Horus: Ancient Deity and Modern Science

• Horus is an ancient Egyptian divine name that now designates a range of scientific instruments, computational frameworks, and datasets across multiple disciplines.
• In astronomy and astrophysics, Horus links mythological records like the Cairo Calendar with modern high-resolution spectrographs and gravitational lens systems to study celestial phenomena.
• Horus also represents advanced detectors in nuclear astrophysics, robust frameworks in federated learning and blockchain forensics, and comprehensive seismic catalogs, highlighting its versatile application in research.

Horus is an ancient Egyptian divine name that, in modern technical literature, also designates several distinct instruments, computational systems, datasets, and astrophysical objects. The term appears in historical astronomy as a deity associated with the variable star Algol, in observational astronomy as both a spectrograph and a gravitational-lens designation, in nuclear astrophysics as a $\gamma$-ray spectrometer, in seismology as a homogenized earthquake catalog, and in computer systems research as the name of multiple frameworks for resource estimation, federated learning, mixed-reality robot supervision, autonomous-vehicle sensing, and blockchain attack analysis (Jetsu et al., 2016).

1. Ancient Egyptian Horus and the Cairo Calendar

In the Cairo Calendar, Horus is treated as an active divine agent within a hemerological system that assigns good and bad prognoses to the days of the Egyptian civil year. Statistical analyses of the calendar detect a period of about $2.850$ days in the prognoses, alongside a lunar period of $29.6$ days. The phase formalism used in that work is $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$, with ephemerides $t_0=0.53,\ P=2.850\,\mathrm{d}$ for Algol and $t_0=3.50,\ P=29.6\,\mathrm{d}$ for the Moon. The interpretation advanced in the literature is that the $2.850$-day signal corresponds to Algol’s period in that epoch and that Horus received this title because Egyptian scribes described celestial phenomena as the activity of gods (Porceddu et al., 2018).

Within that framework, Horus is the strongest mythological representation of the Algol signal. The relevant study reports for Horus a contribution $z_x = +3.5$ with Rayleigh-test significance $Q_z = 0.03$ and $n_G = 14$, with lucky Horus prognoses clustering near the bright phase and several unlucky Horus prognoses appearing just after eclipse. The same study argues that the Cairo Calendar thereby preserves the oldest surviving documentation of a variable star’s periodicity, while Seth is tied more closely to the lunar cycle and especially to dark lunar phases (Jetsu et al., 2016).

The historical-astronomical interpretation is supported by observational arguments. Algol’s primary eclipse produces a naked-eye brightness drop from about $2.850$0 to $2.850$1 for roughly $2.850$2 hours, and the calendar analyses emphasize the recurring “3+3+13 = 19” day pattern and the $2.850$3-day repetition expected for a $2.850$4-day period. This suggests that Egyptian hour-watchers could have recognized a stable night-time cadence without modern instrumentation (Porceddu et al., 2018).

2. HORuS on the Gran Telescopio Canarias

HORuS, the High Optical Resolution Spectrograph, is a moderate/high-resolution optical echelle spectrograph on the $2.850$5 m Gran Telescopio Canarias at the Nasmyth-B focal plane. It provides nearly continuous coverage from $2.850$6 to $2.850$7 Å in $2.850$8 spectral orders at $2.850$9, uses a $29.6$0 microlens integral field unit of $29.6$1 arcsec feeding optical fibers to a pseudo-slit, and employs a $29.6$2 Fairchild CCD with $29.6$3 pixels and 16-bit electronics. Wavelength calibration uses Th–Ar hollow-cathode lamps, and the custom pipeline chain performs reduction tasks including bias subtraction, order tracing and extraction, and wavelength calibration. The instrument is explicitly described as not being thermally, optically, or mechanically stabilized (Tabernero et al., 2020).

The instrument’s first exoplanet transmission-spectroscopy demonstration targeted 55 Cnc e. One transit was observed on 12 Dec 2018, with four $29.6$4 s exposures followed by twenty $29.6$5 s exposures over about $29.6$6 h. The study reports residuals for the Na I doublet difference spectrum of $29.6$7, establishing an upper limit one order of magnitude more stringent than previous literature limits, and an H$29.6$8 upper limit of $29.6$9. The analysis also showed residuals consistent with photon noise and argued that giant planets with transmission depths above $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$0 at the $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$1 level could be easily detected with HORuS under comparable conditions (Tabernero et al., 2020).

HORuS was subsequently used for two transits of KELT-7b, where it provided $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$2 spectroscopy over approximately $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$3–$\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$4 Å. In that study, HORuS data supported revised stellar and planetary parameters, a combined Doppler-tomographic obliquity measurement of $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$5 deg, and a derived 3D obliquity $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$6 deg (or $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$7 deg). The transmission-spectroscopy analysis yielded corrected upper limits of $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$8 for Na I D, $\phi = \mathrm{mod}\big(\frac{t-t_0}{P},1\big)$9 for H$t_0=0.53,\ P=2.850\,\mathrm{d}$0, $t_0=0.53,\ P=2.850\,\mathrm{d}$1 for Mg I b, $t_0=0.53,\ P=2.850\,\mathrm{d}$2 for Ca II H&K, and $t_0=0.53,\ P=2.850\,\mathrm{d}$3 for Li I after accounting for Rossiter–McLaughlin, center-to-limb variation, and stellar-activity contamination (Tabernero et al., 2022).

In stellar archaeology, HORuS enabled the high-resolution confirmation of the thin-disk r-II star LAMOST J020632.21+494127.9. The GTC/HORuS spectrum covered $t_0=0.53,\ P=2.850\,\mathrm{d}$4–$t_0=0.53,\ P=2.850\,\mathrm{d}$5 Å at $t_0=0.53,\ P=2.850\,\mathrm{d}$6 with $t_0=0.53,\ P=2.850\,\mathrm{d}$7 at $t_0=0.53,\ P=2.850\,\mathrm{d}$8 Å, and the analysis derived $t_0=0.53,\ P=2.850\,\mathrm{d}$9, $t_0=3.50,\ P=29.6\,\mathrm{d}$0, $t_0=3.50,\ P=29.6\,\mathrm{d}$1, and $t_0=3.50,\ P=29.6\,\mathrm{d}$2. The star showed $t_0=3.50,\ P=29.6\,\mathrm{d}$3, $t_0=3.50,\ P=29.6\,\mathrm{d}$4, and $t_0=3.50,\ P=29.6\,\mathrm{d}$5, the latter reported as the highest measured among known RPE stars listed in that paper, and its orbit was characterized as thin-disk with eccentricity $t_0=3.50,\ P=29.6\,\mathrm{d}$6 (Xie et al., 2024).

The R-Process Alliance’s fifth data release extended HORuS work to faint halo stars. That campaign obtained snapshot spectra for 41 stars with magnitudes $t_0=3.50,\ P=29.6\,\mathrm{d}$7, exposure times of $t_0=3.50,\ P=29.6\,\mathrm{d}$8–$t_0=3.50,\ P=29.6\,\mathrm{d}$9 s, and $2.850$0 between $2.850$1 and $2.850$2 per pixel near $2.850$3 Å. The study reports the discovery of five CEMP stars, one limited-$2.850$4 star, three $2.850$5-I stars, four $2.850$6-II stars, and six Mg-poor stars, and it uses those abundances to discuss [Mg/Eu] trends, neutron-capture patterns, and a possible globular-cluster escapee at $2.850$7 (Bandyopadhyay et al., 2024).

3. HORUS in nuclear astrophysics and detector simulation

A separate entity, written in all capitals, is HORUS, the High efficiency Observatory for $2.850$8-Ray Unique Spectroscopy at the University of Cologne. It is a high-efficiency HPGe $2.850$9-ray spectrometer coupled to the 10 MV FN tandem ion accelerator and designed for in-beam measurements of absolute and partial cross sections for charged-particle induced reactions at astrophysical energies. The array consists of up to 14 HPGe detectors mounted on the faces and corners of a cube, with six positions optionally equipped with active BGO Compton-suppression shields, and provides five laboratory angles relative to the beam axis: $z_x = +3.5$0, $z_x = +3.5$1, $z_x = +3.5$2, $z_x = +3.5$3, and $z_x = +3.5$4 (Netterdon et al., 2014).

The nuclear-astrophysics implementation emphasizes efficiency calibration, coincidence capability, and angular-distribution analysis. HORUS uses event-by-event list-mode data acquisition, supports $z_x = +3.5$5–$z_x = +3.5$6 coincidence matrices, and enables extraction of partial cross sections through fits of the form

$z_x = +3.5$7

Commissioning on $z_x = +3.5$8Y(p,$z_x = +3.5$9)$Q_z = 0.03$0Zr yielded total cross sections from $Q_z = 0.03$1 mb at $Q_z = 0.03$2 keV to $Q_z = 0.03$3 mb at $Q_z = 0.03$4 keV, in excellent agreement with previous measurements (Netterdon et al., 2014).

A later technical update described a redesigned target chamber, in-beam Rutherford backscattering diagnostics, and sensitivity down to the nb region. In that work, HORUS measured $Q_z = 0.03$5Nb(p,$Q_z = 0.03$6)$Q_z = 0.03$7Mo total cross sections of $Q_z = 0.03$8 at $Q_z = 0.03$9 keV, $n_G = 14$0 at $n_G = 14$1 keV, and $n_G = 14$2 at $n_G = 14$3 keV, values reported as higher than previously published data by a factor of about $n_G = 14$4 (Heim et al., 2020).

HORUS is also part of the combined SONIC@HORUS setup, in which the SONIC silicon particle spectrometer is embedded within the HPGe geometry for particle–$n_G = 14$5 coincidence measurements. That system provides up to 12 silicon $n_G = 14$6–$n_G = 14$7 telescopes with total solid-angle coverage of about $n_G = 14$8, supports event-by-event Doppler correction, and was used to study reactions such as $n_G = 14$9Mo(p,p$2.850$00), $2.850$01Sn(d,X), $2.850$02Ni(p,p$2.850$03), and $2.850$04Mo(p,p$2.850$05) (Pickstone et al., 2017).

Two simulation tools also bear closely related names. G4Horus is an open-source Geant4 application for efficiency calculations in the Cologne HORUS array, including full-energy-peak, single-escape, and double-escape efficiencies up to about $2.850$06 MeV, with detailed CAD-derived chamber geometries and explicit modeling of detector dead regions (Mayer et al., 2020). Independently, HORUS was used as an HPAD detector simulation program in a study of AGIPD design choices for X-ray Photon Correlation Spectroscopy at the European XFEL; that work found that aperturing is not beneficial at low intensities, becomes beneficial above approximately $2.850$07 photon per $2.850$08, and that $2.850$09 pixels outperform $2.850$10 pixels above approximately $2.850$11 photons per $2.850$12 (Becker et al., 2011).

4. Horus in computing, robotics, and digital infrastructure

Several recent systems papers use Horus as the name of an algorithmic or operational framework rather than an instrument.

Name	Domain	Core function
Horus	GPU resource management	analytical training-memory estimator
Horus	Federated learning	LoRA-based robust FL under poisoning and heterogeneity
HORUS	Mixed reality robotics	MR interface for teams of mobile robots
Horus	Autonomous driving	fusion of infrastructure and on-board sensors
Horus	Blockchain forensics	detection and tracing of smart-contract attacks

In GPU scheduling, Horus appears as a representative analytical memory estimator for training-aware resource management. It performs a lightweight closed-form pass over model statistics, requires detailed architectural specifications, and is intended for pre-execution memory sizing to avoid OOM during collocation. A 2026 comparative study reports that Horus has negligible overhead and acts as a conservative upper bound, but is hardware-dependent, misses allocator behavior and framework-level optimizations, and systematically overestimates memory consumption across MLP configurations (Yousefzadeh-Asl-Miandoab et al., 19 Feb 2026).

In federated learning, Horus is expanded as Heterogeneity-Oblivious Robust federated learning Under hyper-heterogeneity and poisoning attackS. It inserts LoRA adapters into empirically stable layers, aggregates only adapter parameters, computes a Heterogeneity-Oblivious Poisoning Score from the singular spectra of LoRA-A, and applies projection-aware aggregation based on dominant singular vectors. Experiments across 54 dataset–attack–heterogeneity settings report first place in 49 settings and second place in the remaining 5, with average gains of $2.850$13 on CIFAR-10, $2.850$14 on CIFAR-100, and $2.850$15 on FMNIST over the strongest baseline (Zhang et al., 5 Aug 2025).

In mixed-reality robotics, HORUS denotes “Holistic Operational Reality for Unified Systems,” an MR interface running on Meta Quest 3 for supervising teams of Husarion ROSbot 2.0 robots. It provides a mini-map Ground Station, robot-specific status and visualization panels, goal and waypoint assignment, and two teleoperation modes. In the reported user study, the HORUS condition completed the search task in a mean time of 8:42 versus 11:17 for a teleoperation-only baseline, was approximately $2.850$16 faster overall, and achieved a System Usability Scale mean of $2.850$17 versus $2.850$18 (Adekoya et al., 3 Jun 2025).

In autonomous-vehicle safety, Horus is a proof-of-concept system that fuses on-vehicle and infrastructure-based sensing. It combines steering commands from a Raspberry Pi camera and two fixed webcams through confidence-aware weighting and was evaluated on a small-scale robotic testbed. The study reports that periodic on-vehicle outages of about $2.850$19–$2.850$20 s at a scaled speed of about $2.850$21 km/hr caused crashes, whereas the fused system could tolerate sensors that fail $2.850$22 of the time and still navigate safely when confidence-weighted averaging was used (Seshan, 2020).

In Ethereum security, Horus is a full pipeline for post-deployment attack detection, quantification, and tracing. It replays historical transactions with Geth tracing, converts execution traces into Datalog facts, uses Soufflé for logic-driven attack detection, and stores asset flows in Neo4j for graph analysis. On Ethereum transactions up to May 2020, the framework identified 1,888 attacked smart contracts and 8,095 adversarial transactions, and it was further demonstrated on the Uniswap and Lendf.me attacks (Torres et al., 2021).

5. The “Eye of Horus” in modern astrophysics

The “Eye of Horus” is the nickname of the gravitational lens system HSC J142449−005322, discovered in the Hyper Suprime-Cam Subaru Strategic Program. The system consists of a very massive early-type lens galaxy at $2.850$23 and two spectroscopically confirmed source planes at $2.850$24 and $2.850$25. It was described as the first double source-plane system with spectroscopic redshifts of both sources, and lens modeling showed that some lensed features could not be reproduced by a smooth potential without added substructure (Tanaka et al., 2016).

The environment of the Eye of Horus was later studied with XMM–Newton. That work detected two extended X-ray sources: a main cluster centered on the lens and a northeast cluster about $2.850$26 arcsec away. Under hydrostatic assumptions, the main cluster was assigned $2.850$27 and the northeast cluster $2.850$28. The projected mass contribution of the main cluster within the Einstein radius was estimated as $2.850$29, corresponding to $2.850$30–$2.850$31 of the total enclosed lensing mass, whereas the northeast cluster contributed only about $2.850$32 (Tanaka et al., 2019).

A dedicated MMT/Binospec campaign subsequently confirmed that the field contains two distinct galaxy clusters rather than a physically associated pair. The main cluster lies at $2.850$33 and the northeast cluster at $2.850$34; they are separated by about $2.850$35 arcsec on the sky but have a line-of-sight velocity offset of about $2.850$36, which the authors interpreted as a line-of-sight projection rather than a merger. The same survey characterized both brightest cluster galaxies as old and massive, with ages from $2.850$37 to $2.850$38 Gyr and stellar masses in the range $2.850$39–$2.850$40 (Di et al., 2023).

6. HORUS as a seismic catalog and as a recurrent scholarly label

In seismology, HORUS denotes the Homogenized Instrumental Seismic catalog of Italy, maintained by INGV and covering 1960 to the present. A 2025 study used the Italy HORUS dataset to formalize the EEPAS and PPE earthquake-forecasting models as inhomogeneous Poisson point processes, reproduce published Italy results within about one hour, and implement a fully automated pipeline from raw catalog to parameter estimation and pyCSEP evaluation. For the reproduced setup, the paper used a testing region of 177 contiguous cells, a learning period of 1990–2011, a pseudo-prospective testing period of 2012–2021, and thresholds such as $2.850$41 or $2.850$42 and $2.850$43 depending on the experiment (Chung et al., 15 Dec 2025).

Across these literatures, the name is attached to fundamentally different objects. The expansions include High Optical Resolution Spectrograph, High efficiency Observatory for $2.850$44-Ray Unique Spectroscopy, Heterogeneity-Oblivious Robust federated learning Under hyper-heterogeneity and poisoning attackS, Holistic Operational Reality for Unified Systems, and Homogenized Instrumental Seismic catalog of Italy, while the Eye of Horus and the Cairo Calendar preserve explicitly astronomical and mythological uses (Tabernero et al., 2020). This distribution suggests that “Horus” functions less as a single research lineage than as a reusable scholarly label spanning instrumentation, algorithms, field names, and historical interpretation.