Analyst

PAPER

Cite this: DOI: 10.1039/c7an02046e

Received 19th December 2017,Accepted 9th February 2018

DOI: 10.1039/c7an02046e

rsc.li/analyst

Quantification of isomerically summedhydrocarbon contributions to crude oil by carbonnumber, double bond equivalent, and aromaticityusing gas chromatography with tunable vacuumultraviolet ionization†

Jeremy A. Nowak, *a Robert J. Weberb and Allen H. Goldsteinb,c

The ability to structurally characterize and isomerically quantify crude oil hydrocarbons relevant to refined

fuels such as motor oil, diesel, and gasoline represents an extreme challenge for chromatographic and

mass spectrometric techniques. This work incorporates two-dimensional gas chromatography coupled to

a tunable vacuum ultraviolet soft photoionization source, the Chemical Dynamics Beamline 9.0.2 of the

Advanced Light Source at the Lawrence Berkeley National Laboratory, with a time-of-flight mass spectro-

meter (GC × GC-VUV-TOF) to directly characterize and isomerically sum the contributions of aromatic

and aliphatic species to hydrocarbon classes of four crude oils. When the VUV beam is tuned to 10.5 ±

0.2 eV, both aromatic and aliphatic crude oil hydrocarbons are ionized to reveal the complete chemical

abundance of C9–C30 hydrocarbons. When the VUV beam is tuned to 9.0 ± 0.2 eV only aromatic hydro-

carbons are ionized, allowing separation of the aliphatic and aromatic fractions of the crude oil hydro-

carbon chemical classes in an efficient manner while maintaining isomeric quantification. This technique

provides an effective tool to determine the isomerically summed aromatic and aliphatic hydrocarbon

compositions of crude oil, providing information that goes beyond typical GC × GC separations of the

most dominant hydrocarbon isomers.

Introduction

Crude oil is a complex mixture containing tens of thousandsof distinct organic species,1,2 dominated primarily by hydro-carbons (∼90%).3 The composition of a crude oil variesdepending on its origin and degree of weathering,4,5 as hydro-carbons become photooxidized, biodegraded, or subjected toabiotic environmental processes. While the range of crude oilhydrocarbons has been observed up to C80 with a double bondequivalent (NDBE)

6 chemical class up to 30,2 refined fuels suchas diesel, gasoline, and motor oil mostly consist of hydro-carbons below C30 in a NDBE range of 0–10.7–9 Hydrocarbons inthese refined fuels increase in aromatic and aliphatic com-pound complexity as the carbon number (NC) and the potential

number of isomers of the compound both increase, particu-larly as NC ≥ 9. This study therefore focuses on the characteriz-ation of those C9–C30 hydrocarbons of crude oils pertaining tothe aforementioned refined fuels that are amendable to separ-ation in our GC system.

There have been many chromatographic and mass spectro-metric advancements to characterize both crude oils and theirfuel derivatives as a function of hydrocarbon content. Liquidchromatographic (LC) separations distinguish resins, aromaticcompounds, saturated compounds, and acidic compounds incrude oil,10,11 but are insufficient for separating compounds ofincreased alkylation in a higher boiling range. LC also doesnot isomerically quantify compounds of a specific NC

9,10 andthe analysis processes are often laborious, time-consuming,and prone to sample loss.

Ultrahigh resolution Fourier transform ion cyclotron massspectrometry (FT-ICR MS) allows the identification of uniqueelemental compositions in crude oil between C10–C80 andNDBE = 0–30.1,2,4,5 FT-ICR MS analysis of crude oil is typicallyperformed using electrospray ionization (ESI), which targetspolar compounds,5,12 or atmospheric pressure photoionization(APPI), which targets nonpolar compounds and sulfur contain-

†Electronic supplementary information (ESI) available. See DOI: 10.1039/c7an02046e

aDepartment of Chemistry, University of California, Berkeley, California 94705, USA.

E-mail: jnowak01@berkeley.edubDepartment of Environmental Science, Policy and Management, University of

California, Berkeley, California 94705, USAcDepartment of Civil and Environmental Engineering, University of California,

Berkeley, California 94705, USA

This journal is © The Royal Society of Chemistry 2018 Analyst

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

.

View Article OnlineView Journal

ing species.13 While these soft ionization methods retain mole-cular ions necessary for molecular formula identification, theyhave difficulty ionizing pure hydrocarbons without significantion suppression or reduced ionization efficiency. Even whenpure hydrocarbons are ionized in FT-ICR MS,12 this techniqueis unable to offer structural characterizations of hydrocarbonsor differentiate hydrocarbon isomers in a single NDBE chemicalclass at a particular NC. Attempts at combining ion mobility/mass spectrometry (IM/MS) with FT-ICR MS have providedlimited isomeric separation of predominantly heteroatom con-taining crude oil components rather than separation of purehydrocarbons.14,15 Furthermore, FT-ICR MS relies on charac-terizing molecular formulas in terms of relative abundances ina mass spectrum rather than quantifying them based on massfractions of a particular NC or NDBE.

1,2,4,5

Gas chromatographic (GC) techniques are traditionallyused in gasoline, diesel, and motor oil analyses when there isno necessity to characterize compounds up to C80 and NDBE =30, as these heavier non-GC amendable species are rarelyfound in the aforementioned fuels.7–9 Traditional GC tech-niques are often limited to amendable hydrocarbons up toC30, due to the temperature limitations of columns and trans-fer lines (∼320 °C). GC typically couples a volatility separationof the hydrocarbons with a flame ionization detector (FID) orelectron ionization mass spectrometer (EI MS).16,17 Due to thecrude oil’s or fuel’s chemical complexity, much of the organicmass co-elutes as an unresolved complex mixture (UCM) in theresulting chromatogram,18 which complicates mass spectrainterpretation and compound identification. EI, a hard ioniza-tion technique (∼70 eV), imparts a large excess of energy toorganic molecules, resulting in fragmentation that complicatescompound identification on a molecular level.16–18 GC hasbeen combined with FT-ICR MS,19 but EI fragmented com-pounds >155 amu, making molecular identification of heaviercompounds difficult. GC has also been previously combinedwith atmospheric pressure chemical ionization (APCI) andFT-ICR MS,20 but the results illustrated limited isomeric separ-ation of hydrocarbons with the same elemental compositionand retention time as well as limited characterization of struc-tural contributions to a particular NDBE class of one NC.

Two-dimensional gas chromatography with electron ioniza-tion mass spectrometry (GC × GC-EI-MS) enhances compoundseparation and peak capacity by over 10-fold compared toGC-MS.21–23 The addition of a second polar column in GC ×GC-EI-MS allows aromatic compounds of the same volatility asaliphatic compounds to separate in the second dimension ofthe chromatogram, facilitating the acquisition of cleaner massspectra and more easily quantifiable peak areas. However,even the most advanced GC × GC separation has its limitationsin the characterization of oils and fuels. As the number ofcarbon atoms in molecules increases, the number of possiblehydrocarbon constitutional isomers becomes exponentiallylarger (>103 isomers of alkanes with NC = 15 and >105 isomersof alkanes with NC = 20),24 making it difficult to resolve theseindividual species even with high resolution GC × GC-EI-MS.While cycloalkanes and alkylbenzenes are more easily identifi-

able in the second dimension of a GC × GC chromatogramthan they are in a chromatogram using traditional GC-MS withEI, only the most concentrated species are quantifiable and itremains difficult to quantify the total range of isomers of ahydrocarbon class. It can therefore be challenging to use GC ×GC-EI-MS to achieve a comprehensive quantitative summary ofhydrocarbon distributions that takes into account the contri-bution of structural isomers based on volatility, polarity, andchemical structure.

Further complicating advanced GC × GC separation is thefact that at NDBE ≥ 4, both pure aromatic, pure aliphatic, andheavily alkylated aromatic hydrocarbons are present with thesame chemical formulas in crude oil and fuels. It becomesincreasingly difficult to separate and quantify how muchmaterial is aliphatic versus aromatic as NC and NDBE increasedue to the aforementioned exponential increase in isomersand the similar polarities of heavier aliphatic compounds andheavily alkylated aromatic compounds.25

In order to quantify both the isomeric hydrocarbon contentand the aromatic distributions of crude oil hydrocarbons rele-vant to refined fuels, this proof-of-concept study incorporatesGC × GC coupled with tunable synchrotron vacuum ultravioletphotoionization and time-of-flight mass spectrometry (GC ×GC-VUV-TOF).25,26 The VUV beam at the Advanced LightSource at the Lawrence Berkeley National Laboratory providesa soft ionization source (8–30 eV, less than the 70 eV ionizationenergy of typical “hard” electron ionization) with high photonflux (∼1016 photons per s)25 to reduce hydrocarbon fragmenta-tion and enhance the signal of the molecular ion compared tothat of EI. This technique allows complete classes of GC-amendable hydrocarbons to be integrated, summed, andcharacterized based on molecular weight, NC, degree ofbranching, and NDBE,

25,26 which is represented for only hydro-carbons in this study by

NDBEðCxHyÞ ¼ X � Y=2þ 1: ð1Þ

Unlike other soft photoionization methods combined withGC that rely on a rare gas excimer lamp to ionize petroleumalkanes at only one ionization energy,27,28 the VUV beam inthis study can be tuned to different photoionization energies.While previous studies have also incorporated tunable syn-chrotron VUV photoionization mass spectrometry to analyzeheavy oils and aromatic crude oil hydrocarbons,29–31 thoseinvestigations did not combine chromatographic resolutionwith soft ionization to isomerically sum both aliphatic andaromatic hydrocarbon fractions of those C9–C30 compoundsrelevant to fuels. This proof-of-concept investigation combinesGC × GC-MS with tunable synchrotron VUV photoionizationenergies to determine the NC dependent and isomericallysummed aromatic and aliphatic fractions, as a function ofmass loading, of GC-amendable hydrocarbons of four crudeoils from different locations. VUV radiation at 10.5 eV waschosen as the ionization energy to ionize all aliphatic and aro-matic hydrocarbons, thereby simultaneously maximizing thesignal from the molecular ion needed for quantification and

Paper Analyst

Analyst This journal is © The Royal Society of Chemistry 2018

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

minimizing signal interferents from fragment ions. Ionizationenergies greater than 10.5 eV result in greater compound frag-mentation and reduced molecular ion signal, making it moredifficult to accurately and isomerically sum both the aromaticand aliphatic hydrocarbon fractions. VUV radiation at 9.0 eVionizes aromatic species without significantly ionizing ali-phatic hydrocarbons,32 which allows the molecular ion signalof aromatic hydrocarbons to be maximized while minimizingboth fragmentation of aromatic hydrocarbons and interferingions from aliphatic compounds. Aromatic fractions, represent-ing those compounds that contain at least one aromatic func-tional group, of isomerically summed hydrocarbon molecularformulas can then be determined based on the response ratiosof hydrocarbon signals obtained at 9.0 eV versus 10.5 eVionization energy (R9.0/10.5). This tunable soft ionizationmethod allows immediate quantification, isomeric summar-ization, and aromatic characterization of those crude oil hydro-carbons relevant to refined fuels. The ability to efficientlyquantify and distinguish isomerically summed aliphatic fromaromatic fractions of one hydrocarbon mass and NDBE isuseful to characterize crude oil hydrocarbons for environ-mental modeling, fingerprinting of the hydrocarbon com-ponent of fuels for industrial applications, examining environ-mental weathering pathways, and understanding the impactsof oil spills in the environment.34

Experimental sectionOil samples

Crude oil samples were obtained from four sources: the NorthSea, the Gulf of Mexico, Texas, and Azerbaijan. The oils weredissolved in chloroform (HPLC grade, Sigma-Aldrich, St Louis,MO, USA) prior to analysis. Densities were determined byweighing a known volume of the original crude oil (MettlerAB104 analytical balance, Toledo, Columbus, OH, USA). APIgravities (in degrees) of the oils were calculated using theequation below, where SG is the specific gravity of crude oil:

API Gravity ¼ 141:5=SG-131:5: ð2Þ

GC × GC-VUV-TOF

Oil dissolved in chloroform was directly injected, without priorfractionation or chromatographic separation, via a septumlesshead into a liquid nitrogen cooled inlet (−40 °C) with a fusedsilica liner (CIS4, Gerstel Inc, Linthicum, MD, USA). Thesample was transferred to the gas chromatograph (GC, Agilent7890, Santa Clara, CA, USA) by heating the liner to 320 °C at10 °C s−1. Analytes were separated based on volatility using anonpolar column (60 m × 0.25 mm × 0.25 μm Rxi-5Sil-MS;Restek, Bellefonte, PA, USA) followed by polarity using amedium-polarity second dimension column (1 m × 0.25 mm ×0.25 μm Rtx-200MS; Restek, Bellefonte, PA, USA), with ahelium gas flow rate of 2 mL min−1. A dual-stage thermalmodulator (Zoex, Houston, TX, USA) consisting of a guardcolumn (1 m × 0.25 mm Rxi; Restek, Bellefonte, PA, USA) cryo-

genically focused the first column effluent and rapidly heatedthe analyte every 2.4 s via a 400 ms hot jet pulse into thesecond column. The GC program began at 40 °C, ramped at3.5 °C min−1 to 320 °C, and was held isothermal for a final10 minutes. These temperatures provide an upper limit ofamendable hydrocarbons of C30. The second column wasplaced in a secondary oven maintained 15 °C above the mainoven temperature. This system created an orthogonal separ-ation of the analyte based on volatility in the 1st-dimensionand polarity in the 2nd-dimension.

The analyte from the second column was transferred fromthe GC to a 170 °C ion source via a 270 °C transfer line. Thesetemperatures reduce the thermal energy imparted to thehydrocarbons and subsequently reduce compound fragmenta-tion.25,26 Analytes were ionized with 10.5 ± 0.2 eV or 9.0 ± 0.2 eVVUV photoionization with a photon flux of ∼1016 photons pers and detected using a time-of-flight mass spectrometer (TOF,Tofwerk, Thun, Switzerland) with m/Δm50% ≈ 4000. The VUVbeam was from the Chemical Dynamics Beamline 9.0.2 of theAdvanced Light Source at the Lawrence Berkeley NationalLaboratory. Mass spectra were collected at 100 Hz and dataprocessing was performed using custom code written in Igor6.3.7 (Wavemetrics).26 To correct for thermal transfer efficiencyof high- and low-volatility analytes through the GC column, aseries of perdeuterated alkanes (Sigma-Aldrich, St Louis, MO,USA) from C10–C30 were used as an internal standard in 10.5eV analyses to establish a relationship between transferefficiency and retention time. Perdeuterated polycyclic aro-matic hydrocarbons (PAHs) (Sigma-Aldrich, St Louis, MO, USA)were used to correct for thermal transfer efficiency in the 9.0eV analyses (when alkane ionization is negligible).

Data visualization for isomeric summarization

Fig. S-1a† illustrates a chromatogram of molecular ion massversus first dimension retention time of the Gulf of Mexico oilobtained using GC × GC-VUV-TOF at 10.5 eV ionization energy.In this figure the second-dimension separation is not visible.Increasing molecular mass is observed as the GC retentiontime increases, as is to be expected with heavier compounds ofdecreasing volatility. Fig. S-1b† illustrates the mass spectrumas a function of chromatographic retention time for theisomers of hydrocarbons with 20 carbon atoms in the Gulf ofMexico oil, showing their classification by mass and NDBE

under 10.5 eV VUV ionization. NDBE = 0 represents alkanes andNDBE = 6 represents a mixture of hexacycloalkanes and aro-matic species. PAHs of NDBE ≥ 7 were also observed, but arenot shown in Fig. S-1b† for simplicity. The volatility separationbased on retention time successfully distinguishes aliphaticcompounds and PAHs of equal molecular mass. Additionally,hydrocarbon volatility of NDBE = 0 compounds increases withbranching of the straight-chain alkane, therefore branchedisomers elute earlier in the chromatogram than the straight-chain hydrocarbon.7 The enhancement of the molecular ionunder VUV ionization, combined with chromatographic separ-ation by volatility, distinguishes the branched isomers fromthe straight-chain alkane of the same NC. Fig. S-1c† demon-

Analyst Paper

This journal is © The Royal Society of Chemistry 2018 Analyst

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

strates this concept by showing the single ion chromatogram(SIC) of pentadecane (m/Q = 212, C15H32) of the Gulf of Mexicooil, where the straight-chain alkane and its branched isomersare clearly distinguished.

In addition to retaining molecular ions necessary for hydro-carbon classification and quantification, Fig. S-2† illustratesthe advantages of GC × GC-VUV-TOF over traditional GC ×GC-EI-MS. Fig. S-2a and S-2b† illustrate the complete GC ×GC-EI-TOF chromatogram at 70 eV ionization energy and thecomplete GC × GC-VUV-TOF chromatogram at 10.5 eV ioniza-tion energy, respectively, of mass versus first dimension reten-tion time of the Gulf of Mexico oil. While the isomers ofheavier molecular weight and lower volatility compounds areobservable and quantifiable at 10.5 eV, they are not quantifi-able at 70 eV. Fragmentation of hydrocarbon isomers is alsoincreased in the chromatogram at 70 eV, as illustrated by theincrease in signal between m/Q = 100–200. Fig. S-2c† furtherillustrates the advantages of GC × GC-VUV-TOF over traditionalGC × GC-EI-MS by depicting m/Q = 326 (C24H38 NDBE = 6)hydrocarbon isomers using the two techniques. In the GC ×GC-EI-MS total ion chromatogram in Fig. S-2c,† only a fewmost concentrated isomers are identifiable as individual peakscorresponding to peaks in the single ion GC × GC-VUV-TOFchromatogram of m/Q = 326. Most of the signal for C24H38 isfrom the extremely large number of unresolved isomerseluting between 55 and 70 minutes, which are not separableusing GC × GC-EI-MS. It is extremely difficult to separate orresolve the numerous isomers present at much lower concen-trations (>105 isomers could be present with NC = 20).24 It isdifficult, time consuming, and perhaps impossible to isomeri-cally separate and quantify all of these hydrocarbons,especially as NC further increases, to achieve complete chemi-cal characterization of a compound class. GC × GC-VUV-TOFrepresents these C24H38 compounds in Fig. S-2c† as a singlechromatographic trace using the molecular ion. By integratingacross the trace and summing the signal of the complete rangeof isomers of a particular hydrocarbon molecular ion, weachieve the ability to quantify the complete mass of theseisomers. This methodology extends across all GC-amendablemolecular weights and compound classes, allowing us to iso-merically sum compounds more easily and to a greater analyti-

cal extent than even in the most advanced GC × GC-EI-MSseparation.

Differences in chromatograms at 9.0 eV and 10.5 eV

All aliphatic and aromatic compounds are ionized using GC ×GC-VUV-TOF with 10.5 eV VUV energy to reveal the chemicalcomposition of crude oil hydrocarbons from C9–C30 with NDBE

= 0–10. Under 9.0 eV ionization energy, only the aromaticspecies are ionized. Fig. 1 shows the mass spectrum as a func-tion of chromatographic retention time for the isomers ofhydrocarbons with 20 carbons atoms (C20) in the Gulf ofMexico oil, showing their classification by mass and NDBE

under 10.5 and 9.0 eV VUV ionization (1a and 1b, respectively).All NDBE = 0–10 hydrocarbons are ionized in the 10.5 eV chro-matogram, while only NDBE = 4–6 aromatic compounds andPAHs (NDBE = 7–10, not shown here for simplicity) are ionizedin the 9.0 eV chromatogram. There is some detection of ali-phatic compounds at 9.0 eV, but the ionization efficiency is<5% of that at 10.5 eV and is therefore considered negligible inour analyses. Fig. S-3† shows two-dimensional chromatogramsof a known standard hydrocarbon mixture to further illustratethe use of tunable VUV to selectively ionize hydrocarbonsdepending on molecular structure.

Calculating NC-dependent aromatic fractions of NDBE classes

In this proof-of-concept study, aromatic fraction is defined asthe fraction of isomers of a hydrocarbon molecular formula,based on NC and NDBE, that contain an aromatic functionalgroup. An aromatic fraction of 1 represents isomericallysummed species that contain no aliphatic character, such asmethyl naphthalenes or anthracene. We assumed that com-pounds of NDBE = 4 with fewer than 14 carbon atoms areentirely aromatic (aromatic fraction = 1). As the aromatic frac-tion decreases from 1, aliphatic contributions to an isomeri-cally summed compound increase in the form of alkyl sidechains or pure cycloalkanes. Isomers of a hydrocarbon mole-cular formula with an aromatic fraction <1 therefore canconsist of pure aromatic compounds, aliphatic compounds,mixtures of compounds containing fused aromatic and cyclicrings, etc. An aromatic fraction of 0 represents an isomericallysummed molecular formula that contains no aromaticity, and

Fig. 1 Mass spectrum as a function of chromatographic retention time for the isomers of hydrocarbons with 20 carbon atoms (C20) in the Gulf ofMexico Oil, showing their classification by mass and corresponding double bond equivalents (NDBE) under (a) 10.5 and (b) 9.0 eV VUV ionization.

Paper Analyst

Analyst This journal is © The Royal Society of Chemistry 2018

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

therefore consists entirely of alkanes, including branchedalkanes and cycloalkanes.

The response ratio of hydrocarbon signals obtained by inte-grating across a molecular formula at 9.0 eV versus 10.5 eVionization energy (R9.0/10.5) is indicative of the aromatic frac-tion in a specific NDBE as a function of NC. It should be notedthat R9.0/10.5 ≠ 1 even if all isomers of a molecular formula arecompletely aromatic. For instance, C10H14 represents isomersof alkylated benzenes with ten carbon atoms. While a combi-nation of double bonded cyclic compounds could theoreticallypossess a chemical formula of C10H14, the likelihood offinding such compounds in crude oils or refined fuels is low,as they do not generally contain alkenes.7,35 R9.0/10.5 of theseC10H14 alkylated benzenes is still less than 1 due to thedecreased photoionization efficiencies of aromatic hydro-carbons at 9.0 versus 10.5 eV ionization energies.33 As thelength of the alkyl side chain group in heavier NDBE = 4–6 com-pounds increases, there is an increasing trend in ionizationenergy that leads to a lower response at 9.0 eV and thusa lower R9.0/10.5.

32,33 Therefore, there is a NC dependence inR9.0/10.5 for alkylated aromatic compounds.

To calibrate for R9.0/10.5 of the hydrocarbons as a functionof NC and NDBE in the oils, authentic standards (Sigma-Aldrich, St Louis, MO, USA), including alkanes, branchedalkanes, cycloalkanes, alkylbenzenes, hopanes, steranes,PAHs, and alkylated PAHs were analyzed at 9.0 eV and 10.5 eV.The ratio of these hydrocarbon signals is referred to asRstandard. Fig. 2 shows R9.0/10.5 for a standard mixture of hydro-carbons consisting of the aforementioned species.Compounds with NDBE < 4 in this standard mixture are notaromatic and have a R9.0/10.5 < 0.05, corroborating the fact thatthe photoionization probability of aliphatic compounds isextremely low at 9.0 eV. R9.0/10.5 for alkylbenzenes decreaseswith increasing NC and becomes indistinguishable fromR9.0/10.5 for tetracyclic steranes above C25. PAHs have a R9.0/10.5of 0.45–0.50 independent of molecular weight. To estimate the

response ratios between 9.0 eV and 10.5 eV energies for purelyaliphatic components of NDBE = 4, 5, and 6 (Raliphatic), weextrapolated the trend in response ratios observed in NDBE = 0,1, 2, and 3, which cannot include aromatic rings. Using the NC

dependent values for the aromatic and aliphatic hydrocarbonsignal responses, the aromatic fraction of an isomericallysummed molecular formula is calculated as

Aromatic Fraction ¼ ½R9:0=10:5 � Raliphatic�=ðRstandardÞ ð3Þwhen NC in an isomerically summed NDBE = 4 molecularformula exceeds 25, the aromatic hydrocarbon isomers have asimilar R9.0/10.5 to that of tetracyclic steranes and therefore thearomatic fraction cannot be discerned by this method.

Results and discussionCrude oil chemical composition

The calculated densities and API gravities of the four crudeoils used in this study are given in Table 1. Based on the APIgravities, the Azerbaijan and Texas oils are classified as lightand the Gulf of Mexico and North Sea oils are medium-heavy.All four oils were initially analyzed using GC × GC-VUV-TOFwith 10.5 eV photoionization energy. Fig. 3 illustrates the dis-tribution of each oil’s GC-amendable isomerically summedhydrocarbon mass fractions (measured in milligram per kilo-gram injected oil) as a function of both NC and NDBE.∼60–65% of the measured mass fractions of the hydrocarbonsin the Texas and Azerbaijan oils are between the C9–C19 hydro-carbons. The mass fraction distributions of the hydrocarbonsin the North Sea and Gulf of Mexico oils are similar, with∼50% of the mass fractions split between C9–C19 and C20–C30

species. The Texas and Azerbaijan oils contain ∼70% of thetotal measured hydrocarbon mass fraction within completelyaliphatic compound classes, NDBE = 0–3. Only ∼60% of thetotal measured hydrocarbon mass fractions of the North Seaand Gulf of Mexico oils belongs to NDBE = 0–3.

GC × GC-VUV-TOF at 10.5 eV ionization energy also providesseparation of crude oil alkanes based on their degree ofbranching. Alkanes can be classified by retention time basedon the degree of branching BX, where x represents the numberof alkyl branches from methyl (B1) to hexamethyl (B6). Forsimplicity, each branch is assumed to represent a methylgroup, rather than ethyl, propyl, etc. Fig. S-4† illustrates theisomerically summed mass fractions of the branched alkanesin each of the oils as a function of NC and degree of branching

Table 1 Densities and API gravities of the Gulf of Mexico, North Sea,Azerbaijan, and Texas oils

Density (g cm−3) API gravity (°)

Gulf of Mexico oil ∼0.92 ∼22.0North sea oil ∼0.91 ∼22.5Azerbaijan oil ∼0.85 ∼35.0Texas oil ∼0.83 ∼40.0

Fig. 2 Ratio of hydrocarbon signal responses at 9.0 and 10.5 eV ioniza-tion energies (R9.0/10.5) for a standard mixture of hydrocarbons, includingalkanes, alkenes, cycloalkanes, ketones, hopanes, steranes, alkylben-zenes, and polycyclic aromatic hydrocarbons (PAHs).

Analyst Paper

This journal is © The Royal Society of Chemistry 2018 Analyst

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

(B1–B6). B1 branched alkanes dominate the branched alkanesin the Texas oil. The branched signatures of both the Gulf ofMexico and North Sea oils exhibit similar mass distributions,corroborating the similarities in the hydrocarbon compo-sitions of these two oils seen in Fig. 1. The Azerbaijan oil has asimilar branched alkane distribution to that of the North Seaand Gulf of Mexico oils, but the Azerbaijan oil contains sub-stantially higher mass fractions of B4–B6 branched alkanesbetween C19–C30.

Response ratios at 9.0 eV and 10.5 eV ionization energies

Fig. S-5, S-6, and S-7† display overlayed SICs with arbitraryunits of intensity at 9.0 eV and 10.5 eV ionization energies forC18H30 (NDBE = 4), C20H32 (NDBE = 5), and C20H30 (NDBE = 6),respectively, for all four oils. The signal peaks that are presentin the 10.5 eV SICs, but not in the 9.0 eV SICs, represent ali-phatic isomers that are not ionized at 9.0 eV. R9.0/10.5 of theintegrated signal of the summed isomers of NDBE = 4, 5, and6 hydrocarbons was calculated for each oil and plotted as afunction of NC in Fig. S-8.† R9.0/10.5 decreases with increasingmolecular weight independent of an oil’s source, indicatingthat aliphatic contributions in the form of alkyl groups, cyclicrings, etc. to these isomerically summed compound classesincrease with NC.

As previously discussed and illustrated, traditional GC ×GC-EI-MS separates aliphatic and aromatic compounds basedon both volatility and polarity, but the exponential increase inthe number of constitutional isomers as NC increases makes

quantification of the whole chemical class by NC and NDBE

challenging and time-consuming. Only concentrated com-pounds are identifiable and quantifiable in GC × GC space asNC increases, therefore it becomes increasingly difficult to iso-merically sum and distinguish aliphatic compounds from aro-matic compounds as molecular weight increases. In order totest our structural separation method based on ionizationenergies, Table S-1† displays the ratio of aromatic componentsto the total aromatic plus aliphatic components of nine iso-merically summed hydrocarbon molecular formulas in theNorth Sea oil, spanning a range of molecular weights. Thisratio was calculated using the second-dimension separation ofGC × GC-EI-MS and tunable GC × GC-VUV-TOF. As NC growsfrom NC = 16 to NC = 24 in NDBE = 4, 5, and 6, it becomesdifficult to separate and quantify isomers of a given hydro-carbon. The calculated ratios using the GC × GC-EI-MS separ-ation differ from those calculated using tunable GC ×GC-VUV-TOF by ∼35–65% as NC increases. This comparison ofexperimental techniques illustrates the increased fraction ofisomers which are not measured using GC × GC-EI-MS as theoil’s chemical complexity increases. The mass closure affordedby summing the isomers of hydrocarbons of higher molecularweights, as well as the structural separations afforded by thetunable ionization energies, present tunable GC ×GC-VUV-TOF as an analytical tool that can efficiently andquantitatively characterize sums of hydrocarbon isomersfound in crude oil to a greater extent than GC × GC-EI-MS sep-aration can alone.

Fig. 3 Distribution of GC-amendable isomerically summed hydrocarbon mass fractions as a function of both the number of carbon atoms anddouble bond equivalents (NDBE) for oils from (a) Azerbaijan, (b) North Sea, (c) Texas, and (d) Gulf of Mexico.

Paper Analyst

Analyst This journal is © The Royal Society of Chemistry 2018

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

NDBE = 7–10 response ratios

At 10.5 eV, the SICs of molecular weights pertaining to NDBE =8 hydrocarbons contain fragments of larger aliphatic hydro-carbons, making it difficult to discern the total pure NDBE =8 hydrocarbon content in the GC × GC-VUV-TOF chromato-gram. Fig. S-9† illustrates this concept using the SIC of m/Q =196 (C15H16 NDBE = 8) at 10.5 eV. Portions of the chromato-graphic trace align with retention times of heavier straight-chain and branched alkanes. The C15H16 molecular mass alsocorresponds with fragment ions from alkanes such as C17H36

and C20H42, following the loss of hydrogens and alkyl groups.Therefore at 10.5 eV, the chromatographic trace of C15H16

includes signal from C15H16 isomers and aliphatic fragmentsof higher molecular weight species. At 9.0 eV, these aliphaticcompounds are not ionized and the interference from frag-ment ions is negligible. While the 9.0 eV data can detect thetotal aromatic contribution to a NDBE = 8 molecular formula,

an aromatic fraction of this isomerically summed molecularformula cannot be calculated due to the interfering fragmentions in the 10.5 eV signal.

Fig. 4b further illustrates this concept by overlaying theGulf of Mexico oil’s C15H16 (NDBE = 8) SICs (with arbitraryunits of intensity) at both 10.5 eV and 9.0 eV ionizationenergies. Branched aliphatic hydrocarbons fragment morethan straight-chain aliphatic compounds, thus the largerfragments in the 10.5 eV trace in Fig. 4b correspond to frag-ments of branched compounds of a higher molecular weightwhose retention times overlap with those of the lower mole-cular weight C15H16 isomers. At 9.0 eV, the aliphatic com-pounds are not ionized and the NDBE = 8 SIC consists purelyof aromatic hydrocarbons with no observable interferencesfrom fragmentation. This observation was consistent in allmolecular formulas pertaining to NDBE = 8 hydrocarbons inall four oils.

Fig. 4 Single ion chromatograms (SICs) with arbitrary units of intensity (A.U.) of Gulf of Mexico oil for (a) C15H18 (NDBE = 7), (b) C15H16 (NDBE = 8), (c)C15H14 (NDBE = 9), and (d) C15H12 (NDBE = 10) at both 10.5 and 9.0 eV ionization energies. The C15H16 NDBE = 8 SIC at 10.5 eV includes signal fromC15H16 isomers and aliphatic fragments of higher molecular weight species (e.g. C17H36, C19H40, C20H42) with overlapping retention times. The NDBE

= 7, 9, and 10 SICs at 10.5 eV contain no observable interferences from fragmentation.

Analyst Paper

This journal is © The Royal Society of Chemistry 2018 Analyst

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

Fig. 4a, c, and d display the overlapping 9.0 eV and 10.5 eVSICs (with arbitrary units of intensity) for C15H18 (NDBE = 7),C15H14 (NDBE = 9), and C15H12 (NDBE = 10) of the Gulf ofMexico oil. Unlike in the SIC of C15H16 at 10.5 eV, there wereno observable interferences from fragmentation of heavier ali-phatic compounds in these PAH SICs at 10.5 eV and R9.0/10.5values could be estimated. This observation was consistent inall molecular formulas pertaining to NDBE = 7, 9, and 10 com-pounds in all four oils. Fig. S-10† shows R9.0/10.5 of the inte-grated signal of the summed isomers of NDBE = 7, 9, and 10compounds for all four oils as a function of NC. All of the oilshave a R9.0/10.5 of ∼0.4–0.6 in the aforementioned compoundclasses independent of NC. There is no decrease in R9.0/10.5 aswas observed in the NDBE = 4, 5, and 6 R9.0/10.5, implying thatNDBE = 7, 9, and 10 hydrocarbons in these oils consist of purePAHs with minimal aliphatic isomers.

Aromatic fraction calculations

The measured R9.0/10.5 of NDBE = 4, 5, and 6 hydrocarbons,along with Raliphatic and Rstandard, is used to calculate the NC-dependent isomerically summed aromatic fraction of eachNDBE in each oil using eqn (3).

Fig. 5a–c present the isomerically summed aromatic frac-tions of NDBE = 4, 5, and 6 hydrocarbons of each oil as a func-tion of NC. The Gulf of Mexico oil contains the largest fractionof isomerically summed aromatic hydrocarbons of the fouroils used in this study across all NC in NDBE = 4–6. This oil con-tains ∼1.5–2.0 times higher fractions of isomerically summedaromatic material across the NDBE = 4, 5, and 6 compoundclasses when compared to those of the other three oils. TheAzerbaijan oil has the lowest isomerically summed aromaticfraction in the NDBE = 4 class, while the Texas oil has a∼1.5–2.0 times higher isomerically summed aromatic fractionthan that of the North Sea oil in the C15–C25 range of NDBE = 4.The North Sea, Texas, and Azerbaijan oils all possess similarisomerically summed aromatic fractions at NDBE = 5 across allNC. At NDBE = 6, the isomerically summed aromatic fractionfrom C15–C25 in the North Sea oil is ∼1.3–1.6 times that of theTexas and Azerbaijan oils, revealing that the North Sea oil isstructurally similar to the Gulf of Mexico oil in the NDBE = 6class and consists of more aromatic material than the Texasand Azerbaijan oils in the NDBE = 6 class.

This study shows that as molecular weight increases inNDBE = 4–6, the fraction of isomers of a hydrocarbon molecularformula that are purely aromatic decreases. This patternimplies that hydrocarbon structures consisting of benzenes oralkylated benzenes decrease in concentration as NC growsacross NDBE = 4–6, and structures consisting of fused cyclicrings increase in concentration in the same scenario. It cantherefore be concluded from this study that in NDBE = 4–6, iso-merically summed heavier hydrocarbons are more likely toconsist of fused cyclic rings and cycloalkanes than pure alkyl-ated benzenes or a combination of aromatic functional groupsand cyclic rings. Therefore, aliphatic contributions to the NDBE

= 4–6 compound classes in all of these oils become more

prevalent as molecular weight increases across one isomeri-cally summed compound class.

Conclusions

Tunable soft VUV photoionization coupled to GC × GC-MSoffers a fast, efficient, and unique analytical technique tofurther characterize and quantify the chemical complexity ofcrude oil hydrocarbons in the molecular range relevant to

Fig. 5 Aromatic mass fractions as a function of carbon number for thefour oils, indicated by color, in three classes of double bond equivalents:NDBE = 4, 5, and 6 (a, b, and c, respectively).

Paper Analyst

Analyst This journal is © The Royal Society of Chemistry 2018

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

refined fuels such as gasoline, diesel, and motor oil beyondthat of traditional GC × GC-EI-MS separation. Soft ionization at10.5 eV allows the full range of isomeric hydrocarbons to besummed as a function of NC, NDBE chemical class, and theamount of branching. By altering the energy of the VUV beambetween 10.5 eV and 9.0 eV, we successfully ionize hydro-carbon isomers with NDBE≥4 into aliphatic and aromatic frac-tions as a function of NC while maintaining the ability to iso-merically sum mass fractions. This study provides a novelchromatographic and mass spectrometric tool to constrainand strengthen existing models of crude oil hydrocarboncharacterization with applications to oil spills, microbial trans-formations, and environmental impacts. Tunable soft ioniza-tion is therefore a powerful analytical tool which can now beapplied to isomerically sum and further characterize crude oilhydrocarbons beyond that of traditional compound class andcarbon number labels.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge the Energy Biosciences Institute (EBI) forfunding this work under project 86217. We also acknowledgeKevin Wilson and Bruce Rude for their assistance at theChemical Dynamics Beamline 9.0.2 at the Advanced LightSource, Coty Jen for GC × GC-VUV-TOF instrumental setup andtransport, and Professor John Coates for access to the oils andadvice.

References

1 A. G. Marshall and R. P. Rodgers, Petroleomics: the nextgrand challenge for chemical analysis, Acc. Chem. Res.,2004, 37(1), 53–59.

2 A. G. Marshall and R. P. Rodgers, Petroleomics: chemistryof the underworld, Proc. Natl. Acad. Sci. U. S. A., 2008,105(47), 18090–18095.

3 K. A. P. Colati, G. P. Dalmaschio, E. V. R. de Castro,A. O. Gomes, B. G. Vaz and W. Romão, Monitoring theliquid/liquid extraction of naphthenic acids in braziliancrude oil using electrospray ionization FT-ICR mass spec-trometry (ESI FT-ICR MS), Fuel, 2013, 108, 647–655.

4 H. Chen, A. Hou, Y. E. Corilo, Q. Lin, J. Lu,I. A. Mendelssohn, R. Zhang, R. P. Rodgers andA. M. McKenna, 4 years after the Deepwater Horizon Spill:Molecular transformation of Macondo well oil in Louisianasalt marsh sediments revealed by FT-ICR mass spec-trometry, Environ. Sci. Technol., 2016, 50(17), 9061–9069.

5 B. M. Ruddy, M. Huettel, J. E. Kostka, V. V. Lobodin,B. J. Bythell, A. M. McKenna, C. Aeppli, C. M. Reddy,R. K. Nelson, A. G. Marshall and R. P. Rodgers, Targeted

petroleomics: analytical investigation of Macondo well oiloxidation products from Pensacola Beach, Energy Fuels,2014, 28(6), 4043–4050.

6 F. W. McLafferty and F. Turecek, Interpretation of MassSpectra, University Science Books, Mill Valley, CA, 4th edn,1993.

7 D. R. Worton, G. Isaacman, D. R. Gentner, T. R. Dallmann,A. W. H. Chan, C. Ruehl, T. W. Kirchstetter, K. R. Wilson,R. A. Harley and A. H. Goldstein, Lubricating oil dominatesprimary organic aerosol emissions from motor vehicles,Environ. Sci. Technol., 2014, 48(7), 3698–3706.

8 R. C. Prince, R. P. Parkerton and C. Lee, The primaryaerobic biodegradation of gasoline hydrocarbons, Environ.Sci. Technol., 2007, 41, 3316–3321.

9 F. Adam, F. Bertoncini, D. Thiébaut, S. Esnault, D. Espinatand M. C. Hennion, Towards comprehensive hydrocarbonsanalysis of middle distillates by LC-GCxGC, J. Chromatogr.Sci., 2007, 45, 643–649.

10 A. Demirbas and O. Taylan, Removing of resins from crudeoils, Pet. Sci. Technol., 2016, 34, 771–777.

11 C. A. Islas-Flores, E. Buenrostro-Gonzalez and C. Lira-Galeana, Fractionation of Petroleum Resins by Normal andReverse Phase Liquid Chromatography, Fuel, 2006, 85(12),1842–1850.

12 V. V. Lobodin, P. Juyal, A. M. McKenna, R. P. Rodgers andA. G. Marshall, Tetramethylammonium hydroxide as areagent for complex mixture analysis by negative ion elec-trospray ionization mass spectrometry, Anal. Chem., 2013,85(16), 7803–7808.

13 A. M. McKenna, J. T. Williams, J. C. Putman, C. Aeppli,C. M. Reddy, D. L. Valentine, K. L. Lemkau,M. Y. Kellermann, J. J. Savory, N. K. Kaiser, A. G. Marhsalland R. P. Rodgers, Unprecedented ultrahigh resolutionFT-ICR mass spectrometry and parts-per-billion mass accu-racy enable direct characterization of nickel and vanadylporphyrins in petroleum from natural seeps, Energy Fuels,2014, 28(4), 2454–2464.

14 F. A. Fernandez-Lima, C. Becker, A. M. McKenna,R. P. Rodgers, A. G. Marshall and D. H. Russell, Petroleumcrude oil characterization by IMS-MS and FTICR MS, Anal.Chem., 2009, 81, 9441–9947.

15 M. Farenc, Y. E. Corilo, P. M. Lalli, E. Riches, R. P. Rodgers,C. Afonso and P. Giusti, Comparison of atmosphericpressure ionization for the analysis of heavy petroleumfractions with ion mobility-mass spectrometry, EnergyFuels, 2016, 30, 8896–8903.

16 Z. Wang and M. Fingas, Differentiation of the source ofspilled oil and monitoring of the oil weathering processusing gas chromatography-mass spectrometry,J. Chromatogr., A, 1995, 712(2), 321–343.

17 C. M. Reddy and J. G. Quinn, GC-MS analysis of total pet-roleum hydrocarbons and polycyclic aromatic hydro-carbons in seawater samples after the North Cape oil spill,Mar. Pollut. Bull., 1999, 38(2), 126–135.

18 H. K. White, L. Xu, P. Hartmann, J. G. Quinn andC. M. Reddy, Unresolved complex mixture (UCM) in coastal

Analyst Paper

This journal is © The Royal Society of Chemistry 2018 Analyst

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online

environments is derived from fossil sources, Environ. Sci.Technol., 2013, 47, 726–731.

19 J. E. Szulejko and T. Solouki, Potential AnalyticalApplications of Interfacing a GC to an FT-ICR MS:Fingerprinting complex sample matrixes, Anal. Chem.,2002, 74, 3434–3442.

20 M. P. Barrow, K. M. Peru and J. V. Headley, An addeddimension: GC atmospheric chemical ionization FTICR MSand the Athabasca oil sands, Anal. Chem., 2014, 86, 8281–8288.

21 Z. Y. Liu and J. B. Phillips, Comprehensive 2-dimensionalgas chromatography using an on column thermal modu-lator interface, J. Chromatogr. Sci., 1991, 29(6), 227–231.

22 R. K. Nelson, B. S. Kile, D. L. Plata, S. P. Sylva, L. Xu,C. M. Reddy, R. B. Gaines, G. S. Frysinger andS. E. Reichenbach, Tracking the weathering of an oil spillwith comprehensive two-dimensional gas chromatography,Environ. Forensics, 2006, 7(1), 33–44.

23 J. Gros, C. M. Reddy, C. Aeppli, R. K. Nelson,C. A. Carmichael and J. S. Arey, Resolving biodegradationpatterns of persistent saturated hydrocarbons in weatheredoil samples from the Deepwater Horizon disaster, Environ.Sci. Technol., 2014, 48(3), 1628–1637.

24 A. H. Goldstein and I. E. Galbally, Known and unexploredorganic constituents in the earth’s atmosphere, Environ.Sci. Technol., 2007, 41(5), 1514–1521.

25 G. Isaacman, K. Wilson, A. Chan, D. Worton, J. Kimmel,T. Nah, T. Hohaus, M. Gonin, J. H. Kroll, D. R. Worsnopand A. H. Goldstein, Improved Resolution of HydrocarbonStructures and Constitutional Isomers in ComplexMixtures Using Gas Chromatography-Vacuum Ultraviolet-Mass Spectrometry, Anal. Chem., 2012, 84(5), 2335–2342.

26 D. Worton, H. Zhang, G. Isaacman, A. Chan, K. Wilson andA. Goldstein, Comprehensive Chemical Characterization ofHydrocarbons in NIST Standard Reference Material 2779Gulf of Mexico Crude Oil, Environ. Sci. Technol., 2015,49(22), 13130–13138.

27 R. Geibler, M. Saraji-Bozorgzad, T. Streibel,E. Kaisersberger, T. Denner and R. Zimmermann,Investigation of different crude oils applying thermal ana-lysis/mass spectrometry with soft photoionization,J. Therm. Anal. Calorim., 2009, 96, 813–820.

28 S. Wohlfahrt, M. Fischer, M. Saraji-Bozorgzad,G. Matuschek, T. Streibel, E. Post, T. Denner andR. Zimmermann, Rapid comprehensive characterization ofcrude oils by thermogravimetry coupled to fast modulatedgas chromatography-single photon ionization time-of-flightmass spectrometry, Anal. Bioanal. Chem., 2013, 405, 7107–7116.

29 J. Chen, L. Jia, L. Zhao, X. Lu, W. Guo, J. Weng and F. Qi,Analysis of Petroleum Aromatics by Laser-Induced AcousticDesorption/Tunable Synchrotron Vacuum UltravioletPhotoionization Mass Spectrometry, Energy Fuels, 2013, 27,2010–2017.

30 W. Guo, Y. Bi, H. Guo, Y. Pan, F. Qi, W. Deng and H. Shan,Infrared laser desorption /vacuum ultraviolet photoioniza-tion mass spectrometry of petroleum saturates: a newexperimental approach for the analysis of heavy oils, RapidCommun. Mass Spectrom., 2008, 22, 4025–4028.

31 L. Jia, J. Weng, Z. Zhou, F. Qi, W. Guo, L. Zhao and J. Chen,Note: Laser-induced acoustic desorption/synchrotronvacuum ultraviolet photoionization mass spectrometry foranalysis of fragile compounds and heavy oils, Rev. Sci.Instrum., 2012, 83, 026105.

32 T. Adam and R. Zimmermann, Determination of singlephoton ionization cross sections for quantitative analysisof complex organic mixtures, Anal. Bioanal. Chem., 2007,389(6), 1941–1951.

33 NIST Chemistry WebBook, NIST Standard ReferenceDatabase Number 69 (http://webbook.nist.gov/).

34 G. T. Drozd, D. R. Worton, C. Aeppli, C. M. Reddy,H. Zhang, E. Variano and A. G. Goldstein, Modeling com-prehensive chemical composition of weathered oil follow-ing a marine spill to predict ozone and potential secondaryaerosol formation and constraint transport pathways,J. Geophys. Res.: Oceans, 2015, 120(11), 7300–7315.

35 D. Mao, H. Van De Weghe, R. Lookman, G. Vanermen,N. De Brucker and L. Diels, Resolving the unresolvedcomplex mixture in motor oils using high-performanceliquid chromatography followed by comprehensive two-dimensional gas chromatography, Fuel, 2009, 88(2), 312–318.

Paper Analyst

Analyst This journal is © The Royal Society of Chemistry 2018

Publ

ishe

d on

12

Febr

uary

201

8. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Ber

kele

y on

21/

02/2

018

17:5

7:06

. View Article Online