硬碳薄膜中氢含量，薄膜厚度等物理性质检测方案(激光拉曼光谱)

智能文字提取功能测试中

Carbon 02 Derivation of Physical Parameters from RamanSpectra of Hard Carbon Films The Raman：spectra of elemental carbonmaterialsareknown tobeesensitiveetopolymorphy. For hardcarbonfilms.，thespectraoff amorphous andddiamond-likecarbons can be band-fit to separate thecontributions of the“graphitic carbon" (G band)from the"disordered carbon" (D band). Figure1 shows a typical spectrum of a hard carbonfilm. The spectral behaviour of carbon filmshas been empirically correlated with thin filmphysicalpropertiessuch as hardness.durability，opticallttransparency，electricalconductivity，'， thermal conductivity andcorrosion resistance， and can be of use forprediction of these properties without extensivealternative testing. In recent!yearsthe computer hard discmanufacturing industry has implemented hardcarbon coatings on all disc media as a meansof protecting the magnetic media with hard，non-brittle films.i. The coatings provide wearprotectionagainst head sliders repeatedlydragging and slapping on the disk surfaceduring the start and stop cycles of normaloperation. Usually the coatings contain some amount ofhydrogen which issadded to improve thecorrosion protection of the underlying magneticlayer， toincreasethe filmhardnessforimproved wear resistance， and to optimiseinteraction withthe lubricants； in order toeliminate friction problems. New films underdevelopment often contain nitrogen as well inorder to reduce the surface resistivity andhence static charging without compromisingfilm hardness. Combinations of hydrogen andnitrogen inn thelayerscanimprove tthecompatibility of the magnetic data storagemedia with different types of read/write heads. The DiskRam has been engineered to facilitateand significantly accelerate the acquisition ofthe Raman spectra of hard carbon overcoatson disk media. Subsequent data reduction automates the derivationof tthee physicalpropertiesof interest. Curve-fitting thespectrum with two carbon bands (the D and Gbands)， additional nitrogen bands if required，and a baseline achieves these goals.Themethodology used for the spectral reductionwill be reviewed in this note. Figure 1. Raman spectrum of Carbon Filmrecorded and band-fit on the DiskRAM Description Of Raman Spectra The Raman spectrum of an amorphous pure，carbon film(as well as films doped withhydrogen and/or nitrogen) is composed of theD and G bands sitting on top of a broadluminescent baseline. The G band usually occurs between 1480 and1580 cm-1， while the D band position appearsbetween 1320 and 1440 cm-1. The bands areusually overlapping and the actual positionsare to some degree dependent on the laserexcitation wavelength. Using red excitation such as 632.8 nm of theHelium Neonn laser resultsinna smallerseparation of the two spectral components andconsequently in more ambiguity of the curve fitresult. To increase the spectral separation ofthe D and G bands， it is common to use anexcitation wavelength in the green. For this the532 nm line of a diode-pumped， frequency-doubled， NdYag laser or the 514.5 nm line of an Argon ion gas laser are employed. TheNdYag laser is the more energy efficient of thetwo lasers and can be run without externalcooling fans. This can be a valuable feature，when cleanroom compatibility is of importance. More recent experiments also focus on the useof the blue Argon ion line at 488 nm， whichresults in an even better spectral resolution ofthe two carbon bands. Films containingnitrogen show an additional weak band at 2180cm-1， corresponding to a nitrile CN triple bondstretching vibration. Over a definedconcentration range of the nitrogen in thesputter atmosphere and hence in the depositedcarbon film， there is a linear response in theintensity of the band at2180)cm-1.Mathematical methods for extracting nitrogenconcentrations from the Ram an signatures arecurrently under development. Nitrogen-doping of the carbon films causes asecond， less visible effect on the Ramanspectra. While the peak position of the D bandremains relatively stable withincreasingnitrogen concentration in the film， the G bandposition decreases significantly， i.e. the twoRaman bands increasingly overlap with highernitrogen concentration. While the G bandposition principally represents a fairly sensitivemeasure of the nitrogen concentration， otherfilm processingparameters，，mainly thesubstrate temperature during film deposition，also influences it. Films containing constantamounts of hydrogen or nitrogen show a nearlylinear relationship of the G peak position withdeposition temperature [2]. A linear regression of the experimental datayields a variation of ca.1.0.l cm-1//o°C.Furthermore. Electron Energy LOSSSpectroscopy (EELS) data indicate that filmssputtered at higher temperature (higher G peakposition) have lower sp3 and higher sp2carbon content. It has been shown by abrasionand contact start/stop measurements， thathigher sputter temperatures yield films withdecreasingmechanical performance，whichcorrelates with a higher sp2 carbon (i.e.，softer)content. Figure 2. Relationship between Raman spectraand excitation wavelength as determined byfluorescent background With the curve fitting techniques commonly inuse it is almost impossible to decouple thedifferent parameters influencing the Ramanpeaks positions， such that they can be used tosimultaneously monitor nitrogen content of thefilmand substrate temperature) duringdeposition. In a recent research effort， weapplied advanced mathematical methods toanalyse the response of the Raman spectra asafunction of the variation in processingparameters as a whole. Initial results of these Chemometrics methodsare very conclusive， highlighting the use ofsuch techniques for more sophisticated dataanalysis (fig.3). The nitrogen content can bedetermined with an absolute error of ±0.1%within a range of 7 to 24 %. In addition， anoverall better accuracy of quantitative resultsfor other parameters can be obtained by usingmultivariate data analysis since more spectralinformation is taken into account. Thus， the thickness of a nitrogenated overcoatwithin a range of 85 to 116 A can be derivedwith an absolute error of ± 0.75 A all over thisrange. The improved accuracy of tlthesemeasurements is of prime importance due tothe continuously decreasing size ofthesestructures. Figure 3. Results of multivariate analysis Carbon films preparedin ahydrogen- ormethane-containing atmosphere ：show aphotoluminescence background in addition tothe Raman bands. The photoluminescence is due torecombination of electron hole pairs within spbonded clusters in an sp° bonded amorphousmatrix. The intensity of the photoluminescencetends to increase with increasing hydrogencontent in the film， due primarily to thesaturation of non-radiative recombination sites[2]. A spectrum acquired using blue 488 nmexcitation shows the luminescence to peak atabout 600 nm. Using red excitation at 632.8nm places the Raman spectrum close to thetop of the broad luminescence peak， resultingin essentially a flat background. Refer to Figure2. For a film of given thickness the intensity ofthis luminescent background is proportional tothe amount of hydrogen in the film. However toavoid the influence of the film thickness on themeasure ofthe rhydrogen content，it isadvisable to use a laser line in the green orblue， such that the Raman spectrum is on theflank of the luminescence peak. By exciting with a laser wavelength on therapidlyrisingorfallingg slopeetheofluminescence peak， theeslopeetheluminescentbackground canbe used topredict the amount of hydrogenation. Theslope， as opposed to the total intensity of thebackground， is independent of the carbon filmthickness. Bandfitting The Raman Spectrum For the Raman analysis of the hard carbon filmproperties one usually limits the spectral regionof analysis to the range between 800 and 1950cm-1 (for nitrogenated films up to 2250 cm-1).The spectra can be fitted with Gaussian orLorentzian band profiles， or aGaussian/Lorentzian sum function (usually theGaussian/Lorentzian sum function works best).To fit the baseline the DiskSpec softwarepermits one to use a polynomial of first，secondorthird order(to accommodate smalldeviations of the baseline from linear， it is bestto use a polynomial of second order). Once therecipe for the curve fitting is defined in thesoftware， all subsequently collected spectrawill be submitted to this curve fitting routine.The parameters derived from the curve fitting，such as peak intensities， widths， positions andbackground slope， can be automatically stored in a spreadsheet matrix format. Alternatively，the DiskSpec software allows the data streamto be routed to an external terminal forstatistical process control (SPC). Deriving Hydrogen Content There are mainly two different effects of anincrease in hydrogen content on the Ramanspectral response. For hydrogenconcentrations between 20 and 40 % in thecarbon film， the photoluminescence peakaround 600 nm increases exponentially. Withgreen or blue excitation， the Raman spectrumis superimposed on the rising flank of this verybroad luminescence background. The slope ofthis background canbe used asa verysensitive measure for the hydrogen content inthe carbon film. Although the： luminescent background andhence the spectral baseline is not always linear，the slope can be estimated in first instance byselecting two frequency positions that span theD and G bands (for instance， 800 and 1950cm-1) and defining a slope derived from theintensities， y1 and y2， at these frequencies.The user has the ability to freely select the waythe slopeeisdefined for correlationn withhydrogen content. Possible slope definitionsare： In any case it is important to work with relativeintensities， rather than with absolute values，since absolute values are dependent oninstrumental and sample parameters such asthe laser power， the optical alignment， theintegration time and the carbon film thickness.Using relative intensities cancels these factorsout， since all spectral intensity values will beequally influenced by the above parameters.Note that the classical definition of the slope is(y2-yl)/(x2-xl)1)(where xi is theRamanwavenumber value) is not an effective way todefine the slope for this application becauseany instrumental drift in intensity will effect thespectral intensities； (y values)， but not thespectral frequencies (x values). It has beenfound that if the natural logarithm of the slope，calculated from one of the definitions above，isplottedVs. % hydrogenncontent.，1therelationship is near linear between 20 and40 % film hydrogen content. For films with ahydrogen content of less than 20 %， theintensity ratio of the D to G bands (ID/IG) can be empirically correlated to the hydrogenconcentration. For those manufacturers whoare interested in predicting hydrogen contentfrom 0 to 50%， methods combining slope andratio are potentially useful. Calibration Transfer Forr manufacturersPperforminggSPCcwithseveral DiskRams it is sometimes necessary tocorrect for small instrument-to-instrument. variations， in order to assure that all DiskRamsgive the same results for the same sample.Instrumental variations are unavoidable andwill happen even though all opticalcomponents are "identical". It is not possible toacquire exactly identical components! TheDiskSpec software allows for the introductionof standardisation coefficients. In order toderive these coefficients， we suggest analysingon all DiskRam tools a series of three or morediskssshowing，a widevariationiinntheparameter to be measured. When plottingactual versus measured parameters， everyinstrument can be standardised to give thesame result for the same sample. Most of thetime this can be done with a simple correction，such as the inclusion of a factor to correct theslope and a coefficient to correct for the offset.Refer to Figure 4. Figure44..Slope tfactor vs. %hydrogenrecorded on 3 different DiskRAM tools. HORIBAJOBIN YVON Film Thickness The film thickness can be derived from theintensity of the G band normalised for theintegration time. In order to compensate forinstrumental effects， the G band intensityshould furthermore be ratioed to the intensityof a reference sample. The DiskSpec softwareallows for a separate entry of spectra collectionand curve fit parameters for reference samplesdifferent from carbon films. It is advisable touse an alternative to carbon films as referencesince the amorphous carbon tends to adsorbcomponents from the ambient atmosphere，resulting in spectral changes over time. It issuggested to use a silicon wafer cut andpolished to the same dimensions as the diskmedia as a reference that does not change itsspectrum in time. The frequency of thereference measurements can be defined in thesoftware by the timeout between references or，alternatively， by the number of actual carbonfilmnmeasurements betweenreferences.Whenever one of these two conditions is met.the operator is prompted to make a referencemeasu；urreemment. By comparing the actualthickness of a well-characterised carbon filmwith that measured， a correlation factor can bedetermined. Subsequently the system is ableto measure the film thickness of unknownproduction samples in A's. Conclusion The DiskRam has been designed to automatethe collection of Raman spectra from hardcarbon coatings on computer hard disk mediaand the extraction of parameters that are wellcorrelated with the properties of the films. Theextracted information is output in spreadsheetformat for SPC at a manufacturing facility. 1. Impactt otfsRaman sSpectroscopyonTechnologically Important Forms of ElementalCarbon， Jobin Yvon application note # A.N.082. Photoluminescence and Ramanspectroscopy in hydrogenated carbon films， B.Marchon， J. Gui， K. Grannen， G. Rauch， J.Ager， S. Silva and J. Robertson， to bepublished intheIEEE Transactions onMagnetics (1997) ( France： H ORIBA Jobin Yvon S.A.S.， 231 rue de Lill e ， 59650 Villeneuve d'Ascq. T e l：+33 (0)3 2059 18 00， F ax ： +33 ( 0)3 20 59 18 08. Email ： raman@jobinyvon.fr www.jobinyvon.fr ) ( USA： H ORIBA Jobin Yvon Inc.， 3880 Park Avenue， Edison， NJ 08820-3012. Tel：+1-732-494-8660， Fax：+1-732-549-2571.Email ： raman@jobinyvon.com w ww.jobinyvon.com Japan： H ORIBALtd.， JY Optical Sales Dept.， 1 - 7-8 Higashi-kanda， Chiyoda-ku， Tokyo 101-0031 Tel： + 81 ( 0)3 3861 8231， Fax： +81 (0)3 3861 8259 . Email： raman@horiba.com Germany： +49(0)62518475-0 I taly： +39 02 57603050 UK： + 4 4 (0)20 8204 8142 ) China： +86(0) 10 68492216 ORIBAExplore the future The DiskRam has been designed to automate the collection of Raman spectra from hard carbon coatings on computer hard disk media and the extraction of parameters that are well correlated with the properties of the films. The extracted information is output in spreadsheet format for SPC at a manufacturing facility.