Improving your spectrometric analysis of aluminum.

Four viewpoints reveal methods and problems that emerge in spectrometer techniques for aluminum alloys.

Working collectively, members of the AFS Aluminum Division's Molten Metal Processing Committee (2-G) delved deeper into the analysis of aluminum alloys and problems that might develop using spectrometer techniques. As a panel at th 1993 AFS Casting Congress, committee members covered topics such as principles, theory and sampling and analysis techniques that must be controlled to achieve accurate and reproducible results.

Participating in the panel discussion were Norman J. Brooks, Reynolds Metals Co.; David J. DeWyse, CMI Precision Mold, Inc.; Frank DeHart, Stahl Specialty Co.; and the author. Highlights from the four presentations follow.

Current Instrumentation

Brooks explained the principles of spectrometric analysis, as well as some of the more commonly used equipment. He touched on recent technical advances in instrumentation and described the actual analysis mechanism.

Brooks pointed out that the simple act of striking a controlled arc against the specimen was the start of the analytical method, and that the capture and measurement of energy emitted from that arc burn constituted the essence of the analytical technique. This is illustrated in Fig. 1, and shows how the emitted energy is converted into a concentration level by photomultiplier tubes that feed information to a controlling computer for conversion to the common elemental weight percentages in foundry alloys.

Today's instrumentation, he said, can produce an estimate of the sample material's composition in 10-15 seconds. Near real-time production decisions ca be made to adjust the chemical composition to achieve the best combination of mechanical properties or foundry characteristics affecting castings. Thirty elements commonly found in aluminum foundry alloys can be routinely measured with current spectrometers, with an accuracy and reproducibility unheard of 10 years ago.

The accuracy of spectrochemical measurements is conventionally measured by comparing the deviation of the quantitative expression to the level defined in an accepted reference material. Most commonly, the reference materials are defined by wet chemical determinations, which are subject to variation and rang between analysts. The reproducibility and accuracy of the spectrochemical measurement is a function of the concentration, and may be degraded significantly at low elemental percentages.

Discussing repeatability of spectrochemical measurements, Brooks said successiv measurements won't always produce the same value, and multiple burns are usuall made and averaged to develop the reported analysis. Excessive range in the individual burns is normally a sign of a problem in the sample or its preparation.

He described some of the new alloy sorters and cautioned that they vary greatly in their portability and analytical capability. Brooks recommended that if they are being considered for purchase, they should be evaluated in the environment in which they will perform and be checked for accuracy on samples that are representative of the material requiring analysis. He emphasized that all of th units are useful only as alloy sorters, and aren't designed to produce quantitative analyses.

Brooks concluded that today's spectrometers:

* are capable of accuracies of 1-3% of the amount present (for major alloying elements);

* offer repeatability of no worse than 1-2% on properly prepared samples;

* have detection limits in the range of 0.0001-0.001% for routinely determine elements.

The instruments can make a major contribution to improved process control in casting operations, he said.

Standards, Curve Construction and Calibration

DeWyse pointed out that the basic spectrometer--as received from the manufacturer--cannot just be plugged into a wall outlet, turned on and expected to perform quantitative analytical functions. The spectrometer must be "taught" how to interpret the signals and data that it will receive. This is accomplishe by inputting a library of information relating signal intensities to chemical concentration for any particular element.

To do so, the relationship between optical emission intensity and concentration must be determined using certified standards, which are available from several sources. The standards for foundry alloys are usually cast materials produced under rigorous control to achieve a high level of homogeneity.

The standards are then sent to at least five laboratories for wet chemical analysis, and the values derived (usually including some range) are averaged to produce the analysis listed in the certification documents.

DeWyse noted that standards should be selected with two distinct purposes in mind. The first consideration is to provide complete coverage throughout the concentration range applicable for all elemental wavelength channels included i the spectrometer hardware.

The second is to match the alloy types cast in the foundry. It is wise to have at least one standard whose analysis closely matches each of the aluminum alloy being analyzed. Also, its composition should include modifying elements for the silicon phase (if used).

Once standards and reference materials have been selected and gathered, the calibration process can begin. The first step involves an operation called "profiling," which consists of manually tweaking the instrument alignment to achieve maximum light intensity on the photomultiplier tubes. Calibration is continued with burns on the chosen standards and the recording of the spectral intensities developed relative to the chemical elements of interest. The goal i to convert these intensities to ultimate concentrations in the sample.

Once the selected standards have been burned and the data stored to disk, the calibration curve fitting can begin. While most modern computer software will develop the curves, there is a good deal of judgment and observation that requires input of the analyst. The object is to develop the best possible fit, and this is accomplished by using a weighted system to emphasize the standards that more closely match the composition being analyzed. In some cases, it may b necessary to delete a standard from the curve fit if the analysis seems to be too far from the balance of the data base.

DeWyse gave several examples of curve construction to define analytical curves. He noted interelement interferences can occur when other elements in a sample emit radiation at wavelengths that contribute significant intensity to the analyte channel.

If such conditions exist, the intensity contributed by the interfering matrix element may report an excessively high analyte concentration. These interferences can be noticed during calibration if some data points fall significantly above (or below) a smooth curve through a majority of the points.

Figure 2 illustrates such an interference, showing a calibration curve for antimony (Sb) on the 2598.06A spectral line, which receives energy from an intense adjacent iron line at 2599.37A.

Standards containing high levels of iron show an abnormal displacement from standards with lower levels of iron, and suggest false higher concentrations of Sb if curve adjustments were not made. DeWyse said most modern software package have provisions for interelement corrections, and can easily be made if the nee is identified.

He also said standardization and recalibration are needed because of changes in the optics, the excitation source, processing spectrometer electronics, and eve ambient room temperature or humidity. These changes can cause drifts in the intensity ratios, when compared to those recorded during the initial calibration. If uncorrected, errors will occur and false concentration readings will result.

Table 1. Coupon Preparation Parameters

                                            Levels
Variables                     +                                 -

Mold Temp.                 570F (299C)                     300F (149C)
Metal Temp.               1450F (788C)                    1350F (732C)
Quench                       Yes                               No
Surface Finish   [less than] 70 [[micro]inch]   [greater than] 150 [[micro]inch
Depth of Cut                 18%                               36%
Gas Level(*)                 0-2                               7-9

* Gas Level per Stahl Specialty Co. Comparison Chart

The drift is corrected by using a procedure called "Two-Point Standardization," which measures standards at high and low concentrations of elements of interest and makes appropriate adjustments to the curves so accurate analyses may be mad at ambient conditions. "Single-Point Standardization" (also known as Type Recalibration) is used to improve analysis estimates in a particular alloy family.

DeWyse stressed the need for foundries to maintain equipment and establish good practices. He noted that accurate analyses are possible and are a valuable tool to any foundry operation.

Spectrochemical Coupons

DeHart discussed the importance of following good practices in the production and preparation of sample coupons for analysis. He said ASTM E716 defined "Standard Practices for Sampling Aluminum and Aluminum Alloys for Spectrochemical Analysis," but the influence of those practices upon the actual analysis results developed isn't commonly recognized.

For example, several mold types are defined as acceptable, and practices are suggested that include preheating the mold, filling at a uniform rate, machinin the sample to a constant depth and surface finish, and analyzing the sample in preferred locations, depending upon mold design. While all are important, other factors--such as alloy type, pouring temperature, mold temperature and gas leve in the melt being sampled--can influence the results of the analysis estimate.

DeHart defined the results of a detailed investigation at Stahl Specialty Co. o the importance of the ASTM defined practices and the influence of the foundry parameters. Three basic alloy types were evaluated in the effort: 319.1, A356.2 and a special eutectic composition. The work was done with commercially available production melts, and all samples were poured from dry hearth reverberatory furnaces with 3000-1b holding capacities.

The initial work concentrated on evaluating the impact of the ASTM parameters o mold preheat, amount of stock removal and surface finish of the machined samples. The two most common mold types used in the industry (Type A--a vertically parted sample mold with a top pour position, and Type B--a horizontally poured sample with a center sprue position) were equipped with thermocouples to quantify the mold temperature during sample production.

Consecutive samples were poured as rapidly as possible, and the increase in mol temperature was tracked for each type of mold, with samples analyzed at different temperatures. Because of differences in mold mass, Mold B heated up much more quickly than Mold A, possibly influencing the analytical result.

Another analysis studied stock removal's impact on analysis uniformity. The AST recommendations of 14-22% removal produced almost equal results with both Molds A and B, and confirmed a more significant effect upon the eutectic composition than either 319.1 or A356.2 alloys. Similarly, smooth surface finishes consistently produced more uniform readings than rough finishes.

The second half of the Stahl work, DeHart reported, evaluated the impact of foundry process parameters on the analyses generated. The variables listed in Table 1 were evaluated with a Taguchi Design of Experiment technique, and by averaging the % relative standard deviation (RSD). Eight foundry trials were conducted, using an L-8 orthogonal array, with spectrochemical analyses conducted in four different labs to minimize bias from analyses techniques.

While the data was still understudy, it was apparent that control of the castin parameters was also important to developing uniform and consistent analysis results, DeHart said. The eutectic alloy composition exercised the most sensitivity to variations in foundry practice, with mold and metal temperatures found to be important, as well as depth of cut and gas level. Additional work o the subject is planned.

DeHart observed that the precision of analysis is improved by casting samples i molds in the lower range of operating temperature and developing a good surface finish while removing less than 22% stock from the sample coupon. He also noted that the 319.1 and A356.2 alloys are more tolerant of changes in practice than the eutectic alloys.

Good Specifications

Spectrometric analysis is only an estimate of the chemical composition of the sample. For that matter, any chemical analysis is only an estimate of the composition, since there are almost always differences in the numbers when any two laboratories compared results. This is true of the wet chemistry analyses used to define "Certified Standards," as well as many routine production analyses. While steps can be taken to improve the reproducibility of the estimates, they are still only an estimate and not an absolute value.

Ways to improve the accuracy and reproducibility of the analyses include operator training, proper maintenance and calibration of the equipment, establishing and following consistent sampling and preparation practices, and confirming results with regular cross-checks with other laboratories. ASTM doesn't offer standards of performance for checks between laboratories, but mos large corporations have established their own criteria for acceptance. Table 2 lists a typical set of accepted tolerances for spectrochemical comparisons between two laboratories.

Table 2. Selected Examples of Generally Accepted Tolerances for Spectrochemical
Comparisons (Two Laboratories)

Range (%)                       Tolerance

0.001-0.009                     [+ or -]0.002
0.041-0.10                      [+ or -]0.010
0.76-1.00                       [+ or -]0.05
1.51-2.00                       [+ or -]0.07
5.51-6.00                       [+ or -]0.17
[less than]7.01                 [+ or -]3%

Although the panel said consecutive burns on the same sample would never produc the same analysis result, it didn't point out the magnitude of the variation that might be expected. Figure 3 shows a graphical depiction of the results of 50 sequential burns on the same standard, while using two different spectrometers located in the same lab. Unit 2 produced analyses with a standard deviation of less than half of Unit 1, and also developed a slightly higher average silicon level (+0.14%).

While the results from Unit 2 appear to offer a higher level of "accuracy" and be the preferred result (using the criteria of Table 2), the analyses would be considered equivalent because the actual range between instruments was about 1.5%.

Most foundries and diecast shops recognize the limitations of analytical equipment, and specify their materials with realistic ranges on analysis and chemical limits. "Off-sets" are frequently built into the specifications to produce analyses that are consistently inside of their customer limits. In that manner, castings are poured with an adequate margin of safety of conformance to the governing specification. Therefore, bitter disagreements over a difference in 0.01-0.05% analytical result are avoided.

The economic impact of establishing restrictive specification limits, however, should be considered. In any manufacturing process, restrictive tolerances always add cost to the product. The direct benefit developed should be carefull assessed to ensure that a return on the investment is achieved.

The achievement of metal chemistry control should be one of the first steps in establishing melt process control, and should then be followed by addressing other melt parameters such as cleanliness, modification level, grain refinement and gas level.