applets.

Overhang

← All apps

About Overhang

Stack identical bricks at the edge of a table. How far out past the edge can the pile reach? With four bricks the top one can sit entirely beyond the table. With enough bricks it can reach as far as you like, and with bricks used as counterweights it can reach much further than the famous textbook answer.

PLAY: drag bricks out of the toy box and stack them. After every move the page checks, properly, whether the stack can stand; if it can't, it falls. The rule along the table edge reads the overhang in brick lengths. STEADY HAND holds the stack while you build (some good stacks can't be built one brick at a time without help), and LET GO lets it stand or fall. SHOW FORCES draws the pushes between the bricks. Light bricks hold up the furthest brick; darker ones are counterweights.

HARMONIC is the classic stack, one brick on each, built the way it is usually taught. BEYOND shows stacks that beat it, and some that look as if they should stand but can't. CHALLENGES sets four puzzles. COPY LINK copies a link to what you're looking at, including a stack you built in PLAY.

THE STORY

The one-brick-on-another stack appears in physics and engineering textbooks from the middle of the 19th century. J. G. Coffin posed the question in the American Mathematical Monthly in 1923, and Martin Gardner made it famous in his Scientific American column in November 1964, calling the fact that the overhang can be as large as you like the infinite-offset paradox. It has been called the Leaning Tower of Lire and the book-stacking problem.

The trick is that each brick, counted from the top, sticks out 1/2, 1/4, 1/6, 1/8... of a length past the one below. That adds up to half the harmonic series, which grows without limit, but very slowly: one length takes 4 bricks, two take 31, three take 227, and ten take 272,400,600.

For a long time that was taken to be the best you can do. It is the best you can do if the bricks form a single tower, one brick on each and nothing used as a counterweight. In 1979 Ainley found that four bricks can reach (15 − 4√2)/8, about 1.168, using one as a counterweight. Mike Paterson and Uri Zwick then searched by computer for the best stacks of up to 30 bricks, and showed that stacks built like a brick wall reach out in proportion to the cube root of the number of bricks, far beyond the harmonic stack's logarithm. Their paper "Overhang" appeared in the Monthly in January 2009. Later that year Paterson, Yuval Peres, Mikkel Thorup, Peter Winkler and Zwick showed in "Maximum overhang" that the cube root is the true limit: no balanced stack of n bricks reaches 6 times the cube root of n. The two papers shared the MAA's David P. Robbins Prize in 2011.

HOW IT WORKS

The page uses Paterson and Zwick's definition of balance. The bricks are identical, 1 long, perfectly rigid and smooth (no friction), and lie in rows. Two touching bricks can push on each other but not pull, and with no friction every push is straight up or down. A spread of pushes along a contact can always be replaced by two pushes, one at each end of the contact. The stack is balanced if some choice of these end pushes, none of them negative, makes the upward and downward forces on every brick add up to its weight and makes their turning effects about its middle cancel. That is a linear programme, which the page solves with its own simplex method after every move. To show where a stack would fail, the same programme is allowed to pull and asked to pull as little as possible: the bricks it has to pull on are the ones that come away.

The model of balance is exact for ideal bricks; the page solves it in ordinary computer arithmetic, so it allows a pull of up to a hundred-billionth of a brick's weight to cover rounding. The fall is only a picture: the bricks that come away tip about the edge they lean on and drop, and the rest are checked again. Stacks that reach the furthest are always right on the point of tipping (the harmonic stack is balanced exactly on every edge), so here "balanced" includes "just balanced". Real bricks have rounded edges and friction, and a real stack needs a little margin.

The best stacks for 2 to 7 bricks in BEYOND are this page's own: random stacks improved step by step by solving linear programmes, then checked with the balance test. They match Paterson and Zwick's published values for each of those sizes (for 8 bricks the same search fell short, so the page doesn't offer one); theirs come from searching every arrangement, and they say they are confident of them but have no formal proof. Values for more bricks are theirs, labelled as published.

FURTHER WATCHING AND READING

The rule along the table is marked in brick lengths, 0 at the table edge.