Energy Level Splitting

Resolve one degenerate quantum energy into two or more distinct eigenvalues by changing the Hamiltonian so that its action within the degenerate subspace is not proportional to the identity.

Version

v2 · 2026-09-06 · History

Domain-specific #

1767

Origin domain

quantum physics

Subdomain

spectral structure and degenerate perturbation

Aliases

Level Splitting, Splitting of Energy Levels, Degeneracy Lifting

Core Idea

Energy level splitting is the lifting, complete or partial, of a degeneracy in a quantum Hamiltonian's spectrum. Before the relevant interaction or control is included, two or more linearly independent states share one energy eigenvalue. After the Hamiltonian changes, its restriction to that formerly degenerate state space distinguishes directions that previously had the same energy, producing two or more distinct energy eigenvalues. The abstraction turns a vague statement—“the interaction changes the energy”—into a definite spectral event: multiplicity is reduced and a resolvable separation appears.

Let an unperturbed Hamiltonian \(H_0\) have an eigenvalue \(E_0\) with an \(m\)-dimensional eigenspace \(D\). Write the changed Hamiltonian as \(H(\lambda)=H_0+\lambda V\). In first-order degenerate perturbation theory, the decisive object is not a diagonal matrix element in an arbitrarily chosen old basis. It is the Hermitian operator \(W=P_DVP_D\) restricted to \(D\), where \(P_D\) projects onto the degenerate subspace. Its eigenvalues \(w_a\) give the first-order branches

\[ E_a(\lambda)=E_0+\lambda w_a+O(\lambda^2). \]

Distinct \(w_a\) split the level at first order; repeated \(w_a\) preserve a residual degeneracy. If \(W=cI_D\), every state in \(D\) receives the same first-order shift and there is no first-order splitting, though higher orders may split it. MIT's official Quantum Physics III notes derive precisely this need to diagonalize the perturbation inside the degenerate subspace and call the resulting eigenvectors the “good” zeroth-order states.^[1]

The same identity can be found by exact diagonalization, numerical spectral computation, or experiment; perturbation theory is one method for predicting it, not part of the phenomenon's definition. External magnetic and electric fields, spin-dependent interactions, tunneling or intersite coupling, molecular distortion, and crystal environments can all provide the differentiating Hamiltonian term. The retained invariant is one quantum energy with multiplicity > 1 -> a physically specified Hamiltonian change acts nonuniformly within that state space -> fewer degeneracies and multiple energy branches.

Structural Signature

The abstraction has eight mandatory roles:

• Reference Hamiltonian. A stated \(H_0\) defines the spectrum relative to which “unsplit” is meaningful.
• Degenerate level. An energy \(E_0\) has eigenspace dimension \(m>1\). Similar numerical values are not enough; degeneracy is equality of eigenvalues within the model and resolution being used.
• Degenerate subspace. The set \(D\) of states sharing \(E_0\) is the space on which the lifting mechanism must be tested.
• Differentiating term. A field, interaction, coupling, distortion, relativistic correction, boundary change, or other contribution changes the Hamiltonian.
• Restricted action. \(P_DVP_D\), or the corresponding exact effective Hamiltonian, acts nonuniformly on \(D\). A scalar multiple of the identity shifts the whole level but does not split it at that order.
• Selected combinations. Diagonalizing the restricted action chooses orthogonal combinations of the old degenerate states that connect continuously to the new eigenstates.
• Separated branches. The resulting eigenvalues differ, fully or partially lifting the original multiplicity.
• Comparison scale. A coupling, field strength, distortion amplitude, or other parameter relates the branches to the reference limit and permits a splitting magnitude such as \(\Delta E=E_+-E_-\).

The recognition test is therefore stronger than “two energy values are visible.” One must identify a common parent level in a reference Hamiltonian and show how the added term resolves its state multiplicity. Conversely, the term remains valid when the splitting is too small to resolve experimentally, provided the Hamiltonian predicts distinct eigenvalues on the relevant scale. “Complete” means all formerly equal branches become distinct; “partial” means the perturbation partitions an \(m\)-fold level into groups that retain internal degeneracy.

What It Is Not

• Not an arbitrary energy shift. If \(P_DVP_D=cI_D\), all states move together. The common level changes but its multiplicity does not.
• Not perturbation theory. Perturbation theory is an approximation framework. Splitting is a spectral outcome that exact diagonalization, ab initio computation, or measurement may establish even where a small-parameter expansion is unsuitable.
• Not restricted to diagonal perturbations in the old basis. Off-diagonal matrix elements often select new linear combinations. Treating arbitrary old basis vectors separately gives basis-dependent and potentially wrong “corrections.”
• Not necessarily symmetry breaking. Lowering a symmetry often lifts symmetry-protected degeneracy, but accidental degeneracies can split without that story, and a perturbation preserving the relevant protecting symmetry may leave degeneracy intact.
• Not the same as an avoided crossing. An avoided crossing is parameter-dependent level repulsion near where uncoupled branches would cross. It can be understood through a two-level splitting mechanism, but level splitting also occurs away from crossings and need not be described as avoidance.
• Not automatically spectral-line splitting. Energy levels belong to stationary-state spectra. Observed lines correspond to allowed transition-energy differences; selection rules, populations, linewidths, and splitting of both endpoints determine the visible pattern.^[2]
• Not fine structure, hyperfine structure, Zeeman effect, Stark effect, crystal-field splitting, or Jahn–Teller splitting as a whole. Each names a specific interaction or setting that can instantiate the more general outcome.
• Not merely two eigenvalues. Two unrelated nondegenerate levels do not constitute the splitting of one degenerate parent.

Scope of Application

Energy level splitting is used throughout quantum mechanics wherever a reference degeneracy is resolved. In atomic physics, magnetic fields produce Zeeman sublevels and electric fields produce Stark structure. NIST describes a weak-field Zeeman level of angular momentum \(J\) as splitting into magnetic sublevels labeled by the \(2J+1\) possible values of \(M\), with field-dependent energy shifts.^[2] Relativistic and spin-dependent corrections generate fine structure; nuclear-spin coupling generates hyperfine structure. Those mechanisms differ, but analysts ask the same questions: what was degenerate, which Hamiltonian term distinguishes the states, which quantum numbers remain good, and what branch separations result?

In molecular physics and quantum chemistry, tunneling between classically equivalent configurations creates symmetric and antisymmetric combinations with different energies. Feynman's treatment of ammonia uses the two localized nitrogen configurations, their coupling, and the resulting pair of stationary energy states as the basis of the ammonia maser discussion.^[3][4] The Jahn–Teller theorem provides another route: certain nonlinear molecules in electronically degenerate states are unstable to symmetry-lowering nuclear displacements, and the distortion removes or reduces electronic degeneracy.^[5]

In condensed-matter and materials settings, crystal fields split atomic-like orbital multiplets, spin–orbit coupling reorganizes degeneracies, tunneling splits finite-well states, and fields resolve spin or orbital branches. In quantum devices, the controlled splitting between two computational states sets transition frequencies and control times. The abstraction also applies to effective Hamiltonians derived for a low-energy manifold, provided the reference subspace and approximations are stated.

The scope does not include every separation between measured peaks, every classical normal-mode frequency shift, or generic eigenvalue bifurcation in a nonquantum operator. Those have related mathematics but lack literal quantum-energy roles. Nor should “splitting” conceal a model choice: a level can be exactly degenerate in an idealized Hamiltonian, weakly split in a more complete one, and experimentally unresolved because the gap is smaller than linewidth or instrument resolution.

Clarity

Use a six-question diagnostic:

1. What is the reference? Name \(H_0\), the control-parameter limit, and the supposedly common energy \(E_0\).
2. What is degenerate? Give the dimension and basis-independent characterization of \(D\), not just state labels.
3. What changes the Hamiltonian? Identify the physical term or coupling and the approximation under which \(H=H_0+V\) is used.
4. Does it distinguish states in \(D\)? Calculate or characterize \(P_DVP_D\). If it is scalar, no first-order splitting occurs.
5. What survives? Report the new branches, their multiplicities, selected eigenstates, and whether residual symmetry protects any degeneracy.
6. How is it observed? Connect energy differences to transition frequencies, selection rules, populations, and resolution rather than equating eigenvalues with peaks automatically.

For a two-state subspace, write the effective Hamiltonian as

\[ H_{\mathrm{eff}}=E_0I+ \begin{pmatrix} \delta & t\\ t^* & -\delta \end{pmatrix}. \]

Its eigenvalues are \(E_\pm=E_0\pm\sqrt{\delta^2+|t|^2}\), so the gap is \(2\sqrt{\delta^2+|t|^2}\). A diagonal bias \(\delta\) distinguishes the basis states; an off-diagonal coupling \(t\) selects superpositions; either can lift the degeneracy. This exact result also shows why “read the diagonal entries” is not a general diagnostic.

Manages Complexity

The abstraction compresses a wide catalog of named effects into one spectral workflow. Instead of memorizing each phenomenon as an isolated fact, the analyst locates the reference multiplet, constructs the effective operator on it, diagonalizes that smaller object, and tracks the remaining degeneracies. An \(m\)-fold problem embedded in a large Hilbert space often reduces, at leading order, to an \(m\times m\) Hermitian matrix. The reduction preserves the part that decides splitting while postponing irrelevant details.

It also separates three questions that are commonly conflated: cause, spectral result, and observable signature. A magnetic field is a cause; distinct \(M\)-dependent energies are the level splitting; a collection of polarized transition components is a possible spectroscopic signature. This separation makes comparisons possible across atomic, molecular, solid-state, and engineered two-level systems.

Finally, splitting size becomes a compact measurable parameter. A gap can diagnose interaction strength, field magnitude, tunneling amplitude, distortion, or symmetry protection. But the inverse inference is model-dependent: equal gaps do not establish equal causes, and an unresolved line does not prove exact degeneracy.

Abstract Reasoning

The structural signature licenses several bounded inferences:

1. If \(P_DVP_D\) has \(r\) distinct eigenvalues, the \(m\)-fold parent divides into \(r\) first-order energy groups; multiplicities of the \(w_a\) give residual degeneracies.
2. If the restricted perturbation is proportional to the identity, there is a common first-order shift but no first-order splitting. One must inspect higher-order effective terms before claiming protection to all orders.
3. If an off-diagonal coupling is nonzero in a two-state model, the eigenstates are generally superpositions of the reference states and the minimum exact gap is nonzero unless another parameter cancels the entire traceless part.
4. If a symmetry commutes with the full Hamiltonian and enforces a multidimensional representation or protected pairing, perturbations respecting that symmetry cannot arbitrarily separate all partners. Breaking or reducing the protection can permit splitting.
5. A measured transition pattern constrains differences between upper and lower energies, not either level in isolation. Several assignments may fit until selection rules and polarization are included.
6. A gap that varies linearly with weak field is consistent with a first-order field coupling; nonlinear behavior can signal mixing, higher-order terms, or a regime where the original quantum numbers cease to be good.
7. Label continuity must be handled carefully near strong mixing. Branches can exchange their dominant old-basis character even while the exact eigenvalues vary smoothly.

These are diagnostic consequences of the mechanism, not universal promises about convergence, line visibility, or the physical origin of \(V\).

Knowledge Transfer

Literal transfer occurs across atomic, molecular, condensed-matter, nuclear, and quantum-information problems because all retain Hamiltonian energy eigenvalues, a degenerate reference space, and a differentiating interaction. The detailed vocabulary changes—multiplet, tunneling doublet, crystal-field level, spin branch, qubit frequency—but the recognition test stays intact.

The underlying linear-algebra skeleton transfers more widely: a repeated eigenvalue of an operator separates when an operator change acts non-scalar on its eigenspace. That portable skeleton belongs to Eigenvalue and Eigenvector and to spectral perturbation theory. It does not make Energy Level Splitting a prime. In vibration analysis, covariance matrices, graph spectra, or stability theory, the separated quantities are not quantum energies and the physical interpretation, observability rules, and state-space constraints differ. Calling all eigenvalue separation “energy-level splitting” would erase the domain information that makes the abstraction useful.

Perturbation Theory supplies a transferable computational strategy, while Symmetry Breaking supplies one common causal explanation. Neither is an alias. One can calculate a splitting exactly without perturbation theory; a small perturbation can shift without splitting; symmetry may be reduced without the specific parent degeneracy under study; and degeneracy can be accidental.

Examples

• Two-state bias. Start from \(H_0=E_0I\). Adding \(V=\epsilon\sigma_z\) gives \(E_0\pm\epsilon\). The old twofold degeneracy becomes two nondegenerate levels with gap \(2|\epsilon|\). The chosen eigenstates coincide with the \(\sigma_z\) basis.
• Two-state tunneling. For localized states \(|L\rangle,|R\rangle\) of equal reference energy, an off-diagonal coupling \(t\) gives symmetric and antisymmetric combinations with energies \(E_0\pm|t|\). Ammonia inversion is the canonical molecular example; the transition between the pair underlies maser operation.^[3][4]
• Weak-field Zeeman splitting. An atomic level with angular momentum \(J\) separates into \(M\)-labeled magnetic sublevels. To leading order, their shifts depend on \(gM\mu_BB\); distinct \(M\) values normally give distinct branches. NIST emphasizes both the energy shifts and the observed Zeeman patterns used to infer \(J\) and \(g\).^[2]
• Jahn–Teller lifting. A nonlinear molecule in an appropriate degenerate electronic state distorts. The nuclear displacement changes the electronic Hamiltonian, reduces the symmetry, and separates the electronic energies; stabilization competes with elastic distortion cost.^[5]
• Partial splitting. An \(m=4\) level can have restricted perturbation eigenvalues \((a,a,a,b)\). It divides into a triply degenerate level and a singlet rather than four distinct levels. MIT's perturbation notes explicitly illustrate first-order and higher-order partial lifting.^[1]
• Negative case: common shift. If \(V=cI_D\), every state in the degenerate subspace moves from \(E_0\) to \(E_0+c\). Energy changed, but no splitting occurred.
• Negative case: unrelated lines. Two optical peaks from two independent transitions are not evidence that one parent level split unless a Hamiltonian model, parameter dependence, or selection-rule analysis identifies the common parent.

Structural Tensions

• Ideal degeneracy vs. complete Hamiltonian. A convenient baseline exposes the mechanism; omitted weak interactions may mean the real system was never exactly degenerate.
• Basis labels vs. selected eigenstates. Familiar state labels aid interpretation; off-diagonal coupling can make the true states mixtures that invalidate those labels.
• Symmetry explanation vs. accidental degeneracy. Symmetry predicts robust multiplicity and selection rules; not every equality of energies is symmetry-enforced.
• Perturbative clarity vs. strong coupling. A small-\(\lambda\) expansion cleanly identifies leading terms; exact or numerical diagonalization is required when mixing is strong or denominators become unsafe.
• Complete vs. partial lifting. Analysts often say “the degeneracy is lifted” loosely; the restricted operator may leave multiplets and require higher-order or additional interactions.
• Energy spectrum vs. observed spectrum. Eigenvalue differences are model outputs; linewidth, lifetime, population, selection rules, polarization, and instrumental resolution decide what appears.
• Label continuity vs. eigenstate exchange. Following a branch by energy is easy; following it by physical character becomes ambiguous when states mix.

Structural–Framed Character

Energy Level Splitting is framed, with an estimated aggregate of (0.82). Its core is formal: a repeated eigenvalue is resolved because a changed Hermitian operator is non-scalar on the old eigenspace. Yet literal use is constituted by quantum commitments—Hamiltonians, energy eigenstates, superposition, physical interactions, quantum numbers, transitions, and spectroscopic evidence. The abstraction transfers robustly within quantum subdomains but only mathematically outside them. It is neither a culturally contingent practice nor a substrate-independent prime.

Structural Core vs. Domain Accent

The structural core is multiplicity -> nonuniform restricted operator -> diagonalization -> separated eigenvalue branches + residual multiplicities. Linear algebra owns that skeleton. The domain accent supplies why the operator is a Hamiltonian, why its eigenvalues are energies, which interactions are allowed, how states and symmetries are labeled, and how transition frequencies reveal gaps.

Remove the domain accent and the candidate becomes repeated-eigenvalue perturbation, already expressible through Eigenvalue and Eigenvector plus Perturbation Theory. Preserve it and one can reason correctly about Zeeman sublevels, tunneling doublets, orbital multiplets, and molecular distortions while respecting selection rules and resolution. The Encyclopedia should therefore retain the quantum abstraction as domain-specific and route its portable residue upward rather than promoting the whole term to prime status.

Instantiates / Related Primes

• Eigenvalue and Eigenvector is the minimal structural parent. An energy level is a Hamiltonian eigenvalue; degeneracy is eigenvalue multiplicity; splitting is the changed eigenvalue structure after diagonalization.
• Perturbation Theory is a principal computational relation. Degenerate perturbation theory obtains leading branch energies from \(P_DVP_D\), but splitting can be exact and perturbation theory can produce common shifts without splitting.
• Symmetry Breaking explains many instances in which a symmetry-lowering interaction distinguishes previously equivalent states. It is related rather than mandatory.
• Symmetry explains protected degeneracy and why residual multiplets may survive a permitted perturbation.

Only the first is proposed as a DAG parent. The others are explanatory neighbors and should not multiply inheritance edges.

Relationships to Other Abstractions

Current abstraction Energy Level Splitting Domain-specific

Parents (1) — more general patterns this builds on

• Energy Level Splitting is a kind of Eigenvalue And Eigenvector Prime

  Eigenvalue and Eigenvector is the minimal structural parent.

Hierarchy paths (2) — routes to 2 parentless roots

• Energy Level Splitting → Eigenvalue And Eigenvector → Linearity

Show alternative path (1)

Neighborhood in Abstraction Space

Energy Level Splitting sits in a sparse region of the domain-specific corpus (76^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Quantum States & Thermal Dynamics (12 abstractions)

Nearest neighbors

• Energetic Space — 0.85
• Eigenstate Thermalization Hypothesis — 0.85
• Stable Yang–Mills–Higgs Pair — 0.83
• Schrödinger Equation — 0.82
• Mixed Quantum–Classical Dynamics — 0.82

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With

Perturbation Theory decomposes and approximates a changed problem; Energy Level Splitting is one possible spectral result. Symmetry Breaking is loss of invariance and may yield splitting, but the two do not entail each other generally. Eigenvalue and Eigenvector is the substrate-independent linear-algebra structure; the candidate specializes its repeated-eigenvalue change in a quantum Hamiltonian.

Also distinguish degeneracy lifting from a common-mode level shift; an energy gap from the event that created it; level repulsion and an avoided crossing from all forms of splitting; and an observed spectral-line multiplet from the energy levels and allowed transitions that generate it. Fine, hyperfine, Zeeman, Stark, crystal-field, spin–orbit, tunneling, and Jahn–Teller effects should be recorded as mechanisms or named instances, not unrestricted aliases.

References

[1] Barton Zwiebach, “Chapter 1: Time-Independent Perturbation Theory,” Quantum Physics III, MIT OpenCourseWare, Spring 2018. Official lecture notes; §§1.2–1.3 derive degenerate perturbation theory, good-basis selection, and first- and higher-order lifting. registry ↩a ↩b

[2] National Institute of Standards and Technology, “Atomic Spectroscopy—Zeeman Effect”, Atomic Spectroscopy Compendium. Official reference for weak-field magnetic sublevels, energy shifts, \(g\) factors, and observed Zeeman patterns. registry ↩a ↩b ↩c

[3] Richard P. Feynman, Robert B. Leighton, and Matthew Sands, The Feynman Lectures on Physics, Vol. III, Chapter 8, “The Hamiltonian Matrix”, 1965, especially the ammonia inversion two-state treatment. registry ↩a ↩b

[4] Feynman, Leighton, and Sands, The Feynman Lectures on Physics, Vol. III, Chapter 9, “The Ammonia Maser”, 1965. registry ↩a ↩b

[5] H. A. Jahn and E. Teller, “Stability of Polyatomic Molecules in Degenerate Electronic States. I. Orbital Degeneracy,” Proceedings of the Royal Society A 161 (1937), 220–235. Primary statement and analysis of distortion associated with electronically degenerate molecular states. registry ↩a ↩b