Anal. Chem. 1982, 5 4 ,
curve had good linearity and could be used for quantitative analysis. In the Case of an Unknown a. Ideally, the measuring conditions of hole 1 and hole 2 of multimode cavity should be equal. However, hole 1and hole 2 were slightly different from each other in practice. Hence the factors a and 0 had to be determined. Although a and 0 are proper values for the cavity used, a is considered to vary with change in the shape and the shift of the position of the sample tube adaptors, even if the cavity is the same!. Accordingly, the value of a will be changed by exchanging or resetting the adaptors. The computation method in the case of an unknown a is as follows: Extract the square root of the product of eq 5 and eq 6 on each side of the equation, obtaining eq 81. As this equation
:=d-
(ha + ha/P)(h,' .- h,'/P) (ha - ha/P)(h,' + h,'/P)
(9)
does not include the factor a , the intensity ratio can be calculated without knowing the value of a,although it is necessary to do the measurements twice. That is, the signal intensity is the geometrical mean of the ratios obtained from eq 5 and 6 when the value of a is unknown.
CONCLUSION The two-step integral method using a multimode cavity has been found to be a good method for quantitative measurements of ESR signal. Ai intensity of an unknown sample can be decided as some-fold of a known amount of suitable standard substance, regardless of the measuring conditions and Q value of the sample. This makes it possible to compare signal intensities measured under different conditions at different times.
1853-1858
1853
For precise measurement, it is desirable that volumes of sample be not so large and amounts of standard and sample be not so different from each other. In such cases, measurements can be carried out within only a few percent error. Standards including various values of spin number are required for quantitative analysis over a wide range of sample amount. If attention is paid to the above, the present method will be useful for quantitative analysis using ESR. LITERATURE CITED (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19)
Guilbault, G. G.; Meisel, T. Anal. Chem. 1969, 4 7 , 1100. Moyer, E. F.; McCarthy, W. J. Anal. Chim. Acta 1969, 4 8 , 79. Gullbault, G. G.; Moyer, E. S. Anal. Chem. 1970, 4 2 , 441. Meisel, T.; Guilbault, G. G. Anal. Chim. Acta 1970, 5 0 , 143. Fujiwara, S.; Tadano, H.; NakaJima,M. Bull. Chem. SOC.Jpn. 1970, 4 3 , 3023. Yamamoto, D.; Fukumoto, T.; Ikawa, N. Bull. Chem. SOC.Jpn. 1972, 45, 1403. Yamamoto, D.; Ikawa, N. Bull. Chem. SOC.Jpn. 1972, 45, 1405. Yamamoto, D.; Ozaki, F. Bull. Chem. SOC. Jpn. 1972, 4 5 , 1408. Bryson, W. G.; Hubbard, D. P.; Peake, B. M.; Simpson, J. Anal. Chim. Acta 1975, 77, 107. Warren, D. C.; Fltzgerald, J. M. Anal. Chem. 1977, 49, 250. Bryson, W. G.; Hubbard, D. P.; Peake, B. M.; Slmpson, J. Anal. Chim. Acta 1978, 96, 99. Bryson, W. G.; Hubbard, D. P.; Peake, B. M.; Simpson, J. Anal. Chim. Acta 1980, 116, 353. Sakane, Y.; Salto, K.; Matsumoto, K.; Osajima, Y. Bunsekl Kagaku 1981. ..- ., 30 .., 3023. - - --. Nakano, K.; Tadano, H.; Oshima, M. Nippon Kagaku Kaishi 1972, 2453. Nakano, K.; Tadano, H.; Suglmoto, S. BunsekiKagaku 1978, 27,256. Goldberg, Ira B.; Crown, H. R.; Robertson, W. M. Anal. Chem. 1977, 4 9 , 962. Goldberg, Ira B.; Crown, H. R. Anal. Chem. 1977, 4 9 , 1353. Goldberg, Ira B. J . Magn. Reson. 1978, 32,233. Nakano, K.; Tadano, H.; Sugimoto, S. Nlppon Kagaku Kaishi 1979, 885.
RECEIVED for review January 18,1982. Accepted June 1,1982.
Quantitative Room-Temperature Phosphorescence with Internal Standard and Standard Addition Techniques Malcolm W. Warren 111,' James P. Avery,* and Howard V. M a l m ~ t a d t " ~ University of Illinois, School of Chemical Sciences, 1209 W. California St., Urbana, Illinois 6 180 1
Greatly improved preclslon and accuracy are demonstrated by the methods of internal standard and standard addltlon wlth room-temperature phorrphorescence of adsorbed organic molecules. An Internal standard Is used to increase the preclsion of the measurement by a factor of I O , typlcally to 1-3 % from 10 to 20%. A two-point standard addition technlque greatly Improves the accuracy. For example, in a sample containing sodlum acetate the errors due to shifts In spectral distribution and intenslty are decreased from greater than 100% to less than 8%. Shifts in spectral intenslty and dlstributlon are accounied for wlth spectra calculated with factor analysis assumlng no previous knowledge of the spectral dlstrlbution or intensity. This allows measurement of anaiytes in "real" sample matrlces without the use of compllcated standards or a prevlous knowledge of ?he spectral distrlbutlon or Intenslty.
'The Dow Chemical Co., Analytical Laboratories, Building 1603, Midland, MI 48640. University of Colorado, Department of Electrical Engineering, Campus Box 425, Boulder, CO 80309. 3Pacific and Asia Christian University, P.O.Box YWAM, Kailua-Kona, HI 96740.
Room-temperature phosphorescence provides selective and sensitive detection of many organic compounds (1-3). Unfortunately, quantitative measurements have seldom been reported because of poor precision and accuracy. Quantitative measurements reported in the literature are for components in simple, well-defined sample matrices or use complicated sample preparation or preseparation methods. There are no reports of analyses made in sample matrices that change the spectral characteristics of the phosphore in an unknown manner. Also, there has not been a thorough evaluation of factors that affect accuracy in determining a specific analyte in complex matrices. The combination of internal standard and standard addition improves the precision for manual preparation of the sample on paper by a factor of 10, from about 10-20% to 1-3%, and provides accuracy to better than 8% ( 4 ) . The use of an internal standard decreases errors caused by nonreproducible sample handling and measurement parameters. Standard addition allows the measurement of components in a sample matrix that affects the excitationemission intensity and spectral distribution of the phosphorescence. Early papers on room-temperature phosphorescence indicated that it is free from the problems associated with
0003-2700/82/0354-1853$01.2!5/00 1982 American Chemical Society
1854
ANALYTICAL CHEMISTRY, VOL. 54, NO. 11, SEPTEMBER 1982
2 . 6
I
Y)
MASTER
MICROPROCESSOR
I
0)
E
5
.I 5
4
I
3
E
2
.s
Figure 2. Phosphorescence experimental measurement system.
a J l 0 450
500
550
600
650
Wavelength (nml
Figure 1. Emission spectra of 5 mM 1-hydroxy-2-naphthoicacid in (EM = emission maximum): 1 M LiOH (EM = 520 nm); 1 M NaOH (EM = 530 nm); 1 M KOH (EM = 535 nm); 1 M NaOH and 1 M NaAc (EM = 525 nm); 1 M NaOH and 1 M NaI (EM = 530 nm).
phosphorescence a t liquid nitrogen temperatures while retaining the fundamental advantages of phosphorescence. In addition, the use of 3-10 p L sample volumes was demonstrated. Typical precision for analyte weights less than 1 pg were about 10-20% for samples spotted on paper manually ( 4 , 5 ) . An automated paper method was demonstrated to be capable of about 5% precision (5). Precisions of 3% were reported for a manual sodium acetate procedure (6). The newly discovered micelle stabilized room-temperature phosphorescence provides precisions of 6-10% (7,8). The relatively poor precision has, in general, made room-temperature phosphorescence unattractive as an analytical method. The chemical matrix of the sample can change the phosphorescence excitation-emission distribution, intensity, and lifetimes (9,10). Heavy atoms and differences in the matrix salt composition can cause shifts in intensity and spectral distribution as shown in Figure 1. Therefore making accurate determinations of concentration is normally difficult without a knowledge of the composition of the sample matrix. Internal Standard and Standard Addition in Phosphorescence. Only a few attempts have been reported using an internal standard in phosphorescence (4, 11, 12). The authors were the first to report the use of standard addition or a combination of internal standard and standard addition with room-temperature phosphorescence ( 4 ) . In 1961, Freed and Vise used an internal standard to decrease the relative standard deviation of phosphorescence measured a t liquid nitrogen temperatures (12). Percent relative standard deviations without a monitor were about 15% vs. 2-8% with the monitor. Hollifield and Winefordner later demonstrated that the use of a rotating sample cell improved precisions from 10 to 20% to 1-470 without the use of an internal standard (13). In 1979 Goering and Pardue used the method of internal standard with room-temperature phosphorescence (11). Without an internal standard a 2.5-pg sample varied in intensity between samples from 6 to 19%. With an internal standard, the precision improved to 2-4% for samples containing 100-1000 ng. Data presented indicate that the linear range for the compounds extends to less than 500 ng. The high cost of this vidicon array instrument limits its application.
EXPERIMENTAL SECTION Apparatus. The experimental measurement system consists of a 200-W Xe-Hg source, excitation and emission monochromators (14),a reference beam splitter and 1P28 photomultiplier tube, a sample module, and an emission 1P28 photomultiplier tube as shown in Figure 2. The system is controlled by a master microprocessor running programs written in BASIC and assembly language. The excitation and emission monochromators are controlled by separate microprocessors, which communicate over
Figure 3. Sample holder for paper support.
serial lines with the master microprocessor (14). Input and output devices include a teletypewriter (TTY) for communication between the operator and the master microprocessor, a thermal plotter/printer (TPP), and a video monitor (TV) for display of experimental data and calculated results. The sample module is flushed with dry nitrogen and contains a sample holder. The sample is attached to the sample holder as shown in Figure 3. Reagents. Salicyclic acid (Analytical, Mallinckrodt), 1hydroxy-2-naphthoic acid, 2-hydroxy-1-naphthoicacid (Technical, Aldrich), 3-hydroxy-2-naphthoic acid, 1-naphthoic acid, 2naphthoic acid (Analytical, Eastman Organic Chemicals), p aminobenzoic acid (Reagent, Sigma), sodium hydroxide, and sodium acetate (Analytical, J. T. Baker Chemical Co.) are all used as received. Procedure. All samples used for quantitative determinations are dissolved in 1M sodium hydroxide. Five-microliter aliquots of sample are spotted on Whatman Number 1 filter paper and dried under a heat lamp and dry nitrogen. The filter paper is then attached to the sample holder, placed in the sample module and rotated with a small dc motor at several thousand revolutions per minute. After the sample is placed in the sample chamber, the microprocessor automatically collects and analyzes the data. The sample spot covers only a portion of the sample holder. This leads to a modulated signal, so the phosphorescence emission is integrated for several seconds t o eliminate the effect of this modulation. The intensities of the analyte and the internal standard are measured. The spectral distribution of the analyte and internal standard must be calculated from the excitationemission matrix of (a) the sample [SI, (b) the sample plus the internal standard [S IS], and (c) the sample plus the internal standard and added standard analyte [S + IS + SA]. The phosphorescence intensities for the analyte and internal standard are spectrally stripped from the total phosphorescence emission spectrum using factor analysis (15). The intensity of the analyte is then divided by the intensity of the internal standard which provides the normalized analyte intensity. For standard addition the normalized analyte intensity is plotted VI. the weight of standard analyte added and the analyte concentration computed from the intercept on the weight axis. Calculation of Excitation-Emission Distribution. The phosphorescence excitation-emission distribution of the analyte and internal standard are calculated by factor analysis (15). The use of factor analysis to spectrally strip the intensities of spectra using standard spectra was developed by Warner, Christian, Davidson, and Callis (16). It is based on rank analysis as developed by Weber (17). Rank analysis allows the determination of the number of components in the excitation-emission matrix. Rank analysis as described by Weber requires that the absorption spectra of the components be statistically different from each other and that the intensities be greater than the experimental error (17). Factor analysis can be used to determine the number of
+
ANALYTICAL CHEMISTRY, VOL. 54, NO. 11, SEPTEMBER 1982
components required to span the excitation-emission data spectrum (15). The excitation-emission matrix or data matrix [D] is organized with the excitation wavelengths in columns and emission wavelengths in rows. The matrix elements are proportional to the photocurrents at the corresponding excitation-emission wavelength pair. Wavelengths iue selected to cover both the analyte and internal standard with several points. The matrix has c columns and r rows where c is less than or equal to r. When r is greater than c, the calculations provide the same results but require larger covariance matrices. The data analysis time increases with increasing covariance matrix size. Normally c is 6 and r is 9 for the analyses described in this paper. Principle factor analysis is composed of four steps: (a) calculation of the covariance matrix [Z], (b) calculation of the eigenvectors C, and eigenvalucas A,, (c) calculation of residue matrices [R,],and (d) determination of the number of primary factors n. The covariance matrix [Z] in calculated by prernultiplying the data matrix [D] by its transpose [D]’. [Z] = [D]’[D] The eigenvectors that span the data space also span the covariance matrix. Eigenvectors are calculated by using an iterative procedure until C, is self-consistent according to eq 2. As a [ZIC, = AIC, (2) starting approximation the elements of the eigenvector C, on the left side of the equality sign are set equal to one divided by the square root of the number of columns in the covariance matrix. The eigenvalue is equal to the square root of the sum of the squares of the elements of the vector produced by the multiplication of the covariance matrix and the vector on the left side of eq 2. The normalized eigenvector on thle right side of the equation is calculated by dividing the eigenvalue into each element produced by this multiplic,iltion. The normalized eigenvector is substituted for C, on the left side of the equality sign. The process is then repeated until C, becomes self-consistent. Self-consistency is taken to occur when each C, on the right side of eq 2 differs from the C, on the left side at the end of a iteration by less than 0.01%.
The residual matrix is calculated by subtracting the eigenvalue times the eigenvector timets the eigenvector transpose as in eq 3. The resulting residual matrix is substituted into eq 2 in place [R,] [Z] - A,C,C,’ (3) of the covariance matrix. The next eigenvalue and eigenvector are calculated by the same process used to calculate the previous eigenvalues and eigenvectors. This process is repeated, substituting [R,] for [Z] in eq 2 and 3 until c eigenvectors and c eigenvalues have been extracted. The number of primary factors is determined by comparing the real error (RE),calculated with eq 4, to the root-mean-square -1
error (square root of the mean of the variance of the data points) for the experimental data. ‘When the RE is approximately equal to the root-mean-square error, all of the primary factors have been accounted for. The RE is calculated by assuming different numbers of primary factors. The calculation is started by assuming that only one primary factor exists (Le., n = 1). RE is calculated and compared to the root-mean-square error. If RE is greater than the root-mean-square error, not all of the primary factors have been accounteld for. The number of prime factors, n, is incremented and the procedure is repeated until all of the intensity is accounted for by experimental error. The excitation-emission matrix of the sample is measured with and without the internal aitandard. The excitation-emission matrix without the internal. standard is then multiplied by a scaling factor, FIs [IS] [Q + IS] - FIs[S] (5) The scaling factor is varied )untilthe rank of the sample with the
1855
Table-I. Possible Sources of Error for Room-Temperature Phosphorescence on Filter Paper 1 standard/sample preparation (a) pipetting of standard or sample (b) pipetting of reagents (c) dilution transfer of sample to substrate 2 (a) pipetting volume accuracy ( 5 p L ) (b) positional accuracy of sample spotting (c) the chromatography effect ( d ) directional wicking of the sample in paper 3 drying of sample (a) primary drying step (b) secondary drying step 4 placement of sample in measurement system 5 measurement system noise internal standard [S + IS] minus the sample without the internal standard [SI times FIsis one indicating that all of the intensity except that due to the internal standard has been subtracted. In practice, this is accomplished by varying FISuntil the second eigenvalue is at a minimum. The resultant matrix is the excitation-emission matrix for the internal standard [IS]. The scaling factor FIS normalizes the spectral intensity of the components common to [S + IS] and [SI. This corrects for nonreproducible sample preparation and handling. The excitation-emission distribution matrix of the added standard analyte, [SA], is tal.. culated by subtracting the sample plus the internal standard [S + IS] times Fsafrom the sample plus the internal standard and added standard analyte [S + IS + SA].
[SA] = [S + IS
+ SA] - F S A [ S + IS]
(6)
The calculated excitation-emission matrices [SA] and [IS] are then used to spectrally strip the excitation-emission matrices of the analyte and internal standard from the total excitationemission matrices of samples containing different amounts of the standard analyte and constant amounts of internal standard. The spectral intensities of the internal standard and the analyte are spectrally stripped by subtracting the calculated excitation-. emission matrices from the excitation-emission matrices by varying the factor F, in each of the following equations until the number of prime factors decreases by one for each of the data matrices produced by [Dl] = [S + IS] - Fl[IS]
(7)
[Dz] = [S + IS] - Fz[SA]
(8)
[Ds] = [S + IS
+ SA] - Fs[IS] [Dd] = [S + IS + SA] - F4[SA]
(9) (10)
Fl,F2,F3, and F4are the intensity due to [IS] in [S + IS], [SA] in [S + IS], [IS] in [S + IS + SA], and [SA] in [S + IS + SA], respectively. F2/Flgives the normalized intensity of the analyte and F4/F3gives the normalized intensity of the analyte plus the added standard. These factors are then used to calculate the standard addition curve for the sample matrix. Mathematically, this is condensed into eq 11. CS is the concentration of the added CA = CS(Fz/Fi)/(F4/F3 - F2/F1)
(11)
standard and CA is the concentration of the analyte in the sample. Calculation of the concentration of the analyte takes about 1min after data collection. It may be possible to apply this method to experimental data obtained with high intensity pulsed sources (i-e.,nitrogen lasers) by using emission-decay matrices.
RESULTS AND DISCUSSION Error Budget. Considering the sample preparation and measurement procedure for samples spotted on filter paper, it is possible to break down each step into several sources of error as in Table I. Only in a few cases is it possible to assign an error magnitude. If care is taken in the preparation of sample and standard solutions, the error resulting from this step will be less than
1856
ANALYTICAL CHEMISTRY, VOL. 54, NO. 11, SEPTEMBER 1982
0.270. The transfer of the sample to the paper substrate requires that an accurate volume of the sample be spotted in a consistent location on the paper. Because in most measurement systems excitation and emission spectra of the entire sample spot are measured, differences in sample volume and location, due to differences in sensitivity across the measurement axis, can be a problem. In addition, preferential wicking occurs along the grain of the filter paper. As the sample spreads out on the paper the distribution of the analyte is not even. This effect is first noted by Winefordner and could be a chromatography-type effect (5). In some classes of compounds the spot is more strongly phosphorescent in the center and in others a t the edge. Drying the sample spot requires two steps (1-4). In the primary drying step, the sample is placed under dry nitrogen and a heat lamp. In this step most of the water is driven out of the sample. If the sample is dried too long under the heat lamp, the paper substrate appears to break down, lowering the measured phosphorescence intensity. If the sample is not dried long enough under the heat lamp, the phosphorescence never reaches the maximum possible intensity. In the secondary drying step the sample is dried under dry nitrogen alone which removes the remaining water in the sample. Phosphorescence emission intensity during the secondary drying step reaches a steady state after 5 min of drying. The intensity remains nearly constant for over 20 min during measurement. Measurement system drift can cause errors of several percent. If sample drying can be accurately controlled, the remaining causes of poor precision are mechanical in nature except for the chromatography effect and wicking. This suggests the usefulness of an internal standard to decrease the large relative standard deviations encountered in room-temperature phosphorescence. If the internal standard is chemically similar to the analyte, the differences caused by the chromatography effect and wicking should be small. Selection of an Internal Standard. The selection of an internal standard for spectroscopy requires chemical equivalence and spectral separation of the analyte and standard. Chemical equivalence requires the use of compounds that are in the same chemical class. In addition, it can be easily argued that chemical equivalence requires that the positions of the substituent groups have an equivalent relationship with each other. If, as has been postulated, room-temperature phosphorescence is a surface adsorption process (1-3), parallels can be made with adsorption chromatography. The energy of interaction of the surface active groups is dependent on both the type of group and the degree of overlap of the group with the surface sites ( I S ) . The degree of overlap of the surface sites and the active sites on the molecule depend on (a) the geometry of the active sites on the molecule and the surface, (b) surface contour, and (c) the chemical nature of groups involved (18). The aromatic ring system of the molecule also interacts strongly with the surface active sites. Thus, differences in aromatic ring number will change the degree of overlap and energy of interaction of the internal standard with respect to the analyte. For aromatic molecules to show spectral separation generally requires the use of an aromatic system that is smaller or larger than the aromatic system of the analyte. The use of a smaller or larger ring system for spectral separation creates a chemical difference that results in different energies of interaction for the analyte and the internal standard. One possible solution to this problem might be to add substituent groups that have low energies of interaction but change the phosphorescence excitation-emission spectrum of the internal standard from that of the analyte. For example, the addition of a methyl group (either 3 or 5 ) to salicylic acid
P ; 2
::
r &
I
0
350
400
450
500
550
600
650
Wavelength n m
Figure 4. Phosphorescence emission spectra for salicylic acid and
its internal standard candidates. changes the excitation-emission spectrum slightly. Unfortunately, the spectral shift is small and becomes statistically insignificant a t low concentrations of the internal standard and analyte. Accuracy. The errors due to mechanical operations, measurement system drift, etc. lead to imprecision in replicate analyses. Although an internal standard can solve these types of problems, it is unable to eliminate errors caused by changes in the sample matrix. If the chemical nature of the sample is different from the standards, the relative intensities of the analyte, the internal standard, and the background can change with respect to each other. Thus the ratio of the analyte intensity to the internal standard intensity changes. These changes in different matrices can vary as much as several hundred percent. Typical chemical species that cause changes in spectral intensity and/or distribution include sucrose, sodium acetate, silver nitrate, and sodium iodide. All of these compounds tend to increase the measured phosphorescence intensity (12,15). Sucrose and sodium acetate appear to increase the stability of the sample matrix increasing the phosphorescence lifetimes (12). Silver nitrate and sodium iodide change the phosphorescence by a heavy atom effect (13). The heavy atom effect tends to change the spectral distribution of the phosphorescence excitation-emission wavelengths, increases the quantum yield of intersystem crossing, increases the quantum yield for phosphorescence, and decreases the phosphorescence lifetime. These phenomena indicate that quantitative phosphorescence is difficult without the use of standard addition in an unknown sample matrix. Samples. In this work several molecular and matrix systems are used to demonstrate the principles of internal standard and standard addition with room-temperature phosphorescence. These systems include salicylic acid and p-aminobenzoic acid in 1 M sodium hydroxide with and without 1 M sodium acetate added and salicylic acid in blood serum. Salicylic Acid. The selection of an internal standard for salicylic acid requires a ring system larger than benzene. The hydroxy carboxylic acid derivatives of naphthalene or anthracene are possibilities. Chemical equivalence requires that the hydroxy and carboxylic acid groups are in adjacent positions. A search of commercially available compounds indicates three possible internal standards for salicylic acid: 1-hydroxy-2-naphthoic acid, 2-hydroxy-1-naphthoic acid, and 3-hydroxy-2-naphthoic acid. The phosphorescent excitation and emission spectra are obtained for salicylic acid and the three possible internal standards are shown in Figure 4. Salicylic acid is spectrally separate from the three possible internal standards. 1-Hydroxy-2-naphthoic acid is chosen because it had the strongest phosphorescence.
ANALYTICAL CHEMISTRY, VOL. 54, NO. 11, SEPTEMBER 1982
Table 11. Results for Several Sample Determinations analyte salicylic acid 1 M NaOH sample 1 M NaAc matrix 1-hydroxy-2-naphthoicacid internal standard wt of internal standard added 50 ng actual analyte w t 1 7 2 ng wt of added standard1 172 ng measd analyte w t 165 ng 4
re1 error, %
p-aminobenzoic acid 1 M NaOH 1 M NaAc 1-naphthoic acid 50 ng 171 ng
171 ng 180 ng 5
9 t
z
8
0
200
salicylic acid in serum serum and 1 M NaOH 1-hydroxy-2-naphthoicacid 50ng 50ng 50ng 50ng
150ng 54ng 8
100ng 150ng 97 n g 3
150ng
150ng 158ng 5
A
iI
,/
3 k
1857
400
600
800
Sample \Neigh1 Spotted On Paper ( ng 1 ~-
Figure 5. Salicylic acid working curves. Wavelength nrn
The calibration curves for salicylic acid shown in Figure 5 indicate a linear region at low concentrations. The curves become nonlinear a t high concentrations. The precision of the measurements, exprlessed as percent relative standard deviations, ranges from 6 to 23% without an internal standard and 1 to 3% with the internal standard. Addition of 1M sodium acetate to the samples causes a shift in the ratio of the spectral intensity of thle salicylic acid to the internal standard of about 150%. The relative standard deviations of samples containing sodium acetate remain at about 2%. Therefore the internal standard improves precision but does not correct for chemical differences in the matrix. Thus, a standard addition technique is required to correct for errors in accuracy caused by the matrix. Standard addition for salicylic acid in 1 M sodium hydroxide and 1 M sodium acetate determined the sample weight of a 172-ng sample to be 165 ng. This is an error of 4%. p -Aminobenzoic Acid. The logical internal standard for p-aminobenzoic acid is 1-amino-4-naphthoic acid. Unfortunately, this compound is not commercially available. Since the synthesis of this compound would probably require the use of 1-naphthalamine ,and the naphthallamines are carcinogens, it is felt that syntlhesis would be unwise. Because the internal standard appears to correct for mechanical but not chemical nonreproducibility as demonstrated by salicylic acid, exact chemical equivalence may not be necessary. Therefore the internal standard must be selected from a related class of compounds. 1-Naphthoic acid and 2-naphthoic acid were tested as possible internal standards for p-aminobenzoic acid. 1-Naphthoic acid was seliected on the basis of emission intensity as shown in Figure 6. Calibration curves are obtained that are linear up to 700 ng for p-aminobenzoic acid as shown in Figure 7. Relative standard deviations without an internal standard ranged between 5 and 16% and with an internal standard from 1 to 3%. When 1M sodium acetate is added to the sample, the ratio of intensity of the p-aminobenzoic acid to the internal standlard increased by 100% indicating that standard addition is rlequired. With a two-point standard addition method the mea,aured amount of the analyte is 180 ng while the actual weight is 171 ng. This i s an error of 5%. Salicylic Acid in Blood Serum. Several problems can limit the accuracy and precision of room-temperature phos-
Figure 6. Phosphorescence emission spectra for p-aminobenzoic acid and its internal standard candidates.
c m
‘C w
3-
ly
c
EP
a
6 a
Sample Weigh1 Spotted On Paper ( ng )
Figure 7. p-Aminobenzoic acid working curves. phorescence measurements of drugs in blood serum. These problems include: (a) distortion of the excitation-emission distribution and intensity because of the complex nature of the matrix, (b) high phosphorescence background from other species in serum in the spectral region of interest, and (c) quenching of the phosphorescence of the analyte. An exact survey of the effects for the many components that make-up blood serum on the analyte would be almost impossible. Blood serum is a complex mixture containing: (a) electrolytes and gases, (b) carbohydrates, (c) proteins, (d) amino acids, (e) porphyrins, (f) lipids, (8) vitamins, etc. (19). Each of these classes of components is composed of many different compounds. Fortunately, a background scan of blood serum in 1 M NaOH indicates that only emission from tryptophane causes problems in determining the amount of analyte. The standard working range for salicylic acid in the clinical laboratory is between 10 mg/100 mL and 50 mg/100 mL (19). This is equivalent to 0.5-2.5 pg in the samples prepared in this study. Because the linear range extends from 0.0 ng to about 400 ng for salicylic acid a dilution by a factor of 10 is possible, which decreases the effect of the serum matrix during the measurement. Additionally, it is possible to measure the concentration of salicylic acid in blood serum at concentrations
1858
Anal. Chem. 1982, 5 4 , 1858-1862
lower than is conventionally possible. Serum samples were spiked with normal therapeutical levels of salicylic acid and then diluted by a factor of 10. Sodium hydroxide was added to make the sampies 1 M in sodium hydroxide. Sample weights of 50, 100, and 150 ng of salicylic acid are determined to contain 54, 97, and 158 ng for errors of 8, 3, and 5%, respectively. The results for this paper including sample compositions have been tabulated in Table 11. They indicate that this method is capable of accuracies of better than 8% for several analytes in "real" matrices. It can be concluded that precision and accuracy are greatly improved by using the methods of standard addition and internal standard with room-temperature phosphorescence. Two effects limit this approach to quantitative analysis using room-temperature phosphorescence: the background emission and limits on the ratio of the analyte and internal standard intensities. The background emission appears to have little effect on the calculation of the excitation-emission matrices. The effect of the background on the calculation of the integrated spectral intensities is more important. If the analyte and background have limited overlap, there will be no adverse effect. If the analyte and background have a pronounced overlap the background will add to the integrated spectral intensity of the analyte and its apparent concentration. If the spectral intensities are widely different, the photon noise of the more intense component will tend to bury the weaker component. This will distort the calculated spectrum of the weaker component and affect the calculated spectral intensity. Statistically the division of a large uncertain number
information by a small reaction number leads to a larger error than if the components have equal intensities.
LITERATURE CITED Parker, R. T.; Freedlander, R. S.; Dunlap, R. B. Anal. Chim. Acta 1980, 119, 189. Parker, R. T.; Freedlander, R. S.;Dunlap, R. B. Anal. Chim. Acta 1980, 120, 1. Vo Dihn, T.; Winefordner, J. D. Appl. Spec. Rev. 1977, 73, 261. Warren, M. W.; Avery, J. P.; Malmstadt, H. V. 1980 Meeting, Federatlon of Analytical Chemistry and Spectroscopy Societies, Paper number 311, Oct 2, 1980. Vo-Dlhn, T.; Walden, G. L.; Wlnefordner, J. D. Anal. Chem. 1977, 49, 1126. Von Wandruszka, R. M. A.; Hurtubise, R. J. Anal. Chem. 1978, 48, 1784. Cline-Love, L. J.; Skrilec, M.; Habarta, J. G. Anal. Chem. 1980, 52, 754. Skrilec, M.; Cline-Love, L. J. Anal. Chem. 1980, 50, 1559. Nlday, G. J.; Seybold, P. G. Anal. Chem. 1978, 50, 1577. Vo-Dlnh, T.; Hooyman, J. R. Anal. Chern. 1979, 51, 1915. Goerlnger, D. E.; Pardue, H. L. Anal. Chem. 1979, 51, 1054. Freed, S.; Vlse, M. H. Anal. Biochem. 1983, 5 , 338. Hollifield, H. C.; Wlnefordner, J. D. Anal. Chem. 1988, 40, 1759. Warren, M. W.; Avery, J. P.; Malmstadt, H. V. J. Autom. Chem. 1981, 3, 76. Mallnowsky, E. R.; Howery, D. G. "Factor Analysis in Chemistry"; Wiley: New York, 1980. Warner, I. M.; Christian, 0. D.; Davidson, E. R.; Callis, J. B. Anal. Chem. 1977, 49, 564. Weber, G. Nature (London) 1981, 190, 27. Kiselev, A. V.; Yashin, Y. I. "Gas-Adsorption Chromatography"; Plenum Press: New York, 1969. Tietz, N. W. "Fundamentals of Clinical Chemistry;" W. 8. Saunders: Philadelphla, PA, 1976.
RECEIVED for review December 23, 1981. Resubmitted and accepted June 10,1982. The authors acknowledge the support provided by Grant HEW-PHS GM 21984.
Determination of Zirconium, Molybdenum, Hafnium, and Tungsten in Niobium and Tantalum by X-ray Fluorescence Spectrometry after Separation from Matrix and Preconcentration Harald Knote and Vlllam Krlvan" Sektlon Analytlk und Hochstrelnigung, Unlversitat Ulm, N 26, D 7900 Ulm/Donau, Federal Republic of Germany
A method for the determlnatlon of Zr, Mo, Hf, and W In nloblum and tantalum Is described. This method is based on the removal of the matrlx elements by extractlon with dlantipyrylmethane In dlchiorethane from HF medium followed by the collection of the elements to be determlned on an anion exchange paper suitable for X-ray fluorescence measurement. The recovery studies were carrled out wlth the radiotracer technique. The recoverles for the indlvldual elements wlth regard to the whole procedure are between 83.1 f 2.3% (Mo) and 93.9 f 1.4% (Hf). Limits of detection for all four elements ranged from 0.2 to 0.25 pg/g. This method was applied to the analysis of sampes of different purity grade and the results are compared with those obtalned by activation analysls.
The refractory metals niobium and tantalum have become materials of great scientific and technological significance. A number of the relevant properties depend decisively on the trace impurities (1, 2). Molybdenum and tungsten are the
most significant impurities in niobium and tantalum. Important metallic impurities also include zirconium and hafnium. Although several methods have been developed for the determination of these four trace constituents in niobium and tantalum,there is still an urgent need for sensitive and reliable techniques suited for routine analysis. Optical emission spectrography using a dc arc and involving the conversion of the metals into oxides before the analysis can detect these elements only at the 10 pg/g level ( 3 , 4 ) . By addition of certain substances, e.g., AgCl and CdS in the determination of tungsten (5),or by separations (6),the limits of detection can be improved by about 1order of magnitude. Atomic absorption spectrometry is not a suitable technique for the determination of these elements mainly because of low sensitivity and carbide formation if a graphite furnace is used. I t was therefore applied as the flame technique only to the determination of molybdenum providing a limit of detection of about 60 pg/g (7). Spectrophotometry combined with separations and/or preconcentrations seems to be the most widely used technique for the determination of Zr, Mo, Hf, and W in niobium and tantalum (8-12). Also in these tech-
0003-2700/82/0354-1858$01.25/00 1982 American Chemical Society
