Problems of Quantitative Chemical Spectral Analysis *

Walter Gerlach (Munich)

The task of chemical spectral analysis consists in determining the chemical composition of a material from its spectrum. In many cases spectral analysis serves as an aid to chemical analysis; moreover, it makes possible the solution of problems that are in general insoluble by the methods of chemistry. The latter include cases in which it is necessary to carry out a complete analysis in the presence of very small amounts of material, or else to establish, as quantitatively as possible, the presence of minute impurities. Such minimal masses may constitute a uniformly distributed contamination; however, it is also possible that the material contains very small inclusions, grains, or something similar, whose nature and composition must be determined. The problem of finding uniformly distributed foreign substances includes above all the task of detecting small concentrations of heavy poisonous metals in solutions, for example mercury, lead, copper*. In what follows, some special problems** will be considered, which were

* Naturwissenschaften, 1931, transl. by E. Brumberg.

** Cf. R. Willstätter, Naturwissensch., 18, 868, 1930.

*** In a report delivered by me on November 21, 1930, at the Munich Physical Society, only the problems dealt with by Scheibe, Ruxardt, van Calker, and myself were considered. Therefore it is necessary to point, for example, to the important experiments of Scheibe (Erlangen) and his collaborators, who, in processing spectrograms, made use chiefly of direct photometric methods.

have recently been solved by means of chemical spectral analysis and which cannot be solved otherwise; consequently, such problems are those whose solution by chemical analysis had been regarded by experts as experimentally impracticable.

The following example characterizes the occasion that prompted us several years ago to take up these problems. In processing an alloy, one factory encountered a number of difficulties: the material exhibited brittleness not previously observed, despite the fact that chemical analysis gave the same composition for the “bad” and the “good” specimens. The presence of negligible impurities, which was considered to be the cause, could not be established with certainty. However, the very first photographs of the spark spectrum revealed traces of lead, which were entirely absent in the “good” alloy. Thus not only was the cause of the low quality found, but it was also ascertained that the material which had proved unsuitable had been obtained from another factory.

This small success gave us the resolve to test the general applicability of the method also for solving quantitative problems, although, for example, Kayser in his book on spectroscopy (Handbuch der Spektroskopie) considers such a method completely inapplicable. The problem consists in determining the quantitative content of a certain metal in the base material from the intensity of spectral lines. The difficulties are due to two circumstances: 1) the absolute intensity of spectral lines is practically unmeasurable, and 2) the relative intensity of lines is not independent of the discharge conditions.

I. Method

We followed the following course: the content \(x\) of an impurity metal \(Z\) in the base metal \(G\) is determined by comparing the intensity of the spectral line \(\lambda_Z\) with the line of the base substance \(\lambda_G\), avoiding as far as possible the case \(\lambda_Z = \lambda_G\). For the alloy \(G + n\%Z\), such pairs of lines \(\lambda_Z, \lambda_G\) are found which at \(a, b, c\ldots\%Z\) in \(G\) have equal intensity. These special points we call a corresponding pair of li-

WALTER GERLACH

the lines. \(x\) is determined by means of objective or else subjective photometric interpolation. The influence of the discharge conditions is reduced by selecting for comparison such lines whose intensity depends little on these conditions. In addition, definite discharge conditions are established, not by electrical indicators, but directly by optical-spectroscopic ones. This is achieved by such a choice of the electrical quantities of the spark circuit that there is a quite definite ratio of the intensities of the arc and spark lines*. For adjustment we choose lead, whose arc line 2657 is, in intensity, approximately equal to the neighboring spark line 2562. Here visual comparison is sufficient, since again what is required is the establishment of equality (or even approximate equality).

If a certain salt is being analyzed for the metal \(Z\), it is dissolved, and to the solution a known mass of another metal \(M\) is added, so that the corresponding pair, previously obtained for the combination \(M + n\%Z\) on solid electrodes, is selected. For, on the basis of all the cases investigated, it should be regarded as reliable that the ratio of the intensities of the corresponding pair of lines of \(M\) and \(Z\) does not depend on whether alloys are evaporated during the passage of the spark between solid electrodes, or by means of a spark produced between gold electrodes and a solution of the alloy in acid. However, the absolute intensity changes somewhat and, in particular, a dependence on the acid anion is observed; thus, for example, there exists the following sequence, arranged in decreasing intensity and almost always valid:

\[ \mathrm{M \cdot NO_3 > M \cdot J > M \cdot Br > M \cdot Cl > M \cdot SO_4.} \]

However, the ratio \(\lambda_M : \lambda_Z\) always remains constant,

* The method has greater similarity to the determination of “spectroscopic parallax” in astrophysics. A closer method is found in the monograph by W. Gerlach and E. Schweitzer: Die quantitative chemische Emissionsspektralanalyse, Leipzig, Leopold Voss, 1930.

II. Limits of Sensitivity of the Method

There are many indications in the literature that the limit of detectability, i.e. the minimum concentrations at which recognition of an element is still possible, under equal conditions of observation, is a function of the position of the element in the periodic system. In particular, it is asserted that it decreases downward along a vertical column of the system. Rutkhardt carefully investigated this proposition. Fluctuations in detectability always proved possible to attribute to differences in the strength of the background of the bands, or else to external conditions. On the contrary, there is a distinct difference between elements situated on the right and on the left in the periodic system, and this difference is physically self-evident: it is connected with the complex structure of the spectrum, with multiplet terms. For example, with a certain setup in a solution of copper salt it was still possible to prove the presence of \(0.001\%_a\)* of lead and only \(0.01\%_a\) of nickel (relative to Cu).

The limit of detectability also depends very strongly on the method of excitation of the light. In general it varies around \(0.001\%_a\). In this area there still remains very much unfinished work. Thus Rutkhardt found that for arsenic (and also antimony) the worst method of excitation is the spark in air between solid electrodes, somewhat better is the spark in solution, and most advantageous of all is the spark in an argon atmosphere.

Another question: what absolute magnitude of mass is sufficient for spectroscopic detection? Here the sensitivity of the method differs for different metals and, above all, naturally depends on the luminosity of the spectrograph and (as we noted above) on the method of excitation of the light, the method of obtaining the spark. We carried out several control experiments by the method of Bäil and Ami, by electrolytic deposition of a known mass of metal (calculated according to Faraday’s law) on another base metal. In this case the limit proved to lie between

* \(\%_a\) — atomic percent.

$10^{-7}$ and $10^{-9}$ g. The question was also investigated for solutions of salts. For example, to obtain a spark between liquids, the electrodes used for filling required $4\ \mathrm{cm}^3$ of solution. The presence of the Sr 4077 line could still be established if the solution, in addition to calcium nitrate (15%), contained another $2.4 \cdot 10^{-6}$ g of strontium. Since no more than $1\ \mathrm{cm}^3$ of liquid evaporated during the experiment, we obtain, as the smallest mass still sufficient to prove the presence of Sr in the solution, a value less than $6 \cdot 10^{-7}$ g.

It should be noted that the salt solution must have sufficient conductivity, which can be achieved by adding some other salt that has been preliminarily tested spectroscopically.

III. Local Analysis

In practice, the analysis of the smallest inclusions in a metal is often required. We received for investigation a considerable number of gold, silver, and platinum objects in which the tiniest grains or cracks had been found (mostly during the polishing of valuables). In all cases the analysis was successful: in one case it was a matter of grains in gold consisting of iridium and osmium; in another case, in a plate of noble metal rolled with insufficiently clean rolls, there proved to be grains of hard solder. Further, there was a case of Si + C, which had evidently entered during melting from the crucible. A fracture of a platinum plate that appeared during rolling could be attributed to the presence of silicon, which was found exclusively at the sites of natural fractures. It should be noted that in most cases chemical analysis—even qualitative—proved completely powerless.

A university analytical laboratory investigated the weathering of a refractory stone that had become brittle. After unsuccessful attempts to carry out a chemical analysis, the residue of the powder was given to us. The powder was dissolved in hydrochloric acid, and a spark spectrogram was obtained in this solution, which at once revealed, along with Mg, Cr, Ca, etc., considerable quantities of vanadium.

By means of microanalysis or local analysis, a number of scientific problems are also solved: thus we were able to show that a slight addition of lead (up to 0.04%) makes gold brittle, since it is deposited at the edges of the grains, forming a brittle lead–gold bond. It can also be shown that, when small quantities of Cu and Ag are added to gold in the presence of lead, the copper together with the lead precipitates at the grain edges*, whereas the silver remains uniformly distributed in the gold. The method also gave good results in testing Tamann’s resistance limits.

IV. Testing the Purest Metals

For both scientific and technical questions, the purity of metals will long remain a pressing problem. Especially important is the case of testing for the absence of iron. Any material usually contains iron introduced by the processing machines; this iron, however, is easily removed from the surface. A brief washing in acid or, better, evaporation of surface contaminants by passing a spark before making the exposure is evidently sufficient for this purpose. Indeed, we found that copper wire of any diameter, as well as gold and silver, after such treatment prove to be entirely free of iron. The situation is different with platinum. We have not yet had platinum in which the presence of iron could not be demonstrated. It is remarkable that the wire under investigation showed different iron contents in different places. This circumstance led us to suppose that in most cases we are dealing with grains of iron oxide enclosed in the platinum. Analysis of platinum for iron plays a very substantial role for technically important catalyst gauzes; we believe,

* As is known, gold is purified of lead by the addition of copper and subsequent heating: the lead evaporates together with the copper (“Quartieren”).

that chemical analysis, at contents of \(0.1\%\) and less, is already insufficiently reliable. At the present time, magnetic measurements of metals and metallic alloys should not be carried out without spectroscopic testing of all samples. In this connection, an essential circumstance is the slight expenditure of material and the possibility of examining different places of a single object, or different grains of the powder under investigation.

The purity of lead—in particular the absence of bismuth—plays a decisive role in accumulator production. Here the circumstances are very favorable: it is possible to establish the presence in lead of even \(1 \cdot 10^{-4}\%\) Bi. Otherwise, up to the present time the situation has been with the analysis of copper for arsenic and of silver for tellurium; in this case the sensitivity limit of the method (with the spark method) is no higher than \(0.1\%\). Here further work is needed to find other, more advantageous methods of spark excitation.

We have never yet encountered absolutely pure platinum metals. Although the difference between “physically” and “chemically” pure platinum is immeasurably great, nevertheless in the former it was always possible to demonstrate the presence of other platinum metals, as well as of iron and nickel. All platinum metals are always found together. In the purest iridium and rhodium, specially prepared for us, it was always possible to demonstrate, along with platinum, rhodium, iridium, and palladium, the presence also of copper, iron, and nickel.

Recently one firm supplied us with very pure nickel, obtained from nickel carbonyl and then remelted in vacuum. In evaluating the results of magnetic measurements obtained for this nickel, the content in it of a considerable amount of carbon is of great importance. But the very proof of the presence of carbon must be carried out with extreme caution: to obtain the last carbon line on the photograph, the carbon dioxide contained in the breath and entering the spark during the exposure is sufficient. It is even easy to carry out a qualitative analysis of a gas for carbon (for example, for the compounds CO, CO\(_2\), CH).

In general, in chemical analysis,* a conclusion about the purity of a substance still says nothing if one is also interested in small traces that are still detectable spectroscopically. We once had silver tested, available in a quantity sufficient for carrying out chemical analysis. However, the latter gave no results, whereas spectral analysis revealed the presence of gold, silicon, tin, zinc, and thallium. Here, of course, the amounts in question are those whose order of magnitude is no more than 0.001 At%.

Particularly many difficulties are presented by testing the purity of metals rich in lines: not infrequently, even with precise measurements of wavelengths, it is difficult to establish whether an extremely weak spectral line is the last line of an impurity or a weak line of the principal metal. Then the suspicious line should be compared as obtained in different experiments; the following method is especially helpful: whereas the intensity of very weak lines of the principal metal, as a rule, changes strongly when the discharge conditions are changed (change of the self-induction or capacitance of the oscillatory circuit), the “last lines” of impurities at low concentrations are, within considerable limits, independent of them. Here, however, caution must be observed so as not to change the brightness of the background and thereby not to change the sensitivity of the apparatus.

V. Explanation of deviations of atomic weights

New experiments by Hönigschmid provide an important problem of analysis: calcium of various origins was investigated in order to attempt to explain deviations in the atomic weight of individual samples by the existence of a heavier isotope. The material was available in very small quantities; chemical analysis indicated the chemical purity of the substance, but from two samples there was obtained, for the atomic weight of calcium, a value equal, in round numbers, to 40.2

* We intentionally distinguish between metals given as chemically pure and metals whose chemical analysis has shown “complete” purity.

instead of 40.085. Spectroscopic examination revealed the presence of a small amount of strontium and the absence of barium. Strontium had escaped chemical analysis—the purification methods employed and considered sufficient proved unsatisfactory. Thus, by using spectral-analytical control, it is possible to investigate various methods for separating Ca and Sr. Here, however, the question arose: should the results of the measurements obtained be rejected when the strontium content is so insignificant, or should the presence of the heavier isotope be accepted? To resolve this question, the Sr content in the objects studied was determined from Gönigschmid’s experimental material (by comparison with a calcium salt having a definite Sr content) (Table I, first vertical column), and the “apparent” atomic weight of impure calcium was calculated from the amounts of strontium thus found and the normal atomic weight of calcium, taking the latter to be 40.085 (second vertical column).

TABLE I

A comparison of the “spectroscopic atomic weight” and the atomic weight determined chemically* shows that the presence of a heavier isotope of calcium should be regarded as unproven (Table II).

TABLE II

* Cf. Hönigschmid u. Kempter, ZS. f. anorg. Chem., in press (chemical determination); K. Ruthardt, ZS. f. anorg. Chem., 1931, in press (spectroscopic determination).

It follows from this that, when there are insufficient quantities of material, resorting to spectroscopic methods can avoid a dangerous error.

Undoubtedly, the further testing of purification methods used in analytical chemistry holds much promise; we would also like to investigate methods for purifying Hg, since a significant Cu content was recently found in mercury purified by the Institute.

VI. Detection of Mercury

In recent years the question of the physical and mental injuries caused by mercury has occupied many minds. We investigated what minimum quantities of Hg can still be detected by spectral-analytical methods. The solution of this problem was greatly advanced by the application of the new method of exciting a spark in a high-frequency circuit*.

We developed two procedures for determining the presence of Hg in a liquid. A small amount of copper salt is added to the liquid under investigation, and then Cu and Hg are precipitated by adding hydrogen sulfide. The black precipitate is filtered off with a small filter, which is then used as an electrode when the spark is excited in a high-frequency circuit. The sulfide is first converted by the action of HCl into the chloride and is then evaporated in the spark. For the smallest mass of Hg present on the filter and sufficient for detecting mercury, we obtained a value of \(4\gamma\) (the spectrograph had a light intensity of 1:10).

The following procedure gives more: Hg is deposited electrolytically from the solution onto a small tin foil \(0.01\) mm thick. The first experiments showed the possibility of qualitatively determining approximately \(0.002\%\) Hg in Sn. The electrolytically amalgamated tin foil is then evaporated in the spark; in this way a tin–mercury spectrum is obtained, so that, by comparing the Sn and Hg lines, on the basis of

* W. Gerlach u. E. Schweitzer, ZS. f. anorg. Chem., 1931 (in press).

comparative tables one can calculate the quantitative content of Hg relative to the known mass of Sn and thereby find the Hg content in the solution under examination. In one of the cases a tin foil was used, having a surface area of \(0.5\ \mathrm{cm}^2\) and a weight of \(3.5\ \mathrm{mg}\). To prove the presence of Hg, \(0.07\gamma = 7 \cdot 10^{-8}\ \mathrm{g}\) Hg is sufficient, since here the sensitivity limit lies at \(0.002\%\) of the total weight. Thus the proof of the presence of mercury in a liquid is carried out by means of electrolytic concentration of the substance. The fact that Hg is electrolytically deposited completely from solution onto tin was also investigated spectroscopically: the values obtained in the examination of a solution of known Hg concentration were compared with the values obtained for tin–mercury alloys with a known mercury content.

The examples given are sufficient to demonstrate the versatility of the method of qualitative and quantitative chemical spectral analysis. It should, however, be pointed out that such a delicate method requires particularly careful execution and, above all, a critical attitude toward the results obtained. It should also not be forgotten that every special problem requires a corresponding experimental procedure. With these reservations, the principle of the method may be recommended as an aid in physical as well as in chemical, metallographic, and biological investigations.