Semiclassical Dynamics of Bloch Electrons

Semiclassical Bloch dynamics represents a quantum state localized in both real space and crystal momentum by a center $(\mathbf r_c,\mathbf k_c)$ that moves through one isolated band. In the ordinary, nongeometric approximation, a carrier of signed charge $q$ obeys

$$\dot{\mathbf r}_c = \frac{1}{\hbar} \nabla_{\mathbf k} \varepsilon_n(\mathbf k_c), \qquad \hbar\dot{\mathbf k}_c = q \left( \mathbf E + \dot{\mathbf r}_c\times\mathbf B \right).$$

For an electron, $q=-e$ with $e>0$; for a hole described relative to a filled band, $q=+e$. The first equation is a property of the band dispersion. The second is the Lorentz-force law for crystal momentum. Neither equation supplies a collision time, a conductivity, a Landau spectrum, or a guarantee that one band remains isolated.

Electromagnetic equations on this page use SI units. In Gaussian units the magnetic Lorentz term carries the corresponding factor of $1/c$.

Here $\mathbf k_c$ is a wave-vector coordinate on the Brillouin torus, not the electron’s mechanical momentum divided by $\hbar$. The construction below explains the packet’s localization, its motion between collisions and the conditions under which an isolated band is sufficient.

Required background. Band Theory Overview supplies Bloch states, band energies, Brillouin-zone periodicity, and the distinction among model, Kohn–Sham, and coherent quasiparticle bands.

Helpful background. Wave Packets and Classical Trajectories supplies generic localization, spreading, and trajectory criteria, while Effective Mass supplies the local band Hessian. General Berry Curvature is needed only for the geometric branch below.

An Isolated-Band Wave Packet

Let a periodic Hamiltonian have normalized cell-periodic eigenvectors

$$H(\mathbf k) |u_{n\mathbf k}\rangle = \varepsilon_n(\mathbf k) |u_{n\mathbf k}\rangle,$$

and Bloch states $|\psi_{n\mathbf k}\rangle$. A packet made from one band is

$$|W_n\rangle = \int_{\mathrm{BZ}}[d\mathbf k]\, a(\mathbf k,t) |\psi_{n\mathbf k}\rangle, \qquad \int_{\mathrm{BZ}}[d\mathbf k]\, |a(\mathbf k,t)|^2 = 1,$$

Use delta-normalized Bloch states, $\langle\psi_{n\mathbf k}|\psi_{m\mathbf k'}\rangle=(2\pi)^d\delta_{nm}\delta^{(d)}(\mathbf k-\mathbf k')$, and $[d\mathbf k]=d^dk/(2\pi)^d$. Cell-periodic inner products use unit cell-average normalization. Other overall normalizations give the same packet when the measure and amplitudes are transformed together. The envelope is concentrated in a contractible patch around $\mathbf k_c$. Within that patch one may define

$$\mathbf k_c = \int_{\mathrm{BZ}}[d\mathbf k]\, |a(\mathbf k,t)|^2\mathbf k, \qquad \mathbf r_c = \langle W_n|\mathbf r|W_n\rangle.$$

The first formula is not a global average of angular coordinates across an arbitrary Brillouin-zone cut; it is meaningful because the packet is narrow in a local chart. If $a=|a|e^{-i\gamma}$, the narrow-packet center has the local form

$$\mathbf r_c \simeq \nabla_{\mathbf k}\gamma(\mathbf k_c) + \boldsymbol{\mathcal A}_n(\mathbf k_c), \qquad \boldsymbol{\mathcal A}_n = i \langle u_{n\mathbf k}| \nabla_{\mathbf k}u_{n\mathbf k}\rangle.$$

The two terms change oppositely under a Bloch-state gauge transformation, so their sum is gauge invariant. This is a local construction: a packet can be assembled patchwise even when no globally smooth eigenvector exists over the full Brillouin zone.

Let $a_{\mathrm{lat}}$ denote a representative lattice spacing, $L_{\mathrm{pert}}$ the shortest spatial scale on which the external perturbation changes appreciably, and $\Delta r_i$, $\Delta k_i$ the rms standard deviations along a chosen pair of conjugate directions. A controlled local packet requires a window such as

$$a_{\mathrm{lat}} \ll \Delta r \ll L_{\mathrm{pert}}, \qquad L_{\mathrm{pert}}^{-1} \ll \Delta k \ll a_{\mathrm{lat}}^{-1}, \qquad \Delta r_i\Delta k_i \gtrsim \frac12.$$

The packet spans many cells, so microscopic lattice oscillations are averaged into a band description, but it remains small enough to sample the perturbation locally. It must also remain in one spectrally isolated band over its significant momentum region and over the trajectory it will explore. A narrow packet centered far from a degeneracy can be valid even if the band touches another band elsewhere in the Brillouin zone; a trajectory through that touching cannot.

A Gaussian envelope has nonzero mathematical tails throughout the Brillouin zone, so literal support is too strong for a practical isolation test. Fix a small discarded-weight tolerance $\delta_a$ and declare a significant-weight region $\mathcal K_a(t)$ such that

$$\int_{\mathrm{BZ}\setminus\mathcal K_a(t)} [d\mathbf k]\, |a(\mathbf k,t)|^2 < \delta_a.$$

Whenever this page refers to the packet’s retained momentum region, it means the declared $\mathcal K_a(t)$ together with the reported tolerance, not the literal support of an infinite-tailed model envelope.

Group Velocity in a Bloch Band

Without external perturbations, each momentum component accumulates the phase $\exp[-i\varepsilon_n(\mathbf k)t/\hbar]$. Expanding that phase around $\mathbf k_c$ translates the packet envelope at

$$\mathbf v_n(\mathbf k_c) = \frac{1}{\hbar} \nabla_{\mathbf k} \varepsilon_n(\mathbf k_c).$$

More precisely, before replacing the packet by its center,

$$\dot{\mathbf r}_c = \int_{\mathrm{BZ}}[d\mathbf k]\, |a(\mathbf k,t)|^2 \mathbf v_n(\mathbf k).$$

The center formula follows when $\mathbf v_n$ varies little across the support. Packet spreading and distortion depend on higher derivatives and are not encoded by the center alone.

For a local Schrödinger Hamiltonian,

$$H(\mathbf k) = \frac{(\mathbf p+\hbar\mathbf k)^2}{2m} +V(\mathbf r),$$

the Hellmann–Feynman theorem gives

$$\frac{1}{\hbar} \frac{\partial\varepsilon_n}{\partial k_i} = \frac{1}{\hbar} \left\langle u_{n\mathbf k}\middle| \frac{\partial H(\mathbf k)}{\partial k_i} \middle|u_{n\mathbf k}\right\rangle = \left\langle \frac{p_i+\hbar k_i}{m} \right\rangle.$$

Thus the gradient of the band energy is the cell-averaged physical velocity for that Bloch state. It need not be parallel to $\mathbf k$, and it is periodic under $\mathbf k\mapsto\mathbf k+\mathbf G$. A flat band has vanishing ordinary group velocity even though its Bloch eigenvectors may carry nontrivial geometry.

Crystal Momentum Under Force

For an electric field alone, the acceleration theorem is

$$\hbar\dot{\mathbf k}_c = q\mathbf E.$$

A uniform field therefore translates the packet through reciprocal space at a rate whose direction already contains the carrier-sign convention. An electron moves opposite $\mathbf E$ in $\mathbf k$ space. In a reduced-zone description, crossing a chosen zone face merely changes the representative by a reciprocal vector; it is not automatically an interband transition.

In slowly varying electromagnetic fields, the ordinary coupled equations are

$$\begin{aligned} \dot{\mathbf r}_c &= \frac{1}{\hbar} \nabla_{\mathbf k} \varepsilon_n(\mathbf k_c), \\ \hbar\dot{\mathbf k}_c &= q \left[ \mathbf E(\mathbf r_c,t) + \dot{\mathbf r}_c\times \mathbf B(\mathbf r_c,t) \right]. \end{aligned}$$

An additional smooth scalar energy $U(\mathbf r)$ contributes $-\nabla_{\mathbf r}U$ to the second equation. If $U=q\phi$ already represents the electrostatic energy, its force is $q\mathbf E$ and must not be counted twice.

With no additional scalar energy $U$, the work–energy relation follows directly:

$$\begin{aligned} \frac{d\varepsilon_n}{dt} &= \nabla_{\mathbf k}\varepsilon_n \cdot\dot{\mathbf k}_c \\ &= q\, \dot{\mathbf r}_c\cdot\mathbf E. \end{aligned}$$

The magnetic part vanishes because $\dot{\mathbf r}_c\cdot(\dot{\mathbf r}_c\times\mathbf B)=0$. A magnetic field redirects the trajectory without doing work. If an additional static $U(\mathbf r)$ is present, the corresponding relation is $d(\varepsilon_n+U)/dt=q\mathbf E\cdot\dot{\mathbf r}_c$; its force changes the band energy while transferring energy to or from $U$.

Uniform Electric Fields and Bloch Oscillations

Set $\mathbf B=0$ and take $\mathbf E$ constant. Then

$$\mathbf k_c(t) = \mathbf k_c(0) + \frac{q\mathbf E}{\hbar}t,$$

with reciprocal-lattice equivalence understood. Because $\varepsilon_n(\mathbf k)$ is periodic on the Brillouin torus, the velocity repeats whenever the trajectory advances by a reciprocal vector. In one dimension, a perfectly coherent isolated band therefore produces Bloch oscillations rather than indefinite acceleration. No physical reflection at a zone boundary is required.

In more than one dimension, the crystal-momentum point itself recurs only when the field direction is commensurate with the reciprocal lattice: for some nonzero reciprocal vector $\mathbf G$,

$$\frac{q\mathbf E T}{\hbar} = \mathbf G.$$

An incommensurate direction generally gives quasiperiodic velocity, although a special dispersion can have a shorter velocity period or constant components without momentum recurrence. Even a closed reciprocal-space path can carry an ordinary transverse drift. The one-dimensional example below has an unambiguous period and zero displacement over a complete cycle.

Worked one-dimensional tight-binding band

For lattice spacing $a$ and hopping energy $J>0$, consider

$$\varepsilon(k) = -2J\cos(ka).$$

The group velocity and accelerated wave vector are

$$v(k) = \frac{2Ja}{\hbar}\sin(ka), \qquad k(t) = k_0+ \frac{qE}{\hbar}t.$$

One reciprocal period $2\pi/a$ is traversed in the Bloch period

$$T_B = \frac{2\pi\hbar}{|qE|a} = \frac{h}{|qE|a}.$$

Integrating the velocity gives

$$\begin{aligned} x(t)-x(0) &= \frac{\varepsilon[k(t)]-\varepsilon(k_0)}{qE} \\ &= \frac{2J}{qE} \left[ \cos(k_0a)-\cos(k(t)a) \right]. \end{aligned}$$

The peak-to-peak excursion is $4J/|qE|$, and the time average of $v$ over one full period is zero. This is a coherent single-band result, not a generic prediction of zero current in a real crystal. Scattering interrupts the cycle, interband transfer leaks amplitude into other bands, and geometric terms can add transverse motion. Those effects require their own declared regimes.

For $qE>0$, $k_0=0$ and the cosine band, the reduced-zone representative jumps from $\pi/a$ to $-\pi/a$. The velocity and real-space position remain continuous; the jump is a coordinate change. Here $v_{\max}=2Ja/\hbar$ and the plotted displacement is $\xi=qE[x(t)-x(0)]/(2J)$. These are the narrow-packet center formulas. A finite momentum width averages the velocities and reduces the excursion, as the first exercise shows.

Uniform Magnetic Fields and Band Orbits

Set $\mathbf E=0$ and take $\mathbf B$ constant. The equations imply two invariants:

$$\frac{d\varepsilon_n}{dt} = 0, \qquad \frac{d}{dt} \left( \widetilde{\mathbf k}_c\cdot\mathbf B \right) = 0.$$

Here $\widetilde{\mathbf k}_c(t)$ is a continuous lift of the trajectory to repeated-zone reciprocal space. The dot product is not a globally single-valued coordinate on the Brillouin torus because adding $\mathbf G$ can change it. The lifted trajectory is the intersection of a periodically continued constant-energy surface with a plane perpendicular to $\mathbf B$; the physical torus path is its projection modulo reciprocal vectors. The signed charge fixes its orientation. Compact closed lifts and open lifts have different dynamical consequences; an open lift can project to a closed winding loop on the torus. Their classification and extremal areas belong to Fermi Surface.

For $\mathbf B=B\hat{\mathbf z}$, the transverse real- and reciprocal-space displacements satisfy locally

$$\mathbf r_{c\perp}(t)-\mathbf r_{c\perp}(0) = \frac{\hbar}{qB} \hat{\mathbf z}\times \left[ \widetilde{\mathbf k}_{c\perp}(t)- \widetilde{\mathbf k}_{c\perp}(0) \right].$$

Weak field has more than one meaning. The magnetic length

$$\ell_B = \sqrt{ \frac{\hbar}{|q|\,|\mathbf B|} }$$

should be large compared with microscopic lattice scales for a local band trajectory. Equivalently, for a representative cell cross-section $A_c$, the flux test is

$$\frac{|q|\,|\mathbf B|A_c}{h} \ll 1.$$

The field-driven orbit must also avoid appreciable transfer to nearby bands. Separately, discrete Landau levels become spectrally resolved when a representative spacing $\hbar\omega_c$ is comparable to or larger than both thermal broadening $k_{\mathrm B}T$ and a disorder linewidth $\Gamma$. At that point a continuous phase-space trajectory is no longer a sufficient description of the spectrum. Landau Levels in Solids derives the resolved material spectrum. Quantum Oscillations connects orbit quantization to measurements, and Magnetic Translations explains the field-modified translation algebra.

Acceleration and local band curvature

Differentiate the group velocity along the trajectory. With the local inverse curvature tensor

$$\left( \mathsf M_n^{-1} \right)_{ij} = \frac{1}{\hbar^2} \frac{\partial^2\varepsilon_n} {\partial k_i\partial k_j},$$

the ordinary acceleration is

$$\dot v_i = \sum_j \left( \mathsf M_n^{-1} \right)_{ij} F_j, \qquad \mathbf F = \hbar\dot{\mathbf k}_c.$$

This equation is local: both the Hessian and the force are evaluated along the path. Force and acceleration need not be parallel in an anisotropic band, and a constant force need not produce constant acceleration in a nonparabolic band. Negative electron-band curvature is not by itself a positive carrier charge. Effective Mass distinguishes the masses extracted from different responses; Holes derives the positive-charge description near a full band.

Berry curvature and orbital magnetic moment

The ordinary equations depend only on $\varepsilon_n(\mathbf k)$. A wave packet also samples how $|u_{n\mathbf k}\rangle$ changes across momentum space. With the connection convention introduced above, define

$$\boldsymbol\Omega_n = \nabla_{\mathbf k} \times \boldsymbol{\mathcal A}_n.$$

For an isolated nondegenerate band in spatially uniform fields, the leading geometric form is

$$\dot{\mathbf r}_c = \frac{1}{\hbar} \nabla_{\mathbf k} \widetilde{\varepsilon}_n - \dot{\mathbf k}_c\times \boldsymbol\Omega_n,$$

where $\boldsymbol\Omega_n$ is the Berry curvature and

$$\widetilde{\varepsilon}_n = \varepsilon_n - \mathbf m_n\cdot\mathbf B$$

includes the band orbital magnetic moment. The force equation remains coupled to $\dot{\mathbf r}_c$, so the system must be solved consistently. In an inhomogeneous field, $-\nabla_{\mathbf r_c}\widetilde{\varepsilon}_n$ must also enter $\hbar\dot{\mathbf k}_c$; the uniform-field preview does not include that force.

The same first-order expansion changes the phase-space factor to

$$D_n = 1 - \frac{q}{\hbar} \mathbf B\cdot\boldsymbol\Omega_n.$$

At this order the orbital moment, phase-space measure, occupations and magnetization currents can all affect a measured response. Include them consistently rather than adding an anomalous velocity alone. Berry Curvature in Band Dynamics derives these material corrections, while Hall Effect distinguishes intrinsic and scattering-dependent transverse responses. Nonzero local curvature alone does not determine a Chern number or Hall coefficient.

When One Band Is Not Enough

An isolated band must remain separated over the declared significant-weight region, not merely at its center. A useful instantaneous gap is

$$\Delta_{\mathrm{iso}}(t) = \min_{\substack{ \mathbf k\in\mathcal K_a(t)\\ m\ne n }} \left| \varepsilon_m(\mathbf k) - \varepsilon_n(\mathbf k) \right|.$$

One needs $\Delta_{\mathrm{iso}}>0$ along the full trajectory. Gap size alone is not enough; the rate at which the basis changes also matters. For $m\ne n$, define the local adiabatic diagnostic

$$\eta_{mn}(\mathbf k,t) = \frac{ \hbar \left| \dot{\mathbf k}_c\cdot \langle u_{m\mathbf k}| \nabla_{\mathbf k}H(\mathbf k) |u_{n\mathbf k}\rangle \right| }{ |\varepsilon_m(\mathbf k)-\varepsilon_n(\mathbf k)|^2 }, \qquad \eta_{\max}(t) = \max_{\substack{ \mathbf k\in\mathcal K_a(t)\\ m\ne n }} \eta_{mn}(\mathbf k,t).$$

Single-band following requires $\eta_{\max}\ll1$ together with a discarded tail small enough that it does not change the requested observable. This instantaneous ratio is a useful local warning, not by itself a rigorous long-time error bound; smoothness, accumulated duration, and packet spreading must also be controlled. At an exact degeneracy, an individual-band eigenvector is not a unique smooth object and this ratio is singular; the retained object must be a multiband subspace.

Near an isolated avoided crossing, write the local projected Hamiltonian in a diabatic basis as

$$H_{\mathrm{loc}}(\mathbf k) = \frac{\delta(\mathbf k)}{2}\sigma_z + \Delta\sigma_x.$$

The minimum adiabatic gap is $2|\Delta|$. The band-dynamics problem supplies the diabatic sweep rate

$$\alpha = \left. \frac{d\delta}{dt} \right|_{\mathbf k_c} = \dot{\mathbf k}_c\cdot \nabla_{\mathbf k}\delta, \qquad \mathbf B=0 \ \Longrightarrow\ \alpha = \frac{q}{\hbar} \mathbf E\cdot \nabla_{\mathbf k}\delta.$$

Landau–Zener Transition derives the transition probability and explains the finite-time and linear-sweep assumptions. In a magnetic field, two nearby semiclassical orbits can undergo analogous field-driven transfer across a small band gap. This magnetic breakdown changes the orbit network and cannot be repaired by continuing a single-band path through the junction. Multiple junctions can also interfere coherently.

The principal one-band failure modes are distinct:

Failure	Lost assumption	Required change
exact or symmetry-enforced degeneracy	unique isolated eigenline	propagate the degenerate subspace with a multiband Hamiltonian
electric-field interband transfer	adiabatic following through an avoided crossing	use Landau–Zener or a full time-dependent multiband calculation
magnetic breakdown	orbit remains on one sheet	use a coupled orbit network with junction amplitudes
resolved orbital quantization	continuous phase-space spectrum	use Landau-level or magnetic-translation methods
packet splitting across branches	one localized center represents the state	propagate multiple packets or the full state

Strong field is therefore not one universal threshold. Interband transfer, magnetic breakdown, and resolved intraband quantization compare the field with different gaps, velocities, broadenings, and length scales.

Disorder, Collisions, and Spectral Coherence

The collisionless equations describe one packet between scattering events. A transport measurement instead samples an ensemble of occupied packets and a collision mechanism. A relaxation time $\tau$, mean free path $\ell\sim|\mathbf v|\tau$, and quantities such as $\omega_c\tau$ are not determined by $\varepsilon_n(\mathbf k)$ alone. Boltzmann Transport introduces distributions and collision integrals to calculate conductivity.

Smooth disorder can enter locally as an additional force if it varies on scales large compared with the packet. Atomic-scale or strong disorder scatters over a broad momentum range, can mix bands, and may destroy the trajectory through localization. Treating every collision as an instantaneous random reset is a model whose regime and conserved quantities must be stated.

For an interacting system, the propagated object must be a coherent quasiparticle pole or comparably sharp spectral ridge, not merely a Kohn–Sham eigenvalue or a broad maximum. A useful packet requires a lifetime long enough to traverse the distance and time being modeled and a residue and linewidth compatible with a band label. Coherent spectral weight can coexist with an incoherent background; when no sharp excitation survives, there is no unique $\varepsilon_n(\mathbf k)$ whose gradient defines a long-lived carrier velocity. Spectral Functions explains these poles, linewidths and continua.

Exercises

1. Finite-width Bloch oscillations

In the cosine band, take an initial Gaussian momentum distribution with center $k_0$ and variance $s^2$, narrow enough to use a local chart. A uniform field shifts every momentum by $qEt/\hbar$. Show how the packet-averaged velocity differs from the center value. What happens to the peak-to-peak displacement?

Solution

The Gaussian characteristic function gives

$$\langle e^{ika}\rangle =e^{ik_c(t)a}e^{-a^2s^2/2}.$$

A wrapped Gaussian on the Brillouin circle gives the same periodic moment. Therefore

$$\langle v(t)\rangle =\frac{2Ja}{\hbar}e^{-a^2s^2/2} \sin\!\left(k_0a+\frac{qEa}{\hbar}t\right).$$

The period remains $T_B=h/(|qE|a)$, but the peak-to-peak displacement is $4J e^{-a^2s^2/2}/|qE|$. For $as=0.25$, the amplitude is about $0.969$ times the center approximation. The packet’s changing spatial width does not invalidate this expectation-value result in the ideal single-band model; additional forces, collisions or interband transfer can.

2. A closed momentum path with transverse drift

For a square-lattice band

$$\varepsilon(\mathbf k)=-2J_x\cos(k_xa)-2J_y\cos(k_ya),$$

take $\mathbf B=0$ and $\mathbf E=E\hat{\mathbf x}$. Show that the momentum returns modulo a reciprocal vector after $T_B$, but the real-space path need not close.

Solution

$k_x(t)=k_{x0}+qEt/\hbar$ while $k_y=k_{y0}$ is constant. The $x$ velocity averages to zero over a full reciprocal period. The transverse velocity is constant:

$$v_y=\frac{2J_ya}{\hbar}\sin(k_{y0}a),\qquad \Delta y=v_yT_B.$$

Thus the center drifts unless $\sin(k_{y0}a)=0$. This motion requires no Berry curvature. The electric field does no work through the transverse displacement, so it is consistent with the band energy returning after each cycle.

3. Magnetic invariants

For constant $\mathbf B$ and $\mathbf E=0$, prove that both $\varepsilon_n(\mathbf k_c)$ and $\widetilde{\mathbf k}_c\cdot\mathbf B$ are conserved, where $\widetilde{\mathbf k}_c$ is a continuous repeated-zone lift. What geometric curve does the packet follow before projection to the Brillouin torus?

Solution

The energy derivative is $q\mathbf v\cdot(\mathbf v\times\mathbf B)=0$. For the second,

$$\frac{d}{dt} (\widetilde{\mathbf k}_c\cdot\mathbf B) = \frac{q}{\hbar} (\dot{\mathbf r}_c\times\mathbf B) \cdot\mathbf B = 0.$$

The lifted path lies simultaneously on the periodically continued surface $\varepsilon_n(\widetilde{\mathbf k})=\text{constant}$ and on a plane of constant component parallel to $\mathbf B$. The physical path is its projection to the Brillouin torus. Whether the lift is compact and closed or instead open is a property of the band geometry, not of the plotting convention. A projected path can close modulo a reciprocal vector while its lift remains open, so torus closure alone does not define a compact cyclotron orbit.

4. Adiabaticity at an avoided crossing

Let

$$H_{\mathrm{loc}}(k) = \frac{\beta(k-k_0)}{2}\sigma_z + \Delta\sigma_x,$$

with $\Delta\ne0$. A uniform electric field drives the one-dimensional packet with $\mathbf B=0$. Derive the diabatic sweep rate $\alpha$, the minimum adiabatic gap and the local adiabatic ratio at the crossing. State the assumptions needed to use a Landau–Zener probability.

Solution

Here $\delta(k)=\beta(k-k_0)$ and $\hbar\dot k_c=qE$, so

$$\alpha = \frac{d\delta}{dt} = \beta\dot k_c = \frac{\beta qE}{\hbar}.$$

The eigenvalue separation is $\sqrt{\beta^2(k-k_0)^2+4|\Delta|^2}$, so its minimum is $2|\Delta|$. At the crossing, the adiabatic states are $\sigma_x$ eigenstates. Since $\dot H=\alpha\sigma_z/2$, its off-diagonal matrix element has magnitude $|\alpha|/2$, giving

$$\eta(0)=\frac{\hbar|\alpha|}{8|\Delta|^2} =\frac{|\beta qE|}{8|\Delta|^2}.$$

A small local ratio supports adiabatic following but is not a complete long-time bound. To use Landau–Zener Transition, verify that $\mathcal K_a$ samples only this crossing, other bands are remote, the diabatic detuning is locally linear, $\Delta$ is approximately constant, coherence survives the passage, and the initial and final windows are far enough from the crossing for the chosen asymptotic or finite-window formula.

References

• N. W. Ashcroft and N. D. Mermin, Solid State Physics, Holt, Rinehart and Winston, 1976 — standard band dynamics, transport, and magnetic-orbit treatment.
• E. I. Blount, “Formalisms of Band Theory,” Solid State Physics 13, 305–373, 1962, doi:10.1016/S0081-1947(08)60459-2 — crystal-momentum and wave-packet formalism.
• M.-C. Chang and Q. Niu, “Berry Phase, Hyperorbits, and the Hofstadter Spectrum: Semiclassical Dynamics in Magnetic Bloch Bands,” Physical Review B 53, 7010, 1996, doi:10.1103/PhysRevB.53.7010.
• W. V. Houston, “Acceleration of Electrons in a Crystal Lattice,” Physical Review 57, 184, 1940, doi:10.1103/PhysRev.57.184 — uniform-field evolution and interband-transition boundary.
• C. Kittel, Introduction to Solid State Physics, 8th ed., Wiley, 2004 — textbook treatment of Bloch electrons, effective mass, and semiclassical motion.
• M. P. Marder, Condensed Matter Physics, 2nd ed., Wiley, 2010 — modern pedagogical treatment of electron dynamics in bands.
• R. M. Martin, Electronic Structure: Basic Theory and Practical Methods, Cambridge University Press, 2004, doi:10.1017/CBO9780511805769 — band objects, effective Hamiltonians, and electronic-structure context.
• G. Panati, H. Spohn, and S. Teufel, “Effective Dynamics for Bloch Electrons: Peierls Substitution and Beyond,” Communications in Mathematical Physics 242, 547–578, 2003, doi:10.1007/s00220-003-0950-1 — controlled isolated-family dynamics for slowly varying scalar and vector potentials.
• J. R. Reitz, “Magnetic Breakdown in Metals,” Journal of Physics and Chemistry of Solids 25, 53–58, 1964, doi:10.1016/0022-3697(64)90161-1.
• D. Shoenberg, Magnetic Oscillations in Metals, Cambridge University Press, 1984 — magnetic orbits, breakdown, and oscillatory phenomena.
• S. H. Simon, The Oxford Solid State Basics, Oxford University Press, 2013 — accessible derivations of band velocity, acceleration, and Bloch oscillations.
• G. Sundaram and Q. Niu, “Wave-Packet Dynamics in Slowly Perturbed Crystals: Gradient Corrections and Berry-Phase Effects,” Physical Review B 59, 14915–14925, 1999, doi:10.1103/PhysRevB.59.14915 — systematic isolated-band wave-packet dynamics.
• G. H. Wannier, “Dynamics of Band Electrons in Electric and Magnetic Fields,” Reviews of Modern Physics 34, 645–655, 1962, doi:10.1103/RevModPhys.34.645 — field-driven band motion and its limits.
• D. Xiao, M.-C. Chang, and Q. Niu, “Berry Phase Effects on Electronic Properties,” Reviews of Modern Physics 82, 1959–2007, 2010, doi:10.1103/RevModPhys.82.1959 — review of geometric semiclassics and its response applications.
• C. Zener, “Non-Adiabatic Crossing of Energy Levels,” Proceedings of the Royal Society A 137, 696–702, 1932, doi:10.1098/rspa.1932.0165 — canonical avoided-crossing transition.