1 Fundamental concept
1.1 Definition
Avoided crossing is a behavior seen when two energy levels (or resonance frequencies) become close as an external control parameter is varied, yet the levels do not meet. Instead, the proximity is followed by a reversal in character: each eigenstate continuously evolves from one “uncoupled” state into the other, and a minimum energy separation remains at the would-be intersection. The phenomenon reflects that the system’s states are not independent; they are linked by an interaction that mixes them.
1.2 Level repulsion
A defining feature is level repulsion: as the control parameter approaches the resonance value, the eigenvalues veer away from each other rather than crossing. In an energy-versus-parameter plot, two curves approach and then separate, leaving a characteristic gap at the closest approach. The gap is typically smallest near the nominal crossing point and grows as the interaction becomes less able to mix the states efficiently away from resonance.
1.3 Role of coupling
The interaction responsible for avoided crossing arises when the Hamiltonian contains off-diagonal terms in a basis where the two relevant states would otherwise be distinct. These coupling terms permit transitions between the basis states, so the true eigenstates are superpositions. The strength of that coupling controls the minimum gap: stronger coupling generally produces a larger separation between the two eigenvalues near the resonance condition.
1.4 Comparison with true crossing
A “true crossing” occurs when two eigenvalues become equal and the eigenvectors remain distinct without any mixing—typically requiring additional symmetry or an absence of coupling between the corresponding subspaces. In many realistic systems, such symmetry conditions are not met, so the crossing is lifted. Avoided crossing therefore provides a robust qualitative expectation: whenever two states of the same symmetry mix through a nonzero interaction, their energies repel and a degeneracy is avoided.
2 Theoretical background
2.1 Two-level systems
Many avoided-crossing discussions begin with a reduced two-level model. One assumes two relevant basis states, each with an energy that depends approximately linearly on an external parameter near the resonance. The coupling between the basis states is taken as approximately constant over that narrow range. Despite its simplicity, this model reproduces the essential gap and eigenstate evolution observed in more complex systems.
2.2 Adiabatic and diabatic representations
A key theoretical distinction is between how states are labeled and followed as the parameter changes.
2.2.1 Adiabatic energy surfaces
In the adiabatic representation, one diagonalizes the full Hamiltonian at each parameter value. The resulting eigenstates vary smoothly, and their energies form the “adiabatic surfaces.” Avoided crossings appear directly in these adiabatic energies: the gap is a consequence of diagonalization in the presence of coupling.
2.2.2 Diabatic crossing picture
In the diabatic representation, one instead keeps a basis where the coupling is minimized but not fully diagonalized at each parameter. In this picture, uncoupled energies may cross, and the coupling appears in off-diagonal elements. The avoided crossing then becomes interpretable as the effect of those off-diagonal terms on what would be a degeneracy in the uncoupled (diabatic) limit.
2.3 Matrix models
Avoided crossing can also be formulated using generic parameter-dependent matrix models. The simplest Hermitian two-by-two Hamiltonian captures the phenomenon, while larger matrices extend it to multiple interacting levels. In these models, the qualitative outcome is governed by the eigenvalue structure of Hermitian operators: when two branches approach and the coupling between their corresponding eigenvectors is nonzero, degeneracy is typically removed, producing repulsion rather than intersection.
3 Physical interpretation
3.1 Energy level interaction
At the resonance region, the two bare states become nearly degenerate. Because they are not truly independent, the interaction allows the system to lower its energy by forming superpositions. The energies then split: one branch shifts upward and the other shifts downward relative to the uncoupled crossing point, ensuring that equality is no longer achieved.
3.2 State mixing
The eigenstates on either side of the avoided crossing differ in composition. Near the gap, each adiabatic eigenvector is a mixture of the two diabatic (or bare) states. As the parameter passes the resonance region, the dominant component of the eigenstate gradually swaps from one basis state to the other. This continuous change is the physical meaning behind “continuation” of states through the resonance region.
3.3 Gap formation
The minimum separation between eigenvalues at the closest approach is the gap. In the idealized two-level case, the gap is directly tied to the magnitude of the coupling matrix element. Thus, the size of the avoided crossing serves as a diagnostic of interaction strength: a small gap indicates weak coupling between the relevant states, while a larger gap signals stronger mixing.
3.4 Resonance behavior
Away from resonance, each eigenstate tends to recover its bare character because the energy difference between the basis levels becomes large compared with the coupling. Consequently, the avoided crossing is localized: the most pronounced repulsion and mixing occur near the parameter value where the unperturbed levels would coincide. This localization explains why spectroscopic lines often appear to “split” only in a narrow region of tuning.
4 Mathematical description
4.1 Eigenvalue analysis
Consider a two-level Hamiltonian in a diabatic basis: \[ H(\lambda)= \begin{pmatrix} E_1(\lambda) & V \\ V^\ast & E_2(\lambda) \end{pmatrix}, \] where \(\lambda\) is an external parameter and \(V\) is the coupling. The eigenvalues are \[ E_{\pm}(\lambda)=\frac{E_1(\lambda)+E_2(\lambda)}{2}\pm
| \sqrt{\left(\frac{E_1(\lambda)-E_2(\lambda)}{2}\right)^2+ | V | ^2}. |
|---|
\]
| When the bare energies approach each other so that \(E_1(\lambda)\approx E_2(\lambda)\), the square-root term remains finite because of \( | V | ^2\), preventing equality of \(E_+\) and \(E_-\). The minimum gap occurs where \(E_1(\lambda)=E_2(\lambda)\) and equals \(2 | V | \) in this simplified model. |
|---|
4.2 Hamiltonian formulations
In many applications, one rewrites the Hamiltonian using Pauli matrices: \[ H(\lambda)=\bar{E}(\lambda)I + \Delta(\lambda)\sigma_z + \Re(V)\sigma_x - \Im(V)\sigma_y, \] with \(\Delta(\lambda)=(E_1-E_2)/2\). This form highlights that the eigenvalues are determined by the magnitude of the effective “field” vector \((\Re(V),-\Im(V),\Delta)\). Avoided crossing then corresponds to how this vector’s components vary with \(\lambda\), ensuring the eigenvalue splitting never collapses when the transverse components (the coupling) are nonzero.
4.3 Perturbative treatment
| For weak coupling, one treats \(V\) as a small perturbation near the region where \(E_1\) and \(E_2\) are close. Far from resonance, standard nondegenerate perturbation theory yields small corrections proportional to \( | V | ^2/(E_1-E_2)\). Close to resonance, that expansion must be handled more carefully, and the two-level diagonalization provides the correct nonperturbative expression for the gap. |
|---|
4.4 Dependence on coupling strength
4.4.1 Weak coupling limit
| When \( | V | \) is small, the avoided crossing is narrow: the curves approach closely but remain separated by a small minimum gap of order \(2 | V | \). The eigenstates exhibit rapid but not discontinuous transformation only in a limited parameter window where the bare energy difference is comparable to \( | V | \). |
|---|
4.4.2 Strong coupling limit
| When \( | V | \) is large compared with the scale over which \(E_1(\lambda)-E_2(\lambda)\) varies, the mixing is substantial across a broad interval of \(\lambda\). The eigenvalue splitting becomes dominated by the coupling term, and the distinction between “before” and “after” resonance becomes less tied to where bare levels would cross, since the eigenstates are strongly hybridized throughout. |
|---|
5 Applications
5.1 Molecular physics
In molecules, electronic or vibrational states can approach each other as internuclear geometry changes. Coupling between different potential energy surfaces leads to avoided crossings, which affect reaction dynamics, nonradiative transitions, and vibrational energy level structure. The gap location and size influence transition probabilities and can guide the identification of coupled modes or electronic states.
5.2 Atomic and nuclear spectroscopy
Spectra often reveal level splitting near regions where two configurations become nearly degenerate. Configuration interaction—an effective mixing due to residual interactions—produces avoided crossings in the dependence of energies on quantum parameters or effective mean-field variables. In such settings, the observed separation between lines can be interpreted as the manifestation of coupling between nearly resonant configurations.
5.3 Condensed matter systems
In solids, bands and discrete levels interact through perturbations such as spin-orbit coupling, lattice distortions, or external fields. When two bands of the same symmetry approach, hybridization yields anticrossing behavior that is central to interpreting band structure and quasiparticle dispersions. Avoided crossings also play a role in understanding how carriers evolve in parameter-tuned systems like heterostructures.
5.4 Coupled oscillator models
Coupled oscillators provide an accessible classical analogy to level repulsion. Two modes with frequencies that depend on a tuning parameter can display mode splitting when a coupling is introduced. Although the underlying mathematics for eigenfrequencies resembles the quantum two-level case, the physical meaning is different: energy gaps correspond to differences in normal mode frequencies rather than quantized eigenenergies.
6 Related phenomena
6.1 Level repulsion
Level repulsion is the umbrella term describing the general tendency of eigenvalues of coupled systems to separate as parameters change. Avoided crossing is a prominent instance of this behavior, especially when the plot of eigenvalues exhibits a clearly defined minimum separation.
6.2 Conical intersections
Conical intersections occur in multidimensional parameter spaces of molecular electronic structure, where two adiabatic potential surfaces become degenerate at specific geometries. Locally, the surfaces form a cone, and the system can exchange character when moving around the intersection. Compared with avoided crossing, conical intersections involve true degeneracies, but avoided crossings are the common “lower-dimensional projection” outcome when the full space is not explored or when additional coupling prevents degeneracy.
6.3 Landau-Zener transitions
Landau-Zener theory describes the probability that a system transitions between adiabatic eigenstates when a parameter is swept through an avoided crossing at a finite rate. The key ingredients are the sweep rate, the minimum gap, and the coupling-controlled mixing. This framework connects the static eigenvalue repulsion picture with dynamics and transition likelihood.
6.4 Tunneling effects
In systems with potential barriers, tunneling can effectively couple states that would otherwise be separated. When tunneling couples nearly degenerate configurations, avoided crossings arise in the combined description: the barrier-mediated coupling creates the off-diagonal interaction that lifts degeneracy. Thus, tunneling can be viewed as a mechanism that generates the coupling responsible for the gap.
7 Experimental observation
7.1 Spectroscopic signatures
Avoided crossing typically appears in spectroscopy as line splitting or as an apparent repulsion between resonance peaks when an external parameter (such as field strength or geometry) is tuned. Rather than two lines merging into one at degeneracy, the peaks exchange intensity or exhibit anticorrelated shifts consistent with eigenstate mixing across the resonance region.
7.2 Measurement techniques
Experimental setups often combine tunable control with high-resolution detection. Examples include adjusting magnetic or electric fields, using laser-based spectroscopy to probe transitions, or employing scanning techniques to vary system parameters such as cavity detuning. Data analysis then fits observed resonances to coupled-mode or two-level models to extract the coupling strength and the minimum gap.
7.3 Interpreting avoided crossings
Interpreting measurements involves determining which modes or levels are coupled and how the eigenstates evolve. Common approaches include fitting eigenvalue curves to coupled Hamiltonians, comparing intensity changes to predicted mixing angles, and validating whether the observed gap scales with an independently varied coupling mechanism. When multiple levels contribute, more sophisticated multi-state models may be required, but the avoided-crossing logic remains a guiding principle.
8 Historical development
8.1 Early theoretical studies
The foundational mathematics of coupled eigenvalues traces to early work on perturbations and degeneracy lifting in quantum systems. As spectroscopic techniques improved, the need to understand why resonances do not cross became clearer, leading to systematic treatment of two-level mixing and eigenvalue splitting in simplified models.
8.2 Modern applications
Modern research places avoided crossing within broader frameworks such as diabatic/adiabatic dynamics, nonadiabatic transition theory, and band-structure engineering. It is routinely used to interpret anticrossing in engineered quantum devices, to analyze molecular potential surface connectivity, and to model coupled-mode behavior in engineered classical and quantum oscillatory systems.