This was an interesting problem to me because it showed me how my intuition could be misled by the geometry of the problem. One would think by looking at the problem (I did anyway) that there could be a centering force. It's obvious that the magnet would experience an attractive force before it gets fully between the bars - calculation with femm agrees with this if you move the magnet a bit to the left. I think the position it is in for this problem is right where it hits the fuzzy zone where that force is disappearing. To convince myself of this, I grabbed the magnitude and normal components of B along the right and left side contours where the force line integral is done and put them in a spreadsheet. Then I calculated the x-component of the force from the stress tensor for right and left sides and added them up. The graph shows the right contour contribution (red) and the left contour contribution (grey) with the total x component of force in black. It illustrates what David mentioned in the previous post about two large but opposite sign contributions summing to zero (or close to it). 2 notes: this illustrates force up to a multiplicative constant since I didn't multiply by path lengths. Also, I didn't include the x-component of force from the top and bottom contours, since they are of equal magnitude but opposite sign, and I assumed from the symmetry of the problem that they would cancel. Based on this result though, I could see how the answer might be down in the noise so to speak. I tend to trust this result, but I would be interested in seeing some pics of Greg's actual setup and magnets. I know the NdFeB magnets are highly anisotropic and very strong; one possible source of stray force to look out for would be if the magnetization direction of the magnet were not completely perpendicular to the bar surfaces. Does the actual setup show a true centering force in the center of the bar; i.e. does the force direction switch signs as you roll past the middle? Rob
Attachment:
jpg00000.jpg
Description: JPEG image
