Thought I'd type a bit about this
So you want to improve your metal? Well first that depends on what it is. Then you need to know what to do to it, and then you need to know how to do it!
This will cover fluxes, slags and melt chemistry.
Aluminum isn't that easy to twiddle because it's so reactive. An attempt to burn out impurities will burn the most reactive metal, which in most cases is the aluminum itself. You might be able to use fractional crystallization and pour off other impurities, but that's difficult at best, and considering the usual use of grain refiners, unlikely to work very well!
Magnesium, and other reactive metals
Magnesium can be burned out because it is more reactive than aluminum. All the alkali and alkaline earth metals can be removed this way, but magnesium and lithium are the only ones in use as alloying elements (and a high-strength lithium alloy is probably worth more as stock than as scrap). Hydrogen, not an alloying component but a nuisance gas dissolved in the melt, can also be removed by oxidation. The most effective oxidation is with chlorine gas, often sourced from an unstable salt like calcium hypochlorite, which breaks down when submerged in the hot melt. Because aluminum chloride is unstable at this temperature, the chlorine will react with other elements in preference to the aluminum; in order, lithium, magnesium and hydrogen will be consumed, forming the respective salts (or in the case of hydrogen, acid gas). Beware of excess unreacted chlorine gas, particularly common with low-quality home made hardware.
Zinc has a low boiling point of 1665°F and, though it is very soluble in its alloys (brasses are a good example, many melting in excess of 1665°F), it has a high vapor pressure and is prone to evaporating from them. I don't know how the aluminum oxide surface layer or a flux cover would affect the evaporation. It would likely take several hours at yellow heat to remove much, and by the nature of the process, you can never remove all of it. Fortunately zinc has little effect under 5%, although I haven't tested it in combination with silicon.
Silicon alloys are ideal for casting: they have high strength right out of the mold, fluidity is high and some are particularly wear-resistant. Unfortunately, average scrap aluminum is deficient (made for extrusion or rolling rather than casting) and silicon to add is hard to come by. I'm working on a small smelter to produce silicon, to alloy with aluminum and produce pound quantities of 50% master alloy. More accessible attempts such as reducing silica with aluminum metal using a molten salt "solvent" have met with poor results, contaminated with iron (from the crucible).
The typical flux for aluminum is a chloride salt melt, mixed in the right combination to give a melting point less than aluminum's. Right on the edge is 41%wt NaCl, 59%wt KCl. 5% of a fluoride, such as CaF2, can be added to improve action. CaCl2 (anhydrous) can be mixed with NaCl, KCl or both, in the correct proportions (I don't have the ratios on hand) to give an even lower melting flux, possibly suitable even for fluxing zinc alloys. I'm told calcium and fluoride ions improve the performance of the flux, so making use of CaF2 and CaCl2 would be advantageous.
The amount of flux used varies; there is no "correct" amount. To clean up a relatively clean melt, you might throw an ounce or two on top, wait for it to melt, stir and be done with it. To break the apparent surface tension of metal and oxide you might do the same for a much dirtier melt. To completely seperate metal and oxide or provide an oxidation cover, you can melt a pound or two of flux over the metal, stir and free the oxides from the metal. After a proper treatment like this, the oxide appears dark and floats in the salt like fine sand in water.
Because water vapor appears to be soluble in molten salt, hydrogen may be present. Degassing, either by chlorination or bubbling inert gas (nitrogen, CO2, argon, etc.) should be performed before pouring. Hydrogen may also be introduced by disturbing the oxide layer (of unfluxed melts) by incessant stirring, whether in crucible or reverberatory furnaces.
Though most copper alloys work over wide ranges of composition, some elements are detrimental or mutually exclusive to the properties of it.
Aluminum bronzes are some of the strongest copper alloys available, but aluminum and common bronze elements (particularly tin and lead) are mutually exclusive. (I haven't tested this, but I haven't seen any alloy listed with aluminum and tin or lead together in any more than trace quantities.) As with aluminum alloys, you can't remove lead or tin from aluminum bronze because they are much less reactive than aluminum. However, aluminum as an impurity in conventional bronze alloys can be removed by oxidation. Instead of adding a gas, such as air or oxygen, the high temperature lets us use metal oxides instead. The dissolved aluminum reacts with the oxide, releasing the metal (which dissolves in the melt) and forming aluminum oxide which floats to the top and combines with the slag layer.
Zinc evaporates just as with aluminum alloys, and unfortunately happens in melting brasses and bronzes containing it. In a half hour melt, at most 10% of the zinc content is lost, not much, though it may look like a lot because zinc oxide is a very white substance and makes a thick smoke. If you want to remove zinc, it can be oxidized (as with aluminum, any metal oxide less reactive will work; iron oxide, lead oxide, copper oxide, etc.) or evaporated.
Same goes for copper as goes for aluminum, except a little goes farther (3-6% as compared to 10-25% in aluminum) and is typically alone, or optionally with a few percent zinc, and lead for free machining. Silicon can be burned out but as it's usually a specialty alloy, I don't know why you'd want to.
Because copper is so unreactive, all alloying elements (except the precious metals, which are rarely present in scrap) can be oxidized, including zinc, manganese, iron, nickel, tin, lead and so on. The best way (untested) is to add copper oxide to an unfluxed melt. On stirring it in, it partially dissolves (copper metal has some solubility for its oxide and oxygen gas) and immediately reacts with the more reactive elements, oxidizing them. Add oxide until no further reaction occurs; depending on temperature, there will be about 2% of oxygen dissolved in the melt. Skim the oxides as best you can; lead oxide will be liquid and may flux the other oxides. To remove the oxygen, add ground charcoal or 3-5% worth of zinc (if you don't mind a small amount of zinc in the final product) and add flux.
The standard flux for standard copper alloys is composed of silica, which because it has a high melting point, is fluxed with oxides like soda, lime and boric oxide. This glass melt adheres to and dissolves oxides present in and on the melt, cleaning it. It also provides a cover (if enough is used), protecting the melt from oxidation and to some extent, reducing zinc loss. The perfect slag for copper is one which is viscous enough to be skimmed with a rod before pouring (or which adheres to the crucible rather than pouring out with the metal), yet thin enough that it flows just as the metal starts melting, protecting it right away, and that it readily absorbs oxides in the melt, keeping it clean. You might start with a base of borax, silica, lime and optionally, soda, magnesia and other fluxes like zinc oxide and ferrous oxide, depending on the composition of the melt (for instance, zinc in the melt will react with iron oxide, adding iron to the melt and removing zinc). After testing this, you can add more flux to thin the mixture or silica or alumina (clay is a source of both) to thicken it.
For exotic alloys like beryllium and aluminum bronzes, these metals will react with a conventional slag. One foundryman told me potassium chloride (KCl) is used in these situations, but I've tried it before and it evaporates way too fast to be any possible benefit at all. (Like alcohol, chloride salts have a high vapor pressure and evaporate quickly, even below the boiling point.)
Occasionally, you may encounter hydrogen in your copper alloy melts. Supposedly, this can be removed by oxidation, as with aluminum alloys. I don't recommend trying chlorine gas (or a generator like calcium hypochlorite) with something as hot as copper, however, copper chloride (especially cuprous chloride, CuCl, which does not absorb moisture, unlike cupric chloride) should be useful for sourcing free chlorine. Similarly, copper oxide supposedly can be used, as hydrogen is more reactive than copper and will reduce the oxide to metal, releasing water vapor (steam). The problem with using copper salts is alloying elements, which are more reactive than hydrogen, should in theory react in preference to the hydrogen.
Though not encountered much in the backyard foundry, I will still add some here on iron and its alloys. This will be a bit disjointed because there are so many alloys that are all suitable endpoints; there are fewer "wrongs" compared to the alloys above, which I also have experience with.
Iron comes in two nearly distinct animals: steel and cast iron. Steel has carbon typically less than 1% and certainly no more than 1.5%, and up to 30% metal alloy content. Cast iron has more than 2% carbon and typically 1 to 3% silicon.
Most of the elements used in alloying iron are more reactive than it, so its chemistry is pretty rich. It also melts at a temperature where everything is loose and bumping around; if reactions can happen, they will; unlike aluminum, which despite its high reactivity, only lazily reacts with metal oxides (unless powdered).
Carbon is the essential component of all iron alloys. Too much and it becomes hard and brittle, but add some silicon and you are now in cast iron territory. Just the right amount and it is hardenable to great strength. "Too little" and you just get a soft alloy, mild steel, with good ductility and malleability.
Silicon tempers the carbide-forming nature of iron and carbon when cooled quickly, allowing a final product with elemental carbon dispersed in the alloy, leaving a soft, very machinable, but relatively weak and brittle product. By reducing silicon and controlling solidification, or adding such tweakers as copper, carbides can be reintroduced as pearlite and the strength goes up. With large amounts of silicon and carbon (up to 5% each), silicon carbide crystals begin to form, producing a strong, heat-resistant alloy. Alone, you get silicon steel which, though hard and somewhat brittle, is valuable for its good magnetic properties.
Manganese is to iron what zinc is to copper. First, it deoxidizes the metal due to its high reactivity (comparable to zinc's). Second, it strengthens and gives impact resistance, with alloys up to 10% being used for bulldozer shovels, rock crusher jaws and ballistic armor.
Chromium and carbon go together well, in fact too well. Just 1% chromium added to a steel with less than 0.4% carbon will cause it to harden like a 1% steel! This and other similar carbide-forming elements are what allow slower-hardening alloys (oil and air quench hardened, rather than water quenched), without exceeding the 1.5% carbon limit of steel. These warp less when hardened, saving time and money in final grinding and finishing. Large amounts produce a protective oxide layer of Cr2O3, much as aluminum is coated with a protective layer of Al2O3, imparting corrosion resistance.
Of course, 10 to 20% chromium would make an impossibly hard metal to deal with no matter what the carbon content! This is where nickel comes in: nickel does not form carbides as chromium and iron do, and in fact prohibits it enough that the carbon content can be within realistic limits while the metal still being ductile. Such alloys are called stainless steels, because the chrome oxide layer makes them stain less than an average steel alloy. (Stain *less*, not stainproof! It will still rust, particularly under corrosive conditions, and the surface needs to be passivated after working or else it may rust even under ordinary conditions.)
Nickel is also used, up to 55% (that is, more than the iron content!), in cast irons known as Ni-resist.
I don't know what use copper has in iron alloys, but as mentioned, up to 2% copper will prefer pearlite to ferrite in cast irons, imparting strength and reducing ductility. This works on gray or ductile cast ironsm (which are essentially the same).
Now here's an odd one. A look at the phase diagram shows these two do not interact at all, in fact if you could have both molten at the same time (magnesium ordinarily would boil off), the magnesium would float on top clearly as a seperate layer like oil on water! And yet, a very small amount (typically around 0.05%) makes a very dramatic change in the character of cast iron: instead of forming flakes of graphite, it crystallizes as approximately round nodules instead. The iron around these nodules can be deformed now, without breaking! Gray iron has suddenly become ductile iron, through a subtle but very significant change in the microstructure. Cerium I'm told also works in the same way, and calcium can be added to get a halfbreed between flat flakes and round nodules (seminodular cast iron).
Another advantage of reactive metals like Mg and Ca is they form very high melting point salts with sulfur and phosphorous, which generally embrittle the iron. The result is, in an iron which contains 0.05% S, you need to add 0.1% Mg to get the required 0.03 to 0.05% Mg content. Magnesium also evaporates from the melt, just as zinc from bronze.
That's all I'm edit/adding for now, I need a break from the keyboardski...