UV Filter Atlas — Spectra & Photodegradation of Sunscreen Active Ingredients

01

Main Things You Need to Know

Eight short answers, FAQ-style. Click the chevron on any card to expand the longer technical version. Read these first to make sense of the rest of the atlas — especially if some of the chart numbers seem surprising.

F·01 — what reaches your skin

How much of solar UV gets to ground level, and why does the small UVB fraction matter so much?

~95% of solar UV at sea level is UVA (320–400 nm); only ~5% is UVB (290–320 nm). UVC (<280 nm) is fully blocked by atmospheric ozone. The 5% UVB punches above its weight because the erythema action spectrum peaks at 297 nm.

▸

Solar irradiance at the top of the atmosphere extends from extreme UVC through near-IR, but ozone removes essentially all UVC and ~70% of UVB before it reaches sea level. UVA passes through nearly unattenuated.

At noon equator under clear sky (AM 1.5), total solar UV is roughly 30–40 W/m²: UVB ~1–2 W/m², UVA ~28–38 W/m², visible ~500 W/m², near-IR ~700 W/m². So in absolute energy terms, visible and IR dominate the solar load; UV is a small slice.

Biological relevance comes from action spectra — different wavelengths have very different per-photon damage. The CIE 1987 erythema action spectrum (the FDA SPF standard) peaks at 297 nm and drops by ~10× every 30 nm into UVA. So the small UVB fraction does most of the direct DNA damage, even though UVA is far more abundant.

Solar irradiance at sea level (AM 1.5) CIE 1987 erythema action spectrum (×10⁵, normalized)

The action spectrum × solar irradiance product peaks ~305 nm — explaining why erythemal SPF is dominated by 290–315 nm even though UVA is far more energetically abundant.

F·02 — biological effect

What does each UV band actually do to skin?

UVB damages DNA directly (sunburn → mutation → skin cancer). UVA-2 (320–340 nm) drives immediate pigment darkening and indirect DNA damage via reactive oxygen species. UVA-1 (340–400 nm) penetrates deepest and causes photoaging — collagen breakdown, elastosis. HEV blue light (400–500 nm) contributes to melasma, especially in skin of color.

▸

Penetration depth: UVB is ~70% absorbed by the stratum corneum; UVA-2 reaches the basal epidermis; UVA-1 penetrates 1–2 mm into the dermis where collagen and elastin fibers live; HEV passes through into the dermis as well.

Effects by band, at a glance

Band	Burning	Photoaging	Skin cancer	Pigmentation
UVB 290–315 nm	PRIMARY	minor (epidermal only)	PRIMARY (direct DNA damage)	delayed tan (2–3 days)
UVA-2 315–340 nm	minor (~1/100 UVB potency)	moderate	indirect via ROS	immediate pigment darkening
UVA-1 340–400 nm	minimal direct erythema	PRIMARY (dermal collagen breakdown)	indirect; accumulates with chronic exposure	persistent darkening; melasma driver
HEV 400–500 nm	delayed only (skin of color, Mahmoud 2010)	emerging evidence (ROS pathway)	minor	strong driver in Fitzpatrick IV–VI

Burning (erythema)

Sunburn is dominated by UVB. The CIE 1987 erythema action spectrum peaks at 297 nm and drops by ~10× per 30 nm into UVA. So even though UVA is ~20× more abundant than UVB at sea level, UVB does ~80–90% of the erythemal work. UVA-1 alone struggles to produce visible erythema in fair skin even at high doses. SPF testing weights the UV dose by this action spectrum, which is why SPF essentially measures UVB protection.

Photoaging (wrinkles, elastosis, leathery texture)

Photoaging is dominated by UVA-1. UVB stays mostly in the epidermis (the upper 0.1 mm, dead and constantly turning over); UVA-1 penetrates to the dermis (1–2 mm) where collagen and elastin live, and where you can't replace damaged structural proteins easily. UVA-1 generates singlet oxygen and superoxide, which activates matrix metalloproteinase-1 (MMP-1, collagenase) and elastase, gradually digesting the extracellular matrix. The clinical signature is solar elastosis — the leathery, deeply wrinkled appearance of chronically sun-exposed skin in lighter complexions.

Practical implication: SPF doesn't tell you about photoaging. A high-SPF sunscreen with poor UVA-1 coverage (bare-bones US formulations with low avobenzone, no Mexoryl 400 / Tinosorb / DHHB) will let through most of the photoaging-relevant UV even when it stops sunburn. The UVA-PF rating (PA+ to PA++++) and FDA "broad spectrum" critical-wavelength claim exist specifically to distinguish anti-burn from anti-aging coverage.

Skin cancer

UVB is the main driver of basal-cell and squamous-cell carcinomas. Direct UVB absorption by DNA produces cyclobutane pyrimidine dimers (CPDs) and 6-4 photoproducts at adjacent pyrimidines, and the C→T transition at TT or CC dipyrimidine sites is the well-known UVB mutational signature seen in tumor sequencing. UVA contributes via indirect oxidative DNA damage (8-oxoguanine), and chronic UVA exposure is increasingly implicated in melanoma — UVA penetrates to the basal melanocyte layer and accumulates over decades. Most epidemiological evidence still points to UVB as the dominant carcinogenic component, but UVA's contribution to melanoma risk is non-zero and is the reason "broad-spectrum" claims have grown stricter over time.

Pigmentation and melasma (especially in skin of color)

This is where HEV (visible) light becomes clinically important — and it's a relatively recent realisation in dermatology. Three pigmentation responses with different drivers:

Immediate pigment darkening (IPD): appears within minutes of exposure, fades within an hour. Caused mostly by UVA-2 (315–340 nm). Photo-oxidation of pre-formed melanin.
Persistent pigment darkening (PPD): appears within hours, persists for days. Driven by UVA-1 (340–400 nm). The basis of the UVA-PF / PA-rating test method.
Delayed tanning + melasma: develops over days to weeks; sustained pigmentation. Driven by UVB (delayed tan) and HEV/visible (melasma in Fitzpatrick IV–VI). Mahmoud 2010 (J Invest Dermatol 130:2092) was the canonical demonstration that visible light alone induces sustained erythema and pigmentation in dark skin types but not in fair ones — which is why tinted mineral sunscreens with iron oxides became the recommended approach for melasma in skin of color (Castanedo-Cazares 2014).

The action spectrum for melanogenesis extends to ~440 nm — well into visible light, beyond what any UV filter alone can block. This is the case for tinted sunscreens specifically: they do something pure-UV sunscreens cannot.

Vitamin D — the inverse case

Vitamin D synthesis peaks around 295–300 nm UVB — almost exactly where the erythema action spectrum peaks, and exactly where SPF blocks hardest. Hence the perennial concern: high-SPF sunscreens block the same UVB that makes vitamin D. In practice, a few minutes of incidental sun exposure on incidental skin (hands, face during a walk) is enough for most people; sunscreen use does not reliably cause vitamin D deficiency in healthy populations who get any outdoor exposure at all. But it's worth knowing the chemistry maps onto a real biological tradeoff.

Bottom line for sunscreen choice

If your concern is burning + skin cancer, SPF and UVB coverage matter most — most US-monograph organic stacks do this well. If your concern is photoaging, look for high UVA-PF and ideally Mexoryl 400 or Tinosorb-class coverage extending to 380–400 nm — newer EU/JP formulations dominate here. If your concern is melasma or post-inflammatory hyperpigmentation in darker skin, you also need a tinted formulation (iron oxides) — pure UV sunscreens, no matter how high the SPF, won't address visible-light-driven pigmentation.

F·03 — what SPF measures

What does an "SPF 50" label actually represent?

SPF is the ratio of minimum erythema dose on protected vs unprotected skin. Because erythema action peaks in UVB, SPF dominantly measures UVB protection. UVA-PF is the parallel UVA metric (PA+/++/+++/++++ in Japan). FDA "broad spectrum" requires critical wavelength ≥370 nm.

▸

Test method: a panel of subjects, sunscreen applied at 2 mg/cm², exposed to a calibrated UV source. The dose at which erythema appears 16–24 h later, divided by the same dose unprotected, gives SPF. Higher SPF = bigger ratio.

Diminishing returns: SPF 30 blocks ~97%, SPF 50 blocks ~98%, SPF 100 blocks ~99% of erythemal UV. The marginal gain from "50 → 100" is only 1 percentage point at the standard test thickness — but that 1 pp doubles your safe outdoor time at peak sun if you've applied perfectly.

UVA-PF (PPD method) is the same kind of ratio for UVA-induced persistent pigment darkening. Japan's PA system: PA+ ≈ UVA-PF 2–4, PA++ ≈ 4–8, PA+++ ≈ 8–16, PA++++ ≥ 16. EU labels broad-spectrum if UVA-PF ≥ SPF/3.

FDA "broad spectrum" uses the critical wavelength instead — the wavelength below which 90% of the absorbance area lies. ≥370 nm earns the label. The Spectrum Builder shows this metric live.

F·04 — chart units

What does the chart's Y-axis number mean (and not mean)?

E(1%, 1cm) is the absorbance you'd measure in a 1 cm spectrophotometer cuvette at 1% w/w concentration — the cosmetics-industry-standard reporting metric. Real cosmetic films are ~14 µm thick, 500× thinner, so the chart's "10⁵" peaks correspond to real-film absorbances of ~10² ≈ 99% blockage at λmax. Switch the Spectrum Builder to "Lab test" or "Real-world" mode for direct % transmittance.

▸

Beer-Lambert: A = ε × c × ℓ. In cosmetics convention, E(1%, 1cm) = ε × 10/MW. So if you see ε = 35,000 M⁻¹cm⁻¹ for avobenzone (MW 310), you can convert: E1% = 35,000 × 10 / 310 = 1,130. At 3% in a 1 cm cuvette, you'd measure A = 1,130 × 3 × 1 = 3,390 (off-scale; in practice diluted further to ~0.001% to fit a real instrument).

Real cosmetic film thickness from a 2 mg/cm² application at density ~1 g/cm³: ℓ = 2e-3 g·cm⁻² / 1 g·cm⁻³ = 2e-3 cm = 14 µm (about a third the thickness of household cling film). Multiply chart values by 0.002 to get real film absorbance.

For inorganics (ZnO/TiO₂/iron oxides), there's no meaningful molarity because they're particles, not dissolved species. The atlas uses a calibrated effective E1% derived from real published mass extinction k(λ), with values like ZnO peak ≈ 160 (vs Croda patent 12–20 L/g/cm × 10 = 120–200, central 160).

The peak chart value isn't literally what you'd measure — it's the linear-stacking prediction. Particles aggregate at high loading (the f(c) factor in the Spectrum Builder corrects for this), and skin-film roughness adds another ~30% attenuation loss. The dropdown methodology in the Spectrum Builder section walks through every correction.

F·05 — application gap

Why do real users get less protection than the SPF on the label?

The FDA SPF test applies 2 mg/cm² of formulation. Most consumers apply 0.5–1 mg/cm² (a quarter to half), and Beer-Lambert is linear in path length. Realised SPF is roughly 25–50% of the labeled value. SPF 50 at half-thickness ≈ effective SPF 12–18.

▸

2 mg/cm² is the dose used in clinical SPF testing — it's about 35 ml of sunscreen for a full body. Empirical measurements (Petersen & Wulf 2014; Cole 2010) consistently find consumers apply 0.5–1 mg/cm² in real conditions. The "more reapplication, not higher SPF" mantra is grounded in this gap.

Even coverage matters too. Sunscreen that pools in finger ridges and skin creases doesn't protect the bare patches. Spray formats look like they cover everything but a recent meta-analysis showed actual coverage is often spotty.

Reapplication: every 2 hours of sun is the FDA recommendation. In water/sweat, reapply after 40–80 minutes regardless of "water-resistant" claims. Real-world SPF decay is dominated by sweat dilution and physical removal, not by photodegradation of the active filters (which most users would assume from sun exposure).

The Spectrum Builder's "Real-world" film mode (0.75 mg/cm²) captures this — switching from Lab to Real-world visibly raises the transmittance curves because the film is 2.7× thinner.

F·06 — physical vs chemical

Is "mineral sunscreen" really physical, or is that a marketing story?

Both organic and inorganic UV filters work primarily by absorption. Cole et al. 2016 measured only 4–5% UV reflection from ZnO/TiO₂ films — the other 95% is absorbed via electronic band-gap transitions (semiconductors) or π→π* / n→π* transitions (organic chromophores). The "mineral reflects, chemical absorbs" framing is a marketing convenience, not physics.

▸

Inorganic UV filters (ZnO, TiO₂) are wide-bandgap semiconductors. Below their bandgap energy (Eg ~ 3 eV for both), they're transparent. Above the bandgap, electrons get excited from the valence to conduction band, dissipating the energy as heat. ZnO bandgap 3.37 eV → cutoff 368 nm; rutile TiO₂ 3.0 eV → 413 nm. Mie scattering by the particles contributes a few percent of total attenuation but is small.

Organic filters are small molecules with conjugated π systems. UV photons promote electrons across HOMO-LUMO gaps; the excited state relaxes to ground state via internal conversion (heat) — usually within 1 picosecond for ESIPT-class filters (most photostable ones), much slower for the photolabile cinnamate/dibenzoylmethane filters that need stabilization.

The real practical distinction isn't "physical vs chemical" but: (a) skin penetration — small organic filters can penetrate (and Matta JAMA 2019/2020 showed plasma levels exceed FDA toxicology thresholds at typical use), well-coated nano-inorganic generally don't; (b) photostability — most newer organics are photostable, avobenzone is famously not, and inorganic crystal lattices are essentially indestructible; (c) aesthetic feel — high-refractive-index inorganics (especially TiO₂ at n=2.7) cause visible white cast at high loading; organics don't.

F·07 — iron oxides for HEV

Why do iron oxides show up in tinted sunscreens, if they're not UV filters?

Iron oxides (CI 77491/77492/77499) are FDA color additives, not regulated UV filters. They appear in tinted mineral sunscreens because their bandgaps fall inside the visible spectrum (Fe₂O₃ 2.1 eV, FeOOH 2.5 eV). 1–3% iron-oxide loading blocks 60–70% of HEV (400–500 nm) vs 10–20% untinted — clinically relevant for melasma, post-inflammatory hyperpigmentation, and idiopathic photodermatoses.

▸

Three pigments are typically blended: red hematite (α-Fe₂O₃, CI 77491, bandgap 2.1 eV → cutoff ~590 nm), yellow goethite (α-FeOOH, CI 77492, 2.5 eV → ~496 nm), and black magnetite (Fe₃O₄, CI 77499, semi-metallic broadband absorber). A 3:1 yellow:red blend at ~3% gives a skin-tone-neutral tint that absorbs across UV plus the entire 400–500 nm HEV band.

Why this matters: the visible-light action spectrum for melanogenesis (Mahmoud 2010, Lyons 2021) extends to ~440 nm, and is sustained much longer in Fitzpatrick IV–VI. Castanedo-Cazares 2014 demonstrated tinted sunscreens prevent melasma recurrence better than non-tinted ones in matched trials. For idiopathic photodermatoses with VL action spectrum (chronic actinic dermatitis, polymorphous light eruption in some patients), tinted formulations are first-line.

Iron oxides are the only widely-available cosmetic ingredient that delivers measurable visible-light protection. Pigmentary (≥150 nm) TiO₂ does too, but at the cost of substantial white cast.

F·08 — photodegradation

Why does avobenzone need a "stabilizer" — and what's the catch with the most common one?

Avobenzone — the only FDA-approved UVA-I filter — degrades ~36% in 1 hour of sun unless stabilized, typically by octocrylene. Octocrylene then decomposes to benzophenone (an IARC 2B carcinogen) over shelf life. Most newer EU/JP filters (Tinosorb M/S, DHHB, Mexoryl SX/XL/400, EHT) are intrinsically photostable via ESIPT (excited-state intramolecular proton transfer).

▸

Avobenzone exists in two tautomers: enol (357 nm UV-A peak, photostable in protic solvent like methanol) and keto (265 nm UV-B peak, photoreactive). The diketo triplet state undergoes [2+2] cycloaddition with the cinnamate alkene of octinoxate, irreversibly destroying both filters — which is why avobenzone + octinoxate is a known formulation incompatibility.

Octocrylene rescues avobenzone via triplet-triplet energy transfer (TTET): it absorbs the diketo triplet's excited-state energy and dissipates it harmlessly. But Downs et al. 2021 (Chem Res Toxicol) showed octocrylene undergoes retro-aldol decomposition over storage to produce free benzophenone — measured at 6–186 mg/kg in fresh products and up to 435 mg/kg after FDA accelerated aging.

The structural problem: the FDA OTC sunscreen monograph has been frozen since 1999. Newer photostable filters (Tinosorb S/M, DHHB, Mexoryl 400) are EU/AU/JP approved with decades of safety data but stuck in TEA pending. The December 2025 FDA proposed order on bemotrizinol at 6% is the first crack in the logjam.

ESIPT-class filters dissipate UV in <1 picosecond — they convert UV photons to heat via a proton hop along an intramolecular hydrogen bond, then back to ground state, all before any chemistry can happen. Tinosorb M/S, DHHB, Mexoryl XL, the benzophenones (BP-3, BP-4, BP-8) all use this mechanism. They are essentially indestructible by sunlight at typical formulation loading.

02

Spectral Coverage Map

Each bar is one filter's approximate absorption band — Gaussian-shaped, peak opacity at λmax, fading toward the wings. Sorted by peak wavelength. Iron oxides extend deep into visible (HEV / blue light through near-IR).

UVB-dominant UVA-II / dual UVA-I anchor Inorganic semiconductor Iron oxide (HEV / visible)

03

λmax × Photostability — At a Glance

Each filter is rendered as a horizontal smear bar — its width and Gaussian fade represent its absorption band, with peak opacity at λmax. Vertical position is half-life under solar UV (log scale). Hover any filter to see the uncertainty range: the half-life is highly conditions-dependent, and the dotted bracket shows the range across vehicles and dose conditions for that filter's confidence tier. Inorganics and iron oxides sit on the dashed "∞" line at top — their lattice does not photodegrade.

UVB-dominant organic UVA-II / dual UVA-I anchor Broad-spectrum next-gen Inorganic (ZnO/TiO₂ — off-scale) Iron oxides (visible/HEV; off-scale)

04

Spectrum Builder — How Filters Combine

Real sunscreens stack 4–7 filters at different concentrations to fill the gap between UVB and UVA-I. Pick a preset or click filters in the legend to add them; drag the sliders to change concentrations.

Math, calibration, and known limitations →

How the curves and metrics are computed

Each filter contributes E(1%, 1cm)(λ) × f(c) × concentration% at every wavelength. Beer-Lambert says absorbances add, so the bold curve is the sum across all selected filters. E(1%, 1cm) is the cosmetics-industry-standard metric — the absorbance of a 1% w/w solution at 1 cm path length, as reported in BASF/DSM/Symrise/3V Sigma datasheets.

Concentration nonlinearity f(c)

Inorganics (ZnO, TiO₂, iron oxides): per-filter f(c) = 1 − α·tanh((c−c₀)/scale), with parameters fit to real SPF and HEV-blockade measurements. ZnO α=0.70 / c₀=2 / scale=10 (validated 22% mean SPF error on 9 mineral formulations). TiO₂ α=0.62 / scale=8. Iron oxides saturate fastest (scale 1.5–4%) — pigment-grade particles opacify films at lower loading.

Organics: default f(c) α=0.92 / c₀=0.5 / scale=2.5. Captures self-quenching at high concentration, partial dissolution / micelle formation, and the cuvette-to-real-film discrepancy. Calibrated against Sayre 2005 in-vitro SPFs of mono-filter formulations: 5% octinoxate predicted SPF 6 vs measured 5–8; 5% bemotrizinol predicted 9 vs 7–12.

SPF estimator

SPF = ∫E_solar(λ)·E_ery(λ)·dλ / ∫E_solar(λ)·E_ery(λ)·10^−A(λ)·dλ, integrated over 290–400 nm. E_ery is the CIE 1987 erythema action spectrum (the FDA SPF testing standard). E_solar is AM 1.5 sea-level solar irradiance. The same integration restricted to 320–400 nm gives UVA-PF; "critical λ" is the wavelength below which 90% of the absorbance area sits.

Known limitations

1) Mineral SPFs are over-predicted by ~10×. The CIE 1987 action spectrum drops sharply past 340 nm, so mineral filters' soft band-edge leak past 380 nm doesn't hurt SPF in the math even though it does in real cosmetic films (where particle aggregation, film roughness, and thickness inhomogeneity matter). Capped at "100+" on display.

2) Stacked organic formulations are sometimes under-predicted by ~3×. Per-filter f(c) double-counts when many filters share the total chromophore load. The "US classic" preset (25% total organic load across 4 filters) lands at SPF ~12 in the math vs ~30–50 measured.

3) Real-world application is 0.5–1 mg/cm², half to a quarter of the FDA-standard 2 mg/cm² test thickness. Switch the film mode below to see the difference (Beer-Lambert is linear in path length).

4) The chart's Y-axis literal value (E(1%,1cm) × concentration%) is what you'd measure in a 1 cm spectrophotometer cuvette at the displayed concentration. Real cosmetic films are ~14 µm thick (500× thinner), so multiply by 0.002 to get film-level absorbance.

Best read as a relative comparison of formulations, not an exact SPF predictor.

Presets 📄 📄

Learn more: How important is UVA-1? What about HEV / visible light? What does this Y-axis really mean? Real-world vs lab SPF?

Film: methodologySPF integrates erythemal-weighted UV transmittance per CIE 1987: SPF = ∫E_solar(λ)·E_ery(λ)dλ / ∫E_solar·E_ery·10^(-A(λ))dλ. Film absorbance A(λ) = Σ E1%(λ)·f(c)·c·ℓ_cm summed across selected filters. Per-filter f(c) damping is fit to real SPF/HEV measurements: ZnO 22% Badger predicted 47 vs measured 40, etc. Limitations: (1) The CIE 1987 action spectrum understates UVA-1 sensitivity, so mineral sunscreens (which leak past 370 nm) are over-predicted. (2) Stacked organic formulations are sometimes under-predicted because per-filter f(c) double-counts when many filters share the total load. (3) Real SPF testing has a measurement ceiling around 100; "100+" should be treated as "very high" rather than a precise number. (4) Real-world application of 0.5–1 mg/cm² gives roughly 25–50% of lab SPF. The chart is best read as a structured comparison of formulations, not an exact SPF predictor.

Add another filter…

05

Filter Directory

Click any card for full citations, photodegradation data, and per-region regulatory limits. Use search and chips to filter.

36 of 36 filters

Range Photostability Region

06

Photodegradation Kinetics — Photolabile Filters

Quantitative loss data with primary peer-reviewed citations. Solvent and UV dose are stated because the same molecule can be photolabile in cyclohexane and photostable in methanol.

Filter	Loss	Conditions	Photoproducts	Source
Avobenzone (alone)	~36% loss / 1 h sun	Neat, no stabiliser	14 species — arylglyoxals, benzils, benzoic-acid derivatives, dibenzoylmethane radical	Schwack & Rudolph 1995, J Photochem Photobiol B 28:229
Avobenzone (4%)	~77% loss / 25 MED	Film, no stabiliser	same as above	Bonda 2008 (Sunscreens, CRC Press)
Avobenzone (cyclohexane)	"appreciable"	Aprotic solvent	photoiso. + degradation	Mturi & Martincigh 2008, J Photochem Photobiol A 200:410
Avobenzone (methanol)	"essentially photostable"	Protic — H-bond stabilises enol	none significant	Mturi & Martincigh 2008
Octinoxate (OMC)	~10% loss / 35 J·cm⁻² UVA	EtOH	trans → cis isomer; dimers	Tarras-Wahlberg et al. 1999, J Invest Dermatol 113:547
Octinoxate + Avobenzone	Both filters lost faster	Film	[2+2] / Paternò–Büchi cycloadduct; ESR-detectable radicals	Sayre, Dowdy, Gerwig, Shields, Lloyd 2005, Photochem Photobiol Sci 4:699
PABA	photoallergenic + mutagenic photoproducts	Aqueous / EtOH	4-aminobenzaldehyde + radicals	Knowland et al. 1993, Mutat Res 304:39
Octocrylene → Benzophenone	avg 39 mg/kg fresh; 75 mg/kg accelerated; max 435	Retro-aldol over shelf life	benzophenone (IARC 2B)	Downs et al. 2021, Chem Res Toxicol 34:1046

07

Photostabilization — How Avobenzone Survives

Real sunscreens are mixtures. The single most important formulation question is keeping avobenzone alive, since it is the only FDA-approved UVA-I filter.

08

Cross-Cutting Findings

What emerged from comparing the data across all 30 filters.

F·01

"Physical" vs "chemical" sunscreen is largely a marketing fiction. Cole, Shyr & Ou-Yang (2016) measured only 4–5% UV reflection from ZnO/TiO₂. Both inorganic and organic filters absorb UV — the difference is that semiconductors absorb via electronic-band-gap transitions while organics absorb via π→π* transitions.Cole et al. 2016, Photodermatol Photoimmunol Photomed 32:5

F·02

The avobenzone–octocrylene–benzophenone trilemma is a structural consequence of the FDA monograph freeze. US sunscreens needing UVA-I have effectively one filter (avobenzone), which is photolabile, so most are stabilised with octocrylene, which decomposes to benzophenone (IARC 2B) over shelf life. The trilemma disappears once next-generation filters (DHHB, Tinosorb M/S) are FDA-approved.

F·03

ESIPT is the dominant photostability mechanism. Most photostable filters share one trick: excited-state intramolecular proton transfer in <1 picosecond — benzophenones, Tinosorb S/M, DHHB, Mexoryl XL all use it. Avobenzone is the conspicuous exception: its diketo tautomer is photoreactive instead of dissipating to ground state. This is why avobenzone alone fails and ESIPT-class filters succeed.

F·04

The deep UVA-I band (370–400 nm) was poorly covered until Mexoryl 400. Approved EU via Regulation 2020/1684 based on SCCS/1605/19 (Dec 2019), broadly commercialised 2022–2024 as UVMune 400, Mexoryl 400 with λmax 385 nm closes a long-standing gap implicated in photoaging and pigmentary disorders. ZnO and TBPT (with its scattering tail) help partially. This band remains the weakest spot in current US-approved formulae specifically because Mexoryl 400 has no FDA path.

F·05

Solvent matters more than is usually reported. Avobenzone's λmax shifts from 355 nm (cyclohexane) to 363 nm (DMSO), and its photostability ranges from "appreciable degradation" to "essentially photostable" across the same solvent set. Any photostability claim without a stated vehicle is uninterpretable.Mturi & Martincigh 2008, J Photochem Photobiol A 200:410

F·06

The next-generation filters are EU/AU/JP-approved but mostly stuck in FDA limbo. Tinosorb M/S, DHHB, Mexoryl XL/SX/400, EHT, iscotrizinol — all have decades of safety data and are higher MW (no skin penetration) and more photostable. The Dec 2025 FDA proposed order on bemotrizinol (6% GRASE) is the first crack in the logjam after the 2014 Sunscreen Innovation Act produced no approvals through 2024.

F·07

Iron oxides are the only practical visible-light protectant in cosmetics — but they're regulated as colorants (FDA 21 CFR 73.2250; EU Annex IV), not as UV filters. With bandgaps inside the visible spectrum (Fe₂O₃ 2.1 eV, FeO(OH) 2.5 eV, Fe₃O₄ semi-metallic), they absorb HEV (400–500 nm) and beyond. A typical 3% iron-oxide blend in a tinted mineral sunscreen blocks 60–70% of blue light vs 10–20% for the same untinted base — clinically meaningful for melasma, PIH, and idiopathic photodermatoses with visible-light action spectra (Castanedo-Cazares 2014, Mahmoud 2010, Lyons 2021). The visible-light action spectrum for pigmentary disorders is the reason "tinted mineral sunscreen" is now a dermatological category, not just an aesthetic one.

09

Methodology & Provenance

How this was researched and where to verify.

This compendium was researched by AI agents using web search and direct fetches of peer-reviewed literature, regulatory documents (FDA Federal Register, EU Annex VI, SCCS opinions), and manufacturer technical bulletins. Six parallel deep-research subagents produced 37 individual claim files covering the six filter categories, then a synthesis pass produced the master comparison. Every numerical value (λmax, ε, photodegradation rate, regulatory cap) traces to a primary source with author, journal, year, and DOI; values are tagged with confidence tier C1 (verified primary) through C5 (not located).

Confidence tiers

C1 — directly stated in peer-reviewed paper or pharmacopoeia
C2 — credible secondary source (review citing primary)
C3 — derived/calculated from public data (working shown)
C4 — graph-digitised or estimated
C5 — not located

About this atlas

This is an AI-generated reference compendium. Background on how it was built — and on the broader experiment of sharing Claude Code outputs publicly — is at blog.sus.cat / share-your-claude-code-outputs. The 66 source documents (claim files, verification docs, calibration deep-dives, preset-source documentation) are pre-rendered to static HTML and accessible directly: MASTER-SYNTHESIS, SkinAqua source, UVMune 400 source, and the per-category summaries linked from the master file. Each page links to its raw .md source for verification.

AI disclosure

Researched, synthesized, and rendered by AI agents. Primarily AI fact-checked, not personally verified by the author. Citations were spot-checked against PubMed and publisher pages — 2 misattributed citations were corrected in an adversarial critique pass (Stange→Hanson; Hamzavi→D'Ruiz). Numerical values, regulatory statuses, and SPF predictions should be re-verified against primary sources before any clinical, regulatory, or product-formulation use.

Limitations — model and math

Beer-Lambert assumes uniform homogeneous solution; real cosmetic films have particle aggregation, film-thickness inhomogeneity, finger-ridge gaps
Concentration nonlinearity f(c) is per-filter; double-counts when many organics share the load (US classic stack under-predicted ~3×); may under-correct for very high inorganic loadings
Spectrum curves are linear interpolation between literature data points (with exponential 4 nm decay past the data range); could miss sub-nm absorption features
No explicit Mie scattering correction for inorganics — assumed lumped into the calibrated effective E1%
Photodegradation kinetics shown in §06 are simplified single-pass first-order — no formulation matrix effects, no oxygen dependence, no wavelength-resolved photoproduct branching
Particle size correction for inorganics is a single multiplier; doesn't capture full size distribution effects

Limitations — SPF estimator

CIE 1987 erythema action spectrum is the FDA standard, but understates UVA-1 sensitivity in some skin types; mineral sunscreens over-predicted ~10× because the soft TiO₂ band-edge leak past 380 nm doesn't hurt SPF in this metric even though it does on real skin
AM 1.5 solar spectrum is a single approximation; real solar varies 10× by latitude / season / time of day
HEV blockade computation uses flat solar irradiance in 400–500 nm (good enough for relative comparisons)
Display capped at "SPF 100+" — clinical SPF testing has a similar practical ceiling around 100
UVA-PF computation uses an unweighted UVA integral, not the full PPD action spectrum
Critical wavelength uses 1 nm sampling; sub-nm precision not meaningful

Limitations — data quality and sourcing

Mexoryl 400 spectral data all traces back to L'Oréal/BASF sponsor measurements (Marionnet 2021 + SCCS/1605/19) — no independent academic ε determination published as of May 2026
Iscotrizinol Φ_T and Φ_Δ values not located; Bisoctrizole ε mostly manufacturer-derived (BASF Tinosorb M datasheet)
Cinoxate, Padimate O — original ε measurement papers not retrieved during audit; values cited in secondary literature only
Mexenone (BP-10) and BCSA (Mexoryl SD): no modern primary spectra located — values held by structural analogy (C5 confidence)
Iron oxide effective E1% calibrated against ~10 HEV-blockade anchor points (Mahmoud 2010, Ruvolo 2018, Castanedo-Cazares 2014, Lyons 2021, Dumbuya 2020)
ZnO concentration-nonlinearity parameters fit to 9 mineral formulations (Badger, ThinkSport, Aveeno + industry benchmarks); 22% mean SPF error on validation set
Coral-reef toxicity for benzophenones other than BP-3 — analogous Downs-style assays not located

Limitations — scope

36 filters covered; some Korea- or China-only modern filters not included
Photostabilizer interactions covered in §07 but not factored into the SPF math
After-sun repair products, antioxidant boosters (Vitamin E, C, ferulic acid), and DNA-repair enzymes not modeled
Filter interactions with skincare actives (retinol, AHAs, vitamin C) not addressed
Octocrylene → benzophenone shelf-life kinetics (Downs 2021) noted but not implemented in the SPF model — predicted SPF doesn't degrade with age
DHHB's DnHexP phthalate impurity noted but not explored in depth
Half-life values are conditions-dependent; a single value per filter is an approximation

Limitations — visualization

Y-axis E(1%, 1cm) requires mental conversion to film absorbance (multiply by 0.002 for a 2 mg/cm² film); the Spectrum Builder offers Lab / Real-world modes for direct transmittance
Log scale on most charts can mislead users unfamiliar with logarithmic axes
Color choices not formally tested for full color-blind safety; the UV-band palette uses violet/pink/amber which may collapse for some forms of dichromacy
Modal SVGs use a single x-axis range per filter; comparing two filters with different ranges requires opening each modal separately
Mobile experience tested at 375 px width but very narrow viewports (<320 px) may have minor layout issues

Limitations — real-world translation

Application thickness varies widely in real users (0.3–1.5 mg/cm²) vs the FDA-standard 2 mg/cm² test; switch the Spectrum Builder to "Real-world" mode for the typical-use case
Coverage is rarely 100% even with diligent application — finger-ridge gaps reduce effective protection
Sweat / water resistance varies by formulation; not modeled (use the brand's water-resistance claim as the practical guide)
Reapplication kinetics (every 2 h, or after sweating/swimming) not modeled — the chart assumes a fresh single application
Skin type (Fitzpatrick I–VI) affects baseline MED but not modeled; visible-light damage in particular is more relevant for IV–VI
Geographic / seasonal UV exposure varies by 10× and not factored — the chart assumes mid-latitude noon clear-sky AM 1.5
Indoor light through windows: UVB blocked, UVA-1 partially transmitted (not modeled)
Product expiration / shelf life reduces filter potency over time, especially for octocrylene-stabilized formulations (Downs 2021 retro-aldol decomposition)

Limitations — regulatory currency

Regulatory status is a snapshot of May 2026 — jurisdictions change rapidly
FDA Dec 2025 bemotrizinol Proposed Order at 6% GRASE is recent and not yet final (expected June 2026)
Reef-ban legislation (Hawaii Act 104, Florida Keys, Palau, USVI, Aruba, Bonaire) may evolve
EU 4-MBC ban effective May 2026 — formulations may still be in distribution channel as of the atlas date
Asia-Pacific markets (Thailand, Vietnam, Indonesia) less thoroughly covered than US/EU/JP/KR/AU