Well I'll butt in on this one...Sorry.
The following will not help anyone make better knives
In terms of bonding, I'd say that is better to think of a spectrum or continuous range of possibilities instead of discrete boxes that represent the extremes that everything must be pigeon-holed into.
Table salt is a 3-D array of ions held together by ionic attraction.
Graphite is a stack of strongly covalently-bonded 2-D sheets of carbon atoms held together in the pile by weaker non-covalent bonding. It's very weakly held together in one dimension. which allows for easy slippage of the sheets--that accounts for its lubricity.
Diamond is a 3-d array of covalently bound carbon atoms.
Polymers can be essentially infinite strings of atoms held together by covalent bonds. Sometimes they branch and connect to other strings sometimes they don't. Some polymers conduct electricity down their length through non-localized bonds--this means that electrons can "hop" fron one atom to another. Sometimes linear polymers lie in parallel arrays and electrons can hop across the strands, but not down the length. If electrons can easily move from atom to atom over many atoms, instead of sticking to one atom or a bond between two or more atoms, that is conduction. There are conducting organic polymers that contain no metals.
All of these structures can be so big that they surely stretch beyond the usual conception of "molecule". Some might be called "1-D" wires" or nonisotropic conductors since they have mobile electrons like metals, but they can move only in one or two directions. Can something be a metal in one dimension, but not another?
Now think of a small, rusty bit of iron--really small. So it's a chunk of iron except the outside is covered with oxide. How small before it is a molecule? 10,000 iron atoms? 1,000, 100? What about a cluster of 20 iron atoms stablized by bonds to other elements as well as having iron-iron bonds? If a molecular orbital is shared by all iron atoms and partially occupied, how close to a metal is it? The valende elctrons are free to move about all the iron atoms. Lots of inorganic chemists and materials scientists work with such clusters of atoms as a model for trying to understand some properties of bulk metals, alloys, and other materials.
It all boils down to how electrons distribute themselves around atoms, and whether it is energetically favorable to be localized near a single atom or few, or to spread themselves over a larger volume. Most of the electrons of larger atoms like iron stay localized around the atom, whether they are in complexes of 1 metal atom or a chunk of metal. Simplistically (but good enough for this discussion), the highest energy electrons or valence electrons participate in bonding of any kind, whether localized, or completely delocalized as in the conducting electrons of a metal.
So, don't throw out what you know about "molecules", the basic interactions of atoms and electrons underlies the entire business of matter, unless it's high energy plasma obtained by heating stuff to zillions of degrees and ripping atoms apart. It just gets wierd when things are BIG,and maybe nobody told you.
It also gets more complicated when transition elements like iron with partially occuped d-orbitals are involved, which many never encounter in much detail outside of specialized courses.
With a lot of physics, quantum mechanics, and more brains than I have, a crude band theory of metals can, and has been be derived from the molecular orbital theory
extended to infinite arrays of regularly spaced atoms. One way to think of it is that when a molecule gets big enough and crosses the fuzzy line to bulk material, the energy levels and volumes that electrons that electrons effectively occupy are quite different than those of molecules--They are delocalized over a large volume and occur in "bands" with near identical energy. There are energy gaps between the different bands and the bands are made up of many very,very closly spaced energy levels, spaced so close it is effectively a continuum within a band. (It's a continuum because the array of atoms is so big--the energy between levels is smaller than ambient thermal energy--no I will not and cannot attempt to discuss super-conductivity--'nother kettle of fish.)
Think of a bulk material as a "super-molecule" with these bands (big "fuzzy" "super-molecular-orbitals" that encompass the whole chunk of metal) instead of discrete molecular obitals. If all the valence electrons occupy a band that is completly filled, no electron movement is possible due to the Pauli exclusion principle and it is an insulator. If an empty band is slightly higher in energy above a filled valence band, then it is easy to energetically promote electrons to the empty band, which then is partially filled--that's one way to view a semi-conductor, since the promoted electrons are free to move about the "super-molecular-orbital" made up of a continuum of closly spaced energy levels that represent the unfilled valencies of all the atoms, but they have to be kicked up there. If the valence band is partially filled, then it's a conductor--the electrons can move without any addition of energy. If the lower part of an empty band overlaps the upper part of a filled (or obviously unfilled) valence band it's also a conductor. The larger the number of unfilled, energetically accessable close-energy levels the better the conductor.
A simple treatment would use an periodic array of atomic orbitals, a more complex one would use a periodic array of symmetry-optimized hybrid atomic orbitals that reflect the symmetry of the lattice (If you don't know what that means, don't worry, but don't ask!). Gross picture is similar.
With a big enough piece of metal, there will be defects in the lattice. Bigger and it will be polycrystalline--a bunch of smaller crystals firmly bonded together, but not all in line or parallel. Single crystals of metals exist, but they are very,very tiny. Then add inclusions of different morphology (structure compsed of same atoms), or chemical composition bonded together (like pearlite). Then worry about how big they are, whether there is a grain, etc. To me, these are the things that is metalurgy.
In ferrite, I would suspect that the carbon-iron bonds are made between a symmetry-optimized subset of of the Fe (d3) (and maybe s4) orbitals and the (s2)3(p2) C hybrid orbitals. In other words, the valence band is split into two--part of the Fe electrons are in a band that is used for bonding to C (and maybe Fe) The remaining Fe valence electrons are in one or more other bands. Depending upon the spacing of these bands, some or all may overlap. If the electrons that constitute the Fe-C bonds are lower in energy than the other valence electrons, separated from the remaining valence electron bands by an adequate energy level and totally filled, then they will not be conduction electrons. If all of the above don't apply, then they will be conduction electrons, and "metallic". That's how I see it anyway.
Anway, just wanted to say what anybody knows probably isn't
wrong. The basics still apply ... theoretically. It's just gets a lot more complicated when things get big, bigger than what is commonly thought of as a "molecule". Clearly, as things get big enough, or alloys get complicated enough, it's just too much work to follow this approach and useful results can be obtained by simpler or more empirical methods, such as considering a pool of "free electrons", and it is true that the concepts of bonding taught in basic chemistry and organic chemistry are not extended to materials as I have tried to describe--with-out that extension, an inacurate picture results, much the same a describing astrophysics without using relativity. For real life metalurgical problems I suspect follwing the approach described here might be like describing the dumping of sand from a dumptruck by using the Newtonian dymanics for each individual grain of sand.
But the places in bewtween the extremes are where some fascinating new stuff is being developed, like how many atoms does it take to make a transistor, and the field of nanotechnology. Also think of the advances made recently in weather modeling and climate prediction made possible by combining fundamental priciples with massive brute force computing--viewing real life problems from this angle may be possible one day.
Sorry to ramble , I'm just a dumbazz chemist,waxin philosophical after a few beers. Apologies if it makes no sense. But in theory, it's possible to skin the cat starting at either end.
OK Mete, rip me apart!
Back to normal programming.
That does make things perfectly clear!