journal of chromatography a - xiamen university€¦ · monolithic columns [29], however, were...

7
Journal of Chromatography A, 1533 (2018) 136–142 Contents lists available at ScienceDirect Journal of Chromatography A jo ur nal ho me pag e: www.elsevier.com/locate/chroma Dual polyhedral oligomeric silsesquioxanes polymerization approach to mutually-mediated separation mechanisms of hybrid monolithic stationary and mobile phases towards small molecules Jiao Su a , Limin Yang a,, Qiuquan Wang a,b,a Department of Chemistry & the Key Laboratory of Spectrochemical Analysis and Instrumentation, College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, 361005, China b State Key Laboratory of Marine Environmental Science, Xiamen University, Xiamen, 361102, China a r t i c l e i n f o Article history: Received 3 October 2017 Received in revised form 27 November 2017 Accepted 12 December 2017 Available online 14 December 2017 Keywords: Hybrid monolithic stationary phase Polyhedral oligomeric silsesquioxane Separation mechanism Small molecule Nano-LC a b s t r a c t Hybrid monolithic stationary phase based HPLC is a typical example of practices in separation science. In this study, we developed a dual polyhedral oligomeric silsesquioxanes (POSS) polymerization approach to the preparation of a hybrid monolithic stationary phase of tri-porous structure and various surface chemistry. N-phenylaminopropyl-POSS (PA-POSS) and glycidyl-POSS (EP-POSS) were exemplified to demonstrate effective mutually-mediated separation mechanisms of the hybrid monolithic stationary phase and mobile phase towards diverse small molecules. PA-POSS and EP-POSS can be the monomer and/or crosslinker each other. They were polymerized via the epoxy-ring opening reaction to form the poly[(PA-POSS)-(EP-POSS)] (polyPOSS) monolithic stationary phase of 110.6/164.6 Å 3 micropore (as a cube/ball), 10 nm mesopore and 0.95 m macropore with the native siloxane cage and remaining phenyl/epoxy as well as chemically generated positive-chargeable tertiary phenylamine and hydrophilic hydroxyl groups. Such pore-structure and surface chemistry allow us to perform the effective separa- tion of targeted small molecules, such as alkylbenzenes and alkylbenzene ketones, nucleic acid bases and amino acids, as well as phenols and phenolic acids, under reversed-phase, HILIC and mixed mode (polarity, size-exclusion and hydrogen-bonding) by just changing the molar ratio of POSS-precursors, and the composition and pH of a mobile phase as well. We believe that the approach developed herein can be extended to fabricate other kinds of hybrid monolithic stationary phases that are suitable for the separation of biomacromolecules and chiral molecules when choosing the existed POSS and/or designing new POSS with the substituted pendant groups of different physicochemical properties. © 2017 Elsevier B.V. All rights reserved. 1. Introduction Today, monoliths are becoming ever more popular as chro- matographic stationary phases [1]. Their high permeability and thus fast separation of targeted analytes, without paying for the high back-pressure arising from a small bead-packed stationary phase, render them practical usefulness even in a conventional HPLC system. In general, the organic polymer-based monoliths that consist of interconnected microglobules [2] and the inor- Corresponding authors: Department of Chemistry & the Key Laboratory of Spectrochemical Analysis and Instrumentation, College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, 361005, China. E-mail addresses: [email protected] (L. Yang), [email protected] (Q. Wang). ganic silica-based monoliths of mesoporous skeleton-mediated bicontinuous porous structure [3] feature the rapid separation of large biopolymers [4–6] and small molecules [7], respectively. When a silanized organic and/or an organically modified siliceous monomer or crosslinker is used, the resulting organic-inorganic hybrid monolith combines their respective advantages, to a great extent, such as wider pH stability, enhanced structure and surface chemistry, demonstrating a superior ability for the faster separa- tion of both small and large molecules [8–11]. Among the various organosilicon compounds, polyhedral oligomeric silsesquioxane (POSS) is a three-dimensional cubic nanomaterial with thermally and chemically robust siloxane cage and eight pendant organic arms [12,13]. It is conceivable that the inherent nanoscale silox- ane cage (sub-nanometer Si···Si cage diagonal and nanometer diagonal distance of the pendant arms) can offer native microp- https://doi.org/10.1016/j.chroma.2017.12.033 0021-9673/© 2017 Elsevier B.V. All rights reserved.

Upload: others

Post on 21-Oct-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

  • Dts

    Ja

    Ub

    a

    ARR2AA

    KHPSSN

    1

    mthpHt

    SE

    (

    h0

    Journal of Chromatography A, 1533 (2018) 136–142

    Contents lists available at ScienceDirect

    Journal of Chromatography A

    jo ur nal ho me pag e: www.elsev ier .com/ locate /chroma

    ual polyhedral oligomeric silsesquioxanes polymerization approacho mutually-mediated separation mechanisms of hybrid monolithictationary and mobile phases towards small molecules

    iao Sua, Limin Yanga,∗, Qiuquan Wanga,b,∗

    Department of Chemistry & the Key Laboratory of Spectrochemical Analysis and Instrumentation, College of Chemistry and Chemical Engineering, Xiamenniversity, Xiamen, 361005, ChinaState Key Laboratory of Marine Environmental Science, Xiamen University, Xiamen, 361102, China

    r t i c l e i n f o

    rticle history:eceived 3 October 2017eceived in revised form7 November 2017ccepted 12 December 2017vailable online 14 December 2017

    eywords:ybrid monolithic stationary phaseolyhedral oligomeric silsesquioxaneeparation mechanismmall moleculeano-LC

    a b s t r a c t

    Hybrid monolithic stationary phase based HPLC is a typical example of practices in separation science. Inthis study, we developed a dual polyhedral oligomeric silsesquioxanes (POSS) polymerization approachto the preparation of a hybrid monolithic stationary phase of tri-porous structure and various surfacechemistry. N-phenylaminopropyl-POSS (PA-POSS) and glycidyl-POSS (EP-POSS) were exemplified todemonstrate effective mutually-mediated separation mechanisms of the hybrid monolithic stationaryphase and mobile phase towards diverse small molecules. PA-POSS and EP-POSS can be the monomerand/or crosslinker each other. They were polymerized via the epoxy-ring opening reaction to formthe poly[(PA-POSS)-(EP-POSS)] (polyPOSS) monolithic stationary phase of 110.6/164.6 Å3 micropore (asa cube/ball), 10 nm mesopore and 0.95 �m macropore with the native siloxane cage and remainingphenyl/epoxy as well as chemically generated positive-chargeable tertiary phenylamine and hydrophilichydroxyl groups. Such pore-structure and surface chemistry allow us to perform the effective separa-tion of targeted small molecules, such as alkylbenzenes and alkylbenzene ketones, nucleic acid basesand amino acids, as well as phenols and phenolic acids, under reversed-phase, HILIC and mixed mode

    (polarity, size-exclusion and hydrogen-bonding) by just changing the molar ratio of POSS-precursors,and the composition and pH of a mobile phase as well. We believe that the approach developed hereincan be extended to fabricate other kinds of hybrid monolithic stationary phases that are suitable for theseparation of biomacromolecules and chiral molecules when choosing the existed POSS and/or designingnew POSS with the substituted pendant groups of different physicochemical properties.

    © 2017 Elsevier B.V. All rights reserved.

    . Introduction

    Today, monoliths are becoming ever more popular as chro-atographic stationary phases [1]. Their high permeability and

    hus fast separation of targeted analytes, without paying for theigh back-pressure arising from a small bead-packed stationary

    hase, render them practical usefulness even in a conventionalPLC system. In general, the organic polymer-based monoliths

    hat consist of interconnected microglobules [2] and the inor-

    ∗ Corresponding authors: Department of Chemistry & the Key Laboratory ofpectrochemical Analysis and Instrumentation, College of Chemistry and Chemicalngineering, Xiamen University, Xiamen, 361005, China.

    E-mail addresses: [email protected] (L. Yang), [email protected]. Wang).

    ttps://doi.org/10.1016/j.chroma.2017.12.033021-9673/© 2017 Elsevier B.V. All rights reserved.

    ganic silica-based monoliths of mesoporous skeleton-mediatedbicontinuous porous structure [3] feature the rapid separationof large biopolymers [4–6] and small molecules [7], respectively.When a silanized organic and/or an organically modified siliceousmonomer or crosslinker is used, the resulting organic-inorganichybrid monolith combines their respective advantages, to a greatextent, such as wider pH stability, enhanced structure and surfacechemistry, demonstrating a superior ability for the faster separa-tion of both small and large molecules [8–11]. Among the variousorganosilicon compounds, polyhedral oligomeric silsesquioxane(POSS) is a three-dimensional cubic nanomaterial with thermallyand chemically robust siloxane cage and eight pendant organic

    arms [12,13]. It is conceivable that the inherent nanoscale silox-ane cage (sub-nanometer Si···Si cage diagonal and nanometerdiagonal distance of the pendant arms) can offer native microp-

    https://doi.org/10.1016/j.chroma.2017.12.033http://www.sciencedirect.com/science/journal/00219673http://www.elsevier.com/locate/chromahttp://crossmark.crossref.org/dialog/?doi=10.1016/j.chroma.2017.12.033&domain=pdfmailto:[email protected]:[email protected]://doi.org/10.1016/j.chroma.2017.12.033

  • J. Su et al. / J. Chromatogr. A 1533 (2018) 136–142 137

    F prepa(

    oplaapApsdhothm,w

    ig. 1. Schematic diagram of pore distribution (a), selected POSS structure (b) andd).

    res, and the reactive pendant organic groups undertake furtherolymerization to form mesoporous skeletons in resulting mono-

    iths. Macropores can be also formed when suitable porogensre employed. Such POSS-based monoliths are expected to have

    tri-porous structure with multifunctional characteristics if theendant arms have different physicochemical properties (Fig. 1a).ctually, POSS-based hybrid monolithic capillary column was pre-ared via the free radical polymerization between the methacrylubstituted-POSS (POSS-MA) and N-(2-(methacryloyloxy)ethyl)-imethyloctadecylammonium bromide. Such a monolithic columnad fair mechanical and pH stability and column efficiency for notnly small molecules but also model proteins as well as the BSAryptic digested mixture [14]. In a similar way, various POSS-based

    ybrid monoliths appeared using POSS-MA crosslinker, and theonomers including hydrophilic neutral [15], hydrophobic alkyl-

    perfluoroalkyl- and phenyl-substituted compounds [16–19] asell as alkyl- or perfluorinated phenyl-substituted ionic liquids

    ration of the polyPOSS hybrid monolith (c) as well as the small molecules studied

    [20–22]. Other chemical reaction-induced polymerization, such asamine-epoxy, thiol-epoxy and thiol-ene reactions [23–28], werealso employed for initiating a progressive phase separation pro-cess in order to improve structure homogeneity of the synthesizedmonoliths with highly ordered 3D skeletal structure. All thesemonolithic columns [29], however, were prepared using one POSScrosslinker and a small organic monomer. We hypothesize thatdifferently chemical modified POSS may be the monomer and/orcrosslinker each other [30], and thus more regular porous struc-ture and plentiful surface chemistry can be expected. Herein, weselected two fully substituted POSS, N-phenylaminopropyl-POSS(PA-POSS) and glycidyl-POSS (EP-POSS) (Fig. 1b), to demonstratea dual POSS-polymerized hybrid monolithic approach for the first

    time towards the separation of small molecules via the mutually-mediated separation mechanisms by adjusting the molar ratio ofPOSS-precursors and the composition and pH of the mobile phaseused.

  • 1 ogr. A

    2

    2

    PUbtiaMhbahMguislC

    2

    caetpdGlspfit

    iawUdmBN(tt

    2

    cv((paw1bot

    38 J. Su et al. / J. Chromat

    . Experimental

    .1. Chemicals

    N-Phenylaminopropyl-POSS (PA-POSS) and glycidyl-POSS (EP-OSS) were purchased from hybrid plastics (Hattiesburg, MS,SA). The distances of adjacent Si-Si, planar diagonal Si-Si andody diagonal Si-Si are respectively 3.1 Å, 4.4 Å and 5.4 Å in bothhe POSSs. The length of Si-R is 10.4 Å in PA-POSS and 9.0 Ån EP-POSS (Fig. 1b, calculated from ChemBio3D Ultra 12.0). (3-minopropyl)trimethoxysilane (APTMS), polyethylene glycol (PEG,w = 20,000 Da), L-histidine, L-phenylalanine, L-tryptophane, 2,6-

    ydroxy benzoic acid, 3,4-hydroxy benzoic acid, 3,5-hydroxyenzoic acid, sorbic acid, p-hydroxy benzoic acid, benzoic acidnd alkylbenzene ketones were obtained from Aladdin (Shang-ai, China). Alkylbenzenes were purchased from Sigma (St. Louis,O, USA). Cytosine, uracil and thymine were obtained from San-

    on (Shanghai, China). HPLC grade acetonitrile (ACN) and methanolsed were obtained from Merck (Darmstadt, Germany). Water used

    n all the experiments was doubly distilled and purified by a Milli-Qystem (Millipore, Milford, MA, USA). Other reagents of at least ana-ytical grade used were bought from Sinopharm Chemical Reagento., Ltd (Shanghai, China).

    .2. Instruments

    Fused-silica capillary (75 �m i.d. and 375 �m o.d.) was pur-hased from Yongnian Refine Chromatography Ltd. (Hebei, China)nd used to prepare the monolithic capillary columns. The nano-LCxperiments were carried out on a Prominence Nano LC sys-em (Shimadzu, Japan) equipped by two Shimadzu LC-20AD Nanoumps, a Shimadzu CBM-20A system controller, a MU701 UV-VISetector with a 6 nL capillary fiber optic flow cell (Shimadzu-L, Japan) and a micro valve injector with a 4 nL inner sample

    oop (VICI, USA). The Shimadzu LC-solution chromatography work-tation was used for data acquisition. All the experiments wereerformed at room temperature (24 ◦C). All the mobile phases wereltered through a 0.22 �m membrane and degassed under sonica-ion before use.

    The morphology of the polyPOSS capillary columns was stud-ed using a Zeiss Sigma FE-SEM instrument (Zeiss, Germany). Theverage mesopore size and Brunauer-Emmett-Teller surface areaere determined on a Micromeritics Tristar 3020 (Norcross, GA,SA) through nitrogen adsorption/desorption. The macropore sizeistribution of the monolith synthesized was measured on a Pore-aster 60 mercury intrusion apparatus (Quantachrome, Boynton

    each, FL, USA). All the IR measurements were carried out on aicolet IR360 Spectrometer (Thermo Electron, USA). A Vario EL III

    Elementar, Germany) was used for elemental analysis. A SDT Q600hermo gravimetric analyzer (TA, USA) was used to characterizehermal stability of the monoliths.

    .3. Pretreatment of the capillary columns

    Before preparation of the polyPOSS hybrid monolithic capillaryolumns, inner surface of the capillary was firstly cleaned and acti-ated for effective attachment of the POSS skeleton using acetone30 min), H2O (30 min), 1 M NaOH (12 h), H2O (30 min), 1 M HCl12 h), H2O (30 min) and acetone (1 h) in sequence with a syringeump, and then dried under a nitrogen stream at room temper-ture overnight. Subsequently, a 50% APTMS/THF (V/V) solutionas pumped through the capillary at a flow rate of 3 �L/min for

    h. After both ends of the capillary were sealed with silicone rub-er septa, the capillary was submerged in an 80 ◦C water bathvernight. Finally, the capillary was rinsed by acetone to flush outhe unreacted residuals and dried by a nitrogen stream again. The

    1533 (2018) 136–142

    internal surface of such pretreated capillary was modified withamine groups for anchoring later formed polyPOSS hybrid mono-lith.

    2.4. Preparation of polyPOSS hybrid monolithic capillary columns

    Homogeneous prepolymerization solution consisting of PA-POSS, EP-POSS and porogens was manually injected into thepretreated capillary using a syringe, and then both ends of the filledcapillary were sealed with silicone rubber septa. The filled capillarywas incubated at 130 ◦C for 18 h in a muffle furnace and then nat-urally cooled down to room temperature. The obtained monolithiccolumn was flushed with MeOH using an HPLC pump in order toremove the unreacted residuals. In parallel, the corresponding bulkpolyPOSS hybrid monoliths were prepared in glass vial under thesame conditions for characterizing the prepared polyPOSS hybridmonoliths. The bulk hybrid monoliths were cut into smaller pieces,extracted with methanol overnight in a Soxhlet apparatus, and thendried in a vacuum at 80 ◦C overnight before further measurements.

    3. Results and discussion

    3.1. Characterization of the polyPOSS hybrid monoliths

    One-pot method via amine-epoxy ring-opening polymerizationwas applied to prepare the polyPOSS hybrid monolithic capillarycolumns as we previously did [31,32]. The fabrication process andcondition can be found in Experimental section, and is schemati-cally illustrated in Fig. 1c. FT-IR study indicated that the stretchingvibration peaks at 907 and 847 cm−1 (C O in the epoxy group) ofEP-POSS and 3410 cm−1 (N H in the aromatic secondary aminegroup) became unnoticeable, while a broad stretching peak at3407 cm−1 (O H) appeared in the synthesized polyPOSS hybridmonolith when the molar ratio of PA-POSS to EP-POSS was 1/1,in addition to the peaks remained at 1600 and 1505 cm−1 (C Cin the benzene ring), and 1117 cm−1 (the Si O Si in the POSScage), suggesting the effective ring opening nucleophilic substitu-tion reaction between the phenylamine in PA-POSS and the epoxyin EP-POSS. Furthermore, the results obtained from elemental anal-ysis indicated that C/N ratio increased from 10.88 to 15.56 whenstoichiometric ratio of PA-POSS to EP-POSS decreased from 1.75/1to 1/1.7 in polymerization system (Table S1), again confirming thesuccessful polymerization.

    As we know, components of solvent mixtures (porogens) cangreatly affect the macroporous structure and related permeabil-ity of prepared stationary phase. Therefore, proper porogens werecarefully selected considering the factors of polarity, solvation andespecially solubility (the solubility parameter, ı) [30,33]. The ıvalues of DMF (ı = 20.87 J1/2 cm−3/2) and PEG (19.25 J1/2 cm−3/2)calculated by means of group contribution [34] are close to22.12 J1/2 cm−3/2 (PA-POSS) and 19.93 J1/2 cm−3/2 (EP-POSS), butH2O (47.87 J1/2 cm−3/2) [35] differs significantly. Therefore, aternary mixture of H2O, DMF and PEG for preparing the polyPOSShybrid monolithic capillary columns was used as the porogensto adjust the macropore and mesopore structures, in additionto the mass percentage (%) of the porogens in the prepoly-merization solution and the molar ratio of PA-POSS to EP-POSS(Table S2). Actually, the preparation of a monolith is a process ofpolymerization-induced phase separation, and the final morphol-ogy is a competition result between the kinetics of polymerizationand phase separation [36]. When keeping DMF constant (80%) for

    obtaining a homogeneous solution of PA- and EP-POSS, the bal-ance between the poor solvent H2O and good solvent PEG showeda remarkable influence on the morphology of the resulting mono-lithic columns (Fig. 2A–C). When the mass percentages of both H2O

  • J. Su et al. / J. Chromatogr. A 1533 (2018) 136–142 139

    Fig. 2. SEM images of the polyPOSS hybrid monoliths prepared under the conditions of a mass percentage of 80% DMF, 10% H2O and 10% PEG; 75% porogens in thep ). 80%7 . 80%,1

    addtfpitss1tsI1coatttooplma4fp3(c

    repolymerization solution; and the molar ratio of PA-POSS to EP-POSS = 1.75/1 (A0%; 1.75/1 (D). 80%, 8% and 12%; 80%; 1.75/1 (E). 80%, 8% and 12%; 85%; 1.75/1 (F)/1.5 (I). 80%, 8% and 12%; 75%; 1/1.7 (J).

    nd PEG were 10%, a center-hollow monolith (column A) formedue to that the poor solvent H2O limited the actually homogeneousistribution of the POSS monomer/crosslinker in the apparentlyransparent prepolymerization solution. As the decrease of H2Orom 10 to 6% while the increase of PEG 10–14%, continuedorous monoliths were obtained with the average macropore size

    ncreased from 0.95 �m to 3.5 �m (columns B and C). Althoughhe large macropores could provide good permeability, they causelow mobile phase mass transfer. In general, optimal macroporeize should be within 0.5–1 �m [37]. In the case of 8% H2O and2% PEG (column B), the average macropore size was 0.95 �m, andhe measured BET mesopore size was 10 nm, which was corre-ponding to the size formed by ca. 4 PA-POSS/EP-POSS (Fig. 1a).t should be noted that the native sub-nanometer siloxane cage,10.6 Å3 as estimated by ChemBio3D Ultra 12.0 considering as aube and/or 164.6 Å3 as a ball, which were related to the microp-re of the resulting monolith, could not be reflected effectively by

    conventional BET measurement, but still existed and contributedo the tri-porous structure of the polyPOSS hybrid monolithic sta-ionary phase. Moreover, mass percentage (%) of the porogens inhe prepolymerization solution, which controls the concentrationf the POSS precursors and thus the polymerization rate and theccurrence of phase separation as well, was investigated. The fasterolymerization rate and relative retarded phase separation under

    ess percentage (70%) of the porogens led to detachment of theonolith from the capillary wall because of interface stress [38],

    lthough the fine skeleton (ca. 300 nm) and smaller macropore (ca.00 nm) formed (column D). As the porogens percentage increasedrom 70 to 85%, the polymerization rate decreased and relative early

    hase separation occurred, resulting in the skeleton size from ca.00 to 850 nm and growing macropore from ca. 400 nm to 8 �mFig. 2D, B, E and F). Generally, 75% porogens was optimal in thease of column B (470 nm). On the other hand, in spite of the

    , 8% and 12%; 75%; 1.75/1 (B). 80%, 6% and 14%; 75%; 1.75/1 (C). 80%, 8% and 12%; 8% and 12%; 75%; 1.5/1 (G). 80%, 8% and 12%; 75%; 1/1 (H). 80%, 8% and 12%; 75%;

    insignificant difference in terms of the morphology and porousstructure under optimal preparation conditions (Fig. 2B, G, H, I andJ) owing to the same reactivity of the eight fully substituted pen-dant arms on PA-POSS and/or EP-POSS, the free phenylamine orepoxy groups remained in the monoliths when the molar ratio ofPA-POSS to EP-POSS was changed from 1.75/1 to 1/1.70. It shouldbe pointed out that unreacted phenylamine and/or epoxy groupsmight remain in the polyPOSS monolith even in the case of PA-POSS/EP-POSS = 1/1, because of possible steric hindrance, one of thepossible reasons. They may act different surface chemistry underthe sophisticated mobile phase of different composition and espe-cially the pH. We thus selected column B (1.75/1, guaranteeing offree phenylamine remained) and J (1/1.70, free epoxy) as examplesin the following chromatographic behavior study. Furthermore, themechanical stability of column B and J was investigated using purewater, methanol and acetonitrile (Fig. S1). Good relationships (thecorrelation coefficient R2 > 0.99) between the flow rate from 5 to30 �L/min (corresponding to 4.7 to 28.3 mm/s linear velocity) andthe backpressure from 0.3 to 23.4 MPa using a conventional HPLCpump were observed, indicating good mechanical stability of themonoliths. These monoliths also presented good thermostabilitybecause significant weight loss began until 350 ◦C when heatedunder N2 atmosphere (Fig. S2), implying that they might be usedas the stationary phase in gas chromatography. Additionally, in thecase of column B, its reproducibility was evaluated in term of RSD(%, n = 3) for the retention factors (k) of toluene as a model analyte(thiourea as the void time marker) on nano-LC. The RSDs of run-to-run, day-to-day and batch-to-batch were 0.33%, 3.98% and 6.85%,respectively. Meanwhile, neither significant decrease of column

    efficiency nor obvious column deterioration was observed evenafter hundreds of continuous injections, demonstrating the highstability of the synthesized polyPOSS hybrid monolithic capillarycolumn.

  • 140 J. Su et al. / J. Chromatogr. A 1533 (2018) 136–142

    Fig. 3. Typical reversed-phase chromatographic separation of alkylbenzenes and alkylbenzene ketones on the polyPOSS hybrid monolithic capillary column B. Columnsize: 75 �m i.d. ×22.5 cm in length; mobile phase: 60% ACN and 40% H O; flow rate: 200 nL/min; UV detection at 214 nm. Analytes: (1) thiourea, (2) toluene, (3) ethylben-z ) proh

    3m

    3a

    caltdmpecsgBtazAat24t(attflhcanv(epptcaatia

    2

    ene, (4) propylbenzene, (5) butylbenzene, (6) pentylbenzene, (7) acetophenone, (8eptanophenone.

    .2. Chromatographic performance of the polyPOSS hybridonolithic capillary columns on nano-LC

    .2.1. Chromatographic separation of alkylbenzenes andlkylbenzene ketones using neutral mobile phases

    As we know, the siloxane cage and chemically generatedrosslinking part (Fig. 1c), which include the tertiary phenylaminend hydroxyl groups in the synthesized polyPOSS hybrid mono-ithic stationary phase, contribute to the surface chemistry andhe corresponding interactions with an analyte, controlling theistribution of the analyte between the stationary phase andobile phase. The cubic siloxane cage is very hydrophobic, its

    olarity (logP) reaches 7.66 as calculated by molinspiration prop-rty calculator on the website http://www.molinspiration.com; therosslinking part is also hydrophobic as a whole (logP = 3.14). Ithould be noted that there are unreacted PA (phenylaminopropyl)roups (logP = 2.67) existing in the stationary phase of column

    due to the excess PA-POSS used. Such a surface chemistry ofhe monolithic stationary phase was thus expected to perform

    reversed-phase chromatographic separation towards alkylben-enes using a relative polar and neutral mobile phase composed ofCN and H2O. The obtained results indicated that the alkylbenzenesnd thiourea were well separated in the sequence according toheir polarity (thiourea, logP = −0.46; toluene, 2.39; ethylbenzene,.85; propylbenzene, 3.24; butylbenzene, 3.80; pentylbenzene,.31) (Fig. 3a), and their retention factors decreased along withhe increase of ACN content in the mobile phase from 55 to 70%Fig. 3b), demonstrating a typical reversed-phase separation mech-nism based mainly on the hydrophobic and �-� interactions withhe siloxane cages and the aminophenyls. The plate heights ofhiourea and toluene were 7.4 and 6.4 �m under 60% ACN at theow rate of 200 nL/min. Moreover, the alkylbenzene ketones, whichave the carbonyl besides the same phenyl and alkyl substitutesompared to their corresponding alkylbenzene counterparts, werelso baseline separated according to their polarity of acetophe-one (logP = 1.84), propiophenone (2.34), butyrophenone (2.90),alerophenone (3.40), caprophenone (3.91) and heptanophenone4.41) (Fig. 3c). It should be pointed out that the chemically gen-rated hydroxyl and the remaining hydrogen in the secondaryhenylamine groups on the polyPOSS hybrid monolithic stationaryhase may theoretically offer additional hydrogen bonding interac-ions. Such hydrogen bonding interactions with the oxygen in thearbonyl of the molecules should result in higher retention of thelkylbenzene ketones, but the results observed indicated that the

    lkylbenzene ketones have shorter retention time (less retention)han their alkylbenzene counterparts. Clearly, the hydrogen bond-ng between the carbonyl oxygen and the hydrogen in hydroxylnd phenylamine was much more compromised by H2O and ACN

    piophenone, (9) butyrophenone, (10) valerophenone, (11) caprophenone, and (12)

    in the mobile phase; on the other hand, the inductive effect of thecarbonyl in the alkylbenzene ketone molecules contributed a cer-tain extent to the weaker hydrophobic and �-� interactions withthe stationary phase, resulting in less retention.

    3.2.2. Chromatographic separation of polar small molecules usingacidic mobile phases

    Compared with column B, free EP containing epoxy group(logP = 1.01) exists on the monolithic stationary phase of column J,in addition to the siloxane cages and chemically generated tertiaryphenylamine and hydroxyl groups. The redundant epoxy groupscan be hydrolyzed under acidic medium to form adjacent hydroxyls(logP = −0.59) [39]; moreover, the tertiary phenylamine (pKa = 5.8)can be protonated under an acidic condition. These mean that thesurface chemistry of column J will be changed by merely adjust-ing pH of the mobile phase to be used. Separation of cytosine(logP = −1.61), uracil (−1.09) and thymine (−0.6) was first per-formed using water (−0.29) as the mobile phase (Fig. 4a). UnderpH 5.31, not only majority of the tertiary phenylamine (logP = 2.04)(ca. 31% are protonated, −1.12) but also the epoxy (0.52) remainedintact in the stationary phase, the separation sequence of thenucleic acid bases was again dominantly according to their polar-ity. Similar results were obtained from the separation of histidine(His, logP = −3.00), phenylalanine (Phe, −1.23) and tryptophan (Trp,−1.08) using water (pH 6.81) as the mobile phase (Fig. 4b). The res-olution factors RHis/Phe and RPhe/Trp were 2.57 and 2.38, and the plateheights 31.2 (His), 24.9 (Phe) and 40.3 �m (Trp). While their sepa-ration became more efficient with very sharp peaks and shorterretention time under pH 1.30; RHis/Phe and RPhe/Trp increased to3.65 and 5.17, and the plate heights reached 4.0 (His), 4.7 (Phe)and 10.2 �m (Trp). These observations suggested that the chem-ically generated hydrophilic hydroxyls and tertiary phenylaminethat were totally positive-charged at pH 1.30 works on the separa-tion of the three amino acids, which were also positively chargedunder such an acidic condition considering their pI of 7.64 (His),5.91 (Phe) and 5.88 (Trp). One could speculate that electrostaticrepulsion and hydrophilic interaction become significant due to thesurface chemistry change resulted from pH-switch of the mobilephase, and thus fast mass transfer rates (C-term = 3.01 ms for His,2.87 Phe and 3.93 Trp) (Fig. 4c). Such a pH-switchable surface chem-istry from predominantly hydrophobic to significantly hydrophilicstatus provides an opportunity to separate other polar compoundsunder HILIC mode.

    We thus selected the typical HILIC mechanism probing com-

    pounds of formamide (logP = −0.89), thiourea (−0.46), dimethyl-formamide (DMF) (−0.27) and toluene (2.39) to see whether themonolithic stationary phase of column J has HILIC characteristicsor not. As the increase of ACN percentage from 40% to 80% in the

    http://www.molinspiration.comhttp://www.molinspiration.comhttp://www.molinspiration.comhttp://www.molinspiration.com

  • J. Su et al. / J. Chromatogr. A 1533 (2018) 136–142 141

    Fig. 4. Separation of nucleic acid bases (a); amino acids (b) and their corresponding H-u plots (c); HILIC probe compounds (d); phenols (e) and phenolic acids (f) on thep n lenga ACN (

    ataaTrlwcmtem

    nr0rtscmfdoq(tA(SamD1zie

    olyPOSS hybrid monolithic capillary column J. Column size: 75 �m i.d. ×22.5 cm ind 1.30); (c) water (pH 1.30); (d) ACN (40% and 80% containing 0.1% TFA); (e) 20%

    cidic mobile phase containing 0.1% TFA, we observed conversion ofhe retention time of toluene and thiourea (Fig. 4d), demonstrating

    typical HILIC behavior. The four compounds were efficiently sep-rated in the sequence of toluene, DMF, formamide and thiourea.his sequence was also somewhat contributed by the electrostaticepulsion effect, because of the positive-charged tertiary pheny-amine in the stationary phase, meanwhile DMF and formamide

    ere also positively charged. The plate heights under 80% ACNontaining 0.1% TFA were 11.1 (toluene), 8.41 (DMF), 8.86 (for-amide) and 9.79 �m (thiourea). Such a nano-LC performance of

    he prepared polyPOSS hybrid monolithic capillary column J wasven comparable to that of CEC using the silica-based polypeptideonolithic capillary column [32].Compared with the amino acids and nucleic acid bases, phe-

    ols and phenolic acids as well as benzoic acid and sorbic acidemain intact (do not ionize) under the mobile phase containing.1% TFA. They are not affected significantly by the electrostaticepulsion effect arising from the positive-charged phenylamine onhe stationary phase while the hydrophobic interactions becameignificant, which may drive them entering into the siloxaneages (ca. 110.6/164.6 Å3). The siloxane cages contribute to theicropore structure of the hybrid monolithic stationary phase as

    ore-mentioned, playing an additional size-exclusion effect as evi-enced by the polarity- and size-dependent separation sequencef phloroglucin (logP = 0.43; molecular volume = 108.10 Å3), hydro-uinone (0.98; 100.08 Å3), resorcinol (0.95; 100.08 Å3), catechol0.99; 100.08 Å3) and phenol (1.46; 92.06 Å3) using 20% ACN con-aining 0.1% TFA as the mobile phase (Fig. 4e). Further increasingCN content in the mobile phase led to shorter retention time

    Fig. S3), demonstrating a mixed mode separation mechanism.imilar results (Figs. 4f and S3) from the separation of phenoliccids including 2,6-dihydroxybenzoic acid (2,6-DBA; logP = 1.39;olecular volume = 127.08 Å3), 3,4-dihydroxybenzoic acid (3,4-BA; 0.88; 127.08 Å3), 3,5-dihydroxybenzoic acid (3,5-DBA; 0.82;

    27.08 Å3) and p-hydroxybenzoic acid (1.37; 119.06 Å3) and ben-oic acid (1.85; 111.06 Å3), except sorbic acid (0.97; 111.03 Å3) thats a linear unsaturated acid, generally confirmed the size-exclusionffect of the siloxane cages again. It should be noted that, how-

    th; UV detection at 214 nm. Mobile phase: (a) water (pH 5.31); (b) water (pH 6.810.1% TFA) and (f) water (0.1% TFA) at the flow rate of 200 nL/min.

    ever, separation order of the three hydroxy phenols (hydroquinone,resorcinol and catechol) of similar polarity and size but differenthydroxyl substitution positions indicated that the positive-chargedphenylamines and hydroxyls influence their separation; these mys-tic effects became more remarkable in the case of 2,6-DBA thatwas first eluted among the three phenolic acids studied. It issimilar in size with its analogues of 3,4-DBA and 3,5-DBA, andmore hydrophobic, thus should be theoretically eluted later. Onepossible reason for its first elution might be its intramolecularhydrogen bonding while intermolecular hydrogen bonding is dom-inant in the other two phenolic acids. This can be reflected byobservations of not only their changing separation order but alsoless retention even co-eluted (Fig. S3) along with the increase inACN content because the non-protonic solvent ACN affects muchmore remarkably on the intermolecular hydrogen bonding thanthe intramolecular one. A mixed mode separation with multipleinteractions should again apply to this situation, suggesting thatthe novel stationary phase with tri-porous structure and multiplesurface chemistry provides flexibility to improve the resolution ofparticular small molecules using a sophisticated mobile phase.

    4. Conclusions

    We developed a dual POSS-polymerized approach to preparepolyPOSS hybrid monolithic stationary phases. The two nanoscalePOSS with PA and EP modifications offer more surface chemistry,in addition to the resulted tri-porous monolithic structure, allow-ing us to perform a more effective separation of the targeted smallmolecules on reversed-phase through HILIC to mixed mode mech-anisms via adjusting the molar ratio of POSS precursors while thecomposition and pH of a mobile phase. Not limited to the smallmolecules studied here, this polyPOSS approach can be expectedto fabricate other kinds of hybrid monolithic stationary phases thatare suitable for biomacromolecules and chiral molecules separation

    when choosing the existed and/or designing new POSS with thependant groups of different structure and chemical property as wellas chirality. Moreover, we believe that the thermostability of themonolithic stationary phases obtained via such an approach ren-

  • 1 ogr. A

    dp

    A

    etaaNPU

    A

    t1

    R

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [38] G. Guiochon, Monolithic columns in high-performance liquid

    42 J. Su et al. / J. Chromat

    ers their potential usefulness as gas chromatographic stationaryhases in the near future.

    cknowledgments

    This work was financially supported by the National Natural Sci-nce Foundation of China (Grants 21535007, 21475108, 21275120),he National Basic Research 973 Program (Grant 2014CB932004)nd National Science and Technology Basic Work (2015FY111400)s well as the Foundation for Innovative Research Groups of theational Natural Science Foundation of China (Grant 21521004),rogram for Changjiang Scholars and Innovative Research Team inniversity (PCSIRT, Grant IRT13036).

    ppendix A. Supplementary data

    Supplementary material related to this article can be found, inhe online version, at doi:https://doi.org/10.1016/j.chroma.2017.2.033

    eferences

    [1] F. Svec, Y. Lv, Advances and recent trends in the field of monolithic columnsfor chromatography, Anal. Chem. 87 (2015) 250–273.

    [2] C. Viklund, F. Svec, J.M.J. Fréchet, K. Irgum, Monolithic, ”molded”, porousmaterials with high flow characteristics for separations, catalysis, orsolid-phase chemistry: control of porous properties during polymerization,Chem. Mater. 8 (1996) 744–750.

    [3] K. Nakanishi, N. Tanaka, Sol-gel with phase separation. Hierarchically porousmaterials optimized for high-performance liquid chromatographyseparations, Acc. Chem. Res. 40 (2007) 863–873.

    [4] F. Svec, J.M.J. Fréchet, Continuous rods of macroporous polymer ashigh-performance liquid chromatography separation media, Anal. Chem. 64(1992) 820–822.

    [5] T. Umemura, Y. Ueki, K.-i. Tsunoda, A. Katakai, M. Tamada, H. Haraguchi,Preparation and characterization of methacrylate-based semi-micromonoliths for high-throughput bioanalysis, Anal. Bioanal. Chem. 386 (2006)566–571.

    [6] M. Chen, L. Li, B. Yuan, Q. Ma, Y. Feng, Preparation and characterization ofmethacrylate-based monolith for capillary hydrophilic interactionchromatography, J. Chromatogr. A 1230 (2012) 54–60.

    [7] H. Minakuchi, K. Nakanishi, N. Soga, N. Ishizuka, N. Tanaka, Effect of domainsize on the performance of octadecylsilylated continuous porous silicacolumns in reversed-phase liquid chromatography, J. Chromatogr. A 797(1998) 121–131.

    [8] F. Hoffmann, M. Cornelius, J. Morell, M. Froeba, Silica-based mesoporousorganic-inorganic hybrid materials, Angew. Chem. Int. Ed. 45 (2006)3216–3251.

    [9] T. Zhu, K.H. Row, Preparation and applications of hybrid organic-inorganicmonoliths: a review, J. Sep. Sci. 35 (2012) 1294–1302.

    10] J. Ou, Z. Liu, H. Wang, H. Lin, J. Dong, H. Zou, Recent development of hybridorganic-silica monolithic columns in CEC and capillary LC, Electrophoresis 36(2015) 62–75.

    11] Z. Zhang, Y. Hao, J. Ding, S. Xu, B. Yuan, Y. Feng, One-pot preparation of amixed-mode organic-silica hybrid monolithic capillary column and itsapplication in determination of endogenous gibberellins in plant tissues, J.Chromatogr. A 1416 (2015) 64–73.

    12] D.A. Loy, K.J. Shea, Briged polysilsesquioxanes-highly porous hybridorganic-inorganic materials, Chem. Rev. 95 (1995) 1431–1442.

    13] D.B. Cordes, P.D. Lickiss, F. Rataboul, Recent developments in the chemistry ofcubic polyhedral oligosilsesquioxanes, Chem. Rev. 110 (2010) 2081–2173.

    14] M. Wu, R. Wu, R. Li, H. Qin, J. Dong, Z. Zhang, H. Zou, Polyhedral oligomericsilsesquioxane as a cross-linker for preparation of inorganic-organic hybridmonolithic columns, Anal. Chem. 82 (2010) 5447–5454.

    15] X. Lin, X. Wang, T. Zhao, Y. Zheng, S. Liu, Z. Xie, Electroneutral silica-basedhybrid monolith for hydrophilic interaction capillary electrochromatography,J. Chromatogr. A 1260 (2012) 174–182.

    16] H. Zhang, J. Ou, Y. Wei, H. Wang, Z. Liu, H. Zou, A hybrid fluorous monolithiccapillary column with integrated nanoelectrospray ionization emitter for

    [

    1533 (2018) 136–142

    determination of perfluoroalkyl acids by nano-liquidchromatography-nanoelectrospray ionization-mass spectrometry/massspectrometry, J. Chromatogr. A 1440 (2016) 66–73.

    17] X. Xiong, Z. Yang, Y. Li, L. Xiao, L. Jiang, Y. Chen, M. Ma, B. Chen, Preparation ofa polyhedral oligomeric silsesquioxane-based perfluorinated monolithiccolumn, J. Chromatogr. A 1304 (2013) 85–91.

    18] J. Ou, Z. Zhang, H. Lin, J. Dong, M. Wu, H. Zou, Preparation and application ofhydrophobic hybrid monolithic columns containing polyhedral oligomericsilsesquioxanes for capillary electrochromatography, Electrophoresis 33(2012) 1660–1668.

    19] Z. Liu, J. Ou, Z. Liu, J. Liu, H. Lin, F. Wang, H. Zou, Separation of intact proteinsby using polyhedral oligomeric silsesquioxane based hybrid monolithiccapillary columns, J. Chromatogr. A 1317 (2013) 138–147.

    20] N. Zhang, L. Zhang, X. Qiao, Y. Wang, H. Yan, L. Bai, Facile one-pot preparationof an imidazolium embedded C8 hybrid monolith using polyhedral oligomericsilsesquioxane for capillary liquid chromatography, RSC Adv. 5 (2015)91436–91440.

    21] X. Qiao, N. Zhang, M. Han, X. Li, X. Qin, S. Shen, Periodic imidazolium-bridgedhybrid monolith for high-efficiency capillary liquid chromatography withenhanced selectivity, J. Sep. Sci. 40 (2017) 1024–1031.

    22] Y. Shan, L. Qiao, X. Shi, G. Xu, Preparation and evaluation of a novel hybridmonolithic column based on pentafluorobenzyl imidazolium bromide ionicliquid, J. Chromatogr. A 1375 (2015) 101–109.

    23] Z. Zajickova, Advances in the development and applications of organic-silicahybrid monoliths, J. Sep. Sci. 40 (2017) 25–48.

    24] H. Lin, J. Ou, Z. Zhang, J. Dong, H. Zou, Ring-opening polymerization reactionof polyhedral oligomeric silsesquioxanes (POSSs) for preparation ofwell-controlled 3D skeletal hybrid monoliths, Chem. Commun. 49 (2013)231–233.

    25] H. Lin, L. Chen, J. Ou, Z. Liu, H. Wang, J. Dong, H. Zou, Preparation ofwell-controlled three-dimensional skeletal hybrid monoliths via thiol-epoxyclick polymerization for highly efficient separation of small molecules incapillary liquid chromatography, J. Chromatogr. A 1416 (2015) 74–82.

    26] H.B. He, B. Li, J.P. Dong, Y.Y. Lei, T.L. Wang, Q.W. Yu, Y.Q. Feng, Y.B. Sun,Mesostructured nanomagnetic polyhedral oligomeric silsesquioxanes (POSS)incorporated with dithiol organic anchors for multiple pollutants capturinginwastewater, ACS Appl. Mater. Interfaces 5 (2013) 8058–8066.

    27] H.B. He, D. Chen, B. Li, J.P. Dong, T.Y. Bo, T.L. Wang, Q.W. Yu, Y.Q. Feng,Fabrication of enrofloxacin imprinted organic-inorganic hybrid mesoporoussorbent from nonomagnetic polyhedral oligomeric silsesquioxanes for theselective extraction of fluoroquinolones in milk samples, J. Chromatogr. A1361 (2014) 23–33.

    28] H. Lin, J. Ou, Z. Liu, H. Wang, J. Dong, H. Zou, Facile construction ofmacroporous hybrid monoliths via thiol-methacrylate Michael addition clickreaction for capillary liquid chromatography, J. Chromatogr. A 1379 (2015)34–42.

    29] X.Q. Qiao, R. Chen, H.Y. Yan, S.G. Shen, Polyhedral oligomericsilsesquioxane-based hybrid monolithic columns: recent advances in theirpreparation and their applications in capillary liquid chromatography, TrAc97 (2017) 50–64.

    30] Z.D. Xu, L.M. Yang, Q.Q. Wang, Different alkyl dimethacrylate mediated stearylmethacrylate monoliths for improving separation efficiency of typicalalkylbenzenes and proteins, J. Chromatogr. A 1216 (2009) 3098–3106.

    31] H.R. Xu, Z.D. Xu, L.M. Yang, Q.Q. Wang, Öne-potp̈reparation of basic aminoacid-silica hybrid monolithic column for capillary electrochromatography, J.Sep. Sci. 34 (2011) 2314–2322.

    32] L.C. Zhao, L.M. Yang, Q.Q. Wang, Silica-based polypeptide-monolithicstationary phase for hydrophilic chromatography and chiral separation, J.Chromatogr. A 1446 (2016) 125–133.

    33] O. Okay, Macroporous copolymer networks, Prog. Polym. Sci. 25 (2000)711–779.

    34] R.F. Fedors, A method for estimating both the solubility parameters and molarvolumes of liquids, Polym. Eng. Sci. 14 (1974) 147–154.

    35] F. Gao, Y. Ji, R. Han, B. Chen, Estimation and determination of the solubilityparameter of 1,5-diazido-3-nitrazapentane, Chin. J. Energ. Mater. 21 (2013)612–615.

    36] J. Li, Z. Du, H. Li, C. Zhang, Porous epoxy monolith prepared via chemicallyinduced phase separation, Polymer 50 (2009) 1526–1532.

    37] N. Tanaka, H. Kobayashi, K. Nakanishi, H. Minakuchi, N. Ishizuka, MonolithicLC columns, Anal. Chem. 73 (2001) 420–429.

    chromatography, J. Chromatogr. A 1168 (2007) 101–168.39] R. Mallik, T. Jiang, D.S. Hage, High-performance affinity monolith

    chromatography: development and evaluation of human serum albumincolumns, Anal. Chem. 76 (2004) 7013–7022.

    https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033https://doi.org/10.1016/j.chroma.2017.12.033http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0005http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0010http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0015http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0020http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0025http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0030http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0035http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0040http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0045http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0050http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0055http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0060http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0065http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0070http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0075http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0080http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0085http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0090http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0095http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0100http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0105http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0110http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0115http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0120http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0125http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0130http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.com/S0021-9673(17)31812-5/sbref0135http://refhub.elsevier.c