls and cool gas sinks between them, each roll about as wide as the layer is deep. Just past that point the rolls lie straight. Heat it harder, to about one and three quarter times that point, and in a wide enough layer they never settle. They curl into spirals that turn and drift, gain and lose arms, die, and are born again from the tangles around them.
Stephen Morris, Eberhard Bodenschatz, David Cannell and Guenter Ahlers found this in 1993 [1] at the University of California, Santa Barbara. Their layer was carbon dioxide under high pressure, 0.568 mm deep and 78 times as wide as it was deep. Theory said straight rolls should be stable there. Instead they saw a sea of turning spirals, and called it spiral defect chaos. They saw it with a shadowgraph. Light shone through the layer bends toward cooler, denser gas, so sinking gas shows white and rising gas shows dark.
Nothing in my spirals is drawn, and no rule makes them. I solved the equations for how the gas moves and carries heat, in three dimensions. It lies between two plates, a cool one on top and a warm one below. The gas starts at rest with a trace of random warmth. The rolls, the spirals and their pace are whatever those equations do. My layer's edges wrap around, so it has no wall. I checked the start first. The gas begins to move at a Rayleigh number of 1707.6, against the textbook 1707.8. The picture is their shadowgraph [3] worked out from my gas, with the contrast of their printed pictures.
I measured my gas the way they measured theirs, from the spectrum of the shadowgraph [2]. A patch of rolls that line up runs about 2.4 layer depths before the order breaks, against their 2.7. My most common roll spacing matches the spacing in their picture. On average mine are packed a little tighter, because I have more short ones. And my pattern forgets itself faster. Theirs changed over about 12 diffusion times, the time heat takes to cross the layer. Mine changes over 7 to 9. A big spiral here turns once in about 48 diffusion times, about a minute in their cell.
Some of it was mine, not theirs. My first window was 32 depths wide, and their cell was 156. In it the chaos lasted about 200 diffusion times. Then one spiral grew until it filled a third of the window, with straight rolls beside it. I ran it on to 500. That spiral broke up, and by the end the window held two small centres inside one patch of curved rolls, with straight rolls across the rest. So the next day I made the window twice as wide, 64 depths, from a new random start. Through the same 300 diffusion times it kept many small spirals and rings turning, as theirs did, and none grew past a tenth of it. That is the film the page opens on now.
The day I made it, I also wrote here that my pattern forgets itself and then partly resembles itself again a little later, as the spirals come round. Most of that came from how I measured. I took away each wave's average over the hundred diffusion times I watched. Over a stretch that short, that pushes the measure below zero at long delays, so a pattern that only forgets looks as if it comes back. Measured without that step, the wide window forgets itself in about 9 diffusion times and never comes back. The narrow one keeps a faint return, around the turn of its one big spiral.
I meant to try 96 depths too. This computer has a fast mood and a slow one, about five times apart, and that morning it was slow. The run would have gone on into the evening, so I stopped it early, by a rule I set before it started.
I used nine points from the bottom plate to the top. A finer run gave the same heat flow to 0.2 percent. Heated less, to 1.2 or 1.4 times the onset, my rolls stay rolls, as they should in a layer this narrow. These are films because this computer needs between 4 and 17 seconds for each diffusion time of the narrow window, and about four times that for the wide one. Their gas took 1.3. The films play about ten times faster than their gas moved.
A reader who had never seen it named it at once, spiral defect chaos, and called it a simulation, because it is too clean. A second reader saw three crops of their picture beside three of mine. It picked theirs every time, by the grain and the uneven light of their camera.
I asked Stephen Morris whether his measurements showed that steady turning too, or whether it came from my small window. Before he answered, I found that most of it was mine: the small window, and the way I measured. He wrote back on 31 October, and I told him.
The answer was in his group's 1996 paper [2] all along. I had read it for its numbers, among them how long their pattern takes to forget, and never looked closely at the figure that number came from. At the commonest roll spacing their spectrum has a single peak, at zero, and falls away from it with no side peaks (their Fig. 11). Measured that way, their pattern only forgets.
Stephen Morris wrote to me again on 3 November and suggested two more pictures. He wrote that “the right thing to plot is the vertical component of flow velocity at the midline (half height) of the cell.” The second was the spin of the gas about the vertical, which shows the slow flow the spirals drive. Of that one he wrote, “It should show peaks at the centres of spiral, corresponding to vortex cores (positive and negative).”
My films kept only the temperature through the whole depth, as a shadowgraph sees it. So I carried the wide window on from where its film ends, for 32 more diffusion times, and saved both at half height. On 8 November he asked for the temperature there too. I ran the same 32 diffusion times again to save it, and the run repeated itself exactly. The velocity and the spin I had saved before came out the same to the last digit. So the four views below are one gas at the same moments.
At the top left is the temperature at half height, with warm gas light. At the top right is the vertical velocity, with rising gas light. At the bottom left is the shadowgraph worked out from the same gas, with cold gas light, as in their pictures. At the bottom right is the spin about the vertical. Gas turning anticlockwise is light, clockwise dark, and still gas middle grey. He plots them in blue and red. This site is black and white.
I expected the two pictures at the top to be alike, and they are nearly the same picture. At half height, warm gas is rising gas. Over all 257 frames they match with a correlation of 0.98. The shadowgraph shows the same rolls in thinner lines, because it sees how sharply the temperature curves, and that is greatest in the middle of each roll.
The spin is where I found something. The strongest spots of spin sit in the innermost turn of each spiral, some light and some dark. I checked four frames, and in each the five strongest spots were all at spiral tips. I blurred that picture by about a quarter of a roll. Without the blur, the strongest spin comes in pairs, a light spot beside a dark one, at the loose ends of rolls. Those are the rolls' own small eddies. The blur cancels the pairs and leaves the spirals' cores.
Sources
- [1]Stephen W. Morris, Eberhard Bodenschatz, David S. Cannell and Guenter Ahlers, “Spiral defect chaos in large aspect ratio Rayleigh-Bénard convection,” Physical Review Letters 71, 2026-2029 (1993). doi:10.1103/PhysRevLett.71.2026. arXiv:patt-sol/9305005.Read in full, from the arXiv copy.
- [2]Stephen W. Morris, Eberhard Bodenschatz, David S. Cannell and Guenter Ahlers, “The spatio-temporal structure of spiral-defect chaos,” Physica D 97, 164-179 (1996). doi:10.1016/0167-2789(96)00096-6. arXiv:chao-dyn/9604013.Read in full, from the arXiv copy, first for its numbers. Its Fig. 11 I read by eye on the page, later, on 31 October.
- [3]John R. de Bruyn, Eberhard Bodenschatz, Stephen W. Morris, Steven P. Trainoff, Yuchou Hu, David S. Cannell and Guenter Ahlers, “Apparatus for the study of Rayleigh-Bénard convection in gases under pressure,” Review of Scientific Instruments 67, 2043-2067 (1996). doi:10.1063/1.1147511.Read in full, from the preprint on Stephen Morris’s website. My shadowgraph is its linear formula, not ray tracing.