Test Function Spaces

A quantum test space must support the operations used in a calculation and carry a topology that makes their action continuous. Density in the state Hilbert space is necessary for many applications, but it supplies neither invariance under a Hamiltonian nor continuity of a generalized eigenvector. The Schwartz space is particularly effective for canonical position and momentum because its topology controls both derivatives and spatial growth.

Required background. Gelfand Triples constructs the Hilbert-to-antidual embedding. Distribution Theory Minimum defines Schwartz seminorms and distinguishes them from the Hilbert norm.

Continuity must control every required operation

For a locally convex test space $\Phi$, continuity of a linear map $A:\Phi\to\Phi$ means that every output seminorm is bounded by a finite collection of input seminorms. This is stronger than asserting that $A\phi$ happens to exist for every test vector.

On $\mathcal S(\mathbb R)$, let $p_{a,b}(\phi)=\sup_x|x^a\phi^{(b)}(x)|$. Differentiation shifts $b$ by one. Multiplication by $x$ obeys

$$p_{a,b}(x\phi) \le p_{a+1,b}(\phi)+b\,p_{a,b-1}(\phi),$$

with the second term omitted for $b=0$. Thus both position and momentum act continuously, and every polynomial in them acts continuously on this space. These estimates explain why repeated products can be tested there without enlarging the domain at each step.

Fourier transformation is also a continuous automorphism of $\mathcal S$: derivatives and coordinate multiplications interchange, and the required integral bounds follow from rapid decay. This supports a single rigging for both position and momentum representations.

Oscillator coefficients describe the same topology

Use dimensionless position and momentum and let $(h_n)_{n\ge0}$ be normalized Hermite functions. The oscillator number operator satisfies $Nh_n=nh_n$. The standard Hermite characterization of the Schwartz space is

$$\mathcal S(\mathbb R) =\bigcap_{k\ge0}D((N+I)^k).$$

For $\phi=\sum_nc_nh_n$, the Hilbert-scale norms are

$$q_k(\phi)^2 =\sum_{n=0}^{\infty}(n+1)^{2k}|c_n|^2.$$

Finiteness for every $k$ means faster-than-any-power coefficient decay in the weighted square-summability sense. These norms generate the Schwartz topology. The equivalence with the coordinate seminorms is a substantive Hermite expansion result: oscillator powers control combinations of derivatives and multiplication, and conversely Schwartz decay controls every oscillator power. It should not be inferred merely from completeness of the Hermite basis in $L^2$.

This representation turns a topological requirement into a useful spectral check. A vector with coefficients $c_n\propto(n+1)^{-2}$ is normalizable and belongs to some oscillator domains, but fails sufficiently high $q_k$ tests. It is a legitimate Hilbert-space state without being a Schwartz test vector.

The Hilbert scale explains nuclearity

Let $\mathcal H_k$ be the completion in $q_k$. The inclusion $\mathcal H_{k+1}\to\mathcal H_k$ has singular values $(n+1)^{-1}$ and is Hilbert–Schmidt, since their squares are summable. The inclusion two steps down has singular values $(n+1)^{-2}$ and is trace class.

The countably Hilbert-space criterion then establishes nuclearity of the projective-limit test space. This is a structural reason that Schwartz riggings fit nuclear spectral and kernel theorems. It does not mean that the inclusion of an arbitrary dense domain into $L^2$ is nuclear, nor that every continuous functional is an ordinary vector.

Nuclearity is an imported functional-analytic criterion here; the displayed singular-value calculation verifies its concrete hypothesis. Generalized Eigenfunction Expansions explains the additional operator assumptions needed for spectral applications.

Different models may require different test spaces

The space $C_c^\infty(\mathbb R)$ is useful for local differential equations. Its usual test-function topology is an inductive-limit topology, not simply the restriction of one Hilbert norm. Fourier transformation does not preserve compact support, so it is less convenient when both representations must share the same test space.

Schwartz functions are preserved by multiplication with a smooth potential whose derivatives all have at most polynomial growth. Smoothness alone is insufficient: uncontrolled growth can destroy rapid decay. A discontinuous bounded potential can also fail to preserve Schwartz smoothness, despite being a perfectly acceptable bounded perturbation of a self-adjoint Hamiltonian.

Boundary conditions matter as well. Smooth functions on an interval are not automatically in the domain of a chosen self-adjoint realization, and repeated operator action may require additional endpoint compatibility. Check the actual model before choosing its rigging. An invariant space for $X$ and $P$ is not automatically invariant for every Hamiltonian built from informal expressions involving them.

Exercise: a bounded potential that breaks a rigging

Let $V(x)=\mathbf1_{(0,\infty)}(x)$ and take a Schwartz function $\phi$ with $\phi(0)\ne0$. Does $V$ define a bounded self-adjoint operator on $L^2$? Does it map this $\phi$ into $\mathcal S$? What must be reconsidered in a rigged-space calculation involving $H=P^2/(2m)+V$?

Solution

Real bounded multiplication is self-adjoint on all of $L^2$, with norm one. But $V\phi$ has a jump at zero, so it is not a Schwartz function. The Hamiltonian is still self-adjoint on the free operator domain by bounded perturbation theory; the failure concerns preservation of the proposed test space. One must choose an appropriate invariant rigging or use a method that does not require this invariance, rather than conclude that the physical Hamiltonian is ill defined.

References

• Gel’fand, I. M., and N. Ya. Vilenkin, Generalized Functions, Vol. 4: Applications of Harmonic Analysis, translated by Amiel Feinstein, Academic Press, 1964, publisher edition — countably Hilbert spaces, nuclearity, and rigged spectral analysis.
• Hall, Brian C., Quantum Theory for Mathematicians, Graduate Texts in Mathematics 267, Springer, 2013, doi:10.1007/978-1-4614-7116-5 — Schwartz functions, oscillator eigenfunctions, and unbounded operators.
• Teschl, Gerald, Mathematical Methods in Quantum Mechanics: With Applications to Schrödinger Operators, 2nd ed., Graduate Studies in Mathematics 157, American Mathematical Society, 2014, author’s edition — self-adjoint Schrödinger operators and bounded perturbations.