Conditions on Acceptable Wavefunctions

Every quantum chemistry calculation performed today — whether a Hartree–Fock computation on a small molecule or a density-functional theory simulation of a battery cathode material — implicitly builds its basis functions (Gaussian-type orbitals, Slater-type orbitals, plane waves) to already satisfy these acceptability conditions, so that the resulting molecular orbitals are automatically well-behaved. Gaussian basis functions, universal in computational chemistry packages such as Gaussian, ORCA, and Psi4, are chosen partly because their smooth, rapidly decaying, square-integrable form makes numerical integration (needed for millions of two-electron repulsion integrals) tractable, while still respecting continuity and normalizability exactly. In spectroscopy, the requirement of single-valuedness under $\phi \to \phi + 2\pi$ is exactly what quantizes the projection of angular momentum in rotational spectroscopy, giving integer or half-integer quantum numbers that show up directly in microwave and rotational Raman spectra used to determine bond lengths of small molecules in the gas phase (important in atmospheric chemistry and astrochemistry, where rotational transitions of species like CO reveal the composition of interstellar clouds). In materials science, particle-in-a-box reasoning built entirely on these boundary/regularity conditions is the working model for quantum dots — nanocrystals of semiconductor material (CdSe, PbS, and similar) whose electronic confinement energies, and hence the color of light they emit, are tuned by controlling the physical size of the "box." The same square-integrability and boundary-matching logic used here for the idealized 1-D box is what predicts how quantum dot emission color shifts with particle size, directly informing the design of quantum-dot displays and biomedical fluorescent tags. In drug design and enzymology, tunneling calculations for hydrogen or proton transfer across finite energy barriers — relevant to enzyme catalysis and to understanding kinetic isotope effects — rely on exactly the finite-step matching conditions (continuity of $\psi$ and $d\psi/dx$) worked out here, since tunneling probabilities are computed by matching oscillatory and exponentially decaying pieces of $\psi$ across the barrier boundaries.

Ask the tutor: Why is a kink in $d\psi/dx$ forbidden at a finite step but allowed at an infinite wall — can you walk through the integral argument again with a specific numeric potential?

Consider what a wavefunction is actually asked to do. According to the Born interpretation, $|\psi(\mathbf{r})|^2 d\tau$ is the probability of finding a particle in the infinitesimal volume $d\tau$ around $\mathbf{r}$. A probability must be a single, definite, non-negative real number at each point — there cannot be two different probabilities of finding the electron at the same location. This immediately rules out multivalued functions. If $\psi$ depended on an angular coordinate $\phi$ in a way that gave $\psi(\phi) \ne \psi(\phi + 2\pi)$, then walking once around a circle and returning to the same physical point would hand you a different probability density than you started with — a logical contradiction, since $\phi$ and $\phi + 2\pi$ label the identical point in space. Continuity is the next demand, and it follows from asking that probability not have "cliffs." Suppose $\psi(x)$ jumped from one finite value to another at some point $x_0$. Then $|\psi(x_0)|^2$ would be ambiguous: approaching from the left gives one probability density, approaching from the right gives another, and the point $x_0$ itself has no well-defined density at all. Physically, particles do not have a well-defined probability of being found in a shrinking neighborhood of $x_0$ that suddenly changes character; matter distributions and probability densities are continuous in every physical theory we trust, and quantum mechanics is no exception. The derivative condition is more subtle and is best understood through the machinery of the Schrödinger equation itself. The time-independent Schrödinger equation reads

Solving for the second derivative,

For this equation to hold at every point, $d^2\psi/dx^2$ must exist and be finite wherever $V(x)$ is finite. But a function can only have a finite second derivative if its first derivative is continuous — a kink (a sudden change of slope) in $\psi$ corresponds to a Dirac-delta spike in $d^2\psi/dx^2$, which is not finite and therefore violates the Schrödinger equation unless $V(x)$ itself contains a matching delta-function or an infinite step at that point. This is why $d\psi/dx$ must be continuous everywhere the potential is finite, but is allowed to have a kink at a point where $V(x) \to \infty$ (such as the infinite walls of a particle-in-a-box), because there the equation is not even required to hold in the usual sense inside the forbidden region. Square-integrability is the condition that ties everything back to the requirement of total probability equal to one. If

then no finite normalization constant $N$ can make $\int |N\psi|^2 dx = 1$; the function describes a particle that is, loosely speaking, "infinitely delocalized" and cannot represent a real, localized quantum state. For bound states, this forces $\psi \to 0$ as $x \to \pm\infty$, and it forces us to discard any mathematically valid solution of the Schrödinger equation that blows up at infinity — even though such solutions solve the differential equation perfectly well for every value of $E$.

The four acceptability conditions can be stated compactly. A physically acceptable wavefunction $\psi$ must satisfy:

where: $\mathbf{r}$ ranges over all physically accessible configuration space, and "single-valued" means that any two paths reaching the same physical point must give the same $\psi$. Reading: if $\phi$ and $\phi+2\pi$ describe the same physical angle, $\psi(\phi) = \psi(\phi + 2\pi)$ is mandatory, not optional.

where: $x_0$ is any point in the domain, and the two one-sided limits must both equal the function's actual value there. Reading: $\psi$ has no jumps; probability density $|\psi|^2$ is therefore also continuous.

where: the two one-sided derivatives at $x_0$ must agree whenever the potential does not itself become infinite there. Reading: finite potential jumps (finite step, finite square well) leave $d\psi/dx$ continuous; only an infinite wall permits a kink.

where: $C$ is any finite positive number (later absorbed into a normalization constant $N = 1/\sqrt{C}$). Reading: the total "amount" of $|\psi|^2$ integrated over all space must be finite so that the wavefunction can be normalized to represent exactly one particle with total probability 1. Taken together, these four statements define the space of acceptable (sometimes called "well-behaved" or "admissible") wavefunctions, a strict subset of all mathematical solutions to the Schrödinger equation. The next two derivations show explicitly how this filtering mechanism operates: first in the general logic of matching solutions across a boundary, and second in the concrete case of the particle-in-a-box, where imposing these conditions single-handedly generates the quantized energy ladder $E_n \propto n^2$.

Example 1 — Foundational: testing three candidate functions on $[0,\infty)$. Three trial functions are proposed as possible wavefunctions for a particle confined to $x \ge 0$ with $\psi(x)=0$ enforced for $x<0$:

Test each against the four conditions. Function $\psi_a = N_a e^{-x}$. It is single-valued (one output per input), continuous everywhere on $x\ge0$, and its derivative $-N_a e^{-x}$ is continuous. Check square-integrability:

This is finite for any finite $N_a$, so $\psi_a$ passes all four tests; it is acceptable, and normalization requires $N_a^2/2 = 1$, i.e. $N_a = \sqrt{2}$. Function $\psi_b = N_b x^{-1/2}$. At $x=0$ this function diverges to infinity — it is not even finite, let alone continuous, at the boundary point. Moreover,

diverges logarithmically at both the $x\to 0$ end and the $x \to \infty$ end. $\psi_b$ fails both the finiteness/continuity condition and square-integrability; it cannot describe any physical state. Function $\psi_c = N_c e^{2x}$. This function is single-valued, continuous, and smooth everywhere — but it grows without bound as $x\to\infty$:

The integral diverges, so no finite $N_c$ can normalize $\psi_c$. It fails square-integrability even though it satisfies every other condition — a clean illustration that all four conditions are independent hurdles, and passing three out of four is not enough.

Example 2 — Applied: a particle-in-a-box with asymmetric parameters, checking the derivative condition at an internal step. Consider an electron (the kind of confinement used as a crude model for a conjugated polyene chromophore) in a box of width $L = 0.65\,\text{nm}$ with an added finite interior potential step at $x = L/2$ of height $V_0 = 1.20\times 10^{-19}\,\text{J}$ (about $0.75$ eV), used to model a slight electronic asymmetry along the chain. We are told that in the region $x < L/2$ the wavefunction for some particular quantized state has the form $\psi_1(x) = A\sin(k_1 x)$ with $k_1 = 4.10\times 10^{9}\,\text{m}^{-1}$, and in the region $x>L/2$ it has the form $\psi_2(x) = C\sin(k_2 x + \delta)$ with $k_2 = 3.55\times10^{9}\,\text{m}^{-1}$. The acceptability conditions tell us exactly what must be matched at $x_0 = L/2 = 3.25\times10^{-10}\,\text{m}$. Because $V_0$ is finite (not an infinite wall), Step 2 of the first derivation applies in full: both $\psi$ and $d\psi/dx$ must be continuous at $x_0$. Concretely, this means solving

simultaneously for the two unknown ratios ($C/A$ and $\delta$) — exactly the standard matching procedure used to build finite-well and step-potential solutions in the same way as for a finite square well. Plugging in $k_1 x_0 = (4.10\times10^9)(3.25\times10^{-10}) = 1.3325$ rad and $k_2 x_0'$ appearing in the second piece, one finds a unique $(C/A, \delta)$ pair — the point being that both equations are legitimately imposed here (unlike at an infinite wall, where only the first would survive). If a student proposed a trial solution with a kink in slope at $x_0$ while $V_0$ remains finite, that solution would violate the Schrödinger equation at $x_0$ itself and must be rejected regardless of how well it fits the boundary conditions elsewhere.

Example 3 — Exam/competition level: a deliberately flawed trial wavefunction for the 1-D box, testing all four conditions at once. A student proposes the following un-normalized trial function for a particle in a box on $0 \le x \le L$ (and zero elsewhere), with $L = 1.00\times10^{-9}\,\text{m}$:

Determine whether $\psi_{\text{trial}}$ is an acceptable trial function (in the variational sense) by the four conditions — noting explicitly that it is not required to be an eigenfunction of $\hat H$, only to be an acceptable member of the function space. Single-valuedness: for each $x$ there is exactly one output; passes trivially. Continuity: at $x=0$, $\psi_{\text{trial}}(0) = N\cdot 0 \cdot L^2 = 0$, matching the exterior value of $0$. At $x=L$, $\psi_{\text{trial}}(L) = N\cdot L \cdot 0 = 0$, again matching the exterior value of $0$. The function is a polynomial inside the box, manifestly continuous there. Passes. Continuity of the derivative: differentiate inside the box,

At $x=0^+$: $d\psi/dx = N\cdot L \cdot L = NL^2$. At $x=L^-$: $d\psi/dx = N\cdot 0\cdot(L-3L) = 0$. Outside the box, $\psi \equiv 0$ so $d\psi/dx \equiv 0$ on both sides. At $x=L$ the derivative matches ($0=0$) — consistent with an infinite wall, where a mismatch would in fact have been allowed, and here it happens to vanish anyway. At $x=0$, however, the interior derivative is $NL^2 \ne 0$ while the exterior derivative is $0$: there is a slope mismatch (kink) at $x=0$. Because the wall at $x=0$ is also an infinite potential (particle-in-a-box), this mismatch is permitted by the exception established in the first derivation — the kink at an infinite wall does not disqualify the function. So the trial function actually passes condition 3 in the relevant physical sense (kinks are allowed exactly at infinite walls, which is both boundaries of this box). Square-integrability: compute

Substituting $u = x/L$, $x = uL$, $dx = L\,du$:

The integral $\int_0^1 u^2(1-u)^4 du$ is a Beta function, $B(3,5) = \dfrac{2!\,4!}{7!} = \dfrac{2\cdot 24}{5040} = \dfrac{48}{5040} = \dfrac{1}{105}$. So

which is manifestly finite for any finite $N$ and finite $L$. Passes. Verdict: $\psi_{\text{trial}}$ satisfies all four acceptability conditions and is a legitimate trial function — e.g., usable in a variational-method estimate of the ground-state energy — even though (as one can check) it is not an exact eigenfunction of $\hat H$ for the box, since it does not equal $\sin(\pi x/L)$. Acceptability and being an exact energy eigenfunction are different, non-equivalent requirements; the four conditions define the arena, not the winning move.

The acceptability conditions did not arise from one single discovery; they crystallized as physicists and mathematicians worked out how to make Schrödinger's new wave mechanics consistent with a probabilistic reading of $\psi$. The pivotal figure for connecting rigor to physical interpretation is Max Born, who in 1926 first proposed that $|\psi|^2$ should be read as a probability density — a suggestion so consequential that it retroactively forced every acceptable wavefunction to obey exactly the regularity conditions catalogued here. Born was a German theoretical physicist working in Göttingen, at the heart of the "matrix mechanics" school built with Werner Heisenberg and Pascual Jordan; when Erwin Schrödinger's wave mechanics arrived as an apparently different formulation of quantum theory, Born recognized that Schrödinger's $\psi$ could carry the statistical content that matrix mechanics handled more abstractly. His 1954 Nobel Prize citation credits exactly this: "for his fundamental research in quantum mechanics, especially for his statistical interpretation of the wave function." Born's interpretive leap is what makes single-valuedness, continuity, smoothness, and square-integrability mandatory rather than merely convenient simplifying assumptions. Before Born's probabilistic reading, $\psi$ could have been treated as a purely formal auxiliary quantity, and one might have tolerated wilder mathematical behavior. Once $|\psi|^2 d\tau$ is asserted to be an actual, physically measurable probability, every one of the conditions in this article becomes non-negotiable: they are the mathematical translation of "probability densities must be single, finite, continuous, and must add up to 1."