Chemistry · Essay 1 · 14.essays.01

The Chemical Bond — Many-Body Phenomenon vs Effective Two-Center Object

Why the Lewis bond and the molecular orbital aren't the same thing — and what either one is, ontologically.

The Question Behind the Question

Ask a working chemist what a chemical bond is, and you will get an answer that depends on what they are about to do. If they are about to predict the geometry of a small organic molecule, the bond is a shared pair of electrons sitting between two nuclei. If they are about to explain why O $_{2}$ is paramagnetic, the bond is a particular pattern of occupation of delocalized molecular orbitals. If they are about to compute a binding energy, the bond is whatever the wavefunction is doing — the word is mostly absent from the calculation, replaced by a many-electron Hamiltonian and a self-consistent field. If they are writing a teaching answer for a first-year student, the bond is the dash between the letters in H–H.

These are not glosses of the same thing. They are different ontologies, and chemistry as a working discipline has not picked between them. It uses each in the regime where it predicts, and it does not pause much over the contradictions. The contradictions are real. A bond that is a localized pair of electrons in one description and a property of a delocalized many-electron wavefunction in another is not the same kind of object across the descriptions. It cannot be. If we are honest about what chemistry is doing when it teaches and predicts and synthesizes, we have to admit that the bond is being used as several different things at once.

This essay is about that situation. Not about resolving it — the resolution, if there is one, is not in the working chemist's hands and probably not in anyone's — but about seeing it clearly. The chemical bond is the place where chemistry has its hardest ontological problem, and chemistry has chosen, sensibly, not to solve it but to use it. Understanding what the bond is requires understanding the pictures, where each picture comes from, what each picture predicts, and what each picture cannot say. It also requires understanding why a discipline can run on contradictory pictures for a century without breaking.

The companion units to this essay are 14.05.02 MO theory for homonuclear diatomics, which builds the molecular-orbital picture for the diatomic case; 15.04.02 SN1 vs SN2 mechanism, which is the cleanest case where the bond's existence becomes a question about time; and the philosophical essays in §20, particularly the ones on reflexivity and on what counts as a real object inside a working scientific practice. The mathematical companion is 03.09.02 Clifford algebra, which gives the symmetry language inside which atomic orbitals (and therefore bonds) live. None of these resolve the question. They are different angles on it, and the angles are the point.

The Lewis Bond

The picture most chemists meet first is older than the modern atom. Gilbert Newton Lewis published it in 1916 in a paper called The Atom and the Molecule ^{[Lewis 1916, pending+pointer]}, in which he proposed that the chemical bond is two electrons shared between two atoms, and that atoms tend to be most stable when they have eight electrons in their outer shell — the octet rule. He drew bonds as dots between letters, and he formalized in those drawings something practicing chemists had been doing intuitively for half a century: representing molecules as graphs, with atoms as vertices and bonds as edges.

This is, on its face, an astonishingly successful picture. Almost the entire formal apparatus of organic chemistry — structural formulas, oxidation states, the prediction of molecular geometry from electron-pair repulsion (VSEPR), the curly-arrow language of mechanism — works inside the Lewis picture. A first-year chemistry student who can draw Lewis structures can predict the geometry, polarity, and approximate reactivity of essentially every molecule made of main-group elements with surprising accuracy. The picture is not just a teaching tool. It is the conceptual scaffold on which the working knowledge of synthetic chemists is built. When a chemist plans a synthesis, they think in Lewis structures. The arrows they push are Lewis-bond arrows.

Lewis's 1923 textbook Valence and the Structure of Atoms and Molecules ^{[Lewis 1923, pending+pointer]} consolidated the picture and made it the standard. By the late 1920s, the language of "shared electron pairs" was the lingua franca of chemistry.

But the picture is also, on inspection, wrong in specific ways. Three failures are worth naming, because they sharpen what kind of object the Lewis bond is.

The first failure is O $_{2}$ . The Lewis structure of dioxygen wants to put a double bond between the two oxygens, with two lone pairs on each — an octet on each atom. This predicts a closed-shell, diamagnetic molecule. The experimental fact is that O $_{2}$ is paramagnetic. It is attracted to a magnet. It has two unpaired electrons. No Lewis structure captures this. The Lewis picture, when applied to dioxygen, predicts the wrong magnetic ground state of one of the most common molecules in the universe.

The second failure is electron-deficient compounds, most famously diborane B $_{2}$ H $_{6}$ . The Lewis picture insists each bond is two electrons between two atoms. Diborane has 12 valence electrons, and there is no way to arrange them as two-electron-two-center bonds and account for all the connectivity. The molecule exists. It is stable. It has three-center-two-electron bonds — bridge bonds where two electrons are shared across three atoms — which the Lewis picture, taken strictly, cannot draw.

The third failure is aromaticity. Benzene cannot be drawn as a single Lewis structure. The two Kekulé structures (alternating single and double bonds) misrepresent the real molecule, which has six equivalent C–C bonds, each somewhere between a single and a double. Linus Pauling's response to this was resonance: the real molecule is a superposition of the two Lewis structures, neither of which is the truth alone. But "superposition of Lewis structures" is already an admission that no single Lewis structure is the bond.

What these failures share is structural. The Lewis bond is a localized two-center two-electron object. When the underlying reality is delocalized — when electrons are shared across more than two atoms, when the bond-pattern is not a definite count of paired electrons in definite places, when the magnetic behavior reveals unpaired spins — the Lewis picture loses contact with what the molecule is doing. The picture's strength and its weakness are the same thing: it commits, in advance, to a particular metaphysics of locality and pairing. Where the molecule obeys that metaphysics, the picture is excellent. Where the molecule disobeys it, the picture lies.

This is the first lesson. The Lewis bond is not the bond. The Lewis bond is a picture of the bond that happens to be predictive across a wide range of cases. The bond is whatever is making the picture mostly correct, and whatever is making it sometimes wrong.

Valence Bond Theory

Linus Pauling, in The Nature of the Chemical Bond ^{[Pauling 1939, pending+pointer]}, gave the Lewis bond its quantum-mechanical formalization. Pauling's project, beginning in the late 1920s and culminating in the 1939 book, was to show that the Lewis picture was not just an empirical regularity but had a foundation in the new quantum mechanics of Schrödinger and Heisenberg.

The basic move was this. An atom has orbitals — solutions of the Schrödinger equation for an electron in the field of the nucleus. When two atoms come close, the orbitals on one can overlap with the orbitals on the other. The bond is the overlap region: two electrons, one from each atom, sharing a region of space where both orbitals are simultaneously nonzero. The energy of the system is lowered when the electrons occupy this shared region, because each electron experiences attraction to both nuclei.

This is the Lewis bond in quantum-mechanical clothing. Two electrons. Two atoms. A localized region of shared space. Pauling's contribution was to make the picture quantitative — to compute bond energies, to predict bond lengths, to explain bond angles. He introduced hybridization (the mixing of s and p orbitals on the same atom to produce sp $^{3}$ , sp $^{2}$ , and sp hybrid orbitals that point in the right directions for the observed molecular geometries), resonance (the superposition of valence-bond structures to handle delocalization), and electronegativity (the quantitative measure of an atom's pull on shared electrons) as the working tools of the theory.

Valence-bond theory was, for the 1930s and 1940s, the dominant quantum-mechanical theory of bonding. It was the theory that the Pauling–Wheland school of thought built American structural chemistry on. And it remains the theory that organic-chemistry pedagogy uses, even when it doesn't name itself as such. When an organic chemist says "this carbon is sp $^{2}$ hybridized," they are speaking valence-bond theory. When they say "the bond is formed by overlap of an sp $^{3}$ orbital on carbon with a 1s orbital on hydrogen," they are speaking valence-bond theory.

The theory is, at its core, the Lewis bond made quantum-mechanical without losing the Lewis bond's locality commitments. The shared pair is real, the localized bond is real, the hybrid orbital is a real construction belonging to a real atom. Pauling's book is, in this sense, the most ambitious defense of the Lewis bond as a fundamental object that has ever been mounted. Pauling believed in chemical bonds. The book reads like a man who is confident, page after page, that what he is describing is a kind of thing in the world.

The trouble is that valence-bond theory, for all its predictive power, is not the quantum-mechanical theory that won. The theory that won is molecular-orbital theory, which begins from a completely different ontology.

Molecular Orbital Theory

Friedrich Hund, Robert Mulliken, and John Lennard-Jones, working in the late 1920s and through the 1930s, developed an alternative picture that dispensed with the localized bond entirely ^{[Mulliken 1932, pending+pointer; Lennard-Jones 1929, pending+pointer]}. In molecular-orbital theory, the question "what is the bond between atom A and atom B" is not asked, because the answer is: there isn't one.

The MO picture goes like this. A molecule, considered as a quantum-mechanical system, has its own set of orbitals — molecular orbitals — that extend over the entire molecule, not over individual bonds. These molecular orbitals are constructed (in the standard linear-combination-of-atomic-orbitals approximation, LCAO) by taking linear combinations of the atomic orbitals of the constituent atoms. The molecular orbitals are then filled with electrons according to the Aufbau principle, just as atomic orbitals are filled in an atom.

There is no localized bond. There is a molecular wavefunction, and within that wavefunction, certain orbitals are bonding (they have constructive overlap between atoms and concentrate electron density between nuclei) and certain orbitals are antibonding (they have a node between atoms and concentrate electron density outside the internuclear region). The bond order between two atoms is computed as half the difference between the number of electrons in bonding orbitals and the number in antibonding orbitals — a derived quantity, not a primitive one. The bond identity — the bond as a thing — has been dissolved into the molecular wavefunction.

The MO picture handles the three Lewis failures cleanly. O $_{2}$ is paramagnetic because, when you build its MO diagram, the two highest-occupied electrons land in degenerate π* antibonding orbitals and Hund's rule puts them in separate orbitals with parallel spin. The unpaired electrons fall out of the theory without any special pleading. Diborane's three-center-two-electron bonds are routine in the MO picture, which has no commitment to two-center localization. Benzene's six equivalent C–C bonds are the natural consequence of a delocalized π system whose lowest three MOs are filled.

Through the 1950s and 1960s, with Charles Coulson's Valence ^{[Coulson 1952, pending+pointer]} as a major bridge text and with the rise of computational chemistry, MO theory became the dominant theoretical framework. By the time Roald Hoffmann and Kenichi Fukui (and, before them, Woodward and Hoffmann) used frontier molecular orbital theory to explain pericyclic reaction selectivity in the 1960s, work that won them the 1981 Nobel Prize, the MO picture had become the language of theoretical chemistry. Valence-bond theory survived in teaching but was no longer the cutting edge.

And here the contradiction sharpens. We now have two theories. One says the bond is a localized two-electron object between two atoms. The other says the bond is a derived bookkeeping quantity computed from a delocalized many-electron wavefunction. Both are correct in their domains of application. Both make quantitative predictions. They cannot both be telling us what the bond is.

The Two Theories Side by Side

What does it mean for two theories to disagree about the ontology of their central object and both be predictively successful? This is the question that the philosophy of chemistry has spent the last forty years working on. It is not a question chemists themselves spend much time on, because the disagreement does not show up at the bench. It shows up when you try to say what chemistry is about.

Robin Hendry, whose work in the metaphysics of chemistry ^{[Hendry 2017, pending+pointer]} is one of the most careful recent treatments, frames the situation this way. Valence-bond theory and molecular-orbital theory are not really competing theories of the same object; they are different approximations to the full quantum-mechanical treatment of molecules, and they emphasize different features of the molecular wavefunction. VB theory builds the wavefunction starting from products of atom-centered functions and then introduces correlation through resonance among VB structures. MO theory builds the wavefunction starting from delocalized molecular orbitals and then (in the more sophisticated treatments) introduces correlation through configuration interaction or coupled-cluster expansions. In the limit of full configuration interaction — that is, in the exact solution — VB and MO give the same answer. They are mathematically interconvertible.

This is a real point, and it dissolves some of the contradiction. The two theories are different starting points for the same calculation. Each starting point has its own intuitions about what is primitive. Neither starting point is privileged by the underlying quantum mechanics; both are approximation strategies that emphasize different aspects of the exact wavefunction.

But the dissolution is partial. The question "what is the bond" is not the same as the question "what does the full wavefunction look like." If the bond is something the chemist points to and the calculation does not, then the bond is not in the wavefunction. It is in the picture the chemist is using to interpret the wavefunction. The two pictures are pictures of different aspects of the same underlying reality, but the bond — the object — is in the picture, not in the underlying reality.

Eric Scerri, whose work on the philosophy and history of chemistry includes both The Periodic Table: Its Story and Its Significance ^{[Scerri 2007, pending+pointer]} and substantial writing on chemical reduction, has argued that chemistry sits in an unstable relationship to physics: the rules of chemistry are consistent with quantum mechanics but cannot be cleanly derived from it without already importing chemical concepts (orbitals, bonds, atoms-in-molecules) that quantum mechanics, in its bare form, does not contain. Bensaude-Vincent and Stengers, in A History of Chemistry ^{[Bensaude-Vincent and Stengers 1996, pending+pointer]}, make a related historical argument: chemistry has, throughout its history, refused to dissolve into physics, and the chemical bond is the central object on which the refusal turns.

The philosophical literature here is not unanimous. There are reductionists who hold that chemistry is, in principle, fully reducible to quantum mechanics and that the bond is a derived concept of the underlying physics; there are autonomists who hold that chemistry has its own ontology, irreducible to physics, and that the bond is one of its primitive objects. The cleanest reductionist view says: there is no bond, there is only the molecular wavefunction, and the bond-talk is heuristic shorthand. The cleanest autonomist view says: the bond is a real chemical object, the molecular wavefunction is the physics-side description, and there is no obligation to reduce one to the other.

Most working chemists, I would guess, do not hold either of these clean views. They hold something stranger: that the bond is a real object for the purposes for which they treat it as real, and not real otherwise. This is not philosophical sloppiness. It is, I think, a sophisticated practice that the philosophy of science is only beginning to name.

When Does a Bond Exist?

A second angle on the question comes from chemistry's own mechanism literature. Consider the cleanest case, treated in detail in the companion unit 15.04.02 SN1 vs SN2 mechanism. A nucleophile attacks a carbon bearing a leaving group. In the SN2 mechanism, the bond from the carbon to the leaving group breaks at the same time as the bond from the carbon to the nucleophile forms. At the transition state — the highest-energy point along the reaction coordinate — the carbon is partially bonded to both the incoming nucleophile and the departing leaving group simultaneously. The transition state has, in the curly-arrow picture, fractional bonds.

What is a fractional bond, ontologically? In the Lewis picture, a bond is two electrons. You can't have one and a half electrons. The picture has no internal vocabulary for partial bonds. In the MO picture, fractional bond order is fine — it falls out of the delocalized wavefunction at the transition-state geometry — but then the question becomes whether the transition state has any bonds in the localized sense at all.

The SN1 mechanism makes the question sharper still. In SN1, the leaving-group bond breaks completely before the nucleophile arrives. There is an intermediate carbocation — a carbon with three bonds and a positive charge — that exists for some finite time before the nucleophile attacks. Between the moment the leaving group departs and the moment the nucleophile attaches, what is the bond count of the central carbon? Three, evidently, since the new bond hasn't formed and the old one is gone. But the carbocation is unstable; it is on its way to forming a fourth bond. Is the potential fourth bond, in any sense, a present feature of the molecule?

The question is not idle. It maps onto the question of whether the bond is a state of the molecule (something the molecule has, presently, as a property) or a predicted future (a pattern the molecule is moving toward). These are different ontologies. A bond as a present state is the Lewis-VB picture. A bond as a structural feature of the wavefunction that varies continuously with geometry is the MO picture. The MO picture handles the transition state cleanly because it does not require bonds to be present-state objects.

There is a third way to think about this, due to Richard Bader and his school: the atoms in molecules (AIM) analysis ^{[Bader 1990, pending+pointer]}. Bader argued that the electron density of a molecule — the scalar field $ρ (r)$ that says how much electron density is at each point in space — is a real, observable quantity, and that the topology of this scalar field defines real chemical objects. Atoms in molecules are regions of space bounded by surfaces across which the gradient of $ρ$ is zero. Bonds in molecules are bond critical points — saddle points of $ρ$ — connected to atomic basins by bond paths, lines of maximum density linking nuclei.

The AIM picture has a striking property. It defines bonds from the electron density, which is computed from the wavefunction, which can be computed from first-principles quantum mechanics. The bond is, on this view, derived from the wavefunction (so the reductionist is satisfied), but it is also a real topological feature of an observable density (so the autonomist is satisfied — the bond is not just a useful fiction). The AIM bond is a saddle point. It is there or it isn't.

A related approach, Natural Bond Orbital analysis (Weinhold and collaborators) ^{[Reed Curtiss Weinhold 1988, pending+pointer]}, goes in a different direction: it takes a delocalized MO wavefunction and projects it onto a localized basis of Lewis-like bonds. The NBO analysis answers the question "what would this molecule look like if you insisted on writing it as a Lewis structure?" and quantifies how much of the wavefunction the resulting structure captures. The leftover — the part that doesn't fit — is interpreted as delocalization, hyperconjugation, and other corrections to the localized picture.

These analyses are not new theories of bonding. They are interpretive tools applied to the existing wavefunction to extract bond-like objects from it. They are the working chemist's machinery for moving between the delocalized many-body description and the localized two-center description, depending on which one they want to reason in.

The Bond as Effective Object

Step back from the specific theories and ask the structural question. What kind of object is the bond?

The modern computational chemist's working answer is something like this. The fundamental description of a molecule is the many-electron wavefunction $Ψ (r_{1}, r_{2}, \dots, r_{N})$ , a function of all electron coordinates, governed by the Schrödinger equation with the molecular Hamiltonian. The bond is not in the Hamiltonian. The Hamiltonian contains nuclei, electrons, and Coulomb interactions; it does not contain bonds. The bond is a feature that emerges from the wavefunction when you analyze the electron density, or project onto localized orbitals, or examine the topology of $ρ$ , or compute bond orders by any of several conventions. It is a renormalized description of the underlying many-body physics — a description at a coarser resolution, in terms of effective two-center objects that capture most of the chemistry while suppressing most of the detail.

This is the same kind of structural move that physics makes constantly. Quasiparticles in condensed matter are not in the Hamiltonian; they emerge from collective excitations of the underlying many-body system, and they are excellent objects for predicting transport, magnetism, and superconductivity at energy scales below the bare-electron scale. Phonons are not in the Hamiltonian. Cooper pairs are not in the Hamiltonian. Effective field theories at the LHC scale do not contain the quarks and gluons of QCD as fundamental degrees of freedom; they contain pions and nucleons, which are emergent. The bond, on this view, is chemistry's quasiparticle: a stable, predictable, transferable feature of an effective description of the many-body wavefunction.

If this view is right, several things follow. The bond is real in the sense that it is a robust feature of an excellent effective theory, but it is not real in the sense of being a fundamental constituent of the molecular Hamiltonian. The Lewis bond is correct where the effective theory in which it lives is correct, and wrong where the effective theory breaks down. The MO description is closer to fundamental but is itself an approximation — the orbitals of MO theory are mean-field objects that ignore electron correlation. The "true" wavefunction has neither localized bonds nor independent molecular orbitals; it has a correlated many-electron structure of which both pictures are projections.

This is a satisfying view, intellectually. It dissolves the apparent contradiction (the pictures are different projections of the same underlying object) while preserving the working utility of each (the projections are excellent at what they project). It is the view that most computational chemists, asked to reflect, would probably endorse.

But it has a cost the philosophical autonomist will press. If the bond is just an emergent feature of the wavefunction, then the practitioner who pushes Lewis arrows is, in some sense, doing something that is not fundamentally about the molecule but about a particular projection of it. The arrows track the projection, not the thing. The autonomist will want to know whether the projection itself is privileged in some way that grounds the chemist's practice as more than instrumental.

The Structural Realist's Reply

There is a reply available, due in spirit to the structural realism of John Worrall, James Ladyman, and others ^{[Worrall 1989, pending+pointer; Ladyman and Ross 2007, pending+pointer]}, and developed for chemistry specifically by Hendry and others. The reply goes like this.

Even granting that the localized bond is an emergent, effective object — granting fully that there is no two-electron pair sitting in a definite region of space inside a benzene molecule — the patterns of bond properties across the millions of molecules chemistry has studied are so robust that something real is being captured by the bond concept. Bond lengths are transferable: a C–H bond is roughly 1.09 Å in essentially every organic molecule, varying by a few hundredths of an angstrom in ways that themselves correlate predictably with electronic environment. Bond energies are transferable: a C–C single bond contributes roughly 348 kJ/mol to the enthalpy of formation across an enormous range of molecules. Vibrational frequencies are transferable: a C=O bond stretches at 1700 cm $^{- 1}$ , give or take fifty wavenumbers depending on context, in every carbonyl-containing molecule.

These transferabilities are facts. They are facts about real molecules, measured in real laboratories, and they are not facts that the bare quantum-mechanical Hamiltonian volunteers. The Hamiltonian gives you a different molecule with a different wavefunction each time you change the atoms. The transferability — the structural regularity that says "C–H bonds are kind of the same across molecules" — is something that emerges when you slice the wavefunction in the right way. The bond concept is the name of that slice. It is not nothing.

The structural realist's claim is that the pattern of bond properties is more real than the bond as a substance. The bond is not a thing in the molecule, but the pattern is a real feature of how molecules behave, and the bond concept tracks that pattern with high fidelity. To deny the reality of the bond is to deny the reality of the pattern, which is empirically unsupported. To affirm the bond as a fundamental constituent is to overcommit — the bond is not in the Hamiltonian. The middle position, the structural-realist position, is that the bond is real as a structural feature and not real as a substance, and this is a coherent and defensible ontology.

This is, I think, the closest thing to a settled answer that the philosophy of chemistry has produced. It is not unanimous. There are still hard reductionists and there are still hard autonomists. But the structural-realist middle is the position that takes both the success of the Lewis picture and the priority of the wavefunction seriously, without forcing one to win.

What the Practitioner Does

Return to the working chemist. They use the Lewis picture when they are predicting geometry, mechanism, and reactivity. They use the MO picture when they are explaining magnetism, photochemistry, or spectroscopy. They use the wavefunction when they are computing energies and properties. They use the AIM density when they are arguing about whether a particular interaction (a hydrogen bond, a halogen bond, a stacking interaction) counts as a bond. They do not pause much over the contradictions, because the pictures are tools, and a working tradesperson does not need to defend the metaphysics of their hammer.

There is something here that maps cleanly onto the structure of the philosophical essays in §20. The reflexivity essay (20.essays.01) argues that what is disclosed depends on the depth at which it is being seen — that the seer and the seen are coupled. The chemical bond is a case in point. The bond as Lewis pair is what discloses to a Lewis-trained seer attending to mechanism. The bond as MO bond order is what discloses to an MO-trained seer attending to magnetism. The bond as AIM saddle point is what discloses to a density-analyst attending to topology. None of these seers is wrong. They are each seeing a real feature of the underlying molecular reality at the resolution of their picture. Different seer, different disclosure.

This is not relativism. The molecule is the same molecule. The wavefunction is the same wavefunction. The disclosed features are different, and the differences track the apparatus of the disclosing, not arbitrary preference. A Lewis-picture chemist who claimed O $_{2}$ was diamagnetic would simply be wrong, and any other chemist could check, by hanging O $_{2}$ over a magnet, that they were wrong. The pictures are constrained by the molecule. They are not free constructions.

What the practitioner does — the move that I think is genuinely sophisticated and that the philosophy of chemistry has only partially named — is to move fluidly between pictures depending on what they are trying to predict, while holding, in the background, the awareness that no single picture is the bond. The bond is what causes the pictures to coordinate. It is what makes the Lewis prediction agree with the MO prediction agree with the AIM density, when they all apply, and it is what makes them diverge, in characteristic ways, when one of the pictures is being pushed past its domain.

The practitioner does not need to name this awareness to operate from it. Most chemists could not articulate the structural realist position in two paragraphs. But they enact it, by using each picture without confusing it for the thing it depicts. The pictures are not the bond. The bond is what makes the pictures predictive when they are predictive. The chemist who knows this — who uses the pictures while knowing they are pictures — is doing a sophisticated thing that looks, from the outside, like just doing chemistry.

There is a closing move here that the philosophical essays would recognize. The pictures are not the seen. They are the apparatus through which the seen discloses. A practitioner who has internalized this can use the pictures fully, even passionately, even with the conviction that comes from years of seeing them work — and still know that what they are working with is a useful disclosure, not the underlying real. This is, I think, what the better chemists do. It is what the practice teaches over time, even when the practice does not name it. The bond is real enough to push arrows over and unreal enough to admit it is not in the Hamiltonian. Holding both at once is the working ontology of modern chemistry.

Closing

The chemical bond is the place where chemistry's relationship to physics is most exposed and most unresolved. It is also the place where chemistry's relationship to itself as a discipline — its working pictures, its teaching traditions, its everyday practice — is most visible. The Lewis bond is the picture that chemistry teaches first and reaches for most often. The MO picture is the picture that won the twentieth-century theoretical argument. The wavefunction is the picture that the computers run. The AIM density is the picture that the topology-minded interpreters extract from the wavefunction to recover something Lewis-shaped. None of these is the bond. Each of them is a way of disclosing the bond.

What this essay has tried to do is make the situation visible without resolving it. Resolving it would require either deciding that one of the pictures is the truth and the others are useful errors, or deciding that the bond is not a real object at all but a useful fiction. The working practice of chemistry has rejected both decisions and instead settled into something more delicate: the bond is real as a structural feature of an effective description, not real as a substance in the Hamiltonian, and the discipline operates by holding both of these at once. The pictures coordinate, mostly, and where they fail to coordinate, the failure is informative.

The companion unit 14.05.02 builds the MO picture for the diatomic case in technical detail; that unit and this essay together give the first complete pass over what the bond is, mechanically and ontologically. The companion unit 15.04.02 shows the bond in motion — being formed and broken in the transition state — and exposes the temporal version of the ontology question. The mathematical companion 03.09.02 Clifford algebra is the language in which the orbital symmetries underneath both VB and MO pictures live; the Pauli matrices and the small Clifford algebra $Cl_{0, 3}$ are the structural backbone of the orbital theory that makes either picture possible. The philosophical companions in §20 give the broader frame: this is one instance of a recurring pattern in which the practitioner uses pictures fully while knowing they are pictures.

What a chemist means when they say "the bond between these two atoms" is something like: there is a localized pattern of electron density between these two atoms, robust across geometric perturbations and transferable to other molecules, that I can predict and manipulate using a particular set of pictures, none of which is the underlying physics but all of which approximately agree, in this case, on a feature I can call a bond. This is not what a first-year textbook says. But it is, I think, what the practice is doing when it is doing well. The bond is the place where the pictures coordinate, and where they coordinate is exactly where they predict, and where they predict is exactly the regime in which chemistry, as a working discipline, has authority over molecules. That is the bond's ontological status: real where it predicts, not present where it doesn't, and known by working chemists to be both at once.

Cross-references. This essay synthesizes across 14.05.02 MO theory for homonuclear diatomics (the MO picture in technical detail), 15.04.02 SN1 vs SN2 mechanism (the bond-during-transition-state question), 03.09.02 Clifford algebra (the symmetry language for orbitals), and the philosophical essays in §20 (the pictures-vs-disclosure structure). It pairs with the broader chemistry-section discussions of 14.02 Chemical bonding I and 14.05 Chemical bonding II, and with anticipated future essays on chemistry's autonomy from physics and on the role of effective theories in scientific practice. Outbound hooks to 20.essays.NN (philosophy synthesis space) are proposed but not pinned to specific essays; the resolution will happen during the cross-domain audit at end of Wave 1.

Citation status. All references in this essay are pending+pointer: — the reference/ archive carries no chemistry material as of Wave 1 chem production, and the citations here are author/year placeholders to be resolved during the chemistry-side sourcing pass (CHEMISTRY_PLAN.md §5.3). The key sources to obtain are: Lewis 1916 (The Atom and the Molecule); Lewis 1923 (Valence and the Structure of Atoms and Molecules); Pauling 1939 (The Nature of the Chemical Bond); Mulliken's 1930s papers on MO theory; Lennard-Jones 1929 on the LCAO method; Coulson 1952 (Valence); Bader 1990 (Atoms in Molecules: A Quantum Theory); Reed, Curtiss, and Weinhold 1988 on NBO analysis; Hendry 2017 (The Metaphysics of Chemistry, or related work); Scerri 2007 (The Periodic Table: Its Story and Its Significance); Bensaude-Vincent and Stengers 1996 (A History of Chemistry); Worrall 1989 on structural realism; Ladyman and Ross 2007 (Every Thing Must Go).