A lumen is a daylight unit. The photopic curve behind it peaks at 555nm, where the cone-dense fovea is most sensitive when light levels are high. At night the eye hands part of the work to rods, the sensitivity peak slides toward 507nm, and a fixture can deliver more useful light than its lumen figure suggests. Comparing two products by lumen output alone therefore compares the wrong quantity for the job.
Two Curves, One Unit
The photopic luminous efficiency function V(lambda) is the CIE 1924 standard observer, peaking at 555nm. The scotopic function V'(lambda) peaks at 507nm and describes rod response. Both are normalised to 1.0 at their own peak, so the same radiant power at 507nm scores far higher on the scotopic curve than on the photopic one. The lumen is defined only on V(lambda), which is why scotopic lumen is not a unit in the SI chain but still a useful engineering ratio.
Moving between the two curves is the Purkinje shift. A red object that looks bright by day can look near black at dusk, while a blue object gains relative brightness. Anyone who has watched a garden go flat after sunset has seen the photopic system hand over.
The Mesopic Middle
Most real outdoor and industrial levels sit between the two regimes. Roadway, car park and security lighting commonly land between 0.01 and 3 cd/m2, where cones and rods both contribute. CIE 191 defines a mesopic photometry system with an adaptation-dependent weighting, and the weighting changes with the light level itself. That is the awkward part: the same fixture does not convert from photopic to mesopic lumens at a fixed factor.
| Regime | Luminance level | Sensitivity peak | Receptors |
|---|---|---|---|
| Photopic | Above about 3 cd/m2 | 555 nm | Cones |
| Mesopic | 0.01 to 3 cd/m2 | 507 to 555 nm | Cones and rods |
| Scotopic | Below about 0.01 cd/m2 | 507 nm | Rods |
The S/P Ratio and the Numbers It Moves
The scotopic to photopic ratio, shortened to S/P, compares scotopic lumens with photopic lumens for one spectrum. A warm phosphor white in the 2700K to 3000K range measures roughly 1.2 to 1.4. A cool white near 5700K to 6500K commonly reaches 1.8 to 2.2, because it carries more energy near 500nm where rods peak. The figures move with the exact spectrum, so a product claim should be read against the measured spectrum (to be verified per product).
The practical consequence is ranking. Two fixtures with the same photopic lumens can deliver noticeably different scotopic lumens, and at low light levels the one with the higher S/P can read brighter to the eye even though the meter says the same number.
Where This Changes a Decision
For security and perimeter work, spending photons near 480nm to 520nm is efficient for rod vision. Warm sources with low S/P lose on that metric, which is one reason cool white dominates in low-level exterior applications. The counterweight is glare and discomfort: a fixture with a high S/P and a bright, small aperture can be unpleasant at low adaptation levels, and blue-rich light scatters more in fog and rain.
For interior and task lighting the picture reverses, because levels sit well inside the photopic regime and colour rendering dominates. Two checks are worth running before a spec is frozen. First, ask what the actual maintained luminance will be at the task, not the initial lux at the source. Second, if the project is exterior and the levels are low, request spectral power distribution data rather than a single lumen figure.
Key Takeaways
- The lumen is defined on the photopic curve, which peaks at 555nm.
- Rod vision peaks at 507nm and takes over as levels fall, so the ranking of sources can shift at night.
- The mesopic band from 0.01 to 3 cd/m2 is where most exterior lighting lives, and its weighting is level dependent.
- S/P ratio is the practical shortcut for comparing night time performance, but it must be read against a measured spectrum.
- For low level exterior work, ask for maintained luminance and spectral data rather than lumen output alone.
Related reading
