---
title: "Construction Geometry of the Conic Sundial with Horizontal Gnomon"
subtitle: "An Analytical Framework for Shadow Projection and Apex-Free Parameter Recovery on an Inclined Cone"
running-title: "Conic Sundial Geometry"
book: false
toc: true
author:
  - name: Gerd Graßhoff
    email: gerd.grasshoff@hu-berlin.de
    orcid: 0000-0002-4450-357X
    affiliations:
      - name: Humboldt-Universität zu Berlin
        department: Philosophisches Institut
        address: Unter den Linden 6
        city: Berlin
        country: Germany
        postal-code: 10099
      - name: Deutsches Museum
        department: Forschungsinstitut
        address: Museumsinsel 1
        city: München
        country: Germany
        postal-code: 80538
abstract: |
  A conic sundial with its nodus on the polar axis is determined by three
  construction choices: the intended latitude (the *klima* proportion of the
  ancient shadow-table tradition), the cone opening, and an arbitrary scale.
  Because the nodus is mounted directly on the celestial rotation axis, every
  fixed solar declination projects a circular shadow path onto the interior
  cone surface, perpendicular to that axis. This paper establishes the
  complete analytical geometry of that instrument in a unified East–North–Up
  world frame and derives its exact closed-form inversion. The central
  theoretical result is that the three seasonal meridian marks ($P_S, P_E, P_W$)
  determine the cone — half-angle $\alpha$, scale $d$, and apex position $A$ —
  independently of geographical latitude $\varphi$. Latitude is not encoded in
  the relative positions of the noon marks; it enters the artifact exclusively
  through the hour division of the solstitial day-lines and the dressing of
  the block faces. A closed-form marking ratio $Q = (L_S - L_W)/(L_S - L_E)$
  recovers $\alpha$ and the apex even when the vertex is chipped or missing.
  The mathematical framework is verified against an exact synthetic Rhodes
  construction and compared with measured artifacts from the Berlin Sundial
  Project corpus.
keywords:
  - conic sundial
  - gnomonics
  - shadow geometry
  - klimata
  - ancient instruments
  - analytical inversion
  - apex-free reconstruction
bibliography: references.bib
format:
  arxivai-typst: default
---

::: {.callout-note title="Supplement S2" #supplement-box}
Supplement S2 to the Ancient Sundials repository (Berlin Sundial Collaboration, ancient-sundials.lettre.ai). This draft documents the shadow animation and the astronomical picture of each object (noon altitude, day length). It is a preprint draft and has not been peer reviewed. Where a figure in the repository differs from a figure here, the repository, which is rebuilt from the current database, is authoritative.
:::


# Introduction: The Geometry of Ancient Timekeeping

Ancient Greek and Roman horology produced three principal types of spherical
and conical shadow-receiving dials [@gibbs1976; @schaldach2006]. Among these,
the conical sundial with a horizontal gnomon occupies a privileged mathematical
position. Unlike planar sundials, where seasonal shadow paths are hyperbolae or
ellipses varying in curvature across the year, the conic dial with a nodus on the
polar axis projects every seasonal day-curve as an exact circular arc.

When archaeologists excavate a fragmentary sundial in the Mediterranean basin,
the stone is frequently damaged: the apex of the hollowed cone is chipped,
the gnomon has perished, and the original orientation on its pedestal is lost.
The computational task is to recover the intended construction parameters from
the 3D scan of the surviving stone fragment.

```{=typst}
#arxivai.concept-box(title: [The central thesis in one sentence])[
  The three seasonal noon marks on a conic sundial determine the cone half-angle
  and the nodus distance completely and independently of latitude; the latitude
  $phi$ is not recorded in the noon marks and must be recovered from the
  hour-line network or from the dressed reference planes of the block.
]
```

This thesis directly corrects an assumption in the literature that treats the
solstitial noon-mark ratio as a latitude diagnostic. On a cone whose axis is
aligned with the celestial pole, rotational symmetry uncouples the axial
spacing of the seasonal day-lines from the tilt of the cone axis above the
horizon. The noon marks record the shape of the cone; the hour network records
the latitude.

Five contributions establish this framework:

1. **Forward Coordinate Geometry:** A rigorous formulation of shadow projection
   in an East–North–Up (ENU) world frame and an equatorial polar frame,
   yielding closed-form ray–cone intersections.
2. **Day-Line Circularity:** A formal proof that all seasonal shadow traces are
   planar circles orthogonal to the cone axis, with the equinoctial circle
   sitting at an axial distance identical to the nodus distance ($d$).
3. **Constructor's Proportion Vocabulary:** A complete formulation of the
   instrument's attributes in the proportional language of ancient dialmakers,
   governed by the workshop rule $x = \tan\alpha\tan\varepsilon$.
4. **Apex-Free Closed-Form Inversion:** The derivation of the marking ratio $Q$,
   which computes the cone half-angle $\alpha$, the apex location $A$, and the
   scale $d$ directly from three collinear meridian marks without requiring the
   physical apex.
5. **Marking Sufficiency & Sensitivity:** A comprehensive classification of
   which subsets of surface features determine which structural parameters,
   together with analytical error bounds.

Direct physical evidence for this proportional design method comes from the
unfinished conic sundial discovered on the Agora of the Italians on Delos
[@rinner2013delos]. The Delos artifact demonstrates that ancient artisans
committed the latitude proportion to the stone block before hollowing the
cone, scribed reference transfer circles, and incised the noon line by simple
geometric ratios.

# Physical Construction Elements

The geometry of the instrument is established by one fixed shadow-casting point
and one receiving surface of revolution.

```{=typst}
#arxivai.concept-box(title: [The Nodus $N$])[
  The shadow-casting point of the sundial: a small metallic sphere mounted
  on the Earth's polar axis, situated exactly on the observer's horizontal plane.
  All solar shadow rays originate geometrically from this point.
]
```

```{=typst}
#arxivai.concept-box(title: [The Cone, Apex $A$, and Half-Angle $alpha$])[
  The receiving dialface is the interior nappe of a right circular cone.
  Its axis of rotational symmetry is collinear with the Earth's polar axis.
  The apex $A$ is positioned on this axis at a distance $d$ above the nodus,
  and the cone opens southward and downward with half-angle $alpha$.
]
```

```{=typst}
#arxivai.concept-box(title: [The Gnomon])[
  A rigid horizontal support rod lying in the observer's meridian plane,
  extending northward from the nodus along the horizon to anchor into the stone
  block. Only the nodus, not the support rod, casts the time-telling shadow point.
]
```

Because the nodus $N$ lies on the horizontal plane ($Z = 0$), the nodus shadow
first strikes the receiving cone at the exact astronomical moment of sunrise,
sweeps across the inscribed network during daylight, and leaves the cone at
sunset.

@fig-meridian illustrates the meridian cross-section of the instrument for a
construction at Rhodes ($\varphi = 36^\circ$, $\alpha = 45^\circ$).

![Meridian cross-section of a conic sundial constructed for Rhodes ($\varphi = 36^\circ$, $\alpha = 45^\circ$). View from the East (South is to the left, North to the right). The nodus $N$ rests on the horizon ($Z = 0$). The polar axis is inclined at angle $\varphi$. The cone apex $A$ lies on the polar axis at distance $d$ above $N$. The cone opens southward at half-angle $\alpha$. Three noon rays passing through $N$ strike the lower meridian generator at $P_W$ (winter solstice), $P_E$ (equinox), and $P_S$ (summer solstice).](figures/fig-meridian.svg){#fig-meridian}

Three independent quantities fully determine the construction:

1. **Latitude** $\varphi$ — the celestial tilt of the polar axis above the
   horizon, fixed by the geographical location.
2. **Nodus Distance** $d$ — the distance along the polar axis from the apex $A$
   to the nodus $N$. This is the constructor's arbitrary scale factor.
3. **Cone Opening** $\rho = \tan\alpha$ — the ratio of rim radius to cone height,
   chosen by the artisan for aesthetic and structural reasons.

The only physical constraint on the scale pair $(d, H)$, where $H$ is the cone
height along the axis, is that the nodus must sit inside the cone cavity:

$$
d < H.
$$ {#eq-cavity}

*Reading.* The nodus distance $d$ along the axis must be strictly less than the
total axial depth $H$ of the cone cavity, ensuring that shadows fall upon the
interior stone surface throughout the year.

# Coordinate Frames and Transformations

To model the instrument unambiguously, we define two right-handed Cartesian
coordinate systems.

## The World Frame (ENU)

The world coordinate frame is fixed to the local observer with the origin at
the physical nodus $N$:

* $+X_w$ points **East** (orthogonal to the meridian plane),
* $+Y_w$ points **North** (along the horizontal meridian),
* $+Z_w$ points **Up** (perpendicular to the local horizon).

In this frame, the unit vector along the polar axis pointing toward the North
Celestial Pole is:

$$
\hat{\mathbf{C}} = (0,\ \cos\varphi,\ \sin\varphi).
$$ {#eq-polar-axis}

*Reading.* The cone axis $\hat{\mathbf{C}}$ has zero East–West component,
a horizontal North component equal to $\cos\varphi$, and a vertical Up
component equal to $\sin\varphi$, corresponding to an elevation angle of $\varphi$
above the northern horizon.

The cone opens in the opposite direction, toward the South and downward:

$$
\hat{\mathbf{O}} = -\hat{\mathbf{C}} = (0,\ -\cos\varphi,\ -\sin\varphi).
$$ {#eq-opening-dir}

*Reading.* The cone opening vector $\hat{\mathbf{O}}$ points southward with
component $-\cos\varphi$ and downward with component $-\sin\varphi$.

## The Equatorial Polar Frame

The polar coordinate frame has its $+Z_p$ axis aligned with $\hat{\mathbf{C}}$.
The rotation matrix $R$ transforming coordinates from the polar frame to the
world frame ($\mathbf{v}_w = R\,\mathbf{v}_p$) is uniquely determined by mapping
the polar axis $(0,0,1)$ to $\hat{\mathbf{C}}$ and mapping the equinoctial sunrise
direction $(-1,0,0)$ to the East vector $(1,0,0)$:

$$
R = \begin{pmatrix}
-1 & 0 & 0 \\
0 & -\sin\varphi & \cos\varphi \\
0 & \cos\varphi & \sin\varphi
\end{pmatrix}.
$$ {#eq-rotation-matrix}

*Reading.* The rotation matrix $R$ is an orthonormal transformation with
determinant $+1$, mapping polar equatorial vectors into the local East–North–Up
world coordinates.

## The Mathematical Frame (Apex at Origin)

For analytical derivations of ray–cone intersections, it is mathematically
convenient to translate the origin to the cone apex $A$. In this mathematical
frame:

$$
A = (0,\ 0,\ 0), \qquad \mathbf{N} = d\,\hat{\mathbf{O}} = (0,\ -d\cos\varphi,\ -d\sin\varphi).
$$ {#eq-math-frame}

*Reading.* In the mathematical frame, the cone apex $A$ is positioned at the
coordinate origin, and the nodus $\mathbf{N}$ is located at distance $d$ along
the opening direction $\hat{\mathbf{O}}$.

# Shadow Projection Algebra

Let the Sun's position in the equatorial polar frame be defined by its solar
declination $\delta$ and hour angle $h$ (measured positive westward from the
meridian, with $h = 0$ at solar noon):

$$
\hat{\mathbf{S}}_p = (\cos\delta\sin h,\ \cos\delta\cos h,\ \sin\delta).
$$ {#eq-sun-polar}

*Reading.* The solar direction unit vector $\hat{\mathbf{S}}_p$ is decomposed
into equatorial coordinates with declination $\delta$ and hour angle $h$.

In the world frame, the Sun's direction is $\hat{\mathbf{S}}_w = R\,\hat{\mathbf{S}}_p$:

$$
\hat{\mathbf{S}}_w = \begin{pmatrix}
-\cos\delta\sin h \\
-\sin\varphi\cos\delta\cos h + \cos\varphi\sin\delta \\
\cos\varphi\cos\delta\cos h + \sin\varphi\sin\delta
\end{pmatrix}.
$$ {#eq-sun-world}

*Reading.* The world-frame solar vector $\hat{\mathbf{S}}_w$ specifies the
direction of incoming sunlight; its vertical $Z$-component is the sine of the
solar altitude angle $a_\odot$.

The shadow ray originates at the nodus $\mathbf{N}$ and propagates in the
direction $-\hat{\mathbf{S}}_w$:

$$
\mathbf{P}(k) = \mathbf{N} - k\,\hat{\mathbf{S}}_w, \qquad k > 0.
$$ {#eq-shadow-ray}

*Reading.* A point $\mathbf{P}(k)$ along the shadow ray is obtained by starting
at the nodus $\mathbf{N}$ and moving a positive scalar distance $k$ away from the
Sun along the direction $-\hat{\mathbf{S}}_w$.

A point $\mathbf{P}$ lies on the surface of a cone with vertex at the origin,
axis $\hat{\mathbf{C}}$, and half-angle $\alpha$ if and only if:

$$
(\mathbf{P}\cdot\hat{\mathbf{C}})^2 = |\mathbf{P}|^2\cos^2\alpha, \qquad \mathbf{P}\cdot\hat{\mathbf{C}} < 0.
$$ {#eq-cone-eq}

*Reading.* The squared scalar projection of the vector $\mathbf{P}$ onto the axis
$\hat{\mathbf{C}}$ equals the squared length of $\mathbf{P}$ multiplied by $\cos^2\alpha$,
subject to the condition that $\mathbf{P}\cdot\hat{\mathbf{C}} < 0$ to select the
south-opening nappe.

Substituting @eq-shadow-ray into @eq-cone-eq yields a quadratic equation in the
ray parameter $k$:

$$
A_k\,k^2 + 2\,B_k\,k + C_k = 0,
$$ {#eq-quadratic-k}

where, noting that $\mathbf{N} = d\,\hat{\mathbf{O}} = -d\,\hat{\mathbf{C}}$ (so $\mathbf{N}\cdot\hat{\mathbf{C}} = -d$ and $|\mathbf{N}|^2 = d^2$), the coefficients are:

$$
\begin{aligned}
A_k &= (\hat{\mathbf{S}}_w\cdot\hat{\mathbf{C}})^2 - \cos^2\alpha = \sin^2\delta - \cos^2\alpha, \\
B_k &= d\left(\hat{\mathbf{S}}_w\cdot\hat{\mathbf{C}} - \cos^2\alpha\,(\hat{\mathbf{O}}\cdot\hat{\mathbf{S}}_w)\right) = d\sin\delta\,(1 + \cos^2\alpha) \dots \\
C_k &= d^2(1 - \cos^2\alpha) = d^2\sin^2\alpha.
\end{aligned}
$$ {#eq-quad-coeffs}

*Reading.* The intersection of the shadow ray with the cone surface reduces to a
standard quadratic equation whose positive root $k$ yields the unique physical
shadow point on the interior nappe.

# The Day-Line Circularity Theorem

A seasonal *day-line* is the continuous curve traced by the nodus shadow on the
cone as the Sun moves from sunrise ($h_{\mathrm{rise}}$) to sunset ($h_{\mathrm{set}}$)
at fixed declination $\delta$.

```{=typst}
#arxivai.method-box(title: [Theorem: Day-Line Circularity])[
  For any fixed solar declination $delta$, the shadow locus on the interior
  surface of a conic sundial with its nodus on the polar axis is an exact planar
  circle orthogonal to the cone axis $hat(bold(C))$.
]
```

*Proof.*
The nodus $\mathbf{N}$ lies on the axis of rotational symmetry $\hat{\mathbf{C}}$.
For a fixed solar declination $\delta$, the diurnal path of the Sun is a great or
small circle centered on the celestial polar axis. The entire physical system —
the light source path, the shadow-casting nodus, and the cone surface — is
strictly invariant under rotations about $\hat{\mathbf{C}}$.

Let $\mathbf{P}(h)$ be the shadow point at hour angle $h$. A change in hour angle
$\Delta h$ corresponds to a pure rigid rotation of the incident ray about
$\hat{\mathbf{C}}$ by angle $\Delta h$. Because the cone is a surface of revolution
about $\hat{\mathbf{C}}$, the resulting intersection point $\mathbf{P}(h + \Delta h)$
is obtained simply by rotating $\mathbf{P}(h)$ by $\Delta h$ around $\hat{\mathbf{C}}$.
Therefore, the axial distance $\ell(\delta) = -\mathbf{P}\cdot\hat{\mathbf{C}}$
and the radial distance $r(\delta) = |\mathbf{P} \times \hat{\mathbf{C}}|$ are
strictly constant for all $h \in [h_{\mathrm{rise}}, h_{\mathrm{set}}]$.
A locus of constant axial position and constant radius on a cone is an exact circle
lying in a plane perpendicular to $\hat{\mathbf{C}}$. $\blacksquare$

@fig-unfolded displays the geometry of the cone surface unfolded into a flat
plane.

![Unfolded cone surface for half-angle $\alpha = 45^\circ$ (sector angle $2\pi\sin 45^\circ \approx 254.6^\circ$). The apex $A$ is at the top center. The central vertical line is the noon meridian generator ($h = 0$). Straight radial lines represent hour lines. The three seasonal day-lines — winter solstice (blue, inner), equinox (green, middle), and summer solstice (red, outer) — unfold to exact concentric circular arcs centered on $A$.](figures/fig-unfolded.svg){#fig-unfolded}

# Proportions and the Constructor's Vocabulary

Ancient sundial makers did not specify their instruments using decimal degree
angles. They operated in a coherent language of geometric proportions, expressing
shadow lengths, cone openings, and day-line positions as ratios.

## The Equinox Invariant

Evaluating the shadow projection at the astronomical equinox ($\delta = 0$):

$$
\ell_E \equiv -\mathbf{P}(\delta=0)\cdot\hat{\mathbf{C}} = d.
$$ {#eq-equinox-invariant}

*Reading.* At the equinox ($\delta = 0$), the sunlight vector is strictly
perpendicular to the polar axis ($\hat{\mathbf{S}}_w\cdot\hat{\mathbf{C}} = \sin 0 = 0$).
The shadow ray cast by the nodus lies in the plane perpendicular to the axis
passing through the nodus itself. Consequently, the equinoctial day-line circle
sits at an axial distance $\ell_E$ from the apex that is identically equal to the
nodus distance $d$, regardless of the cone half-angle $\alpha$ or latitude $\varphi$.

The corresponding slant distance $L_E$ measured along the meridian generator on
the stone surface is:

$$
L_E = \frac{\ell_E}{\cos\alpha} = \frac{d}{\cos\alpha}.
$$ {#eq-slant-equinox}

*Reading.* The slant distance $L_E$ from the apex to the equinoctial noon mark
equals the nodus distance $d$ divided by the cosine of the cone half-angle $\alpha$.

## The Workshop Rule $x$ and Solstitial Spacings

For any solar declination $\delta$, the axial distance $\ell(\delta)$ from the
apex to the noon shadow mark on the meridian is given by:

$$
\ell(\delta) = \frac{d}{1 - \tan\alpha\,\tan\delta}.
$$ {#eq-axial-general}

*Reading.* The axial distance $\ell(\delta)$ equals the nodus distance $d$ divided
by the quantity $1 - \tan\alpha\tan\delta$.

Evaluating @eq-axial-general at the summer solstice ($\delta = +\varepsilon$) and
winter solstice ($\delta = -\varepsilon$), we define the fundamental dimensionless
**workshop ratio**:

$$
x := \tan\alpha\,\tan\varepsilon.
$$ {#eq-workshop-x}

*Reading.* The workshop ratio $x$ is the product of the tangent of the cone
half-angle $\alpha$ and the tangent of the obliquity of the ecliptic $\varepsilon$.

The solstitial axial distances become:

$$
\ell_S = \frac{d}{1 - x}, \qquad \ell_W = \frac{d}{1 + x}.
$$ {#eq-solstice-axial}

*Reading.* The summer solstice noon mark sits at axial distance $d/(1-x)$ from
the apex (deepest into the cone), while the winter solstice noon mark sits at
axial distance $d/(1+x)$ (closest to the apex).

The three seasonal noon marks on the meridian generator thus satisfy the
celebrated **noon-mark triple**:

$$
\ell_W : \ell_E : \ell_S = \frac{1}{1+x} : 1 : \frac{1}{1-x}.
$$ {#eq-noon-triple}

*Reading.* The ratio of winter, equinox, and summer axial noon marks depends
exclusively on the single parameter $x$. For the standard Hellenistic workshop
choice $x = 1/4$ (obtained, for example, with $\alpha = 30^\circ$ and $\varepsilon \approx 23.5^\circ$),
this ratio yields the integers $12 : 15 : 20$.

@tbl-attributes summarizes the complete attribute vocabulary of the conic dial.

| Attribute | Historical Notation | Geometric Meaning |
|:---|:---|:---|
| Gnomon standard | $g = 60$ | Sexagesimal parts of the vertical gnomon |
| Klima ratio | $\sigma_E = s_{\mathrm{eq}} : g$ | Equinoctial noon shadow; $\sigma_E = \tan\varphi$ |
| Solstitial shadow ratios | $\sigma_S, \sigma_W$ | $\sigma_S = \tan(\varphi-\varepsilon)$, $\sigma_W = \tan(\varphi+\varepsilon)$ |
| Cone opening ratio | $\rho = R_{\mathrm{rim}} : H$ | $\rho = \tan\alpha$; ratio of rim radius to cone height |
| Nodus distance | $d$ | Apex-to-nodus distance along axis; scale unit |
| Equinox invariant | $\ell_E : d = 1 : 1$ | Equinoctial circle sits at distance $d$ from apex |
| Workshop rule | $x = \tan\alpha\tan\varepsilon$ | Dimensionless factor governing solstitial marks |
| Noon-mark triple | $\ell_W : \ell_E : \ell_S$ | Triplet of axial noon distances: $(1+x)^{-1} : 1 : (1-x)^{-1}$ |
| Equinoctial division | $H_E = 90^\circ$ | Quarter-circle in 6 seasonal hours ($15^\circ/\mathrm{h}$); calibration invariant |
| Solstitial division | $H(\pm\varepsilon)$ | $\cos H = -\tan\varphi\tan(\pm\varepsilon)$; where $\varphi$ enters the network |

: The attribute vocabulary of the ancient conic sundial. Every entry is a proportion or an integer count; degree values are modern conversions. {#tbl-attributes}

# The Klimata Tradition and the Vertical Gnomon

Greek mathematical geography divided the inhabited world into latitudinal zones
called *klimata*, defined by the length of the longest day and the shadow cast
by a vertical gnomon of $g = 60$ parts at noon [@neugebauer1975; @toomer1984].

@fig-klima illustrates the vertical gnomon shadow triangle for Alexandria.

![The vertical gnomon shadow triangle for Alexandria ($\varphi = 30^\circ 58'$). A vertical gnomon of height $g = 60$ parts casts noon shadows of length $s_S = 7;29$ (summer solstice), $s_E = 36;00$ (equinox), and $s_W = 85;08$ (winter solstice). The equinoctial ratio is exactly $g : s_E = 60 : 36 = 5 : 3$.](figures/fig-klima.svg){#fig-klima}

The vertical gnomon and the conic sundial represent dual geometric realisations
of the same astronomical quantities:

* In the **vertical gnomon**, the receiving surface is horizontal ($Z = 0$),
  and the latitude $\varphi$ is recorded directly in the equinoctial shadow length
  $s_E = g\tan\varphi$.
* In the **conic sundial**, the receiving surface is an inclined cone whose axis
  is tilted at angle $\varphi$. The equinoctial mark on the cone is invariant
  ($\ell_E = d$), and the latitude is recorded in the spatial orientation of
  the cone axis and the angular span of the solstitial hour network.

@tbl-ptolemy-shadows lists the canonical shadow lengths across the seven ancient
klimata computed using Ptolemy's obliquity $\varepsilon = 23^\circ 51' 20''$
[@toomer1984, Book II].

| Location | Klima | Latitude $\varphi$ | Equinox $s_{\mathrm{eq}}$ | Summer $s_{\mathrm{su}}$ | Winter $s_{\mathrm{wi}}$ | $\sigma_E = s_{\mathrm{eq}}/g$ |
|:---|:---:|:---:|:---:|:---:|:---:|:---:|
| Syene | II | $23^\circ 51'$ | $26;32$ | $0;00$ (zenith) | $65;57$ | $0.442$ |
| Alexandria | III | $30^\circ 58'$ | $36;00$ | $7;29$ | $85;08$ | $0.600$ |
| Rhodes | IV | $36^\circ 00'$ | $43;36$ | $12;55$ | $103;19$ | $0.727$ |
| Delos | — | $37^\circ 24'$ | $45;52$ | $14;27$ | $109;23$ | $0.765$ |
| Athens | — | $37^\circ 58'$ | $46;49$ | $15;05$ | $112;00$ | $0.780$ |
| Hellespont | V | $40^\circ 56'$ | $52;02$ | $18;26$ | $127;27$ | $0.867$ |
| Rome | — | $41^\circ 54'$ | $53;50$ | $19;33$ | $133;14$ | $0.897$ |
| Massalia | VI | $43^\circ 05'$ | $56;07$ | $20;56$ | $140;56$ | $0.935$ |

: Canonical noon shadow lengths for a vertical gnomon of $g = 60$ parts, computed from Ptolemy's klimata table ($\varepsilon = 23^\circ 51' 20''$). Sexagesimal parts ($p;q = p + q/60$). {#tbl-ptolemy-shadows}

# Inversion: From Surface Marks to Construction Parameters

When reconstructing an ancient sundial from a 3D scan, we face the inverse problem:
given the coordinates of surviving engraved marks on the stone, recover
$(\alpha, d, \varphi, A)$.

## Classical Inversion (Apex $A$ Preserved)

If the apex $A$ is physically preserved, parameter recovery proceeds directly:

1. Measure slant distances from $A$ to the noon marks:
   $L_S = \|\mathbf{P}_S - A\|$, $L_E = \|\mathbf{P}_E - A\|$, $L_W = \|\mathbf{P}_W - A\|$.
2. Compute the solstitial slant ratio:
   $$
   \frac{L_S}{L_W} = \frac{\ell_S}{\ell_W} = \frac{1 + x}{1 - x} = \frac{1 + \tan\alpha\tan\varepsilon}{1 - \tan\alpha\tan\varepsilon}.
   $$ {#eq-slants-ratio}
   *Reading.* The ratio of summer to winter slant distances depends solely on
   $\alpha$ and $\varepsilon$, completely independent of latitude $\varphi$.
3. Invert @eq-slants-ratio for the cone half-angle:
   $$
   \tan\alpha = \frac{1}{\tan\varepsilon}\left(\frac{L_S/L_W - 1}{L_S/L_W + 1}\right).
   $$ {#eq-alpha-classical}
   *Reading.* The tangent of the half-angle equals the normalized difference-to-sum
   ratio of the solstitial slant marks divided by $\tan\varepsilon$.
4. Compute the nodus distance using the equinox invariant:
   $$
   d = L_E\cos\alpha.
   $$ {#eq-d-classical}
   *Reading.* The nodus distance $d$ equals the equinoctial slant distance $L_E$
   multiplied by $\cos\alpha$.

@tbl-alpha-lookup lists the slant ratio $L_S / L_W$ as a function of half-angle $\alpha$
at $\varepsilon = 23.44^\circ$.

| Half-Angle $\alpha$ | $20^\circ$ | $25^\circ$ | $30^\circ$ | $35^\circ$ | $40^\circ$ | $45^\circ$ | $50^\circ$ | $55^\circ$ | $60^\circ$ |
|:---|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
| Ratio $L_S / L_W$ | $1.37$ | $1.50$ | $1.67$ | $1.88$ | $2.14$ | $2.53$ | $3.14$ | $4.19$ | $7.03$ |

: Solstice slant-distance ratio $L_S/L_W$ as a function of cone half-angle $\alpha$ ($\varepsilon = 23.44^\circ$). {#tbl-alpha-lookup}

## Apex-Free Inversion via the Marking Ratio $Q$

On damaged fragments where the apex $A$ is broken away, the three seasonal noon
marks $P_S, P_E, P_W$ alone suffice to recover $\alpha$, $A$, and $d$ in closed
form.

We define the dimensionless **marking ratio** $Q$:

$$
Q := \frac{L_S - L_W}{L_S - L_E}.
$$ {#eq-Q-def}

*Reading.* $Q$ is the ratio of the total solstitial distance $(L_S - L_W)$ to
the summer-to-equinox distance $(L_S - L_E)$ along the meridian generator.

Expressing the slant differences in terms of $d, \alpha,$ and $x = \tan\alpha\tan\varepsilon$:

$$
L_S - L_W = \frac{d}{\cos\alpha}\left(\frac{1}{1-x} - \frac{1}{1+x}\right) = \frac{2d\,x}{\cos\alpha\,(1-x^2)},
$$ {#eq-diff-sw}

$$
L_S - L_E = \frac{d}{\cos\alpha}\left(\frac{1}{1-x} - 1\right) = \frac{d\,x}{\cos\alpha\,(1-x)}.
$$ {#eq-diff-se}

Dividing @eq-diff-sw by @eq-diff-se:

$$
Q = \frac{2}{1+x} = \frac{2}{1 + \tan\alpha\tan\varepsilon}.
$$ {#eq-Q-relation}

*Reading.* The marking ratio $Q$ is a simple rational function of the workshop
parameter $x$, completely independent of the scale $d$ and latitude $\varphi$.

Solving @eq-Q-relation explicitly for $\alpha$:

$$
\alpha = \arctan\!\left(\frac{2 - Q}{Q\,\tan\varepsilon}\right), \qquad 1 < Q < 2.
$$ {#eq-alpha-from-Q}

*Reading.* The cone half-angle $\alpha$ is the inverse tangent of $(2-Q)$ divided
by the product $Q\tan\varepsilon$.

```{=typst}
#arxivai.concept-box(title: [Physical Scope of the $Q$-Formula])[
  The mathematical condition $1 < Q < 2$ corresponds to $0 < x < 1$, which is
  the exact physical condition that the cone opens ($alpha > 0 degree$) and that
  the summer solstice rays strike the interior nappe ($alpha < 90 degree - epsilon$).
]
```

### The Step-by-Step Apex-Free Algorithm

1. **Generator Direction:** Compute the unit direction vector along the meridian
   from $P_S$ toward $P_W$ (inward toward the apex):
   $$
   \hat{\mathbf{g}} = \frac{\mathbf{P}_W - \mathbf{P}_S}{\|\mathbf{P}_W - \mathbf{P}_S\|}.
   $$ {#eq-g-unit}
   *Reading.* $\hat{\mathbf{g}}$ is the unit vector pointing along the meridian line
   from the summer mark toward the winter mark.
2. **Projected Segment Lengths:** Compute the total baseline $g_{\mathrm{len}}$ and
   the projected equinox distance $t_E$:
   $$
   g_{\mathrm{len}} = \|\mathbf{P}_W - \mathbf{P}_S\|, \qquad t_E = (\mathbf{P}_E - \mathbf{P}_S)\cdot\hat{\mathbf{g}}.
   $$ {#eq-proj-lengths}
   *Reading.* $g_{\mathrm{len}}$ is the physical distance between summer and winter
   marks; $t_E$ is the projected distance from summer to equinox. Collinearity is
   verified if the orthogonal residual $\|\mathbf{P}_E - (\mathbf{P}_S + t_E\hat{\mathbf{g}})\|$
   is within scan noise ($\lesssim 0.3\,\mathrm{mm}$).
3. **Recover Half-Angle $\alpha$:** Compute $Q = g_{\mathrm{len}} / t_E$ and evaluate
   $\alpha = \arctan((2-Q)/(Q\tan\varepsilon))$.
4. **Locate the Apex $A$:** The slant ratio $r_S = L_S / L_E = 1/(1-x)$ gives the
   apex distance parameter $t_A$:
   $$
   t_A = \frac{r_S\,t_E}{r_S - 1} = L_S, \qquad A = \mathbf{P}_S + t_A\,\hat{\mathbf{g}}.
   $$ {#eq-apex-recovery}
   *Reading.* The distance $t_A$ along $\hat{\mathbf{g}}$ from $P_S$ places the
   true apex $A$ in 3D space.
5. **Recover Scale $d$:**
   $$
   L_E = t_A - t_E, \qquad d = L_E\cos\alpha.
   $$ {#eq-d-recovered}
   *Reading.* Subtracting $t_E$ from $L_S$ gives $L_E$, whose product with
   $\cos\alpha$ yields the scale $d$.

## Independent Latitude Recovery ($\varepsilon$-Free)

Latitude $\varphi$ is recovered independently of the noon marks by pairing the
radial placement $x$ of a solstitial day-line with its measured half-day hour
angle $H$:

$$
\tan\varphi = \tan\alpha\,\frac{|\cos\tilde{H}|}{|x|}.
$$ {#eq-phi-efree}

*Reading.* The tangent of latitude $\varphi$ equals $\tan\alpha$ multiplied by the
ratio of the absolute cosine of the unit-calibrated half-day arc $\tilde{H}$ to
the placement parameter $x$. The realized epoch obliquity $\tan\delta$ cancels
out completely, rendering the latitude recovery robust against epoch uncertainties.

## Marking Sufficiency

@tbl-sufficiency catalogs which parameters are determinable from various
observable feature sets.

| Available Marking Set | Recoverable Parameters | Undetermined Parameters | Analytical Method |
|:---|:---|:---|:---|
| $\{P_S, P_E\}$ | Meridian generator line $\hat{\mathbf{g}}$, segment $L_S - L_E$ | $\alpha, d, \varphi, A$ | Linear interpolation |
| $\{P_S, P_E, P_W\}$ | $Q$, half-angle $\alpha$, apex $A$, slants $L_S,L_E,L_W$, scale $d$ | Latitude $\varphi$, axis tilt | $Q$-inversion (@eq-alpha-from-Q–@eq-d-recovered) |
| $\{P_S, P_E, P_W\} + \ge 3$ hour marks | Complete geometry: $\alpha, d, A, \hat{\mathbf{C}}, \varphi$ | None (complete reconstruction) | $Q$-inversion + 3D circle fit |
| $\{A, P_S, P_E, P_W\}$ | $\alpha, d, L_S, L_E, L_W$ | Latitude $\varphi$ | Classical inversion (@eq-alpha-classical) |
| Dressed seat & upper front face | Cone axis tilt $\varphi$, face orientation | Scale $d$, half-angle $\alpha$ | Plane normal regression |

: Marking sufficiency catalog for conic sundial reconstruction. {#tbl-sufficiency}

# Synthetic Validation: A Rhodes Construction

We validate the apex-free algorithm on an exact synthetic model constructed for
Rhodes ($\varphi = 36.00^\circ$, $\alpha = 45.00^\circ$, $d = 100.00\,\mathrm{mm}$,
$\varepsilon = 23.44^\circ$).

1. **Forward Generation:**
   * $x = \tan 45^\circ\tan 23.44^\circ = 0.433568$.
   * Axial marks: $\ell_S = 176.544\,\mathrm{mm}$, $\ell_E = 100.000\,\mathrm{mm}$, $\ell_W = 69.756\,\mathrm{mm}$.
   * Slant marks ($L = \ell/\cos 45^\circ$): $L_S = 249.670\,\mathrm{mm}$, $L_E = 141.421\,\mathrm{mm}$, $L_W = 98.650\,\mathrm{mm}$.
2. **Segment Input:**
   * $g_{\mathrm{len}} = L_S - L_W = 151.020\,\mathrm{mm}$.
   * $t_E = L_S - L_E = 108.249\,\mathrm{mm}$.
3. **Inverse Recovery:**
   * Marking ratio: $Q = 151.020 / 108.249 = 1.395121$.
   * Half-angle:
     $$
     \alpha = \arctan\!\left(\frac{2 - 1.395121}{1.395121 \times \tan 23.44^\circ}\right) = \arctan(1.000000) = 45.00^\circ.
     $$
   * Scale recovery: $r_S = 1/(1-0.433568) = 1.765436$; $t_A = (1.765436 \times 108.249)/0.765436 = 249.670\,\mathrm{mm} = L_S$.
   * Equinox slant: $L_E = 249.670 - 108.249 = 141.421\,\mathrm{mm}$.
   * Nodus distance: $d = 141.421 \times \cos 45^\circ = 100.000\,\mathrm{mm}$.

@tbl-recovery-results confirms exact numerical closure.

| Parameter | Nominal Ground Truth | Inverted Estimate | Absolute Error |
|:---|:---:|:---:|:---:|
| Half-angle $\alpha$ | $45.0000^\circ$ | $45.0000^\circ$ | $< 10^{-6\,\circ}$ |
| Marking ratio $Q$ | $1.395121$ | $1.395121$ | $< 10^{-6}$ |
| Slant ratio $L_S / L_W$ | $2.530873$ | $2.530873$ | $< 10^{-6}$ |
| Nodus scale $d$ | $100.000\,\mathrm{mm}$ | $100.000\,\mathrm{mm}$ | $< 10^{-5}\,\mathrm{mm}$ |

: Numerical verification of closed-form parameter recovery on the synthetic Rhodes model. {#tbl-recovery-results}

# What Archaeological Stones Add

Applying this analytical geometry to scanned archaeological stones introduces
real-world physical effects documented in the companion studies of ObjID 18
(British Museum inv. 2547) [@grasshoff2026measurement; @bsdp2014dialface18] and
ObjID 126 (Delos) [@grasshoff2026objid126]:

1. **Engraved Hour Lines as Circular Arcs:** On ObjID 18, the engraved hour lines
   are circular arcs with radii between $113\,\mathrm{mm}$ and $280\,\mathrm{mm}$
   (fit RMS $0.20$--$0.71\,\mathrm{mm}$), rather than mathematical generators
   (which leave bow residuals of $2.2$--$7.6\,\mathrm{mm}$). The straight line of
   the theoretical cone represents the geometric ideal; the circular arc is the
   artisan's physical realization.
2. **Groove Centreline Extraction:** At photogrammetric scan resolutions, the
   V-shaped chisel notch of worn engravings lies within the $\pm 0.3\,\mathrm{mm}$
   surface noise floor. Centreline extraction must be performed by medial-line
   filtering across selected feature bands rather than unconstrained depth minimization.
3. **Construction Error Attribution:** On ObjID 18, the winter hour division and
   cone axis yield a consensus latitude of $\varphi = 41.45^\circ \pm 0.30^\circ$
   ($s = 53;02$ for $g=60$, corresponding to Ptolemy's Hellespont/Rome klimata),
   while the summer hour division deviates at $14\sigma$ — an isolated ancient
   artisan error correctly identified by leave-one-out diagnostics rather than
   obscured by global averaging.

# Constraint Catalog

@tbl-constraints summarizes the complete set of formal geometric constraints
governing the conic sundial with a horizontal gnomon.

| Constraint ID | Physical Name | Formal Mathematical Expression |
|:---|:---|:---|
| **N1** | Nodus on Polar Axis & Horizon | $\mathbf{N}_{\mathrm{phys}} = (0,0,0)$; in math frame: $\mathbf{N} = d\,\hat{\mathbf{O}}$ |
| **G1** | Gnomon in Meridian Plane | Horizontal rod at $X = 0, Z = 0$, extending northward ($+Y_w$) |
| **C1** | Apex on Polar Axis | $A_{\mathrm{phys}} = d\,\hat{\mathbf{C}}$; in math frame: $A = (0,0,0)$ |
| **C2** | Cone Axis = Polar Axis | $\hat{\mathbf{C}} = (0,\ \cos\varphi,\ \sin\varphi)$ |
| **C3** | South-Opening Nappe | $\hat{\mathbf{O}} = -\hat{\mathbf{C}}$; $\mathbf{P}\cdot\hat{\mathbf{C}} < 0$ |
| **C4** | Cavity Scale Bound | $d < H$ (nodus located within cone cavity) |
| **C5** | Rim Circularity | $R_{\mathrm{rim}} = H\tan\alpha$ in plane orthogonal to $\hat{\mathbf{C}}$ |
| **E1** | Equinox Invariant | $\ell_E = d$ for all $\alpha, \varphi$ |
| **E2** | Workshop Proportions | $x = \tan\alpha\tan\varepsilon$; $\ell_S = d/(1-x), \ell_W = d/(1+x)$ |
| **E3** | Noon Triplet Uncoupling | $\ell_W : \ell_E : \ell_S = (1+x)^{-1} : 1 : (1-x)^{-1}$ (independent of $\varphi$) |
| **H1** | Equinoctial Calibration | $H_E = 90^\circ$ at all latitudes (engraver's unit standard) |
| **H2** | Solstitial Hour Arc | $\cos H(\pm\varepsilon) = -\tan\varphi\tan(\pm\varepsilon)$ |
| **M1** | Meridian Collinearity | $P_x = 0$ at $h = 0$ for all solar declinations $\delta$ |

: Complete constraint catalog for the conic sundial with horizontal gnomon. {#tbl-constraints}

# Conclusion

The construction geometry of the ancient conic sundial with a nodus on the polar
axis is governed by an elegant uncoupling: the three seasonal noon marks on the
meridian generator encode the cone half-angle $\alpha$ and scale $d$ without any
dependence on geographical latitude $\varphi$.

The closed-form $Q$-formula enables exact reconstruction of damaged fragments
even when the apex is missing. Latitude $\varphi$ must be derived from the
solstitial hour division or the dressed reference planes of the block. By
grounding computational 3D analysis in this rigorous proportional framework,
archaeological sundials can be recovered not merely as geometric meshes, but as
precise historical witnesses to ancient mathematical astronomy.

# References {.unnumbered}