The variety of applications of ultraviolet (UV) light and the consequent need for accurate UV measurements
have increased enormously over the last 20 years. In some cases, the UV radiation from a source is of inter
(e.g., tanning booths and solar radiation). At other times, the action or chemical reaction initiated by UV
irradiation of a system is of interest (e.g., water purification,UV curing, and semiconductor photolithography).
Finally, UV radiation has a cumulative deleterious effect on biological systems; there are consequently health
safety requirements for the accurate measurement of UV radiation.
Considerable effort has been made to produce simple instrumentation to meet these wide-ranging UV
measurement needs. The typical UV meter or radiometer is composed of a number of simple optical elements,
as shown in Fig. 1. The incident radiation passes through an aperture that limits the active area of the system.
diffuser is often placed after the aperture and is used to improve the angular response and spatial uniformity
The instrument. An optical filter is then employed to select the spectral region of the incident optical
radiation that strikes the detector.
To fully understand the accuracy of such a UV meter, the optical properties of its components and the spectral
responsivity should be known as well as the relative spectral distribution of the source. Additionally, the UV
meter will seldom perform ideally, and out-of-band, non-linear, and non-ideal geometric or spatial response
must be characterized to achieve the lowest uncertainties. However, most UV meters are supplied from the
manufacturer with a calibration at a specific wavelength, and only a nominal wavelength band is specified. In
addition, the spectral distribution of the source being measured is often unknown. The purpose of this paper
is to illustrate that considerable thought must be given to the utilization and calibration of these simple
devices order to understand and minimize measurement errors
2. Sources of Error
It is important to define, at the outset, the physical quantity that is to be measured and the level of uncertainty
needed to achieve the measurement goals. The measurement requirements for the UV meter can be
very different: spectrally integrated irradiance (W/cm2) in the UV-A (315 nm to 400 nm) or UV-B (280 nm to
315 nm) regions as in the case of solar irradiation; a single wavelength dose or exposure (J/cm2) as in the
case of semiconductor photolithography; or an effective or weighted dose (Effective J/cm2) as in the case of
biological action spectra. The sources of error in optical radiation measurements described here are not new
to radiometry.These errors in addition to measurement techniques and procedures are well documented in the field of
photometry. However, these topics are less well known in the UV radiation measurement community,
especially among novice users of UV measurement instruments.Due in part to increasing UV applications, recent
publications specifically address UV meter calibration and characterization [2, 3]. In the following, we discuss
common sources of error in UV radiation measurements,including out-of-band contributions to the signal,
non-ideal geometric properties (non-ideal cosine response in the meters), and poor matching to a defined
action spectrum. Other sources of error have been discussed in the literature and will not be discussed here.
These include environmental factors such as temperature and humidity,which can lead to wavelength-dependent
responsivity changes in UV meters. In addition, UV radiation itself induces aging of the optical elements of meters.
Finally, optical detectors used in UV meters have a finite range over which they have an output signal linearly
proportional to the incident irradiance. UV meters should be tested to verify that they are in the linear range both
for the irradiance level used in practice as well as for the smaller levels typically used for calibration.
3. Out-of-Band/Non-Ideal Responsivity
An ideal meter would have a well-defined responsivity within a specific spectral region and zero responsivity
outside of this region. For example, an ideal UV-A meter would have a constant responsivity from 315 nm
to 400 nm and no response outside of this region.Figure 3 shows the spectral responsivity, determined
in monochromatic radiation, of two broadband UV meters used in semiconductor photolithography to
determine the total exposure of a photoresist to 365 nm radiation from a filtered mercury source [4]. These
meters have a maximum responsivity in the 365 nm region, and the responsivity then decreases to a much
smaller, though non-zero, value at longer wavelengths. The instruments demonstrate differing amounts of
increased responsivity in the near infrared (IR), with Meter A showing responsivity 2 to 3 orders of magnitude
larger than Meter B in the 700 nm to 1000 nm spectral region. The increased IR responsivity is due to
increased transmission in the IR by the glass filters, and because silicon photodiodes have their peak response
in the near IR. The increased responsivity observed at wavelengths shorter than 300 nm is caused by fluorescence
of the diffuser, which then re-emits longer wavelength radiation that passes through the filter to the
photodiode. This was verified in Meter A by placing the diffuser between the filter and the photodiode. This
effectively eliminated the responsivity near 275 nm.For monochromatic radiation measurements near
365 nm, the out-of-band response is not important and both meters can make measurements with little error.
Many real optical sources that are assumed monochromatic,such as lasers, often emit radiation
at additional wavelengths. If the source to be measured emits flux at wavelengths below 300 nm or above
680 nm, the 365 nm radiation could be overestimated and measurements with these two meters will
disagree. Although these UV meters were designed to measure monochromatic radiation, they are very similar to UV
meters designed and used for broadband UV radiation.To illustrate these errors, we compare the signal
produced by the two UV meters from four typical sources with different spectral power distributions: a
mercury arc lamp, a quartz-tungsten halogen lamp (ANSI designation, FEL), a deuterium lamp, and a
xenon arc lamp. The relative spectral distribution of each source is shown in Fig. 4.
Using Eq. (1), we compare the integrated in-band irradiance signal with the out-of-band signal. The
in-band signal is the product of the spectral distribution of the source and the meter responsivity, integrated over
the spectral region from 315 nm to 400 nm. The out-ofband response is the integral of the product summed
over the 200 nm to 315 nm and 400 nm to 1000 nm spectral ranges.
