Determination of As, Sb and Se in

Difficult Environmental Samples by

Hydride Generation

Application Note

Atomic Absorption

Author Introduction
Lucinda M. Beach Hydride generation techniques are used routinely for the determination of As, Se, and
Sb in environmental samples. Advantages include high sensitivity, simplicity and rela-
tive freedom from interferences. The current method for the hydride generation deter-
mination of As and Se, as outlined in the USEPA manual SW-846, utilizes zinc slurry
methodology. A batch type method is described. Batch type hydride generators are
inherently slow and prone to interferences. These problems can be overcome with
continuous flow systems which are faster, simpler to use and more reliable. Also,
interferences are more easily managed.

The USEPA is updating SW-846 with alternative hydride methods for As and Se,
while establishing a new method for Sb. The new methods will include the use of
NaBH4 rather than zinc slurry reduction and the utilization of continuous flow rather
than a batch type apparatus. Lockheed Engineering and Sciences Company (LESC),
Las Vegas, NV, wrote and evaluated the new methods. They also investigated the
use of complexing reagents such as L-cystine to reduce transition metal interfer-
ences. The Agilent AA Resource Center in Wood Dale, Illinois was involved in the
analysis of the referee samples. Although LESC utilized a “homemade” continuous
flow apparatus, all of the work done at Agilent utilized the Agilent VGA-76 Vapor
Generation accessory.

The VGA-76 features a continuous flow technique in which samples and liquid
reagents are pumped and allowed to mix. The gaseous reaction products are then
swept into an absorption cell located in the optical path of the AA. The VGA-76 can
be readily connected to the PSC-56 programmable sample changer to provide total
automatic presentation of up to 67 samples plus up to five calibration standards.
Sixty determinations per hour can be made with 1% or better RSD. The VGA-76 has
been described in more detail in Varian AA-at-Work No. 38 [1].
This paper discusses the hydride generation analysis of sev- Table 1. Method 3050
eral complex environmental samples. Numerous factors influ-
1. Transfer 1.0–2.0 g of homogeneous sample to beaker.
ence accuracy. Sample digestion methods and preparation
procedures are of critical importance. 2. Add 10 mL 1:1 HNO3, heat at 95°C for 10–15 minutes.
3. Cool, add 5 mL conc. HNO3. Reflux for 30 minutes then evaporate to
Digestion must adequately solubilize the elements of interest 5 mL.
without loss by volatilization. Preparation procedures that 4. Add 2 mL water and 3 mL H2O2, heat sample.
guarantee the proper analyte oxidation state and eliminate 5. Continue to add 1 mL H2O2 aliquots until clear.
the nitric acid and inter-element interferences are needed. 6. For flame: Add 5 mL HCl and 10 mL water. Heat 15 minutes, dilute to
100 mL.
Five solid samples and a blank water sample were selected
7. For furnace: Heat to reduce volume to 5 mL. Cool, dilute to 100 mL.
for this study. The samples were:

Sample #1 – NIST 1645 River Sediment

Table 2. Method 3010
Sample #2 – USEPA (WP286) Electroplating Sludge
1 Transfer 100 mL aliquot to a beaker.
Sample #3 – USEPA (1-884) Hazardous Waste Soil
2 Add 3 mL HNO3, heat and reduce volume to 5 mL.
Sample #4 – USEPA (EC-4) Industrial Sludge 3 Repeat step 2 until digestate is clear.
Sample #5 – NIST 1633 Fly Ash 4 Wash beaker and add 1:1 HCl. Heat 15 minutes.
5 Cool, wash beaker and filter if necessary.
Sample #6 – Water Blank
6 Adjust to 100 mL final volume with Type 2 water.
These were very difficult samples containing high levels of
potential interferents.
Two 1.0 gram portions of each sample were prepared. One
Experimental portion was digested “neat”. The second portion was spiked
with the analytes of interest before digestion and was identi-
Methodology fied with an ‘S’ suffix in the sample name. Spiking levels were
adjusted so that each sample contained, in addition of the
The samples were digested according to SW-846 Method 3050 indigenous levels, 100 µg/L Sb, 40 µg/L As, and 10 µg/L Se.
(solid samples) and Method 3010 (water samples). The diges- The aqueous digestion is outlined in Table 2. It was used to
tion procedures are outlined in Table 1 and Table 2. Each solid digest the blank water sample. This sample was also spiked
digestion (Method 3050) required the addition of 1.0 gram of with 100 µg/L Sb, 40 µg/L As, and 10 µg/L Se.
sample plus 10 mL of 1:1 nitric acid into a 250 mL beaker. The
beaker was covered with a ribbed watchglass and the contents These digested samples were provided to Agilent by LESC.
heated and refluxed without boiling at 95 °C for 10–15 minutes. Further preparation was necessary for vapor generation. The
The contents were allowed to cool. Five mL of additional nitric procedure for As and Sb is listed in Table 3. The samples were
acid was added and the mixture was refluxed under the watch- further diluted 1:10 with additional reagents. The final solution
glass cover for 30 minutes. This step was repeated once, after contained 50% HCl, 2% urea, and 1% KI. As and Sb determina-
which the volume was reduced to approximately 5 mL. The tions can be done with the same sample solution. The addition
mixture was allowed to cool. Two mL of Type 2 water and of KI guarantees that As and Sb are in the +3 oxidation state
3 mL of 30% H2O2 were added and the mixture was heated and urea minimizes the nitric acid interference [2].
until effervescence subsided. More hydrogen peroxide was
Table 3. Sample Preparation for As and Sb
added in 1 mL portions until the sample appearance remained
unchanged. At this point, the furnace analysis option of One sample preparation for both elements 5 mL sample placed in 50 mL volumetric.
Method 3050 was followed, wherein the acid-peroxide- sample 25 mL HCl added.
mixture was evaporated to approximately 5 mL. The beaker Mixed and cooled. 5 mL 20% urea added.
contents were filtered along with Type 2 water rinsings of the
Mixed and cooled. 2.5 mL 20% KI added.
beaker into a 100 mL volumetric flask, and diluted to volume
Diluted to mark with D.I. water.
with Type 2 water. During the study, LESC found it necessary to
concentrate the original digests to near-dryness after addition Analyzed for As and Sb after 30 minutes.
of an extra 5 mL of concentrated HCl. This extra step effectively
removed peroxide interferences.

2
The Se determination required a separate sample preparation The results of As determinations in the digested samples are
which is outlined in Table 4. Again, the digested samples listed in Table 6. Recall that each of the six samples had a
were diluted 1:10 with additional reagents. The final solution pre-digestion spike of 40 µg/L As (designated with an “s”).
contained 60% HCl and 0.4% urea. The 30 minute hot water The difference between the unspiked and spiked sample was
bath step was required to reduce Se+6 to Se+4. The sample often not equal to the spike level. This variability was proba-
degassed with the addition of urea. Dissolved gas can cause bly due to error introduced during digestion and sample han-
poor precision in hydride analysis. Allowing the samples to dling rather than errors in the instrument analytical method.
degas before analysis improved precision. Graphite furnace results are shown for comparison. There
was good agreement between the two methods except for
Table 4. Sample Preparation for Se
sample 2. Standard additions calibration was not used. There
5 mL sample placed in 50 mL volumetric. 30 mL HCL added.* appeared to be a matrix interference in the graphite furnace
Placed in hot water bath (boiling) for 30 minutes.
analysis of this sample (the electroplating sludge).
Samples cooled. Table 6. Arsenic Results (µg/L)
1 mL 20% urea added.
Sample Hydride Graphite
Samples cooled and periodically shaken to degas over next 30 minutes. generation furnace
Samples diluted to mark with D.I. water and analyzed for Se after they 1 245 260
had stopped degassing. 1s 281 294
* For standard preparation 3 drops of HNO3 were also added. 2 143 116
2s 195 150
3 54 56
3s 79 82
The VGA-76 was configured with the same conditions for all 3 4 45 47
elements. The acid channel contained concentrated HCl and 4s 75 78
the uptake rate was 1 mL/min. The reductant channel con- 5 527 530
tained 0.3% NaBH4 and 0.5% NaOH and the uptake rate was 5s 582 595
6 0 0
1 mL/min. The prepared sample uptake rate was
6s 36 40
6–7 mL/min. (The exact uptake rates were not determined.
They remained constant for samples and standards.) A spike recovery study was done and the results are shown in
Table 7. The sample solutions were spiked with 8 µg/L As
(samples 5 and 5s had been further diluted 1:2). Recoveries of
Results the 8 µg/L As were excellent, ranging from 87% to 108% or
7.0 to 8.6 µg/L. This indicated minimal interference in the
Arsenic VGA analysis of properly prepared samples.
The VGA-76 As calibration is shown in Table 5. The sensitivity
Table 7. Arsenic Spike Recovery Study
and precision were excellent. Although the sample digest
was diluted 10-fold in our sample preparation procedure, the Recovery
excellent sensitivity of the VGA-76 allowed the determination Expected Found of 8 ng/mL
of less than 1 µg/L. For example, 0.2 µg/L As would result in Sample results (ng/mL) spike (%)
an absorbance of 0.006. 1 32.5 33.1 108
1s 36.1 36.6 106
Table 5. Arsenic Calibration VGA-76 2 22.3 22.6 104
2s 27.5 28.2 109
Conc. Mean 3 13.4 14.0 108
Sample (ppb) %RSD abs 3s 15.9 16.5 108
Blank 0.0 0.008 4 12.5 12.7 102
4s 15.5 16.0 106
Standard 1 2.0 2.0 0.062
5* 34.4 33.3 87
Standard 2 5.0 0.9 0.148 5s* 37.1 37.0 99
Standard 3 10.0 0.6 0.262 All samples diluted 1:10
Standard 4 20.0 1.0 0.455 * Samples 5 and 5s further diluted 1:2

Standard 5 40.0 0.4 0.700

3
Antimony Table 10. Antimony Results (µL)

The VGA-76 antimony calibration is shown in Table 8. The sensi- Hydride Graphite
tivity was excellent. The highest standard was 20 µg/L Sb and it Sample generation furnace
gave an absorbance of 0.846. The precisions of 3 replicate read- 1 64 66
ings were excellent (< 1.5%). With this sensitivity and precision, 1s 109 102
0.1 µg/L Sb could be easily determined. 2 30 25
2s 68 54
Table 8. Antimony Calibration VGA–76 3 16 17
3s 37 32
Conc. Mean 4 3 4
Sample (ppb) %RSD abs 4s 42 41
5 4 4
Blank 0.0 0.000
5s 40 36
Standard 1 2.0 1.5 0.086
6 0 0
Standard 2 5.0 0.7 0.216
6s 105 102
Standard 3 10.0 0.5 0.425
Standard 4 20.0 1.5 0.846

A spike recovery study was done and these results are listed in
Shown in Table 9 are the absorbance and concentration results Table 11. The diluted samples solutions were spiked with
of the actual samples. Sample preparation involved a 1:10 dilu- 8 µg/L Sb. Recoveries of the 8 µg/L spike were excellent ranging
tion. Even with the dilution the precisions in the VGA determi- from 90% to 101% or 7.2 to 8.1 µg/L. There was no indication of
nations were 3 %RSD or better for all samples except the blank interference in the VGA analysis of the prepared samples.
(sample 6). Sample 4 contained 3.0 µg/L Sb (0.3 µg/L Sb with
dilution). The precision was excellent at 2.9 %RSD. Table 11. Antimony Spike Recovery Study

Table 9. Antimony Results VGA-76 Recovery

Expected Found of 8 ng/mL
Conc. Mean Sample results (ng/mL) spike (%)
Sample (ppb) %RSD abs 1 14.4 14.5 101
1 6.4 2.2 0.276 1s 18.9 18.1 90
1s 10.9 1.1 0.463
2 3.0 1.4 0.130 2 11.0 10.8 98
2s 6.8 2.3 0.295 2s 14.8 14.1 91
3 1.6 1.6 0.071
3 9.6 9.5 99
3s 3.7 0.2 0.160
4 0.3 2.9 0.015 3s 11.7 11.6 99
4s 4.2 1.6 0.182 4 8.3 8.0 96
5 0.4 3.0 0.017
4s 12.2 11.8 95
5s 4.0 1.0 0.171
6 0.0 51.9 0.001 5 8.4 8.1 96
6s 10.5 0.8 0.445 5s 12.0 11.3 91
All samples diluted 1:10 + 8 ng/mL spike

The results for Sb in the digested samples are listed with

graphite furnace results in Table 10. Each of the six samples had
Selenium
a predigestion spike of 100 µg/L Sb (designated with an “s”).
This predigestion spike was not recovered for any of the samples The selenium concentrations in the samples were very low.
except number 6 which was a blank water. This indicated severe The selenium predigestion spike concentration was also low
losses of Sb in the digestion procedure. at 10 µg/L. Therefore, a low level calibration was used and is
shown in Figure 1. The precisions for three replicate readings
of the standards were excellent at 1.6 %RSD or better. A
predigestion spike of 10 µg/L resulted in a spike concentra-
tion of 1.0 µg/L Se in the analysis solution after the 10 fold
dilution from sample preparation. This concentration can be
easily determined by the VGA and would give an absorbance
of approximately 0.030.

4
 Table 13. Selenium Spike Recovery Study

Recovery
Expected Found of 2 ng/mL
Sample results (ng/mL) spike (%)
1 2.4 2.5 105
1s 2.9 3.0 105
2 2.5 2.6 105
2s 3.4 3.4 100
3 2.6 2.7 105
3s 3.2 3.3 105
4 2.2 2.1 95
4s 2.7 2.6 95
Figure 1. Low level selenium calibration VGA-76.
5 5.7 5.6 95
5s 6.3 6.2 95
All samples diluted 1:10 + 2.0 ng/mL spike
Results for Se in the digested samples are listed in Table 12.
There was good agreement between hydride generation and Data Comparison from Two Hydride
graphite furnace determinations. The predigestion spike of
10 µg/L Se was not recovered except in the blank water
Generation Systems
sample (number 6). Again, this reflected problems which In the following portion of this paper the Agilent methodology
occurred during digestion. and results will be compared to that of Lockheed Engineering
Table 12. Selenium Results (µg/L) and Sciences Company (LESC). Hydride generation systems,
reagent concentrations, and results will be considered.
Hydride Graphite
Sample generation furnace Hydride methodologies are compared in Table 14. LESC uti-
1 4 5 lized a four channel system, however only three channels
1s 9 9 were used. The samples were introduced through channel 1
2 5 7 at a flow rate of 4.2 mL/minute. The reductant channel or
2s 14 13
channel 2 was used to introduce a solution of 4.0% NaBH4.
3 6 5
3s 12 11 The acid channel was not used. The fourth channel was used
4 < 2.0 < 0.5 to add KI for As and Sb determinations. The Agilent VGA-76
4s 7 7 conditions are also listed. The VGA-76 used 3 channels with
5 37 32 the KI added to the sample solutions for As and Sb determi-
5s 43 37
nations. Both systems used sample solutions of high HCl con-
6 0 0
6s 9 10 tent. (LESC further prepared the Method 3050 solutions by
adding HCl to make the analysis solutions 40% HCl.) The
A spike recovery study was also done for selenium. The
Agilent VGA-76 sample flow rate was almost double that of
sample solutions were spiked with 2.0 µg/L Se. Recoveries of
the LESC’s system. LESC used a NaBH4 solution 13 times
the 2.0 µg/L Se were excellent, ranging from 95% to 105% or
more concentrated than that used in the VGA-76. This solu-
1.9 to 2.1 µg/L Se. The results of this study are shown in
tion was also added at twice the rate of the VGA-76.
Table 13.
Considering that the LESC sample flow rate was half that of
Agilent, the ratio of NaBH4 to sample volume was 52 times
greater in the LESC system than in the Agilent system. LESC
added the KI solution automatically while the Agilent method
involved addition of 1% KI to the samples.

5
Table 14. Comparison of Hydride Conditions Table 15. Arsenic Data Comparison mg/L

Channel LESC Agilent

AFT BEF
Sample 40% v/v HCl 50% v/v HCl Agilent Aglient LESC LESC
4.2 mL/min 7 mL/min Sample GFAA hydride hydride hydride
Reductant 4.0% NaBH4 0.3% NaBH4 1 260 245 267 281
2.1 mL/min 1.0 mL/min 1s 294 281 310 315
Acid — Conc HCl 2 116* 143 167 203
1.0 mL/min
2s 150* 195 196 243
KI (As, Sb only) 20% *
0.5 mL/min 3 56 54 40 340

* 1% in samples 3s 82 79 64 212
4 47 45 36 164
LESC added L-cystine to the prepared samples making the final
4s 78 75 58 259
analysis solution 3% L-cystine.* This resulted in much better recov-
eries in studies of transition metal interferences. LESC had metal 5 530 527 505 1326
precipitation problems and severe interferences with high concen- 5s 595 582 559 624
trations of transition metals. With the conditions used in the VGA-76 6 0 0 1.6 SATD
there was no improvement with L-cystine in this study. Higher 6s 40 36 28 1018
NaBH4 concentrations are associated with greater transition metal * Standard additions calibration not performed
interferences [2]. With higher levels of NaBH4 L-cystine appears to
The antimony results are shown in Table 16. Again some of
suppress the metal interferences in As and Sb determinations.
the results without the extra evaporation step are high
L-cystine could not be used in the determination of Se. It caused
(BEF LESC HYD). Agilent results by graphite furnace and
complete suppression of the signal.
hydride compare fairly well with LESC’s “AFT” results.
Analytical results from both labs are compared in Table 15. Table 16. Antimony Data Comparison mg/L
Agilent results by hydride generation and graphite furnace are
shown. Two data sets are shown for LESC: before and after an AFT BEF
extra evaporation step at the end of the digestion. It appeared Agilent Agilen LESC LESC
that residual H2O2 caused a synergistic interference with Sample GFAA hydride hydride hydride
nitric acid resulting in a large background signal that could 1 66 64 59 63
1s 102 109 95 107
not be corrected for by the AA instrument used. The results
2 25 30 23 43
listed in the “BEF” (before) column are often very high. This 2s 54 68 49 79
problem was not encountered with the VGA-76 and the 3 17 16 13 25
Agilent SpectrAA-20 and SpectrAA-400. An extra evaporation 3s 32 37 27 48
step was not necessary. The samples were diluted further 4 4 3 8.2 28
4s 41 42 33 60
than LESC, however, detection limits were not compromised
5 4 4 4.4 46
due to the excellent sensitivity of the instrumentation. The 5s 36 40 28 25
extra dilution minimized the H2O2 interference and results 6 0 0 1.8 ND
compare very well with the LESC “AFT” (after) results. In the 6s 102 105 90 192
graphite furnace determination of As in sample 2 matrix inter-
ferences were encountered. Standard additions calibration
was not done and lower results were obtained. Detection A comparison of selenium results is shown in Table 17. Only
limits were not determined, therefore no results were one set of data from LESC was obtained for selenium. LESC
reported for sample 6. Sample 6s was the blank water spiked was not able to use L-cystine for the selenium determination
with 40 µg/L As. The problems that LESC was having with due to severe signal suppression. To minimize the transition
high background are apparent in the “BEF” results. With metal interference the digested solutions were diluted 10 fold.
sample 6 the background corrector was overranged or “satu- This dilution minimized the H2O2/nitric acid gaseous back-
rated”’ in the As determination. LESC’s “AFT” results were ground interferences. Therefore, the extra evaporation of the
considerably lower. Agilent results by graphite furnace and digest was not necessary. There was excellent agreement
hydride compare fairly well with LESC’s “AFT” results. between all three methods of analysis.
* Warning Care should be taken when L-cystine is added to high concentration
HCl solutions. Excessive degassing can result in the “boiling over” of strongly
acidic solutions. Adequate face and eye protection is required.

6
Table 17. Selenium Data Comparison mg/L obtained by graphite furnace and VGA-76 hydride analysis and
LESC’s data obtained after an extra evaporation step
Agilent Agilent LESC
Sample GFAA hydride hydride
(As and Sb). There was excellent agreement in the Se data.
1 5 4 4.3
1s 9 9 9.5 Conclusion
2 7 5 6
2s 13 14 17.5 Numerous factors affect the accuracy of the hydride generation
3 5 6 5.5
3s 11 12 11.5
technique. The digestion procedure can be a source of analyte
4 < 0.5 <2 <3 loss. Residual HNO3 and H2O2 can result in interferences. The
4s 7 7 8 nitric acid interference can be eliminated by the use of urea.
5 32 37 34.5 Residual H2O2 can be removed by an extra evaporation step at
5s 37 43 41 the end of the digestion or simple dilution can minimize the
6 < 0.5 <2 <3
6s 10 9 10.5
interferences encountered.

Transition metal interferences can be minimized by the use of

In this study, predigestion spikes often were not recovered. L-cytine, the use of high HCL concentrations and low NaBH4 con-
The percent recoveries of the predigestion spikes are sum- centrations. Agilent VGA-76 results were obtained without the use
marized in Table 18. They are calculated from the difference of L-cystine. Comparable results were obtained without L-cystine
between results obtained from the spiked and unspiked with the VGA76 methodology. Of course, different hydride generator
samples. The % recovery is the percentage this number is of conditions can result in different conclusions in interference studies.
the actual spike. The average percent recovery of the 5 sam- Nonetheless, with proper sample preparation, hydride generation
ples is shown. Sample 6s data was a blank water solution can be a very accurate technique for difficult samples.
and is shown separately. Good recoveries were obtained
from the blank solution for all three elements. Recoveries
were poorer with real samples. Only 25–36% of the spiked
Acknowledgement
Sb was recovered. The digestion used was known to result The author appreciates the work of Dave Dobb and Jeff Lindner
in poor antimony recovery due to poor solubility and reten- of Lockheed Engineering & Sciences Company of Las Vegas and
tion of antimony in the non-solubilized portion of the thanks them for doing the digestions and providing the final
sample. The digestion appeared to result in fairly good report of their evaluation of hydride methods for the USEPA.
recoveries of As. Although Se recoveries were good for the
spiked blank solution, they were poor for the real samples,
the percent recoveries ranging from 56–74%.
Table 18. % Recoveries of Predigestion Spikes

Agilent Agilent AFT LESC

GFAA hydride hydride

_Arsenic
x (1S–5S) 100 90 66
6S 100 90 66
Antimony
_
x (1S–5S) 30 36 25
6S 102 105 88
Selenium
_
x (1S–5S) 56 66 74
6S 100 90 105
Accurate results are very dependent on digestion procedures
and sample handling that occurs before instrumental analysis.
In the analysis of difficult solid environmental samples this
may be the greatest source of error. It was the purpose of this
study to investigate the accuracy of the NaBH4 hydride tech-
nique and not the digestion procedures. For the most part
there was very good agreement between Agilent’s data
7
References
1 K. Brodie, B. Frary, B. Sturman and L. Voth, March 1983,
Varian Instruments at Work No. 38.
2 L. M. Voth-Beach and D. E. Shrader, Spectroscopy, 1,
November 1985, pp 60–65.

For More Information

For more information on our products and services, visit our
Web site at www.agilent.com/chem

www.agilent.com/chem
Agilent shall not be liable for errors contained herein or
for incidental or consequential damages in connection
with the furnishing, performance, or use of this material.

Information, descriptions, and specifications in this

publication are subject to change without notice.

© Agilent Technologies, Inc., 1992

Printed in the USA
November 1, 2010
AA105