Direct Determination of Fluoride in Sea Water with a Fluoride Selective Ion Electrode by a Method of Standard Additions C. J. Rix and A. M. Bond* Department of Inorganic Chemistry, University of Melbourne, Parkville 3052, Victoria, Australia
J. D. Smith Marine Chemistry Laboratory, School of Chemistry, University of Melbourne, Parkville 3052, Victoria, Australia
In natural water systems, variable matrices can provide considerable problems In analytical methods based on calibration procedures with standard solutions. For the determination of fluoride in sea water by a selective Ion electrode, a modified method of standard additions Is shown to provide a direct and simple means for determining fluoride without the need for addition of buffers and/or complex Calibration procedures. Data obtained under a variety of conditions are presented to validate the method uslng both synthetlc and natural sea water. The presence of fluoride complexing metal ions such as aluminum(lll) and iron(l1l) was shown not to Interfere In the determination.
The determination of fluoride in sea water has been a subject of considerable interest to Marine chemists for many years (1-6). Currently two major approaches to the problem are employed: (i) the spectrophotometric method of Greenhalgh and Riley ( 2 )which is based upon the color of the lanthanum alizarin complexone, and (ii) the potentiometric methods described by Brewer et al. (4), Warner ( 3 ) ,and Windom (6) in which measurements of the fluoride concentration (or activity) are made with a fluoride selective ion electrode, after treatment of the sample with an acetate or metal complexing buffer (7). Published potentiometric methods for determining fluoride in sea water appear to be unnecessarily complicated in several respects. Reagents, such as a total ionic strength adjustment buffer, TISAB, are added to the sample at high concentrations t o set the p H and ionic strength and to liberate the majority of fluoride bound in metal complexes. The possibility of introducing solution contamination when using high concentrations of TISAB is always a potential hazard and, if this step can be avoided, this will obviously be advantageous. Furthermore, a significant drawback with most current methods is the difficulty of establishing an acceptable or representative matrix in which to prepare standards, as the actual matrix is often unknown in natural waters. This aspect of fluoride determination in sea water has received considerable attention, and extensive and elaborate procedures have been used to match the sample and standard matrices (4). The method of standard additions is frequently employed in environmental analysis for samples of variable matrix. For example, the method of anodic stripping voltammetry (8) is readily implemented in this mode where a parameter (peak height) is directly proportional to concentration. The difficulty of using the method of standard additions in potentiometry arises from the logarithmic rather than linear relationship of concentration (activity) to potential. Thus, this is one of the techniques not having an absolute reference point for zero concentration. Nevertheless, we have exploited a modified form of standard additions to directly determine the concentration of fluoride a t the environmental pH, low pH, and in the presence of metal ions and complexing agents. Our method of data treatment is similar to that originally de1236
ANALYTICAL CHEMISTRY, VOL. 48, NO. 8, JULY 1976
scribed by Gran (9) for the exact determination of potentiometric end points. Rechnitz (IO), Baumann ( 1 1 ) and Craggs e t al. (12) have already described procedures employing the method of standard additions in selective ion electrode potentiometry, but no data are available for natural waters using this direct method except for the work of Liberti and Mascini (13) who determined fluoride in mineral waters. Our procedure is exceedingly simple and the method of data treatment provides a linear plot, the slope and intercept of which give independent measures of the original concentration of fluoride in the sample. Total fluoride concentration is determined, despite the fact that the activity of the fluoride ion is the parameter actually monitored by the fluoride selective ion electrode.
EXPERIMENTAL Reagents. All chemicals used were of reagent grade purity. Standard fluoride solutions were prepared from weighed sodium fluoride dried in an oven at 110 O C for 4 h. Synthetic sea water, containing no added fluoride,was prepared as described by Riley and Chester ( 1 4 ) . TISAB was prepared as described in the Orion Handbook (7),except that NazEDTA ( 2 gb.) replaced CDTA, and an additional 7 g/l. of ammonium citrate was added. Artificial levels of iron and aluminum were obtained by addition of ferric ammonium sulfate or potassium aluminum sulfate, respectively. Sea water samples having a salinity of 35.21%0were collected from Point Lonsdale, Victoria, Australia, filtered through a 0.45qtMillipore filter, and stored in polyethylene containers at ambient temperature. Analyses were performed on the samples within 3 weeks of collection. Unless otherwise indicated, natural pH was used to store and determine the fluoride concentrations. When required, acidification was made with acetic or hydrochloric acid. Instrumentation. An Orion model 94-09A fluoride selective ion electrode was used throughout. Potentials of this electrode were measured (with an accuracy of f 1mV) relative to a saturated calomel electrode using a Radiometer PHM-62 standard pH meter. The pH of the solutions was measured with the same instrument using a glass membrane electrode. Method for Determining Fluoride. A standard solution of 500 ppm fluoride was prepared by dissolving the appropriate amount of NaF in distilled water. Immediately after preparation, this solution was transferred to a polyethylene container. Standard solutions of 100 and 10 ppm were prepared by sequential dilution and stored in polyethylene bottles. A 100-ml aliquot of the sea water sample was placed in a 250-ml Nalgene beaker and a Teflon coated stirring bar added. The fluoride electrode and calomel reference electrode were then inserted into the solution and the potential difference was measured. Aliquots of the standard fluoride solution were then added from a 10-mlmicroburet and the potential was difference measured after each addition. Except for one fluoride determination (the sea water containing a high concentration of aluminum),potential readings became stable (flmV) within 2 to 3 min. The fluoride concentration was determined as described in the Theory section.
THEORY The potential of a reversible selective 3n electrode is given by the expression:
RT
E=E,+-ln[FF
1
Emf
-20
t
E(mV)
Figure l a . Theoretical plot of emf vs. log [x] over the concentration
range where the amount of x added is approximately equal to the amount (a) of x originally present in the sample
Figure 2a. Actual emf vs. log [F-] plot for data obtained from a sea water sample
Emf
/ /
/
/
c.0
(Y) P
Figure lb. Portion of the emf vs. log [x] plot showing the choice of
6
parameters for data treatment
where E is the measured potential difference, and E , is a constant incorporating terms for the activity coefficient of fluoride when measurements are made a t constant ionic strength. Other symbols have their usual meaning. If the unknown concentration of fluoride present in the sample is a and the known concentration of fluoride added is x , the equation of the E-x curve generated by the standard addition experiment is
2.303RT E=E,+log ( a x ) F Figure l a shows the theoretical hyperbolic plot of E vs. log x . Clearly when added x a, the Nernstian response is expected. Thus, a hyperbolic curve characterized by two approximately linear regions, one of slope 0, and the other of slope 2.303 RT/F, is expected: These regions are labeled (i) and (ii) in Figure la. The equation of the tangent to (i) is
+
E = E , + - 2.303RT log a F
(a >> x )
(3)
and to (ii)
2
Flgure 2b. Plot of z vs. y showing the linear extrapolation to obtain a.
Note the value of a is independent of the chosen reference point. Fluoride concentrationsdetermined from intercepts (1.40 f 0.01) mg I.-' and from the slopes (1.40 f 0.01) mg I.-'
tion for a which requires only a single linear extrapolation is expected to be more accurate. The curved part of Figure l a corresponds to the potential readings influenced by both a and x . Figure l b shows that an arbitrary fixed reference point X I , now represented by c can be chosen on this curve and potentials AE measured relative to it: A t c
E1 = E ,
2.303RT +log ( a + c) F
(5)
At x 2 , where x 2 is a variable, now represented by the symbol z,
2.303RT E2=EC+log ( a F
+
2)
Subtraction of Equation 5 from Equation 6 gives
E = E , + - 2*303RTlog x F
(x
>> a )
(4)
The intersection of these two limiting straight lines occurs when log x = log a and the value of a , the unknown fluoride concentration, may be determined readily from this point (See Figure la). This approach is valid only when the electrode gives a Nernstian response over the concentration range under consideration. The concentration of fluoride in sea water is approximately 1.4 ppm or 7 X 10-5 M (14), and the fluoride electrode has been shown to give a Nernstian response a t this level and indeed 3 orders of magnitude lower (15). This graphical procedure gives the approximate concentration of fluoride by inspection, but a direct graphical solu-
2 303RT a +z E2 - E l = AE = -logF a+c FAE Letting y = antilog 2.303RT and rearranging Equation 7 yields ~
z = y(a
+ c) - a
(8)
Hence, a plot of z vs. y should be linear with an intercept of -a (at y = 0) and a slope of (a c j. Thus, under interference free conditions, both the intercept and slope should give the same value for a, allowing an internal check for the consistency of the data. Figure 2a shows a typical plot of E vs. log [F-] and
+
ANALYTICAL CHEMISTRY, VOL. 48, NO. 8, JULY 1976
1237
Table I. Determination of Fluoride in Sea Water under a Variety of Conditions to Demonstrate the Reliability of the Method [ F I T found mg I. --I
Sample
Point Lonsdale, Victoria, Australia
1
1.40
t
0.05
2
11.35 t 0.05a 1.30 t 0.05
3
1.40 i 0.05
4
1.30
5 6
1.28 i 0.05 1.36 i 0.05
7
1.43 t 0.10
8
1.39 2 0.05 1.37 i 0 . 0 5 a
9
1.30 i 0.05
10
1.37 i 0.05
11
1.28
i
t
Conditions
pHE8 1.5 ppm spike a d d e d . [ F ] ~ determined = 2.80 t 0.05 ppm 4.0 ppm spike added.[ FIT determined = 5.40 i 0.05 ppm 8.6 ppm spike a d d e d . [ F ] ~ determined = 9.90 i 0.05 ppm pH = 5.50 (HCl added) 5 mg potassium alum added/100 ml sea water 2.7 ppm A13+ 50 mg potassium alum added/ 100 ml sea water L= 27 ppm A13+.Equilibration time of electrode very long 50 mg potassium alum and 3 g ammonium citrate added/100 ml sea water. 37 mg ferric alum added to 100 ml sea water and pH adjusted t o 3 by adding few drops of glacial acetic acid. Soln L= 40 ppm Fe3+ 50 ml TISAB added/100 ml sea water Calculated via addition of TISAB and calibration vs. synthetic sea water pH 5.8 pH 4.3
0.05
0.10
Synthetic
0.0068 i 0.0008 0.0053 i 0.0008 a This determination was performed approximately three weeks after the first. Figure 2b illustrates a plot of z vs. y for the same data, using various chosen values of c. Several aspects of the above equations should be clearly recognized. The electrode strictly senses the activity of fluoride. However, for a 1:l electrolyte (NaF) added to a matrix of high ionic strength such as sea water, the approximation that concentration is proportional to activity over the concentration range of interest is valid and, hence, only for very accurate work would it be necessary to compute activities. This point is adequately presented in the work of Craggs et al. (12)in which the characteristics of a calcium electrode have been examined by a procedure related to ours. It has been well established that the fluoride electrode is highly selective in its response to the activity of fluoride with the only common interferant being hydroxide (16). Nevertheless, although sea water has a uniform pH of 8, possible hydroxide interference was shown not to be a problem as indicated by the results presented in Table I. Complexation of fluoride by metal ions in sea water has previously been overcome by the addition of TISAB solution. This reagent is presumed to release the bound fluoride by preferential complexation of the metal ions with EDTA type ligands present in the TISAB. Examination of the metal ions present in sea water (14) suggests that magnesium is the major species forming fluoride complexes since it has a Pl-value of about 20 mol-l 1. (MgF+) a t the appropriate ionic strength (17)and is present a t about 0.05 M. The following considerations for the simple case when only MgF+, Mg2+,and F- are considered demonstrate that even this species is unlikely to interfere. For the equilibrium
+
P1
Mg2+ F- s MgF+ the stability constant is given by
(9)
(10) 1238
ANALYTICAL CHEMISTRY, VOL. 48, NO. 8, JULY 1976
Also [MgF+] + LF-1 = [FIT where [FITis the total fluoride concentration. From Equations 10 and 11
Pi
(11)
[Mg2+][F-] + [F-] = [FIT
(12)
or
Thus, if 1 >> P1[Mg2+],then
[F-I = [FIT
(144
or, if ,&.[Mg2+] = constant, i.e. [Mg2+]= constant, then [F-] = constant [FIT
(14b)
[MgF+l + [Mg2+l = [ M g l ~
(16)
so that
or
Since and [ M g l is ~ almost constant during the titration and [ M g l ~ >> [F-]T it follows from Equations 11and 16 that [Mg2+]>> [F-] and that [Mg2+] N ( M g l ~ constant throughout the titration. Thus Equation 15b is applicable and despite the fact that the fluoride electrode responds to free fluoride, the method of standard additions provides a measure of the total fluoride rather than the uncomplexed fluoride as might be naively anticipated. A complete treatment considering complexes of all the major and minor ions in sea water confirms
0.06
0.0 2
-2.0
t-', /-5-
0 0.02~
Figure 3b. Plot of z vs. y for data trapolation Figure 3a. Plot of emf water sample
vs. log [F-] for data obtained from synthetic sea
the conclusion that the total fluoride concentration will be obtained using the method of standard additions. Baumann (11,15) used a similar treatment to explain the ability of the fluoride selective ion electrode to determine total fluoride in solutions of high acidity where the undissociated HF, rather than the free fluoride ion, is the predominant species.
RESULTS AND DISCUSSION Applying the graphical methods described above, excellent agreement with the theoretical predictions for the shape of the E vs. log [FIaddedplot was obtained with sea water under a variety of conditions as demonstrated in Figures 2a and 2b. Thus, the method of fluoride determination was shown to be valid. Results for sea water of 35.21%0salinity are presented in Table I, and the average value of 1.35 f 0.05 mg/l. is in excellent agreement with previously published data (2-5,14). The first result in Table I is the value found directly. The next set of three results shows that recovery of fluoride after deliberate addition to spiked samples is excellent. The fifth result is that obtained on an acidified sample. The sixth, seventh, eighth, and ninth sets of data show that the expected fluoride concentration is still obtained after deliberate addition of aluminum or iron in the form of their alums. Aluminum(II1) and iron(II1) form very strong fluoride complexes (18) and, provided that sufficient time is allowed for equilibration (as also noted by Baumann (15)for very low fluoride concentrations), total fluoride could still be determined accurately by the method of standard additions even though the artificial concentration levels of aluminum and iron were three orders of magnitude greater than those normally occurring in sea water (14). This observation is noteworthy, since the work of Liberti and Mascini (13) suggests that the direct method cannot be used for fluoride in the presence of high relative concentrations of aluminum(II1) or iron(II1) aquo species. I t is likely that this complication is avoided in sea water [even with relatively high concentrations of Al(II1) and Fe(III)] by the large excess of chloride ions which partially complex with any aluminum(II1) or iron(II1) present to produce chloro complexes of the metals, thus sequestering their influence on the fluoride concentration. Data obtained after addition of TISAB to release bound fluoride and using the calibration method further confirm the validity of the direct method. The data in Table I demonstrate the absence of hydroxide ion interference, the absence of interference caused by metal ion complexation and indicate that total rather than free fluoride is determined by the method of standard addi-
in Figure 3(a) showing the linear ex-
tions. Figure 3a shows a plot of potential vs. the logarithm of the concentration of added fluoride in synthetic sea water. The slope of the linear portion of the graph was 58.8 mV per decade change in fluoride ion concentration, and this value was used in the calculations of fluoride concentration in sea water. A t lower levels of fluoride M), it might be assumed that the electrode behaves in a non-Nernstian fashion. However, calculations of the kind suggested in this paper (Figure 3b) reveal that the deviation from linearity can satisfactorily be attributed to residual fluoride resulting either from reagent contamination or from the electrode itself a t a concentration of about 3 X M. The level of fluoride in sea water is a t least two orders of magnitude higher than this residual level, hence, contamination does not significantly influence the results. Subsequent to developing the method for determining fluoride described in this paper, a survey of waters of widely varying salinity collected from around the southeast coast of Australia confirmed the generality of the method. Furthermore, when handling large numbers of samples, computerized (least-squares fit of Equation 8), rather than graphical analysis, is considered to be substantially superior. The precision of the data reported in Table I is largely limited by the accuracy of the potential measurement (71 mV). Using instrumentation capable of measuring to 0.1 mV, and with leastsquare fitting of data, average deviations on replicate determinations were routinely found to be