&EPA UnrtedStates EnviionnwntBi Protection Afrnoapheric Research and Exposure AssMSfiMnt L&boFBtofy RMMR*) Tnangk. Park, NC 27711 January 1994 Research and Dovdopment EPA 600/R-94/016 A Preliminary Evaluation of the Scitec MAP-3, Warrington Microlead I, and Princeton Gamma-Tech XK-3 Portable X-Ray Fluorescence Spectrometers ------- January 1994 U.S. EPA OPPTS Chemical Library EPA West Room 3379 MC7407T 1200 Pennsylvania Ave. NW Washington DC 20460-000! A Preliminary Evaluation of the Scitec MAP-3, Warrington Microlead I, and Princeton Gamma-Tech XK-3 Portable X-Ray Fluorescence Spectrometers Prepared by E. D. Estes D. L. Hardison C. O. Whitaker J. D. Neefus W. F. Gutknecht Center for Environmental Measurements and Quality Assurance Research Triangle Institute Research Triangle Park, North Carolina 27709-2194 EPA Contract No. 68-D1-0009 RTI Project No. 91U-6960-187 Prepared for Ms. Sharon Harper, Work Assignment Manager Atmospheric Research and Exposure Assessment Laboratory Office of Research and Development U.S. Environmental Protection Agency Research Triangle Park, North Carolina 27711 ------- DISCLAIMER The information in this document has been funded wholly or in part by the United States Environmental Protection Agency (USEPA) under EPA Contract No. 68- Dl-0009 to the Research Triangle Institute. It has been subjected to the Agency's peer and administrative review, and it has been approved for publication as an EPA document. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. ------- ACKNOWLEDGEMENTS This document was prepared under the direction of Ms. Sharon L. Harper, Atmospheric Research and Exposure Assessment Laboratory (AREAL), U.S. Environmental Protection Agency (USEPA), Research Triangle Park, NC. Special acknowledgement is given to Mr. Jack Suggs, AREAL/USEPA, Research Triangle Park, NC, Mr. John Scalera, Office of Pollution Prevention and Toxics, U.S. Environmental Protection Agency, Washington, DC and Dr. Mary McKnight, National Institute of Standards and Technology, Gaithersburg, MD, for their careful review. n ------- EXECUTIVE SUMMARY The potential impact on health from environmental lead has resulted in increased interest in lead exposure by Federal, State and local government agencies. As a result, programs committed to sampling and analysis of lead are increasing nationwide. Public housing authorities are required, by 1994, to randomly inspect all their housing projects for lead-based paint. Currently, the most common approach to screening housing for the presence of lead in paint is the use of the portable X-ray fluorescence (XRF) detector, which is relatively inexpensive (0.1 - 0.2 hours/test for labor), gives rapid results, and can be used with minimal damage. Inconclusive XRF measurements must be confirmed with field-collected samples back in the laboratory using a more accurate analytical method such as atomic absorption spectrometry (AAS) or inductively coupled plasma emission spectrometry (ICP). The accuracies and precisions of portable XRF instruments for measurements of lead-in-paint concentrations near the U.S. Department of Housing and Urban Development (HUD) action level for abatement (1.0 mg/cm2) have not been well established. Therefore, at the request of the U.S. Environmental Protection Agency (USEPA), the Research Triangle Institute (RTI) performed a limited evaluation of three portable XRFs currently being used for field measurements: the Scitec MAP-3, the Warrington Microlead I, and the Princeton Gamma-Tech (PGT) XK-3. The purpose of the evaluation was to compare these very different, commercially available portable XRFs, on as equal a basis as possible, for their biases and precisions when measuring lead in paint films at concentrations ranging from 0 to 7.16 mg/cm2. Each instrument was used to make a series of measurements in order to determine the effects of sample concentration and substrate on the precision and bias. The samples chosen for the study were standard latex and oil-based paint films prepared by RTI. The lead concentrations of the seven paint films ranged from nominally 0 to 7.16 mg/cm2 as confirmed by digestion using a modification of NIOSH Method 7082 and analysis by inductively coupled plasma emission spectrometry. Eight different substrates of varying densities were chosen to represent the types of building materials likely to be encountered in the field. The measurement plan called for triplicate 111 ------- measurements of each paint film on each substrate using each of the three portable XRF instruments according to the manufacturer's instructions. The order of measurement for the Warrington and PGT instruments followed a randomized test matrix. Because of the availability of the instrument, the measurements for the Scitec XRF were completed before the test matrix was developed and therefore were not performed in the same order. Based upon the data collected during this limited evaluation, it appears that all three instruments measure lead in paint more accurately (as determined by inspection of bias) on 1/2" plasterboard than on the other substrates tested. Predicted biases for the Scitec XRF across the test sample concentration range varied from -0.6 mg/cm2 at 1.6 mg/cm2 for 5" concrete to 3.3 mg/cm2 at 0.3 mg/cm2 for solid tinderblock Predicted biases for the Warrington XRF ranged from -0.03 mg/cm2 at 1.6 mg/cm2 for plasterboard to 1.1 mg/cm2 at 0.3 mg/cm2 for concrete. Predicted biases for the PGT XRF ranged from 0.1 mg/cm2 at 0.3 mg/cm2 for 4" cinderblock to -0.9 mg/cm2 at 1.6 mg/cm2 for 1/8" aluminum. IV ------- TABLE OF CONTENTS Section Page Disclaimer i Acknowledgements ii Executive Summary iii Table of Contents v List of Tables vii List of Figures ix 1.0 Introduction 1 1.1 Background 1 1.2 Objectives/Purpose of the Study 2 1.3 Study Design 2 2.0 Overview of the Portable XRF 5 2.1 Instrument Basics 5 2.2 Available Instruments 7 3.0 Laboratory Tests of the Portable XRF Units 9 3.1 Introduction 9 3.2 Scitec Map-3 9 3.2.1 Description 9 3.2.2 Experimental Procedure 11 3.3 Warrington Microlead I 11 3.3.1 Description 11 3.3.2 Experimental Procedure 11 3.4 Princeton Gamma-Tech XK-3 12 3.4.1 Description 12 3.4.2 Experimental Procedure 13 ------- TABLE OF CONTENTS (continued) Section Page 4.0 Results and Discussion 14 4.1 Experimental Precision and Bias 14 4.2 Linear Regression Analyses 14 4.2.1 Hypothesis Testing 18 4.2.2 Confidence Intervals 27 4.3 Predicted Bias at Selected Lead Concentrations 27 4.4 Scitec Map-3 31 4.5 Warrington Microlead 31 4.6 Princeton Gamma-Tech XK-3 32 5.0 Conclusions and Recommendations 33 5.1 Conclusions 33 5.2 Recommendations 33 6.0 References 35 Appendix A: XRF Data VI ------- LIST OF TABLES Table Page 1 Randomized test matrix - order (1-56) for testing each combination of sample and substrate 3 2 Test procedures 4 3 Portable XRF characteristics 10 4 Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: Scitec 15 5 Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: Wanington 16 6 Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: PGT 17 7 1/2" Plasterboard 19 8 3/4" Plywood 20 9 1/8" Aluminum 21 10 1/8" Steel 22 11 4" Cinderblock 23 12 Solid cinderblock 24 vn ------- LIST OF TABLES (continued) Table Page 13 Brick 25 14 5" Concrete 26 15 HUD Guidelines for Pb testing using portable XRF instrumentation 29 Vlll ------- LIST OF FIGURES Figure Page 1 Comparison of experimentally generated slopes for portable XRF/substrate *y combinations across the lead concentration range of 0 to 7.16 mg/cm 28 2 Comparison of predicted biases at 1 mg Pb/cm2 for portable XRF/substrate combinations 30 IX ------- SECTION 1 INTRODUCTION 1.1 BACKGROUND The adverse health effects resulting from exposure of young children to environmental lead have caused increased public concern in recent years. Studies have shown that chronic exposure even to low levels of lead can result in impairment of the central nervous system, mental retardation, and behavioral disorders.1'2 Although young children are at the greatest risk, adults may suffer harmful effects as well.3 Numerous programs involving research on the toxicity and bioavailability of lead, the environmental monitoring of lead, lead abatement, and clearance testing of housing are currently active. Examples of past projects include the U.S. Department of Housing and Urban Development (HUD) National Survey,4 the Urban Soil Lead Abatement Demonstration Project5 and method evaluation studies sponsored by the U.S. Environmental Protection Agency (EPA), and abatement programs carried out by Maryland, Massachusetts, and many other groups. The major sources of exposure to lead in housing units are paint, dust, and soil. Currently, lead-based paint is receiving emphasis by HUD6 as the principal source for lead contamination and exposure. House dust contaminated with lead from deteriorating interior paint and from tracked-in exterior paint has been implicated as the most common route of exposure for young children.7 It is particularly hazardous when painted walls, woodwork, and furniture are accessible to young children to touch and to chew. Public housing authorities are required, by 1994, to randomly inspect all their housing projects for lead-based paint.8 Currently, the most common approach to screening housing for the presence of lead in paint is the use of the portable X-ray fluorescence (XRF) detector, which is relatively inexpensive (0.1 - 0.2 hours/test for labor), gives rapid results, and can be used with minimal damage.9'10 Inconclusive XRF measurements must be confirmed with field-collected samples back in the laboratory using a more accurate analytical method such as atomic absorption spectrometry (AAS) or inductively coupled plasma emission spectrometry (ICP).6 ------- 1.2 OBJECTIVES/PURPOSE OF THE STUDY The biases and precisions of portable XRF instruments for measurements of lead- in-paint concentrations near the HUD action level for abatement (1.0 mg/cm2)6 have not been well established. Therefore, at the request of the EPA, the Research Triangle Institute (RTI) has evaluated three of the portable XRFs currently being used for field measurements: the Scitec MAP-3, the Warrington Microlead I, and the Princeton Gamma-Tech (PGT), XK-3. The purpose of the evaluation was to compare these very different, commercially available portable XRFs, on as equal a basis as possible, for their biases and precisions when measuring lead in paint on particular substrates. Biases were calculated by comparing lead concentrations of the standards as measured by each XRF to the lead concentrations measured by digestion/ICP analysis.11 Another goal of the project was to determine a mechanism or protocol for comparison and evaluation of current and future portable XRFs on as equal a basis as possible. 1.3 STUDY DESIGN RTI's evaluation of portable XRF technology was performed using a phased approach. When the experiments were begun in 1990, the Scitec MAP-3 was the portable XRF that was most widely used for field measurements of lead in paint at "scattered-site" housing. However, at that time, no paint film reference materials were available for testing the responses of the XRF to various concentrations of lead. Therefore, RTI developed a procedure for preparing standard latex and oil-based paint films containing known amounts of lead.12 The lead concentrations of the paint films that were prepared ranged from nominally 0 to 7.16 mg/cm2 as confirmed by digestion using a modification of NIOSH Method 708211 and analysis by ICP. This range represents that typically encountered in actual field measurements.4 Different substrates of varying densities were chosen to represent the types of building materials likely to be encountered in housing units. A Scitec MAP-3 portable XRF rented from Scitec Corporation was used to make triplicate measurements for the various paint film/substrate combinations. ------- In 1992, two additional instruments, the Warrington Microlead I and the Princeton Gamma-Tech XK-3, were available to RTI for testing. Using the information gained in the Stitec evaluation, a measurement plan was developed for these two.XRFs to collect data that would allow comparison of the three instruments. Mr. William Boyce of Scitec Corporation had observed a hysteresis effect when the XRF was used to measure lead in paint on a low density substrate after it was used to measure lead in paint on a high density substrate. Hysteresis effects also may be observed when painted surfaces containing low levels of lead are measured immediately following the measurement of surfaces containing high levels of lead.13 Therefore, the measurement sequence was not ordered by lead concentration or substrate density. Instead, a randomized test matrix, presented in Table 1, was developed to determine the order of the measurements so that the hysteresis effects, if present, would be randomized. No tests were performed to determine the presence or the magnitude of the hysteresis effects. The measurement plan called for triplicate measurements of each combination of paint film and substrate for the Warrington and PGT instruments. In order to perform a measurement, the test Table 1. Randomized Test Matrix - Order (1-56) for Testing Each Combination of Sample and Substrate Paint Sample ID (cone., me/cm2) Substrate 1/2" Plasterboard 3/4" Plywood Cinderblock, Std 1/8" Steel 1/8" Aluminum 5" Concrete Brick Cinderblock, Solid 20D1 (7.16) 18 24 17 41 18 27 34 3 20D2 (6.20) 48 5 54 20 43 14 13 4 20B2 (1.27) 37 35 40 26 2 29 28 50 20A2 (0.52) 44 32 10 9 22 23 8 6 17B (0.00) 7 30 16 47 55 53 33 56 120-2-B (0.07) 12 52 39 36 45 21 42 46 120-5-B (4.%) 51 15 1 38 11 25 49 31 ------- paint film was placed against the substrate, and an attempt was made to position the XRF probe on the center of the film so that the measurement would be made on nominally the same area each time. The triplicate measurements were made consecutively with no repositioning of the instrument between readings. A "measurement" for each particular instrument was performed as defined in Table 2. Scitec recommends a 60-second reading for a "test" and a 240-second reading for "confirmation".14 For this study, an intermediate measurement time of 120 seconds was chosen because a previous study indicated that there were no significant differences in the biases and precisions for 60- and 120-second reading times.15 The 4-minute reading would only be performed in the field to confirm levels within the action range. The manufacturer's instructions for Warrington XRF defines a measurement as the average value for the lead concentration that is displayed when the trigger is depressed for three read cycles (approximately 15 to 20 seconds for a new source). The manufacturer recommends that at least three measurements per test location should be made before any conclusions are drawn concerning the presence or absence of a lead hazard.16 The PGT manufacturer recommends that a measurement be made by taking three readings (15 to 20 seconds each for a new source), recording the values, and manually averaging them.17 Table 2. Test Procedures Instrument Scitec Warrington PGT new Measurement Procedure Take one 120 sec reading Take one 3-cycle reading (15-20 sec)*, temporarily halting after each cycle to record the value displayed; manually average Take 3 readings (15-20 sec each)*, record values and manually average # of Measurements 3 excitation source ------- SECTION 2 OVERVIEW OF THE PORTABLE XRF 2.1 INSTRUMENT BASICS The portable XRF consists of an excitation source, a detector, and some means of specifically determining the lead X-rays emitting from the sample.18 Characteristics of the three portable XRFs used in the study described in this report are summarized in Table 3, Section 3. The excitation source for portable units consists of a radioactive source. Examples of source materials include 241Am, 109Cd, 244Cm, 57Co, and 55Fe. These emit X-rays or gamma rays of sufficient energy to cause excitation of the target atoms. Different isotopes are preferred for different ranges of elements because of the need to match the energy of the excitation source emission with the energy needed to cause excitation and X-ray emission (fluorescence) from the element of interest. 55Fe, for example, is well suited for the light elements Al - V (K-shell X-rays) and Nb - Sb (L-shell X-rays). 109Cd is well suited for excitation of lead L-shell X-rays, while 57Co is well suited for both lead K- and L-shell X-rays. (The K and L designations describe X-rays produced from electron transitions between the K and L shells, respectively, and the higher level shells; the L X-ray is of lower energy while the K X-ray is of higher energy and therefore can penetrate deeper.)19 Several types of detectors are available.19 The solid-state detector used in the Scitec XRF consists of the semiconductor material SiLi. When an X-ray passes through the material, the X-ray creates free electrons and "holes." When a voltage is applied across this detector, and an X-ray passes through the detector, a pulse of current flows across the detector. The number of free electrons and "holes" produced, and subsequently the magnitude of this current pulse, is a function of the energy of the X- ray, which, in turn, is specific to the emitting element. Specific detection of lead is accomplished by measurement of the magnitude of the current pulse and relating this pulse magnitude to a standard. This is done with a pulse height analyzer, that is, complex electronic circuity which differentiates current pulses by their magnitude, deals ------- with overlapping pulses, and keeps track of the number of pulses of each magnitude. The number of pulses at a particular magnitude is directly related to the concentration of the fluorescing element. The Warrington XRF uses a scintillation counter composed of a cesium thallium iodide doped crystal to convert the X-ray photon into a light photon. When radiation interacts with the crystal, the transmitted energy excites the iodine atom and raises it to a higher energy state. When the iodine atom returns to its ground electronic state, the energy is re-emitted as a light pulse in the ultraviolet region which is then absorbed by the thallium atom and re-emitted as fluorescent light. The light is detected by a 1/2" photomultiplier tube. The detector for the PGT XRF is a proportional counter. This is a xenon-filled tube across which a high voltage potential is applied. When an X-ray enters the tube, the xenon gas is ionized, producing ion pairs. The charged particles migrate toward the appropriate electrodes under the voltage gradient. Each electron acquires sufficient energy to produce another ion pair upon collision with another xenon atom. This process is repeated many times so that each original X-ray entering the counter results in a large number of electrons traveling toward the anode. An electronic pulse is produced that is proportional to the number of electrons reaching the anode and, accordingly, the energy of the X-ray. The simpler XRF units are preset to count only pulses equivalent to lead X-ray energies; these are called "direct reading" instruments. The more complex instruments, called "spectrum analyzers," take and store the entire range of X-ray emissions in a small on-board computer that performs the task of relating current pulse amplitude and numbers of pulses to information stored from standards. This computer will automatically identify the elements and determine their concentrations. A major limitation of XRF is that the efficiency of production of the fluorescing X-rays is very dependent upon the paint matrix and the substrate. The numbers of X- rays produced will depend upon (1) the ability of the exciting or source X-rays to penetrate the matrix and reach the atoms of interest; (2) secondary X-rays (created from the interaction of the primary excitation X-rays with other elements in the sample) reaching and causing fluorescence of the element of interest; (3) primary excitation X- ------- rays being scattered by matrix and substrate components so as to hit the atoms of element of interest from all angles; and (4) the ability of the X-rays emitted from the atoms of interest to escape from the matrix and reach the detector without being electromagnetically reabsorbed or scattered by the matrix. In the laboratory, great care is taken to match the matrix of the standards and unknowns. This is very difficult in the field, and impossible with instruments used to perform direct readings off walls, floors, etc. Two principal approaches are being taken to deal with this substrate problem when analyzing lead-containing paint. In the first, which provides for a partial correction, a reading of the surface is taken. Then the paint is scraped off the surface and a second reading is taken to obtain a background reading, which is subtracted from the first. This substrate correction approach was applied for the Warrington Microlead I and Princeton Gamma-Tech XK-3 instruments. In the second approach, information is taken from the total X-ray emission spectrum, and, using proprietary software, a substrate correction is calculated. The software includes mathematical variables determined from measurement of standard paint films placed on different substrates, e.g., wood, plasterboard, metal, etc. The Scitec MAP-3 instrument evaluated in this study uses this second approach. 2.2 AVAILABLE INSTRUMENTS At the time of these evaluations in 1992, three manufacturers were widely recognized as having commercially available XRFs designed for in-situ paint testing. The instruments that were tested by RTI and addresses of these manufacturers are as follows: Princeton Gamma-Tech XK-3 Princeton Gamma-Tech Inc. 1200 State Road Princeton, NJ 08540 (609) 724-7310 Warrington Microlead I Warrington Corporation 2205 West Braker Lane Austin, TX 78758 (800) 233-9491 ------- Scitec MAP-3 Scitec Corporation 1029 Kellogg Street Kennewick,WA 99336 (509) 783-9850 8 ------- SECTION 3 LABORATORY TESTS OF THE PORTABLE XRF UNITS 3.1 INTRODUCTION The Scitec, Warrington, and PGT portable X-ray fluorescence units (XRF) were employed to measure lead in paint using standard paint films and selected substrate materials. These instruments were pre-calibrated by the manufacturer and were not recalibrated by RTL The paint films, containing known concentrations of lead, were prepared by RTI from new latex and oil-based paints and pure lead salts. The concentration of lead in each paint film was confirmed by digestions using a modification of NIOSH Method 708211 and ICP spectrometry. The characteristics of the three XRFs are discussed in the following sections and summarized in Table 3. 3.2 SCITEC MAP-3 3.2.1 Description The Scitec X-ray fluorescence unit consists of the Ambient Scanner and the MAP-3 Console with battery power and cable. The MAP-3 Console is basically a portable computer that contains the calibration data and has memory for more than 1000 entries, including measurement values, identification numbers, analysis times and methods chosen. The data can be retrieved by connecting the console to an IBM PC-XT. The Ambient Scanner houses the radioactive 57Co source. The system should be returned to Scitec once a year for source replacement and recalibration. After connection of the Ambient Scanner to the console, the appropriate values for calibration program, identification number, and analysis time may be chosen. The surface to be analyzed is positioned against the rubber boot and shutter of the scanner. The shutter then is opened to expose the sample to the source, and the analysis is controlled from the console. ------- Table 3. Portable XRF Characteristics Type Source (Halflife: Days) Decay Compensation Detector Shell detected (KeV) Substrate Correction Potential Interferences Range of Calibration Substrates Used for Calibration Calibration Materials Strength of Source Used for Study (Original Strength) Recommended Source Replacement Scitec Warrington MAP-3 1 Microlead 1 Spectrum Analyzer 57Co (276) — SiLi K (74.96) L (10.55) Proprietary Algorithm K=Au, Bi, Hg, Tl L=As, Cd, Cu, Mo, Zn 0 - 6.19 mg/cm2 >30 Old HUD/NIST & RTI prepared 37.8 mCi (40 mCi) Once /year Direct Reader 57Co (276) Source date in software Scintillation Counter K (74.96) Scrape K, same 0.6 - 2.00 mg/cm2 up to 15 Old HUD/NIST 3.4 mCi (10.4 mCi) Once/ 14 months PGT XK-3 Direct Reader 57Co (276) Ta foil internal ref Xe gas, prop, counter K (74.96) Scrape K, same 0.6 - 2.99 mg/cm2 none Old HUD/NIST 9.1 mCi (10.4 mCi) Once/year ------- 3.2.2 Experimental Procedure The Scitec MAP-3 unit used by RTI in this evaluation contained a new 57Co source and had been calibrated by the manufacturer over the range of 0 to 6.19 mg/cm2 of lead using films prepared by RTI and the old HUD/NIST films. The XRF was used in the "universal" substrate correction mode to make three 120-second measurements each of selected paint films placed one at a time on selected substrates. The Scitec measurements were made prior to the development of the randomized test matrix (Table 1) and, therefore, were not made in the order specified in the matrix. In addition, several paint film/substrate combinations specified in the matrix (Table 1) had not been made when the Scitec system was being evaluated. Measurement data are reported in Appendix A. 3.3 WARRINGTON MICROLEAD I 3.3.1 Description The Warrington Microlead I is a portable, microcomputer-based XRF with a 57Co source. The instrument has built-in "Zero Standard" and "Lead Standard" functions so the user can establish a zero reference using a "standard" substrate and check that the analyzer reads the lead concentration correctly with a "standard" paint film provided by Warrington. The background density is displayed with each reading to alert the inspector to changes in substrate that might affect the accuracy of the lead reading. Up to 1000 pieces of information (lead readings, project numbers, apartment numbers, times and dates, for example) can be stored and subsequently printed on a portable printer. Warrington recommends that the source be changed every 14 months. The probe is placed against the test surface and the trigger is pulled until the analyzer beeps and displays the lead reading. Successive readings are averaged automatically. Warrington recommends that the trigger be depressed through three successive cycles to obtain a lead measurement. 3.3.2 Experimental Procedure The Warrington XRF used in this evaluation was borrowed from Dr. Mary McKnight of the National Institute of Standards and Technology. It contained a used 11 ------- source with a strength of -3.4 mCi. Some measurements were performed in the "auto" mode (by depressing the trigger through three read cycles to obtain an automatically- calculated average) prior to the development of the randomized test matrix. In a memorandum dated 3/26/92 to the RTIXRF team, the EPA Work Assignment Manager recommended that the randomized tests be performed in the "manual" mode (i.e., the trigger would be released after each read cycle so that the lead concentration read during that cycle could be recorded and an average of the three readings be calculated manually to represent one measurement). The reason for this modified procedure was that Warrington's definition of a single reading was interpreted to be three count cycles while PGT's single reading is one count cycle. This "halted" reading procedure would allow comparison of the XRFs on a more equal basis. When the analyst began to work through the randomized test matrix, he made the decision to omit film/substrate combinations that had already been measured in the automatic mode. However, since these measurements were made at a different time and in a different mode than the randomized measurements, they were not included in this evaluation and do not appear in the Warrington data presented in Appendix A. A discussion with a representative of Warrington corporation revealed that the precision obtained using the automatic mode will be better than the precision obtained using the manual mode.20 When the trigger is depressed through three read cycles, a "running average" is generated continuously. When the trigger is released after each cycle, the data from the previous cycle is not included in the following cycle. In addition to the paint film/substrate measurements, measurements were made of each substrate alone so that the randomized data could be corrected for substrate effects. The "zero standardize" procedure described in the user's manual was performed at the beginning, middle, and end of each analysis day. 3.4 PRINCETON GAMMA-TECH XK-3 3.4.1 Description The Princeton Gamma-Tech XK-3 utilizes a 57Co scintillation source with a tantalum foil internal reference device to automatically correct for source decay. The instrument's internal programming determines exposure time by referring to the count 12 ------- for the 73Ta foil. When the 73Ta counter discriminator/channel is full, the instrument triggers a shutter release. One charge of the battery pack will provide 8 hours of operation. The manufacturer recommends that the source be changed once a year at the PGT factory. The user presses the face-plate of the instrument against the surface to be analyzed, pushing down firmly on the handle. Depression of the spring-loaded handle opens the shutter. After 15 to 20 seconds, a reading appears on the digital display. PGT recommends taking three successive readings and manually averaging the results to obtain a lead measurement. 3.4.2 Experimental Procedure The PGT XRF used in this evaluation was borrowed from Mr. Jim Hayes of the North Carolina State Government Health Services Section. The 57Co source had been replaced by Warrington Corporation and was new (first half-life). Mr. Hayes reported that replacements had been made by Warrington previously and that the XRF performed properly each time such a replacement was made.21 Three readings were taken manually and averaged to obtain one lead measurement. In addition to the paint film/substrate measurements, measurements were made of each substrate alone so that the randomized data could be corrected for substrate effects. Measurement data are reported in Appendix A. 13 ------- SECTION 4 RESULTS AND DISCUSSION 4.1 EXPERIMENTAL PRECISION AND BIAS Data from the triplicate measurements of each paint film/substrate combination using each of the three XRFs are presented in Appendix A. The effects of lead concentration and substrate on the precision and bias of XRF measurements can be determined by examination of the data presented in Tables 4 through 6. The tables are ordered by decreasing lead concentration in the paint films and by increasing substrate density as determined by the Warrington XRF. The relative standard deviation (RSD) of each set of triplicate measurements, an indicator of precision, is calculated as RSD = (SD/x) x 100 where X is the average value of the three measurements and SD = N The percent difference (% Diff) of the XRF measured value from the ICP measured value, an indicator of bias, is calculated as % Diff ICP value - x ICP value / x 100 The data show that, in general, precision and bias improve as the lead concentration in the paint film increases. No trend with respect to substrate was apparent from visual inspection of the data; no statistical tests for substrate effects were performed. 4.2 LINEAR REGRESSION ANALYSES Linear regression analyses were performed on the experimental XRF data presented in Appendix A using a standard linear calibration computer program.22 14 ------- Table 4. Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: Scitec ICP Anal. mg/cm 7.16 6.20 4.96 1.27 0'.52 0.0 I/ Plaste RSD 8.7 15.1 NA 10.0 35.3 — 2" rboard % Diff -3.2 11.3 NA -21.3 -16.7 — 3/ Plyv RSD 9.2 4.9 5.0 24.1 33.3 — 4" rood % Diff -5.0 -4.3 -38.2 -50.1 -42.3 — I/ Alum RSD 8.0 9.1 NA 14.3 20.0 — 8" inum % Diff 6.1 20.4 NA -16.0 -3.8 — I/ St< RSD 3.7 11.6 NA 20.8 17.3 173 8" :el % Diff 0.6 18.3 NA 15.5 28.2 — 4 Cinde RSD 5.6 3.7 NA 142 40.0 63.0 it rblock % Diff -5.5 8.6 NA -79.0 -3.8 — Sc Cinde RSD 1.4 4.0 NA 8.5 56.8 5.1 )lid .rblock % Diff 15.0 28.5 NA 275 451 — Br RSD 5.6 12.9 NA 173 173 — ick % Diff -10.6 3.8 NA -92.1 -93.6 — e Con RSD 18.0 16.3 NA 96.6 — — >" crete % Diff -29.7 -13.4 NA -63.3 -100 — ------- Table 5. Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: Warrington ICP Anal. mg/cm 7.16 6.20 1.27 0.52 0.07 0.0 I/ Plaste RSD 0.8 2.3 1.4 18.6 NA 1530 2" rboard % Diff -9.2 -4.1 0.0 24.4 NA — 3/ Plyv RSD 0.6 1.0 8.2 16.5 NA 28.9 4" wod % Diff -10.8 -4.6 11.5 23.1 NA — I/ Alum RSD 1.7 1.9 9.3 42.4 NA 31.4 8" inum % Diff -3.6 2.6 11.0 75.0 NA — I/ St RSD 2.5 0.3 6.0 10.1 22.3 32.9 8" eel % Diff 1.7 4.0 55.1 176 1267 — 4 Cinde RSD 1.0 1.3 22.7 — 13.7 30.3 M rblock % Diff -1.7 2.7 27.8 — 424 — Sc Cinde RSD 2.5 3.8 12.1 45.5 149 13.9 did rblock % Diff -5.4 0.8 5.8 42.9 105 — Br RSD 2.0 2.6 6.4 8.2 31.5 28.2 ck % Diff -5.5 6.1 40.9 101 467 — < Con RSD 6.1 0.0 7.7 5.7 12.2 12.2 >" crete % Diff -6.1 11.9 74.8 235 1362 — ------- Table 6. Relative standard deviations and percent differences for triplicate measurements of lead in paint films using portable XRF: PGT ICP Anal. mg/cm 7.16 6.20 4.96 1.27 0.52 0.07 0.0 I/ Plaste RSD 5.8 6.3 5.5 16.5 5.9 32.9 — 2" rboard % Diff -15.1 -10.6 11.3 -3.9 -1.9 424 — 3/- Plyw RSD 1.0 4.9 3.4 25.8 54.1 24.7 68.7 4" /ood % Diff -20.8 -16.5 -39.2 -48.3 3.2 -300 — I/ Alum RSD 1.7 5.3 6.8 67.9 NA 20.1 163 8" inum % Diff -27.1 -26.2 -17.9 -65.1 NA -976 — I/ St< RSD 1.3 2.0 6.7 14.6 243 8.2 2.3 8" :el % Diff -18.8 -6.8 8.3 -68.8 -103 1110 — 4 Cinde RSD NA 2.6 3.2 37.8 71.6 82.0 86.7 it rblock % Diff NA -21.6 23.6 -23.4 10.3 224 — Sc Cinde RSD 2.4 7.7 3.9 41.4 32.8 208 456 >lid .rblock % Diff -21.3 -26.9 9.2 -38.8 -46.8 -243 — Br RSD 6.6 7.9 10.9 28.2 39.3 121 24.2 ick % Diff -26.2 -28.5 -26.9 -16.8 -42.3 -371 — < Con RSD 0.6 3.9 10.2 29.3 6.0 4015 17.2 >" crete % Diff -34.6 -28.9 -6.3 -24.7 -44.2 114 — ------- Regression parameters were calculated for each substrate for each portable XRF using the model y = a + bx, where a is the intercept and b is the slope. The data were entered as x,y pairs where x is the lead concentration of the paint film measured in the laboratory using acid digestion/ICP and y is the concentration measured by XRF. A least squares analysis assumes that the independent variable, x, is measured without error. The laboratory analysis of the paint films showed some variability. However, least squares analysis can be used if the error of x is less than one-tenth of the average scatter of the x's from their mean.23 Examination of the paint film laboratory analysis results revealed that this condition was met. Results of the linear regression analyses are presented in Tables 7 through 14. 4.2.1 Hypothesis Testing After the regression values were computed, tests were performed to determine slopes that are statistically equal to one at the 95 percent confidence level according to the following: b = 1 if b-1 (se)b < t, /i 1 - a/2, n-., where b = slope, (se)b = standard error of the slope and ti-^/2, n.2 = critical value with (n-2) degrees of freedom. Similarly, tests were performed to determine intercepts that are statistically equal to zero at the 95 percent confidence level according to the following: a = 0 if a fee). < *1 - a/2,n - 2 where a = intercept, (se)a = standard error of the intercept and *!—12, n-2 = critical value with (n-2) degrees of freedom. Slopes that are statistically equal to one and intercepts that are statistically equal to zero are indicated by asterisks in Tables 7 through 14. 18 ------- Table 7. %" Plasterboard No. of Data Pairs Slope SDslope Intercept SD •-^intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated bias* and 95% confidence interval at:" 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 1.6 mg/cm2 Scitec 15 1.043* 0.051 -0.108** 0.216 0.9851 2.96 0.539 -0.09 ± 1.35 -0.07 ± 1.34 -0.06 ± 1.34 -0.05 ± 1.33 -0.04 ± 1.33 Warrington Warrington (uncorrected) || (corrected) 15 0.915 0.014 0.413 0.060 0.9985 0.30 0.172 +0.39 ± 0.37 +0.34 ± 0.37 +0.33 ± 0.37 +0.30 ± 0.37 +0.28 ± 0.37 15 0.915 0.014 0.103** 0.060 0.9985 0.30 0.172 +0.08 ± 0.37 +0.04 ± 0.37 +0.03 ± 0.37 -0.01 ± 0.37 -0.03 ± 0.37 PGT (uncorrected) 21 0.812 0.032 0.386 0.129 0.9859 2.80 0.494 +0.33 ± 0.91 +0.24 ± 0.91 +0.20 ± 0.90 +0.14 ± 0.90 +0.09 ± 0.90 PGT (corrected) 21 0.812 0.032 0.096** 0.129 0.9859 2.80 0.494 +0.04 ± 0.91 -0.05 ± 0.91 -0.09 ± 0.90 -0.15 ± 0.90 -0.20 ± 0.90 \D * Slope (b) = 1 at the 95% level of confidence ** Intercept (a) = 0 at the 95% level of confidence a (se). t 1 - afZ, n - 2 f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentraUons chosen by reference to HUD Guidelines (see Table 15). ------- Tables. 3/4" Plywood No. of Data Pairs Slope SDslope Intercept sn JLJ intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated bias* and 95% confidence interval at:" 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 n 1.6 mg/cm Scitec 18 0.934* 0.056 -0.341** 0.248 0.9721 5.5 0.689 -0.36 + 1.53 -0.39 + 1.52 -0.41 + 1.51 -0.43 + 1.50 -0.45 + 1.50 Warrington (uncorrected) 15 0.873 0.014 0.651 0.060 0.9983 0.34 0.182 +0.61 + 0.38 +0.55 + 0.38 +0.52 + 0.37 +0.49 + 0.37 +0.45 + 0.37 Warrington (corrected) 15 0.873 0.014 0.301 0.060 0.9983 0.34 0.182 +0.26 + 0.38 +0.20 + 0.38 +0.17 + 0.37 +0.14 + 0.37 +0.10 + 0.37 PGT I PGT (uncorrected) || (corrected) 21 0.799 0.031 0.231** 0.125 0.9863 2.72 0.487 +0.17 + 0.88 +0.07 + 0.88 +0.03 + 0.87 -0.03 + 0.87 -0.09 + 0.87 21 0.799 0.031 -0.199** 0.125 0.9863 2.72 0.487 -0.26 + 0.88 -0.36 + 0.88 -0.40 + 0.87 -0.46 + 0.87 -0.52 + 0.87 N> O * Slope (b) = 1 at the 95% level of confidence b - 1 (se)b Intercept (a) = 0 at the 95% level of confidence — (se) ll - «/2. n - 2 t 1 - «/2. o -2| f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 9. 1/8" Aluminum No. of Data Pairs Slope SDslope Intercept SDintercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:ft 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 1.6 mg/cm2 Scitec 15 1.137 0.044 -0.117** 0.186 0.9906 1.87 0.428 -0.08 + 1.17 -0.01 + 1.16 +0.02 + 1.15 +0.06 + 1.15 +0.10 + 1.14 Warrington Warrington (uncorrected) | (corrected) 15 0.931 0.020 0.267 0.086 0.9970 0.60 0.243 +0.25 + 0.54 +0.21 + 0.53 +0.20 + 0.53 +0.18 + 0.53 +0.16 + 0.53 15 0.931 0.020 1.387 0.086 0.9970 0.60 0.243 +0.37 + 0.54 +0.33 + 0.53 +0.32 + 0.53 +0.30 ± 0.53 +0.28 + 0.53 PGT (uncorrected) 18 0.773 0.031 0.188** 0.136 0.9875 2.48 0.476 +0.12 + 0.86 +0.01 + 0.85 -0.04 + 0.85 -0.11 + 0.85 -0.18 + 0.84 PGT (corrected) 18 0.773 0.031 -0.492 0.136 0.9875 2.48 0.476 -0.56 + 0.86 -0.67 + 0.85 -0.72 + 0.85 -0.79 + 0.85 -0.86 + 0.84 * Slope (b) = 1 at the 95% level of confidence b - 1 (se)h ** Intercept (a) = 0 at the 95% level of confidence a (se). ll - «/2, n - 2 - a/2. n - 2 f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 10. 1/8" Steel No. of Data Pairs Slope SDslope Intercept "^intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:ft 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 1.6 mg/cm2 Scitec 15 1.064* 0.046 0.115** 0.199 0.9879 2.41 0.486 +0.13 ± 1.24 +0.17 + 1.23 +0.18 + 1.23 +0.20 + 1.22 +0.22 + 1.22 Warrington (uncorrected) 18 0.893 0.013 -0.575 0.051 0.9983 0.34 0.177 -0.61 + 0.36 -0.66 + 0.36 -0.68 + 0.36 -0.71 + 0.36 -0.75 + 0.36 Warrington (corrected) 18 0.893 0.013 0.895 0.051 0.9983 0.34 0.177 +0.86 + 0.36 +0.81 + 0.36 +0.79 + 0.36 +0.76 + 0.36 +0.72 + 0.36 PGT (uncorrected) 21 0.917* 0.043 0.484 0.175 0.9797 4.02 0.591 +0.46 + 1.24 +0.42 + 1.23 +0.40 + 1.23 +0.38 + 1.22 +0.35 + 1.22 PGT (corrected) 21 0.917* 0.043 -0.756 0.175 0.9797 4.02 0.591 -0.78 ± 1.24 -0.82 + 1.23 -0.84 + 1.23 -0.86 + 1.22 -0.89 + 1.22 N) N) * C Slope (b) = 1 at the 95% level of confidence ** Intercept (a) = 0 at the 95% level of confidence ll - a/2. n - 2 (se). t 1 - a/2, n - 2 f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 11. 4" Cinderblock No. of Data Pairs Slope SDslope Intercept "^intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:ft 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm 1.3 mg/cm2 1.6 mg/cm2 Scitec 15 1.017* 0.055 -0.155** 0.236 0.9814 3.69 0.601 -0.15 + 1.48 -0.14 + 1.47 -0.14 + 1.46 -0.13 + 1.46 -0.13 + 1.45 Warrington Warrington (uncorrected) | (corrected) 15 0.929 0.019 1.796 0.082 0.9972 0.56 0.239 +1.77 + 0.53 +1.74 + 0.52 +1.72 + 0.52 +1.70 + 0.52 +1.68 + 0.52 15 0.929 0.019 0.486 0.082 0.9972 0.56 0.239 +0.46 + 0.53 +0.43 + 0.52 +0.41 + 0.52 +0.39 + 0.52 +0.37 + 0.52 PGT (uncorrected) 18 0.717 0.028 0.405 0.091 0.9883 2.33 0.388 +0.32 + 0.64 +0.18 + 0.64 +0.12 + 0.64 +0.04 + 0.63 -0.05 + 0.63 PGT (corrected) 18 0.717 0.028 0.155** 0.091 0.9883 2.33 0.388 +0.07 + 0.64 -0.07 + 0.64 -0.13 + 0.64 -0.21 + 0.63 -0.30 + 0.63 NJ OJ * Slope (b) = 1 at the 95% level of confidence b - 1 (se)b ** Intercept (a) = 0 at the 95% level of confidence —5- y(se)B ll - «/2. n - 2 ll - a/2. n -2 f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 12. Solid Cinderblock No. of Data Pairs Slope SDslope Intercept "^intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:ft 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 *+ 1.6 mg/cm Scitec 15 0.697 0.075 3.434 0.321 0.9323 13.08 1.132 +3.34 + 2.01 +3.19 + 1.99 +3.13 + 1.99 +3.04 + 1.98 +2.95 + 1.97 Warrirtgton Warrington (uncorrected) | (corrected) 18 0.934 0.020 1.851 0.078 0.9964 0.72 0.260 +1.83 + 0.56 +1.80 + 0.55 +1.79 + 0.55 +1.77 + 0.55 +1.75 + 0.55 18 0.934 0.020 0.251 0.078 0.9964 0.72 0.260 +0.23 + 0.56 +0.20 + 0.55 +0.19 + 0.55 +0.17 + 0.55 +0.15 + 0.55 PGT (uncorrected) 21 0.749 0.032 0.401 0.130 0.9832 3.33 0.539 +0.33 + 0.92 +0.20 + 0.91 +0.15 + 0.91 +0.07 + 0.91 0.00 + 0.91 PGT (corrected) 21 0.749 0.032 -0.189 0.130 0.9832 3.33 0.539 -0.26 + 0.92 -0.39 + 0.91 -0.44 + 0.91 -0.51 + 0.91 -0.59 + 0.91 NJ Slope (b) = 1 at the 95% level of confidence b - 1 (se)b Intercept (a) = 0 at the 95% level of confidence a (se). - «/2, n - 2 ll - «/2. n - 2 f Bias is calculated as XRF predicted value minus the known concentration of the paint film. ft Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 13. Brick No. of Data Pairs Slope SDslope Intercept SDintercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:" 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 « 1.6 mg/cm Scitec 1 Warrington 1 (uncorrected) 15 1.019* 0.056 -0.495** 0.238 0.9812 3.72 0.604 -0.49 + 1.49 -0.48 + 1.48 -0.48 + 1.47 -0.47 + 1.47 -0.46 + 1.46 18 0.918 0.021 2.118 0.081 0.9959 0.82 0.276 +2.09 + 0.58 +2.05 + 0.58 +2.04 + 0.57 +2.01 + 0.57 +1.99 + 0.57 Warrington (corrected) 18 0.918 0.021 0.528 0.081 0.9959 0.82 0.276 +0.50 + 0.58 +0.46 + 0.58 +0.45 + 0.57 +0.42 + 0.57 +0.40 + 0.57 PGT (uncorrected) 21 0.721 0.036 0.373 0.146 0.9772 4.51 0.627 +0.29 + 1.03 +0.15 + 1.03 +0.09 + 1.03 +0.01 + 1.02 -0.07 + 1.02 PGT (corrected) 21 0.721 0.036 -0.227** 0.146 0.9772 4.51 0.627 -0.31 + 1.03 -0.45 + 1.03 -0.51 + 1.03 -0.59 + 1.02 -0.67 + 1.02 N) Ul * Slope (b) = 1 at the 95% level of confidence b - 1 (se)b ** Intercept (a) = 0 at the 95% level of confidence a (se). - a/2. n - 2 ll - a/2. n - 2 Bias is calculated as XRF predicted value minus the known concentration of the paint film. t Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- Table 14. 5" Concrete No. of Data Pairs Slope SDslope Intercept ^intercept Corr. Coef. Residual Variance, % Standard Error of Estimate of X from Y Estimated biast and 95% confidence interval at:ft 0.3 mg/cm2 0.8 mg/cm2 1.0 mg/cm2 1.3 mg/cm2 1.6 mg/cm Scitec 15 0.809 0.058 -0.279** 0.248 0.9682 6.26 0.783 -0.34 ± 1.55 -0.43 + 1.54 -0.47 + 1.54 -0.53 + 1.53 -0.58 + 1.53 Warrington (uncorrected) 18 0.854 0.030 2.288 0.119 0.9901 1.97 0.430 +2.24 + 0.85 +2.17 + 0.84 +2.14 + 0.84 +2.10 + 0.84 +2.05 + 0.84 Warrington (corrected) 18 0.854 0.030 1.098 0.119 0.9901 1.97 0.430 +1.05 + 0.85 +0.98 + 0.84 +0.95 + 0.84 +0.91 + 0.84 +0.86 + 0.84 PGT (uncorrected) 21 0.669 0.030 0.370 0.120 0.9820 3.57 0.557 +0.27 + 0.85 +0.11 + 0.84 +0.04 + 0.84 -0.06 + 0.84 -0.16 ± 0.84 PGT (corrected) 21 0.669 0.030 -0.150** 0.120 0.9820 3.57 0.557 -0.25 + 0.85 -0.41 + 0.84 -0.48 + 0.84 -0.58 + 0.84 -0.68 + 0.84 K) * Slope (b) = 1 at the 95% level of confidence b - 1 (se)b ll - «/2, n - 2 Intercept (a) = 0 at the 95% level of confidence a (se). t 1 - a/2, n - 2 I f Bias is calculated as XRF predicted value minus the known concentration of the paint film. tf Lead concentrations chosen by reference to HUD Guidelines (see Table 15). ------- 4.2.2 Confidence Intervals The slopes of the calibration curves with their 95 percent confidence intervals are plotted in Figure 1 for each portable XRF and each substrate. The graph shows that the slopes are not significantly different from one for the Scitec XRF on plywood, plasterboard, 4" cinderblock, brick and steel and for the PGT XRF on steel. However, the standard deviations for PGT and Scitec XRFs generally are greater than the standard deviations for the Warrington XRF. Also, the PGT and Scitec XRFs show more scatter in the slopes for the various substrates than does the Warrington. If the confidence interval of a given slope overlaps the center estimate of another slope, there is probably no difference between the two slopes. However, overlapping of intervals where neither interval overlaps the center estimate of the other is not a strong indication that the slopes are equivalent24 4.3 PREDICTED BIAS AT SELECTED LEAD CONCENTRATIONS The linear regression equations were used to calculate predicted values for y (the lead concentration measured by XRF) at various observed values of x (the "true" lead concentration of a paint film as measured by ICP). The values of x were chosen by reference to the HUD Guidelines6 as summarized in Table 15. The predicted values were used to estimate the biases, and their 95 percent confidence intervals, of the lead concentrations as measured by XRF compared to the laboratory analyzed concentrations. A negative bias indicates that the concentration measured by XRF is predicted to be lower than the "true" lead concentration in the paint film; conversely, a positive bias indicates that the concentration measured by the XRF is predicted to be higher than the "true" lead concentration. The estimated biases and their 95 percent confidence intervals for each XRF/substrate combination at 1 mg Pb/cm2 are plotted in Figure 2. This lead concentration level was chosen because it is the HUD level of hazard. The biases include zero for all combinations except the Warrington XRF on steel and concrete and the Scitec XRF on solid cinderblock. The 95 percent confidence intervals for both the Scitec and PGT XRFs are generally larger than those for the Warrington XRF. 27 ------- PGT H Plywood —i Plasterboard Aluminum 4" Cinderblock i Brick i •- -• 1 Concrete i • 1 Solid Cinderblock Steel Warrington Plywood i—•—i Plasterboard i • 1 Aluminum i • 1 4" Cinderblock i • 1 Brick i—•—i Steel -• 1 Concrete i • 1 Solid Cinderblock Scitec H Plywood 1 Plasterboard Aluminum -" 4" Cinderblock Brick 1 Steel Concrete Solid Cinderblock i i 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 Slope of Calibration Curve (with 95% Confidence Interval) 1.3 Figure 1. Comparison of experimentally generated slopes for portable XRF/substrate combinations across the lead concentration range of 0 to 7.16 mg/cm2. 28 ------- Table 15. HUD Guidelines for Pb Testing Using Portable XRF Instrumentation Type of Dwelling Multi-family Multi-family Scattered Site Scattered Site XRF Type Direct Reading* Spectrum Analyzer Direct Reading* Spectrum Analyzer XRF Measured Pb Level Considered Negative (mg/cm2) <0.5 <0.8 ** <0.3 XRF Measured Pb Level Requiring Abatement (mg/cm2) >1.6 > 1.3 > 1.6 > 1.3 * Concentrations are substrate corrected. Pb levels of 1.5 mg/cm2 or less require confirmatory testing. »* 29 ------- -2 -1 - 1/2' Plasterboard Ph Wl Sh P = PGT W = Warrington S = Scitec - 3/4' Plywood Ph W Sh 1/8" Aluminum PI •- Sh 0) O 8 0> 1/8' Steel PI— W| - 4' Cinderblock Ph -• 1 Sh Solid Cinderblock PI Wh Sh - Brick Ph W| Sh • 5* Concrete Ph Wh Sh -2 -1 Bias (with 95% Confidence Intervals), mg/cm2 Figure 2. Comparison of predicted biases at 1 mg Pb/cm2 for portable XRF/substrate combinations. 30 ------- Again, if the interval of a given bias overlaps the center estimate of another bias, there is probably no difference between the two biases. However, overlapping of intervals where neither interval overlaps the center estimate of the other is not a strong indication that the biases are equivalent. 4.4 SCITEC MAP-3 The Scitec instrument was evaluated using the "universal" calibration installed by the manufacturer to cover the range from 0 to 6.19 mg/cm2. In an earlier study, to determine whether the Scitec results could be improved by applying a substrate correction, the Scitec was used to measure the bare substrate alone for wood, plasterboard, 4" cinderblock, brick, aluminum siding and 1/8" steel.15 All substrates gave readings very dose to 0 mg/cm2 (0.00 to 0.03 mg/cm2). Therefore, no substrate correction was applied for this evaluation. The standard paint films yielded correlation coefficients ranging from 0.93 for solid cinderblock to 0.995 for 1/8" steel. The standard error of the estimate of x from y ranged from 0.30 mg/cm2 for 1/8" steel to 1.13 mg/cm2 for solid cinderblock. The bias of the predicted values ranged from -0.6 mg/cm2 at 1.6 mg/cm2 for 5" concrete to 3.3 mg/cm2 at 0.3 mg/cm2 for solid cinderblock. The bias of the predicted values for solid cinderblock is nearly an order of magnitude greater than the bias for any other of the substrates tested. With the exception of 4" cinderblock, all substrates showed a change in predicted bias with a change in concentration of lead; the predicted bias for 4" cinderblock was nearly constant at -0.15 to -0.13 mg/cm2 over the range 0.3 to 1.6 mg/cm2. The data in Table 4 show that, in general, the Scitec precision improves as the lead concentration in the paint film increases; no trend with respect to substrate was observed. 4.5 WARRINGTON MICROLEAD I The Warrington instrument was evaluated by first measuring the lead concentration in the paint films placed on the various substrates and then measuring the "apparent lead concentration" of the substrates alone so that a substrate correction could be applied. The correlation coefficients were consistent, ranging from 0.990 for concrete to 0.998 for plasterboard, plywood and steel. The standard error of the estimate of x 31 ------- from y was nearly constant, ranging from 0.2 mg/cm2 for plasterboard, plywood and steel to 0.4 mg/cm2 for concrete. The data that had not been corrected for substrate effects showed a positive predicted bias for all substrates except 1/8" steel. Application of a substrate correction decreased the magnitude of the predicted bias for all substrates except 1/8" aluminum and 1/8" steel. For 1/8" steel, the substrate correction changed the sign of the predicted bias from negative to positive (from -0.68 mg/cm to +0.79 mg/cm2 at 1.0 mg/cm2). The largest predicted biases observed were for the more dense substrates cinderblock (4" and solid), brick and concrete. The predicted biases ranged from -0.03 mg/cm2 at 1.6 mg/cm2 for plasterboard to 1.1 mg/cm2 at 0.3 mg/cm2 for concrete and were found to change with a change in concentration for all substrates. The data in Table 5 show that, in general, the Warrington precision improves as the lead concentration in the paint film increases; no trend with respect to substrate was observed. 4.6 PRINCETON GAMMA-TECH XK-3 The PGT instrument was evaluated using the same substrate correction procedure as described in Section 4.5 for the Warrington XRF. The correlation coefficients were consistent, ranging from 0.977 for brick to 0.988 for 1/8" aluminum and 4" cinderblock. The standard error of the estimate of x from y ranged from 0.39 mg/cm2 for 4" cinderblock to 0.63 mg/cm2 for brick. Application of a substrate correction generally changed the sign of the predicted bias from positive to negative and increased the magnitude of the bias. The predicted bias after correction ranged from 0.1 mg/cm2 at 0.3 mg/cm2 for 4" cinderblock to -0.9 mg/cm2 at 1.6 mg/cm2 for 1/8" aluminum; The predicted bias was found to change with a change in concentration for all substrates; the largest changes were observed for concrete and 4" cinderblock. The data in Table 6 show that, in general, the PGT precision improves as the lead concentration in the paint film increases; no trend with respect to substrate was observed. 32 ------- SECTION 5 CONCLUSIONS AND RECOMMENDATIONS 5.1 CONCLUSIONS Based upon the data collected during this limited evaluation of the Scitec, Warrington, and Princeton Gamma-Tech portable XRFs, using the experimental conditions and test design described in Section 1.3, it appears that all instruments measure lead in paint films more accurately (as determined by inspection of bias) on 1/2" plasterboard than on the other substrates tested. The Warrington XRF gave a lower standard error of the estimate of x from y and a lower residual variance than the Scitec and PGT instruments for each of the eight substrates. Also, the standard deviations of the slope and intercept for the Warrington XRF were lower than those for the Scitec and PGT XRFs for all substrates, with one exception; the Warrington and PGT XRFs had the same standard deviation of the slope for 5" concrete. The predicted biases for the three instruments on each of the eight substrates are not significantly different except for two cases: (1) the bias of the Scitec XRF is predicted to be significantly higher than the biases of the PGT and Warrington XRFs on solid cinderblock and (2) the bias of the Warrington XRF is predicted to be higher than that of the PGT XRF on 1/8" steel and 5" concrete. The data show that, in general, precision and bias improve as the lead concentration in the paint film increases. No trend with respect to substrate was apparent from visual inspection of the data. 5.2 RECOMMENDATIONS The data presented in this report should be considered preliminary. Several improvements have been made to the portable XRFs and their calibration procedures. It is recommended that the evaluation be repeated with care being taken to ensure that the instruments are compared on as equivalent a basis as possible with a minimum number of tests. This would include the following: 1) Careful selection of instruments so that source age is similar for XRFs from each of the three manufacturers. 33 ------- 2) Testing of multiple XRFs (at least 3) from each manufacturer to obtain more representative data. 3) Careful definition of "measurement" so that each XRF is being used as intended by the manufacturer to obtain data that are directly comparable. 4) Following a randomized matrix, in the order specified, for each XRF. 5) Using the new NIST-certified standard paint films (SRM-2759) and any other characterized research materials that have become available since this limited evaluation was performed. 6) Refining the test matrix to minimize the number of substrates tested. 34 ------- SECTION 6 REFERENCES 1. Agency for Toxic Substances and Disease Registry. The Nature and Extent of Lead Poisoning in Children in the United States: A Report to Congress, U.S. Department of Health and Human Services. 1988. 2. Grand, L.D. and J. M. Davis. Effects of Low Level Lead Exposure on Pediatric Neurobehavioral Development: Current Findings and Future Direction. In: Smith, M.S., L.D. Grant, and A.I. Sors, eds. Lead Exposure and Child Development: An International Assessment. London. Kluwer Academic Publishers, pp. 49-115,1989. 3. Goyer, R.A. Toxic Effects of Metals. In: Klassen, CD., M.O. Amdur, and J. Doull, eds., Casarett and Doull's Toxicology, Third Edition, Macmillan, New York. 1986. 4. U.S. Department of Housing and Urban Development. Comprehensive and Workable Plan for the Abatement of Lead-Based Paint in Privately Owned Housing, A Report to Congress. Washington, D.C., December 1990. 5. Elias, Robert. Overview of the Urban Soil Lead Abatement Demonstration Project. In Proceedings of U.S. EPA Symposium on Urban Soil Lead Abatement, Research Triangle Park, NC, August, 1992. (in preparation) 6. Lead-Based Paint: Interim Guidelines for Hazard Identification and Abatement in Public and Indian Housing, Office of Public and Indian Housing, Department of Housing and Urban Development, April 1,1990. 7. Abbritti, G., G. Muzi, C. Cicioni, M.P. Accattoli, T. Fiordi, and P. Morucci. Effects of Low Doses of Lead on Children's Health. In: Caroli, S., G.V. lyenzar, and H. Muntau, eds., Bioelements: Health Aspects, Ann. Inst. Super. Sanita. 25(3): 437- 488,1989. 8. Lead-Based Paint Poisoning Prevention Act, 42 U.S.C. 4822 (d)(2)(A), 1971. Amended 1991. 35 ------- 9. McKnight, M.E., W.E. Byrd, and W.E. Roberts, Measurine Lead Concentration in Paint Using a Portable Spectrum Analyzer X-Rav Fluorescence Device. National Institute of Standards and Technology, NISTIR W90-650, May 1990. 10. Estes, E.D., and W.F. Gutknecht. Workshop Report: Identification of Performance Parameters for Portable X-Ray Fluorescence Measurement of Lead in Paint. EPA 600/R-93/130, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina, 1993. 11. Binstock, D.A., D.L. Hardison, P.M. Grohse, and W.F. Gutknecht. Standard Operating Procedures for Lead in Paint by Hotplate- or Microwave-based Acid Digestions and Atomic Absorption or Inductively Coupled Plasma Emission Spectrometry. EPA 600/8-91/213, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina, 1991 (19 pp). Available from NTIS, Springfield, VA; NTIS PB92-114172. 12. Gutknecht, W.F., L.L. Hodson, H.H. Dutton, and D.A. Binstock. Preparation of Lead-in-Paint Standard Films. EPA Contract No. 68-D1-0009, Report in preparation. 13. International Organization of Legal Metrology. First Preliminary Draft Recommendation on Portable and Transportable X-Ray Fluorescence Spectrometers for Field Measurements of Hazardous Elemental Pollutants. OIML Reporting Secretariat PS17/RS5: U.S.A., October 1992. 14. Scitec MAP-3 Operator's Manual. Available from Scitec Corporation, 1029 Kellogg Street, Kennewick, Washington 99336. Tel. No. (509) 783-9850. 15. Hardison, D.L., Whitaker, C.O., Neefus, J.D., Estes, E.D., and Gutknecht, W.F., "Evaluation of Portable X-Ray Fluorescence Spectrometer for Measurement of Lead in Paint, Soil and Dust, EPA 600/A-92/245, U.S. Environmental Protection Agency, Research Triangle Park, NC1992. Available from NTIS, Springfield, VA; NTIS PB93-121010. 16. Warrington Microlead I Operator's Manual. Available from Warrington Corporation, 2205 West Braker Lane, Austin, Texas 78758. Tel. No. (800) 233-9491. 36 ------- 17. Princeton Gamma-Tech XK-3 Operator's Manual. Available from Princeton Gamma-Tech Inc., 1200 State Road, Princeton, New Jersey 08540. Tel. No. (609 724-7310. 18. Woldseth, R. X-Ray Energy Spectrometry. Kevex Corporation, Burlingame, CA 1973. 19. Willard, H. H., L. L. Merritt, Jr., J. A. Dean, and F. A. Settle, Jr. Instrumental Methods of Analysis, Sixth Edition. D. Van Nostrand Company, New York, NY, 10020, 1981 (pp. 249-260). 20. Personal Communication with Gary Stafford, Warrington, Corporation, Tel. No. (512) 251-7771. December 20,1993. 21. Personal Communication with Jim Hayes, State of North Carolina Environmental Health Services Section, Tel. No. (919) 733-2884. June 4, 1993. 22. Hartley, T. F. Computerized Quality Control. John Wiley & Sons, New York, NY, 1987 (pp. 16 - 28). 23. Daniel, C, F. S. Wood and J. W. Gorman. Fitting Equations to Data, Computer Analysis of Multifactor Data for Scientists and Engineers. Wiley-Interscience, New York, New York, 1971 (p. 32). 24. Nelson, Lloyd S. Evaluating Overlapping Confidence Intervals. Tournal of Quality Technology, 21(2): 140-141, 1989. 37 ------- APPENDIX A XRF DATA: Scitec Warrington Uncorrected Warrington Substrate Corrected PGT Uncorrected PGT Substrate Corrected ------- Scitec mg/cm2 Sample ID 20D1 20D2 120-5-B 20B2 20A2 120-2-B 17B ICP Analyzed Cone 7.16 6.20 4.96 1.27 052 0.07 0.00 1/2" Plasterboard 7.0 6.3 75 5.7 75 75 N/A 0.9 1.1 1.0 0.3 0.4 0.6 N/A 0 0 0 3/4" Plywood 6.1 7.0 73 6.1 6.1 5.6 2.9 3.2 3.1 05 0.6 0.8 0.2 0.4 0.3 N/A 0 0 0 1/8" Aluminum 8.3 72 7.3 6.7 8.0 7.7 N/A 1.1 1.2 0.9 05 0.4 0.6 N/A 0 0 0 1/8" Steel 7.5 7.1 7.0 6.5 73 82 N/A 1.2 1.4 1.8 0.6 0.6 0.8 N/A 0.1 0 0 4" Cinderblock 7.2 6.6 65 7.0 65 6.7 N/A 0.1 0 0.7 0.3 05 0.7 N/A 05 0.5 0.1 Solid Cinderblock 8.3 8.3 8.1 8.1 7.6 8.2 N/A 4.7 5.2 4.4 4.0 3.6 1.0 N/A 3.7 4.1 3.9 Brick 6.8 63 6.1 6.7 7.1 55 N/A 03 0 0 0 0 0.1 N/A 0 0 0 5" Concrete 4.9 42 6.0 6.1 4.4 5.6 N/A 0.9 05 0 0 0 0 N/A 0 0 0 N/A - Not Analyzed ------- Warrington Randomized Test - Unconnected Data mg/cm2 Sample ID 20D1 20D2 20B2 20A2 120-2-B 17B ICP Analyzed Cone 7.16 6.20 1.27 052 0.07 0.00 1/2" Plasterboard 6.87 6.80 6.77 6.23 6.13 6.40 1.60 157 157 0.97 0.83 1.07 N/A 0.10 0.50 0.37 3/4" Plywood 6.70 6.73 6.77 6.33 6.23 6.23 1.93 1.67 1.70 0.87 1.03 1.07 N/A 0.80 0.77 0.60 1/8" Aluminum 6.90 6.77 6.67 6.13 6.23 6.37 1.43 1.27 1.17 0.87 1.13 0.37 N/A 0.40 0.43 0.17 1 1/8" Steel 5.93 5.90 5.60 5.00 4.97 4.97 0.40 0.47 0.63 0.03 -0.20 0.07 -0.27 -0.60 -0.67 -0.30 -0.63 -0.87 4" Cinderblock 8.27 8.40 8.37 7.77 7.60 7.67 333 2.60 2.87 N/A 1.67 1.73 1.63 1.93 1.83 2.23 Solid Cinderblock 853 8.40 8.20 7.87 7.60 8.07 2.83 2.87 3.13 220 2.73 2.10 150 1.83 1.90 2.00 2.07 2.13 1 Brick 823 850 833 837 8.07 8.07 337 350 327 2.60 2.73 257 1.93 1.90 2.13 220 1.97 227 5" Concrete 837 757 7.80 8.13 8.13 8.13 323 3.43 357 2.83 3.03 2.93 227 2.07 230 220 2.00 220 A - Not Analyzed ------- Warrington Randomized Test - Substrate Corrected Data mg/cm2 Sample ID Substrate only 20D1 20D2 20B2 20A2 120-2-B 17B ICP Analyzed Cone _ 7.16 6.20 1.27 052 0.07 0.00 1/2" Plasterboard 0.31 6.56 6.49 6.46 5.92 5.82 6.09 1.29 1.26 1.26 0.66 052 0.76 N/A -0.21 0.19 0.06 3/4" Plywood 0.35 6.35 6.38 6.42 5.98 5.88 5.88 158 1.32 1.35 052 0.68 0.72 N/A 0.45 0.42 0.25 1/8" I! 1/8" Aluminum Steel | -0.12 7.02 6.89 6.79 6.25 6.35 6.49 155 139 1.29 0.99 1.25 0.49 N/A 052 055 0.29 -1.47 7.40 7.37 7.07 6.47 6.44 6.44 1.87 1.94 2.10 150 127 154 1.20 0.87 0.80 1.17 0.84 0.60 4" Cinderblock 1.31 6.96 7.09 7.06 6.46 6.29 636 2.02 1.29 156 N/A 0.36 0.42 0.32 0.62 052 0.92 Solid Cinderblock 1.60 6.93 6.80 6.60 6.27 6.00 6.47 1.23 127 153 0.60 1.13 050 -0.10 0.23 030 0.40 0.47 053 Brick 159 6.64 6.91 6.74 6.78 6.48 6.48 1.78 1.91 1.68 1.01 1.14 0.98 034 031 054 0.61 038 0.68 5" Concrete 1.19 7.18 638 6.61 6.94 6.94 6.94 2.04 124 238 1.64 1.84 174 1.08 0.88 1.11 1.01 0-81 1.01 N/A - Not Analyzed ------- PGT Randomized Test - Unconnected Data mg/cm2 Sample ID 20D1 20D2 120-5-B 20B2 20A2 120-2-B 17B ICP Analyzed Cone 7.16 6.20 4.96 1.27 052 0.07 0.00 1/2" Plasterboard 6.77 6.20 6.13 6.03 5.43 6.03 3.77 3.73 3.43 153 1.70 1.30 0.77 0.80 0.83 0.77 0.67 0.53 0.07 0.30 0.50 3/4" Plywood 6.07 6.17 6.07 5.50 5.43 5.90 3.33 3.47 353 0.90 1.13 1.23 0.83 0.77 1.30 0.33 0.27 0.27 0.10 0.23 0.37 1/8" Aluminum 5.93 5.80 5.97 5.07 553 5.17 3.47 3.60 3.23 1.47 0.93 0.97 N/A -0.07 0.17 0.10 0.43 0.40 0.80 1/8" Steel 7.10 6.97 7.10 7.03 6.90 7.13 3.73 4.07 4.00 1.67 1.67 157 1.20 1.20 1.27 050 0.60 050 050 050 0.47 4" Cinderblock N/A 5.17 5.20 4.97 353 3.73 3.70 1.40 0.80 1.47 1.17 0.93 037 0.27 053 0.63 0.40 0.70 033 Solid Cinderblock 6.37 6.10 6.20 4.77 5.13 5.47 330 3.47 327 1.03 1.67 1.40 0.80 0.97 0.83 0.73 037 037 0.10 0.63 0.80 Brick 5" Concrete 1 5.90 553 6.23 5.00 4.70 5.40 2.93 3.40 2.93 150 1.47 2.00 0.77 0.93 1.00 033 0.23 0.67 020 0.07 027 5.17 520 523 5.13 4.83 4.83 333 2.83 2.97 133 130 1.80 0.80 0.83 0.80 0.97 033 023 0.07 020 0.10 N/A - Not Analyzed ------- PGT Randomized Test - Corrected Substrate Data mg/cm2 Sample ID Substrate only 20D1 20D2 12Q-5-B 20B2 20A2 120-2-B 17B ICP Analyzed Cone 7.16 6.20 4.96 1.27 0.52 0.07 0.00 1/2" Plasterboard 0.29 6.48 5.91 5.84 5.74 5.14 5.74 3.48 3.44 3.14 1.24 1.41 1.01 0.48 0.51 0.54 0.48 0.38 0.24 -0.22 0.01 0.21 3/4" Plywood 0.43 5.64 5.74 5.64 5.07 5.00 5.47 2.90 3.04 3.10 0.47 0.70 0.80 0.40 0.34 0.87 -0.10 -0.16 -0.16 -0.33 -0.20 -0.06 1/8" Aluminum 0.68 5.25 5.12 5.29 4.39 4.85 4.49 2.79 2.92 2.55 0.79 0.25 0.29 N/A -0.75 -0.51 -0.58 -0.25 -0.28 0.12 1/8" Steel 1.24 5.86 5.73 5.86 5.79 5.66 5.89 2.49 2.83 2.76 0.43 0.43 0.33 -0.04 -0.04 0.03 -0.74 -0.64 -0.74 -0.74 -0.74 -0.77 4" Cinderblock 0.25 N/A 4.92 4.95 4.72 3.28 3.48 3.45 1.15 055 1.22 0.92 0.68 0.12 0.02 0.28 0.38 0.15 0.45 0.08 Solid Cinderblock 059 5.78 551 5.61 4.18 454 4.88 2.71 2.88 2.68 0.44 1.08 0.81 021 0.38 024 0.14 -0.22 -0.22 -0.49 0.04 021 Brick 0.60 5.30 4.93 5.63 4.40 4.10 4.80 233 2.80 233 0.90 0.87 1.40 0.17 033 0.40 -0.27 -0.37 0.07 -0.40 -053 -0.33 5" Concrete 052 4.65 4.68 4.71 4.61 431 431 2,81 231 2.45 0.81 0.78 128 028 031 028 0.45 -0.19 -029 -0.45 -032 -0.42 N/A - Not Analyzed ------- |