Laplace’s Equation and Field/Potential Estimation Alan Murray Estimation? Often we can “spot” some equipotentials (electrodes) Can we infer others from these? Can we create field lines and strengths from the equipotentials? How accurate is this? How useful is this?! Alan Murray – University of Edinburgh Estimation – the principle Given some equipotentials Can we infer the rest? And then the field lines? 5V 1V 2V 3V 0V E-Field lines 4V Alan Murray – University of Edinburgh General idea Let’s say we know the height of b, c, d and e’s bums. e Can we estimate the height of a’s bum? b a c The “estimate” is 100% accurate d Ha’s bum = ¼(Hb’s bum + Hc’s bum + Hd’s bum + He’s bum ) Alan Murray – University of Edinburgh General idea Let’s say we know the height of b, c, d and e’s bums. Can we estimate the height of a’s bum? The “estimate” is 100% accurate Ha’s bum = ¼(Hb’s bum + Hc’s bum + Hd’s bum + He’s bum ) Alan Murray – University of Edinburgh General idea Let’s say we know the height of a, c, d and e’s bums. Can we estimate the height of a’s bum? Now the estimate is not 100% accurate How could it be made more accurate? Ha’s bum = ¼(Hb’s bum + Hc’s bum + Hd’s bum + He’s bum ) Alan Murray – University of Edinburgh Now for some Maths … E V , D E , .D .V V , V , x, y , z 2 2 2 V 2 in full, V ( x , y , z ) This is Poisson’s Equation, which describes the variation of V(x,y,z) in space when a charge density ρ(x,y,z) is present. V 0 2 in full, V ( x , y , z) 0 2 This is Laplace’s Equation, which describes the variation of V(x,y,z) in space when no charge is present. Alan Murray – University of Edinburgh Method … V4 V1 Divide the region of interest into a grid … finer grid means more accuracy and more calculation! … then approximate 2V V3 Alan Murray – University of Edinburgh V2 Look at neighbouring grid points 2 d 2Vx d Vy In 2D, V 2 0 2 dx dy 2 2 2 d Vx d Vy d dV dx 2 dy 2 dx dx Approximate dV dx dV dx dV dy dV dy a b c d V x V x V y V y dV dV , dx dy V0 V1 V2 V0 V0 V3 b c d d dV dy dy a V4 (d) V1 (a) V0 ? (c) V3 V4 V0 Alan Murray – University of Edinburgh (b) V2 Look at neighbouring grid points 2 d 2Vx d Vy In 2D, V 2 0 2 dx dy 2 2 2 d Vx d Vy d dV dx 2 dy 2 dx dx Approximate d dV dx dx 0 d dV dx dx dV dx d dV dy dy 0 b dV dy d d dV dy dy d dV , dy dy dV dx (d) V1 (a) c V4 V0 V0 V3 Alan Murray – University of Edinburgh V0 ? (c) V2 V0 V0 V1 a dV dy V4 V3 (b) V2 Look at neighbouring grid points d 2V So, tidying things up, dx 2 0 V2 V1 2V0 d 2V , 2 dy 2 0 V4 V3 2V0 2 2 d 2Vx d Vy V2 V1 2V0 V4 V3 2V0 V1 V2 V3 V4 4V0 V 2 2 2 2 dx dy 2 2 Prof. Laplace says that V 0, so:2 V1 V2 V3 V4 4V0 0 2 V1 V2 V3 V4 V0 in 2D 4 V1 V2 V3 V4 V5 V6 V0 in 3D 6 Alan Murray – University of Edinburgh Look at neighbouring grid points V0 is the average of its 4 neighbours in 2D V0 is the average of its 6 neighbours in 3D Intuitively “correct” … and supported by theory. Is it useful? Often we have some equipotentials • Fixed by conducting plates/electrodes/etc. Use Laplace (approximately) to calculate the other potentials roughly Repeat until accuracy is achieved • Or you go blue in the face! Alan Murray – University of Edinburgh Worked Example : Coarse Grid, First Guesstimates 100V 40V 40V 40V 50V 50V 0V -40V 0V Alan Murray – University of Edinburgh -100V Worked Example : Coarse Grid, Iterate … 100V 50V =48V =38V 40V (100+50+40+0)/4 (100+48-40+40)/4 0V -40V 40V 50V 0V Alan Murray – University of Edinburgh -100V Worked Example : Coarse Grid, Iterate … 100V =38V =48V 47V 50V 50V -4V -13V Alan Murray – University of Edinburgh -11V -100V Worked Example : Coarse Grid, Iterate … 100V =38V =48V 47V 50V 50V -4V -13V Alan Murray – University of Edinburgh -11V -100V Iteration : Guidelines Do not spent ages making initial estimates • All values will converge • Sensible values will converge faster Always use the most recent values of neighbouring potentials If the accuracy required is “to the nearest volt”, stop when the last change to every value is <1V Alan Murray – University of Edinburgh Worked Example : Coarse Grid, Final Values =90° 100V 75V equipotential 48V 50V 45V 45V equipotential -5V -16V E 48V 50V 0V equipotential -5V Alan Murray – University of Edinburgh -100V Can we calculate the electric field .. approximate (-V/x, -V/y)? 100V E 48V 50V Ey = -[100-(-5)]/2 45V 48V Ex = -[45-50]/2 -5V -16V = 0.05 -5V E Alan Murray – University of Edinburgh 50V = (50, -1050)V/m -100V Finite Differences and Laplace This is clearly tedious Not with a computer it’s not • Can make the grid arbitrarily accurate • Can calculate for ages to achieve high accuracy Can include not-zero charge density • and non-zero conductivity • and AC signals This is the technique used when a “closed form” solution is not feasible http://www.see.ed.ac.uk/~afm/teaching/em3/ Examples : MS Excel Alan Murray – University of Edinburgh Finite Differences and Laplace Examples? • Field lines in a MOSFET • Shape and form of “plates” in a cathode ray tube/Scanning Electron Microscope etc • Fields, resistance and inductance of rails on the London Underground This was done with help from Brian Flynn and me, by one of our own M.Eng. Students! Aiming to detect cracks in the line Practically important or esoteric Electromagnetics? The student was able to solve Maxwell#3 “by hand” to calculate inductance. Alan Murray – University of Edinburgh Example : Train Tracks Note adaptive mesh – finer where necessary Alan Murray – University of Edinburgh Example : Train Tracks colour indicates electric field strength E Alan Murray – University of Edinburgh Example : Train Tracks colour indicates current density J Alan Murray – University of Edinburgh