Time-Independent Schrödinger Equation

The time-dependent Schrödinger equation told you that quantum states evolve; it did not tell you how to actually solve anything. This article is where the solving begins. For the enormous class of problems in which the potential energy does not change with time — an electron in an atom, a vibrating molecule, a particle in any trap that just sits there — one sixty-second manoeuvre, separation of variables, splits the full equation into two far friendlier pieces: a trivial clock equation for the time factor, and the equation that will dominate the rest of your course,

\hat H\psi = E\psi.

That is the time-independent Schrödinger equation, and it is a new kind of problem: an eigenvalue problem. You are no longer handed an energy and asked what happens; you are asked which energies the system permits at all, and which spatial shapes go with them. Hand the equation a trapping potential and it returns a discrete menu — $E_1, E_2, E_3, \ldots$ — with one standing-wave profile per entry. That menu is why atoms are stable, why spectral lines are sharp, and why "quantum" means what it means. You arrive with the full law of motion from Time-Dependent Schrödinger Equation, the state object from The Wavefunction, and the probability rules from Probability Density and Normalization. What you take away is the master reduction: solve $\hat H\psi = E\psi$ once, and — as the next article, Stationary States and Separation of Variables, develops — every possible motion of the system is bookkeeping with the solutions.

Key takeaways

When $V$ has no time dependence, seek product solutions $\Psi(x,t) = \psi(x)\,\varphi(t)$ : the full PDE splits into two ODEs joined by one constant, $E$ .
The time piece is universal: $\varphi(t) = e^{-iEt/\hbar}$ — the same phase clock for every problem, spinning at rate $E/\hbar$ .
The space piece is the time-independent Schrödinger equation (TISE): $\hat H\psi = -\frac{\hbar^2}{2m}\psi'' + V\psi = E\psi$ — an eigenvalue problem.
Solving it means finding the allowed energies (eigenvalues) and their spatial profiles (eigenfunctions) together — neither is an input.
Confinement plus normalizability makes the allowed energies discrete for bound systems; unbound systems get a continuum instead.
The separation constant deserves the name "energy" because $\hat H$ is the energy operator: on a product solution it returns exactly $E$ .
Key idea: separation of variables converts quantum dynamics into an eigenvalue problem — find the standing-wave shapes and their energies, and time evolution reduces to attaching phase clocks.

The core insight

For a time-independent Hamiltonian, separation of variables turns the dynamics into the eigenvalue problem $\hat H\psi = E\psi$ ; its solutions are states of definite energy whose time dependence is the pure phase $e^{-iEt/\hbar}$ , and their superpositions generate all possible time evolution.

Where this lives today

Ask a supercomputer what materials science does all day and the answer is: solve the TISE, millions of times. Density functional theory — the workhorse behind battery-electrode screening, catalyst design, and semiconductor engineering — reduces the many-electron problem to a self-consistent loop whose inner step is exactly $\hat H\psi = E\psi$ for one electron in an effective potential; codes like VASP and Quantum ESPRESSO diagonalize that operator in a plane-wave basis, and the resulting eigenvalue menus are the band structures on which chip design rests. The two-state worked example is equally alive. The ammonia doublet drove the first maser (1954); today, engineered double-well systems — superconducting flux qubits, phosphorus donors in silicon — are two-row TISE problems whose splitting $2\Delta$ is tuned by fabrication and read out as a qubit frequency. When a quantum-hardware paper quotes "qubit transition at 5.1 GHz," it is reporting the eigenvalue gap of a deliberately manufactured Hamiltonian — the inversion doublet, industrialized.

The physical picture

Ask a guitar string what it can do, and the honest answer comes in two parts. It can vibrate in certain privileged shapes — the fundamental, the octave, the fifth above that — each shape oscillating at its own fixed frequency while its profile stays put. And it can do everything else, but everything else is just a chord: a sum of those privileged shapes, each ringing at its own rate. The privileged shapes are the string's modes; find them once and you understand every possible motion of the string. The time-independent Schrödinger equation is the mode-finding equation for quantum mechanics. A mode here is a wavefunction whose spatial shape holds still while a single complex clock — the phase factor $e^{-iEt/\hbar}$ — does all the turning, and the energy $E$ plays exactly the role of the mode's frequency, through $\omega = E/\hbar$ . The boundary conditions do the selecting: the wave must die away where the particle cannot escape to, and only special energies produce shapes that manage it. An organ pipe sounds discrete pitches for the same reason a trapped electron has discrete energies — the wave has to fit.

Figure

Separation of variables: one hard PDE becomes shape times clock

The deeper shift is in what kind of question you are answering. The time-dependent equation is a machine: state in, evolution out. The TISE is a filter: it accepts a candidate energy only if some well-behaved shape $\psi$ satisfies the equation at that energy. For a confining potential, almost every candidate fails — the shape blows up at infinity — and the survivors form a discrete ladder. The equation does not just describe the system; it audits it.

Hold both pictures — modes that fit, and an audit that returns a menu.

From the law of motion to the eigenvalue problem

Start from the full law, exactly as Time-Dependent Schrödinger Equation left it. For a single particle of mass $m$ in one dimension,

i\hbar\,\frac{\partial \Psi(x,t)}{\partial t} = -\frac{\hbar^2}{2m}\frac{\partial^2 \Psi}{\partial x^2} + V(x)\,\Psi \qquad (1)

where: $\Psi(x,t)$ — the state; $V(x)$ — a potential that, crucially for everything below, carries no $t$ . That single structural fact — the Hamiltonian $\hat H = -\frac{\hbar^2}{2m}\partial_x^2 + V(x)$ is the same operator at every instant — is the entire licence for what follows. If $V$ depended on time, the separation below would fail at the first step, and you would need the heavier tools of the Quantum Dynamics chapter. The manoeuvre is to look for solutions of a deliberately special form — a fixed spatial profile multiplied by a pure function of time:

\Psi(x,t) = \psi(x)\,\varphi(t) \qquad (2)

where: $\psi(x)$ — the shape, carrying all the $x$ -dependence; $\varphi(t)$ — the clock, carrying all the $t$ -dependence. This is an ansatz, not an assumption about all states: most wavefunctions do not factor this way, and no one claims they do. The strategy — the oldest in mathematical physics — is to find the special factorizable solutions first, and then build every other solution out of them by superposition, which the linearity of (1) permits. The building programme is the next article's business; this article finds the bricks.

Derivation ASeparation of variables — one PDE becomes two ODEs

We are about to perform the single most consequential calculation of the course: inserting the product ansatz (2) into the law of motion (1) and watching the equation fall apart into two ordinary differential equations, one in $t$ alone and one in $x$ alone, hinged by a constant. The move that does all the work is dividing through by $\psi\varphi$ — after that, an inescapable logical step (a function of $t$ equalling a function of $x$ ) forces both sides to be constant. Watch where the time-independence of $V$ gets used: without it, the $x$ -side would smuggle in a $t$ and the argument would die.

Step 1. Insert the ansatz. With $\Psi = \psi(x)\varphi(t)$ , the time derivative touches only $\varphi$ and the space derivatives touch only $\psi$ : $i\hbar\,\psi\,\dfrac{d\varphi}{dt} = -\dfrac{\hbar^2}{2m}\varphi\,\dfrac{d^2\psi}{dx^2} + V(x)\,\psi\varphi$ . Why: partial derivatives become ordinary ones — each factor depends on one variable only, which is the entire point of the ansatz.

Step 2. Divide by $\psi\varphi$ . $i\hbar\,\dfrac{1}{\varphi}\dfrac{d\varphi}{dt} = -\dfrac{\hbar^2}{2m}\dfrac{1}{\psi}\dfrac{d^2\psi}{dx^2} + V(x)$ . Why: this is the sorting step — the left side now depends on $t$ only, the right side on $x$ only. Here is where $V = V(x)$ is load-bearing: a $V(x,t)$ would leave time on the right and the sides could never be disentangled.

Step 3. Argue the hinge. The left side is a function of $t$ ; the right is a function of $x$ . Vary $t$ while holding $x$ fixed: the right side cannot change, so the left side — supposedly free to change — must in fact be constant. Call the constant $E$ . Why: a function of $t$ that equals a function of $x$ for all $t$ and all $x$ can only be a constant; this "separation constant" is the one thread joining the two halves of the problem.

Step 4. Read off the two ODEs. Setting each side equal to $E$ :

i\hbar\,\frac{d\varphi}{dt} = E\,\varphi \qquad (3)

-\frac{\hbar^2}{2m}\frac{d^2\psi}{dx^2} + V(x)\,\psi = E\,\psi \qquad (4)

Why: (3) is first-order with constant coefficient — solvable on sight, once and for all. (4) is the time-independent Schrödinger equation: all the difficulty, and all the physics of the specific problem, has been quarantined here.

Step 5. Compress the notation. Equation (4) is $\hat H\psi = E\psi$ with $\hat H = -\frac{\hbar^2}{2m}\partial_x^2 + V(x)$ — the Hamiltonian acting on the shape returns the same shape, scaled by $E$ . Why: writing it this way exposes the structure that survives into every later chapter: an operator equation whose solutions come in (eigenvalue, eigenfunction) pairs. *Verified: symbolic check (sympy) that $\Psi = \psi(x)e^{-iEt/\hbar}$ satisfies (1) exactly when $\psi$ satisfies (4); the separation steps follow Griffiths §2.1 and Cohen-Tannoudji Ch. III.*

The clock equation deserves its sixty seconds of fame before we hand it its pension.

Go deeper · Derivation B · 2 minSolving the clock — and why the constant deserves the name energy

Derivation BSolving the clock — and why the constant deserves the name energy

Equation (3) is the easiest differential equation in quantum mechanics, and its solution is universal: the same time factor serves every time-independent problem you will ever solve, from the square well to the hydrogen atom. We solve it, fix the normalization convention, and then face the naming question honestly — why should the separation constant $E$ , born as a mere bookkeeping device in Step 3 above, be called the energy? The answer ties back to the operator dictionary built in Time-Dependent Schrödinger Equation.

Step 1. Solve the first-order ODE. $\dfrac{d\varphi}{dt} = -\dfrac{iE}{\hbar}\varphi$ has the solution $\varphi(t) = C\,e^{-iEt/\hbar}$ . Why: a first-order linear equation with constant coefficient always exponentiates; the $i$ makes the exponent imaginary, so $\varphi$ rotates on the unit circle rather than growing or decaying — $|\varphi(t)| = |C|$ forever.

Step 2. Absorb the constant. Fold $C$ into $\psi(x)$ and write the product solution as $\Psi(x,t) = \psi(x)\,e^{-iEt/\hbar}$ . Why: only the product $\psi\varphi$ is physical; the split of constants between factors is pure convention, so we make the clock start at exactly $1$ .

Step 3. Read the clock physically. The phase turns at angular rate $\omega = E/\hbar$ — this is the Planck–Einstein relation $E = \hbar\omega$ from Wave–Particle Duality, reappearing as a derived feature. Why: the separation constant sets the rotation rate of the state's phase, exactly the role frequency plays for a classical mode.

Step 4. Justify the name. Apply the Hamiltonian — the operator that Time-Dependent Schrödinger Equation built as the reader of total energy — to the product solution: $\hat H\Psi = (\hat H\psi)\,e^{-iEt/\hbar} = E\psi\,e^{-iEt/\hbar} = E\,\Psi$ . The energy operator, asked for the state's energy, answers " $E$ , everywhere and always." Why: the constant is not merely analogous to energy; it is the eigenvalue of the energy operator. The full argument that the state has energy $E$ with zero spread — no statistical scatter at all — needs the expectation-value machinery and is completed in Stationary States and Separation of Variables.

Step 5. Note what is deferred. That $E$ must be real (a complex $E$ would make $|\varphi|$ grow or decay and quietly destroy normalization), that $E$ must exceed the minimum of $V$ , and that bound-state $\psi$ can be chosen real — those three sharpenings are owned by Time-Independent Schrödinger Equation in One Dimension, which re-runs this separation for the 1D chapter with full rigour. Why: honest bookkeeping of what has and has not been shown; the first pass takes reality of $E$ on physical faith. *Verified: sympy solves (3) to $Ce^{-iEt/\hbar}$ and confirms $\hat H(\psi e^{-iEt/\hbar}) = E\,\psi e^{-iEt/\hbar}$ given $\hat H\psi = E\psi$ ; conventions per Griffiths §2.1.*

What kind of problem is this? The eigenvalue reading

Equation (4) looks like something you could solve the way you solve any ODE — pick your favourite $E$ , integrate, done. That reading is wrong in the most instructive way. Write it abstractly:

\hat H\,\psi = E\,\psi \qquad (5)

where: $\hat H$ — a fixed linear operator (the system); $\psi$ — an unknown function (the shape); $E$ — an unknown number (the level). Both unknowns must be found together, subject to $\psi$ being physically admissible — normalizable for a bound state, and well-behaved in the sense catalogued later by Admissibility Conditions on Wavefunctions. For generic $E$ the admissible solution simply does not exist; the values of $E$ for which it does are the eigenvalues of $\hat H$ , and the matching shapes are its eigenfunctions. The set of eigenvalues is the system's spectrum — the same word as "spectral lines," and that is no accident: the discrete menu of a trapped electron is what a spectrometer photographs, one line per pair of menu entries. If "operator eats function, returns the same function times a number" feels abstract, there is a finite-dimensional version you already know from linear algebra: a matrix eating a vector. Quantum mechanics uses that version constantly — whenever only a few basis states matter, $\hat H$ is a small matrix, and the TISE is matrix diagonalization. The second worked example solves a genuine two-state system this way, and the abstract theory of eigenvalues and eigenstates gets its own deeper treatment in the Formalism chapter (Eigenvalues and Eigenstates). One vocabulary point saves endless confusion. A product solution $\Psi = \psi_n e^{-iE_nt/\hbar}$ is called a stationary state, because — as the next article proves — every probability you can extract from it is frozen in time. But the equation itself is only the filter; the properties of what passes the filter, and the superposition machinery that turns the menu into full dynamics, belong to Stationary States and Separation of Variables: here, where the equation comes from; there, what its solutions are good for.

The character of the spectrum is decided by whether the particle is trapped, and the full classification — with its curvature arguments and its delta-normalized continuum — is owned by Bound versus Scattering States, already on your shelf. What matters for using the TISE is the headline: confinement discretizes. If $E$ lies below the potential's values at infinity, $\psi$ must decay in the forbidden outskirts, and demanding decay on both sides is one condition too many for generic $E$ — only a discrete ladder survives. If $E$ lies above the confinement edge, oscillatory solutions exist for every $E$ : a continuum, the scattering band of the second figure. The free particle is the cleanest instance of the second case. With $V = 0$ , equation (4) reads $-\frac{\hbar^2}{2m}\psi'' = E\psi$ , solved by $\psi = e^{ikx}$ with

E = \frac{\hbar^2 k^2}{2m} \qquad (6)

where: $k$ — any real wavenumber; every non-negative energy is allowed, and the "eigenfunctions" are the non-normalizable plane waves whose careful handling Bound versus Scattering States set up. The first worked example puts numbers on (6). The quantitative one-dimensional laboratory — wells, barriers, the oscillator — then runs this machinery potential by potential throughout the ONED chapter, beginning with Time-Independent Schrödinger Equation in One Dimension.

Limiting cases

Case	Result	Meaning
$V = V(x,t)$ genuinely time-dependent	separation fails at Step 2	no stationary states; driven-system tools (Quantum Dynamics chapter) required
$V = 0$ everywhere	$\psi = e^{\pm ikx}$ , $E = \hbar^2k^2/2m$ , all $E \geq 0$	continuous spectrum; plane waves, handled with delta normalization
confining $V$ , $E$ below the rim	discrete ladder $E_1 < E_2 < \cdots$	normalizability on both sides over-determines generic $E$ — quantization
$E$ above the confinement rim	oscillatory $\psi$ for every such $E$	scattering continuum; classification per Bound versus Scattering States
$\hbar \to 0$ at fixed classical scales	level spacing $\to 0$	the discrete menu blurs into the classical continuum — correspondence principle

Common trap“Only the separable solutions are physical — a real state must have a definite energy”

Backwards, and worth killing early. The product states $\psi_n e^{-iE_nt/\hbar}$ are the rarest states in nature, not the only ones: a generic wavefunction is a superposition of many of them and has no single energy at all. What makes the separable solutions important is not that states must look like them but that every state can be built from them — they are the alphabet, not the whole language. A twin error hides in the phrase "solve the Schrödinger equation": if you hand equation (4) an arbitrary energy and integrate numerically, you will always get some mathematical solution — it will just blow up at infinity and represent nothing. Solving the TISE means finding the $(E, \psi)$ pairs that survive the admissibility audit, not integrating at the energy of your choice. Keep the two unknowns welded together.

Worked examples

1 · Auditioning a plane wave (confidence builder)

An electron moves freely ( $V = 0$ ) with wavenumber $k = 5.0\times10^{9}\ \mathrm{m^{-1}}$ (wavelength $\lambda = 2\pi/k = 1.26\ \mathrm{nm}$ ). Verify that $\psi = e^{ikx}$ passes the TISE audit, find the eigenvalue $E$ , and give the period of the associated phase clock.

Step 1. Apply the Hamiltonian: $\hat H\psi = -\frac{\hbar^2}{2m}\frac{d^2}{dx^2}e^{ikx} = \frac{\hbar^2k^2}{2m}e^{ikx}$ . The wave comes back unchanged, scaled by a number — it is an eigenfunction, with $E = \hbar^2k^2/2m$ .

Step 2. Numbers: $E = \dfrac{(1.055\times10^{-34})^2(5.0\times10^9)^2}{2\times9.109\times10^{-31}} = 1.53\times10^{-19}\ \mathrm{J} = \boxed{0.95\ \mathrm{eV}}$ .

Step 3. The clock: $\omega = E/\hbar = 1.45\times10^{15}\ \mathrm{rad/s}$ , so the phase completes a revolution every $T = 2\pi/\omega = \boxed{4.3\ \mathrm{fs}}$ .

Step 4. Read the result. Because $V = 0$ confines nothing, every $k$ would have passed this audit — the free spectrum is the continuum $E \geq 0$ of equation (6). The same audition inside a well would have failed for almost every energy; discreteness is confinement's signature, not the equation's.

2 · A genuine two-state TISE: the inversion doublet (exam level)

In many double-well systems — the ammonia molecule's nitrogen, which can sit above or below the H₃ plane, is the classic — only two basis states matter: "left" ( $L$ ) and "right" ( $R$ ), each with energy $E_0$ , coupled by tunnelling through the barrier with strength $\Delta > 0$ . The Hamiltonian is the matrix $H = \begin{pmatrix} E_0 & -\Delta \\ -\Delta & E_0 \end{pmatrix}$ acting on coefficient vectors $(c_L, c_R)$ . Solve this TISE exactly. In ammonia the resulting splitting corresponds to the famous $\nu = 23.87\ \mathrm{GHz}$ microwave line; find $\Delta$ in eV.

Step 1. Set up the eigenvalue problem: $H\mathbf{c} = E\mathbf{c}$ requires $\det(H - E\,\mathbb{1}) = 0$ , i.e. $(E_0 - E)^2 - \Delta^2 = 0$ .

Step 2. Solve: $E_\pm = E_0 \mp \Delta$ — two levels, symmetric about $E_0$ . The eigenvectors: for $E = E_0 - \Delta$ , $\mathbf{c} = \tfrac{1}{\sqrt2}(1, 1)$ — the symmetric combination; for $E = E_0 + \Delta$ , $\mathbf{c} = \tfrac{1}{\sqrt2}(1, -1)$ — the antisymmetric one. Neither eigenstate is "left" or "right": the stationary shapes straddle both wells.

Step 3. The splitting is $E_+ - E_- = 2\Delta$ . Matching it to the microwave photon: $2\Delta = h\nu = (6.626\times10^{-34})(23.87\times10^9) = 1.582\times10^{-23}\ \mathrm{J}$ , so $\Delta = 7.91\times10^{-24}\ \mathrm{J} = \boxed{4.94\times10^{-5}\ \mathrm{eV}}$ — about $49\ \mathrm{\mu eV}$ , a whisper compared with the eV-scale level spacings of the wells themselves.

Step 4. Read the structure. This is the TISE — operator, eigenvalues, eigenfunctions — with the infinite-dimensional function space compressed to two dimensions, and it displays the equation's character perfectly: the allowed energies ( $E_0 \pm \Delta$ ) and the allowed shapes (symmetric and antisymmetric) emerge together, and the "menu" has exactly two entries. A molecule prepared in the left well is a superposition of both eigenstates and oscillates between wells at the Bohr frequency $2\Delta/\hbar$ — but that dynamical story is precisely where Stationary States and Separation of Variables picks up.

Origins · 1 minJoseph Liouville and how this was discovered

Joseph Liouville

1809–1882 · Paris

The mathematics of this article was a French production of the 1830s, ninety years ahead of its physics. Liouville, with Charles-François Sturm, built a complete theory of exactly this species of problem: a second-order differential operator on an interval, boundary conditions attached, and the question for which values of a parameter does an admissible solution exist? Their answers — an infinite discrete ladder of eigenvalues, one more node in each successive eigenfunction, orthogonality between eigenfunctions, and expansion of arbitrary functions in the resulting basis — read today like a table of contents for a quantum mechanics course, written before anyone knew the electron existed. Liouville himself was a prodigious builder of the discipline's infrastructure: he founded the Journal de mathématiques pures et appliquées in 1836 and edited it for four decades, rescuing (among much else) the manuscripts of Galois from oblivion. When Schrödinger's 1926 equation turned out to be a Sturm–Liouville problem, the toolbox was already a century deep — quantization was, mathematically, a solved problem waiting for its physical question.

History

1836

— Sturm and Liouville publish the theory of eigenvalue problems for second-order ODEs: discrete spectra, oscillation theorems, orthogonal eigenfunctions.

1885

— Balmer fits hydrogen's visible lines to a formula; the discrete menu is measured before it is explained.

1910

— Weyl classifies singular boundary-value problems (limit point vs limit circle), extending Sturm–Liouville theory to infinite domains.

1926

— Schrödinger's first Quantisierung als Eigenwertproblem paper poses mechanics as an eigenvalue problem and solves hydrogen from the TISE.

1926

— The oscillator, rotor, and Stark effect fall in the sequels; "eigenvalue" enters the physicist's vocabulary for good.

Going further

Texts. Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., §2.1 · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §2.4 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 4 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. 1, Ch. III · Townsend, A Modern Approach to Quantum Mechanics, 2nd ed., Ch. 6.

Provenance. Derivations and equations follow Griffiths & Schroeter §2.1–2.2, Sakurai & Napolitano §2.4, Shankar Ch. 4, Cohen-Tannoudji Vol. 1 Ch. III, and Townsend Ch. 6. A verifier should recompute: the separation-of-variables steps yielding $\varphi(t) = e^{-iEt/\hbar}$ ; the free-particle eigenvalue $E = \hbar^2k^2/2m$ (here $k = 5.0\times10^9\ \mathrm{m^{-1}}$ → $E = 0.95\ \mathrm{eV}$ , clock period $4.3\ \mathrm{fs}$ ); and the two-state diagonalization $E_\pm = E_0 \mp \Delta$ with the ammonia numbers $2\Delta = h\times23.87\ \mathrm{GHz}$ → $\Delta = 4.94\times10^{-5}\ \mathrm{eV}$ . Constant used: $\hbar = 1.055\times10^{-34}\ \mathrm{J\,s}$ (CODATA). The base article's infinite-well and hydrogen worked spectra are owned in this track by Infinite Square Well and the Hydrogen chapter and are not duplicated here.

Prerequisites. Time-Dependent Schrödinger Equation · The Wavefunction · Probability Density and Normalization. Leads to. Stationary States and Separation of Variables (the properties and the superposition machinery) · Probability Current and the Continuity Equation · then the one-dimensional laboratory, beginning with Time-Independent Schrödinger Equation in One Dimension.

If most states don't factor into shape times clock, why is the product ansatz not a cheat?

Because linearity converts a lucky guess into a complete method. The time-dependent equation is linear: any sum of solutions is a solution. So once the product solutions $\psi_n e^{-iE_nt/\hbar}$ are in hand, every superposition $\sum_n c_n\psi_n e^{-iE_nt/\hbar}$ solves the full equation too — and it turns out (the completeness property, developed from Stationary States and Separation of Variables onward) that every physically admissible initial state can be written this way. The ansatz never claims generic states factor; it claims the factorizable ones span everything, which is a theorem, not a hope.

Where exactly does the separation break if V depends on time?

At the division step. After dividing by $\psi\varphi$ , the right side is $-\frac{\hbar^2}{2m}\frac{\psi''}{\psi} + V$ . If $V = V(x,t)$ , that side depends on both variables, and the "function of $t$ equals function of $x$ " hinge never engages — no constant, no split, no stationary states. Physically: a driven system has no privileged standing-wave modes because the driver keeps redefining what "standing still" means. The tools for that regime — time-ordered evolution, the interaction picture, perturbation theory — are the Quantum Dynamics and Time-Dependent Phenomena chapters' business.

Is E an input or an output? In classical mechanics I choose the energy.

Output — and this is the deepest cultural shift the equation demands. Classically you prepare a particle with any energy you like; the dynamics then respects your choice. Quantum mechanically, for a bound system, the equation plus admissibility dictates the menu of possible sharp energies before you prepare anything; preparation only chooses how the state distributes itself over that menu. For unbound motion the continuum restores something like the classical freedom. The audit picture is the safe one: you propose, the boundary conditions dispose.

Why does the time factor come out complex — shouldn't a physical oscillation be a cosine?

The first-order-in-time structure of the parent equation forces it: $i\hbar\dot\varphi = E\varphi$ has $e^{-iEt/\hbar}$ as its solution, full stop, and Time-Dependent Schrödinger Equation showed that a real wavefunction cannot evolve at all. The complex clock is also exactly why stationary states are stationary: $|e^{-iEt/\hbar}|^2 = 1$ , so the density never learns what time it is. A cosine clock would put the time dependence into $|\Psi|^2$ — real standing waves in classical physics visibly throb; quantum stationary states do not. The complexity is doing physical work.

The same E appears in the clock and in the shape equation. Is that a coincidence of notation?

It is the entire content of the separation. One constant was born at the hinge step, and it went both ways: into the clock as the rotation rate $E/\hbar$ , into the shape equation as the eigenvalue. That double life is what welds "the energy of the state" (an operator statement, $\hat H\psi = E\psi$ ) to "the frequency of the state" (a dynamical statement, $\omega = E/\hbar$ ) — the Planck–Einstein relation reborn as bookkeeping. Two shapes with different eigenvalues carry clocks at different rates, and every observable oscillation in quantum mechanics traces back to that rate difference.

Is the two-state matrix example really the same TISE as the differential equation?

Genuinely the same problem in a smaller wardrobe. The differential operator acts on the infinite-dimensional space of functions $\psi(x)$ ; the matrix acts on the two-dimensional space of coefficient pairs $(c_L, c_R)$ — a deliberate truncation, excellent whenever the neglected states sit far away in energy. "Eigenvalue problem" is the common structure: operator, admissible vectors, the $(E, \text{shape})$ pairs that survive. The Formalism chapter (Hilbert Space and State Vectors, Eigenvalues and Eigenstates) makes the function-space and matrix pictures two coordinates on one geometry; until then, let the $2\times2$ case be your intuition anchor for the infinite one.

What stops me from just integrating the shape equation numerically at any energy I want?

Nothing stops the integration — the solution is what fails. Pick a generic $E$ inside a well, integrate outward, and $\psi$ will diverge exponentially on at least one side: a perfectly good mathematical function representing no physical state, because it cannot be normalized. The numerical method that takes this seriously — integrate from both sides, tune $E$ until the pieces match smoothly — is the "shooting method," and it is precisely a root-search for eigenvalues. The blow-up is not a nuisance; it is the mechanism of quantization made visible on your screen.