Târgu Jiu · Constantin Brâncuși · 1938
The Endless Column and its shadow
Brâncuși's column in Târgu Jiu stands 29.35 m tall: repeated beads facing the sky. Scroll: the Sun goes through a whole year over the real park, and its shadow draws a pattern.
Scroll to move the Sun.
- Clock
- –
- Shadow tip at
- –
- Points to
- –
- Sun height
- –
Târgu Jiu · Constantin Brâncuși · 1938
Do the three works line up with the Sun?
The claim says that Brâncuși set the Table of Silence, the Gate of the Kiss and the Column on a solstice axis. I found no published source that says so. The sources I found say "west–east axis" (the UNESCO evaluation, 2024). The test, its method, its tolerance and its list of events were written down on 13 November 2026, before any angle was computed.
Result: not aligned with the solstice Sun. The three works do lie on one straight line, but none of the solstice or equinox sunrises and sunsets comes closer than 8.5° to it.
The three works lie on one straight line. Measured from the Table to the Column, the line is 1,493 m long and runs at 99.4° (clockwise from north), which is 9.4° south of due east. The Table to the Gate runs at 100.3°, the Gate to the Column at 99.3°, and the Gate lies 2.6 m from the line between the other two.
| Event at Târgu Jiu | Sun azimuth | Tested against | Miss |
|---|---|---|---|
| Midsummer sunrise | 54.6° | sunrise, line to the east | 44.8° |
| Midsummer sunset | 305.4° | sunset, line to the west | 25.9° |
| Midwinter sunrise | 123.2° | sunrise, line to the east | 23.8° |
| Midwinter sunset | 236.8° | sunset, line to the west | 42.6° |
| Equinox sunrise | 89.1° | sunrise, line to the east | 10.3° |
| Equinox sunset | 270.9° | sunset, line to the west | 8.5° |
| Sunrise, 27 October 1938 (exploratory) | 107.1° | sunrise, line to the east | 7.7° |
| Sunset, 27 October 1938 (exploratory) | 252.9° | sunset, line to the west | 26.5° |
The tolerance was fixed at 0.5° (the Sun's half-width plus the position error). Nothing comes within it. The nearest of the six events is the equinox sunset, 8.5° away. Sunrise on 27 October 1938, the date the ensemble was inaugurated, is the one dated event I allowed myself; it misses by 7.7°, and one date on its own would prove nothing. A line pointing in a random direction would match one of the six events within 0.5° about 1.7 % of the time. This one does not.
So the ensemble is straight to within a few metres, runs east-south-east to west-north-west, and is not aligned with a solstice or the equinox. "West–east" is true to within about 9°. After the fact, and not part of the test: the Sun rises along the line on 4 days of 2026, 1 and 2 March, 11 and 12 October. I know of no source that gives those dates a meaning.
Checks on the positions. On satellite imagery (Esri World Imagery, used only to measure) the foot of the Column lies about 3 m from the OpenStreetMap point. The Table and the Gate stand under trees there, so the imagery cannot place them. Wikidata's coordinates (rounded to four decimals for the Table and the Gate) differ from OpenStreetMap by 4 m (Column), 9 m (Table) and 24 m (Gate), and would give a line at 99.7°. Even if both ends were wrong by 25 m, the tolerance would grow only to 1.9°. The real horizon (Copernicus GLO-30 surface model, trees included) rises 0.4° east of the Table and 0.8° west of the Column on the line, too little to move any sunrise or sunset by 8°.
The test as written before it was run
Line: the bearing from the Table of Silence (OpenStreetMap node 3477307934) to the Column (node 925073636); also reported: Table to Gate, Gate to Column, and the Gate's sideways distance from the Table–Column line (collinear if 10 m or less).
Events: sunrise and sunset at the two solstices and at the equinox, flat horizon, upper edge of the Sun, refraction included, tested in both directions of the line; one dated event, 27 October 1938, exploratory only. No other event could be added afterwards.
Tolerance: T = √(0.27² + p²), in degrees, where 0.27° is half the Sun's width and p is the angle that a 5 m position error at each end makes over the 1,493 m line. T comes to 0.47°; the threshold was fixed at 0.5°. Aligned means within the tolerance of an event and collinear. Matching an arbitrary day of the year does not count.
Written on 13 November 2026 at 23:49 UTC. SHA-256 of the file: 126286b5979ec422…
Does the shadow point at the Table?
A common way to put the idea is that on some days the shadow of the Column points at the Table of Silence; I could not trace who says it first. In direction, it does, and on many days: whenever the Sun stands at 99.4°, the shadow lies along the line toward the Table and the Gate. That happens on 226 days of 2026, from 1 March to 12 October, in the morning. At 12:00 by the clock it never does: the shadow then points between 312° and 357°, north-west to north.
Reaching is another matter. The tip of the shadow would lie on the Table on about 5 Mar and 8 Oct, with the Sun 1.1° above a flat horizon. In June the shadow that points along the line is 32 m long.
And a thin column casts a thin shadow. The Sun is a disc 0.53° wide, so a 90 cm module gives a fully dark shadow only up to about 97 m beyond itself, and the 45 cm ends of each module only up to about half that. The shadow cast by the top of the Column is fully dark only when the Sun is higher than 18°. Lower than that it fades with distance: some partial shade can remain, but it is far too faint to see over the 1,493 m to the Table. Houses, trees and the park's own slope, which this page does not model, would stop it sooner.
Your day
Pick your birthday, or any day from 27 October 1938, when the ensemble was inaugurated, to the end of 2026. The picture shows the Column's shadow on the park at 10:00, 11:00, 12:00, 13:00, 14:00 and 15:00 by Romania's clock that day. Your date stays on your device.
Map © OpenStreetMap contributors
What is measured, what is calculated, what is modelled
Published: the height, 29.35 m, and the count of 15 whole modules and 2 half modules, from the evaluation of the UNESCO nomination (ICOMOS, 2024); the module size, 1.80 m tall, 45 cm at the ends and 90 cm at its widest, from the Journal of Engineering Science 26 (2019). The positions of the Column, the Gate and the Table come from OpenStreetMap, read on 13 November 2026.
Calculated: the position of the Sun (Meeus's algorithm, about 0.01° accuracy); Romania's clock for any date since 1938 (the international time-zone database, so a birthday in 1987 uses the clocks of 1987); the shadow, by geometry on flat ground, with the blur that the Sun's width of 0.53° gives it.
Modelled, not measured: the 0.55 m base is drawn as a plain block; the faces are turned square to the line of the three works, and I found no published source for their true direction; the two half modules end on their wide cut, because the modules meet at their narrow ends; the brass colour and shine are a guess; the ground is flat at the Column's level, though the real ground rises a few metres from the Table to the Column; OpenStreetMap has no trees here and no building heights, so none are drawn and buildings are flat; how fast the shadow fades with distance is an approximation, the mean module width (0.68 m) against the Sun's 0.53° disc; every shadow assumes a flat horizon; the sky colours are illustrative.
Brâncuși died in 1957, so the economic rights in his works last until 31 December 2027. Romanian law (Law 8/1996, art. 35(1)(f)) allows the image of a work that stands permanently in a public place to be reproduced without the author's consent, except when the work is the main subject of the image and the use is commercial. This page is not commercial. The permission carries conditions (good practice, no harm to the normal use of the work) and requires the author to be named: Constantin Brâncuși, Endless Column, 1938.
Limits
- The two published heights disagree: 29.35 m (the UNESCO evaluation, with 15 whole modules and 2 half modules; the 0.55 m base is from Sofronie and colleagues) and 29.26 m (an engineering study of the Column's wind response, 2013). The page uses 29.35 m; with 29.26 m every shadow is 0.3 % shorter.
- Flat ground at the Column's level. The real park slopes, houses and trees stand in the way, and none of that is in the drawing. Beyond about 100 m a shadow like this is only a faint blur in reality.
- The Sun's position is accurate to about 0.01° and ignores weather: on a cloudy day there is no shadow at all.
- The clock follows the international time-zone database. The page starts on 27 October 1938, the day the ensemble was inaugurated; for future dates the shadow is geometry only.
- The alignment test covers the directions I wrote down, with one dated event. It cannot show what Brâncuși intended, and it says nothing about other kinds of order in the ensemble, such as its proportions or its relation to the river and the streets.
- Positions come from OpenStreetMap volunteers; the Table and the Gate could not be checked on imagery because trees cover them.
Sources
- Height, modules and the west–east axis: ICOMOS, Evaluation of the nomination Brâncuși Monumental Ensemble of Târgu Jiu, 2024 (UNESCO World Heritage Centre, document 208445).
- Bead size: Journal of Engineering Science 26 (1), 2019, p. 124. The 29.26 m height: G. Solari, International Journal of High-Rise Buildings 2 (3), 2013, p. 197.
- Positions: OpenStreetMap contributors, ODbL, read through Overpass on 13 November 2026 (Column node 925073636, Table node 3477307934, Gate way 269780523, and the roads, paths, parks and buildings drawn). Cross-check: Wikidata Q2628659, Q9031865, Q2443657; Esri World Imagery for measurement only.
- Horizon: Copernicus GLO-30 digital surface model. Sun: J. Meeus, Astronomical Algorithms, chapter 25. Places for the direction lookup: OpenStreetMap.
- The 0.55 m base: Sofronie and colleagues, "Conservation of Gravitational Architecture", proceedings of the 4th International Conference on Conservation of Architectural Heritage.
- Rights: Law 8/1996 on copyright (republished), art. 35(1)(f). Bead width and the engineer's own account: Ș. Georgescu-Gorjan, "Geometria Coloanei infinite", România literară, 1970. No photographs are used.