==> physics/resistors.p <== What is the resistance between various pairs of vertices on a lattice of unit resistors in the shape of a 1. Square with all vertices connected, 2. Platonic solid, 3. N dimensional Hypercube, 4. Infinite square lattice, and 5. between two small terminals on a continuous sheet? ==> physics/resistors.s <== The key idea is to observe that if you can show that two points in a circuit must be at the same potential, then you can connect them, and no current will flow through the connection and the overall properties of the circuit remain unchanged. 1. Square with all vertices connected The circuit between two diagonally opposite vertices of a square with all vertices connected is composed of two circuits in parallel, one of which is the diagonal itself (call this R0) and the other of which is: ---R1--o--R4--- | | | A --| R3 |-- B | | | ---R2--o--R5--- The general formula for 5 resistors connected like that (using the shorthand that 135=R1*R3*R5 and so forth) is: 123 + 124 + 125 + 135 + 145 + 234 + 245 + 345 R= --------------------------------------------- 13 + 14 + 15 + 23 + 24 + 34 + 35 + 45 Then the resistance between the diagonal vertices is 1 / (R + R0). If all the resistances are 1 ohm, the answer is 1/2 ohm. 2. Platonic Solids For the cube, there are three resistors leaving the two "connection corners". Since the cube is completely symmetrical with respect to the three resistors, the far sides of the resistors may be connected together. And so we end up with: |---WWWWWW---| |---WWWWWW---| |---WWWWWW---| | | |---WWWWWW---| | | *--+---WWWWWW---+-+---WWWWWW---+-+---WWWWWW---+---* | | |---WWWWWW---| | | |---WWWWWW---| |---WWWWWW---| |---WWWWWW---| |---WWWWWW---| This circuit has resistance 5/6 times the resistance of one resistor. Same idea for 8, 12 and 20, since you use the symmetry to identify equi-potential points. The tetrahedron is a hair more subtle: *---|---WWWWWW---|---* |\ /| W W W W W W W W W W W W | \ / | \ || | \ | / \ W / \ W / <------- \ W / \|/ + By symmetry, the endpoints of the marked resistor are equi-potential. Hence they can be connected together, and so it becomes a simple: *---+---WWWWW---+----* | | +-WWW WWW-+ | |-| | |-WWW WWW-| 3. Hypercube Think of injecting a constant current I into the start vertex. It splits (by symmetry) into n equal currents in the n arms; the current of I/n then splits into I/n(n-1), which then splits into I/[n(n-1)(n-1)] and so on till the halfway point, when these currents start adding up. What is the voltage difference between the antipodal points? V = I x R; add up the voltages along any of the paths: n even: V = 2{I/n + I/(n(n-1)) + I/(n(n-1)(n-1)) + ... + I/(n(n-1)^[(n-2)/2])} n odd: V = 2{I/n + I/(n(n-1)) + I/(n(n-1)(n-1)) + ... + I/(n(n-1)^[(n-3)/2])} + I/(n(n-1)^[(n-1)/2]) And R = V/I i.e. replace the Is in the above expression by 1s. For the 3-cube: R = 2{1/3} + 1/(3x2) = 5/6 ohm For the 4-cube: R = 2{1/4 + 1/(4x3)} = 2/3 ohm This formula yields the resistance from root to root of two (n-1)-ary trees of height n/2 with their end nodes identified (when n is even; something similar when n is odd). Coincidentally, the 4-cube is such an animal and thus the answer 2/3 ohms is correct in that case. However, it does not provide the solution for n >= 5, as the hypercube does not have quite as many edges as were counted in the formula above. 4. The Infinite Plane For an infinite lattice: First inject a constant current I at a point; figure out the current flows (with heavy use of symmetry). Remove that current. Draw out a current I from the other point of interest (or inject a negative current) and figure out the flows (identical to earlier case, but displaced and in the other direction). By the principle of superposition, if you inject a current I into point a and take out a current I at point b at the same time, the currents in the paths are simply the sum of the currents obtained in the earlier two simpler cases. As in the n-cube, find the voltage between the points of interest, divide by I. As an illustration, in the adjacent points case: we have a current of I/4 in each of the four resistors: ^ | | v <--o--> -->o<-- | ^ v | (inject) (take out) And adding the currents, we have I/2 in the resistor connecting the two points. Therefore V=R x I/2 and effective resistance between the points = R/2 ohm. The equivalent resistance between two diagonally adjacent vertices is 2R/pi. The resistance between two vertices k diagonal units apart is: (2R/pi) (1+1/3+1/5+...+1/(2k-1)). This, together with symmetry and the known R/2 resistance, suffices to determine all the equivalent resistances in this network, once you realize the following fact: if X,Z are any two different vertices, the equivalent resistance between X and Z is the average of the equivalents resistances between X and Z' over the four neighbors of Y. . . . . | | | | R R R R | | | | -R-------|------R--------X-------R-------A-------R-------B-----R- | | | | R R R R | | | | -R-------|------R--------|-------R-------Y-------R-------C-----R- | | | | For instance, if Z is the vertex labeled A in the above diagram we find that that the X-B resistance is (2-4/pi)R (the resistance from X to itself is zero); taking Z=Y and noting that the unlabeled neighbors of Y are equivalent to A,C yields the value (4/pi-1/2)R for the X-C resistance. 5. Continuous sheet The resistence is (rho/dz)log(L/r)/pi where rho is the resistivity, dz is the sheet thickness, L is the separation, r is the terminal radius. cf. "Random Walks and Electric Networks", by Doyle and Snell, published by the Mathematical Association of America. ******************************************** Instructions for Accessing the rec.puzzles Archive INTRODUCTION The rec.puzzles Archive is a list of puzzles, categorized by subject area. Each puzzle includes a solution, compiled from various sources, which is supposed to be definitive. EMAIL To request a puzzle, send a message to archive-request@questrel.com like: return_address your_name@your_site.your_domain send requested_puzzle_name For example, if your net address is "mickey@disneyland.com", to request "geometry/duck.and.fox.p", send the message: return_address mickey@disneyland.com send duck.and.fox To request the index, use: send index To request multiple puzzles, use several "send" lines in a message. Please refrain from requesting the entire archive via email. Use FTP. FTP The entire archive is also accessible via anonymous FTP, from any site which maintains archives of the newsgroups news.answers or rec.answers. The file part01 contains the index. The remaining files contain alternating problem text and solution text for all the puzzles. Some FTP sites are: North America: ftp://rtfm.mit.edu/pub/usenet/news.answers/puzzles/archive ftp://ftp.uu.net/usenet/news.answers/puzzles/archive ftp://mirrors.aol.com/pub/rtfm/usenet/news.answers/puzzles/archive ftp://ftp.cis.ksu.edu/pub/mirrors/news.answers/puzzles/archive Europe: ftp://ftp.cs.ruu.nl/pub/NEWS.ANSWERS/puzzles/archive ftp://src.doc.ic.ac.uk/usenet/news-faqs/news.answers/puzzles/archive ftp://ftp.uni-paderborn.de/doc/FAQ/rec/puzzles Asia: ftp://ftp.hk.super.net/mirror/faqs/puzzles/archive GOPHER From the global home page, the menu choices to access the archives at "cs.ttu.edu" are: North America/USA/Texas/Texas Tech University, Computer Sciences /Entertainment/Games/Puzzles To access "uni-hohenheim.de" your menu choices are: Europe/Germany/University of Hohenheim/Lots of Interesting Stuff /FAQ Frequently Asked Questions/rec/puzzles/archive WAIS wais://xraysgi.ims.uconn.edu:8000/rpa WEB http://xraysgi.ims.uconn.edu/searchform.html By keyword as well as subject. http://einstein.et.tudelft.nl/~arlet/puzzles/index.html Partially HTMLized. http://www.nova.edu/Inter-Links/puzzles.html http://xraysgi.ims.uconn.edu/others.html A list of other sites (maintained by David Moews) THE rec.puzzles ORACLE This is a group of rec.puzzles regulars, who are familiar with the rec.puzzles archive, and who will find your answer there if it exists, or maybe compose an original answer if they are interested enough! At any rate, they promise to respond to your question within two days, and perhaps save you the embarrassment of posting a well-worn question. They will respond within two days even if they do not know the answer to your question. To query the rec.puzzles oracle, send email containing your question to the following address: puzzle-oracle@questrel.com