Representative Sampling and Proper Use of Reference Materials
Report
Oswald U. Anders A Plenary Discussion held at the International Conference: The Modern Trends in Activation Analysis, Munich, Germany, Sept. 15, 1976
Contributors to the Discussion were:
Sampling Methods
O. U. Anders, USA, moderator J. I. Kim, Germany H. I. Al-Shahristani, Iraq R. L. Brodzinski, USA G. F. Clémente, Italy R. Cornells, Belgium J. J. M. De Goeij, Netherlands F. Girardi, Italy G. Guzzi, Italy K. Heydorn, Denmark L. Kosta, Yugoslavia F. Lux, Germany S. Meloni, Italy J. Mitchell, USA J. W. Morgan, USA J. V. Niewodniczanski, Poland R. M. Parr, Austria J. Pauwels, Belgium T. B. Pierce, UK H. Rook, USA B. Sansoni, Germany R. E. Wainerdi, USA
This report is a synopsis of a plenary discussion held at the International Conference: Modern T r e n d s in Activation Analysis in Munich, Germany, on September 15, 1976. Since the subject is of concern to all analytical chemists, it is hoped t h a t this report will help generate a more acute awareness and appreciation of the problems in the scientific community. T h e purpose of the discussion was to provide a forum for the spontaneous exchange of ideas on a subject of general interest to analytical chemists t h a t has taken on new aspects and gained considerable importance through the recent emphasis on trace element analysis and the development of high-sensitivity analytical techniques. T h e session was attended by some 250 scientists from all over the world.
Sampling is the operation of securing portions of an object for subsequent laboratory testing and analysis. Representative samples should always be taken in such a way t h a t a reasonably close knowledge can be obtained about the object being studied. Random selection of samples in many cases cannot accomplish this. Much effort, for instance, was expended to sample the moon in a few locations to gain new knowledge about the materials making up the earth satellite. T h e sites were carefully selected to be representative. If, in a comparable study, the E a r t h were sampled in but one place, and this was a city, the analysis of the samples obtained could lead to the conclusion t h a t the E a r t h is made of concrete and asphalt. T h e number of samples must stand in a relationship to the size of the object being studied and the purpose of the investigation. T h e size of the samples must be sufficient to represent properly the inhomogeneities present in the material. When it is the objective of the study, for instance, to find out how much tin can be obtained per ton of a given ore, it is not import a n t to study the distribution of tin between the individual chunks of ore as mined, but rather find the total tin content per ton of rock. T o obtain this knowledge, a relatively large sample must be taken, which is truly representative of the nonhomogeneous object. For the analysis, one might homogenize and subsample to reduce the bulk of the material, and select a smaller aliquot for the actual tests. If it is, on the other hand, the objective of a study to determine the distribution of copper in liver or trace silver in white lead paint used in paintings, several samples must be taken t h a t are small relative to the object under investigation. Homogenizing a large sample to reduce bulk would, in this instance, obliterate the parameter under investigation.
Dow
Chemical Co. Midland, Mich. 48640
The method of sampling and the selection of sampling sites is thus directly related to the purpose of the study. It is also, however, related to the method of analysis and testing t h a t is subsequently to measure the samples. S a m p l i n g as R e l a t e d to Analytical Method
Many of the modern sophisticated methods have been developed to achieve high sensitivity. They are able to detect very small amounts of the substances of interest. Because of their high sensitivity, they are, however, often not capable of analyzing large samples. When studying inhomogeneous objects, the aliquots must, therefore, be small so t h a t the technique can handle them and yet be representative of the object relative to the purpose of the investigation. If one has a sample representative and able to give the desired information about the object, one can homogenize this sample and take an aliquot of the homogenate, small enough so the analysis method can handle it. T h e other alternative is to perform analyses on many small aliquots of the sample and then average the results. T h e first approach would, of course, obliterate any fine structure and might involve changes in the sample due to loss of volatile elements or the addition of impurities. Such changes in the sample can often be minimized by choice of the proper procedure. They constitute, however, compromise, and their effects on the results must be evaluated. T h e second approach will permit the investigation of microinhomogeneities. Unless, however, the entire representative sample is studied, it will lead to erroneous results since, in a heterogeneous material, the distribution of the results is broad and the true average difficult to obtain. When, in multiple determinations, insufficient amounts of the sample are analyzed, one runs the risk of not finding the element of interest in some of the aliquots. T h e ordinary mathematical
ANALYTICAL CHEMISTRY, VOL. 49, NO. 1, JANUARY 1977 · 33 A
approach to interpret the results is to take the mean of the resulting distribution. In doing so, one implicitly presumes the distribution of the data to fall on a normal curve. This, as was pointed out in the discussion, throws a positive bias onto the result, since there is no negative part to such a distribution, and the non-Gaussian curve is biased toward higher positive values. T h e smaller the sample size used, for such multiple subsampling, the greater the difference of the " m e a n " from the value obtained by a technique able to analyze the entire representative sample with good precision. A microtechnique such as, for instance, the electron probe, analyzing many spots on a sample will, thus, always obtain relatively higher values when the results are averaged, than a technique able to analyze larger amounts. T h e latter, on the other hand, necessarily will lose resolution and detail, when it is important to study microinhomogeneities. Sampling is done for a given purpose and the procedure followed should minimize alterations of the sample due to losses or contamination which would interfere with the analytical technique to be used to test the samples. Sample Operations It is often possible for specific purposes to choose sampling techniques in such a way t h a t the unavoidable changes and contaminations resulting from the sampling operation are irrelevant to the information to be gained, and will not interfere with the analytical method chosen. An alternate approach is to follow a very well-defined and well-studied protocol. Such a protocol would allow even deliberate alterations of the sample. But, since the protocol is well understood, the effects of the sampling operating and the sample preparation steps can be accounted for in the subsequent data interpretation. T h e collecting of general samples t h a t can be used for all possible subsequent studies, and will not be altered in any way by the sampling, is an impossible task. It is only possible to select procedures t h a t will minimize the interferences and changes of the sample with respect to the objectives of the investigation. If it is the intent to obtain samples for many purposes, then these purposes must be well understood and a sampling procedure chosen that will not cause significant interferences for any of the analysis methods. If such is not possible, one must sample the object several times by different procedures so t h a t samples of different preparation can be used independently for the various types of subsequent studies.
rors due to digestion or homogenization of the sample may be encountered and be significant for the results of the study. Sampling for Trace Analysis
Water is always in a container, say a river bed, soil, or a pipe, and some of the material with which the water was in contact is dissolved in it. Water samples are thus, usually collected to study the environmental parameters t h a t have affected the water. When sampling, the water will be exposed to yet another environment, the sampling container, and will be altered by it in certain ways. Some of the material in the water might flocculate or be absorbed on the walls of the sampling container and thereby be lost to the analysis. On the other hand, such deterioration of the sample may be avoided by the addition of chemicals t h a t do not themselves interfere with subsequent tests. Definite changes however, do occur to the sample in either case. These changes must be kept small, with respect to the purpose of the study. T h e objective of the latter and the chosen analytical method determine what constitutes an adulteration of the sample. For water sampling, one might choose a container t h a t would introduce 2 ppb of copper to the sample. If one expects to find copper in the 1-ppm range, there is no problem. T h e same sampling container would, however, completely obliterate the evidence, if the water were to contain only 1 or 2 ppb of the same element. Even if one only freezes a sample, changes can occur. Certain substances might be precipitated or through accumulation of dissolved salts in the last-freezing mother liquor, cause inhomogeneities in the resulting ice. Such changes in the sample may or may not be reversible. It is, of course, possible to avoid some of these errors by employing a method t h a t analyzes the entire object. Whenever t h a t is done, the results describe something t h a t does no longer exist. Even here, however, er-
34 A · ANALYTICAL CHEMISTRY, VOL. 49, NO. 1, JANUARY 1977
When dealing with trace analyses, it is, in fact, at times desirable to take very large aliquots and even the entire object under investigation to obtain as much of the element of interest in the sample as possible so t h a t one can still detect the trace which might otherwise disappear below the limit of detection. In general, it is however, not the intent of a sampling procedure to destroy completely the object under investigation and so obliterate the evidence. It is, in fact, bad practice to destroy completely any sample, if only aliquots need be analyzed. Subsequent investigations, to follow up on interesting results, would be impossible. For this reason, it is often more desirable to subsample and analyze several aliquots, rather than to homogenize the sample when trying to reduce its bulk. In the field of activation analysis, considerable emphasis in the past several years has been placed on proper reporting of error evaluations and the use of statistical techniques for data treatment. Considerable efforts were also made to improve sensitivity and precision of the major measurement steps. It is only now t h a t the field is becoming aware t h a t the errors in the analytical results often are influenced much more by the sampling and sample preparation techniques than by the measurement steps themselves. Once a sample is in the laboratory, the analyst has control. T h e actual sampling, however, must often be left to outsiders, and control of their sampling procedures is difficult. Yet, in studies involving trace elements, the sampling procedures are likely to cause very significant errors. It is for this reason desirable, and often necessary, t h a t the person most intimately familiar with the purpose of the study and the analytical technique carry out the sampling. He then can make sure t h a t no alterations occur to the sample, alterations t h a t might influence the desired results. When reading reports of studies involving trace element analyses, it is often the experimental details describing the sample collection and sample preparation t h a t allow the reader to judge whether the results reported could, in fact, be obtained. T h e recent practice of minimizing the description of experimental detail in journal articles was severely criticized by the participants of the discussion. There are a t present various groups engaged in the writing of sampling procedures. Until their procedures are
published and can be accepted as standards to be quoted, sampling de tails must be included in t h e papers. People involved in reviewing manu scripts to be published were urged to insist on their inclusion.
Reference Materials
SPECTRASPAN IV Low Cost Computer-Aided P l a s m a S p e c t r o m e t e r s for Trace Element M e a s u r e m e n t s
SMI is recognized and respected as a leader in atomic analysis and today counts more plasma emission spectrometers in use in labo ratories and industries worldwide than any other manufacturer. Plasma spectrometry is widely accepted because it combines the best features of atomic absorption and atomic emission techniques.
T h e discrepancies found in t h e re sults of round-robin experiments com paring methodology and techniques used in different laboratories caused considerable concern. Much of the scatter of the data was blamed on sampling procedures and sample han dling to which not sufficient attention had been paid. Reference materials prepared for such studies should be prepared with the specific purposes in mind and these purposes should be clearly stated by the organization issu ing t h e reference material. T h e refer ence material itself must be complete ly described and a minimum sample size representative of the material must be stated. These must be clearly understood also by t h e users. It is only then t h a t analytical techniques inca pable of performing the analysis of t h a t reference material can be avoid ed.
SPECTRASPAN IV — SMI's newest and lowest-cost plasma spec trometer, is a compact, integrated system that performs analyses of elements at trace to major levels including the so-called difficult elements with a minimum of sample preparation. Sequential quantitative analyses of single elements or comprehensive qual itative analysis are available with Spectraspan IV. Conversion between modes of operation can be achieved in seconds. Among the outstanding features of SPECTRASPAN IV are a unique Echelle grating that provides high dispersion of all wavelengths from 1900 to 8000 A in a 4 χ 5 in. area and a built-in microprocessor for ease of operation and increased system effi ciency. SPECTRASPAN IV is an up-to-date analytical system that offers substantial capability at a modest price. For multichannel analysis capability ask for information on Spectraspan III.
A low cost integrated system for sequential analysis
SPECTRAMETRICS INCORPORATED 204 ANDOVER STREET, ANDOVER, MA 0 1 8 1 0 (617)475-7015 CIRCLE 193 ON READER SERVICE CARD
36 A · ANALYTICAL CHEMISTRY, VOL. 49, NO. 1, JANUARY 1977
O s w a l d U . A n d e r s is associate scien tist of the Michigan Division Analyti cal Laboratories and the reactor su pervisor of the Dow TRIGA research reactor. He has published extensively in the field of radiochemistry. Present research interests also include nuclear power plant decontamination. Dr. An ders received his P h D from t h e Uni versity of Michigan in 1957. H e was a member of the Advisory Board of AN ALYTICAL C H E M I S T R Y 1965-67. He
served on t h e committee sponsoring the International Conferences: Mod ern Trends in Activation Analysis, 1968-76, and is presently vice-chair man, chairman elect of the Isotope and Radiation Division of the Ameri can Nuclear Society.
