Your pictures were a big help in understanding what you are doing. I'm delighted to see you experimenting with this. However, I must say that a little studying would go a long way toward improving your setup. People have been building motors and mechanical inverters (commutators) for a very long time. They have learned a lot about them. This knowledge could be applied to what you are doing.
Recognize that all motors are really AC motors. The windings always have AC in them. What people call a "DC motor" is really an AC motor with an internal DC-to-AC inverter. This inverter can be mechanical (commutator and brushes) or electronic (a "brushless DC" motor).
The kind of motor you are making is commonly found in toys. They have a rotor with just 3 windings, and a 3-section commutator to generate the 6-step 3-phase AC waveforms to run it. This design makes pretty crappy motors, which is why you only see them in cheap toys.
In these small motors, arcing isn't a problem due to the low voltages. When designers want higher voltage motors (above about 30v), they found they had to increase the number of rotor windings and commutator bars. They arranged it so the voltage between each bar was less than 30v. At 120v for example, you might have 4 windings and commutator bars in series between the brushes, so each has only 30v across it.
If you try to switch high voltage DC, particularly with an inductive load like a motor winding, you're going to have trouble. You might discover some new solution (aluminum brushes in oil, etc.) but I have a strong feeling that anything that simple has been tried and rejected. After all, we've had some very bright people working on the problem for 100+years.
Look at switch ratings, and see how they change as a function of voltage, current, and AC vs. DC. For example, here is power relay with a pretty complete data sheet showing its life and performance with different loads:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=PB918-ND
Notice that:
- Contacts in series add to the voltage rating (2 in series have 2x the voltage rating, 3 in series has 3x etc.)
- Voltage rating for DC is far lower than AC (440v at 16amps AC, 90v at 16amps DC)
- The higher the current, the shorter the life (10^7 cycles at 1 amp, 10^6 at 4 amps, 2x10^5 at 16 amps etc.)
- DC voltage rating is strongly affected by current (one contact rated 30v at 16a, 50v at 2a, 100v at 0.6a, 300v at 0.25a)
And these are with a resistive load! Arcing is even worse with an inductive load.
This pattern is true for just about all contacts, though the exact ratios will vary. This means that the physical size of the switch you'd need to switch high DC voltage and current for a large motor would be huge! If it had to survive any length of time, it may have to be bigger than the motor itself!
You have to "beat" this problem some other way than brute force.
For example, until the 1950's they built inverters with vibrators. This is basically a relay wired like a "buzzer". Its contacts generated AC to run a transformer, motor, etc. To make it last, they added a "buffer condenser" (capacitor) to resonate the load's inductance with the vibrator's operating frequency. When you get it right, the current happens to resonantly ring to zero just as the vibrator's contacts open; thus arcing is minimized.
Another example is the one I mentioned; driving a synchronous motor with just the right field strength so its current falls to zero just as the switches open.
Still another is a current-fed inverter, where you have a large inductor in series with your DC supply. The inverter switches then momentarily
short the windings rather than open them, preventing the high voltage "kick" that triggers the arcing.
These are old techniques. You'll find them in old textbooks. Start by learning what others have done to deal with the problem, and then work on your own ideas. Otherwise, you'll just be "re-inventing the wheel" (repeating things that have already been tried and failed).
The fundamental problem is excessive arcing as you found. The windings are inductive, and interrupting current in an inductive load causes severe arcing.
