Finer Points of ICP-OES Setup and Operation (Part 1) James Bartos Office of Indiana State Chemist
Challenges with Fertilizers Broad conc ranges from low ppm to upper % level Want to test all nutrient elements at the same time in a single extract Avoid multiple methods Avoid multiple dilutions Easily ionized and difficultly ionized elements in the same digest/extraction solutions Range of matrices from simple water or dilute acids to difficult organic solvents and high salts Investigational allowances (IA s) were created with method(s) optimized for a single analyte ICP is simultaneous, so conditions aren t necessarily optimized for each element
Challenges with Instruments Operation is now completely computer controlled Good: almost anyone can operate Bad: mostly just pointing and clicking; what's going on in background may not be fully understood or watched? e.g. are you looking at intensities or just concentrations? Often configured for robust middle-of-the-road analysis, but one-size-fits-all results in compromises MANY different instrument options, but most have some limitations
Concentration range Wavelength selection Calibration Select Topics Multiple wavelengths per range Line Splitting Rationale General principles Your ICP may have slightly different properties
Calibration Options 1. Single Wavelength per element Easier and less confusing Data/LIMS processing advantages (i.e. one result/element) Can still get good data May not optimize MDL s Can require more secondary dilutions 2. Multiple Wavelengths per element More targeted approach Strong lines for low conc and weaker lines for expanded range Potential for better data (not compromising) Offers confirmation of result More comprehensive use of advanced ICP features Requires more standards as multiple points per curve Potentially large/confusing data output file
Concentration Range Selection Example Fe Similar approach for B, Ca, Cu, Mn, P, K, S, Zn, etc. Limitations for some elements (e.g. As, Pb, Se) Goal: measure from low ppm up to % levels 0.02% = 1 ppm in solution (@ 0.5 g to 100 ml) 20% = 1,000 ppm in solution (@ 0.5 g to 100 ml) Ask vendor 1 to 1,000 ppm? no problem May not mean one wavelength or single calibration Are you willing to do dilutions? How many? How often? Do the samples need to be more concentrated for SUIP or heavy metals analyses?
Wavelength Selection Note: may have fewer wavelength selections with some ICP s
Strategically Select Wavelengths Most intense/sensitive wavelength(s) (e.g. 238.204), may have best MDL, but also narrowest working range Weaker lines will allow for expanded working range, but at the expense of MDL Watch for spectral interference
Basic Considerations Wavelength selection table values are a relative intensity estimate (e.g.1 ppm solution), but that is dependent upon system configuration Rule-of-thumb decent linearity generally up to about 1 million counts Fe wavelength 238.204 = 41,000 counts, so may be linear to about 25 ppm (i.e. 41,000 * 25 ~ 1M) Instrument and detector dependent Data from Agilent 5100 and 720 ICP-OES Suggest making a standard (e.g. 10 ppm), select wavelengths of interest, then start screening
Fe 10 ppm scans 238 Axial is strongest (500K); lowest detection likely; minimal spectral interference; ~ linear to 20 ppm 259 Axial is next strongest line (125K), use to confirm 238?, ~ linear to 80 ppm 238 Radial is ~ 1/10 th the signal of axial view; consider starting calibration at 20 or 80 ppm; ~ linear to 200 ppm 259 Radial is ~ 1/8 th the signal of axial view; consider starting calibration at 80 ppm; ~ linear to 600 ppm
Fe 10 ppm scans 260 Axial ~ 14K net counts; linear to about 700 ppm; could go to 1,000 ppm with quadratic 261 Radial ~ 7K net counts; linear to about 1,400 ppm Strategy: 0 to 20 ppm with Fe 238 Axial 0 to 80 ppm with Fe 259 Axial 20 to 200 ppm Fe with 238 Radial 80 to 600 ppm Fe with 259 Radial 600 to 1,000 ppm Fe with 260 Axial (quadratic) 600 to 1,000 ppm (+) Fe with 261 Radial Evaluate other possibilities
Example of How to Implement: Set Min and Max conc for each wavelength Note: Values are in % with some extrapolation Have to set up a Mean or average column Mean will average appropriate results and can serve as the single result processed by LIMS
Additional Options to Alter Range Pump tube configuration sample vs. internal standard flow rates create a dilution effect Introduction systems and nebulizer type Quadratic curve fit (be careful) Others?
Concerns with Broad Calibration Range? Higher points on curve have disproportionate influence Not just relative value, but the squared term Mainly concern about linear curve fit as quadratic has curvature Misleading: correlation coefficient may look good, but lower concentrations could be compromised Quadratic curve fit may help some Can get better low-level data if split up broad calibration range into segments 1. Can use multiple line approach previously described 2. Can even do this with the same wavelength Select twice and rename
Concern with Point Distribution Serial dilutions can create some problems 10 fold dilutions: 500 ppm, 50, 5, 0.5, 0 Easy, quick and meets 5 points per curve criteria Intense signal linear? Great R 2 = 0.99999 Essentially drawing a straight line from 50 to 500 as lower points have minimal impact on curve
Reality? More points with more even distribution can indicate curvature In this case, selected concentration range is too broad or select different curve fit
Apply Quadratic
Concerns with Dynamic Range Demonstrate that higher points on calibration have disproportionate influence Can apply greater weighting to lower standards, but using what criteria? May not be allowed? 0 to 400 ppm curve Bottom 3 stds (i.e. 3, 5 & 10 ppm) deliberately diluted by 50% Still decent R 2 = 0.9998
Alter Top Standard What happens if alter just top standard by 10%? Can more easily see the curve is bad R 2 = 0.998 What s the point? Higher standards primarily influence the curve and therefore may not best represent the lower calibration range Small errors in the higher standards can negatively impact the accuracy of the lower concentrations
Recommendation? Line splitting Example 0 to 800 ppm using same wavelength Select same wavelength three times Rename: 1. LO Ca183, 2. MID Ca183 3. HI Ca183 (example)
Mid Ca183 Lo Ca183 Good low recoveries, may be even better if use narrower curve
Some Topics for Future? Different introduction systems Exposure length and number Pump tube selection and condition Matrix matching e.g. what about cumulative matrices from stock standards Dilutions - do you compensate for matrix contribution from sample? Internal standards and selection Spectral deconvolution Uptake and rinse times Others??
Summary ICP-OES instruments are well suited for fertilizer analysis, with only limited exceptions Multiple options to cover a broad dynamic concentration range and get both low detection capability and minimize the need for dilutions Requires using ICP features at a higher level Instrument software manages most of this Can always try different options and run in the background Compare results from different configurations Don t need to include result in your mean until validated Doesn t take any extra analysis time Can result in improvement in data quality