
Contents 

Now that this computer analysis has been completed, in which all 35.5 billion 6piece burr assemblies were analyzed, is there anything left to analyze in the way of 6piece burrs? Are there any more interesting designs to discover? The answer is, of course, "Yes". In some areas the computer analysis is incomplete. There are also some situations which are beyond the capabilities of the current computer programs.
Interesting 6piece burr designs which were overlooked or not included in the computer analysis are broken into 3 categories:
In this category are questions which could be answered by computer programs similar to the ones used in the analysis, but were not included because of time constraints or other reasons.
In this category we consider 6piece burr assemblies which are made up of standard pieces, and so are included in the 35.5 billion assemblies analyzed, but can come apart in unusual ways. The program is limited to movements which are linear in one of the orthogonal directions and are a multiple of the cube width in length. What about 1/2 moves, slant moves and twists? Can these nonstandard moves be used to advantage to disassemble one of the standard 6piece burr assemblies?
The "Weave" in the examples section is an example of a cubic
buildup assembly which cannot be disassembled with the "standard"
moves above, but which will fall apart easily if the pieces can be
moved in other ways. The computer programs used in the analysis are
incapable of determining that this puzzle comes apart. Are such
designs possible within the confines of a 6piece burr?
Diagram of The Weave
I posed this question in 1986 (see [5]). A key restriction was that any such move be completely "legal". By this I mean that if the pieces were cut perfectly from a rigid material with no tolerance or extra space whatsoever, then the pieces can be physically separated using the rotations.
At the time I asked this question regarding the 6piece burr, I
thought the answer was probably "no", but I had not really tried to
construct such a design. When the Computer's Choice 4Hole design was
discovered in 1988, I noticed there was a twist movement in the
puzzle. The twist is equivalent to rotating a square within 3 squares
the same size in the shape of an 'L'. See figure below.
In theory, the twist is illegal. In practice, the twist can be made
even with wellmade pieces because the sides of the 'L' hole are different
pieces, so that the 'play' in the pieces is magnified. Although the
move is illegal and does not help the disassembly process if it is
made, this discovery made me realize the possibilities for similar,
legal movements, and I started investigating assemblies with such
movements.
The first design I came up with required a fullylegal twist for disassembly, but unfortunately the solution was not unique (there were different ways to assemble the puzzle which had standard solutions). Imagine my surprise when, about two weeks later, I received a design from Peter Rosler (GERMANY) with the identical twist move in a very similar design. This design was also not unique. By this time, I was running programs on all similar assemblies looking for one with a unique twist solution.
After doing an exhaustive search of about 2,000 assemblies, I
found 2 unique designs, one of which is now called "The Programmer's
Nightmare". Running BURR6 on this design produces "102 assemblies
with no solution". One of the assemblies has the twist move which was
originally planned. However, there are 3 other assemblies which also
appear to have twist moves that will result in disassembly. In two of
these assemblies the twist moves are equivalent to the illegal twist
move described above, and are therefore not legitimate solutions.
With wellmade pieces, these moves cannot be made without noticing
that the pieces 'jam'. On first glance, the twist move in the third
assembly appears to be every bit as legal as the original solution. I
was disappointed when I first found this out, because an equivalent
additional assembly appears in all of the 6piece burr designs with
similar twist moves. However, on closer examination both moves have
additional restrictions in another layer of the assembly. The
original solution turns out to be completely legal, and the third
assembly illegal, but even with wellmade pieces there is little
evidence of the 'jam'. See the following pages for detailed
pictures:
Programmer's Nightmare Solution
Programmer's Nightmare Three Close Solutions
Programmer's Nightmare Rotation in Legal Solution
What about 1/2 moves, in which the moves are linear in the orthogonal directions, but the lengths of the moves are not integral multiples of the cube width? Can they be used to advantage?
In the Programmer's Nightmare, a 1/2 move was combined with a twist to disassemble the puzzle. However, I do not see how 1/2 moves, or other fractional moves, can be used by themselves to advantage. I cannot conceive of an assembly, built up from cubes in a standard 3d lattice, in which fractional moves can accomplish anything that full moves cannot. However, I have been wrong before!
What about slant moves, that is, linear moves in a direction other than an orthogonal direction?
In similar fashion to the 1/2 move argument, I cannot see how slant moves can allow disassembly of a cubic assembly where standard moves cannot do the same. However, a slant move, or a combination of a 1/2 move and a slant move, can replace a number of repeated standard moves that 'jog' a piece in a slanted direction. Hence, the slant move can be used to decrease the number of 'moves' in the solution, but disassembly can still be achieved with standard moves.
In this category are 6piece burrs in which the pieces are not restricted to the cubiccut pieces of Figure 2. We can use slant cuts, which may require that two pieces be moved simultaneously in different directions. We can use rounded crosssections, resulting in twisting moves which are completely legal for these pieces. There are an infinite number of possible piece shapes, and perhaps an infinite variety to the possible uses of such pieces.
A few examples:
Stewart Coffin has recently been experimenting with 'inclined' 6piece burr designs. In some of these designs the crosssections of the pieces are rhombic and in other designs the pieces do not meet each other at right angles. Some of these designs can be analyzed by the same computer programs used in standard 6piece burr analysis, but for other types of designs these programs are useless.
Contents 

