Analytica Chimica Acta 415 (2000) 21–32

Preliminary evaluation of adsorbent-based mercury removal systems forgas condensate

Azman Shafawia, Les Ebdona, Mike Foulkesa,∗, Peter Stockwellb, Warren Cornsba Department of Environmental Sciences, University of Plymouth, Devon PL4 8AA, UK

b PS Analytical Ltd., Arthur House, St. Pauls Cray, Orpington, Kent BR5 3HP, UK

Received 27 February 1999; received in revised form 2 February 2000; accepted 15 February 2000

Abstract

Mercury can be present in gas condensate in its metallic form and/or as organometallic compounds with boiling pointscomparable to that of the range of the condensate. The damage caused to industrial plants, particularly petrochemical plants,by the presence of certain mercury species can be financially crippling especially when unscheduled shut-downs are forced.Speciation of the mercury content of gas condensates is vital in the evaluation of an adsorbent system’s efficiency and forthe development, improvement and monitoring of the performance of newly developed or commercially available mercuryremoval systems. However, with the mercury species content being at such low levels, great demands are made of an analyticaltechnique.

Three commercially available mercury removal systems designated AA, BB, and CC were tested under pilot plant conditionsto evaluate their performance. System AA was a two-stage process involving a species conversion step prior to trapping ofmercury on an alumina-adsorbent impregnated with a metal sulphide. Systems BB and CC were single-stage trapping processesusing sulphide impregnated carbon and molecular sieves, respectively. Both real and substitute condensate sample i.e. hexanecontaining dimethyl mercury (DMM), diethyl mercury (DEM) and dibutyl mercury (DBM) species, were used in the trials.

The determination of total mercury was carried out by a proven method which involved the direct vaporisation withsubsequent adsorption of mercury species onto a gold-coated silica trap maintained at 200◦C. To release metallic mercury,the trap was heated at 900◦C and the analyte determined by atomic fluorescence spectrometry (AFS). The determination ofmercury species in the samples was carried out by direct injection using gas chromatography coupled, via a pyrolysis unit,with atomic fluorescence detection.

All three mercury-removal systems AA, BB and CC showed a reduction in the mercury content of the final condensatestreams. In the two-stage system, AA, the hydrogenolysis reactor converted some of the organomercury present in the gascondensate feed to its elemental form. However, the elemental mercury measured in the liquid product from this first reactorwas only about 30% of the total mercury content. Incomplete conversion of the organomercury species to mercury metalby this reactor may be due to competition between the organomercury species and the unsaturated compounds in the matrixduring the hydrogenolysis reaction. The second reactor, for mercury trapping, was able to adsorb efficiently elemental mercurypresent in the liquid stream (the product from the first reactor) but was unable to remove the organomercury content from thecondensate stream.

∗Corresponding author.E-mail address:mfoulkes@plymouth.ac.uk (M. Foulkes)

0003-2670/00/$ – see front matter © 2000 Elsevier Science B.V. All rights reserved.PII: S0003-2670(00)00838-2

22 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

For the single-stage adsorbent system BB, the efficiency in removing the species DMM, DEM and DBM fromn-hexanehydrocarbon samples was 100% with no indication of mercury present in the product stream over an 8 h continuous run.

For the adsorbent system CC, the efficiency of removal for different mercury species inn-hexane was variable. While DMMand DEM showed a consistent removal range of 50–80%, the DBM was efficiently adsorbed, but only for a short period (>80%at 2 h) and it was rapidly released back into the product stream after 4 h. © 2000 Elsevier Science B.V. All rights reserved.

Keywords:Gas condensate; Mercury speciation; Mercury removal adsorbent; Atomic fluorescence; Gas chromatography; Dialkyl mercury

1. Introduction

A knowledge of the total mercury content and thedifferent species present in gas condensate is ex-tremely important. Mercury in most forms is highlytoxic and, particularly when present as the organomer-cury species, is a cause for environmental concern.Mercury occurs naturally and the quantity and speciesfound is highly dependent upon the source, stage ofprocess, storage and age of the sample [1–4]. How-ever, in general, a value of 1–100 ng ml−1 mercury isexpected.

Mercury may be present in natural gas condensatein its metallic form and/or as organometallic com-pounds with boiling points comparable to that of therange of the condensate [5,6]. The elemental mercurycontent in liquid hydrocarbon or gas condensate maybe in the range of about 10% or less of the total mer-cury in the samples due to the high volatility and thelow solubility of elemental mercury [6]. The damagecaused to industrial plants particularly petrochemicalplants by the presence of certain mercury species canbe financially crippling especially when unscheduledshut-downs are forced. Information on the species con-tent is vital for the development, improvement andmonitoring of the performance of newly developedor commercially available mercury removal systems.Hence, with the mercury species content being uniqueto each type of condensate and at such low levels,great demands are made of an analytical technique.

Various papers report the speciation of mercury inhydrocarbons and gas condensate and a few are tar-geted at particular species of interest [7,8]. A shortreview of the background and methodology employedhas been published [9]. Recently, two papers have ad-dressed the problems of the determination of total mer-cury and total species content directly in condensatesamples using GC-AFS and gold-coated traps, respec-tively [10,11].

Mercury removal systems for both gas and liquidhydrocarbon streams are commercially available.However, removal systems for natural gas conden-sates have not been thoroughly tested under real plantconditions. Only a few units have been installed inreal plants and most of these were on a trial basis.One plant which possessed a mercury removal facil-ity still experienced mercury contamination in theirprocess steams because only mercury metal vapourwas targeted. This was due to little or no information[12] of the mercury species present in the streams.This lack of information leads to difficulties in de-ciding the most suitable removal system on bothtechnical and economic grounds. In general, thislimited knowledge on mercury in condensate arisesbecause:• The amount of mercury and the types of mercury

species present are not easily determined. This isdue to the nature of the samples. Natural gas con-densates are very volatile complex mixtures. Inaddition, the mercury species are volatile and canbe transformed into different forms of mercury,brought about by changes in their environmentsuch as temperature, pressure and the presence ofactive particles or impurities.

• The techniques employed for the determination ofmercury content in different process streams are notstandardised, validated or well documented.It is also noted that problems arise at the process

plant and laboratory level because:• The removal system selected may not perform to

expectation when installed in an actual, real plantenvironment due to unexpected interferences andthe more challenging nature of the real plant envi-ronment.

• Among the plant’s operators and laboratory staff,some may not be aware of the finer technical de-tails required in mercury removal and determina-tion, while others may not confide their expertise.

A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32 23

1.1. Characteristics of the mercury-removal system

The characteristics required of any mercury removalsystem are as follows [4,13,14]:• The removal agent must be highly active towards all

forms of mercury, preferably bonding irreversiblyto the agent (not to be re-released to the treatedstream).

• The removal agent must remain active, i.e. the activesurface must be resistant to blinding (or masking)by components in the stream being treated.

• The removal agent must not be harmful to the enduse of the products, or potentially harmful to down-stream components, i.e. by leaching of chemicalsor particulates.

• The system should ideally be flexible to the de-manding and uncertain nature of conditions of feedand process streams.

• The removal agent should be inexpensive and read-ily available.

• The removal agent should hold the mercury in eithera solid form, or in a liquid form from which it couldbe precipitated by a readily available agent and thenfiltered and disposed of.

1.2. Mercury-removal from natural gas condensate

The removal of mercury from natural gas conden-sate is very different to that from natural gas becauseit is in the liquid phase during the operation and,for a number of sources, the main types of mercurypresent in the condensate are organometallic (>80%).At present, three ‘technologies’ are said to be effec-tive for the removal of total mercury from feeds, des-tined for steam cracking or aromatisation (reforming).There are several manufacturers, but most of the prod-ucts are still under development i.e. at the pilot plantstages. In general, the removal systems used are givenin the following sections.

1.2.1. Sulphide-containing ion exchange resinmaterial

This system claimed to be effective for the removalof elemental and organomercury halide compounds.However, when considering the ion exchange char-acteristics, the effectiveness of the system in remov-ing dialkyl mercury species, which are known to benon-ionic compounds, is not known. At least three

commercially available systems of this type are knownto be offered to the industry [15].

1.2.2. Sulphide-containing aluminaThis system is only capable of adsorbing elemental

mercury. In order for the removal system to includeother species, e.g. organomercury species, a primaryor pre-processing stage is required. This should con-vert all forms of mercury into the metallic species.Hence, this process is in two stages [5,16,17]. The firststage of the process comprises a reactor loaded witha catalyst operating under suitable conditions, in thepresence of hydrogen or high temperature conditions,to convert/break both ionic and organomercury speciespresent in the condensate into metallic mercury. This issubsequently trapped on the sulphide-containing alu-mina in the second stage. However, the efficiency ofthe first reactor to convert ultra trace concentrations ofionic and organomercury compounds into elementalmercury is in doubt because of the lack of speciationinformation post-processing. This is particularly im-portant when the condensate contains alkyl mercurycompounds as the major species.

Other systems available or under developmentinclude, sulphur-containing molecular sieves andsulphur-containing activated carbon.

In this work we have sought to assess the effi-ciency of three commercially available mercury re-moval systems in removing several mercury speciesparticularly the alkyl mercury forms i.e. dimethyl mer-cury (DMM), diethyl mercury (DEM), dibutyl mer-cury (DBM) and diphenyl mercury (DPM) from liquidhydrocarbon and gas condensate. This paper presentsthe pilot plant performance of several commerciallyavailable adsorbents. The adsorbents were designatedadsorbent AA, BB, and CC. In order to critically assessthe efficiency of the BB and CC removal systems, theexperiments were sample-controlled usingn-hexanewith various species of dialkyl mercury. Due to thespecific two-stage process of the AA removal system,the feed sample required had to be a real condensate.

2. Experimental

2.1. Chemicals

DMM, DEM, DBM (Strem Chemical, Mas-sachusetts, USA), and DPM (Sigma Aldrich, Dorset,

24 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

UK), were used in the experimental procedures forcalibration and for spiking into AnalaRn-hexane(Merck, Poole, Dorset, UK) and real condensatesamples.

2.2. Identification of mercury removaladsorbents/catalyst

Two classes of mercury-removal system were iden-tified. The classification was based on the number ofstages that the process required. The adsorbent systemdesignated AA required a two-stage process, i.e. hy-drogenolysis prior to mercury removal. The adsorbentsystems BB and CC only required a single stage i.e.direct adsorption.

For the AA system, a hydrogenolysis catalyst wasused to convert organo- and inorganic mercury to el-emental mercury and a sulphide-containing aluminawas used for the removal of elemental mercury.

The BB mercury-removal system, was a carbon-based adsorbent which contained sulphur as the ac-

Fig. 1. A schematic diagram of the two-stage mercury-removal pilot plant.

tive material. The CC mercury-removal system was amolecular sieve-based adsorbent with undisclosed ac-tive material.

2.3. Testing procedure

2.3.1. Two-stage process (adsorbent AA)A schematic diagram for the pilot plant is shown

in Fig. 1. The pilot unit contains three reactors. Thehydrogenolysis reactor R1 is fed with an upwardflow of condensate and loaded with hydrogenoly-sis catalyst (32 ml, four beds of catalyst) in orderto transform the various mercury species to metal-lic mercury. The other two reactors, R2 and R3contain the trapping media (32 ml) and are used inseparate liquid and gas lines for removing elementalmercury.

Activation of the hydrogenolysis catalyst is requiredprior to processing and this is achieved by reducingthe catalyst with hydrogen (3 l h−1). The effect of thisreduction is to remove the chloride precursor.

A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32 25

Fig. 2. A schematic diagram of the single-stage mercury-removal pilot plant.

At the end of the activation procedure, the reac-tor temperature is reduced to 150◦C under the samehydrogen flow rate. The condensate feed is injectedwhen the temperature is stabilised. The hydrogenoly-sis reactor is maintained at 200◦C and the flow rate forthe actual samples was set to 128 ml h−1 (liquid spacehourly velocity (LSHV) 4 h−1).The operating pressurefor this reactor was 30 bar. For mercury removal, re-actors R2 and R3 are at the same flow rate as the R1reactor, but maintained at ambient temperatures.

The system was conditioned for 48 h before the col-lection of condensate feed samples, samples after theR1 reactor and products from the R3 reactor.

2.3.2. Single-stage processTwo types of adsorbent were tested using the

single-stage pilot plant. These were adsorbents BBand CC. The schematic diagram for the pilot plant isshown in Fig. 2. The pilot plant consists of a singlereactor R1 (diameter 4 cm, length 40 cm) and variableflow liquid pump. The processes were operated at am-bient temperature. The amount of adsorbent loaded

in the reactor was 100 ml. The adsorbent was packedbetween two plugs of carborundum packing material.Prior to the actual run, the system was conditionedfor 6 h by introduction of mercury freen-hexane. The

Table 1GC-pyrolysis-AFS operating conditions for RTx-1 and Cp-Simdistcolumns

Parameters Operating conditions

RTx-1 Cp-Simdist

Gas chromatograph PYE 104 Unicam Varian 3400Injector temperature 125◦C 125◦CColumn carrier gas — argon 6 ml min−1 5 ml min−1

Make-up gas — argon 300 ml min−1 300 ml min−1

Column programmeInitial temperature 40◦C 35◦CRamp temperature 4◦C min−1 5◦C min−1

Final temperature 300◦C 300◦CPyrolysis temperature 800◦C 800◦CShield gas flow rate —argon 600 ml min−1 600 ml min−1

Mercury wavelength 253.7 nm 253.7 nmIntegrator HP 3395 HP 3396

26 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

Table 2Total mercury content in samples by the vaporisation technique-system AA

Feed condensate, Condensate after hydrogenolysis Condensate after mercury trapping reactor,average±S.D. (ng ml−1) reactor, average±S.D. (ng ml−1) average±S.D. (ng ml−1)

26.2±1.3 70.9±8.1a 17.8±1.4

a Based on 6 and 10 replicates.

feed flow rate was set at 400 ml h−1 (LSHV about4 h−1). After completion of the conditioning stage,n-hexane spiked with organomercury species wasintroduced as feed. The system was stabilised for an-other 2 h before collection of samples from both feedand product stages were taken (every 2 h for mercuryspecies determination).

The mercury removal process was carried out fora total of 8 h. After this process, a mercury-freen-hexane feed was introduced and the reactor tubewas reversed. Samples taken after the reactor weremonitored for mercury species content. The experi-ments were performed to assess the capability of theadsorbents at removing particular mercury species,i.e. the major organomercury compounds. The ad-sorbing capacity or break through curve was notdetermined due to time constraints and specificationlimitations.

2.4. The determination of total and mercury speciesin samples

The determination of mercury species in thesamples was carried out using gas chromatography

Table 3Speciation data for feed condensate — system AAa

Peak No. Retention Feed Condensate after hydro- Condensate after Hg0 Speciestime, condensate genolysis reactor (R1) trapping reactor (R3)±0.5 min

Area Mercury content Area Mercury content Area Mercury content(%) (ng ml−1)b (%) (ng ml−1)b (%) (ng ml−1)b

1 1.7 0c 0c 33.4±3.2 23.7 0c 0c Hg0

2 5.5 10.3±0.5 2.7 8.1±1.1 5.8 11.3±0.1 2 DMM3 9.5 43.4±5.1 11.3 39.0±2.9 27.6 50.3±0.7 9 Dialkyl Hg species4 12.5 29.2±5.1 7.7 19.5±1.4 13.8 38.4±1.7 6.8 DEM5 13.5 17.1±1.8 4.5 0c 0c 0c 0c Dialkyl Hg species

Total 26.2 70.9 17.8

a S.D. based on four replicate analyses.b Mercury content calculated based on total mercury analysis (ng ml−1).c Below LOD (60 pg ml−1).

coupled via a pyrolysis unit with atomic fluorescencedetection. The technique utilised the superior separa-tion advantage of gas chromatography and the highsensitivity and element specific detection propertiesof atomic fluorescence detection for mercury speci-ation work. The limit of detection based on 3 S.D.(99.7% confidence) was determined to be 60 pg ml−1

for a 1ml injection. Direct injection, without anypre-treatment, derivatisation or extraction of the sam-ple, was used to avoid problems associated withthe sample type and species having a relatively lowboiling point (<250◦C) and stability [10].

The determination of total mercury was based on asimple and reliable procedure for the determination oftotal mercury in natural gas condensate which elimi-nated the use of chemicals/additives and complicateddigestion procedures. The determinations were car-ried out by the direct vaporisation of samples (usu-ally 0.25–1.0 ml) at 400◦C within a specially designedvolatilisation chamber, with subsequent adsorption ofmercury species by a gold-coated silica trap (P.S. An-alytical, Orpington, Kent, UK) maintained at 200◦C.To release metallic mercury, the trap was heated to900◦C (Galahad unit, P.S. Analytical, UK) and the

A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32 27

Fig. 3. Mercury chromatograms from feed, hydrogenolysis reactor and trapping reactor samples.

28 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

analyte determined by atomic fluorescence spectrome-try (Merlin detector, P.S. Analytical, UK). Full detailsof the methodology employed for the determinationof total mercury in condensates is discussed in [11].In this paper, near to full recoveries (>90%) were ob-tained for up to nine mercury species (including Hg0)when spiked into real condensate samples; the cali-bration slopes being near identical to that of Hg0 formost species. The method limit of detection, based on3 S.D. (99.7% confidence) was determined as 270 pgabsolute (0.25 ml injection).

For the adsorbent AA system, the GC col-umn used for speciation analysis of samples wasa 30 m×0.53 mm i.d. megabore fused silica col-umn coated with 1mm non-polar crossed-linkedpoly-dimethyl siloxane stationary phase (RTx-1)(Restek, Benner Circle, Bellefonte, Pennsylvania,USA). The GC unit used was a PYE Unicam model104 gas chromatograph. The injection system wasconstructed with a deactivated silica-lined adapter forsplitless injection. The column was inserted througha pyrolysis unit (Model CAL 9900, PS Analytical,UK) which was maintained at 800◦C and the outletconnected to the atomic fluorescence detector (Mer-lin, PS Analytical, UK) using 1.65 and 3.18 mm i.d.Teflon tubing. In between the heating block of thepyrolysis unit and the GC oven, a 6.35 mm i.d. coppertube, fitted with a heater, was installed to avoid a coldspot and hence prevent any condensation of elutedspecies. The copper tube also acted as a support forthe fragile column link. A single gas, argon, was used

Table 4Mercury species before and after removal system for adsorbents BB and CCa

Sampling No. Adsorbent BB (carbon-based) Adsorbent CC (molecular sieves-based)

Feed Product Feed Product

Area (×106) Area (×106) Area (×106) Area (×106)

Peak 1 Peak 2 Peak 3 Peak 1 Peak 2 Peak 3 Peak 1 Peak 2 Peak 3 Peak 1 Peak 2 Peak 3(DMM) (DEM) (DBM) (DMM) (DEM) (DBM) (DMM) (DEM) (DBM) (DMM) (DEM) (DBM)

Sample 1, 11:00 hours 9.1 16.4 9.7 0b 0b 0b 8.5 7.9 16.4 3.8 2.9 2.8Sample 2, 13:00 hours 8.9 15.9 8.1 0b 0b 0b 8.1 7.7 16.1 3.9 0b 4.5Sample 3, 15:00 hours 7.9 15.4 9.1 0b 0b 0b 8.8 8.2 15.8 3.9 0b 13.4Sample 4, 17:00 hours 6.8 14.9 9.4 0b 0b 0b 8.1 7.2 16.2 3.8 1.4 29.4

Overall 8.2±1.1 15.7±0.7 8.9±0.7 0b 0b 0b 8.4±0.4 7.8±0.4 16.1±0.2 3.8±0.1 1.1±1.3 12.5±12.1

a Column used: WCOT Ultimetal, with Cp-SimDist stationary phase (1.2mm thickness, i.d. 0.53mm, 10 m length); column flow rate:5 ml min−1 argon; oven temperature: 35◦C (5 min) to 300◦C (5◦C min−1).

b Below LOD (<0.1 area×10−6).

as column carrier gas, make-up gas and shield gasfor the detector. The signals from the detector wererecorded by an integrator, HP 3395 (Hewlett–Packard,Willmington, DE, USA). The operating conditionsused are shown in Table 1.

For adsorbents BB and CC, the column usedwas a WCOT Ultimetal, with Cp-SimDist stationaryphase, 1.2mm thickness, i.d. 0.53 mm, 10 m length(Chrompac, Middleburg, The Netherlands). The in-strumental arrangement was the same as previouslyused with the exception the gas chromatograph em-ployed was a Varian 3400 (Varian, Mulgrave, Victoria,Australia). The operating conditions used for this col-umn are shown in Table 1. The determination of thetotal mercury content in samples fed to the adsorbentAA system was carried out using the vaporisation andtrapping technique.

3. Results and discussion

3.1. Adsorbent AA system

3.1.1. Total mercury by vaporisation-trap-AFSThe total mercury contents for the condensate feed

sample, the sample after the hydrogenolysis reactionR1 and the sample after the mercury trapping reac-tor R3 are shown in Table 2. The results show thatthe system was unable to remove all mercury speciesin the condensate stream. The higher value for theeluant from the reactor R1 may, in part, be due to

A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32 29

the pre-concentration effect of this sample passingthrough it for 48 h; resulting in carry over effects.

3.1.2. Determination of mercury species contentThe feed sample was found to contain four differ-

ent mercury species as shown in Table 3. Two of thesewere identified as DMM and DEM. From the reten-tion times the other two organomercury species wereattributed to alkyl mercury species between DMMand DEM (possibly methyl ethyl mercury (MEM) andmethyl propyl mercury (MPM))[18]. These species

Fig. 4. Chromatogram for mercury species before and after removal system (BB).

were not determined due to the absence of suitablestandards and do not correspond to other species de-termined using this column system i.e. Hg0, DPrM,DBM, MMC, EMC, DPM and PMC. The presence ofMEM in gas condensate has recently been identified[18].

The samples after the hydrogenolysis reactor alsonominally contain four species, Table 3. Three ofthese were identified as elemental mercury, DMM andDEM. The fourth species was also attributed to a di-alkyl mercury species (possibly MEM). These results

30 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

show that the elemental mercury present was a con-version product from another organomercury species.However, the conversion was found to be only about30% of the total mercury in the sample. One otherreason for the low elemental mercury could be thatelemental mercury was carried over to the gas stream(via R2) instead of remaining in the liquid stream.

Condensates can contain large fractions (sometimes>20 wt.%) of unsaturated hydrocarbons in the formof aromatics and olefinic compounds. One possibilityis that the hydrogenation process in R1, targets theunsaturated compounds rather than the organomer-cury itself. Hence, the efficiency of conversion of theorganomercury species to the elemental form is re-

Fig. 5. Example of mercury chromatograms before and after removal unit for adsorbent CC.

duced by these components. One important drawbackto this two-stage process is that the feed condensatehas to be hydrogenated. This changes the grade ofcondensate, in particular the aromatic content, whichdefines its quality. A downgraded product is, there-fore, created in order for partial mercury removal tobe achieved.

The mercury species found in the sample after themercury trapping reactor, Table 3, were DMM, DEMand the two other previously attributed alkyl mercuryspecies. No elemental mercury was detected. This in-dicated that the elemental mercury from the first re-action had been adsorbed by the second trapping re-actor. Examples of the chromatograms for the three

A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32 31

sample stages including the standard mercury speciesare shown in Fig. 3.

3.2. Adsorbent BB

Adsorbent BB was a carbon-based product impreg-nated with a sulphur compound. The species identi-fication in the feed samples and product (after trap-ping) samples are shown in Table 4. Three species oforganomercury (DMM, DEM and DBM) were spikedinto then-hexane (20, 30 and 20 ng ml−1, respectively,as Hg). There were no peaks detected in the productsamples after the reactor i.e. the removal of mercuryspecies spiked into the feed stream was close to 100%efficient. Examples of the chromatograms before andafter the reactor are shown in Fig. 4.

3.3. Adsorbent CC

The performance of this mercury species removingadsorbent is shown in Table 4. The concentration ofspecies inn-hexane was 20 ng ml−1 DMM, DEM andDBM each as Hg. From the results after the reactor,there was some reduction of the DMM peak (from20 ng ml−1 in feed to 9 ng ml−1 in sample after thereactor). The DEM species was detected in some ofthe collected samples with the concentration in therange 4–8 ng ml−1. However, the peak area of DBMwhile showing a reduction during the initial stagesi.e. after 2 h the DBM content had reduced by 83%,then began to increase with reaction time showing adramatic increase after 4 h. One possibility is that theDBM species was temporarily contained in the reactor(chromatographic effect) then re-released to the streamin a similar ‘manner’ to an eluant exiting a column.An alternative explanation to this is an effect due tocompetition for active sites. The overall performanceof the adsorbent in removing DMM, DEM and DBMspecies was 55, 80 and 22%, respectively. Examplesof the chromatograms are shown in Fig. 5.

4. Conclusion

All three pilot plant-tested mercury-removal sys-tems AA, BB and CC showed a reduction in themercury content of the final products. For the ad-

sorbent system AA, the hydrogenolysis reactor wasable to convert some of the organomercury presentin the gas condensate feed to its elemental form.However, the amount of elemental mercury measuredwas only about 30% of the total mercury content.Incomplete conversion of the organomercury speciesto mercury metal may be due to competition be-tween the organomercury species and the unsaturatedcompounds in the matrix during the hydrogenolysisreaction. One other reason could be that elementalmercury was carried over to the gas stream instead ofremaining in the liquid stream. The result from thefirst reactor (hydrogenolysis) also showed a highertotal mercury content compared with the feed con-densate. This could be due to a pre-concentrationeffect resulting in mercury carry-over by the liquidcondensate stream to the adsorber. The second reactor(mercury trapping) was able to adsorb elemental mer-cury present in the stream (the product from the firstreactor) but was unable to remove the organomercurycontent from the condensate stream.

For the single-stage adsorbent system BB, the effi-ciency in removing spikes of DMM, DEM and DBMfrom then-hexane hydrocarbon sample was very high.Removal of the species was 100% with no indicationof mercury present in the product. For adsorbent sys-tem CC, efficiency of removal for the different mer-cury species spiked into then-hexane was variable.Over 50% of DMM was removed using the adsorbent,while the DEM showed a range of 60–80% removal.The DBM was initially adsorbed efficiently (>80% at2 h) but was rapidly released back into the productstream after 4 h.

The above results indicated that speciation of themercury content is vital in the evaluation of an adsor-bent systems efficiency. The performance of an adsor-bent was dependent upon the type of adsorbent and, forthe AA and CC systems, the mercury species presentin the feed stream. The determination of total mercuryonly in feed and product samples in the system willnot give all the necessary information for an unbiasedevaluation to be made of the performance of an adsor-bent.

The results obtained, while indicative of the adsor-bents performance are no guarantee of their efficiencyusing a real sample under full process plant conditionsand should only be used as a guide. Those adsorbentswhich show a suitable potential for removing mercury

32 A. Shafawi et al. / Analytica Chimica Acta 415 (2000) 21–32

species, will require longer pilot plant studies usingvarious gas condensate grades as the feed over a longerplant monitoring period.

Acknowledgements

One of us (A.S.) is indebted to Petronas, Malaysia,for the sponsorship of this research work.

References

[1] M.D. Bigham, Field detection and implications of mercuryin natural gas, SPE Prod. Eng. (1990) 120–124.

[2] G.G. Haselden, The challenge of LNG in the 1980s, Mech.Eng. 103 (1981) 46–53.

[3] W.W. Bodle, A. Attari, R. Serauskas, Considerations formercury in LNG operations, in: Proceedings of the 6thInternational Conference on LNG, Paper 1, Inst. Gas Tech.,Kyoto, Japan, April 1980.

[4] J.E. Leeper, Mercury corrosion in LNG plant, EnergyProcessing, Canada, January–February 1981, pp. 46–51.

[5] P. Sarrazin, C.J. Cameron, Y. Barthel, M.E. Morrison, OilGas J. (1993) 86.

[6] C.J. Cameron, B. Didilion, D. Benayoun, M. Dorbon,Mercury: a trace contaminant in natural gas and natural gasliquids, in: Proceedings of the Eurogas 96, June 1996.

[7] C. Schickling, J.A.C. Broekaert, Appl. Organomet. Chem. 9(1995) 29.

[8] J.P. Snell, W. Frech, Y. Thommassen, Analyst 121 (1996)1055.

[9] W. Frech, D.C. Baxter, B. Bakke, J.P. Snell, Y. Thommassen,Anal. Comm. 33 (1996) 7H–9H.

[10] A. Shafawi, L. Ebdon, M.E. Foulkes, W.T. Corns,P.B. Stockwell, J. Anal. Atom. Spectrom., submitted forpublication.

[11] A. Shafawi, M.E. Foulkes, L. Ebdon, W.T. Corns, P.B.Stockwell, Analyst 124 (1999) 185.

[12] C.J. Cameron, D. Benayoun, M. Dorbon, S. Michelot,M. Robertson, How much mercury and what types ofmercury are in natural gas and associated condensate:how can mercury be eliminated, in: Proceedings of theSeminar on Catalysis, ITB Bandung, Indonesia, 24th October,1995.

[13] J.E. Leeper, Mercury — LNG problem, Hydroc. Proccs.(1980).

[14] M.S.M. Situmorang, M. Muchlis, Mercury problems in theArun LNG plant, in: Proceedings of the 8th InternationalConference on LNG, Session 2, Paper 5, Los Angeles, June1986.

[15] P. Frenken, T. Hubbers, in: Proceedings of the AmericanInstitute of Chemical Engineering Spring National Meeting,Paper 27a, Houston, April 1991.

[16] P. Sarrazin, C. Cameron, Y. Barthel, M.E. Morrison,in: Proceedings of the American Institute of ChemicalEngineering Spring Meeting, Paper 86a, New Orleans, March1992.

[17] M. Yamada, K. Sato, K. Sato, A. Furuta, Stud. Surf. Sci.Catal. 92 (1995) 433.

[18] H. Tao, T. Murakami, M. Tominaga, A. Miyazaki, J. Anal.Atom. Spectrom. 13 (1998) 1085.