pubs.acs.org/Langmuir

Determination of the Electron Escape Depth for NEXAFS Spectroscopy

K. E. Sohn,† M. D. Dimitriou,† J. Genzer,‡ D. A. Fischer,§ C. J. Hawker,†, ) and E. J. Kramer*,†,^

†Department of Materials, University of California Santa Barbara, Santa Barbara, California 93106,‡Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, North

Carolina, 27695, §National Institute of Standards and Technology, Gaithersburg,Maryland 20899, )Departmentof Chemistry and Biochemistry, University of California Santa Barbara, Santa Barbara, California 93106, and

^Department of Chemical Engineering, University of California Santa Barbara, Santa Barbara,California 93106

Received December 1, 2008. Revised Manuscript Received March 12, 2009

A novel method was developed to determine carbon atom density as a function of depth by analyzing the postedgesignal in near-edge X-ray absorption fine structure (NEXAFS) spectra. We show that the common assumption in theanalysis ofNEXAFSdata frompolymer films, namely, that the carbon atomdensity is constant as a function of depth, isnot valid. This analysismethod is then used to calculate the electron escape depth (EED) forNEXAFS in amodel bilayersystem that contains a perfluorinated polyether (PFPE) on top of a highly oriented pyrolitic graphite (HOPG) sample.Because the carbon atom densitites of both layers are known, in addition to the PFPE surface layer thickness, the EED isdetermined to be 1.95 nm. This EED is then used to measure the thickness of the perfluorinated surface layer ofpoly(4-(1H,1H,2H,2H-perfluorodecyl)oxymethylstyrene) (PFPS).

1. Introduction

Knowledge of surface composition of polymeric materials aswell as organic self-assembled monolayers (SAMs) is necessaryfor many applications including antibiofouling coatings forships,1-4 polymer field effect transistors5-10 and polymer light-

emitting diodes (PLEDS),11,12 and biomedical devices.13-20Near-edge X-ray absorption fine structure (NEXAFS) spectroscopyand X-ray photoelectron spectroscopy (XPS) are often employedto determine the composition in the near surface region of asample.21-27 The two techniques are similar, but have distinctdifferences. However, when used in tandem, they provide valu-able information about the surface.

NEXAFS spectroscopy uses tunable, polarized X-rays from asynchrotron source to determine which bonds are present withinthe top 2 nm of the surface as well as their orientation.3,28,29 TheX-rays used in NEXAFS spectroscopy excite a core electron intoan empty molecular orbital, leaving a hole in the core shell. Thishole is filled by an electron from a higher energy level, and anAuger electron is emitted. The number ofAuger electrons that aredetected for a given X-ray energy supplies information about thenumber of core-holes that are created, and therefore the transi-tions that are occurring. Because NEXAFS spectroscopy isfocused on orbitals instead of individual elements, it is possibleto differentiate between C-C and CdC bonds because of thepresence of theπ orbital in the double bond, a capability thatXPSdoes not have.

The typical experimental geometry for NEXAFS is shown inFigure 1, which depicts the tunable soft X-ray beam, the goni-ometer on which the sample is mounted, and the partial electronyield (PEY) detector. The goniometer can be rotated so that theincident X-ray beammakes an angle θ with respect to the sample

*Corresponding author.(1) Finlay, J. A.; Callow, M. E.; Ista, L. K.; Lopez, G. P.; Callow, J. A.

Integr. Comp. Biol. 2002, 42, 1116–1122.(2) Callow, J. A.; Callow, M. E.; Ista, L. K.; Lopez, G.; Chaudhury, M. K.

J. R. Soc. Interface 2005, 2, 319–325.(3) Krishnan, S.; Ayothi, R.; Hexemer, A.; Finlay, J. A.; Sohn, K. E.; Perry,

R.; Ober, C. K.; Kramer, E. J.; Callow, M. E.; Callow, J. A.; Fischer, D. A.Langmuir 2006, 22, 5075–5086.(4) Krishnan, S.; Wang, N.; Ober, C. K.; Finlay, J. A.; Callow, M. E.;

Callow, J. A.; Hexemer, A.; Sohn, K. E.; Kramer, E. J.; Fischer, D. A.Biomacromolecules 2006, 7, 1449–1462.(5) DeLongchamp, D. M.; Sambasivan, S.; Fischer, D. A.; Lin, E. K.;

Chang, P.; Murphy, A. R.; Frechet, J. M. J.; Subramanian, V. Adv. Mater.2005, 17, 2340.(6) Pattison, L. R.; Hexemer, A.; Kramer, E. J.; Krishnan, S.; Petroff, P.

M.; Fischer, D. A. Macromolecules 2006, 39, 2225–2231.(7) DeLongchamp, D.; Kline, R. J.; Lin, E.; Fischer, D.; Richter, L.; Lucas,

L.; Heeney, M.; McCulloch, I.; Northrup, J. Adv. Mater. 2007, 19, 833.(8) Kline, R. J.; DeLongchamp, D. M.; Fischer, D. A.; Lin, E. K.; Heeney,

M.; McCulloch, I.; Toney, M. F. Appl. Phys. Lett. 2007, 90, 052117.(9) Kline, R. J.; DeLongchamp, D. M.; Fischer, D. A.; Lin, E. K.; Richter,

L. J.; Chabinyc, M. L.; Toney, M. F.; Heeney, M.; McCulloch, I. Macro-molecules 2007, 40, 7960–7965.(10) Gurau, M. C.; Delongchamp, D. M.; Vogel, B. M.; Lin, E. K.; Fischer,

D. A.; Sambasivan, S.; Richter, L. J. Langmuir 2007, 23, 834–842.(11) DeLongchamp, D. M.; Ling, M. M.; Jung, Y.; Fischer, D. A.; Roberts,

M. E.; Lin, E. K.; Bao, Z. J. Am. Chem. Soc. 2006, 128, 16579–16586.(12) Park, J.; Yang, R.; Hoven, C. V.; Garcia, A.; Fischer, D. A.; Nguyen,

T.-Q.; Bazan, G. C.; DeLongchamp, D. M. Adv. Mater. 2008, 20, 2491+.(13) Clare, A. S. J. Mar. Biotechnol. 1998, 6, 3–6.(14) Bullock, S.; Johnston, E. E.; Willson, T.; Gatenholm, P.; Wynne, K. J.

J. Colloid Interface Sci. 1999, 210, 18–36.(15) Gan, D. J.; Mueller, A.; Wooley, K. L. J. Polym. Sci., Part A: Polym.

Chem. 2003, 41, 3531–3540.(16) Youngblood, J. P.; Andruzzi, L.; Ober, C. K.; Hexemer, A.; Kramer, E.

J.; Callow, J. A.; Finlay, J. A.; Callow, M. E. Biofouling 2003, 19, 91–98.(17) Schmidt, D. L.; Brady, R. F.; Lam, K.; Schmidt, D. C.; Chaudhury, M.

K. Langmuir 2004, 20, 2830–2836.(18) Yebra, D.; Kiil, S.; Dam-Johansen, K. Prog. Org. Coat. 2004, 50, 75–

104.(19) Lee, S.; Voros, J. Langmuir 2005, 21, 11957–11962.(20) Yagi, S.; Matsumura, Y.; Nomoto, T.; Soda, K.; Hashimoto, E.;

Namatame, H.; Taniguchi, M. Surf. Sci. 2007, 601, 4154–4157.

(21) Ratner, B. D. Surf. Interface Anal. 1995, 23, 521–528.(22) Sodhi, R. N. S.; Sahi, V. P.; Mittelman, M. W. J. Electron Spectrosc.

Relat. Phenom. 2001, 121, 249–264.(23) Castner, D. G.; Ratner, B. D. Surf. Sci. 2002, 500, 28–60.(24) Cheng, Z. Y.; Teoh, S. H. Biomaterials 2004, 25, 1991–2001.(25) Murata, H.; Chang, B. J.; Prucker, O.; Dahm, M.; Ruhe, J. Surf. Sci.

2004, 570, 111–118.(26) Kannan, R. Y.; Salacinski, H. J.; Vara, D. S.; Odlyha, M.; Seifalian, A.

M. J. Biomater. Appl. 2006, 21, 5–32.(27) Cheng, F.; Gamble, L. J.; Castner, D. G. Anal. Chem. 2008, 80, 2564–

2573.(28) St

::ohr, J. NEXAFS Spectroscopy; Springer Series in Surface Sciences;

Springer-Verlag: Berlin/Heidelberg, 1992.(29) St

::ohr, J.; Samant, J. G. J. Electron Spectrosc. Relat. Phenom. 1999,

98-99, 187–207.

Published on Web 4/28/2009

© 2009 American Chemical Society

DOI: 10.1021/la803951yLangmuir 2009, 25(11), 6341–6348 6341

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

surface. There are two possible methods for depth profiling usingNEXAFS spectroscopy. The first is to vary the negative bias onthe entrance grid in front of the PEYdetector. The purpose of thisgrid is to screen out electrons that have lost a significant amountof energy while leaving the sample. As the bias becomes morenegative, more electrons that have lost energy are prevented fromreaching the detector, increasing the surface sensitivity.30Anotherway to depth profile is to vary φ, the electron emission angle fromthe sample normal, just as in XPS (see Figure 1). However,because NEXAFS spectroscopy is sensitive to bond orientation,there are two factors that cause the intensity to vary as a functionofφ (and θ): the orientation of a particular bond, and the depth ofthe bond below the surface of the film. These two effects need tobe separated in order to successfully use NEXAFS spectroscopyas a depth profiling technique.

In addition, the sampling depth, i.e., the electron escape depth(EED), of the technique must be known. Genzer et al. haveexperimentally determined the EED for NEXAFS at two valuesof θ’s where the effects of orientation should be the same fordifferent values of φ, but these results relied on very small signalsfrommethylene groups under a fluorinated SAM. The possibilitythat the surface of the semifluorinated SAM might have beencontaminated by airborne hydrocarbons or siloxanes (common inlaboratory environments containing pumps) was not consid-ered.30 Krishnan et al. calculated the inelastic mean free path(IMFP) for polymeric systems using a method developed byCumpson, but did not verify it experimentally.3,31 The IMFP isa measure of the average distance an electron can travel througha material without being inelastically scattered, and thereforelosing energy, since the probability of inelastic scattering aftertraveling a distance x is given by e-x/IMFP. The IMFP isdependent on the kinetic energy of the outgoing electron, whichforNEXAFS at the CK-edge is anAuger electron with an energyof approximately 263 eV.

In XPS, since the energy of the photoelectron is measuredaccurately, the IMFP can be used directly to compute thephotoelectron yield for a given depth profile as a function of φ.In NEXAFS, since the entrance grid bias (EGB) does notdiscriminate Auger electrons that have lost a small amountof energy from those that have lost none, the EED will be

larger than the IMFP for the Auger electron energy (∼263 eVfor carbon atoms), and it will increase with decreasing negativeEGB.30

In the case of a geometry where the electron detector is notin the same plane as the sample normal and the incoming X-rays,another angle, ω, must be taken into account. This is theangle that the electron detector makes with the plane definedby the incoming X-ray beam and the sample normal (seeFigure 1). In this case, the effective EED will be λ* = λ cos(φ)cos(ω).

In order to determine the EED for NEXAFS, a model systemof highly oriented pyrolytic graphite (HOPG) with a thin surfacelayer of perfluoropolyether (PFPE) has been used. The PFPElayer is Fomblin Z-03 and has the formula CF3-[(O-CF2-CF2)m-(O-CF2)n]-O-CF3 wherem+ n varies between 40 and180 andm/n ranges from0.5 to 2. This systemwas chosen becausetheHOPGhasπbondswith a knownorientation,while the PFPEhas noπ bonds.Additionally important is that the exact thicknessof the PFPE layer can be independently determined using othercharacterization methods.32

The other important aspect of this work is the development ofan analysis method using the postedge data to accurately depthprofile through different types of samples. By fitting the postedgedata prior to normalization, the effects of nonuniform carbonatom density with depth, as well as surface layer thicknesses, canbe determined. Previous work attempting to depth profile usingNEXAFS spectroscopy assumed a constant density as a functionof depth, but our work reported below shows that the changingcarbon atom density must be taken into consideration, and that,in addition, this information provides useful data to constrainmodels of the polymer surface.3,30

2. Theory

At normal X-ray incidence, the number of Auger electronsgenerated from a slice dz below the surface is

Nedz ¼ I0A0μEdz ð1Þwhere μE is the X-ray energy dependent absorption coefficientgiving rise to the Auger electrons, A0 is the area irradiated by theincidentX-rays and I0 is the incident photon flux density.

28 I0A0 isconstant as the sample is tilted because the intensity per unit areadecreases by the same factor that the irradiated area increases asthe sample tilts to more glancing angles, i.e., θ decreases. Theeffective absorption coefficient, μeffective, is not constant, however,as θ is varied:

μeffective ¼ μEsinðθÞ ð2Þ

because the length of the path traveled by a photon through athickness dz of material is dz/sin(θ). When θ 6¼ 90�, eq 1becomes

Nedz ¼ I0A0μEdz

sinðθÞ ð3Þ

and describes the number of Auger electrons created in a slab ofthickness dz within the sample. In writing eq 3, the decrease inX-ray intensity with depth (I(z) = I0e

-μEz/sinθ) is neglected, sinceexperimentally the X-ray absorption length, 1/μE, is much

Figure 1. Depth profiling with NEXAFS spectroscopy. The sam-ple is rotatedwith respect to theX-ray beam to change the effectiveEED, λ*. The angle between the PEY detector and the incidentX-rays is fixed.

(30) Genzer, J.; Kramer, E. J.; Fischer, D. A. J. Appl. Phys. 2002, 92, 7070–7079.(31) Cumpson, P. J. Surf. Interface Anal. 2001, 31, 23–34.

(32) Toney, M. F.; Mate, C. M.; Leach, K. A.; Pocker, D. J. ColloidInterface Sci. 2000, 225, 219–226.

DOI: 10.1021/la803951y Langmuir 2009, 25(11),6341–63486342

Article Sohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

greater than the Auger electron EED.28 The absorptioncoefficient is related to the energy-dependent absorption crosssection by

μE ¼ nvðzÞσE ð4Þwhere nv(z) is the number density of carbon atoms capable ofundergoing the transition described by σE. The number density ofcarbon atoms is determined by

nvðzÞ ¼ FmðzÞNAv

Mð5Þ

where NAv is Avogadro’s number, Fm is the mass density ofcarbon atoms, and M is the atomic mass, i.e., 12 for carbon. Byintegrating eq 1, the number of Auger electrons created through-out the sampling depth can be determined:

I ¼ Ω

4π

Z ¥

0

dzNe

¼ I0A0Ω

4π sinðθÞZ ¥

0

dznvðzÞσEe-

�z=½λ cosðφÞ cosðωÞ�

�ð6Þ

where Ω is the solid angle subtended by the electron detector.2.1. Postedge Intensity Analysis. In general, the absor-

ption cross section, in addition to depending on energy,depends on the angle δ between the vector matrix element, ortransition dipole moment (TDM) of the final orbital state of theexcited atom and the electric field vector of the X-ray beam. Forσ* final states, the TDM points along the bond axis, but for π*

final states, the TDM points in the same direction as the p-component of the final state orbital. For the transitions to thesestates,

σE ¼ σðhνÞ cos2ðδÞ ð7Þ

where hν is theX-ray energy. At energies far above the absorptionedge and any π* or σ* final state, the photoelectron is ejected intothe continuum, and any dependence on the directionality of themolecular orbitals is lost such that

σE ¼ σcontðhνÞ ð8ÞAt a 390 eVX-ray energy, well above the absorption edge, eq 6

can be simplified as

I390 ¼ Ω

4π

I0A0

sinðθÞσcontð390 eVÞZ ¥

0

dznvðzÞe-

�z=½λ cosðφÞ cosðωÞ�

�

ð9Þsince all carbon atoms contribute to the photoelectrons beingejected into the continuum. In addition, if nv(z) is constant as afunction of depth,

I390 ¼ Ω

4π

I0A0

sinðθÞ σcontð390 eVÞnv cosðφÞ cosðωÞλ ð10Þ

and for such a sample, a plot of (I390/I0)[sin(θ)/cos(ω)] versuscos(φ) must yield a straight line through the origin with a slope of(Ω/4π)A0σcont(390 eV)nv. Let us now suppose there is a layer ofthickness t of material with a carbon atom number density, nv,1(i.e., PFPE) on a thick substrate (i.e., HOPG)with a carbon atom

density, nv,2. Integrating eq 9 yields

I390 ¼ Ω

4π

I0A0

sinðθÞ λ cosðφÞ cosðωÞσcontð390 eVÞnv, 2

1-nv, 1

nv, 2

!e-

�t=½λ cosðφÞ cosðωÞ�

�þ nv, 1

nv, 2

24

35 ð11Þ

In principle, dividing the intensity from the layered sample by thatfrom the bare one should allow a determination of λ. Even if thereis a very thin layer of contamination on the “bare” substrate,which is very difficult to avoid given the conditions surroundingthe beamline at a storage ringwith itsmany pumps, fitting data ona plot (I390/I0)[sin(θ)/cos(ω)] versus cos(φ) should allow one toextract values for nv,1 and t/λ for the contaminated substrate andthus the correct values (eq 10) for the truly uncontaminatedsample. If t/[λ cos(φ) cos(ω)], 1, as would be the case when t/λ issmall and cos(φ) is close to 1, we can expand the exponential ineq 11 to yield

I390 ¼ Ω

4π

I0A0

sinðθÞ λ cosðωÞσcontð390 eVÞ"nv, 2 cosðφÞ- ðnv, 2 -nv, 1Þt

λ cosðωÞ þ :::

#ð12Þ

A plot of (I390/I0)[sin(θ)/cos(ω)] versus cos(φ) would thus havethe same slope as for the bare interface but with a negativeintercept. In the limit when λ cos(φ) cos(ω) , t,

I390 ¼ Ω

4π

I0A0

sinðθÞ λ cosðωÞσcontð390 eVÞnv, 1 cosðφÞ ð13Þ

the result is the same as for a film of carbon atom number den-sity nv,1 of infinite thickness. Thus if nv,1 , nv,2 the slope of theplot of (I390/I0)[sin(θ)/cos(ω)] versus cos(φ) will tend toward theone predicted by eq 13 for small cos(φ), giving importantinformation on the carbon atom number density as a functionof depth.2.2. Order Parameter Calculation. By varying the angle

between the TDM for a given orbital transition and the electricfield vector of the polarized X-rays, the orientational orderparameter, S, of the TDM can be determined. When the electricfield vector is parallel to the TDM, a maximum in peak intensityfor that bond is detected. The TDM for a σ-bonded carbon atomlies along the bond axis, while the TDM for a π-bonded carbonatom is perpendicular to the bond axis and parallel to thep orbitals that combine to form the π orbital.28 The orientationdistribution of the TDM can be described by an orientationalorder parameter, S, as described in the Supporting Information(see also refs 29 and 33).

TheNEXAFSspectra need tobenormalized for comparison ofdata at multiple angles and between samples. This is done by firstnormalizing the data by a signal from the electrons emitted from agold grid in the beam after the monochromator, a signal that isproportional to the incident X-ray intensity, I0. This procedureis necessary both because the beam current in the storage ring isdecreasing slowly over time, decreasing the intensity of the peaksfor each transition, and because the beam intensity after themonochromator varies with energy. The next step usually is to do

(33) Li, X.; Andruzzi, L.; Chiellini, E.; Galli, G.; Ober, C. K.; Hexemer, A.;Kramer, E. J.; Fischer, D. A. Macromolecules 2002, 35, 8078–8087.

DOI: 10.1021/la803951yLangmuir 2009, 25(11), 6341–6348 6343

ArticleSohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

a pre-edge background subtraction so that the baseline is zero,followed by a postedge normalization. This postedge normal-ization ensures that the same number of carbon atoms are beingconsidered in each spectrum. If the number density of carbonatoms, nv, is constant from the film surface throughout thedepth of the sample, the post edge intensity may be arbitrarilyscaled to 1.

However, if the number density of carbon atoms changes as afunction of depth, another normalization procedure is required,as is the case with the model system used here. The data for thePFPE/HOPG are normalized by dividing by the intensity of bareHOPG at 390 eV for each angle. The data for the “bare” HOPGshown later indicate that there is a small layer of contaminationon the HOPG, but it is possible to estimate the values for HOPGwithout contamination using eq 11, and these are the values usedfor the postedge normalization. This normalization methodaccounts for the difference in nv between the PFPE and HOPG.The energy at which the normalization was done was chosen hereto be 390 eV since this is above the very strong EXAFS oscilla-tions from the HOPG that arise from backscattering of theemitted photoelectrons from nearest and next-nearest neighborcarbon atoms.

Whether or not the π or σ bonds are oriented, the postedgenormalized intensity for a sample with uniform carbon numberdensity as a function of depth will have the form

I ¼ A þ B cos2ðθÞ ð14Þfor a given transition where A and B are constants.3,28,29 Theseconstants can be used to determine the orientational orderparameter for a given TDM:

S ¼ B

3AP þ Bð15Þ

where P is the polarization of the X-rays (0.85 for beamlineU7A at the NSLS).29,30,33 St

::ohr and Samant have shown that

fz ¼ AP þ B

PItotð16Þ

where Itot is the total integrated intensity.29 From the orienta-tional order parameter defined in eq S-5 (Supporting Informa-tion), substitution and rearrangement yield

A ¼ 1

3ð1-SÞItot ð17Þ

B ¼ SPItot ð18ÞSubstituting these values for A and B into eq 14 yields

I ¼ Itot1

3-S

3þ SP cos2ðθÞ

� �ð19Þ

Substituting these values into eq 6 results in a non-postedgenormalized intensity throughout the depth of the film for a giventransition:

I ¼ Ω

4π

I0A0

sinðθÞZ ¥

0

dznvðzÞσðhνÞ 1

3-SðzÞ3

þ SðzÞP cos2ðθÞ� �

e-z=½λ cosðφÞ cosðωÞ� ð20Þ

where S(z) is the depth dependent order parameter, and λ isthe EED.3 The term nv(z) enters into the equation to account forany variation in the fraction of carbon atoms associated with thetransition and the total carbon atom number density throughoutthe depth of the sample.

3. Experiment

3.1. Preparation of PFPE Films on HOPG. Fomblin Z-03, with a molecular weight of 4 kg/mol, was supplied by SolvaySolexis and used as received. Thin films were dip coated ontoHOPG using dilute solutions (0.04-0.16 wt %) of PFPE inperfluorohexane (Aldrich). Films were dip coated at a rate of 10mm/min using the dip-coating motor from a NIMA Langmuir-Blodgett trough (Coventry, U.K.). The film thickness was mea-sured using X-ray reflectivity and ellipsometry by a methodsimilar to that of Toney et al.32 To prevent hydrocarbon con-tamination from the atmosphere as much as possible, sampleswere stored under high vacuum until characterization. Thechemical structure of Fomblin Z-03 is shown in Figure 2.

3.2. Ellipsometry Measurements. The thickness of thePFPE layer was determined using a Beaglehole Instruments Pico-meter Ellipsometer (Wellington,NewZealand) with aHeNe laser(λ=632.8 nm) that works on the basis of phase modulation.35,36

The value of the refractive index used for FomblinZ-03 is the bulkvalue, n=1.3.32 The thickness of the PFPE layer was determinedusing a previously published method.36

3.3. XPS Characterization. XPS measurements were per-formed using a Kratos Axis Ultra (Manchester, U.K.) withmonochromated aluminum KR radiation at 1486.6 eV. Lowenergy electrons were injected from a filament to provide chargecompensation to the sample. Spectrawere collected at 225Wwitha 40 eV analyzer pass energy. Three scans were taken for eachsample with 0.05 eV resolution and a dwell time of 300 ms. Thesamples canbe rotatedwith respect to the electron energyanalyzerto change the electron emission angle and change the samplingdepth of the film.34

3.4. NEXAFS Spectroscopy Measurements. NEXAFSspectroscopy was performed at the NIST/Dow endstation ofbeamlineU7A at the National Synchrotron Light Source (NSLS)at Brookhaven National Laboratory. The endstation has agoniometer that varies the orientation of the sample with respectto the incoming X-rays. The PEY was recorded with a channel-tron electronmultiplierwith a variableEGB.For allmeasurmentsreported below, the EGB was set to-150 V. Data were collectedin 0.2 eV steps from 250 to 360 eV and in 0.5 eV steps from 360 to410 eV.

4. Results and Discussion

4.1. Determination of the EED.Variable angle ellipsometrywas used to measure the thickness of the PFPE layers on HOPG.The thicknesses of the PFPE samples reported here are 0.88 and0.9 nm. XPSwas used to confirm the presence of the PFPS on theHOPG after dipcoating. The XPS spectra showed peaks at theexpected binding energies, which correspond to the bonds present

Figure 2. Chemical structure of Fomblin Z-03.

(34) Andruzzi, L.; Hexemer, A.; Li, X. F.; Ober, C. K.; Kramer, E. J.; Galli,G.; Chiellini, E.; Fischer, D. A. Langmuir 2004, 20, 10498–10506.

(35) Toomey, R.; Mays, J.; Tirrell, M. Macromolecules 2004, 37, 905–911.(36) Stroumpoulis, D.; Parra, A.; Tirrell, M. AIChE J. 2006, 52, 2931–2937.

DOI: 10.1021/la803951y Langmuir 2009, 25(11),6341–63486344

Article Sohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

in the PFPE. Representative spectra are shown in the SupportingInformation.

NEXAFS spectra were recorded at the carbon K-edge over alarge angular range (20�< θ<125�) and were normalized by theintensity of the incoming X-ray beam, I0. A linear backgroundwas then subtracted based on the slope of the pre-edge region.Since the background and aswell as this slope according to theoryscale as cos(φ)/sin(θ), we find the best straight line fit to a plot ofbackground slope versus cos(φ)/sin(θ) that passes through theorigin and then use the best-fit slopes at each angle to subtract thebackground. Because of the difference in carbon atom densitybetween the HOPG substrate and the surface layers, no postedgenormalization has been performed. Figure 3 shows the HOPGspectra after division by I0 and the background subtraction.Figure 4 shows the PFPE/HOPG data after the same treatment.

The postedge intensities provide information about theconcentration of carbon atoms within the sampling depth ofthe film. Figure 5 shows the intensity at 390 eV for the “bare”HOPG, the PFPE-coated HOPG, and the hypothetical intensityfor clean HOPG as a function of cos(φ). The data for the PFPE-coated HOPG for cos(φ) close to 1 can be fitted with a straightline with a negative intercept, as expected from eq 12. From themass densities of PFPEandHOPG, a value of y= nv,1/nv,2=0.16was determined and, using eq 11, h, where h = t/[λ cos(ω)], wasvaried to produce the best fit to the data. The values of h = 0.51

of the 0.88 nm thick PFPE film and h = 0.61 for the 0.9 nmthick PFPE film, determined by ellipsometry, were used todetermine a value of the EED of λ= 1.95 ( 0.15 nm. This valueof the EED can be compared with an estimated value for theIMFP of 0.6 nm using the method developed by Cumpson for anelectron with a kinetic energy of 263 eV in PFPE, which isreasonable considering the previously mentioned fact that thechanneltron detector, with an EGB = -150 V, does not dis-criminate between Auger electrons that have lost less than-113 eV due to inelastic scattering and those that have not beinelastically scattered.3,30,31

However, it is clear from Figure 5 that the bare HOPG, whichwas exposed to the air at the NSLS for about 1 min between thetime the surface was prepared by cleaving and the time it wasloaded into the vacuum chamber, also has a layer of materialcontaining a lower carbon atom density on the HOPG. Otherexperiments in which the HOPG was exposed to the air of theNSLS for longer times show an increase in the amount ofcontamination. Fitting the data to eq 11 results in a best fit (solidline) with y = 0.3 and h = 0.25. The contamination layer has alower carbon atom density (typical for hydrocarbons) than thePFPE and a thickness of about 0.5 nm. Note also that the cleanHOPG line can be closely approximated by a line with the initialslope of the “bare” HOPG that passes through the origin ofFigure 5, as expected from eq 12.

In order to determine the order parameter for a given transi-tion, the intensity of the peaks for that transition must be treatedin a similar way to the postedge intensities. Equation 21 gives theintensity for the C 1s f πCdC

* transition:

I

I0

sinðθÞcosðωÞcosðφÞ e

t=½λ cosðωÞ cosðφÞ�

¼ ΩA0

4πλnπ/

2σπ/

1

3-Sπ/

3þ Sπ/P cos2ðθÞ

� �ð21Þ

Figure 6 shows the left side of eq 21 as a functionof cos2(θ). Themain difference between these data and those shown for thepostedge is that these data only take into account carbon atoms

Figure 3. Selected NEXAFS spectra for bare HOPG after back-ground subtraction. Spectra were taken in 10� intervals from θ=28� to 98� and 5� intervals from θ= 98� to 128�.

Figure 4. Selected NEXAFS spectra for PFPE coated HOPGafter background subtraction. Spectra were taken in 10� intervalsfrom θ= 28� to 98� and 5� intervals from θ= 98� to 128�.

Figure 5. Intensity at 390 eV after background subtraction forcleanHOPG, “bare”HOPG, and 0.88 nm and 0.90 nm thick PFPEfilmsonHOPG.A0.44nm-thick layer of contamination containingcarbon atoms is responsible for the offset in the HOPG data. The hvalue is a normalized thickness of the overlayer on the HOPGdefined as t/[λ*cos(ω)], while the y value is ratio of carbon atomdensity of the overlayer to that of theHOPG. The experimental un-certainty is estimated tobeapproximately the size of thedatapoints.

DOI: 10.1021/la803951yLangmuir 2009, 25(11), 6341–6348 6345

ArticleSohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

that have π bonds, whereas the postedge data represent all of thecarbon atomswithin the sampling depth. The postedge data showthat there is a thin contamination layer that will attenuate thesignal for the C 1sf πCdC

* transition from the HOPG, and this isaccounted for by adding in an exponential term, eh/cos(φ), whereh is the same as that used to fit the postedge data. Fitting a lineto these data, whereA is the intercept of the fit and B is the slope,allows one to determine the order parameter for the transi-tion using eq 15. Using the fit parameters, an order parameterofSCdC=0.959 is calculated, which is close to the expected orderparameter of SCdC = 1 for perfectly oriented HOPG. If thecontamination contains any CdC bonds, they will contribute tothe signal, but will almost certainly not have the same orientationof the HOPG, and can be the cause of the slightly decreased orderparameter.32 The order parameter for the C 1sf πCdC

* transitioncan also be calculated for the PFPE on HOPG sample. The onlydifference is that there is a thicker layer attenuating the sample,due to the PFPE layer. Figure 7 shows the data for the C 1s fπCdC* transition as a function of cos2(θ); the order parameter is

SCdC = 0.68, indicating that there is contamination on thesample that contain CdC bonds, which decrease the orderparameter from the expected value of 1 for HOPG.

4.2. Measurement of a Fluorinated Surface Layer in a

Homopolymer Film. A model system for depth profiling withNEXAFS is a polystyrene homopolymer with perfluoroalkyl sidechains linked at the para position of the phenyl ring by an etherbond and amethylene group (PFPS). Because of the lower surface

energy of the PFPS side chains, these are expected to segregateto the surface, producing a thin surface layer enriched withperfluoroalkyl bonds. Analysis of the postedge intensities allowsfor determination of the thickness of this layer.

Figure 8 shows spectra for the PFPS after the pre-edge baselinesubtraction. The thickness of the perfluorinated surface layer canbe determined using the same fitting method as was used todetermine the thickness of the contaminated layer on the HOPG.Figure 9 shows a plot of (I320/I0)[sin(θ)/cos(ω)] vs cos(φ), wherethe dashed line represents the predicted values for a polystyrenehomopolymer. For the carbon density of the side chains, the valuedetermined by Starkweather for amorphous polytetrafluoroethy-lene (PTFE)was used, and the density of polystyrene was used forthe underlayer.37 Using the density of PS is not exact becausetherewill be some side chains fromburiedpolymer chains thatwillchange the carbondensity.Using these values for the carbonatomdensities, a thickness of 0.78 nmwas determined for the side-chainrich surface layer. It should be noted that this layer thickness justcorresponds to the layer of lower density, and probably includessome of the phenyl rings.

Figure 10 shows the data from the C 1sf πCdC* transition and

calculated values for “clean” polystyrene as a function of cos(φ).

Figure 8. NEXAFS spectra for PFPS at various incident X-rayangles after pre-edge subtraction.

Figure 9. Intensity of the PEY signal at 320 eV after backgroundsubtraction. Fitting the data shows that the low carbon atomdensity surface layer is 0.64 nm thick.

Figure 6. Intensity of the C 1sf πCdC* transition, which is used to

determine the order parameter, SCdC = 0.959.

Figure 7. Intensity of the C 1sf πCdC* transition for the 0.88 nm

PFPE on HOPG. The order parameter is SCdC = 0.68.

(37) Starkweather, H. W. J. Polym. Sci., Part B: Polym. Phys. 1982, 20,2159–2161.

DOI: 10.1021/la803951y Langmuir 2009, 25(11),6341–63486346

Article Sohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

Fitting the data yields a value of h=0.28, which corresponds to alayer thickness of 0.45 nm. This differs from the thickness of thefluorinated layer determined from the postedge data because thatoverlayer probably contains someof the phenyl rings,whereas thelayer measured here is the layer covering the phenyl rings. Itmakes sense that the overlayer thickness determined here is lessthan that determined from the data at 320 eV. However, there isan additional layer of contamination on the surface of the samplethat can be inferred from the results shown in Figure 11.

Figure 11 shows the data for the intensity of the C 1s fσC-F* transition as a function of cos(φ). If the side chains were

truly at the surface, the data should fit on a line that extrapolatesto zero; however, it is clear that this is not the case, indicating thatthere is some contamination on the surface that attenuates thesignal. In order for the data shown inFigure 11 to fit on a line thatpasses through the origin, an exponential term accounting for theattenuation of the signal with h=0.19 was required, correspond-ing to an overlayer thickness of 0.30 nm. It is unexpected that thiscontamination would be present for a sample with a low surfaceenergy, but it is clear from the data that it is there. The most likelysource of this contamination is from hydrocarbon or siliconepump oil that is in the air in any laboratory and difficult to avoid.

The orientational order parameter for both transitions can bedetermined using eq 15. Both sets of data are shown in Figure 12.The orientational order parameter for the C 1sf πCdC

* transitionis SCdC = - 0.039, indicating a very slight orientation of thephenyl rings perpendicular to the surface. The fitting parametersgive an order parameter of SC-F = 0 for the C 1s fπC-F* transition, indicating that there is no orientation of the

C-F bonds. If the side chains were fully extended, they would

have a length of ∼1-1.4 nm, and the bonds would be oriented.The calculated overlayer thickness of 0.45 nm indicates that theside chains are not fully extended, which is in agreement with thelack of orientation of the C-F bonds.

For comparison, Figure 13 shows theNEXAFS spectra for thispolymer after the postedge data at 320 eV were normalized to 1,which disregards any change in carbon atom density within thefilm thickness. Figure 14 shows the intensities for the twotransitions of interest from the postedge normalized spectra. Bylooking at the data for each transition as a function of cos2(θ), the

Figure 10. Intensity of the C 1s f πCdC* transition for the PFPS

and clean polystyrene.

Figure 11. Data for the C 1sf σC-F* transition with and without

the exponential correction accounting for attenuation of the signaldue to surface contamination. Figure 13. Postedge-normalized spectra where the difference in

carbon atom density is ignored.

Figure 14. Intensity of the C 1s f πCdC* and f σCdF

* transitionsfrom the postedge normalized spectra. Fitting the data for the C1sf πCdC

* yields an overlayer thickness of 0.65 nm, which is largerthan the value calculated using data that are not postedge normal-ized.

Figure 12. Intensity of the C 1sfπCdC* and σC-F

* transitions as afunction of cos2(θ). The order parameters were determined usingeq 15.

DOI: 10.1021/la803951yLangmuir 2009, 25(11), 6341–6348 6347

ArticleSohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y

density variation cannot be ignored. The C 1sf πC-F* transition

data from the postedge normalized spectra do not give anyindication of a surface contamination layer, but it is clear fromthe data shown in Figure 11 that the contamination is present.Fitting the data for the C 1sf πCdC

* transition gives a overlayerthickness of 0.65 nm, which is significantly larger than the0.45 nm value determined from the data that were not postedgenormalized.

As mentioned previously, Genzer et al. used fluorinatedSAMs to experimentally determine the EED by varying theEGB.30 They tracked the C 1s f σC-H

* PEY signal from themethylene units connecting the perfluoroalkyl chains to the silaneunits that anchored the SAMs and calculated an EED ofλ=2.43 nm for an EGB of-150 V. The presence of contamina-tion is the most likely reason for the difference between the EEDcalculated here and that of Genzer et al. The contamination layerwill have the effect of increasing the EED that they determined,which is indeed the case when the value is compared with theresults here.

5. Conclusions

Depth profiling experiments using NEXAFS spectroscopywere performed on PFPE layers dip coated onto HOPG in orderto experimentally determine the EED for NEXAFS using thepostedge PEY intensity, which was possible because the PFPElayer thickness was measured using variable angle ellipsometry.Fitting these data on a graph of (I390/I0)[sin(θ)/cos(ω)] vs cos(φ)lead to the determination of an EED of λ = 1.95 nm. A newmethod for determining the orientational order parameter, S, wasdeveloped for data that had not been postedge normalized. Thisanalysis method also shows that, because of the extreme surfacesensitivity of NEXAFS, care must be taken to minimize samplecontamination. Data from a “bare” HOPG substrate that wasexposed to the air for about 1 min before loading into the samplechamber and being pumped under high vacuum showed a thinlayer of hydrocarbon or siloxane contamination.

Using the postedge values that have only been normalizedby the incident beam intensity is a way to determine the numberdensity of carbon atoms as a function of depth. This isvery important since accurate fitting of the data is required todetermine the thickness of surface layers. Previous work has

assumed that the carbon density remains constant throughoutthe thickness of a polymer film, but here it is shown that this is notthe case.3

Analysis of the postedge intensity as well as the intensity forthe C 1s f πCdC

* transition for a polystyrene homopolymerfunctionalized with a perfluoroalkyl side chain shows a surfacelayer of lower carbon atom density than the bulk of 0.78 nmthickness, which is primarily due to the surface segregation of thelower surface energy side chains. An orientational order para-meter of SCdC = -0.039 was determined, indicating that, as theside chains segregate to the surface, they drag the phenyl ringswith them, causing a very slight orientation of the rings perpen-dicular to the sample surface. The data for the C 1s fσC-F* transition show that there is no preferred orientation of

the side chains (SC-F=0), but that even though the fluorocarbonside chains have low surface energy, there is an ∼0.3 nm-thicklayer of surface contamination.

Acknowledgment. We would like to thank Mahesh Padigalafrom Solvay Solexis for supplying the Fomblin Z-03. MichaelToney (SSRL), and Matthew Mate (Hitachi) are thanked fortheir helpful discussions. Prof. Joe Zasadzinski and his graduatestudent, Youngmin Jin, helped prepare the PFPE films onHOPG. Thanks to Marvin Paik and Kristin Schmidt for helpwith the NEXAFS measurements. Alejandro Parra and GilaStein helped with the ellipsometry measurements and anal-ysis. Primary funding was provided by the NSF GraduateResearch Fellowship and secondary funding by the ONR undergrant number N00014-02-1-0170 and the NSF DMR PolymersProgram under grant number DMR-07-04539. This work madeuse of central facilities at the MRL at UCSB, which is fundedby the MRSEC program of the NSF under grant numberDMR05-20415, and from use of the nanofabrication facili-ties, which is supported by the NSF-NNIN under Award No.44771-7475.

Supporting Information Available: Preparation of4-(1H,1H,2H,2H-perfluorodecyl)oxymethylstyrene andPFPS, orientational order parameter, and XPS results. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

DOI: 10.1021/la803951y Langmuir 2009, 25(11),6341–63486348

Article Sohn et al.

Dow

nloa

ded

by N

OR

TH

CA

RO

LIN

A S

TA

TE

UN

IV o

n Ju

ne 2

9, 2

009

Publ

ishe

d on

May

26,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

la80

3951

y