ORIGINAL PAPER

Andrew O. Simm Æ Craig E. Banks Æ Shelley J. Wilkins

Nikos G. Karousos Æ James Davis Æ Richard G. Compton

A comparison of different types of gold–carbon compositeelectrode for detection of arsenic(III)

Received: 14 September 2004 / Revised: 2 November 2004 / Accepted: 4 November 2004 / Published online: 17 December 2004

� Springer-Verlag 2004

Abstract A study has been conducted using abrasivelymodified basal and edge-plane graphite, carbon-paste,and carbon–epoxy electrodes to create gold–carboncomposite electrodes. Using either nano or micro-sizedgold particles their suitability for use in detecting ar-senic(III) is assessed. It was found that gold arraysprepared from micron-sized particles gave the best per-formance for arsenic detection. In particular micronarrays produced in carbon-paste electrodes with aneasily renewable surface work well for detection of ar-senic, producing a detection limit of 5(±2)·10�9mol L�1, with a high sensitivity of 10(±0.1) A mol�1 L.

Keywords Arsenic(III) Æ Carbon-paste electrodes ÆAnodic stripping voltammetry Æ Gold microarrays ÆGold nanoarrays

Introduction

The claimed mass poisoning of millions of people byarsenic in contaminated drinking water in the Far Easthas been well reported over the last 20 years [1–6].Bangladesh, in particular, has up to 50 million people atrisk in what the World Health Organisation (WHO) callthe worlds worst mass human poisoning disaster [7–9].

The increase in human consumption of contaminatedground water in these areas over the last 20 years isbehind the mass poisoning outbreak. Hundreds ofthousands of wells were dug in remote villages as asource of clean drinking water free from pathogens; it isthought that in Bangladesh as many as 70% [10] of thesewells are contaminated with arsenic levels above theWHO limit of 10 ppb (0.01 mg L�1) [11]. It is necessarytherefore that each of these wells is monitored for ar-senic levels and closed if appropriate.

A variety of well known and accurate analyticaltechniques are easily capable of measuring the arsenicconcentrations in these wells, for example chemilumi-nescence [12], chromatography, spectroscopic methods[13, 14], and various forms of ICP–MS [15, 16]. (For anoverview of these methods see [17, 18].) These methodsare expensive and laboratory-based requiring welltrained technicians to conduct the measurements. Thesheer number of contaminated wells that need to beregularly monitored means that using these laboratory-based methods would simply not be practical. Althoughfield based ‘‘kits’’ based on the colorimetric Gutzeit test[19] are available for estimation of arsenic concentra-tions in well samples, they are not reliable [20] at the lowconcentrations required by the WHO and also requirethe use of highly concentrated acids which produce theextremely toxic gas arsine.

Electrochemical methods hold great promise forportable use in the field. Anodic stripping voltammetry(ASV) of arsenic(III) at gold [21], platinum [21] andmercury [22–24] surfaces has been well documented formany years. Optimisation of the accumulation periodand the use of staircase voltammetry during the strip-ping step combined with either hydrodynamic methodssuch as the dropping mercury electrode [22–26], therotating disc electrode [27, 28], the channel flow cell [29],ultra- [30] and infra- [31] sonicated electrodes ormicroelectrode arrays [32, 33] or nanoparticles [34] en-able extremely low detection limits easily capable ofanalysing samples down to the WHO required levels,Table 1. However while some of these techniques could

A. O. Simm Æ C. E. Banks Æ R. G. Compton (&)Physical and Theoretical Chemistry Laboratory,Oxford University, South Parks Road,Oxford, OX13QZ UKE-mail: Richard.Compton@chemistry.ox.ac.ukTel.: +44-1865-275413Fax: +44-1865-275410

S. J. WilkinsDepartment of Materials, Oxford University,Parks Road, Oxford, OX13PH UK

N. G. Karousos Æ J. DavisChemistry Department, University of Surrey,Guildford, Surrey, GU2 7XH UK

Anal Bioanal Chem (2005) 381: 979–985DOI 10.1007/s00216-004-2960-z

be realistically turned into a portable arsenic detector,the use of expensive gold or platinum surfaces wouldlikely make them prohibitively expensive. The cleaningof the electrode surfaces to a suitable level prior toanalysis might also be troublesome in the field.

Although there has been some success in the elec-trochemical detection of arsenic at other surfaces [35, 36]it seems that gold electrodes in particular are key to thedetection. However although some gold-based methodssuch as those based on micro and nano arrays enablevery low 10�10 mol L�1 detection limits it is difficult tosee how these could be cheaply turned into a practicalproduct. If an electrochemically based method could beused to create a portable detection kit, the use of eithersingle-shot electrodes that might be mass producedcheaply on a large scale and simply be thrown away afterthe analysis has been completed, or electrodes with aneasily renewable surface would seem a sensible step inmaking the product a practical and commercial reality.

In this paper we look at several different ways tocheaply produce micro and nano-sized gold arrays ondifferent carbon surfaces, namely edge plane, basal planepyrolytic graphite (bppg), and graphite powder andcompare their ability to determine arsenic(III) to therequired trace levels.

Experimental

Reagents and chemicals

All chemicals used were of analytical grade and wereused as received without any further purification. Thesewere: sodium (meta) arsenite, (Fluka, +99.0%)—highlytoxic handle with care, synthetic graphite powder (Al-drich, 1–2 lm), spherical gold powder 3.0–5.5 lm

diameter (Alfa Aesar.), gold colloid solution of 5 nmdiameter (Sigma, 0.01% as HAuCl4), mineral oil (Al-drich, CAS 8042-47-5), epoxy resin (Durcisseur MA2),nitric acid (Aldrich, 70%, double distilled PPB/Teflongrade with trace metal impurities in parts per trilliondetermined by ICP–MS).

All solutions were prepared with deionised water ofresistivity not less than 18.2 MW cm (Vivendi watersystems, UK) and degassed using N2 prior to measure-ments.

Instrumentation

Voltammetric measurements were carried out usinga l-Autolab III (ECO-Chemie, Utrecht, The Nether-lands) potentiostat. All measurements were conductedusing a three-electrode cell. The counter electrode was abright platinum wire, with a saturated calomel electrode(Radiometer, Copenhagen, Denmark) as the reference.Basel plane pyrolytic graphite and edge-plane electrodes(4 mm diameter) sealed in Teflon housing were used asworking electrodes (Le Carbone, Sussex, UK). Dispos-able plastic syringes (2 mL) (Gillette) with a 3 mmdiameter opening were used to contain carbon paste andcarbon epoxy electrodes, the syringe enabled simple re-newal of the surface of carbon-paste electrodes. A digitalvoltmeter was used to test the conductivity of manu-factured electrodes, (Cirkit, TM 5315B). Pad-printedelectrodes were obtained using a PE-4C (Pad PrinterEng., Supplied by Pad Print UK, Wellingborough, UK)pad-printing machine from an etched stainless steelplate/closed cup arrangement containing a silver-freecarbon ink (Creative Materials, Tyngsboro, USA). Theelectrodes were typically comprised of five print layersdeposited on standard overhead projection film acetate

Table 1 A summary of differentmethods developed for arsenicdetection

Authors Electrodematerial

Deposition time Limit of detection(mol L�1)

Tomcıket al. [33]

Interdigitatedplatinum array

n/a 7.0·10�6

WHO safe humanconsumption limit1.3·10�7 mol L�1

Svancara et al. [40] Gold-coatedcarbon paste

15 s 4.1·10�8

Prakash et al. [27] Rotating gold disc 60 s 2.7·10�8Simm et al. [30] Ultrasound-assisted

gold disc60 s 1.2·10�8

Huang et al. [29] Portable gold flow cell 180 s 7.0·10�9Simm et al. [31] Gold disc integrated

into a portable sonotrode120 s 3.7·10�9

Forsberg et al. [21] Gold disc 1200 s 2.7·10�9Salimi et al. [35] Iridium oxide-modified

boron-doped diamondn/a 2.0·10�9

Feeney et al. [32] Micro fabricated goldultra microelectrode array

80 s 6.4·10�10

Dai et al. [34] Gold nanoparticle-modifiedglassy carbon

180 s 1.3·10�10

980

sheets. High-resolution scanning electron microscopy(SEM) imaging was performed using a Jeol JSM 6500FFEGSEM (Field effect gun). Conventional secondary-electron imaging was employed for surface topologytogether with backscattered electron detection to obtaincompositional contrast. Energy dispersive X-ray detec-tion (EDX) was used to obtain information about thecomposition of the sample. All samples were carbon-coated prior to SEM analysis.

Results

Gold-modified carbon electrodes

In this section we compare micron and nano-sized goldparticles when immobilised on basal plane and edgeplane carbon electrodes in respect of their ASV responseto As(III). Immobilisation of the powder on an electrodesurface was chosen as an easy way of comparing theperformance of the different powders. Basal plane elec-trodes have a higher resistance than edge-plane elec-trodes, because of the alignment of the graphite sheetswith respect to the electrode surface; for this reason theelectron-transfer kinetics of edge-plane graphite arefaster than those of basal-plane graphite [37].

Colloidal gold-loaded carbon powder was preparedby the method suggested by Ju et al. [38]. Colloidal goldparticles (4 nm) were immobilised on 1–2 lm diametergraphite powder by adding a gold colloid solution(0.01% as HAuCl4) to the graphite powder (2 lL:1 mg).The solution was then thoroughly mixed with thegraphite powder and allowed to completely dry in air forapproximately 12 h. The dried powder was then used toabrasively modify a basal plane pyrolytic graphite elec-trode by rubbing a small amount of the powder into theelectrode surface on filter paper.

Anodic stripping voltammetry experiments using ar-senic(III) in 0.1 mol L�1 nitric acid solution were thencarried out with the modified electrode, using a 60 sdeposition period at �0.5 V followed by a stripping stepat 50 mV s�1. No response was observed to arsenic in0.1–10 lmol L�1 concentration. A small poorly definedarsenic(III) stripping peak was observed at millimolconcentrations. The stripping peak appears at ca. 0.1 V(vs. SCE) which corresponds well with literature values[30, 31].

An amount of 20 lL 0.01% gold colloid solution wasthen placed directly on the surface of a bppg electrodeand left to dry in air. ASV experiments using1 lmol L�1 arsenic(III) additions were then conductedusing a deposition potential of �0.5 V for 60 s. A smallarsenic peak could be seen at ca. 0.1 V (vs. SCE).However it did not become significantly large untilaround 100 lmol L�1. This is a reasonable improvementin sensitivity compared with abrasively modifying theelectrode surface with carbon powder modified withadsorbed gold colloid.

Gold spheres, 3.0–5.5 lm, were first abrasively at-tached to bppg and edge-plane electrodes by rubbing asmall amount on the gold powder into the electrodesurfaces on filter paper. Using a bppg-modified electrodewith a deposition potential of �0.5 V for 60 s no arsenicsignal was visible at the 0.1–100 lmol L�1 levels.Repeating the same experiment, but this time using anedge-plane electrode, resulted in no significant arsenicsignal being visible in the 0.1–10 lmol L�1 level. How-ever a significant well defined arsenic stripping signalwas visible at 0.1 V (vs. SCE) at around 100 lmol L�1

concentration levels. This suggests that the lower back-ground current of edge-plane graphite over basal-planegraphite leads to increased sensitivity at the100 lmol L�1 level over bppg for arsenic stripping at thegold particles. However the extent is not significant en-ough for the method to be analytically useful in thecontext of groundwater evaluation.

Gold spheres, 3.0–5.5 lm, were then mixed with1–2 lm carbon graphite powder, 1:1 by weight. Themixture was then abrasively attached to edge-plane andbppg electrodes by rubbing on filter paper. With thebppg electrode again no significant ASV response toarsenic after a 60 s deposition at �0.5 V was seen using0.2–10 lmol L�1 concentrations. However a relativelysmall stripping peak could be seen at 100 lmol L�1

levels. Using an abrasively modified edge-plane electrodea very small arsenic signal was visible with 0.2–10 lmol L�1 concentrations; however the signal did notbecome significant until 100 lmol L�1 levels, where awell defined stripping signal was seen, Fig. 1. The im-proved sensitivity of gold particles mixed with graphitepowder would suggest that there is better adhesion ofthe gold on the electrode surface probably because of the

Fig. 1 ASV response to 1 lmol L�1 arsenic(III) additions followedby a large 100 lmol L�1 addition. From a micron gold–carbonmodified edge-plane graphite electrode. Potential held at �0.5 Vfor 60 s followed by a stripping step using linear sweep voltamme-try at 50 mV s�1

981

surface roughening. It is also possible that the graphitepowder prevents the aggregation of the gold particles asthey are abrasively attached to the surface.

Gold-loaded carbon electrodes

Now we turn to the construction of solid electrodes inthe form of carbon paste and carbon epoxy electrodescreated from the gold–carbon mixtures investigated inthe previous section.

Carbon epoxy electrodes

Carbon epoxy electrodes where the modified carbonpowder is effectively ‘‘glued’’ in place by the epoxy to

form a solid polishable electrode surface have been de-scribed by Lawrence et al. [39] as a way of preparing analternative renewable carbon surface to that of classicalcarbon-paste-based electrodes. Electrodes were preparedin the manner described using the suggested 3:1 byweight ratio of electrode material (gold and carbonmixture) to epoxy resin. Once prepared however, al-though the electrodes were found to be conductive usinga digital voltmeter, no response to arsenic addition, evenat the millimol level, was seen for either gold colloid orgold micron particle-based carbon epoxy electrodes,suggesting the resin probably coated the gold particlesleaving little or no gold exposed at the electrode surface.

carbon-paste electrodes

Gold-plated carbon-paste electrodes have been de-scribed by Svancara et al. [40] for the detection of ar-senic(III), for which they achieved a detection limit of4·10�8 mol L�1. However in the present report weinvestigate their use with carbon–gold mixtures ratherthan coating the carbon paste to form a gold macrosurface. The gold colloid-modified graphite powder wasprepared as described in Sect. 3.1. When the mixture haddried completely in air, mineral oil was added to themixture (2 graphite:1 mineral oil by weight), forming apaste. This paste was then transferred to a 2-mL plasticsyringe and the surface of the paste in the nozzle useddirectly as the working electrode enabled easy renewal ofthe electrode surface. Before use a small amount of pastewas squeezed out and then rubbed down on sand paper.ASV experiments were unable to detect any response toarsenic even at the millimol level suggesting the oil mayhave completely covered the gold colloid.

The experiment was then repeated using the micronsized gold particles prepared in the same way (2:2:1gold:graphite:mineral oil, by weight). SEM images,

Fig. 2 SEM image of the micron-sized gold carbon-paste electrodesurface at 600· magnification

Fig. 3 Square-wave ASVresponse to successive 20 nmolL�1 arsenic(III) additions.From a micron sized gold–carbon paste compositeelectrode. Potential held at�0.5 V for 120 s followed by astripping step using square-wave voltammetry with theconditions: amplitude 250 mV,frequency 50 Hz, step potential5 mV

982

(Fig. 2) clearly show gold spheres on the electrode sur-face; a surface density of one gold sphere per 131 lm2

was calculated from the images. A well defined arsenicsignal was seen to respond with significant sensitivity toadditions of 1 lmol L�1 arsenic(III), the stripping signalcould be seen at approximately 0.1 V (vs. SCE), despitea large background current from the carbon paste visiblein the 0.2–0.5 V (vs. SCE) region. Square-wave vol-tammetry was then conducted to further enhance thestripping signal and obtain a limit of detection for thedetection of arsenic(III). Using a 5 lmol L�1 concen-tration of arsenic and a 120 s deposition period at�0.5 V (vs. SCE) a stripping signal was obtained at0.1 V. The square-wave measurements were optimisedby first varying the square-wave amplitude from 10 to

250 mV while keeping the frequency constant at 50 Hz.Increasing the amplitude was found to slightly broadenthe stripping peak but also significantly increase thesignal while shifting the stripping peak away from thecarbon paste background signal, therefore an amplitudeof 250 mV was used. Using this value the frequency wasthen varied from 50 to 200 Hz. While increasing thefrequency was found to increase the signal peak heightthere was an associated increase in signal noise thereforethe original value of 50 Hz was used, because no noisewas present using this value. Using optimised square-wave voltammetry the limit of detection (based on 3r)from the arsenic peak height using additions of 20 nmolL�1 was found to be 5(±2)·10�9 mol L�1, with a sen-sitivity of 10(±0.1) A mol�1 L, (�1.9·10�7 A(±2.4·10�8) R=0.9995), the response was found to belinear in the range 20–220 nmol L�1, Fig. 3. The sen-sitivity was monitored over several weeks and was notfound to decrease significantly during that time.

Pad-printed carbon/Au electrodes

‘‘Pad-printed’’ electrodes are a relatively cheap way ofmass producing ‘‘single-shot’’ electrodes for use assensing devices [41]. A conductive carbon-based ink isused to pad print a pattern containing ca. 30 electrodeswith a surface area of 18 mm2 on to acetate sheets. Bymixing gold particles with the ink it is possible to printlarge numbers of carbon–gold array electrodes.

The 3–5.5 lm gold particles were mixed to a com-position of 10% by weight with the conductive carbonink. The ink was then used to ‘‘pad print’’ single-shotelectrodes on to an acetate sheet. Figure 4 shows anSEM image of the pad-printed electrode, micron sizespheres are clearly visible as being well spread out on theelectrode surface with little aggregation. EDAX analysisconfirms the spheres to be gold, Fig. 5, clearly having adifferent composition to the carbon ink. It is estimated

Fig. 5 EDAX image-analysis plots showing the composition of theelectrode surface. In the left hand plot the beam is centred on one ofthe gold particles whereas in the right hand plot the beam is centredaway from the particles on the bare carbon surface

Fig. 4 SEM image of the pad-printed electrode surface at 850·magnification

983

from the SEM analysis that there is one gold sphere per370 lm2 on the electrode surface.

The response to arsenic(III) was tested in 0.1 mol L�1

nitric acid solution. Using a 60 s deposition time at�0.5 V a small peak was visible at ca. 0.1 V with1 lmol L�1 additions. Using square-wave voltammetryto enhance the sensitivity a limit of detection (based on3r) from peak area was found to be 9 (±3)·10�7mol L�1, (2.1·10�8 A (±2.6·10�9) R=0.9887, the lin-ear range was found to be from 1 to 5 lmol L�1. Severalpad-printing runs were tested and all were found toachieve detection limits at this level.

The much higher detection limit achieved with thepad-printed electrodes compared with the performanceof the carbon-paste electrodes might be explained by theSEM images in Fig. 6. The left hand image shows a barepad-printed surface from an electrode composed of onlythe carbon ink and no gold particles, ink globules ca.0.1 lm in diameter can clearly be seen covering theelectrode surface. The right hand image shows a gold–carbon pad-printed electrode, two gold spheres areclearly visible on the electrode surface however the sameglobular ink deposits appear to have completely coveredthe gold surface.

Conclusion

Carbon composite gold electrodes have been shown tobe a valid alternative to solid gold surfaces for arsenicdetection, and several different electrode types have beenexplored. Although most show a response to arsenic,only the carbon-paste-based electrodes were able toachieve a detection limit sufficiently low to be of ana-lytical value.

The attempt to produce gold–carbon electrodes ona large scale via pad printing, although not able toreproduce the sensitivity found in the carbon-pasteelectrodes, did provide a limit of detection close to themagnitude required by the WHO [11] for groundwateranalysis (10 ppb). With further optimisation theymight provide an inexpensive basis for the productionof single-shot arsenic detectors in the future.

Acknowledgements AOS thanks the EPRSC and Abington PartnersSensing for support via an Industrial CASE studentship. CEBthanks the EPRSC for support via a project studentship (GrantGR/R14392/01).

References

1. Chakraborti D, Rahman MM, Paul K, Chowdhury UK,Sengupta MK, Lodh D, Chanda CR, Saha KC, Mukherjee SC(2002) Talanta 58:3

2. Giger W, Berg M, Pham HV, Duong HA, Tran HC, Cao TH,Schertenleib R (2003) Chimia 57:529

3. Mandal BK, Suzuki KT (2002) Talanta 58:2014. Kinniburgh DG, Kosmus W (2002) Talanta 58:1655. Bissen M, Frimmel FH, Christ AG, Aesch S (2003) Acta

Hydrochim Hydrobiol 31:96. Bissen M, Frimmel FH, Christ AG, Aesch S (2003) Acta

Hydrochim Hydrobiol 31:977. Smith AH, Lingas EO, Rahman M (2000) Bull World Health

Organ 78:10938. Feeney R, Kounaves SP (2002) Talanta 58:239. Alam MGM, Allinson G, Stagnitti F, Tanaka A, Westbrooke

M (2002) Int J Environ Health Res 12:23610. Rasul AB, Munir AKM, Hossain ZA, Khan AH, Alauddin M,

Hussam A (2002) Talanta 58:3311. WHO (1993) http://www.who.int/mediacentre/factsheets/fs210/

en/12. Fujiwara K, Tsubota H, Kumamaru T (1991) Anal Sci 7(Suppl

Proc Int Congr Anal Sci 1991 Pt 2) 7:108513. Gong Z, Lu X, Ma M, Watt C, Le XC (2002) Talanta 58:7714. Raman B, Shinde VM (1987) Anal Lett 20:102915. Story WC, Caruso JA, Heitkemper DT, Perkins L (1992)

J Chromatogr Sci 30:42716. Shum SCK, Neddersen R, Houk RS (1992) Analyst 117:57717. Hung DQ, Nekrassova O, Compton RG (2004) Talanta 64:26918. Franesconi KA, Kuehnelt D (2004) Analyst 129:37319. Gutzeit H (1879) Pharm Zeit 24:26320. Greschonig H, Irgolic KJ (1997) Arsenic exposure and health

effects. Chapman and Hall, London, p 11221. Forsberg G, O’Laughlin JW, Megargle RG (1975) Anal Chem

47:158622. Kryukova TA (1940) Zavodskaya Laboratoriya 9:95023. Cozzi D, Vivarelli S (1951) Anal Chim Acta 5:21524. Susic M, Pjescic M (1968) Glasnik Hemijskog Drustva Beograd

33:30325. Li H, Smart RB (1996) Anal Chim Acta 325:2526. Brainina K, Neyman E (1993) Chemical analysis: a series of

monographs on analytical chemistry and its applications.Wiley, New York

27. Prakash R, Srivastava RC, Seth PK (2003) Electroanalysis15:1410

28. Sun Y-C, Mierzwa J, Yang M-H (1997) Talanta 44:137929. Huang H, Dasgupta PK (1999) Anal Chim Acta 380:27

Fig. 6 SEM images of a carbonpad-printed electrode with nogold present (left, 20,000·magnification) and a gold–carbon pad-printed electrode(right, 13,000· magnification)

984

30. Simm AO, Banks CE, Compton RG (2004) ElectroanalysisDOI: 10.1002/elan200403110

31. Simm AO, Banks CE, Compton RG (2004) Anal Chem 76:505132. Feeney R, Kounaves SP (2000) Anal Chem 72:222233. Tomcık P, Jursa S, Mesaros S, Bustin D (1997) J Electroanal

Chem 423:11534. Dai X, Nekrassova O, Hyde ME, Compton RG (2004) Anal

Chem 76:592435. Salimi A, Hyde ME, Banks CE, Compton RG (2004) Analyst

129:936. Zen J, Chen P, Kumar AS (2003) Anal Chem 75:6017

37. Banks CE, Davies TJ, Wildgoose GG, Compton RG (2004)Chem Commun DOI:10.1039/B413177K

38. Ju H, Liu S, Ge B, Lisdat F, Scheller FW (2002) Electroanalysis14:141

39. Lawrence NS, Thompson M, Davis J, Jang L, Jones T,Compton RG (2002) Chem Commun 21(10):1028

40. Svancara I, Vytras K, Bobrowski A, Kalcher K (2001) Talanta58:45

41. Mooring L, Karousos N, Livingstone C, Davis J, WildgooseGG, Wilkins SJ, Compton RG (2004) Sens Actuators B (inpress)

985