Caspar: Viral Geometry & Underground Astrophysics

• Caspar is a dual-use term referring to the Caspar–Klug framework for designing icosahedral viral capsids and a nuclear astrophysics accelerator at SURF.
• In virology, the Caspar–Klug construction subdivides icosahedral surfaces into symmetric triangular lattices, enabling detailed studies of virus shell formation.
• In astrophysics, CASPAR utilizes a 1-MV Van de Graaff accelerator underground to measure low-energy nuclear reactions critical to stellar nucleosynthesis.

The name Caspar appears in two distinct technical contexts. In the geometry of viral shells and symmetry-preserving polyhedral constructions, it refers to Donald L. D. Caspar, whose work with Aaron Klug introduced a 1962 construction for icosahedral virus capsids and related cubic polyhedra. In underground nuclear astrophysics, CASPAR denotes the Compact Accelerator System for Performing Astrophysical Research, a 1-MV Van de Graaff accelerator installed at the Sanford Underground Research Facility (SURF) to measure low-energy nuclear reactions relevant to stellar nucleosynthesis (Brinkmann et al., 2017, Heise, 2017).

1. Principal meanings and historical placement

In the virological and geometric literature, Caspar is one of the two names in Caspar and Klug. Their construction became the standard biologically motivated method for generating highly symmetric spherical structures by subdividing the triangular faces of an icosahedron according to a lattice pattern. In the underground-physics literature, by contrast, CASPAR is the acronym for an accelerator experiment at SURF devoted to stellar-energy nuclear reactions (Brinkmann et al., 2017, Heise, 6 Mar 2026).

Usage	Definition	Domain
Caspar	Donald L. D. Caspar in Caspar–Klug theory	Viral capsids, polyhedral geometry
CASPAR	Compact Accelerator System for Performing Astrophysical Research	Underground nuclear astrophysics
CASPEr	Cosmic Axion Spin Precession Experiment	NMR-based axion and ALP dark-matter search

A persistent historical issue concerns attribution. Later literature often refers to the “Goldberg–Coxeter construction,” but one mathematical analysis argues that, in the form usually used for icosahedral and fullerene-type polyhedra, the construction is actually the Caspar–Klug construction, not Goldberg’s original one. The same source emphasizes that Caspar and Klug worked independently of Goldberg, while Coxeter later described the construction more formally and helped disseminate it in a broader mathematical setting (Brinkmann et al., 2017).

The orthographically similar acronym CASPEr is unrelated to either Donald Caspar or the SURF accelerator. It is a nuclear magnetic resonance experiment seeking to detect axion and axion-like particles through nuclear spin precession (Garcon et al., 2017).

2. Caspar–Klug geometry and quasi-equivalence

The classical Caspar–Klug (CK) framework describes the architecture of many icosahedral viruses in terms of a triangulated lattice on the sphere. In that setting, capsids are built from quasiequivalent protein subunits, the surface is organized into pentamers and hexamers, and allowed capsids have sizes given by the formula

$60T,$

where $T$ is the triangulation number (Roussel et al., 2023).

In the lattice formulation, one works in the Euclidean triangular lattice and chooses integers $a \ge b \ge 0$. The basic Caspar–Klug triangle has vertices

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$

Applied to a triangulated spherical polyhedron, typically the icosahedron, this construction subdivides each triangular face into smaller triangles while preserving the orientation-preserving symmetries of the original triangulation. In the special cases

$b=0 \quad \text{or} \quad a=b,$

mirror symmetries are preserved as well (Brinkmann et al., 2017).

A closely related parametrization uses integer pairs $(m,n)$ and the triangular lattice

$$\mathbb{G}=\{\mathbf{r}_{q_1,q_2}=q_1\mathbf{e}_1+q_2\mathbf{e}_2:(q_1,q_2)\in\mathbb{Z}^2\},$$

with

$$\mathbf{e}_1=(1,0),\qquad \mathbf{e}_2=\left(\frac12,\frac{\sqrt3}{2}\right).$$

For regular tetrahedra, octahedra, and icosahedra, the associated point count is

$N=(V-2)(m^2+n^2+mn)+2,\qquad V\in\{4,6,12\}.$

This formulation makes explicit that the CK construction is not inherently limited to icosahedra; rather, it is a triangular-lattice method whose compatibility with spherical triangulations gives it unusually broad reach (Hamilton, 2021).

The biological significance of CK lies in its quasi-equivalence principle: identical protein subunits occupy similar, though not identical, local environments while maintaining a coherent global icosahedral organization. This became the baseline geometric description for virus-derived protein shells and for many later mathematical generalizations (Konevtsova et al., 2015).

3. Mathematical formalization and generalization of the Caspar construction

One line of work reexamines CK within a broader theory of local symmetry-preserving operations. In this treatment, CK is the clearest example of a local operation that decorates each chamber of a triangulated polyhedron in the same way while respecting its automorphism structure. The same analysis sharply distinguishes CK from Goldberg’s original construction, which works in the hexagonal lattice and the dodecahedral dual picture, and from Fuller’s geodesic-dome approach, which uses subdivided icosahedra but does not formulate the method directly in the Euclidean triangular lattice and therefore does not naturally capture chiral structures (Brinkmann et al., 2017).

A later formalization extends this viewpoint to local orientation-preserving symmetry preserving operations, abbreviated lopsp. Instead of decorating a single chamber, lopsp decorates double chambers, which allows the framework to include chiral constructions that preserve only orientation-preserving symmetries. The defining data are a connected Euclidean tiling $T$ together with rotation centers $T$0 and $T$1 of $T$2 and $T$3, respectively. The theory introduces double chamber decorations, proves a path-independence theorem for the resulting operations, and shows that sufficiently connected lopsp operations define operations on polyhedra (Goetschalckx et al., 2020).

A separate generalization replaces the rigid icosahedral setting by spherical area coordinates (SACs). In that construction, planar barycentric coordinates on a triangular patch are reinterpreted as spherical area coordinates on arbitrary spherical triangles, so that CK-type subdivisions can be applied to any closed triangular mesh. For a sequence of integer pairs

$T$4

the recursive cardinality formula is

$T$5

The quality metric is the mesh ratio

$T$6

with separation distance

$T$7

and covering radius

$T$8

For well-chosen parameters, this method generates point sets with mesh ratios lower than previously reported for $T$9 (Hamilton, 2021).

4. Exceptions, alternatives, and non-Caspar–Klug architectures

Although CK is a powerful blueprint, several important capsid classes are explicit exceptions. Papillomavirus is a canonical example: its capsid contains 72 pentamers and 360 proteins total, which cannot be represented in the standard CK $a \ge b \ge 0$0 scheme and is therefore non-quasiequivalent. In a weighted-graph description of its interaction network, pentamers are vertices and inter-pentamer contacts are edges with three bond types, $a \ge b \ge 0$1, $a \ge b \ge 0$2, and $a \ge b \ge 0$3, satisfying

$a \ge b \ge 0$4

The total capsid energy is written

$a \ge b \ge 0$5

Using an energy-based percolation model, the capsid is found to prefer hole formation before fragmentation, whereas AaLS cages tend to fragment before large holes form, and HK97 behaves more like the fragmentation-prone category. This makes geometry relevant not only for assembly but also for preferred disassembly pathways (Roussel et al., 2023).

Another response to CK’s limits is a quasicrystalline tiling model of spherical viral capsids. Here, the shell is treated as a spherical tiling with proteins located at tile vertices, all tiles having identical edges within each tile type, and the number of tile types kept minimal. The classical CK counting rule is

$a \ge b \ge 0$6

The model was motivated in part by the papovavirus discrepancy: a $a \ge b \ge 0$7 CK shell would require 420 proteins, while experimentally one observes 360, arranged as 72 pentamers. The quasicrystalline construction was presented as a uniform description of both structures that satisfy the CK model and structures that contradict it (Konevtsova et al., 2015).

A further extension shifts attention from subunit placement to the interaction network between capsomers. In this approach, capsomer centres of mass are treated as graph nodes, edges denote capsomer interactions, and the resulting graph is embedded into planar tilings constrained by symmetry. For the AaLS pentamer, this yields exactly six 3D symmetric cage architectures compatible with the observed local interaction rules. Four were used to rationalize known structures—12 pentamers, icosahedral symmetry; 24 pentamers, tetrahedral symmetry; 36 pentamers, tetrahedral symmetry; and 72 pentamers, icosahedral symmetry—and two were given as predictions: 48 pentamers, tetrahedral symmetry and 60 pentamers, tetrahedral symmetry. The same analysis concluded that no AaLS cages with octahedral symmetry are compatible with the observed interaction network (Fatehi et al., 2023).

5. Caspar-derived principles beyond classical capsids

CK logic has also been transplanted to materials systems far from viral capsids. One example is an “inverted CK” framework for programmable assembly of size-controlled triply-periodic polyhedra, discrete variants of the Primitive, Diamond, and Gyroid cubic minimal surfaces. In this setting, the basic patch is a hexagon rather than a triangle, but the construction is again a symmetry-preserving subtriangulation. The resulting structures are denoted $a \ge b \ge 0$8, $a \ge b \ge 0$9, and $v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$0. For the gyroid repeat unit, the unit cell contains $v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$1 triangular particles. Size is characterized by the maximal medial thickness $v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$2, which scales as

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$3

while the number of inequivalent species scales as

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$4

The economy metric

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$5

therefore obeys

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$6

Dynamical assembly simulations show that high-fidelity assembly requires an intermediate degree of flexibility, approximately

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$7

and that off-target states arise through generalized disclinations in hyperbolic crystals (Duque et al., 2023).

CK principles have also been generalized in the Thomson problem. There, the starting point is the CK scaffold with triangulation number

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$8

and particle count

$v_1=(0,0), \qquad v_2=(a,b), \qquad v_3=(-b,a+b).$9

By removing the 12 vertex particles, one obtains trial structures with

$b=0 \quad \text{or} \quad a=b,$0

and by introducing a simplest distortion of the net through a lattice shift $b=0 \quad \text{or} \quad a=b,$1, the change in particle number becomes

$b=0 \quad \text{or} \quad a=b,$2

with

$b=0 \quad \text{or} \quad a=b,$3

In the interval

$b=0 \quad \text{or} \quad a=b,$4

this procedure produced 40 new spherical crystals with the lowest energies seen so far. Their defects were not the usual elongated scars but identical flattened pentagons, and many of the resulting $b=0 \quad \text{or} \quad a=b,$5 values are prohibited in the CK model (Roshal et al., 2014).

Taken together, these developments suggest that CK is best understood as a transferable symmetry principle rather than a closed catalogue of admissible structures. In later work it functions as a scaffold for recursive spherical point sets, programmable negative-curvature materials, and symmetry-broken low-energy spherical crystals (Hamilton, 2021, Duque et al., 2023).

6. CASPAR at the Sanford Underground Research Facility

In uppercase usage, CASPAR stands for the Compact Accelerator System for Performing Astrophysical Research. It is SURF’s dedicated nuclear astrophysics experiment and, according to one review, one of only three deep underground laboratories for nuclear physics in the world. Its core instrument is a 1-MV Van de Graaff accelerator that can deliver high-intensity $b=0 \quad \text{or} \quad a=b,$6A proton and alpha beams in the energy range 150 keV to 1.0 MeV. These beam energies overlap the low-energy regime relevant to hydrogen burning, stellar helium burning, and neutron-production reactions important for interpreting nucleosynthesis channels and the origin of many heavy elements (Heise, 6 Mar 2026).

CASPAR is located at SURF’s 4850-foot level Ross Campus. SURF’s main underground science infrastructure is concentrated on the 4850-foot level, corresponding to about 4300 meters water equivalent (m.w.e.) of shielding, and the specific CASPAR site is listed with an overburden of 4170 m.w.e. At the Ross Campus, a portion of a tunnel and a former maintenance shop were combined to create laboratory space for the experiment. The underground environment is critical because it suppresses cosmic-ray muons and secondary backgrounds such as neutrons, which would otherwise obscure very weak low-energy reaction signals. CASPAR shares the Ross Campus with the BHUC low-background laboratory, and the broader 4850L infrastructure includes the Ross and Yates shafts, power and network redundancy, ventilation, and environmental control (Heise, 2017, Heise, 6 Mar 2026).

The accelerator components were relocated from the University of Notre Dame in Summer 2015. At the time of the 2017 SURF overview, the beamline had been assembled, first beam had been achieved in May 2017, and an initial operations announcement had been made in July 2017, with physics data expected within months. A later review states that, since February 2018, CASPAR has carried out data campaigns using targets including

$b=0 \quad \text{or} \quad a=b,$7

Among the highlighted reactions is

$b=0 \quad \text{or} \quad a=b,$8

relevant to the CNO cycle and to interpretation of the CNO neutrino flux as an independent probe of solar-core metallicity. Operations were temporarily halted in March 2021 due to nearby LBNF construction, resumed for a second phase in Summer 2025, and Phase 2 is expected to last roughly three years until $b=0 \quad \text{or} \quad a=b,$9. A possible third phase has been envisioned for the early- to mid-2030s, potentially with a 200–300 kV high-voltage platform (Heise, 2017, Heise, 6 Mar 2026).

Within SURF’s broader portfolio, CASPAR complements dark-matter, neutrinoless double-beta decay, and neutrino programs by occupying the nuclear astrophysics niche. Its scientific role is to exploit deep underground shielding and accelerator capability to measure reactions at stellar energies, including reactions relevant to the slow neutron-capture nucleosynthesis process, or s-process, in an environment with strongly suppressed backgrounds (Heise, 2017).