Fresenius' Journal of

Fresenius J Anal Chem (1992) 342:15-19

@ Springer-Verlag 1992

A new method for quantitative mass-spectrometric analysis of solutions A. I. Saprykin and Yu. V. Kovalev Institute of Inorganic Chemistry, Siberian Branch of the USSR Academy of Sciences, SU-630090 Novosibirsk, USSR Received March 21, 1991

Summary. The mass-spectrometric method for the analysis of inorganic solutions is not only a very sensitive technique but also one of broad applicability. The study of the behaviour of solutions of mineral acids chilled to - 7 0 ° C under a r.f. voltage from 7 to 15 kV indicated that solutions of hydrochloric acids containing 2 0 - 32% of hydrogen chloride allow to obtain stable ion currents. The equipment for the analysis of solutions on the spark source massspectrometer is designed and the procedure of analysis is described. The detection limits of 47 elements are in the range from 10 7 to 10 -1~ g. The accuracy of the determinations was established by the "added-found" method. The reproducibility is 0 . 1 - 0 . 3 4 for content levels of 1 0 - 7 - - 1 0 -9 g. The analytical possibilities of different spark source mass-spectroscopic methods for the analysis of solutions are compared.

The development of methods for the analysis of inorganic solutions is determined by the large variety of tasks including determinations of impurity contents in aqueous solutions of mineral acids. These tasks range from analyses of environmental objects to contamination control of the reagent chemicals used in microelectronics and analysis of etching solutions by determinations of impurities in semiconductor structures and films. One of the important aspects in the development of these methods relates to the use of chemical concentration in combination with highly sensitive multielement instrumental methods. Belonging to the group of the latter methods is spark-source mass spectrometry (SSMS) which is capable of simultaneously determining up to 70 impurity elements with detection limits of 10- 5 to 10- 7 mass % at a sample consumption of about 10 mg. Combinations of pre-concentration of impurities with SSMS analysis of the concentrates make it possible in a number of cases to lower the detection limits of impurities in high-purity materials by 2 to 4 orders of magnitude [1]. At the present time the SSMS analysis of solutions is mainly performed in two ways: by evaporating samples on a collector [2, 3] or on the surfaces of electrodes made of high purity materials [1, 4] or by freezing a liquid sample and keeping it in the solid state in a nitrogen-cooled electrodecrucible [5]. The lowest detection limits (as low as 10 -11 Offprint requests to .' A. I. Saprykin

mass %) have been achieved by evaporating concentrates of impurities on a carrier surface and analyzing the thin layer of the carrier with the dry residue [1]. The main shortcoming of this technique lies in its low productivity. Preparation of the carrier, evaporation of the sample and measurements usually take not less than 6 - 7 h which greatly limits the use of reference samples in this method and also its application for layer-by-layer analysis of semiconductor structures combined with chemical etching. Analyzing a frozen sample is a simpler procedure but with the use of a single internal standard, the impurities are determined with the accuracy of the respective sensitivity coefficients and the latter may vary by a factor of 2 to 10 depending on the conditions of the analysis and physico-chemical properties of the elements. Recently, liquid metal sources of ions (LMIS) have come into use [6]. They represent a system consisting of an emitter wetted with a liquid metal and an extracting electrode between which a voltage from 6 to 10 kV is applied. A liquid metal source of ions ensures a steady current of ions of up to 100 gA with a relatively narrow energy distribution (about 100 eV). The use of LMIS in the cathode regime has been shown to have a potential for mass-spectroscopic analysis of low-melting-point metals [7]. These studies allowed us to suggest the possibility of obtaining ionic currents from aqueous solutions under high-voltage performance. The main limitation in the use of liquid inorganic solutions in an emitter is associated with the vapour pressure that should not exceed 1 x 10 -4 Torr (1 x 10 .2 Pa) in order to ensure the necessary working vacuum in the ion source of massspectrometer.

Experimental Deionized water and highly pure solutions of nitric and hydrochloric acids, which are usually used to dissolve concentrates of microimpurities in chemical atomic emission and spark source mass-spectrometric analyses, have been chosen as objects for our studies. The temperature dependencies of the vapour pressures of these solutions indicate that below - 7 0 ° C they all have vapour pressures of less than 1 x 10 -4 Torr, enabling them to be put in the ion source of a mass spectrometer. The analyzing solution in an amount of about 0.1 ml was introduced by means of a micropipette into a tantalum emitter (Fig. 1), cooled to liquid nitrogen temperature and quickly introduced into the ion source chamber of the mass spectrometer MS-702 at a distance of 1 . 5 - 3 m m from the extracting opening. For heat removal,

16

20 0

3

4

2

-20 b-

-40 -60

5

4

3

2

1

-80

6

Fig. 1. Equipment for analysis of solutions on a SSMS. 1 tantalum emitter; 2 solution being analyzed; 3 flexible copper heat remover; 4 arm; 5 screw; 6 extracting opening of the mass-spectrometer

1

0

I

i

I

~

I

20

40

60

80

1O0

HCI (%)

Fig. 3. Phase diagram of hydrogen chloride solutions at low temperature. Composition of the solutions: 1 HCI×6 HzO; 2 HCIx 3 H20; 3 HC1 × 2 H20; 4 HC1 × H20

M

Fig. 2. Spark ion source. 1 Tesla transformer; 2 capacitor 50 pF; 3 high voltage diode; 4 emitter with analyzing solution; 5 extracting opening; 6 main slit of mass-spectrometer

the emitter was connected with a flexible copper rod to a container for liquid nitrogen. After a preliminary evacuation of the ion source to about 0.1 Torr the container was filled with liquid nitrogen after which the emitter temperature dropped to - ( 6 5 - 75) ° C within 15 to 20 min. This allowed the ion source chamber to be evacuated to about 1 x 10-s Torr. The emitter temperature was measured prior to the analysis directly in the source under vacuum by means of a chromel-alumel thermocouple whose junction was placed in the sample volume of the emitter filled with deionized water. After evacuation of the chamber an r.f. (1 MHz) voltage with an amplitude from 7 to 15 kV from a spark source (Fig. 2) has been applied between the emitter and the extracting opening of the mass spectrometer, with the negative polarity on the emitter, and the accelerating voltage + 15 kV was switched on. The behaviour of the chilled solutions of mineral acids under an r.f. voltage indicated that solutions of hydrochloric acids containing 2 0 - 3 2 % of hydrogen chloride allow to obtain ion currents that ensure rates of collection of exposures from 1 to 10 nC/min. This is explained by the fact that the melting points of such solutions are below - 70 ° C (Fig. 3) i.e. they remain liquid at the working temperature of the emitter and are ionized under the action of r.f. voltage similar to LMIS. This process continues until the solution is present in the emitter. The pressure in the ion source chamber in this case does not exceed 5 x 10- s Torr. For nitric acid solutions such ionization conditions turned out to be unrealizable since the samples are in the solid state at the • working temperature of the emitter. Attempts to initiate the

process by exciting an r.f. spark discharge between the emitter with the sample and the extracting electrode were not successful, because nitric acid solutions and water froze after termination of presparking. Increasing the emitter temperature resulted in intense sublimation of the solutions and a rapid growth of the pressure in the ion source.

Results and discussion

Thus, it has been established that under the action of an r.f. voltage .of negative polarity on hydrochloric solutions of certain concentrations chilled to - 70 ° C there arises an ion current of a magnitude determined by the amplitude, frequency and the length of the r.f. pulses. The proposed technique of introducing and ionizing solutions may be employed in mass-spectrometric analyses of water, inorganic reagents and etching agents as well as of concentrates of microimpurities soluble in mixtures of acids containing 2 0 32% of hydrogen chloride. To evaluate the analytical possibilities of the proposed procedure, the structure of mass-spectra and the impurity composition of a 6 mol/1 hydrochloric acid of the special purity grade 20-4 have been studied to which 2 x 10- 5 g/ml of strontium has been added as an internal standard. The study indicated the presence of intense lines due to complex ions of the type C1,+, (ClnHm) +, (ClnOm) +, Cln(OH) +, (H20)n(HC1) +, hydroxonium (H20),H + , hydrocarbons (CnHm)+, as well as of ions of the construction materials Ta, Cu, Fe, common impurities K, Na, Ca, Si and their oxides, chlorides and hydrides. To estimate the limits of detection of impurities a statistical processing of the spectra from a blank experiment (a 6 mol/1 solution of hydrochloric acid of the special purity grade 20-4) has been performed taking into account the variation in intensity of the lines from complexes overlapping the analytical lines of ions being determined as the r.f. voltage varies from 7 to 15 kV. In Table 1 are shown the analytical lines of impurity elements, the overlapping ions, the required resolution (M/AM) and the values of the absolute detection limits. The resolving power of the MS702 mass-spectrometer of approximately 3000 is in most

17 Table 1. Analytical lines of impurity elements, superimposing ions, the required resolution (M/A M) and detection limits (Mx,mi,) in the analysis of hydrogen chloride solutions Element Analytical line

Superimposing ions

M/A M

Mx,mln, g

Ag A1 As Au Ba Bi Br Ca* Cd Co Cr Cs Cu*

107 27 75 197 138 209 79, 81 40 111 59 52 133 63, 65

CI~ (H30 x C1)2+ (C12H) + (TaO) +

76500 2300 4200 8400

5 × 10 - 9 7x10 lo 5 x 10 - I ° 8 × 10 -7 4 x 10 -1° 1 x 10 -1° 2x 10 -1° 1 x 10 -8 8x10 lo 1 x 10-11 1 x 10- 7 1 x 10 - l ° 3 x 10 -9

Fe* Ga Ge I In K* Mg Mn Mo Na* Nb* Ni P Pb* Pd Pt Rb Rh Ru Sb Sc Se

56 69 72, 74 127 115 41 24 55 98 23 93 58 31 206 106 194 85 103 102 121,123 45 78

Si*

Sr Sn Ta* Te Ti T1 V W* Y Zn Zr

28

88 118,120 180 128, 130 48 203, 205 51 184 89 64, 66, 68 90

--

--

-

-

(C3H4) +

600 17100 1800 1700 700

C1+ (NaC1 x H) + (C1 x OH) + (C5H3) +, (CsHs) + (C~H8) + (C102) + CI~H30 × (H20)~-(C3H5) + (Nail) +, C~H30 x (H20)~(NaC1) + (NxOH) + (C13H) + (CO2 x H) + -

40 2300 7500 720 -580 2000 600 1800 1000 9500 1100 -

I x 10 -8 5 x 10- lo 2 x 10 7 2 × 10 -8 2 x 10 -1° 2 × 10 -8 1 x 10 9 3 × 10 - 9 1 x 10 -8 1 × 10 -8 2 × 10- 9 5 × 1 0 -11

1 x 10 -8 5x 10 lo 3 x 10 - l ° 8 x 10 lo 2 x 10-11 5 x 10 -9 6x10 -1° 1 × 10 8 2 × 10 . 9

(C2H4) +

1 600

1 × 10 - s

(C120) + -(Cue) + (Clx CH) + -(C10) + (CsH4_8) + (C120) +

3 700

int. standard 5×10 lo 1 x 10 -6 8 × 10 -1° 1×10 .9 i × 10 -1° 2× 10 -7 2 × 10 .9 5 x 10-to 2 x 10 -9 3 × 10 - 9

2600 2000 -2600 600 4100

* Noted by the asterisk are elements whose determination is limited by the blank experiment

cases sufficient to resolve the analytical lines of impurities and multiatomic ions. The detection limits of impurities were determined at exposures of 10 nC. Registration of larger exposures turned out to be unreasonable since with increasing exposures there is a proportional increase in the line intensities of the multiatomic species, the b a c k g r o u n d level and there appear re-discharge bands. As a result, improvement of the detection limits proves possible only for some of the elements with the mass spectra interpretation becoming increasingly complicated.

The reproducibility and accuracy of the proposed method have been checked by the "added-found" method. To accomplish this, stock solutions containing A1, As, Ba, Bi, Cu, Fe, Ga, Na, P, Pb, Sb, Te, T1 and Z n in concentrations of 1 mg/ml have been prepared by dissolving calculated a m o u n t s of metals or oxides in concentrated nitric or hydrochloric acids or mixtures of these. Solutions containing the above mentioned elements in concentrations of 10- 5 _ 10- 6 g/ml (solution No. 1) and 1 0 - 6 _ 10-v g/ml (solution No. 2) were prepared by diluting the stock solutions with 6 tool/1 hydrochloric acid of the special purity grade 20-4. The impurity concentrations in the first solution have been chosen in such a way that their contents exceeded the value of the blank experiment at least by one order of magnitude. The concentration of impurities in the second solution was close to the detection limits. 0.05 ml of the obtained solution was placed into the emitter and the analysis was performed. The mass of the impurities (Mx) was calculated according to formula (1) taking RSCx (relative sensitivity coefficient) to be equal to unity. Gallium has been chosen as the internal standard. M x = Mint.st." rex' Ix" Aint.st. • hx" Ix" RSCx/iat.st.

(1)

mint.st." Iint.st. " A x ' hint.st." fint,st.

where Mint.st. is the mass of the internal standard, g; mx, mint.st" are the atomic masses of the impurity and the internal standard; Ix, Iint.st. a r e the line intensities (after subtraction of the background intensity) of the impurity and the internal standard; Ax, Aint.st. a r e the abundances of the analytical isotopes; hx, hint.st, are the corrections for the mass dependence of the line width and the emulsion sensitivity; RSCx/int.st" a r e the relative sensitivity coefficients of the impurity x with respect to the internal standard. The results of the "added-found" experiments are given in Table 2. The choice of the added elements was determined by our desire to study the behaviour of elements that are substantially different in their physico-chemical properties. Thus the ionization potentials of the selected group of elements are in the range from 4.3 to 10.6 eV and the ionization cross-sections from 3.17 to 6.94 x 10 . 6 c m 2. The absence of a systematic error in the calculations of the concentrations of these impurities with a single internal standard (gallium) suggests that the ionization and atomization of the sample proceed with an equal probability and the RSC's of the elements are close to unity. The somewhat overestimated values of the found concentrations of copper and lead when they were introduced at a level of 4.0 x 10 -9 g and 2.0 x 10 -8 g may be explained by a c o n t a m i n a t i o n resulting from the blank experiment. The relative standard deviation for a large group of impurities (A1, Ba, Bi, Ga, Na, P, Pb, Te, T1 and Sb) is in the interval 0 . 0 8 - 0 . 2 4 and only for some of them (As, Cu, Fe and Zn) this interval is 0 . 3 3 - 0 . 5 6 . It appears that this is also due to the variations in these impurity contents in the blank experiment. I n conclusion, let us compare the main analytical characteristics (detection limits, reproducibility and accuracy) of the k n o w n methods of mass-spectroscopic analysis of solutions: the "frozen drop" method, "thin layer" method and our method. In Table 3 these methods are compared with respect to the absolute detection limits. It can be seen that the lowest detection limits (to 10-12 g) and, hence, the greatest analytical information are exhibited by the "thin

18 Table 2. Results of "added-found" experiments (nl = 6, n2 = 17, P = 0.95)

Element

Na A1 p Fe Cu Zn Ga As Sb Te Ba T1 Pb Bi

Solution No. 2

Solution No. ] added, g

found, g

added, g

found, g

5 x 10 -8 5 x 10 .8 5x10 -8 5x10 -8 l x l 0 -7 1.5 x 10 -7 1 x 10-7 5x10 -8 3 X 10 - 7 3x10 7 2.5 x 10 .7 5x10 -7 5 x 10 .7

(5.9 _+ 1.2) x 10 -8 (3.4 ___ 1.4) x 10 .8 (4.1 + 1.5)x10 -8 (3.6 +_ 1.4)x10 - s (3.1 -t- 1.8)x10 -7 (1.6 + 0.6) x 10 .7 internal standard (3.1 q- 1.5)×10 .8 (2.1 q- 1.0) X 10 . 7 (2.1 + 0.4)x10 -7 (2.5 x 0.2) x 10 -7 (3.9 + 1.0)x 10 -7 (5.7 _+ 1.6) x 10 .7 (5.0 _+ 0.7)x l0 -7

2 × ]0 .9 2x10 -9 4x10 -9 6 x 10 .9 4 x 10-9 2 x 1 0 -9 1.2 x 10 .8 1.2x10 -8 1 x 10 .8 2x10 -8 2 x 10 .8 2x10 -8

(1.4 q- 0.8) × 10 .9 (3.2 _ 0.9)x 10 -9 (6.3 q- 1.2)x10 -9 (7.0 + 1.4) x 10 .9 internal standard (3.5 "1- 1.5) X10 - 9 (1.4 + 0.2) x 10 .8 (9.4 + 1.2)x10 -9 (1.1 + 0.1) x 10 .8 (1.3 _+ 0.3)x t0 -8 (2.8 + 0.4) x 10 -8 (1.8 _+ 0.3) x 10 -8

5×10 -7

Table 3. Comparison of the analytical possibilities of different methods for the analysis of solutions on the spark source mass-spectrometer

Mmln, g

"Thin-layer" method

Our method

"Frozen drop" method

n x 10 .7

Au, Cr, Ge, Ta, V

Bi, Cd, Hf, Hg, I, In, Ir, K, Mg, Mn, Mo, Os, Pb, Pd, Re, Ru, S, Si, Sn, Ta, Te, T1, W

n x l 0 -8

Ca, Fe, I, K, Na, Nb, Pb, Sc, Si, Sr

Ag, A1, As, Ba, Ca, Co, Cr, Cs, Cu, Fe, Ga, Na, Nb, Ni, Rb, Rh, Sb, Sc, Se, Sr, Ti, V, Y, Zn B, Be, P

n x 1 0 -9

Au, Bi, Nb

Ag, Br, Cu, Mg, Mo, Ni, Ru, Se, Ti, Zn

n x l 0 -1°

A1, Ba, Ca, Cd, Fe, K, Mg, Mo, Na, P, Pb, Pt, S, Sc, Ti, T1, W

A1, As, Ba, Bi, Cs, Cd, Ga, In, Pd, Pt, Rb, Sb, Sn, Te, T1, W, Y, Zr

nxl0 -H

Ag, As, Be, Co, Cr, Cs, Ga, Hf, In, Mn, Ni, Rb, Re, Rh, Ru, Sb, Se, Sr, Sn, Te, V, Zn, Zr

Co, P, Rh

layer" method. This is explained by the possibility o f complete utilization of the concentrate of the sample and the performance with large exposures ( ~ 100 nC) without danger o f overlapping o f analytical lines o f impurities with lines o f hydrates, oxides and other multiatomic ions, formed in the r.f. ionization o f the solutions. The p r o p o s e d m e t h o d is somewhat inferior to the "thin layer" m e t h o d in relation to the detection limits, but substantially superior (by 1 to 2 orders of magnitude) to the "frozen d r o p " m e t h o d as a result of more complete and effective utilization o f the mass o f the analytical sample. The reproducibility level is approximately the same for all methods: 0.1 - 0 . 4 7 for the "thin layer" method, 0 . 2 - 0 . 6 for the "frozen d r o p " m e t h o d and 0.1 - 0.56 for our method. As to the accuracy, its value depends on the choice o f the internal standard and the way o f its introduction. F o r example, the use o f isotopic dilution allows a substantial increase in the accuracy o f determination for some impurity elements. It should be noted, however, that with the use o f a single internal s t a n d a r d the p r o p o s e d m e t h o d ensures a narrower interval of the RSC values of 0.8 to 1.5 as c o m p a r e d with 0 . 5 - 2.4 o f the "thin layer" m e t h o d and 0 . 5 - 1 0 o f the "frozen d r o p " method.

The main advantage of the p r o p o s e d m e t h o d is its simplicity and suitability for r a p i d analyses. The time required for measuring one sample does not exceed I h which allows easy analyzation o f a large n u m b e r of samples with detection limits o f up to 10-11 g in one run using specially prepared reference samples to increase the accuracy o f the results. Besides, the m e t h o d is attractive since it opens the technical scope to introduce the solutions directly into the ion source of the mass spectrometer, and in future it seems possible to develop a m e t h o d competitive with the inductively coupled plasma mass spectrometry (ICP-MS). Conclusions

A m e t h o d for analysis o f inorganic solutions and microimpurity concentrates is p r o p o s e d which is based on the ionization o f 20 to 32% hydrogen chloride solutions chilled to - 70 ° C under the action of an r.f. voltage. A device has been designed which allows realization of this m e t h o d on the spark source mass spectrometer. The qualitative composition o f the mass spectra o f hydrogen chloride solutions has been studied and the main peaks o f the complex ions limiting the determination o f impurity elements have

19 been established. The detection limits of 47 elements are in the range from 10 -v to 10 -11 g. The accuracy of the determinations was established by the "added-found" method. The reproducibility is 0 . 1 - 0 . 4 at content levels of 10 - 7 _ 10 - 9 g. The analytical possibilities of different massspectroscopic methods for the analysis of solutions have been compared.

Acknowledgement. The authors are grateful to Dr. I. R. Shelpakova and Dr. G. I. Ramendik for discussions and critical review of this paper. We also would like to thank Professor Dr. G. T61g for help with the publication.

References 1. Shelpakova IR, Saprykin AI, Chanysheva TA, Yudelevich IG (1983) J Anal Chem USSR 38:439 2. Vandelannoote R, Blommaert W, Gijbels R, Van Grieken R (1981) Fresenius Z Anal Chem 309:291 3. Vanderforgh BM, Van Grieken RE (1980) Talanta 27:417 4. Moody JR, Weary ES (1982) Talanta 29:1003 5. Chupakhin MS, Kryuchkova OI (1970) Dokl Akad Nauk USSR 193 : 363 6. Gabovich MD (1983) Usp Phys Nauk USSR 140:137 7. Sikharulidze GG, Burmij GP (1987) Vysokochistye veshestva USSR N2, 171