Why could the earlier version's question not be answered?

Because it asked for the measured irradiance at the joint, and a joint is behind a substrate by definition. A radiometer does not fit into a glue line under a cover glass. Faced with an input nobody could supply, the practical response was to enter the surface reading, which silently assumes the substrate is perfectly transparent and produces a cure time that can be many times too short.

Where do I get the absorbance figure?

From the substrate supplier, at the cure wavelength. It is the one input here worth chasing, because the relationship is exponential: absorbance 0.3 passes 50%, 0.6 passes 25%, 1.0 passes 10%. If nobody can give you a figure, a spectrophotometer trace of the actual substrate is a day's work and settles it permanently for that material.

Is clear plastic transparent to UV?

Usually not, and this catches people constantly. Optical clarity is a visible-light property. Standard polycarbonate is strongly absorbing below about 380 nm, ordinary soda-lime glass cuts UVB and UVC almost entirely, and many clear films carry UV stabilisers whose entire purpose is to absorb the wavelengths you are trying to cure with. A part can look like water and pass almost nothing.

What if the cure time comes out impractical?

Attack the transmission rather than the dwell, because transmission is exponential in absorbance while dwell is only linear in time. A photoinitiator that responds at a longer wavelength the substrate passes can change the multiplier by an order of magnitude. So can a different substrate grade, or curing from the other face if the assembly allows. Doubling the lamp is the expensive way to buy what a wavelength change gives free.

Does scattering help or hurt?

It helps, and this page ignores it, so the estimate is conservative for hazy substrates. Absorbed light is gone; scattered light is redirected and much of it still arrives at the bondline. A filled or frosted substrate whose measured absorbance is high may transmit considerably more usefully than Beer-Lambert alone predicts. Where that matters, a real bond and a pull test beats any calculation.

Why is the safety margin larger here than on other pages?

Because you cannot measure the thing that matters. On an open surface a radiometer verifies the process directly; through a substrate the bondline dose is always inferred, and it depends on an absorbance figure that varies between batches and with thickness. The margin is covering an uncertainty you have no way to close by measurement, which is a different situation from covering lamp ageing you can trend.

UV Curing calculator

UV Adhesive Bond Cure Time Calculator (Through-Substrate)

Work out how long a UV adhesive really needs when the light has to get through something first. Enter the dose the adhesive needs at the glue line, the irradiance measured on the outside of the part, the substrate's absorbance at the cure wavelength, and your safety margin. The calculator applies Beer-Lambert to find what actually reaches the bondline and returns the cure time, the transmission, and how many times longer this is than a surface reading would suggest. That multiplier is the number worth looking at: a substrate that passes half the UV doubles the cure time, and nothing about the part looks different.

What this calculator does

Work out how long a UV adhesive needs when the light has to pass through a substrate to reach the bondline, from the irradiance you can actually measure.
Use it for curing an adhesive through a lens, cover glass or plastic housing, explaining why a bond that met its surface dose still failed, choosing between a substrate grade and a longer cure time, sizing cure time for a stack of layers whose absorbances add, deciding whether a photoinitiator change is worth qualifying.
Work out how long a UV adhesive needs when the light has to pass through a substrate to reach the bondline, from the irradiance you can actually measure.

Formula used

Transmission = 10^(−absorbance)
Irradiance at the bondline = incident irradiance × transmission
Bare cure time = bondline dose target ÷ irradiance at the bondline
Cure time with margin = bare cure time × (1 + safety margin ÷ 100)
Times longer than a surface reading implies = 1 ÷ transmission

Inputs explained

Bondline dose target: The dose the adhesive needs AT THE GLUE LINE, from its datasheet, in a named band.
Incident irradiance at the outer surface: What a radiometer reads on the outside of the part that the UV must pass through. This one you can actually measure.
Substrate absorbance at the cure wavelength: Absorbance of everything between the lamp and the glue line, at the cure wavelength. 0 is water-clear, 0.3 passes half, 1.0 passes a tenth. Ask the substrate supplier; do not assume clear plastic is clear in UV.
Depth-of-cure / process safety margin: Headroom over the bare requirement, covering lamp ageing and substrate batch variation.

How to use the result

Best suited to curing an adhesive through a lens, cover glass or plastic housing, explaining why a bond that met its surface dose still failed, choosing between a substrate grade and a longer cure time, sizing cure time for a stack of layers whose absorbances add, deciding whether a photoinitiator change is worth qualifying.
Ignores scattering. A hazy or filled substrate scatters as well as absorbs, and scattered light still reaches the bondline, so this can overstate the loss. Ignores Fresnel reflection at the entry surface, typically a few percent, which makes the estimate slightly optimistic in the other direction. Assumes absorbance is constant with depth and across the part, which fails for a moulded part of varying thickness. Says nothing about oxygen inhibition at the bond perimeter, where the adhesive meets air and can stay tacky whatever the bondline does. Not a bond-strength prediction. Dose delivered is necessary for cure and is not the same as adhesion.

Common questions

Why could the earlier version's question not be answered? Because it asked for the measured irradiance at the joint, and a joint is behind a substrate by definition. A radiometer does not fit into a glue line under a cover glass. Faced with an input nobody could supply, the practical response was to enter the surface reading, which silently assumes the substrate is perfectly transparent and produces a cure time that can be many times too short.
Where do I get the absorbance figure? From the substrate supplier, at the cure wavelength. It is the one input here worth chasing, because the relationship is exponential: absorbance 0.3 passes 50%, 0.6 passes 25%, 1.0 passes 10%. If nobody can give you a figure, a spectrophotometer trace of the actual substrate is a day's work and settles it permanently for that material.
Is clear plastic transparent to UV? Usually not, and this catches people constantly. Optical clarity is a visible-light property. Standard polycarbonate is strongly absorbing below about 380 nm, ordinary soda-lime glass cuts UVB and UVC almost entirely, and many clear films carry UV stabilisers whose entire purpose is to absorb the wavelengths you are trying to cure with. A part can look like water and pass almost nothing.
What if the cure time comes out impractical? Attack the transmission rather than the dwell, because transmission is exponential in absorbance while dwell is only linear in time. A photoinitiator that responds at a longer wavelength the substrate passes can change the multiplier by an order of magnitude. So can a different substrate grade, or curing from the other face if the assembly allows. Doubling the lamp is the expensive way to buy what a wavelength change gives free.
Does scattering help or hurt? It helps, and this page ignores it, so the estimate is conservative for hazy substrates. Absorbed light is gone; scattered light is redirected and much of it still arrives at the bondline. A filled or frosted substrate whose measured absorbance is high may transmit considerably more usefully than Beer-Lambert alone predicts. Where that matters, a real bond and a pull test beats any calculation.

Last reviewed 2026-08-25.