Date	February, 1992 (Revised December, 1992)
Chemist	Jerry C. Septon
Classification	Backup Data Report ID-125G (ARL)

Branch of Inorganic Methods Development
OSHA Salt Lake Technical Center
Salt Lake City, Utah

Commercial manufacturers and products mentioned in this report are for
descriptive use only and do not constitute advertisements by USDOL-OSHA.
Similar products from other sources can be substituted.

1. Introduction

   The purpose of this work was to validate a simultaneous Inductively Coupled Plasma-Atomic Emission Spectrometer (ICP-AES), an Applied Research Laboratories (ARL) Model 3560 (Fisons, Sunland, CA) for the analysis of industrial hygiene welding fume samples. Previously, the OSHA Salt Lake Technical Center (SLTC) has been analyzing welding fume samples using either a Jarrell-Ash Model 975 AtomComp (upgraded) ICP (Thermo Jarrell-Ash Corp., Franklin, MA) or a Jobin Yvon Model JY-32 ICP (Instruments SA, Edison, NJ).

   The general procedure for the analysis of welding fume samples by ICP has been previously described (12.1.) and evaluated (12.2.). This validation examines the analytical response of the ARL spectrometer to 13 elements commonly found in welding fume operations.

   1. Spectrometer Description

      The ARL spectrometer consists of a 1.0-m pathlength with 1,080 grooves/mm grating and a solid-state radio frequency (R.F.) generator. Standard operating conditions for this instrument were developed through a series of benchmark tests recommended for most analytical applications. Operating conditions used during this evaluation are listed in Table 1. Included in the Table is additional information regarding the computer and software used. This 36 channel instrument is configured for the simultaneous determination of 35 elements shown in the line library of Table 2. Two channels are used for Fe, one wavelength (259.94 nm) being sensitive to dilute concentrations, while the other (271.44 nm) is used for quantitating higher concentrations of Fe. The line library (Table 2) contains information for the line array wavelengths and spectral orders as set up by the instrument manufacturer.

   2. Standard and Reagent Purity
      1. Standards:

         All standard solutions used for instrument calibration or spectral interference determinations were prepared by serial dilution of Instra-Analyzed Atomic Spectral Standards manufactured by J.T. Baker (Phillipsburg, NJ). Standards used during this study were prepared in a 8% HCl/4% H₂SO₄ matrix except as noted in the text. Single-and multielement standards were used to evaluate the instrument.

         Standards used are listed in Table 3.

         ---

         NOTE: Atomic absorption (AA) single-element standards should not be used for interference determinations. These standards may not be acceptable for ICP determinations, because they may be contaminated with other species, such as the alkali and alkaline earth metals.

         ---

         Chemical compatibility of the constituents was taken into account in order to avoid precipitation when preparing the multi-element standard solutions. All solutions used for filter spiking (Section 6 and Section 7) were prepared from standards obtained from Inorganic Ventures, Inc. (Brick, NJ). Both J.T. Baker and Inorganic Venture standards were used to determine interelement corrections.

      2. Reagents:

         The sulfuric acid used for standard and sample preparation was manufactured by J.T. Baker (Instra-Analyzed grade). The hydrochloric acid and hydrogen peroxide used for sample preparation were reagent grade. Deionized water was used for sample and standard preparation.

   3. Instrument Calibration

      Calibration of the instrument was performed using a reagent blank and three sets of multielement standard solutions matched to the sample matrix. The three standard solutions were designated as HIGH1, HIGH2, and HIGH3 with the element mixtures and concentrations shown in Table 3.

      Blank (zero concentration) values were obtained from mixed-acid reagent blanks which consisted of an 8% HCl/4% H₂SO₄ mixture. During the calibration process, linear regression coefficients were calculated by the Digital Equipment Corporation (DEC) 11/53 computer from standard response readings. All calibrations were determined by first order regression only.

2. Experimental Procedure

   An ICP Standard Operating Procedure (SOP) is available for the ARL 3560 ICP, which details the instrument operating procedures used for this evaluation (12.3.).

   As previously mentioned, the analysis of 13 elements found in welding fume operations was evaluated. The 13 elements were: Be, Cd, Co, Cr, Cu, Fe, Mn, Mo, Ni, Pb, Sb, V, and Zn. A total of 21 elements were examined; however, 8 of these were not fully validated and are only determined semi-quantitatively (screened) during routine analysis (Ag, Al, As, Ca, Mg, Se, Si, and Sn).

   ---

   Note: Other elements can be added to or subtracted from this list. The capability for expanding the analysis to other elements is mainly dependent on laboratory instrumentation, and element solubility and stability in the acid matrix used for digestion.

   ---

   Applicable Permissible Exposure Limits (PELs) (12.4.) are listed in Table 4. This Table lists both the Transitional and Final Rule PELs for substances which may be present in welding fume operations and those substances which can be determined by this specific instrument after future validation. The present digestion procedure and evaluation only addresses the 13 welding fume elements listed above. The ability of the instrument to detect, correctly identify, and accurately and precisely measure near the elements' respective PELs is the main concern of this evaluation.

   A systematic set of experiments for the purpose of instrument evaluation was used. The experimental protocol included:

   • 1. investigating and correcting for spectral line interferences
   • 2. establishing analytical Background Equivalent Concentration (BEC), detection limits, and short- and long-term reproducibility for the analysis of samples
   • 3. investigating working ranges for the elements validated
   • 4. determining the overall precision and accuracy for each validated element
   • 5. comparing instrument precision and accuracy with other ICPs
   • 6. comparing instrument performance with other laboratories through the analysis of Proficiency in Analytical Testing (PAT) samples
   • 7. evaluating the acid digestion procedure for the recovery of Pb in a paint sample matrix
   • 8. evaluating the performance of different nebulizers

   These experiments are detailed in Sections 3 through 10.

3. Interferences

   Spectral interferences have been minimized by the careful selection of wavelengths for the line array of the spectrometer. Spectral interferences have also been compensated for by software which performs interelement corrections (IEC). This software assumes a linear relationship between analyte and interferent within specified working range limits. The determination of IEC factors is part of the development procedure for any multielement analytical method which uses an emission spectrometer. Interelement corrections for welding fume-type samples were experimentally determined by identifying and then evaluating the magnitude of the interferences on each analytical line. The interferences were first qualitatively identified through peak scans and then a quantitative correction factor was calculated.

   The ARL scanning mechanism, spectral order, and resolution need to be considered when addressing spectral interferences. These topics are discussed below:

   Scanning mechanism - data presentation, range, and increment

   The evaluated ARL ICP has a mechanical spectral scanning device (this device is called a "SAMI" by the manufacturer) that drives the primary slit using motor steps in order to scan peak profiles over a specific wavelength range.

   As examples, four scans are shown in Figures 1 through 4 . Entries for these Figures are detailed in the Appendix. Additional scans for this instrument can be found in the Quality Assessment (QA) Manual (12.5.). These Figures are presented graphically as peak location (vertical-axis) versus instrument response (horizontal-axis). The actual peak position is reported both in terms of wavelength (nm) and motor steps corresponding to SAMI location. The maximum range of the SAMI is ±1 nm.

   Scans were setup on the SAMI for ±80 motor steps to allow for sufficient resolution of the analyte peak regardless of spectral order. This "window" was sufficient to view any interferences close enough to the analyte peak which may be erroneously identified as the analyte. The scans were performed in 4-step increments and 2-s signal integration time was used at each step.

   Order and resolution

   Spectral resolution is defined as full width at half maximum (FWHM), i.e., at half the height, the measured width of the peak signal. The resolution is dependent upon the grating and spectral-order of the line. Using the grating specified in Table 1, the resolution/spectral order relationship is described below:

   Table of Order Vs. Spectral Resolution

   Order	Wavelength Range (nm)*	Resolution (nm)	SAMI Range (motor steps)
   1st	400-800	0.046	80
   2nd	235-400	0.023
   3rd	165-235	0.015**

   ^* wavelength range is general - slight overlap into the next order are possible

   ^** third order provides the best resolution

   As shown above, a second-order peak has a resolution of 0.023 nm. A ±80 motor step scan of a second-order peak would easily cover a range of 0.046 nm. Figure 3 shows a scan of a second-order line, Al and ±40 steps covers the wavelength from 308.192 to 308.238 nm or 0.046 nm. The ±80 motor step range is also able to cover the narrower resolution for the 3rd order line of 0.015 nm, respectively. An example of a third-order line scan can be seen in Figure 1 (±40 steps on either side of the peak covers a range of 0.03 nm).

   1. Qualitative evaluation of spectral line interferences:

      The identification of interferences was first determined by scanning peak profiles. Three types of peak scans were conducted after the instrument was profiled on the Mn line:

      • 1. the location of each analyte line peak position relative to the Mn (257.61 nm) profile line.

        ---

        Note: "Profiling" is the adjustment of the primary (entrance) slit mechanism to achieve maximum intensity for a specific element(s) in the wavelength array of the ICP. An element having a wavelength in the middle of the array is usually chosen. For profiling this particular instrument, Mn was used.

        ---

      • 2. the aspiration of high concentration (500 to 1,000 µg/mL) single element standard solutions and subsequent analysis using the spectrometer line array to determine potential interferences.
      • 3. the re-scanning of each line during aspiration of solutions containing different concentrations of the interfering element to determine the magnitude of the interference. The varied concentrations of interferent were used to evaluate the linearity of the interference. Solution concentrations up to 1,000 µg/mL of the interfering analyte were used. The 1,000 µg/mL single-element solutions were commercially prepared in various matrices. From these stock solutions, single element standards containing concentrations below 1,000 µg/mL were prepared in deionized water (DI H₂O) or 8% HCl/4% H₂SO₄ .

       

      The following steps were used to qualitatively assess any interferences:

      Step 1: Peak position relative to the Mn profile line
      The spectrometer was profiled with a solution containing 10 µg/mL Mn and then the location of each validated and screened analyte peak position relative to the Mn profile line was evaluated. This was performed in order to determine the closeness of the spectrometer line array to theoretical values.

      Steps 2 and 3: Spectral interference scans
      Single-element standards and a blank solution (DI H₂O) were scanned for each of the 21 elements to determine whether there were any significant interferences coincident or adjacent to the analyte line. The wavelengths for each analytical line included in the array with the corresponding order are listed in Table 2. Additional data from Wavelength tables such as the National Bureau of Standards (12.6.), Winge et al. (12.7.), Boumans (12.8.), Parsons (12.9.), and MIT (12.10.) were used to identify and/or confirm any interfering spectral lines. Two sets of scans were analyzed in order to determine spectral interferences:

      First spectral interference scan

      The first set of scans consisted of scanning a single element standard in the range of 500 to 1,000 µg/mL for all channels in the line array. These scans were used as a screening technique to determine which elements would pose a potential interference for any given spectral line.

      Second spectral interference scan

      After evaluating the results from the first spectral interference scans, those channels showing a potential interferent peak within the resolution of the analyte spectral line were selected for a second scan. These second scans were performed with standard concentrations in a lower range than the first scans to demonstrate the extent of each interference. All the calibration standard solutions in Table 3 were used for this purpose. The extent was determined by examining the relative position of the analyte peak to the interferent peak and the ratio of the response of the interferent to the analyte standard for the particular spectral line.

   2. Quantitative Evaluation of Spectral Line Interferences

      After an interferent was identified, an IEC factor was quantitatively calculated by determining the response on the analyte line from concentrations of 100 to 1,000 µg/mL of the interferent. The procedures used to quantitate each interference are detailed below:

      After visually identifying potential spectral interferences from the peak profiles, the quantitative effects from these interferences were determined by measuring the intensities of each interfering element. This was performed by first calibrating the instrument, aspirating blanks and single element standards, and then noting any quantitative signals produced as interferences on the analyte lines. Blanks were used to determine the baseline signal for each line. Standards and blanks were prepared in DI H₂O in order to minimize any potential chemical interferences from other reagents. The single-element standards contained higher concentrations (>100 µg/mL) than what is normally expected during routine analysis in order to amplify the interfering signal. A determination was performed for each analyte having an interfering element as shown by the scans described in Section 3.1..

   3. Instrument Calibration

      Table 3 contains the concentrations of the multielement standard solutions (HIGH1, -2, and -3) used for calibrating the ICP. The polychromator was calibrated according to the manufacturer's procedures. Multielement standards were used and included all 21 elements to be determined.

   4. Spectral Interference Calculations

      Apparent concentrations resulting from these spectral line interferences on other analyte lines were noted. The spectral line correction factors were then entered into the system software. In some cases, there were multiple corrections for an analyte. The system software limits the number of corrections; therefore, corrections were limited to those affecting the 13 validated elements and only for the most significant corrections. The criteria used to determine significance was based on the detection limit of the affected channel, and the magnitude of the interference. For instance, an interference on the Fe2 line may not appear significant due to the large detection limit for this line. The ability of an element to produce a detectable interference within it's working range was also examined. A minor interference at 100 µg/mL may become significant at an upper working range concentration of 1,000 µg/mL.

      The amount of the interferent correction that needed to be applied to correct for the apparent concentration was determined using the following steps:

      Step 1: Calculation of interelement correction (IEC) factors
      A mathematical matrix was developed using the ARL software for the 13 validated elemental spectral lines after analyzing the single-element standards and determining the apparent concentration of the interferent on each spectral line. The IEC factor was calculated for each interference using the following equation:

      K = B/C

      (1)

      where:

      K = IEC factor

      B = Apparent concentration of the affected element produced by aspiration of the single-element standard.

      C = Concentration of the interferent (single-element standard)

      Step 2: Determination of the correct order of execution of the IEC factors
      To obtain correct concentrations of the analyte when applying IECs, the equation used for correcting the interference is:

      Cc = CUNC - K CINT

      (2)

      where:

      K = IEC factor

      Cc = corrected concn

      CUNC = initial concn

      CINT = concn of the interfering element

      The following guidelines were used:

      • a. The ordered effect of the interference on other channels was evaluated. For example, an interference of Co on the Ni line would be corrected before any Ni interferences on other lines was corrected.
      • b. If such a sequence was not apparent, then each of the interfering elements were ordered by magnitude; i.e. the largest correction was applied first.

      • ---

        Note: During analysis, the ARL ICP software uses an iterative approach for applying IEC factors. According to the equation below, a final concentration for each analyte is iteratively calculated:
      • C_c = C_UNC - S (K_i x C_i)

        (3)

      • where:

        Ki = respective IEC factor

        Ci = concentration from each contributing interference for the analyte

      • The iterative process is continued until the incremental change (each contributing interference) is less than the limit of detection.

        ---

   5. Results

      Spectral scans are shown in Figures 1 through 4. Entries below each scan are further described in the Appendix.

      1. Peak profiles (motor steps "off-profile")

         The peaks for the 21 element line array(s) were first scanned in order to determine how close they were to theoretical values. Figures 1 and 2 contain information demonstrating the nearness of the analyte line to the ideal profile line. In Figure 1, the scan for the 3rd order line of Ni has a peak intensity at 231.603 nm. For Figure 2, the scan for the 2nd order line of Be displays a peak at 313.042 nm. Both of these peaks are within the resolution of the particular order. As shown in Figure 1 and2, the peaks for both Be and Ni are "off-profile" by +2 motor steps from the theoretical wavelengths given in Table 2.

         It has been suggested by a representative of the instrument manufacturer as a general rule that each spectral line should be within ±8 motor steps from the profile peak. If any of the lines are found to exceed this specification, the secondary slit would need adjustment. This guideline is very liberal. For higher order lines having better resolution, peaks exceeding ±3 motor steps could show a drastic change in response for a minimal instrumental drift after calibration. As summarized below, all 22 lines examined for this specific instrument were within ±3 motor steps.

         Motor Steps "off profile"						Element(s)
         3						Cr
         2						Be, Fe2, Ni
         1						Ca, Fe1, Si
         0						Al, Cd, Cu, Mn, Sb, Se, Sn, V
         -1						Ag, As, Co, Mg, Mo, Pb
         -2						Zn

      2. Peak profiles for spectral line interferences

         Since the "off-profile" scans were within the suggested guidelines, scans were next evaluated for potential spectral line interferences. As an example, an interference of Be on Al is discussed:

         Figure 3 shows that a set of three Be standards with concentrations of 500, 750, and 1,000 µg/mL produced peaks above the DI H₂O blank (B), on the Al channel. The Be interference is well within the Al profile; Be peaks were within +1 motor step of the theoretical wavelength for Al.

         A second scan was performed to determine the intensity and position of the Be signal in relation to Al as shown in Figure 4. A calibration standard of 10 µg/mL Al (S) was scanned on the Al channel. In the Figure, the Al peak was 0 steps from the profile line at 308.215 nm, and quantitatively gave a response of about 3 units above the blank (B). The 750 and 1,000 µg/mL Be standards also gave direct overlap peaks at the same wavelength, well within the spectral resolution of the analyte peak. In addition, the 1,000 µg/mL Be standard gave a reading of about 14 units on the Al channel. Assuming linearity, a solution containing approximately 140 µg/mL Be would give an apparent concentration of about 7 µg/mL Al on the Al channel.

      3. Spectral interference table

         A summary of all significant spectral line interferences is shown in Table 5. This Table groups the spectral line corrections in alphabetical order by interfering element. The Table is organized in the following manner: The first row lists two elements. The first element is the channel that is affected by the second element (the interfering element). The IEC factor for the affected element is followed by a qualitative detection limit which was measured by the manufacturer prior to instrument installation.

         ---

         Note: Many of the qualitative detection limits in Table 5 are smaller in magnitude than those shown in Table 6. The two Tables illustrate some degradation in instrument performance over time. The qualitative limits presented in Table 6 were produced about 3 years after the instrument was installed.

         ---

         The manufacturer detection limits are included to illustrate most interference corrections are less than the detection limit when the interferent exists at concentrations near 1 µg/mL.

         As can be seen from Table 5, a correction of 0.056 µg/mL is necessary to compensate for the effect of Be on the Al channel. Elements Mo and V are also interferences on the Al channel.

4. Qualitative Detection Limits and Precision
   1. Qualitative Detection Limit (DL), Short-term Precision, and BEC. The following procedure was used to determine detection limits, short-term reproducibility, and BEC for each element:
      • 1. Aspirate a reagent blank, and perform 11 exposures for each element.
      • 2. Aspirate the multielement standard (c) and perform 3 exposures. (Note: The detection limits were determined after aspirating with a 5-s integration time per exposure, the multielement calibration standards (HIGH 1-3) listed in Table 3.)
      • 3.Use the intensities obtained from aspirating the multielement solution to calculate the average gross intensity and the standard deviation for each element. Perform the same functions for the blank.
      • 4. Calculate short-term precision (%CV). (i.e. for each element in the multielement standard solution, calculate the %CV for the 3 exposures.)
      • 5. Calculate the average net intensity for each element. (i.e. subtract the average blank intensity (Bl_AV) from the average gross intensity.)
      • 6. Calculate the slope (via linear regression) of the calibration curve.
      • 7. Calculate the detection limit. The qualitative DL for the ARL instrument is calculated as:
        Qualitative DL = 2(SD_BL)(c)/(I_S - Bl_AV)

        Where:
        SDBL	=	standard deviation of the blank
        c	=	concentration of the aspirated standard
        IS	=	average intensity of the aspirated standard
        BlAV	=	average intensity of the aspirated blank

      • 8. Calculate the BEC for each element.

        [Note: The BEC is defined as the concentration of an analyte that is equal to the net intensity of the background signal for that analyte (BEC = (m)Bl_AV)].

   2. Results - Short-term Precision, Qualitative DL, and BEC

      The results for short-term instrument precision as the coefficient of variation (%CV), and BEC are reported within Table 6. The same data was used to calculate the qualitative DLs.

      At concentration levels equal to or greater than the BEC, short-term precision should normally be approximately 0.5% to 1.5% CV, provided linearity of the spectral response function is maintained. Standard concentrations used for conducting the precision tests were at or above the BEC value for each element to ensure reproducibility.

      Short-term precision should be <1% for simultaneous ICP instruments. It is not unusual to routinely achieve 0.5% CV for each element routinely analyzed by the ARL model 3560 ICP (12.11..).

      As can be seen in the last column of Table 6, the %CV is <2% for all lines. The excellent short-term precision demonstrated by this instrument is indicative of a stable sample introduction system, specifically the stability of the gas flows and the uniformity of aerosol generation. In addition, a stable RF power supply to the induction coil of the torch is required for good short- term precision.

      The qualitative detection limits ranged from 0.0001 to 0.1 µg/mL.

      ---

      Note: The manufacturer's software algorithms automatically calculate a qualitative detection limit using two times the standard deviation (2SD) among other considerations. See Section 4.1., step 7 for the detection limit calculation. Although the OSHA Inorganic Method Protocol (12.12.) stipulates a qualitative detection limit shall be determined using 3SD in the calculations, 2SD is accepted here to allow for future performance comparisons using the same instrument and software.

      ---

   3. Long-term Reproducibility

      The ability of the instrument to perform in a stable manner over a 4.8-h period was examined. The following procedure was used:

      • 1. Calibrate the instrument with a reagent blank and the multielement standard, HIGH1, as listed in Table 3.
      • 2. Determine concentrations of the elements in the standard approximately every 15 min over a period of 4.8 h without recalibration.
      • 3. Determine the %CVL from each element's concentration readings (%CVL is the precision for each element from 2 exposures at each 15-min interval). Determine the average of the CVLs for all 12 elements for each 15-min period (%CV_MEAN).
      • 4. Calculate the average of all %CV_MEAN values (all 15-min intervals) to yield one final number to characterize the long-term reproducibility of measurement:
        • Long-term reproducibility = (%CV_OVERALL) = S(%CV_MEAN)_i/i
        • Where: i = the respective interval being examined

      A peristaltic pump connected to an automatic sampler was used to assist in introducing each sample into the nebulizer during this test. Routine analysis of industrial hygiene samples and standard calibration are performed using a peristaltic pump. The instrument evaluated in this report has internal temperature compensation and a vacuum system which reportedly increases long-term reproducibility. A constant temperature is maintained within the internal spectral components of the ICP to negate any small changes in ambient temperature.

   4. Results - Long-term Reproducibility

      Exposures were taken for 18 consecutive 15-min intervals. The standard solution selected, HIGH1, contained 12 elements: Ag, Be, Ca, Cd, Cr, Cu, Fe, Mg, Mn, Ni, Pb, and V. The %CV_L for each element was calculated from the two integrations and then the %CV for all 12 elements were averaged together (%CV_MEAN) for each 15-min period. The standard deviation of each average (SD) was also calculated. The tabulated values over the 4-h period, as shown in Table 7, gave a %CV_OVERALL of 0.503. This is less than quoted long-term reproducibility values of 1.1% (for an ARL 3510 ICP) and 0.7% (for an ARL 3580 ICP) as determined by manufacturer's representatives (12.11.).

      Because long-term reproducibility is largely influenced by the ability of the spectrometer to remain stable despite fluctuations in ambient temperature and humidity, this measurement quantifies the degree of drift of the spectrometer. The reproducibility measurement may be used as an indication of the validity of analytical measurement over long periods of time between calibrations. The excellent results obtained from this test demonstrate spectrometer stability and precision. This long-term reproducibility provides freedom from frequent calibration, giving the benefit of improved sample throughput and efficiency.

5. Working Range

   An evaluation to determine the appropriate concentration range for analyzing each element was performed. The linearity of each calibration curve was checked by analyzing several single-element and multielement standard solutions within the range from 5 to 10 times the calculated detection limit up to 1,000 µg/mL. The working range was determined by:

   1. Each spectral line was calibrated with a reagent blank and calibration standard. The calibration standards used are shown within Table 3 as HIGH1, -2, and -3.
   2. Standards above and below the concentration of the calibration standard were analyzed and the corresponding concentrations were calculated from the two-point calibration discussed in Section 5.1..
   3. The working range was determined as:
      • 1. The lower quantitative detection limit (LQD) for this specific ARL instrument is calculated as five times the qualitative detection limit. Each LQD limit value is used to accept or reject the concentration of each element found during a routine analysis of industrial hygiene samples.
      • 2. The upper range (upper quantitative detection limit) is used to determine if samples need dilution to allow quantitation of an element(s) within the linear calibration range. Many of the lines evaluated are linear above 1,000 µg/mL; however, due to the nature of industrial hygiene samples, 1,000 µg/mL was chosen as a tentative cut-off point for the upper working range unless non-linearity was demonstrated. Most air samples taken in industrial hygiene environments do not exceed the upper concentration range of the calibration standards specified (HIGH1-3).
   4. If a standard produced a response greater than 2,000 K-pulses (the intensity unit designated by ARL), a standard giving counts less than this limit was chosen as the upper working standard provided linearity was demonstrated. Above 2,000 K-pulses the detector becomes saturated.
   5. Results

      Table 6 summarizes the working ranges (µg/mL) for lines designated in the array. The "RANGE" column in the Table contains the lower quantitative detection (LQD) limit and the upper range. The quantitative detection limits ranged from 0.001 µg/mL for Be to 1 µg/mL for the less sensitive Fe line.

      The linear range was evaluated using a 5-s integration time. The upper concentration limit ranged from 10 µg/mL for Ca to 1,000 µg/mL for Al, Mg, Pb, and for the less sensitive Fe line. The optimum working range for most elements validated exceeded 100 µg/mL.

6. Precision and Accuracy

   The precision and accuracy of the method were evaluated by analyzing spiked samples. These spiked, multielement samples were prepared by an independent group within the laboratory. Ten samples and a blank were analyzed. These Quality Control (Q.C.) samples were determined using a two-point calibration consisting of standards listed in Table 3 (HIGH1, -2, and -3) and a reagent blank. This experiment was designed to evaluate instrument performance for determining analytes normally found in welding fume operations.

   1. Preparation of Q.C. Samples

      The spiking scheme for the multielement samples was determined in the following manner:

      Because the total amount of each multielement spike would be very large if delivered to a single mixed-cellulose ester (MCE) filter, two filters were used. The elements spiked on the first filter were: Be, Cd, Co, Cu, Pb, and Zn. The remaining elements spiked on the second filter were: Cr, Fe, Mn, Mo, Ni, Sb, and V. The spikes were delivered from stock solutions using calibrated micropipettes. One stock solution was used for the first filter element series; two stocks were used for the elements on the second filter.

      The samples were digested in accordance with the procedures described in reference 12.1.. During the digestion step of the sample preparation, both sets of filters were combined so that they would be analyzed together. All samples were diluted to a 50-mL solution volume.

   2. Analysis

      Using the spectral corrections and analytical parameters previously determined, spiked samples were analyzed using two 5-s integrations. These samples had been previously analyzed using a different type of ICP.

   3. Results

      The precision and accuracy were evaluated in terms of the overall analytical error (AE). Table 8a contains the results and lists the information for each element. The corresponding results in this Table are reported as:

      precision is reported as the CV;
      the average of all found/theoretical results for each analyte reported as Mean Recovery;
      the Range is reported in (µg);
      the overall AE is calculated as:

      ±AE % = 100 [|Mean Bias| + 2(CV)]     95% confidence

      As summarized in Table 8a, the results for the found/theoretical ratios were well within ±0.05. The precision was better than ±0.03 for all elements. The AE was better than ±7% for all elements in the microgram ranges tested. (Note: The Fe2 line results are not included as this is a channel only used for quantitating large concentrations of Fe. The Fe1 channel is normally used for routine determinations).

   4. Results from Routine Analyses

      After the instrument performance was evaluated, a series of 50 Q.C. samples were prepared and analyzed by 11 chemists. The Q.C. samples were analyzed along with routine industrial hygiene samples and were diluted to either 25- or 50-mL solution volumes. The analytical results are presented in Table 8b. As expected during routine day-to-day analysis, results show a decrease in performance when compared to Table 8a results. The element Sb is of particular concern because of the high AE displayed. The signal-to-noise ratio of this element is very low, indicating a lack of sensitivity and is further reflected by a large detection limit. This may account for some of the imprecision noted in the Table 8b results for Sb. Another factor being considered that may account for the lower than expected recovery is the stock solution from which the Sb spikes are prepared. This commercially prepared solution contains Sb and Mo in DI H₂O and may not retain the Sb as well as expected. A different stock solution (acidic) was used to prepare the Sb spikes for samples listed in Table 8a and Table 10.

7. Comparison of Precision and Accuracy on Different ICPs
   1. Procedure

      Quality control sample results were used to compare the performance of other ICPs with the ARL 3560 ICP. The same Q.C. samples shown in Table 8a were analyzed on a Jobin Yvon (JY-32) ICP for comparison with the ARL ICP. A different set of samples was used for instrument comparison with a Jarrell-Ash Model 975 ICP system. Three 15-s integration times were used for the Jarrell-Ash ICP determination of samples, and ten 10-s integrations were used on the JY-32.

   2. Results

      The means, CVs, and overall errors of the 13 evaluated elements are shown in Tables 9 and 10 and summarized below:

      JY-32 ICP

      The Q.C. sample results used to compare the performance of the JY-32 ICP with the ARL 3560 ICP are shown in Table 9. This Table compares the same sample results reported in Table 8a the ARL 3560. The microgram range of spikes on the filters are the same as reported in Table 8a. As can been seen from Table 9, the results for all elements analyzed are:

      ICP	Overall Mean Recovery
      ARL	1.000±0.013
      JY-32	0.967±0.027

      The average AE for all 13 elements analyzed by the ARL was ±4.2% compared with the JY of ±12.9%.

      The Jarrell-Ash (JA) ICP

      Quality control sample results comparing the performance of the JA Model 975 AtomComp (upgraded) with the ARL 3560 ICP are shown in Table 10. As shown, the average recovery for 13 elements was 0.95 for the ARL versus 0.91 for the JA. The average precision (CV) was 0.05 for the ARL compared with 0.06 for the JA. The overall error was ±16% versus ±22% for the ARL and JA, respectively. The overall error for the ARL was less than the JA ICP for all elements except Pb. The most significant improvement in error was for Co; the recovery for this element was 0.76 versus 0.95, giving an overall error of ±35% versus ±7% for the JA and ARL, respectively. (Note: Prior to the evaluation a malfunction of the Co channel on the JA ICP was noted).

      Many of the mass ranges spiked on each filter are lower for samples listed in Table 10 than those reported within Table 8a. The lower amounts for Pb may account for the larger AE value for this element when using the ARL ICP. The quantitative detection limit for Pb on the JA is reported as 0.1 µg/mL compared with 0.15 µg/mL on the ARL ICP. Results from both instruments indicated low recoveries for Pb in these specific Q.C. samples. Because the amount of Pb spiked on the filters was only slightly above the detection limit for Pb on either instrument, this lower range might account for some of the decrease in performance noted when the Pb precision and accuracy results of Tables 9 and 10 are compared for the ARL ICP. Results for the other elements listed in Tables 9 and 10 display good agreement between ICPs and indicate good performance from the ARL.

8. PAT Sample Determinations
   1. Procedure

      Rounds 97 and 100 of Proficiency Analytical Testing (PAT) samples were analyzed using the ARL ICP. The elements contained in these samples were Cd, Pb, and Zn. Samples from these rounds were digested and then analyzed using the ARL ICP. The PAT samples were determined using 25-mL sample volumes with two 5-s integrations.

   2. Results

      The results obtained were compared to the data reported from other participating laboratories. Table 11a gives the upper and lower control limits and the amount found. As can be seen from the data, all found values were within the control limits. Table 11b also gives the PAT reference values of analyte (Cd, Pb, Zn) that were spiked on the filters.

9. Digestion Procedure Evaluation - Pb Recovery from Paint Samples

   The sample digestion procedure was previously evaluated for welding fume-type elements (see reference 12.2. for further details). To examine the ability of the analytical procedure for another potential sample matrix, an experiment was conducted to determine any effect on the recovery of Pb contained in specific paint samples. Occasionally, welding fume samples will be taken during welds of painted surfaces, and a grab sample of the paint used is submitted for ICP analysis along with the air samples taken from the welding fume operation(s). An automotive paint bulk with the specifications described below was selected:

   Chemical Composition of Paint Bulk

   Ingredient	Percent
   Inorganic Ingredients:
   Lead Chromate(*)	15
   Molybdate Orange	10
   Lead (as Pb)	14.71
   Chromium VI (as Cr)	3.00
   Organic Ingredients:
   Light Aromatic Naphtha	<5
   Methyl Isobutyl Ketone	5
   Methyl n-Amyl Ketone	5
   n-Butyl Acetate	15
   1-Methoxy-2-Propanol Acetate	<5
   * Molecular formula not specified by the manufacturer.
   The paint consisted of these components: Catalyst, reducer, and paint base.

   1. Procedure

      Three MCE filters were spiked with the paint bulk as follows:
      10 mL of catalyst and 20 mL of reducer were added to 40 mL of paint, and then mixed. An MCE filter was placed on a micro-balance to determine the tare weight.

      ---

      Note: Welding fume samples are normally collected on MCE or PVC filters. Filters composed of MCE were used for spikes and weights to provide a low amount of residue after digestion. Normally, MCE filters are not recommended for accurate weight determinations because of humidity dependence. Water retention occurs due to the hydroscopic characteristics of the MCE filter and a constant humidity environment is required during weighings. When weighing spikes in a varied humidity environment, more hydrophobic materials such as PVC filters should be used.

      ---

      Each filter was placed in a dessicator for a few minutes, removed, and weighed until a constant value was obtained. A drop of the mixed paint was then added to the filter (mass reported as "wet" in Table 11a). The filter was allowed to air-dry over-night and then re-weighed (weight recorded as "dry"). The difference between the wet and dry weight was calculated and reported as volatile solvents (DIFF³). The volatile solvents were mainly from the catalyst and reducer.

      Samples were digested using the procedure as described in reference 12.1. All samples were diluted so a final matrix of 8% HCl/4% H₂SO₄ was achieved. The samples were analyzed shortly after dilution and also 6 days after preparation to determine if any decrease in Pb recoveries occurred over time.

      In addition, six samples were spiked on polyvinyl chloride (PVC) filters to determine the hexavalent chromium content in the paint. These samples were analyzed by polarography using OSHA Method no. ID-103 (12.13.).

   2. Results

      The amount of Pb or Cr spiked on each filter was calculated using the percent reported in the material safety data sheet (MSDS) and the amount of dry mass on the filter. For example:

      14.7% total Pb was reported to be in the paint, so for filter no. 3 listed in Table 11a, the resultant theoretical amount of Pb would be 0.44 x 0.147 = 0.065 mg. As shown in Table 12b, during the ICP analysis, 0.058 mg of Pb was found and 0.058/0.065 x 100% = 89% was recovered.

      The results can be subject to two sources of error:

      • a. Error from manufacturer's assay:

        Information obtained from the manufacturer revealed that the percentages reported on the MSDS sheet were not assay or analyzed values, but simply the percentage amounts of elements or compounds added during the formulation of the paint. In addition, the value reported for hexavalent chromium was measured as total Cr by the manufacturer. The manufacturer did not determine the valance state of the chromium, but assumes all Cr to be hexavalent.

      • b. Weighing errors:

        The wet weight was not considered to be an accurate weight as the weight decreases very rapidly due to solvent evaporation. As expected, the dry weight was more accurate. The dry weight should be used when considering bulk composition because wet weight will tend to change in relation to the volatility of the substance.

      The results for Pb varied from 80 to 100% for filter sample loadings of 0.065 to 0.463 mg (as Pb). The Cr recoveries ranged from 80 to 95% recovery for filter loadings of 0.01 to 0.09 mg of Cr, with the exception of sample 4 with a recovery of 61%. It is unknown as to why the Cr value was low for this sample.

      Six spiked samples prepared on PVC filters were analyzed by polarography to determine the amount of hexavalent Cr in the paint. The paint was allowed to dry and hardened appreciably before buffer extraction. The sample preparation procedure for the polarographic method (OSHA ID-103) was not able to extract any Cr from the dried paint; therefore, no hexavalent Cr was found by this procedure. According to the polarographic method for hexavalent Cr, after extraction the residual paint sample is digested using mineral acids and analyzed for elemental Cr by atomic absorption spectroscopy. The results by atomic absorption are then reported as "residual total chromium ". The results of this experiment indicate the need to digest the paint sample after extraction and analyze the sample by atomic absorption, or to develop a new extraction procedure to capture all of the hexavalent Cr.

10. Nebulizer Performance Evaluation

    A comparison between two different concentric Meinhard-type nebulizers was conducted in order to evaluate if any difference in analytical sensitivity could occur from nebulizer design differences. Both nebulizers (Type A and C) are composed of borosilicate glass; the difference between the two is that the Type C has a recessed capillary tip and the Type A has a flush tip. The Type C nebulizer has been advertised as capable of aspirating high-solid (dissolved salts) solutions, having lower detection limits, and greater aerosol delivery stability. In addition, the Type C is considered more efficient in sample delivery.

    1. Procedure

       Sensitivity was determined using the ARL software to calculate BEC and DL values for various multielement standard solutions in a 5% HCl or 8% HCl/4% H₂SO₄ acid matrix. The 5% HCl solution was used to determine if nebulizer performance was matrix dependent. The standards used, as listed in Table 3, were (ARL1, OSHA1, -2, and -3). The BEC and DL values were determined using 11 exposures for the blank and 3 exposures for the standards and a 5-s integration time per exposure. A peristaltic pump was used with a flow rate set at 1.54 mL/min, with argon gas pressures for the plasma and carrier set at 21.5 and 30 psi, respectively.

    2. Results

       The results for the Meinhard-type nebulizer comparison ( Table 13) are summarized as follows:

       5% HCl matrix

       The results show that the Type C nebulizer gave improved detection limits over the Type A for this matrix.

       4% H₂SO₄ /8% HCl matrix

       A paired t-test was performed on the BEC values and the results indicate that the nebulizers gave significantly different results. When the elements Cd, Pb, Ag, Be, Ca, Cu, Fe, Mg, Mn, V, Zn, Al, and Se were determined in this acid matrix, the Type A nebulizer gave better detection limits. The elements Co, Mo, Sn, and Si gave about the same results for both the Type A and C nebulizers. Detection limits for As, Cr, Ni, and Sb increased when using the Type A nebulizer. However, when considering the BEC, the Type-C nebulizer values were consistently better than the Type A. Only two exceptions to this were found (Ca in OSHA1 and Si in OSHA2) and appear to be due to random error rather than a significant difference. Both the Ca and Si lines for this instrument exhibit a significant amount of noise during analysis. The Ca line is extremely sensitive while the Si line has very low sensitivity. In addition, the sample solution matrix is not optimized for Si.

       The detection limit results obtained appear to be matrix dependent. The 5% HCl mixture showed an improvement for the C-type nebulizer. For standards contained in the mixed-acid matrix, the A-type nebulizer showed a slight improvement in detection limit over the C-type nebulizer.

11. Discussion

    In conclusion, these experiments demonstrated that the ARL ICP instrument can adequately determine 13 elements commonly found in welding fumes. In addition, the precision and accuracy of the ARL 3560 ICP compared favorably with other in-house ICP spectrometers. Addendum

    ---

    During the evaluation, the existence of a few "bugs" were noted in the software provided by ARL. Although this software (and computer hardware) is no longer available, current users should be made aware of the problems. One problem occurs during repeated calibration of the instrument. During each calibration, an interelement correction is applied if two interfering elements are present in the same calibration standard solution. For a few of the elements having interferences, the respective interelement correction being applied became additive with the next calibration (i.e. if 10 µg/mL Sb was present in a calibration solution containing 10 µg/mL Zn, an interelement correction of 0.03 was applied such that the Sb concentration became 10.03 µg/mL to compensate for the additive effect of Zn on the Sb line. The software would not erase this 10.03 µg/mL calibration concentration and the next calibration would result in Sb being recognized as 10.06 µg/mL, even though there was only 10 µg/mL Sb in the solution). The correction would continue to be additive with each calibration unless the user applied a software program called CSET to "reset" the concentrations to their original values after each calibration. This was somewhat time-consuming to perform and was not widely known by the manufacturer's personnel.

    For other elements, the concentration would be "reset" to it's original concentration for the next calibration. This appeared as a somewhat random occurrence; most of the elements which had an interference present in the same calibration solution were not affected.

    ---

12. References
    • 12.1. Occupational Safety and Health Administration Salt Lake Technical Center: Metal and Metalloid Particulates in Workplace Atmospheres (ICP Analysis) (USDOL/OSHA Method No. ID-125G). In OSHA Analytical Methods Manual 2nd ed. Cincinnati, OH: American Conference of Governmental Industrial Hygienists, 1991.
    • 12.2. Occupational Safety and Health Administration Salt Lake Technical Center: Metal and Metalloid Particulates in Workplace Atmospheres (ICP Analysis) (Backup Data Report) (USDOL/OSHA Method No. ID-125G). In OSHA Analytical Methods Manual 2nd ed. Cincinnati, OH: American Conference of Governmental Industrial Hygienists, 1991.
    • 12.3. Occupational Safety and Health Administration Salt Lake Technical Center: ARL ICP Analysis Reference Guide - OSHA LAB (Comprehensive Version), by J. Septon. Salt Lake City, UT, 1991 (unpublished).
    • 12.4. "Air Contaminants; Final Rule": Federal Register 54:12 29CFR (19 Jan. 1989). pp. 2923-2960.
    • 12.5. Occupational Safety and Health Administration Salt Lake Technical Center: ICP Quality Assessment Manual. Salt Lake City, UT. In progress (unpublished).
    • 12.6. National Bureau of Standards: Tables of Spectral Intensities, Part I-Arranged by Elements, Part II-Arranged by Wavelengths, by W.F. Meggers, C.H. Corliss, and B.F. Scribner (NBS Monograph 145). Washington, DC: Government Printing Office, 1975.
    • 12.7. Winge, R.K., V.A. Fassel, V.J. Peterson, and W.A. Floyd: Inductively Coupled Plasma Atomic Emission Spectroscopy: An Atlas of Spectral Information. 1st ed. New York: Elsevier, 1985.
    • 12.8. Boumans, P.W.J.M.: Line Coincidence Tables for Inductively Coupled Plasma Atomic Emission Spectroscopy. Vol. 1 and 2, 1st and 2nd ed. Elmsford, NY: Pergamon Press, 1980 and 1984.
    • 12.9. Parsons, M.L., A.R. Forater, and D. Anderson: An Atlas of Spectral Interferences in ICP Spectroscopy. New York: Plenum Press, 1980.
    • 12.10. Institute of Technology: MIT Wavelength Tables, Part II-Arranged by Elements. Cambridge, MA: MIT Press, 1982.
    • 12.11. Arellano, S.D., M.W. Routh and P.D. Dalager: Criteria for evaluation of ICP-AES performance. Amer. Lab.: 20-32 (August, 1985.)
    • 12.12. Occupational Safety and Health Administration Salt Lake Technical Center: Evaluation Guidelines of the Inorganic Methods Branch. In OSHA Analytical Methods Manual 2nd ed. Cincinnati, OH: American Conference of Governmental Industrial Hygienists, 1991.
    • 12.13. Occupational Safety and Health Administration Salt Lake Technical Center: Hexavalent Chromium (USDOL/OSHA Method No. ID-103). In OSHA Analytical Methods Manual 2nd ed. Cincinnati, OH: American Conference of Governmental Industrial Hygienists, 1991.

Table 1
Specifications for ARL 3560 Simultaneous AES-ICP

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Appendix
Spectral Scans

Figures 1 through 4 show examples of line profile scans. There can be up to five line entries below each of the figures. The line entries are discussed below:

Peak location standard
The first entry (designated by the letter "S") was the standard used to determine the peak or interference location. This location is reported in terms of wavelength (nm) and is the determinant for calculating the number of motor steps that the peak line (or interference) was found to be "off- profile" from the profiling element.

Second and third standards
The 2nd and/or 3rd entries (indicated as "^*" or "." respectively) in the sample list were standard solutions analyzed for comparison purposes.

Blank standard
The blank solution (indicated by the letter, B) was used to give a baseline value to the scans.

Peak location in relation to profile position
The last entry in the peak scan displays the position found relative to the position profiled with a Mn standard solution. For example, in Figure 1 a 10 µg/mL concentration of Ni contained in the multielement standard solution ARL1 was scanned and it was noted that the Ni line was 2 motor steps away from the profile line.