ARTI-21CR/610-40030-01

DEFINING THE EFFECTIVENESS OF UV LAMPS INSTALLED IN CIRCULATING AIR

DUCTWORK

Final Report

Date Published November 2002

Douglas VanOsdell and Karin Foarde

RTI
P.O. Box 12194
Research Triangle Park, NC 27709-2194

Prepared for the

AIR-CONDITIONING AND REFRIGERATION TECHNOLOGY INSTITUTE
4100 N. Fairfax Drive, Suite 200, Arlington, Virginia 22203

Distribution A Approved for public release; further dissemination unlimited.

 DISCLAIMER

This report was prepared as an account of work sponsored by the Air-Conditioning and
Refrigeration Technology Institute (ARTI) under its HVAC&R Research for the 21st Century
(21-CR) program. Neither ARTI, the financial supporters of the 21-CR program, or any agency
thereof, nor any of their employees, contractors, subcontractors or employees thereof - makes
any warranty, expressed or implied; assumes any legal liability or responsibility for the accuracy,
completeness, any third partys use of, or the results of such use of any information, apparatus,
product, or process disclosed in this report; or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial product, process, or service by trade
name, trademark, manufacturer, or otherwise, does not necessarily constitute nor imply its
endorsement, recommendation, or favoring by ARTI, its sponsors, or any agency thereof or their
contractors or subcontractors. The views and opinions of authors expressed herein do not
necessarily state or reflect those of ARTI, the 21-CR program sponsors, or any agency thereof.

Funding for the 21-CR program provided by (listed in order of support magnitude):
- U.S. Department of Energy (DOE Cooperative Agreement No. DE-FC05-99OR22674)
- Air-Conditioning & Refrigeration Institute (ARI)
- Copper Development Association (CDA)
- New York State Energy Research and Development Authority (NYSERDA)
- Refrigeration Service Engineers Society (RSES)
- Heating, Refrigeration and Air Conditioning Institute of Canada (HRAI)

Available to the public from

U.S. Department of Commerce

National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
(703) 487-4650

Available to U.S. Department of Energy and its contractors in paper from

U.S. Department of Energy

Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831
(423) 576-8401
ARTI-21CR/610-40030-01

DEFINING THE EFFECTIVENESS OF UV LAMPS

INSTALLED IN CIRCULATING AIR DUCTWORK

Final Report

Date Published November 2002

Douglas VanOsdell
Karin Foarde

Prepared for the

AIR-CONDITIONING AND REFRIGERATION TECHNOLOGY INSTITUTE
Under ARTI 21-CR Program Contract Number 610-40030

i
 EXECUTIVE SUMMARY

Germicidal ultraviolet (UVGI) lamps have a long history of use for inactivating microbial
aerosols. Most reports have focused on the control of infectious diseases, such as tuberculosis
(TB), in the occupied spaces of medical facilities. Ventilation duct use of UVGI has been
increasing. In-duct applications are generally more concerned with controlling environmental
organisms than with controlling infections agents. This document reports the results of a project
to investigate the ability of UVGI lamps to inactivate representative environmental microbial
aerosols in ventilation ducts.

Results

During this research, UVGI lamps were experimentally demonstrated to inactivate bioaerosols
composed of vegetative bacteria, bacteria spores, or fungal spores to a reproducible degree under
conditions of fixed dose. Vegetative bacteria were most susceptible to UVGI, with bacteria and
fungal spores being substantially more resistant. The performance equation commonly cited in
the literature for UVGI inactivation was found to generally apply, provided its parameters were
known.

The experimental measurements suggest that the less resistant vegetative organisms, represented
by Staphylococcus epidermidis, can be somewhat protected by organic material coating the
bioaerosol or by increases in humidity. Even with the added resistance, vegetative bacteria
remained susceptible to UVGI relative to the spores. A single lamp inactivated more than 50%
of the bioaerosol. Humidity and/or organic material coating the bioaerosol had much less impact
on inactivation of the spore-formers, which were much more difficult to kill and required use of
more lamps. For example, six lamps were required to inactivate the spore-formers to the same
extent reached using a single lamp on the vegetative bacteria. Overall, the microorganism-
dependent rate constant values determined during this research are similar to the applicable
values.

Of the physical factors that might influence UVGI lamp performance, air temperature and flow
rate, lamp design, and ballast design were found to be most significant. Isolated changes in
humidity had a measurable but very small effect on lamp irradiance. Lamps designated high
output were found to indeed provide higher irradiance than low output lamps, and lamps
designed specifically for lower temperature operation also appeared to perform better under
those conditions. However, the terminology is not completely consistent. UVGI lamps have a
point of maximum output when exposed to various air temperatures and airflow rates. This
optimal operating point is fundamentally related to lamp operating temperature and can be
reached by various combinations of environment and heat transfer rate (principally air
temperature and flow rate) that bring the lamp to the proper operating condition. The reduction
in output due to overcooling at temperatures achievable in a heating, ventilation, and air
conditioning (HVAC) system has been shown to be as much as a factor of two, which could have
a significant effect on kill for bioaerosols.

ii
Conclusions And Recommendations

The key factors for the design of in-duct UVGI systems intended to disinfect moving airstreams
are (1) UV lamp irradiance at operating conditions, (2) microbial resistance data, (3) having a
tested and functioning bioaerosol dose model, and (4) duct wall reflectance. Given this
information, UVGI system designers can make reasonable and responsible performance
estimates for most in-duct applications. (The common lack of test results that provide lamp
output at use temperature and airflow makes independent verification of a vendors UVGI
system design difficult.)

The good news that reliable UVGI system design is available must be tempered by an
understanding of the intrinsic variance of microbial populations. Resistance variations between
strains of the same organism have the potential to be substantial, but very little work has been
done. In consequence of this, UVGI system designers should
Apply safety factors to their designs, particularly as they depart from operating modes for
which they have performance data and field experience.
Know actual lamp output at the most challenging operating condition in the duct.
Avoid relying solely on design equations to determine the performance of their systems.
Actual testing with the contaminants of interest is highly recommended.
Be extremely cautious regarding claims about UVGI systems high levels of inactivation
of pathogenic bioaerosols. While the microbiological science underlying these
conclusions applies to pathogenic bioaerosols as well as environmental organisms, much
greater caution is required in the former case. It would be irresponsible to claim a high
inactivation rate for a pathogenic bioaerosol without substantial testing. Even with
substantial testing, design failures may occur.

In the efforts to counter bioterrorism, the use of UVGI is likely to be proposed with increasing
frequency. It is a low-pressure-drop, nonintrusive technology that has real potential. In that
light, we recommend that test methods and guidance for performance claims regarding
bioterrorism agents be developed by an experienced and broadly based group of technical
experts, users, and manufacturers. Until rigorous and adequate tests have been developed and
performed, claims regarding protection against aerosol bioterrorism agents are suspect. The
potential errors are too large and the consequences too great.

iii
 TABLE OF CONTENTS
EXECUTIVE SUMMARY...........................................................................................................ii

TABLE OF CONTENTS.............................................................................................................iv

1.0 INTRODUCTION............................................................................................................. 1
1.1 BACKGROUND .................................................................................................................. 1
1.2 PROJECT OBJECTIVES ....................................................................................................... 2
1.3 TECHNICAL APPROACH .................................................................................................... 3
1.4 REPORT OVERVIEW .......................................................................................................... 3
2.0 EXPERIMENTAL METHODS AND APPARATUS.................................................... 3
2.1 FULL-SCALE TEST DUCT .................................................................................................. 4
2.2 UVGI LAMP OPERATION AND IRRADIANCE MEASUREMENT ........................................... 6
2.2.1 UV Lamps and Ballasts ........................................................................................... 6
2.2.2 UV Lamp Operating Configurations....................................................................... 6
2.2.3 In-Duct Irradiance Measurements .......................................................................... 7
2.2.4 Apparatus for All-Angle UVGI Measurements ....................................................... 8
2.3 CHEMICAL BYPRODUCTS EXPERIMENTS ........................................................................ 11
2.3.1 Ozone Measurement.................................................................................................... 11
2.3.2 Volatile Organic Chemical Byproducts Measurement ............................................... 12
2.4 MICROBIOLOGICAL MEASUREMENTS ............................................................................. 12
2.4.1 Bioaerosol Disinfection Efficiency Measurement ....................................................... 12
2.4.2 Test Duct Operation .................................................................................................... 13
2.4.3 Test Bioaerosol Generation ........................................................................................ 13
2.4.4 Bioaerosol Sampling ................................................................................................... 13
2.4.5 Microbiological Growth and Enumeration................................................................. 14
2.4.6 Microbial Kill Estimates for Experimental Planning ................................................. 14
2.4.7 Data Reduction............................................................................................................ 15
2.4.8 Bioaerosol QA/QC ...................................................................................................... 16
2.5 IN-DUCT IRRADIANCE MODEL AND DOSE CALCULATIONS ............................................ 16
2.5.1 UVGI Irradiance Field in an Air Duct........................................................................ 17
2.5.2 Particle Movement in Air Ducts.................................................................................. 18
2.5.2 In-Duct UV Dosage Model.......................................................................................... 20
3.0 KEY FACTORS .............................................................................................................. 21
3.1 PHYSICAL FACTORS RESULTS AND DISCUSSION ............................................................. 21
3.1.1 Effect of Duct Air Conditions on UVGI Lamp Output ................................................ 21
3.1.2 Lamp and Ballast Design Effects ................................................................................ 24
3.1.3 Effects of Lamp Soiling ............................................................................................... 25
3.1.4 Reflectance ................................................................................................................... 25
3.1.5 Enclosure and Location Factors ................................................................................. 28
3.2 CHEMICAL BYPRODUCTS RESULTS AND DISCUSSION ..................................................... 30
3.2.1 Ozone........................................................................................................................... 30
3.2.2 Effect of UV lights on VOCs........................................................................................ 30

iv
 3.3 MICROBIOLOGICAL FACTORS RESULTS AND DISCUSSION .............................................. 32
3.3.1 Test Organism Selection.............................................................................................. 32
3.3.2 Growth of Organisms in the Dark............................................................................... 34
3.3.3 Recovery of Damaged Test Organisms ....................................................................... 34
3.3.4 The Influence of Aerosol Composition on Bioaerosol Tests ....................................... 35
3.3.5 The Influence of Relative Humidity on Test Organisms.............................................. 35
4.0 UVGI EFFICACY IN AIR DUCTS ................................................................................... 36
4.1 MEASURED EFFICACY AND DISCUSSION ........................................................................ 36
4.1.1 Vegetative Organisms ................................................................................................. 37
4.1.2 Spore-Forming Bacteria and Fungi............................................................................ 37
4.1.3 Summary of Calculated Microbial Resistance Constants ........................................... 38
4.2 DESIGN OF UVGI SYSTEMS FOR AIR DUCTS ....................................................................... 38
5.0 CONCLUSIONS ............................................................................................................. 39

6.0 RECOMMENDATIONS ................................................................................................ 40

7.0 REFERENCES ................................................................................................................ 40

v
 List of Tables

Table 1. Reflectance of various materials .................................................................................... 26

Table 2. Irradiance measurements with coil in duct..................................................................... 29
Table 3. Effect of Duct Configuration on Microbial Kill ............................................................ 29
Table 4. Chemicals identified in byproduct sampling.................................................................. 31
Table 5. Results of Preliminary Microbial Kill Experiments....................................................... 33
Table 6. Percent kill of the 3 test organims at varying test conditions ........................................ 37
Table 7. Summary of k values for test organisms ........................................................................ 38

List of Figures

Figure 1. 52.2 test rig arranged for bioaerosol tests of UV lamps ................................................. 5
Figure 2. Detail of aerosol injection region of test duct................................................................. 5
Figure 3. Lamp and test duct configurations.................................................................................. 7
Figure 4. Inside view of duct mock-up apparatus showing lamps ................................................. 9
Figure 5. Inside of duct mock-up with guillotine and locator grid .............................................. 10
Figure 6. Reflectance measurement position grid........................................................................ 11
Figure 7. Computed % kill as a function of dose and microbial resistance ................................. 15
Figure 8. Effect of air velocity and temperature on low power lamp irradiance at 1 m........... 22
Figure 9. Effect of air velocity and temperature on 1 m irradiance for high output lamps ...... 22
Figure 10. Measured effect of humidity on irradiance................................................................. 24
Figure 11. Irradiance distribution in galvanized duct .................................................................. 27
Figure 12. Irradiance distribution for black flock paper lined duct ............................................. 27
Figure 13. Ratio of galvanized to black flock irradiance ............................................................. 28
Figure 14. Results of chemical sampling up- and downstream of lamps..................................... 30

vi
1.0 INTRODUCTION

1.1 Background

The use of ultraviolet (UV) lights to disinfect room air and air streams dates to about 1900.
Richard Riley has numerous publications in the 1960s and 1970s that deal with UV disinfection
and has published an extensive review (Riley, 1988), as has Shechmeister (1991). Early work
established that the most effective UV light wavelength was the wavelength range from 220 to
300 nanometers (nm), with the peak effectiveness near 265 nm.

In current commercial practice, UV light of the appropriate wavelength is generated by electrical

discharge through low-pressure mercury vapor, which is enclosed in a glass tube that transmits
UV light. The resulting germicidal lamp produces UV light that has a primary wavelength of
253.7 nm. This wavelength is within the short-wave, or C band of UV light. It is sometimes
abbreviated as UVC and is alternately known as germicidal UV or ultraviolet germicidal
irradiation (both abbreviated as UVGI). UVGI has been shown to deactivate viruses,
mycoplasma, bacteria, and fungi.

To date, most of the use of UVGI for airborne organism control in buildings has focused on
limiting the transmission of very infectious diseases (particularly TB) in occupied spaces such as
medical waiting or treatment rooms. In this application, UVGI lamps are often located in the
upper part of a room (near the ceiling) and must be visually shielded from the occupied space.
Mixing between the breathing zone and the irradiated zone is critical to operation, and mixing
fans are sometimes used. The effectiveness of near-ceiling UVGI lamp use is fairly well known.
Numerous publications are available. In a recent paper that includes an extensive list of
references, Miller and Macher (2000) discuss the use and characterization of occupied-space,
near-ceiling UVGI. Application principles have been published (CDC, 1994), as have guidelines
for system selection and use (CDC 1999a, CDC 1999b.) The occupied space application of
UVGI is not discussed further in this report.

UVGI has been used in ducts for some time, and its use is becoming increasingly frequent as
concern about indoor air quality increases. Lamp manufacturers have published design guidance
documents for in-duct use (Westinghouse, 1982; Phillips, 1985; Phillips, 1992). Recent papers
by Schier and Fencl (1996) and Kowalski and Bahnfleth (1998) have summarized the literature
and discussed in-duct applications. These and other recent papers were based on case studies
and previously published performance data.

For constant and uniform irradiance, the disinfection effect of UVGI on a single microorganism
population can be expressed as follows (Phillips, 1992):

Nt/N0 = exp(- k Eeff t ) = exp(- kDose) (1)

where
N0 = the number of microorganisms at the start,
Nt = the number of microorganisms after any time
t,

1
 Nt/N0 = the fraction of microorganisms surviving,
k = a microorganism-dependent rate constant, cm2/W - s,
Eeff = the effective (germicidal) irradiance received by the microorganism, W/cm2, and
Dose = the product of Eeff t, W s/cm2.

The units shown above are common, but others are used as well, including irradiance in J/m2.
Eq. 1 describes an exponential decay in the number of living organisms as a constant level of
UVGI exposure continues. The same type of equation is used to describe the effect of
disinfectants on a population of microorganisms, with the dose in that case being a
concentration-time product. The fractional kill after time t is (1-Nt/N0). While the applicability
of Eq. 1 is well established for a microorganism exposed to a constant UVGI irradiance, some
investigators have detected fine structure in the decay curves, including onset dose thresholds
and changes in the k values at high dose (Kowalski et al., 2000.) Eq. 1 appears to be adequate
for practical use in ducts.

Eq. 1 shows an apparent distribution of resistance to UVGI within a population, with a fraction
of the microorganisms being killed at a low dose and others at a higher dose. This effect is
thought to be due both to a distribution of resistance inherent to the microorganism population
and the presence of various contaminants or multiple layers of organisms that reduce the
effective UVGI to the resistant microorganisms. Practically, bioaerosols are generated from
active colonies that contain numerous contaminants.

In a duct, the use of Eq. 1 is complicated by the movement of the target microorganisms in the
airstream and the fact that the UVGI irradiance is not constant within the duct. In addition, the
physical parameters of the duct, duct airflow, and UV installation have the potential to affect
both the irradiance and the microorganisms response to it. As is the case with upper-room UV
installation design, the principal design parameters for UVGI in in-duct applications are not
simple and include some uncertainty in the data available to analyze them and secondary effects.
The discussion below expands on these principal and secondary variables.

Recently, the use of UV lamps inside building ventilation ducts has been increasing in both
medical and nonmedical buildings for the control of environmental microorganisms. Therefore,
a need has arisen to better understand the capabilities of such installations. Prior to this research,
public data supporting the design and application of UV products in ducts were very limited.
This research arose out of a desire, on the part of the ventilation industry, to independently
evaluate the claims and design guidance of UV light vendors and manufacturers. Some UV
applications are intended to disinfect the air and others are intended to disinfect surfaces. The
research presented in this report was limited to in-duct use of UV lamps intended to disinfect
flowing air. Only environmental (not pathogenic) microorganisms were included in the test
program.

1.2 Project Objectives

The overall objective of this research program was to obtain reliable scientific data through
laboratory testing to determine the antimicrobial efficacy of UVC in circulating air ducts and the

2
potential ability of these lights to kill or inactivate common microbes and subsequently reduce
indoor air microbiological contamination.

1.3 Technical Approach

To accomplish the project objective, four tasks were undertaken:

1. Identifying the key factors impacting the use of UVC lights in circulating air ducts,
2. Quantifying the impact of these factors on the ability of UV devices to kill or limit
common environmental microorganisms,
3. Developing documented relationships between UVC disinfection efficacy and in-duct use
and design, and
4. Communicating that understanding to the ARI-member audience through appropriate and
useful test methods and design guidance.

Identification of the key factors was accomplished through a literature search and the
reconciliation of published UVC performance with the experimental results obtained during the
research. Quantification of the impact of these factors and the development of documented
relations of UVC performance were linked through the development of a model to calculate
(estimate) the UVC dose received by the organisms in the airstream and thereby link the physical
and microbiological measurements and permit their generalization to other situations.

1.4 Report Overview

Section 2.0 documents the apparatus and methods used during the experimental phase of this
research and also describes the model used to calculate the UVGI dose received by the
bioaerosols during the experiments. Section 3.0 identifies the key physical and microbiological
factors for in-duct UVGI use and briefly summarizes their impact as reported in the literature and
as demonstrated in experimental work from this project. Section 4.0 presents and discusses
results that bear on the primary research objective. Sections 5.0 and 6.0 present conclusions and
recommendations based on this research. References for the report are presented in section 7.0.

2.0 EXPERIMENTAL METHODS AND APPARATUS

The experimental methods and apparatus used during this research consisted of the following:

1. Operation of a test duct with UVGI lamps installed. This operation included
microbiological sampling upstream and downstream of the lamps and control and
measurement of rig temperature, humidity, and flow rate,
2. Installation, operation, and irradiance measurements for the UVGI lamps,
3. Chemical byproducts measurements in the test rig,
4. Test bioaerosol generation, sampling, counting, and disinfection efficiency calculation,
and
5. Use of a bioaerosol UVGI dose model to compute microbial resistance constants, which
allow generalization of the efficacy results to other lamp and duct configurations and
comparison to the literature.

3
2.1 Full-Scale Test Duct

The test duct used during this research is shown in Figure 1. It is a full-scale American Society
of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. (ASHRAE) 52.2 test rig
(ASHRAE, 1999) of recirculating design with airflow control over approximately 600 to 3,000
cfm (17 to 85 m3/min), temperature control between 50 and 85 F (10 and 30 C), and humidity
(3585% RH) control. The airflow through the test duct was measured using a calibrated ASME
flow nozzle. Flow was controlled by modulating the fan motor speed using the computer
interface. Control to the flow set point was automatic with operator tuning as required.
Temperature and humidity control were achieved using a flow-modulated chilled water coil and
modulated steam injector that were also computer-controlled. The test duct was operated at
positive pressure in the sample section to prevent the infiltration of room air aerosol. Air
entering the test section was filtered with High Efficiency Particulate Air (HEPA) filters to
remove ambient aerosol from the airflow.

As shown in Figure 1, the test rig is set up for recirculating operation, which was the operating
mode used for all bioaerosol runs. The dampers were set to draw air into the rig from the
conditioned space and discharge it through the roof for the byproduct chemicals tests.

Figure 2 shows additional details of the aerosol injection region of the test duct. The test
bioaerosol was injected at the location shown in Figures 1 and 2. For all experiments, the test
bioaerosol was nebulized from a liquid solution using a six-jet Collison nebulizer. The upstream
and downstream mixers shown have two parts: an upstream 30.5-cm (12-in.) diameter orifice in
a steel plate and a downstream 15.2-cm (6-in.) diameter target disk centered on the orifice.
Details of the test rig are given by ASHRAE (1999). For the initial screening tests, the test
bioaerosol was injected either directly into the duct (for easily desiccated vegetative bioaerosols)
or through the ASHRAE 52.2 drying tower (for more robust bioaerosols, such as the spores) and
thence into the duct through a Kr-85 aerosol neutralizer. The drying tower operation is described
in ASHRAE 52.2, and provides approximately a 50 s residence time for aerosol conditioning in
0.00189 m3/s (4 cfm) of clean, dry air.

Both injection techniques appeared to give adequate dispersion and mixing of the test aerosol,
which had a mean aerodynamic particle diameter of 2 to 3 m. However, use of the drying
tower was considered to be better practice and all test bioaerosols were injected through the
drying tower for the final tests. Further details of the microbial test aerosol preparation are
provided in Section 2.6.

4
Figure 1. 52.2 test rig arranged for bioaerosol tests of UV lamps.

Figure 2. Detail of the aerosol injection region of the test duct.

Aerosol concentrations were measured upstream and downstream of the test section to obtain the
challenge and penetrating bioaerosol concentrations, respectively. The inlet nozzles of the
upstream and downstream aerosol probes were designed to yield isokinetic sampling conditions
for the 28.3 L/min (1-cfm), single-stage Andersen samplers. The bioaerosol sample ports were
located at points of well defined airflow and well mixed aerosol conditions, as required to
achieve unbiased, representative samples. Mixing and flow at the sample locations were
confirmed by following the procedures in ASHRAE 52.2, which require the same suitable flow
conditions. Aerosol losses through the duct and the 180 bend, while present, were not large
(typically a few percent at 2 m and less than 20% loss at 10 m). The data reduction system
compensated for these losses as described below in the quality assurance (QA) and quality
control (QC) procedures.

5
2.2 UVGI Lamp Operation and Irradiance Measurement

2.2.1 UV Lamps and Ballasts

Two types of commercial UV lamps were utilized during this research. Because the industry
does not use a standard test to measure UVGI lamp output presented in most sales literature, the
lamps are difficult to characterize, except through direct measurement. Performance data are
presented below. Visually, the two test lamps were identical except for the label. Both were
0.61-m (24-in) long, T5 (~ 1.6-cm (5/8-in.) diameter), single-ended lamps intended for insertion
into a duct from outside. The ballasts were connected to the lamps through four-pin power
connectors. Power was provided to the in-duct end of the lamp by small wires running from the
power connector along the outside length of the lamp tube. A flange on the power connector was
used to clamp them in place perpendicular to the duct wall. The two lamp types were

1. Low-output lamps driven by conventional magnetic ballasts. These lamps were obtained
because they appeared (from inspection of vendor sales literature) to be characteristic of
the low-cost end of the in-duct market. They were branded by the vendor and thus could
not be independently traced back to a lamp manufacturer. Discussions with industry
experts led to a characterization of these lamps as low-pressure mercury lamps driven at
low power. In still air at 21C (70F), surface output was computed for a single lamp .
2. High-output lamps driven by electronic ballasts. Based on vendor sales literature, these
lamps were designed for both higher output and also improved output at low temperature
and with air flowing past them. They too were branded by the vendor and could not be
independently traced. An industry expert characterized these lamps as low pressure
mercury lamps that were over-driven to achieve higher output.

2.2.2 UV Lamp Operating Configurations

The UVGI lamps were operated (except during the alternate duct configuration tests) in the test
section shown in the center of Figure 3a, with the lamps in a vertical column perpendicular to the
airflow (crossflow). There were three primary lamp arrangements:

1. One lamp at duct center,

2. Three lamps, one at the center and one above and below on 20.3-cm (8-in.) centers,
3. Six lamps spaced on 10.2-cm (4-in.) centers in a vertical column.

A screen was used to obtain irradiance levels lower than those that could be obtained from a
single low-output lamp. The screen consisted of three concentric layers of common window
screen, rolled into a cylinder and slipped over the lamp.

6
Figure 3. Lamp and test duct configurations

2.2.3 In-Duct Irradiance Measurements

Bioaerosols in a duct may be exposed to UVGI light from all angles, particularly if the duct is
significantly reflective to UVGI. The commonly available detector for UVGI is a solid-state

7
detector whose otherwise broad wavelength response is limited (using filters) to the UVC. A
calibration is used to convert the voltage produced by the detector to UVGI irradiance. Some
filters pass a broad UVC spectrum, while others have a narrower response.

All irradiance measurements conducted during this research utilized an International Light (IL)
IL1700 Research Radiometer fitted with a factory NIST-traceable calibrated IL SED240
detector, QNDS2 filter, and W quartz diffuser. This detector has a 185 to 310 nm wavelength
spectral response. With the diffuser installed, the detector has what is known as a cosine
response characteristic. A true irradiance reading is obtained for light entering perpendicularly
(incident angle of 0) into the detector. The detector measures UV light entering from other
angles as the product of the true irradiance and the cosine of the angle of incidence. Positioned
in the center of the duct and aligned in the direction of flow, the UV detector measures all UV
light striking it at 0 and 70.7% of the UV light striking it at an incident angle of 45. Very little
UVGI is detected as the incidence angle widens to the sides of the detector and, of course, no
UVGI from the back can be detected. All common UVGI radiometers utilize the same operating
principle and have the same limitation.

Use of a standard, cosine-response UVGI radiometer in this research was desirable because it is
the instruments most likely to be used in the ventilation industry. The use of this type of
radiometer required investigation because a bioaerosol particle receives UVGI energy from all
angles, while the UVGI radiometer does not. Should reflected UVGI from the duct walls be a
high fraction of the total UVGI impinging on the bioaerosol particles, single-position
measurements made with these detectors aligned with the duct axis would underestimate the total
dose and be of limited value. Multi-axis measurements with a cosine-response detector would be
time consuming and require consolidation of the data into a total irradiance value.
(Photosensitive chemical methods have been developed for time-integrated UVGI dose
measurement; however, these are also difficult to implement in the field.)

Irradiance measurements within the test duct were made at various distances from the lamps.
The detector was centered in the duct at distances ranging from 25 to 100 cm (10 to 39.4 in.)from
the lamp centerline and aligned along the duct centerline. The instrument cable was run through
a duct wall penetration and connected to the IL1700. All lamp performance measurements were
made after allowing the UV lamps to stabilize at the duct operating conditions.

2.2.4 Apparatus for All-Angle UVGI Measurements

A mocked-up duct section was utilized to measure irradiance from all possible incident angles
and thereby permit evaluation of the errors attendant to use of the radiometer detectors. The duct
mock-up was used in place of the test duct to allow the detector to be rapidly positioned without
exposing the operator to the UVGI. The sections below describe the apparatus and methods.

2.2.4.1 Duct mock-up and all-angle UVGI measurements

The duct mock-up was a 61- by 61-cm (24- by 24-in.) interior dimensions duct section in which
the light from the lamps could be blocked to allow the radiometer to be adjusted without
exposing the operator to the UV light. The wall materials could also be easily changed for

8
reflectance measurements. The mock-up consisted of a box with a replaceable lining. The ends
of the apparatus were covered with grills to prevent accidental exposure. Access into the box
was through two full-depth doors that formed the top of the duct mock-up. The three test lamps
were installed in a vertical column on the right side of the apparatus as shown in Figure 4.
Because the lamps needed to be kept at a stable operating temperature, they could not be turned
off without allowing for a substantial warm-up period. Accordingly, they remained on
throughout the experiment, and a sheet metal guillotine damper (seen in Figure 5) was lowered
before the access door was opened. The relative irradiance level was primary in this experiment,
not the absolute level, and therefore temperature control beyond a stable operating condition was
not required. For all of these experiments, three low-output lamps powered by magnetic ballasts
were positioned perpendicular to the flow direction in the duct mock-up, as shown in Figure 4.
One lamp was at the horizontal center, one 20.3 cm (8 in.) above, and one 20.3 cm (8 in.) below.

Figure 4. Lamps shown in an inside view of duct mock-up apparatus.

The radiometer and detector were described above. The detector was mounted on a rotating and
hinged protractor-equipped base and tripod that allowed the detector to be moved vertically and
aligned in any horizontal or vertical angle. A pencil laser (mounted below, and aligned with, the
detector axis of measurement) was used to accurately position it using the grid on the damper.
While not obvious in the photographs, a 22 acceptance angle cone was mounted on the front of
the detector to prevent confusion when interpreting the all-angle measurement data.

9
Figure 5. Inside of duct mock-up with guillotine and locator grid.

The UV detector was positioned at 50 and 100 cm (19.7 and 39.4 in.)from the plane of the lamps
at the center of the test duct. Irradiance measurements were made every 22.5 around both a
horizontal and a vertical plane, with 0 being the perpendicular from the plane of the lamps to the
detector face.

2.2.4.2 Measurement of irradiance with lined duct

Different materials reflect UV light to a different extent, as further discussed below. The
apparatus was used to allow the measurement of irradiance for both relatively high- and low-
reflectance materials. The purpose was to determine effect of reflectance on the uniformity of
irradiance in the duct and the level of irradiance measured using the radiometer.

10
Figure 6 is a schematic drawing of the measurement locations in the duct mock-up. The
irradiance from three low-output lamps was measured at static flow conditions, which allowed
comparison within this data set. The measurements were made within a single quadrant of the
duct mock-up. This quadrant was divided into 9 segments, each 10.2 (4 in.) on a side (this is the
same grid used in the performance model). Measurements were made with the radiometer
detector facing the lamps at 30, 50, 75, and 100 cm (11.8, 19.7, 29.5, and 39.4 in.) from the plane
of the lamps. The same set of measurements was made for new galvanized metal and for
photographic black flock paper.

Figure 6. Reflectance measurement position grid.

2.3 Chemical Byproducts Experiments

All chemical byproducts measurements were conducted in the full-scale test duct at ambient
temperature and humidity and at a flow rate of 0.472 m3/s (1,000 cfm.) To prevent buildup of
any byproducts, the test duct was operated in open-loop mode, without recirculation. Inlet air
was drawn from conditioned space, filtered ahead of the test section, and discharged through a
roof vent. Six lamps were mounted and operated in crossflow on 10.2-cm (4-in.) centers. They
were allowed to operate for at least 15 minutes prior to beginning each test.

2.3.1 Ozone Measurement

Specific types of UV lamps are used to generate ozone in laboratory settings, which raises
questions about ozone production by UVGI lamps. The 185-nm emission line for mercury is
generated in low-pressure discharge lamps and would generate ozone in air. However, most
low-pressure UVGI lamps intended for use in air utilize glass that does not pass the 185-nm

11
wavelength; this eliminates/minimizes ozone production. Medium-pressure mercury lamps, on
the other hand, may generate ozone. The intent of the ozone generation experiment was to
validate the assertion that no measurable ozone was produced under normal operating conditions.
The challenge of this experiment lay in the large amount of air required to operate normally and
hence the expected low concentration difference between inlet and outlet concentrations.

The potential for UVGI lamps to produce ozone under operating conditions was investigated
using an ambient ozone analyzer (Thermo-Environmental Instruments, Inc., Model 49 UV
Photometric O3 Analyzer, 050 ppm range) to measure ozone upstream and downstream of the
six UVGI lamps. The gas sample was drawn directly into the instrument. The upstream sample
was obtained at the bioaerosol upstream port (2.13 m (7 ft) upstream of the lamps), and the
downstream samples were centered approximately 46 cm (1.5 ft) downstream and at the
downstream bioaerosol sample port [6.1 m (20 ft) downstream.] Tests were run with both the
low- and high-output lamps.

2.3.2 Volatile Organic Chemical Byproducts Measurement

Similarly to the issues regarding ozone production by UV lamps UV light under some conditions
is also known to induce chemical reactions in some volatile organic chemicals (VOCs). The goal
of this experiment was to operate the UVGI lamps under normal duct operating conditions,
challenge them with a cocktail of VOCs, sample upstream and downstream, and look for
chemicals downstream that were not found upstream. As with the ozone experiment, the major
experimental challenge was the level of dilution, which would make detection of low levels of
byproducts very difficult.

The test duct was operated at 0.472 m3/s (1,000 cfm), 22C (72 F), and 50% RH in open loop
mode. A VOC cocktail consisting of approximately equimolar fractions of toluene, hexane, 2-
butanol (commonly known as methyl ethyl ketone, or MEK), and iso-butanol was injected into a
total vaporizer at 0.2 mL/min. The vaporized liquid entered a 57 L/min (2-cfm) hot airstream
that was injected into the test duct, producing an approximately 1.5 ppm VOC challenge.

Air samples were drawn through adsorbent sampling tubes at 0.2 L/min for 90 min to obtain
upstream and downstream samples of the duct air. The sample tubes were located at the duct
center at the location of the bioaerosol upstream and downstream sample ports. Duplicate
samples were drawn; the total duct operating time was approximately 3.5 hours. A field blank
adsorbent tube was also obtained.

The sample tubes were returned (chilled) to the commercial analysis company, where they were
desorbed into a GC/MS and the concentrations of total and individual VOCs were measured.

2.4 Microbiological Measurements

2.4.1 Bioaerosol Disinfection Efficiency Measurement

The basic measurement of bioaerosol kill by UVGI irradiation was based on the following
experimental procedures:

1. Establishment of stable flow and lamp operation in the test duct at the test conditions,

12
2. Generation of a stable test bioaerosol of the desired test organism,
3. Collection of upstream and downstream bioaerosol samples, using one-stage Andersen
biological samplers,
4. Growth and enumeration of microbiological samples, and
5. Data reduction to compute the kill efficiency of the UVGI lamp configuration.

The same general procedures were used for each bioaerosol kill measurement.

2.4.2 Test Duct Operation

The test duct was operated as described in Section 2.2.

2.4.3 Test Bioaerosol Generation

Except for a brief investigation of solid aerosols, all test bioaerosols used during this research
were nebulized from an aqueous organism suspension and injected into the test duct. Depending
on the organism, the bacterial and fungal challenge suspensions were prepared by either

1. Inoculating the test organism onto solid media, incubating the culture until mature,
wiping a wetted sterile swab across the surface of the pure culture, and eluting from the
swab into sterile high-purity deionized [18 megaohm (M)/cm resistance] water to
obtain 15% transmission on the spectrometer, or
2. Purchasing a spore suspension from a commercial vendor and diluting it in sterile water.

Both correspond to a concentration of approximately 1.5 x 107 CFU/mL. The microorganisms

were resuspended in sterile 18 M/cm deionized water. Very high purity water was used to
minimize the particles from sources other than the organisms themselves (e.g., dissolved solids).
For the broth suspensions, the organisms were prepared following the same protocol, but were
resuspended in trypticase soy broth instead of 18 M/cm deionized water. However, in an
attempt to minimize particulate contamination from the broth, the broth was prepared in 18
M/cm deionized water.

The challenge organism suspensions were aerosolized using a Collison modified MRE-type six-
jet nebulizer (BGI, Waltham, MA) at 100 kPa (15 psi) air pressure. The Collison generates
droplets with an approximate volume mean diameter of 2 m. The particle diameter after the
water evaporates depends upon the solids content of the suspension. Particle size is determined
by the size of the suspended particles (if singlets).

During bioaerosol generation, an optical particle counter (OPC) was used to monitor particle
levels in real time to enable control of the experiment within the test duct. The output was sent
to a RS-232 serial port and was recorded on a computer. The OPC was not used for the
efficiency measurements because it cannot distinguish viable from non-viable organisms.

2.4.4 Bioaerosol Sampling

Bioaerosol samples were collected by sequential sampling with the one-stage Andersen
bioaerosol samplers. Simultaneous triplicate samples were obtained during each upstream or
downstream sampling session, and two upstream and two downstream sample sessions were run
on any single experimental condition. The bioaerosol sampling lines were 1.4 cm ID stainless

13
steel lines (5/8-in. OD tube) with dimensions chosen to minimize particle losses during
sampling. The sample lines upstream and downstream had the same configuration in terms of
diameter, horizontal and vertical lengths, and number of bends limit upstream and downstream
differences. Full flow quick-connects were used to connect the bioaerosol samplers to the
upstream and downstream sample lines. Sampling nozzles of the appropriate entrance diameter
were placed on the ends of the sample probes to maintain isokinetic sampling for the test flow
rate.

The one-stage Andersen samplers (Graseby Andersen, Atlanta, GA) are 400-hole multiple-jet
impactors operating at 28.3 L/min (1 cfm.) During sampling, air is impinged on agar in petri
dishes positioned beneath the stages. Only culturable, viable microorganisms were measured.

2.4.5 Microbiological Growth and Enumeration

The bioaerosol sampler plates were allowed to grow and colony-forming units (CFUs) were
enumerated. The growth media used were trypticase soy agar (TSA) for the bacteria and
Sabaurouds dextrose agar (SDA) for the fungi. CFUs were counted shortly after moderate
growth became apparent.

A positive-hole correction is applied to adjust for the probability that more than one viable
microorganism is collected through a sampling hole and combined with other microorganisms to
produce a single CFU (Macher, 1989). The impactor cut-point (d50) is 0.65 m for the one-stage
Andersen.

2.4.6 Microbial Kill Estimates for Experimental Planning

An individual microbiological kill effectiveness test as described above provides the kill in the
form of the ratio Nt/N0 for a given organism, configuration, lamp, and operating condition (flow,
temperature, RH.) Eq. 1 can be used to generalize effectiveness measurements for various UV
lamps and microorganisms by computing the microbial resistance constant, k, provided all the
other terms are known. This requires that the parameters in Eq. 1 be in a measurable range. The
radiometer described above had a sufficient dynamic range to measure Eeff from as many lamps
as was practical to place within the test duct. The exposure duration, t, was calculable as the
bioaerosol path length having meaningful UVGI exposure, divided by the duct velocity.

For the microbial kill experiments, it was important to stay within a measurable range of kill. An
experimental condition that achieves 100% kill cannot be analyzed because the fraction of UVGI
radiation that is wasted cannot be evaluated. Similar, zero kill (below the detection limits of
the impactors) experiments were not useful. The tests were conducted in an expected kill range
of roughly 1095%, although dilutions and long sampling runs extended the range somewhat.
For these reasons, the initial step in setting up the UV lamp systems for these experiments was to
estimate the microbial kills for different configurations of UVGI lamps and duct velocities.

Eq. 1 was used to estimate the UVGI dose required in a duct to achieve a measurable microbial
kill. Figure 7 plots Eq. 1 for a reasonable range of k. Fungi are the most resistant to UVGI,
while vegetative bacteria are easily killed. Figure 7 shows that achieving UVGI doses in the 500
to 2,500 Ws/cm2 range should provide adequate range for the bioaerosol experiments.

14
 Total Dose, uW*s/cm^2
0 2000 4000 6000 8000 10000
100.0%

Percent Kill for indicated "k"

10.0%
k = 3E-5 (fungi)
k = 4E-4 (bacterial spores)

k = 1E-3 (vegetative bacteria)

1.0%

Percent Kill as function of Dose

(%Kill = 1 - exp(-k*dose)

Figure 7. Computed % kill as a function of dose and microbial resistance

Lamp output was directly measured in a duct, and doses in the range of approximately 1,500 to
2,500 Ws/cm2 were found to be readily achievable with three low-power lamps in a 2-m (78.9
in.) exposure distance [1 m (39.4 in.)on either side of the lamps.] The irradiance field was found
to be relatively uniform. The initial bioaerosol experiments were then designed with confidence.
Later experiments were designed applying more sophisticated models of bioaerosol dose in a
flow air duct that are described in the next section.

Figure 7 also shows that increasing dose (with more or better lamps or ballasts or by reducing
velocity) does not proportionately increase kill and will have very little effect at some dose
levels. This also means that a modest dose can do nearly as well as a high dose and achieve high
kill through recirculation and repetitive exposure.

2.4.7 Data Reduction

Measurements were made with the UVGI lamps both ON and OFF to account for losses in
the duct downstream of the lamps. A single measurement run consisted of two sets of three
simultaneous, 1-min, one-stage Andersen upstream samples averaged to give the upstream
challenge count, Cup, and two sets of three downstream samples averaged to provide the
downstream challenge count, Cdown, for both the ON and OFF cases. The fractional
penetrations, PON and POFF, were computed as the respective Cup/Cdown . The inactivation
efficiency was then computed as shown in Eq. 2:

Inactivation efficiency = 100 ( 1 - PON/POFF) . (2)

15
The two replicate runs were compared and averaged to arrive at a final test result for the device,
challenge organism, and operating conditions. The dose exposure model was then used to
compute the k constant for the organism used in the test.

2.4.8 Bioaerosol QA/QC

Multiple quality assurance and control tests were used to maintain high-quality data. The first
line of checks were physical system checks to ensure that the test conditions were as specified.
All measurement instruments were calibrated and maintained as suggested by the manufacturer.
System configuration was checked prior to each test to ensure that the test bench was set up as
specified. Temperature, humidity, duct flow rate, and UVGI irradiance were measured prior to
and following each test.

No standard UV germicidal device exists to allow calibration of the microbiological aerosol

efficiency measurement in the test duct. However, the test duct can be zeroed and spanned by
testing at the 0% and 100% penetration points. In previous filter tests, these tests have been
found critical for maintaining high-quality data. The 100% penetration tests were performed to
provide a relatively stringent check of the adequacy of the overall duct, sampling, measurement,
and aerosol generation system, as well as organism viability. These tests were performed as
normal disinfection efficiency tests, except that the UVGI lamps were not turned on. A perfect
system would yield a measured penetration of 100% for the test bioaerosol. Deviations from
100% efficiency occur because of particle losses in the duct, losses within the test device,
differences in the degree of aerosol uniformity (i.e., mixing) at the upstream and downstream
probes, differences in particle transport efficiency in the upstream and downstream sampling
lines, and loss of viability of the test organism. Results from the no-light tests (POFF correction
factors) were used during data analysis to correct the efficiency measurements obtained during
the filter tests.

The 100% efficiency test was conducted by installing a HEPA filter in the test duct to ensure that
near-zero penetration was measured under the sampling conditions. Measurement of significant
penetration generally indicates sampling line leaks or sample handling problems.

Additional bioaerosol QC tests ensured the purity of the test organism suspension and the
sterility of the media and sterile water.

2.5 In-Duct Irradiance Model and Dose Calculations

While the UVGI irradiance and the microbial kill achieved can be measured as described above,
an important part of the data analysis is the development of a model that allows generalization of
a few measurements to the larger world of air ducts. This section describes the model used in
this research. The general approach applied was to begin with a model for irradiance in a duct
and compare that model to measurements made with the radiometer. The model was then
calibrated such that it predicted the irradiance measured at a few locations. A set of reasonable
bioaerosol trajectories in the duct were then proposed, and the dose received by a particle
following each trajectory was computed step-wise over a distance of 1 m (39.4 in.) This
trajectory dose was used with Eq. 1 to compute bioaerosol survival fractions on that trajectory.

16
The survival fractions were averaged over all of the trajectories to obtain an average survival,
from which an average k could be computed.

2.5.1 UVGI Irradiance Field in an Air Duct

UV light is electromagnetic radiation, and as such is governed by well known physics. All
photocatalytic reactions (smog formation, for instance) and radiant heat transfer follow the same
physical laws. The energy passing through a fixed area decreases according to the inverse square
law, because the irradiance reduction is inversely proportional to the square of the distance from
the source. This assumes a point source and point receptor.

Because commercial UVGI lamps are generally cylindrical sources with finite length and
diameter, the inverse square law is not easily used to compute the UVGI irradiance field within a
duct. An arbitrary location in the duct is irradiated by the whole length of the lamp as well as by
reflected light. In practice, the UVGI field falls off roughly as the inverse of distance from the
lamps. For most lamps and ducts, this means that the biological inactivation takes place within
0.5 to 1.0 m (20 to 40 in.) of the lamps.

Note: Terminology is not uniform within the literature. Phillips (1992) uses irradiance, which
is defined with units of power per unit area (e.g., W/cm2), as does Ryer (1998) in his Light
Measurement Handbook. Irradiance is a different quantity than intensity, which is an angular
measure with units of power per steradian (e.g., W/sr) as stated by Ryer (1988). The UV
literature, however, sometimes uses intensity with the meaning and units of irradiance.
Confusion will be minimized by carefully observing the units that accompany the numbers.

Use of the view factor calculation is an available alternative. Because photocatalytic reactions
and radiant heat transfer have both been studied extensively, means to compute the energy
passing through a unit area have been developed for numerous geometries, including a
cylindrical source and planar receptor. Use of these equations is cited by Phillips (1992) and
Kowalski and Bahnfleth (2000). The intensity of the lamp is assumed to be constant over its
length, and the irradiance can then be calculated at any location in the duct.

The direct UVGI irradiance field at any position can be calculated for cylindrical lamps by using
a view factor equation originally developed for radiant heat transfer from cylinders (Modest,
1993). The applicable view factor is that for a differential planar element to a right-circular
cylinder of finite length, l, and radius, r; normal to element that passes through one end of the
cylinder and is perpendicular to the cylinder axis. The length of the normal between the
cylinder and the element is h. For an arbitrary differential planar element and a UV lamp, the
view factor must be computed for each end of each lamp because the perpendicular passes
through one end of the cylinder. The sum of the view factors for the two ends of the lamp gives
the total view factor for the lamp. The product of the lamp view factor and the lamp surface
intensity gives the total UVGI energy predicted to pass directly through point P.

L 1 1 L X 2H X ( H 1) H 1
VF = tan + tan 1 tan 1 (3)
H L H 2 1 XY Y ( H + 1) H +1

17
where
VF = view factor, either left or right end of lamp,
l = lamp segment length,
r = lamp radius,
h = perpendicular distance from lamp to point, P, in duct,
L = l / r, dimensionless lamp length,
H = h / r, dimensionless distance,
X = (1 + H)2 + L2, and
Y = (1 - H)2 + L2.

VFentire lamp = VFleft + VFright .

Irradiance at P = lamp UVGI surface irradiance VFentire lamp ;

Irradiance at P = (energy into lamp UVGI output efficiency)/lamp surface area .

The calculation was repeated for each lamp and the irradiance added to obtain total irradiance at
each point. Dose calculations have been reported using an assumption of total mixing with
irradiance computed at a fixed distance to use of the view factor equation to compute irradiance
at multiple points along an assumed or computed particle trajectory.

The view factor calculation does not account for reflected light. As stated above, the literature
(in most cases) accounts for reflectance from ducts by multiplying the direct irradiance by a
factor that depends on the duct wall material. Kowalski and Bahnfleth (2000) mention use of a
ray-tracing method that allows direct calculation of the UVGI energy reflected from a surface.
Ray tracing appears to have some potential advantages as a more exact calculational tool, but the
discussion is limited and its overall usefulness relative to more conventional techniques is not
clear. Precision in the UVGI calculations may not limit the design of UVGI systems.

Practically, UVGI lamp output is known to be affected by a number of factors, as discussed

below. The data appear to have trends, but the published literature is incomplete. Because this
project deals with the practical issue of how to ensure that germicidal UVGI installations work
with high reliability, measurements are needed. These measurements can then be correlated with
the microbiological effects.

2.5.2 Particle Movement in Air Ducts

2.5.2.1 Duct airflow pattern

The mean velocity of the ventilation air in the vicinity of UVGI devices is commonly in the
range of 7.1 to 14.2 m/min (250 to 500 ft/min.) In a ventilation air duct that has dimensions of
feet, this velocity results in turbulent flow with a Reynolds number of 20,000 to 50,000.
Depending on the upstream duct design, fittings, and obstructions, this flow may be fairly evenly
distributed in the duct, or significant flow variation may be present. For UVGI, a fairly uniform
velocity distribution across the duct is desirable because it generally provides uniform particle
loading, and irradiating all of the microorganism particles is the goal. As in many other
ventilation applications, an asymmetric flow profile could lead to surprises in UV disinfection.

18
Normal duct flow characterization using Pitot tube traverses or averaging flow measurement
devices is sufficient.

While a turbulent flow condition suggests that considerable mixing is taking place, this may not
be so. The turbulence found in air ducts at moderate Reynolds numbers is composed of
moderate-sized eddies that may travel several meters down a duct without a great deal of mixing.
For example, the author has observed, in a low-speed wind tunnel, neutral density bubbles
moving about within, but staying in, the 15.2-cm (6-in.) square eddy created by an egg-crate
flow straightener over a distance of approximately 8 m (25 ft.) The eddies are not stable, and
form, die, and re-form constantly, which causes some mixing. However, the actual flow path of
an arbitrary particle of fluid is a random path around a relatively straight average line of flight,
and mixing is limited over a the effective range of UVGI lamps. Flow smoothing devices such
turning vanes tend to further reduce mixing.

While the flow is indeed turbulent, to assume that all particles of the fluid are mixed in a short
distance is a mistake. A fluid particle in the corner as the flow enters the UV-irradiated portion
of the duct is likely to stay there as it moves through. For critical applications, the Eefft product
(or dose) must be adequate to kill (or achieve the target level of reduction) in the furthest corner
of the duct. This is consistent with the advice offered by Phillips (1992): The critical factor is
to reach the (dose) threshold at the wall. There the irradiation level is at minimum ....

2.5.2.2 Particle dynamics in airflow

Environmental microbiological particles, while having various shapes and surface
characteristics, follow the same flow dynamics that other airborne particles follow. They are
generally in the range of 1 to 4 m in aerodynamic diameter. Aerosol particles of this size in a
flow field at duct velocities [127 cm/s (250 ft/min)] generally follow the main fluid flow path,
deviating as the vector sum of their momentum and drag force require. The settling velocity of
such particles is on the order of 0.003 to 0.05 cm/s, and the particle velocity is dominated by the
duct airflow. Most will follow the flow closely. (Other forces on particles are less important in
the open duct.) Particle momentum leads to particles modest deviation from the turbulent eddy
movement. Particles near the boundaries of an eddy, however, may deviate sufficiently to enter
the next eddy. This results in particle mixing as well as air mixing, as does the death and
formation of eddies. As mentioned in the previous section, however, the mixing is not a great as
might be assumed.

Particles are not always evenly distributed across a duct. Micrometer-sized particles are known
to stratify in the elbows of ducts, and the bottoms of ducts collect most of the dust deposit.
Common particle sampling guidelines suggest that 10 duct diameters are required to recover a
mixed velocity and particle distribution condition downstream of a flow obstacle. Kowalski and
Bahnfleth (2000) assert that sufficient mixing will occur at these velocities to temper the effects
of a non-uniform velocity profile and make other statements endorsing the assumption of
complete mixing within the UV irradiated zone. Their exact meaning is not clear, but the authors
are probably justifying the use of an average velocity rather than implying that an air duct is a
completely mixed reactor.

19
From a UV system design viewpoint, unless the particles are so concentrated that they
appreciably attenuate the UV radiation, particle position in the duct does not matter so long as
the entire duct is irradiated to the threshold dose. (Position would be an issue under dusty
conditions; under such conditions, the UV lamps would become soiled and the installation would
be limited by soiling, not by its design.) Thus no assumption of particle position is required
provided the duct is fully irradiated with a dose adequate to achieve the desired performance.

On the other hand, particle position would be very important if a UV light or UV device was
installed such that it was not symmetric in the duct. Under such conditions, particle stratification
could lead to significant differences in performance because some microorganisms would
receive a larger dose than others by virtue of their transit path.

2.5.3 In-Duct UV Dosage Model

The model was implemented in a spreadsheet. It was based on the following:

1. Estimation of lamp surface output from the irradiance measured at a few points near the
duct centerline, and use of that output in the view factor equation.
2. Use of the view factor equation to compute the UV irradiance at 100 points along each of
9 straight-line particle trajectories. These 9 trajectories were the centerline of the squares
in 1 quadrant, much as was shown in Figure 6. The product of the time step duration and
the irradiance at each point, added up over the full length of the trajectory, gave the
calculated dose for that trajectory. This amounts to an assumption of complete mixing
within the cell for each time step, but not from time step to time step. A 0.01-s time step
was used, which gave about 80 irradiance calculations along each trajectory in a 1-m
travel distance for each of the 9 cells. By symmetry, the irradiance in each quadrant is
identical, so the 9 cells represented the entire duct. For the dimensions of the test duct,
each trajectory was the centerline of a 10- by 10- by 100-cm (4- by 4- by 39.4-in.) cell.
3. The assumption is made that the microbial particles are uniformly distributed among
these 36 trajectories.
4. Calculation of the accumulated dose for microbial particles passing through each of the
100 UV calculation points as they travel along the 36 trajectories.
5. Calculation of expected kill from literature values of microbial resistance to UV.

The results from the model were then compared to the measured microbial kill within the test rig.

The model is currently as sophisticated as is appropriate for a spreadsheet model. It seems to be

working well, in that the estimated lamp efficiency values and the estimated microbial kill values
agree reasonably well (~20%) with literature values.

20
3.0 KEY FACTORS

3.1 Physical Factors Results and Discussion

3.1.1 Effect of Duct Air Conditions on UVGI Lamp Output

3.1.1.1 Duct air temperature and velocity

Phillips Lighting (1992) states that the resonance line at 254 nm in a low-pressure mercury lamp
is strongest at a particular vapor pressure that occurs in their proprietary UVGI lamps at a still air
ambient temperature of about 68 F (20 C). For their lamps, this gives a lamp wall temperature
of about 104 F (40 C), and the UVGI output is greatest at that temperature. At both higher and
lower temperatures, the UVGI output is reduced, with the output at 50 F (10 C) being about
88% of that at 68 F (20 C). In their product literature, Westinghouse (1985) concurs. They
note that the output of their UVGI lamps, like all other gaseous discharge lamps, diminishes as
the temperature increases or decreases from the design temperature, which for the Westinghouse
lamps is stated to be 80 F (27 C) in still air. They note that the output of one lamp at 40 F (5
C) is only two-thirds of its output at 80 F (27 C.)

Westinghouse (1985) also notes that low temperatures can reduce the operating life of the lamps
and that another of their proprietary lamps responds differently. Another lamp, when operated at
its highest current input, is said to be much less sensitive to ambient temperature changes.
Phillips Lighting (1992) makes a similar argument, saying that their medium-pressure mercury
lamp, relative to the low-pressure lamp, has a higher power density and higher wall temperature
and is less sensitive to ambient temperature fluctuations. High lamp output at the lower
temperatures encountered in ducts is a central argument for the high output UVGI emitter
presented by Scheir and Fencl (1998). Westinghouse (1985) further notes that low operating
temperatures reduce the operating life of their lamps.

The effect of airflow on UVGI lamp output is increased heat transfer that is due to the moving
air. If the air moving past the lamp is ambient or cool, the lamp may be cooled below its
optimum operating point, which reduces output. If the air is warm, the lamp may be heated
above its optimum, which also reduces output. Lamps in the return air of building ventilation
systems are likely to be slightly cooled below their optimum temperature, while lamps
downstream of a cooling coil could be considerably cooled due to a combination of airflow and
low temperature. Lamps designed for operation at low temperatures should also be resistant to
airflow effects.

The results obtained during this research for irradiance as a function of air temperature and
velocity are shown in Figures 8 and 9 for low and high output lamps, respectively. Both
figures show duct center irradiance at 50 cm (19.7 in.) as a function of duct velocity for each of
three duct air temperatures. The overall shape of the curves appears to be the same in all cases.

At a fixed temperature, the irradiance produced by both kinds of lamps was found to increase
from a minimum at no flow (and consequent high lamp temperature) to a maximum at a

21
particular flow rate (the optimum lamp operating temperature), and then to decrease as the flow
rate increased further and the lamp was cooled more than was optimum. Airflow has its
strongest impact at the lowest temperatures, where the available temperature difference (between
the air and the lamp wall) to drive heat transfer is greatest. The apparent reduction in the effect
of flow rate at high flow rates [decreasing negative slope of the lines above 2 m/s (400 ft/min)]
presumably occurs because the temperature difference between the lamp wall and the air is being
reduced and the heat transfer is consequently being reduced.

0.0030

0.0025 14 - 17 C 21 C 30 C
2
Irradiance, W/cm

0.0020

0.0015

0.0010

0.0005

0.0000
0 1 2 3 4
Velocity in Duct, m/s

Figure 8. Effect of air velocity and temperature on low power lamp irradiance at 1 m

0.0030

0.0025

0.0020
2
Irradiance, W/cm

0.0015

0.0010
14 - 17 C 21C 30C
0.0005

0.0000
0 1 2 3 4
Velocity in Duct, m/s

Figure 9. Effect of air velocity and temperature on 1-m irradiance for high output lamps.
This behavior is consistent with literature reports and discussions with manufacturers, and the
shapes of the curves are similar to those presented by Phillips (1992). UVGI lamp output

22
reaches a maximum output at an operating temperature that is determined by lamp mercury
content, fill gases, electrical energy throughput, and other electrical considerations. For any
given lamp/ballast combination, the lamp operating temperature is most affected by flow rate and
temperature.

3.1.1.2 Relative humidity

The literature regarding the effects of humidity on UVGI does not present a completely
consistent picture. Water molecules are known to adsorb UV light, so increased humidity should
reduce the transmittance of UVGI. The effect should be proportional to absolute humidity (mass
of water per volume of dry air), rather than to relative humidity. Within the normal ranges of
temperature and humidity in ventilation systems, the effect would be expected to be modest
because water makes up only a small fraction of the total gas, and that fraction changes little. As
noted by Kowalski and Bahnfleth (2000), studies on this matter are contradictory and
incomplete. Scheir and Fencl (1998) state that humidity is an attenuator to UVGI energy.
Phillips Lighting (1985) provides a graphical correction factor to increase the number of lamps
that are required when the relative humidity exceeds 60%. On the other hand, Phillips Lighting
(1992) does not include this correction factor. The effect of humidity may be different for
surface disinfection than for bioaerosol treatment. Relative humidity above approximately 60%
has been shown to affect the surface properties of powders (Zimon, 1970); this effect is perhaps
due to surface condensation in the interstitial spaces, which might lead to sufficient water to
attenuate the UVGI. An effect on direct transmittance seems unlikely.

To investigate the effect of air humidity on irradiance, the radiometer UV detector was
positioned at a convenient location 1.36 m (4.46 ft) from the lamp centerline, at the center of the
duct. The longer path length from lamp to detector (relative to that used in the study of the
effects of flow rate) was deliberate. The same detector and filter arrangement was used. The test
rig was operated at 0.472 m/s (1,000 cfm) and 24C (74F), and the relative humidity varied
from 35 to 85% RH in three levels in a random order.

The results of this investigation are shown in Figure 10. The effect of humidity is modest but
detectable. Over this short path length, UVGI irradiance decreases linearly with absolute (not
relative) humidity, as would be expected for the direct interaction of water molecules and the UV
photons. References in the UVGI literature stating that the effect of humidity are unclear may be
the result of liquid droplet carryover in some field installations, which would result in much
greater attenuation of the UVGI. However, for droplet-free air, the effect of humidity appears
straightforward, predictable, and modest.

23
 Data 03-20-01 Irr=.000285-.00034*Humid
Lower 95% CI Upper 95% CI

2.86E-04
2.84E-04
2
Irradiance, W/cm

2.82E-04
2.80E-04
2.78E-04
2.76E-04
2.74E-04
0.00 0.01 0.02
Absolute Humidity, g H2O/ g dry air

Figure 10. Measured effect of absolute humidity on irradiance

3.1.2 Lamp and Ballast Design Effects

Lamp design effects have to do with the details of the mercury vapor discharge lamp. Lamp
pressure (Phillips, 1992), electrical current (Westinghouse, 1985), voltage, excitation wave form,
discharge ignition, and internal gas composition (Scheir and Fencl, 1998) are all mentioned as
parameters that can be adjusted to affect lamp output energy and spectrum. Ballasts supply the
electrical power to the UV lamps, and ballasts must be chosen to match the requirements of the
lamps. Conventional ballasts are transformers, with the advantages of low cost and the
disadvantages of low efficiency and weight. Solid-state ballasts are much more efficient and
lighter in weight, although they are more expensive.

The effect of lamp and ballast can be seen clearly by comparing Figures 8 and 9, which clearly
show the higher output for the lamps used in Figure 9. Lamp type and ballast type matters.
However, because no standard rating test is applied to the lamp/ballast systems, it is difficult for
a purchaser to make informed decisions. There are differences beyond the measured irradiance.
Both systems were powered by 110120 V alternating current (AC). The low output lamps,
driven by conventional magnetic ballasts, drew an average of 1.69 A (range from 1.65 to 1.76 A)
at 120 V AC and with no flow in test duct. The high output lamp, driven by solid-state
ballasts, drew 0.39 A at the same no-flow condition, and 0.53 A at 1.27 m/s (250 ft/min) and
21C (69 F.) The high output lamps were also claimed to have a higher power factor than
conventional lamps.

From Figure 8, we see that the maximum output for the low output lamps occurred at less than
0.508 m/s (100 ft/min) at 15C (58 F), at 0.635 m/s (125 ft/min) at 21C (70F), and at about
1.14 m/s (225 ft/min) at 30C (85 F.) The lamps tested are generally specified for use at about
1.27 m/s (250 ft/min) [0.472 m3/s (1,000 cfm) in the duct], so they are normally used at about

24
their maximum output point if the air is at 30C (85 F), but would be below their peak output at
15C (58 F.)

The high-output lamps whose irradiance is presented in Figure 9 were characterized by the
manufacturer as lamps designed to operate at lower temperatures than conventional lamps; they
appear to have an operating characteristic consistent with that design. At 15C (58 F), the
maximum output of about 2.50 x 10-3 W/m2 was reached at about 0.76 m/s (150 ft/min), while
the maximum at 21C (70 F) was reached at 1.27 m/s (250 ft/min) and the maximum irradiance
at 30C (85 F) was reached at about 2.03 m/s (400 ft/min.)

One of the factors about this lamp design that enables it to operate at lower temperatures may be
the higher power load being put into a lamp, relative to the physical size. The high power
lamps are thought to be low-pressure lamps operated at higher power loads than the low output
lamps. Phillips (1992) notes that medium-pressure lamps, operated at appropriate electrical
throughput, have a higher wall temperature and are more resistant to cool temperatures than are
low-pressure lamps.

3.1.3 Effects of Lamp Soiling

As with all lamps, germicidal UV lamps that become soiled suffer a reduction in the delivered
irradiance. Phillips (1992) comments that the lamps must be cleaned at a frequency depending
on the degree of grime in the air to which they are exposed, and the literature concurs. The
installation of lamps downstream of filters is generally recommended. Phillips Lighting (1985)
states that cleaning polishes may adsorb UVGI and recommends that they not be used to clean
and polish reflectors. Westinghouse (1985) recommends using alcohol or ammonia and water
and a clean, lint-free cloth for cleaning; leaving a residue should be avoided.

3.1.4 Reflectance
The importance of duct surface reflectance is that the effective irradiance delivered to the
microorganisms is increased through reflected UV energy. Reflectance specific to UVGI is not
the same as that to visible light. Phillips (1985) and others provide reflectance values for a
number of clean materials. Phillips Lighting (1992), in their design procedure for an axial lamp
mount in a circular duct, takes explicit credit for a reflective duct. They reduce the length of duct
necessary to treat air by (1-R), where R is the reflectance of the duct. Phillips (1985) also allows
credit for reflectance in their design procedure, though using a different and apparently more
conservative design correction. Westinghouse (1982) notes that reflective materials should be
used, but does not appear to take explicit credit for reflectance in their design procedure. Given
the high likelihood of eventual soiling in a ventilation system, caution is advised with regard to
taking credit for reflectance. Table 1 repeats some duct reflectance values from the literature.

25
Table 1. Reflectance of Various Materials
Material Reflectance
Aluminum, untreated surface 4060%
Aluminum, sputtered on glass 7585%
Stainless steel plate 2530%
Chromium plating 39%
Aluminum paint 4075%
Various white oil paints 310%
White baked enamel 510%
Magnesium oxide 7588%
White plastering 4060%
White note paper 25%
Typical duct liner 1%
Note: Reflectance values for a typical duct liner are taken from Schier and Fencl, 1996; values
for all other listed materials are taken from Phillips, 1985.

The values in Table 1 show that reflectance for UVGI is not the same as for visible light.
Aluminum is frequently mentioned as providing a desirable reflector surface. A reflectance
value for a galvanized steel duct was not found in the literature, but experience with UVGI lamps
in galvanized ducts indicates that galvanized steel has some reflectance for UVGI.

The intent of the reflectance measurements made during this research was to evaluate the impact
of reflectance on centerline irradiance measurements and the relative contribution of direct
UVGI and reflected UVGI on those measurements. The materials used were new galvanized
sheet metal and black photographic flock paper, and the basic measurement was irradiance as a
function of position in a mocked-up duct. Duct reflectance was investigated in the duct mock-up
apparatus by measuring the irradiance from low-power lamps with the duct lined with two
materialsnew galvanized steel and photographic background black flock paper. Three low-
power lamps were mounted horizontally across the duct: one at the center, and one each 20.3 cm
(8 in.) above and 20.3 cm (8 in.) below the center. Each lamp was driven by a magnetic ballast
and was allowed to reach an equilibrium output before measurements were started. All
measurements were made at the same static airflow conditions, and therefore the results can be
compared within this data set. Figure 6, above, shows the layout of the measurement points.
The same set of measurements was made for both duct lining materials.

Figure 11 shows the results of this measurement for the galvanized steel-lined duct, and Figure
12 shows the results for the black flock paper-lined duct. In both cases, the measurement points
were laid out as shown in Figure 6, with points 1, 2, and 3 being the row nearest the center, with
point 1 to the outside of the duct and point 3 nearest the center. The sawtooth shape of the
curves, most visible nearest the lamps, results from the point near the outside of the duct
receiving less irradiance than the points near the center. In both figures, the sawtooth shape is
more prominent near the lamps, because measurements near the lamps are dominated by
irradiance directly from the lamps, while farther away, more reflected light reaches the detector.

26
 2.00E-03

1.60E-03

Irradiance, W/cm 2 1.20E-03

8.00E-04

4.00E-04

0.00E+00
1 2 3 4 5 6 7 8 9
Measurement Point

30 cm 50 cm 75 cm 100 cm

Figure 11. Irradiance distribution in galvanized duct.

2.00E-03

1.60E-03
Irradiance, W/cm 2

1.20E-03

8.00E-04

4.00E-04

0.00E+00
1 2 3 4 5 6 7 8 9
Measurement Point
30 cm 50 cm 75 cm 100 cm

Figure 12. Irradiance distribution for black flock paper-lined duct.

The sawtooth is more prominent for the black flock paper-lined duct because essentially all
irradiance comes directly from the lamps.

Figure 13 shows the ratio of total irradiance in the galvanized duct to total irradiance in the black
flock paper-lined duct for all measurement points. It shows that reflected light contributes an
increased fraction of the total irradiance as distance from the lamp increases. Simultaneously,
because of the increase in distance, the total irradiance decreases. The straight line drawn in
Figure 13 is a fit to all the data. The fractional increase due to reflectance is substantial, with the
total irradiance increased by a factor of roughly 1.5 at 30 cm (11.8 in.) and by a factor of 3 at 100
cm (39.4 in.) Over the distance studied, the overall effect is near linear with distance from the
lamps.

27
 3.500

flocked paper-lined duct irradiance, %

Ratio of galvanized-lined duct to
3.000

2.500

2.000

1.500

1.000

0.500

0.000
0 10 20 30 40 50 60 70 80 90 100
Distance from lam ps, cm

Figure 13. Ratio of galvanized to black flock irradiance.

These results apply to lamps mounted across the duct in a crossflow configuration with a clean
galvanized duct. Reflectance effects should be different with the lamps mounted parallel to the
flow.

3.1.5 Enclosure and Location Factors

Objects that are physically in the path of the UV radiation provide shadows and protected
regions in which microorganisms will not be effectively irradiated. Blockage near the UV lamps
is especially important because the high-irradiance regions require the shortest exposure
duration.

Other location factors primarily relate to the temperature and airflow dependencies discussed
above.

3.1.5.1 Irradiance effects

Irradiance at a fixed distance from the UVGI lamps was routinely measured for most operating
conditions. Access was not available to make a measurement with the lamps oriented parallel to
the flow, but irradiance was measured for the crossflow configurations. The position of the
detector relative to the lamps depended on accessibility. Direct comparison is available for the
crossflow case with and without the coil with the detector at 50 cm (1.64 ft) upstream from the
lamps. Measurements were made for three high output lamps with the coil not installed in the
duct (the standard configuration), with the lamps between the coil and the detector (coil
downstream of the detector), and with the coil moved upstream to be between the lamps and the
detector. The air temperature was 22C (71F) and the relative humidity 5559%. Results are
given in Table 2.

28
Table 2. Irradiance Measurements with Coil in Duct
Condition Measured irradiance, W/cm2
No coil installed open duct 0.00252
Coil downstream 0.00260
Coil between lamp and detector (1) 0.00000198
Coil between lamp and detector (2) 0.00000642
Coil between lamp and detector (3) 0.00000161

The first two data rows of Table 2 show that the aluminum coil did not reflect a significant
amount of UVGI energy back to the detector; the irradiance measurement with and without the
coil is essentially the same. Relative to the top two rows, the three bottom rows show that less
than 1% of the UVGI that would be expected to reach the detector penetrates the coil. The cases
with the coil blocking the UV are minor variations of the same measurement. For (1), the
detector was located exactly at the duct centerline. For (2), the detector was slightly [1.9 cm
(0.75 in.) to change the view before making the measurement. For (3), UVGI leakage around
the coil was essentially stopped by sealing the visible light leaks around the coil. Comparison of
(1) and (2) shows that the irradiance measurement is sensitive to detector position with the coil
installed, though most UVGI is stopped in both cases. Light reflecting around the outside edges
of the coil is shown to be a substantial fraction of the total UVGI measured by (3).

3.1.5.2 Results for microbial kill in various configurations

Table 3 compares disinfection efficiency for the four lamp/duct configurations with the same
number of lamps in each. Using two different microbial challenge aerosols (Bacillis subtilis and
Aspergillis versicolor), penetration was evaluated with and without the lamp bank turned on.
Staphylococcus epidermidis was not used as a challenge because three lamps were expected to
produce too high a dose to provide useful data.

The data were analyzed by comparing results for each alternative arrangement to the crossflow
lamp arrangement results; this analysis was required because the alternative configurations were
not compatible with the previously developed model that had been used to estimate dose.
Comparison using the model to compute k values was not possible because the model was
developed for the case of a straight, open duct and was of uncertain usefulness for the other
cases.
Table 3. Effect of Duct Configuration on Microbial Kill

Lamp Number of Bacillis subtilis Aspergillis versicolor

Configuration Lamps % Kill % Kill
Straight duct, 6 99 56
crossflow lamps 3 76 20
Straight duct, 6 67 40
parallel lamps 3 30 17
Straight duct, 6 92 43
coil 3 55 23
6 91 62
Elbow duct
3 55 30

29
3.2 Chemical Byproducts Results and Discussion

3.2.1 Ozone
The results of the ozone production measurements for six high-output lamps mounted in
crossflow at 0.76 m/s (150 ft/min) were nil. The ozone concentration downstream of the lamps
was not different from that upstream. In both cases, the readings were about 28 ppb. Therefore,
under flow conditions, ozone production was not detectable.

3.2.2 Effect of UV lights on VOCs

The experiment used was described in Section 2.3.2. As with the ozone production experiment,
the test was conducted at a normal operating flow rate, not in still air. The results show that no
significant change in the concentration of the injected chemicals was detected, and no chemical
byproducts were clearly detected. Figure 14 includes the results from two upstream and two
downstream tests with the UV lights on; it visibly shows that the upstream and downstream tests
were not different. Table 4 shows the chemicals identified in byproduct sampling.

1000

100
ug/m^3

10

0.1
0 5 10 15 20 25 30 35 40
Compound
Up1 Dwn1 Up2 Dwn2

Figure 14. Results of chemical sampling up- and downstream of lamps

30
Table 4. Chemicals Identified in Byproduct Sampling
ID Chemical Name CAS No ID Chemical Name CAS No
1 1-Hexanol, 2-ethyl 104-76-7 21 Nonyl aldehyde (Nonanal) 124-19-6
2 1-Pentene, 4-methyl 691-37-2 22 Octanal* 124-13-0
3 1- Propanol, 2-methyl (Isobuty 78-83-1 23 Pentane, 2-methyl 107-83-5
alcohol)
4 2-Butanone (Methyl ethyl ketone, 78-93-3 24 Pentane, 3-methyl 96-14-0
MEK)
5 2-Propenal, 2-methyl* 78-85-3 25 (2,6,6-Trimethyl-bicyclo[3.1.1] 80-56-8
hept-2-ene
6 Acetic acid 64-19-7 26 Propanal, 2-methyl (Isobutanal)* 78-84-2
7 Benzaldehyde 100-52-7 27 Toluene (Methylbenzene) 108-88-3
8 Benzene 71-43-2 28 Tridecane 629-50-5
9 Benzene, 1,2,4-trimethyl 95-63-6 29 Xylene (para and/or meta) 106-42-3
10 Benzene, ethyl 100-41-4 30 Xylene, ortho 95-47-6
11 Benzoic Acid* 65-85-0 31 2-Propanol (Isopropanol) 67-63-0
12 Cyclopentane, methyl 96-37-7 32 3-Buten-2-one* 78-94-4
13 Decanal 112-31-2 33 1-ethyl-4-methyl (4-Ethyltoluene) 622-96-8
14 Decane 124-18-5 34 Heptanal (Heptaldehyde) 111-71-7
15 Dodecane 112-40-3 35 Heptane, 2,4-dimethyl 2213-23-2
16 Ethanol 64-17-5 36 Pentane, 2,2,3,4-tetramethyl* 1186-53-4
17 Ethanone, 1-phenyl (9CI) 98-86-2 37 3-Buten-2-one, 3-methyl* 814-78-8
18 Hexane* 110-54-3 38 Acetone 67-64-1
19 1-Methyl-4- 138-86-3 39 Butane, 2-methyl (Isopentane) 78-78-4
(1methylethyl)cyclohexene)
20 Methylene chloride 75-09-2
(Dichloromethane)

31
3.3 Microbiological Factors Results and Discussion

Dosage is defined as the UVGI irradiance received by the microorganism times the duration of
exposure. This is the Eeff
t product of Eq. 1. Eeff is a function of duct position. The most
common unit for dosage is Ws/cm2, but Ws/cm2 and J/cm2 are also used.

Given the theoretical ability to compute particle trajectories, the irradiance received by each
microorganism particle of each size at any given time appears simple, at least in concept, to
predict the disinfection performance of a UVGI installation. To compute the dose, Eeff
t can
be integrated over the particle path, and the fraction of microorganisms killed can then be
estimated. But numerous microbiological considerations must be taken into account.

Although there are multiple tables available that list a series of organisms and the respective
dosages or energies required to kill them, a careful review of the literature shows that these
tables may unintentionally oversimplify the required dosages. Many of the data in these tables
come from the Westinghouse data, which are based on a series of experiments done primarily in
the 1940s. Many of these are truly elegant experiments; however, their goal was a fundamental
understanding of UV and not an evaluation of UVGI lamps. More recent summaries of UVGI
kill (Kowalski and Bahnfleth, 2000) have perpetuated the problem. There were few common
conditions under which the experiments were performed (i.e., some organisms were irradiated
while airborne, others when seeded on agar plates or suspended in buffer). These differences
may explain a great many of the seemingly inconsistent kill dosages reported for the same
organism.

3.3.1 Test Organism Selection

Preliminary microbiological tests were conducted with seven organisms: Serratia marcescens,
Staphylococcus epidermidis, Pseudomonas fluorescens, Bacillus subtilis, Aspergillis versicolor,
Penicillium chrysogenum, and Cladosporium sphaerospermum. The first four are bacteria, and
the last three fungi. These organisms cover a broad range of susceptibility to UV.
S. marcescens, S. epidermidis, and P. fluorescens are vegetative bacteria that are readily killed.
B. subtilis is a spore-forming bacteria, and the three fungi (A. versicolor, P. chrysogenum, and
C. sphaerospermum) also form spores. We expected these organisms to be difficult to kill. Most
of the tests were conducted at approximately 24 C (75F), 50% RH, with the test rig at either
0.472 or 0.694 m3/s (1,000 or 1,470 cfm) for an approach velocity of 1.27 or 1.87 m/s (250 or
368 ft/min.) All of the tests utilized one or more standard UV lamps [nominal 1.6-cm (5/8-in.)
diameter, single-ended, 61-cm (24-in.) length] powered by a large and heavy preheat, self-start
ballast and mounted perpendicular to the airflow.

Depending on the experiment, the lamps were configured in one of four different ways. The goal
was to irradiate all the organisms and remain in an experimental range with reliably measurable
kill. For the three fungi and the B. subtilis, we used either six lamps in a single column on
10.2-cm( 4-in.) spacing or three lamps in a single column on 20.3-cm (8-in.) spacing for high and
low irradiance levels, respectively. For the three vegetative bacteria, we used either one lamp at
the duct center or one lamp covered by three layers of common building screen. In each case,
the irradiance was measured at a known distance from the lamp (on the lamp centerline), and that
irradiance value was used to calibrate the model. The model was then used to compute the mean

32
dose for all the trajectories, and the k factor was computed from this mean dose and measured
microbial kill. A given lamp configuration produces a higher dose at 0.472 m3/s (1,000 cfm)
than at 0.694 m3/s (1,470 cfm), as is reflected in Table 5. This large duct experiment is not a
preferred way to determine k factors because the distribution of dose is fairly broad across the
different trajectories. Computing k values allowed us to compare our full-scale work to the
published literature; the agreement is fairly good in most cases.

As can be seen in Table 5, with this range of irradiance, all organisms can be brought into the
measurable range. Each result is for an individual test or run and they are generally in good
agreement. Use of the screened lamp was not fully satisfactory, but it dramatically demonstrated
the susceptibility of some organisms and the need for the microbiologically informed testing that
was conducted in this project.

Table 5. Results of Preliminary Microbial Kill Experiments

Number of Dose
Organism % Kill k Factor
Lamps W-sec/cm2
1 lamp 1341 99 3.31E-03
Serratia marcesens 1 lamp, 3 screens 369 97 9.16E-03
1 lamp, 3 screens 369 98 9.99E-03
1 lamp 984 81 1.68E-03
1 lamp 984 83 1.77E-03
Staphylococcus epidermidis
1 lamp, 3 screens 329 22 7.71E-04
1 lamp, 3 screens 329 9 2.99E-04
1 lamp 1,341 98 2.90E-03
Pseudomonas fluorescens
1 lamp, 3 screens 369 99.7 1.58E-02
6 lamps 7,509 80 2.17E-04
6 lamps 7,509 90 3.06E-04
Bacillus subtilis
3 lamp 3,197 48 2.07E-04
3 lamp 3,197 53 2.34E-04
6 lamps 7,509 83 2.35E-04
6 lamps 7,509 66 1.45E-04
Aspergillus versicolor
3 lamps 3,197 5 1.51E-05
3 lamps 3,197 16 5.49E-05
6 lamps 7,509 10 1.40E-05
6 lamps 7,509 17 2.46E-05
Penicillum chrysogenum
3 lamps 3,197 0.00
3 lamps 3,197 1 4.57E-06
Cladosporium 6 lamps 7,509 16 2.39E-05
sphaerospermum 3 lamps 3,197 3 8.24E-06
Note: 3 screens indicates that the single lamp was wrapped with three layers of household
screen to reduce irradiance.

The experimental conditions achieved were adequate to conduct useful tests with all of the
microbial organisms. That is, the results fell within a useful experimental range and provided the

33
needed information. The vegetative bacteria (S. marcescens, S. epidermidis, and P. fluorescens)
were readily killed. The spore-forming bacteria (B. subtilis) was, as expected, much more
resistant. The three fungi whose spores were irradiated (A. versicolor, P. chrysogenum and
C. sphaerospermum) were also relatively resistant, although significant kill was achieved by
arranging sufficient UVGI irradiance. The results of the experiments are in general agreement
with previously published microbial kill data and design information. Therefore, we selected the
vegetative bacteria S. epidermidis, the spore-forming bacteria B. subtilis, and the fungus
A. versicolor for the final testing of the lights.

3.3.2 Growth of Organisms in the Dark

All of the plates collected during the testing were incubated in the dark. UV light kills cells by
damaging their deoxyribonucleic acid (DNA). The exact mechanism differs between vegetative
cells and spores. Exposure to UV causes the formation of thymine dimers by
phototransformation in vegetative cells and by spore photoproduct in spores. This results in a
distortion of the DNA strand; replication is then halted. Because cells are commonly exposed to
UV, they have developed a number of repair mechanisms to counteract the UV-induced damage.

Although the phenomenon of UV repair was reported in the late 1940s and 1950s (Kelner,
1949; Dulbecco, 1955), little was understood about the cellular repair mechanisms for UV
damage until the 1960s. Fungal spores, bacteria, and plant, animal, and viral DNA may all be
repaired through photoreactivation with near-UV or visible light. Photoreactivation is only one
of the repair mechanisms available to cells to repair damaged DNA. In addition to light repair,
excision repair, also known as dark repair, and the so called SOS repair mechanism are
important. Spores have an additional mechanism of spore photoproduct repair (Setlow, 1992).
The enzyme spore photoproduct lyase is activated during germination to repair damage that
occurred to the dormant spore (Sun et al., 1994). Excise repair is exactly what it sounds like: the
cell cuts out the segment of DNA with the thymine dimer and synthesizes the new strand to
replace it. The SOS mechanism is particularly interesting because it results in the
misincorporation of DNA bases, which leads to an increased mutation rate for the cell. This
mechanism is responsible for the high mutation rate seen in cells irradiated with UV. Scientists
have used UV to induce mutations in cells for many years. Mutated cells are by no means dead;
while some mutations are lethal, most are not. It is thought that one explanation for the
evolutionary development of the SOS mechanism is that at a cellular level, mutation is preferable
to death. In human cells, this may assist in the development of skin cancers. However, in
microorganisms, some of these mutations may be beneficial or neutral to the cell. The dosages
required to kill an organism are affected by these repair mechanisms, and the dominant
mechanism for a cell can profoundly affect an organisms response to UVGI. An additional
repair mechanism exists for viruses. The infected cell may repair the damaged viral DNA for the
virus if the infected cell contains the repair enzymes.

3.3.3 Recovery of Damaged Test Organisms

The selection of the appropriate recovery media was another factor that needed to be addressed.
Few of the early experiments that are frequently cited for determining the kill dosages of a
specific organism address the fact that damaged organisms will not grow well on standard media

34
after irradiation or any other antimicrobial treatment. The standard bacteriological and
mycological media are generally considered too nutrient rich to effectively recover the damaged
organism. Recovery of environmentally stressed organisms using minimal nutrient media can be
more accurate.

Damaged organisms are notoriously difficult to isolate on standard nutrient media. Therefore,
the selection of the recovery media was thought to be an important factor in this study. We used
media typically used to recover damaged organisms and compared the results with the more
standard media. No differences were found for recovery, so trypticase soy agar was used for the
bacteria and Sabaurouds dextrose agar was used for the fungi.

3.3.4 The Influence of Aerosol Composition on Bioaerosol Tests

Most of the UV inactivation studies done with airborne microorganisms (bioaerosols) have used
pure cultures of one particular organism that were suspended in a liquid diluent and then
nebulized. The overall composition of the aerosol, including the bioaerosol, is important.
Bacteria suspended in phosphate buffer were more susceptible to UV inactivation than those
suspended in broth (Sharp, 1940). Presumably, the more complex organic constituents of broth
provide a physical protection or shield from the UV rays. Also, bacteria in broth are less stressed
and may be naturally more resistant. On the other hand, Beebe (1958) found that in some cases,
dry aerosols of the organism were more resistant to UV than wet (nebulized liquid) aerosols.

Other aerosol characteristics are also important. Bacteria in droplet nuclei are more readily
killed than aggregates. Dust or other particulates associated with the bioaerosol decrease the
efficacy of the UV because of the UVs poor penetration.

Because our goal was to develop a method to test and compare the lights in a test duct, some of
these issues needed to be addressed. To address the influence of the confounding/potentially
protective aerosol characteristics, the test organisms were generated in both water and broth as
described in section 2.4.3. The water media was used as the baseline, and the broth for all of the
potentially protective circumstances.

3.3.5 The Influence of Relative Humidity on Test Organisms

There is little or no agreement in the literature regarding an organisms sensitivity to UV and the
moisture content of the cell. Attempts to correlate the susceptibility of an organism to UV at
different RHs have yielded inconsistent results. Some organisms have shown decreased kill at
RHs greater than 6070%, while others have not. Studies with S. marcescens showed decreased
kill at RHs both greater than 70% and below 40%.

The ducts RH did not appear to affect particle viability (the upstream and downstream counts
with the lamps off were about the same) or particle size (by optical particle counter) between the
upstream and downstream measuring points. Because physical equilibrium with water vapor is
achieved quickly for micrometer-sized particles, this is not surprising.

35
To help elucidate the RH effect on organisms within the confines of this project, experiments
were performed at 55% RH and 85% RH with both water and broth suspensions.

4.0 UVGI EFFICACY IN AIR DUCTS

4.1 Measured Efficacy and Discussion

During this research, UVGI lamps were found to inactivate vegetative bacteria, bacteria spores,
and fungal spores to a reproducible degree under conditions of fixed dose. Further, Eq. 1 was
found to generally apply, provided the dose and the k value were known. This section
summarizes the bioaerosol measurements that were the key findings of the research.

Table 6 summarizes the final results for our three primary test organisms. The test organisms
were suspended in either water or broth to examine the impact of the potentially protective effect
of organic matter (dirt, debris, etc.) on the organisms in these tests. The experiments were
performed at either 55% RH or 85% RH to assess the impact of the potentially protective effect
of humidity on the organisms within the confines of this test method. While the focus of this
project was not to elucidate the nature of these organisms, understanding the various effects was
critical in developing the test and interpreting the results.

36
Table 6. Percent Kill of the Three Test Organisms at Varying Test Conditions

Lamp Water Suspension Broth Suspension

Organism
Intensity 55% RH 85% RH 55% RH 85% RH
# lights # lights # lights # lights

Staphylococcus 1 <1 1 <1 1 <1 1 <1

epidermidis Low 99.1 69 74 15 86 19 63 0
High > 99 83 82 25 ND ND ND ND
# lights # lights # lights # lights
6 3 6 3 6 3 6 3
Bacillus subtilis
Low 83 32 83 35 84 49 89 46
High 99 76 98 75 99 83 99 85
# lights # lights # lights # lights

Aspergillus 6 3 6 3 6 3 6 3
versicolor Low 23 7 36 15 47 25 47 14

High 56 22 74 28 ND ND ND ND
ND = Not Done

4.1.1 Vegetative Organisms

The results suggest that the vegetative organisms, represented by S. epidermidis, became more
resistant to UV as the protective factors increased. Additionally, the results suggest that the
effect may be additive for the vegetative organisms. When comparing a 99.1% kill with one
light at 55% RH to a 74% kill with one light at 85% RH, as seen in Table 6, increased humidity
appeared to be protective. A similar result was seen for the organisms generated in the broth
suspension. At 55% RH, there was an 86% kill with one light, but at 85%, there was a 63% kill.
The increased amount of organic matter in the broth also appeared to be protective. And where
both elevated humidity and organic matter (85% RH, broth suspension) were present, only 63%
of the organisms were killed with one light.

4.1.2 Spore-Forming Bacteria and Fungi

Humidity and suspension composition had much less impact on the spore-formers. It should be
noted that six lights were required to inactivate the spore-formers to levels similar to those seen
with the vegetative bacteria. As stated before, the spore-formers were chosen because they are
much more difficult to kill. For the B. subtilis spores, there did not appear to be a difference in
the percentage killed, regardless of protective factor. For the fungal spore, A. versicolor, there
may have been a small effect. The difference may not be significant in practice; we will
investigate it further as part of the EPA supplemental project. However, this difference is very
interesting because it may point to some fundamental properties of microbial aerosols that cannot
be readily investigated, except in the aerosol state. When microbial aerosols are in contact with

37
agar, as has been the case in most of the reported literature, it would not be possible to
investigate the impact of air RH. Additional tests with a dilute protein solution are under way.
These tests will help elucidate the importance of humidity on kill for the different classes of
organisms.

It is critical to understand the effect of humidity on the various organisms. The literature is
contradictory on the subject. The physical factors work has shown that the lights are not
affected. This makes it critical to clarify the impact of humidity on the organisms.

4.1.3 Summary of Calculated Microbial Resistance Constants

One of the more confusing aspects of comparing the effectiveness of different UV lights is that
manufacturers report different dosages, and therefore, correspondingly different kills. As can be
seen in Table 6, the impact of the high-intensity lights was that a higher percentage of organisms
was killed due to the higher dose. As a check on the test method we developed, it was necessary
to compare our results to those of the same or similar organisms reported in the literature. To
separate out organism effects from light effects, we calculated the k value.

The k value was calculated as shown in Eq. 1 and is defined as the microorganism-dependent
rate constant. Table 7 shows the k values (cm2/mW-s) calculated from data reported in Table 6
for the three primary organisms in the study at both 55% RH and 85% RH. Only the data from
the tests in which the organisms were suspended in water were reported.

Table 7. Summary of k Values for Test Organisms

k values, cm2/mW-s, at indicated humidity

Test Organism
55% RH 85% RH
Staphylococcus epidermidis 2.4 2.0 (n=18) 0.8 0.2 (n=12)

Bacillus subtilis 0.2 0.06 (n=14) 0.2 0.06 (n=14)

Aspergillus versicolor 0.03 0.02 (n=12) 0.06 0.03 (n=12)
n = number of tests

As anticipated, the k values for the three test organisms demonstrate that the vegetative bacteria
was the most readily killed, while the fungal spore was the most difficult to kill. The values we
determined are similar to those reported in the literature. The literature-reported values for
organisms similar to S. epidermidis range from 0.4 to 1.9 cm2/mW-s. For B. subtilis, the values
range from 0.1 to 0.9 cm2/mW-s. The reported k values for various Aspergilli range from 0.01 to
0.05 cm2/mW-s.

4.2 Design of UVGI Systems for Air Ducts

Most of the guidance available from the published literature and manufacturers of UVGI systems
appears to have its origin in the design guidance provided by the UV lamp manufacturers
Westinghouse (1982) and Phillips (1985, 1992). The guidance is based on the validity of Eq. 1,
detailed knowledge of lamp performance under different conditions, and use of either a variation

38
on the inverse-square law or the view factor equation to estimate the effective irradiance in the
space. With the caveat that the microbial resistance constants k should not be taken to have great
precision, the present research generally confirms these design methods. Conversations with
various manufacturers have indicated that some have independently developed models that have
the general characteristics of the model we developed to support this research.

Any design method must begin with knowledge of the output of the UVGI lamps at the expected
operating conditions. For branded lamps, that information is available (at least in part) from the
manufacturer. For an unbranded lamp, few assumptions can be made; the irradiance must be
measured at operating conditions. The design models can then be used.

5.0 CONCLUSIONS

We can draw several conclusions from our studies.

The key factors for the design of in-duct UVGI systems intended to disinfect moving air streams
are (1) UV lamp irradiance at operating conditions, (2) microbial resistance data, (3) functioning
bioaerosol dose model and (4) duct wall reflectance.

UVGI lamps have a point of maximum output when exposed to various air temperatures and
airflow rates. This optimal operating point is fundamentally related to lamp operating
temperature and can be reached by various combinations of environment and heat transfer rate
(principally air temperature and flow rate) that bring the lamp to the proper operating condition.

Different UVGI lamps have different characteristics, and the optimum cooling rate may be
different for each.

The reduction in output due to overcooling at temperatures achievable in an HVAC system has
been shown to be as much as a factor of two, which could have a significant effect on kill for
bioaerosols.

The variance in k values reported in section 4.1.3 for the present work (results reported as mean
and standard deviation) and the ranges of reported k values should be considered to be measures
of real variability in microbial resistance until proven otherwise. Humidity and organic matter
were shown in sections 4.1.1 and 4.1.2 to have protective effects that would be difficult to
distinguish from natural variation in an environmental organism, and significant variation should
be expected. Resistance variations between strains of the same organism have the potential to be
substantial. Very little work has been done. In consequence of this, UVGI system designers
should

Apply safety factors to their designs, particularly as they depart from operating modes
with which they have performance data and field experience.
Avoid relying solely on design equations to determine the performance of their systems.
Actual testing with the contaminants of interest is highly recommended.
Be extremely cautious regarding claims about UVGI systems high levels of inactivation
of pathogenic bioaerosols. While the microbiological science underlying these

39
 conclusions is thought to apply to pathogenic bioaerosols as well as environmental
organisms, much greater caution is required in the former case. Claiming a high
inactivation rate for a pathogenic bioaerosol without substantial testing would be
irresponsible. Even with substantial testing, design failures may occur.

6.0 RECOMMENDATIONS

Because lamp output varies so strongly with temperature and airflow rate, lamp output
measurements should either be made at the expected use conditions or extrapolated to those
conditions.

In the efforts to counter bioterrorism, the use of UVGI is likely to be proposed with increasing
frequency. It is a low-pressure-drop, nonintrusive technology that has real potential. In that
light, we recommend that test methods and guidance for performance claims regarding
bioterrorism agents be developed by an experienced and broadly based group of technical
experts, users, and manufacturers. Until rigorous and adequate tests have been developed and
performed, claims regarding protection against aerosol bioterrorism agents are suspect. The
potential errors are too large and the consequences too great.

7.0 REFERENCES
ASHRAE. 1999. Standard 52.2-1999: Method of Testing General Ventilation Air-Cleaning
Devices for Removal Efficiency by Particle Size. American Society of Heating,
Refrigerating, and Air-Conditioning Engineers, Inc. Atlanta, GA.

Beebe, James M. 1958. Stability of disseminated aerosols of Pasteurella tularensis subjected to

simulated solar radiation at various humidities. Journal of Bacteriology 78:18-24.

Dulbecco, R. 1955. Photoreactivation. In Radiation Biology. Vol. II. A. Hollaender, Ed.

New York: McGraw-Hill. pp. 455-486.

CDC. 1994. Guidelines for Preventing the Transmission of Mycobacterium Tuberculosis in

Health-Care Facilities. Centers for Disease Control and Prevention, Atlanta, GA.

CDC. 1999a. Guidelines for the Application of Upper-Room Ultraviolet Germicidal Irradiation
for Preventing Transmission of Airborne ContagionPart I: Basic Principles. Centers
for Disease Control and Prevention, Atlanta, GA.

CDC. 1999b. Guidelines for the Application of Upper-Room Ultraviolet Germicidal Irradiation
for Preventing Transmission of Airborne ContagionPart II: Design. Centers for
Disease Control and Prevention, Atlanta, GA.

Kelner, A. 1949. Effect of visible light on the recovery of Streptomyces griseus conidia from
ultraviolet irradiation injury. PNAS 35:73-79.

40
Kowalski, W. J., W. P. Bahnfleth, D. L. Witham, B. F. Severin, and T.S. Whittam.
Mathematical modeling of ultraviolet germicidal irradiation for air disinfection.
Quantitative Microbiology 2:249-270.

Kowalski, W. J., and W. Bahnfleth. 1998. Airborne respiratory diseases and mechanical
systems for control of microbes. Heating/Piping/Air Conditioning. July. p34-46, 48.

Miller, S.L., and J. M. Macher. Evaluation of a methodology for quantifying the effect of room
air ultraviolet germicidal irradiation on airborne bacteria. Aerosol Science and
Technology 33:274-295.

Phillips Lighting Division. 1985. Germicidal lamps and applications. Booklet

Phillips Lighting Division. 1992. Disinfection by UV-radiation. Booklet 3222 C34 00671.

Ryer, A. D. 1997. Light Measurement Handbook. Available at http://www.intl-

light.com/handbook/. Newburyport, Maine: International Light, Inc.

Scheir, R., and F. B. Fencl, 1996. Using UVGI technology to enhance IAQ. Heating/Piping/Air
Conditioning 68:109-114, 117-118, 123-124.

Setlow, P. 1992. I will survive: Protecting and repairing spore DNA. Journal of Bacteriology
174:2737-2741.

Sharp, D. Gordon. 1940. The effects of ultraviolet light on bacteria suspended in air. Journal of
Bacteriology 38: 535-547.

Shechmeister, I. L. 1991. Sterilization by ultraviolet radiation. In Disinfection, Sterilization,

and Preservation. S. S Block, Ed. Philadelphia: Lea & Febiger, pp. 553-565.

Sun, Y., K. Palasingam, and W. Nicholson. 1994. High-pressure liquid chromatography assay for
quantitatively monitoring spore photoproduct repair mediated by spore photoproduct
lyase during germination of UV-irradiated Bacillus subtilis spores. Analytical
Biochemistry 221:61-65.

Westinghouse Electric Corp. 1982. Booklet A-8968.

41