Volume 58, Number 12, 2004 APPLIED SPECTROSCOPY 14750003-7028 / 04 / 5812-1475$2.00 / 0q 2004 Society for Applied Spectroscopy

Identification and Measurement of Dirt Composition ofManufactured Steel Plates Using Laser-Induced BreakdownSpectroscopy

DANIEL J. O. ORZI and GABRIEL M. BILMES*Centro de Investigaciones Opticas (CONICET-CIC) y Universidad Nacional de La Plata, Casilla de Correo 124, (1900) La Plata,Argentina

Laser-induced breakdown spectroscopy (LIBS) was used for thecharacterization of the main components of the surface residual dirtproduced in cold-rolled steel plates as a consequence of the manu-facturing stages. At laser fluences between 0.05 J/cm2 , F , 0.30J/cm2, dirt ablation takes place without any contribution from thesubstrate. Results show that the main components of the dirt arefine particles of Fe mostly homogeneously distributed in a thin layerof grease and soaps. In the primary stages of the manufacturingprocess carbon residuals can also be found. By measuring lightemission from the l 5 495.9 nm line of Fe(I) after laser ablation,we developed a real-time on-line method for the determination ofthe concentration of iron particles present in the surface dirt. Theobtained results open new possibilities in the design of real-timeinstruments for industrial applications as a quality control of prod-ucts and processes.

Index Headings: LIBS; Laser-induced breakdown spectroscopy;Surface dirt; Cold-rolled steel plates; Laser ablation; Laser-inducedplasmas; LIPS.

INTRODUCTION

Laser-induced breakdown spectroscopy (LIBS) wasdemonstrated as a useful method for qualitative and quan-titative microanalysis in a variety of matrices in the solid,liquid, and gaseous phases.1–4 As is well known, LIBS isbased on laser ablation of the sample by means of a highfluence laser pulse to produce a plasma. Measuring theemission of the species that make up the plasma, ele-mental analysis of the sample can be performed. For areal-time analysis system, advantages of particular rele-vance of LIBS over more conventional methods are thelack of pre-treatment, single-step excitation of the sam-ples, and the possibility of real-time in situ analysis. Forthese reasons, LIBS is a very attractive technique for on-line industrial analytical quality control that requires sim-pler, cheap, fast, and reliable qualitative and semi-quan-titative analysis.

Various LIBS applications to the steel industry (iden-tification of product composition in solid and moltenphases, remote analysis, carbon content, and quality con-trol of partial processes) were developed during recentyears, providing new tools for real-time and productionline analysis.5–11

Surface dirt identification, measurement, and control inthe millimeter and micrometer scale are problems of greatimportance in many industrial and technological process-es. In the case of metallurgy, during cold-rolling steel

Received 4 May 2004; accepted 29 July 2004.* Author to whom correspondence should be sent. E-mail: gabrielb@

ciop.unlp.edu.ar.

plates manufacturing processes surface dirt may appearas a thin film of grease and solid particles. This dirt isdue to possible malfunctioning of prior stages of the pro-cess and could affect downstream treatments such aspainting.

Existing methods for the identification of the compo-sition and for measurement of the amount of dirt arebased on laboratory procedures. Most of them rely onchemical analysis of the residues. All these methods havethe following drawbacks: they are performed off-line andnot in real time, they are time-consuming, they cannot beautomated, and usually they are restricted to a small por-tion of the sample. Recently12–14 we developed and pat-ented an instrument for the determination of the degreeof cleanliness of cold-rolled steel plates based on laserablation and acoustic detection. The method relies on theablation of the dirt film by means of a short laser pulseand the subsequent measurement of the sound wave emit-ted. The intensity of the sound, proportional to theamount of dirt, provides a direct measurement of thecleanliness of the surface.15

In this paper we show how the measurement of thelight emitted by the plasma produced after laser ablationcan be used for the determination of the constituents ofthe surface dirt. We also present a real-time on-line meth-od for the measurement of the concentration of the maincomponents of the dirt.

EXPERIMENTAL

Figure 1 shows the experimental setup used for laserablation and the analysis of the light emitted by the plas-ma. A Q-switched Nd : YAG laser (Continuum SureliteII) with pulse duration of approximately 7 ns (FWHM)operating at 1064 nm and at 2 Hz repetition frequency isdirected normal on to the sample surface. A neutral den-sity wedge filter allows the energy of the laser pulse tobe changed to give laser fluences (F) ranging from 0 ,F , 4.5 J/cm2. Energy is measured using an energy meterwith a pyroelectric detector (RjP-765, Laser PrecisionCorp.) splitting the laser pulse by means of a beam-split-ter. Taking into account the non-homogeneous spatial en-ergy distribution of the laser pulse and the effective areaof the laser spot, the fluence values calculated in thiswork have an estimated uncertainty of approximately30%.

The samples analyzed were a collection of two typesof steel plates selected from cold-rolled bobbins corre-sponding to two different stages of the production pro-cess. Type two samples (T2) correspond to the end stage

1476 Volume 58, Number 12, 2004

FIG. 1. Experimental setup.

of the production process, and Type one (T1) to a pre-vious intermediate stage. The plates were mounted on amotorized linear translation stage for the measurements.Data was recorded under one of the following conditions:(1) moving the sample perpendicularly to the laser beamat enough speed to assure that the laser beam impingeseach time on a different region of the plate; or (2) thesample is fixed and each shot impinges on the same re-gion of the plate. Under these conditions the ablation pro-cesses can be followed shot by shot.

The spectrum of the plasma was acquired using con-figuration A (see Fig. 1), where the light is focused bymeans of a lens array to a single-mode (0.1 mm opening)fiber optic (FO) attached to a cross dispersion Echellespectrometer (Mechelle Multichannel Instruments) with acharge-coupled device (CCD) incorporated camera. Datawas sent to a computer and processed with appropriatesoftware that gives the whole spectra covering the 350 to1100 nm region with a 0.3 nm spectral resolution. Thiscombination enables simultaneous detection of a largespectral range in a single laser shot. In order to protectthe CCD camera from laser radiation, an infrared (IR)filter is placed between the focusing system and the fiber.

Fluence thresholds for laser ablation and quantitativedetermination of single elements in the dirt were madeusing configuration B (see Fig. 1), in which the light isdetected by means of a 931A-Hamamatsu photomultiplier(PM) and analyzed temporally in a digital oscilloscope.Photomultiplier output was boosted by a preamplifier.This combination has an effective RC time constant ofabout 10 ns. For ablation fluences analysis, the intensityof the light emitted by the plasma was collected directlyby a fiber optic (quartz, 200 mm core) attached to thephotomultiplier. For quantitative determination of single

elements in the dirt, a spectrometer (Jarrell-Ash 82-025,50 cm focal length, Ebert mount) selects only a smallregion of the spectrum around a characteristic line of theelement under study. We call this last configurationMCLE (measurement of a characteristic line emission).

RESULTS AND DISCUSSION

Fluence Values for Dirt Ablation. The analysis of thesurface dirt must be made avoiding any influence of thesubstrate. Then, the first step is to determine the laserfluence values that produce plasma only formed by spe-cies coming from the dirt and not from the steel substrate.For that purpose, the laser fluence must be enough toproduce the ablation of the dirt, but it should be belowthe threshold of ablation and damage of the substrate.

Figure 2 shows the time-integrated light intensity emit-ted by the plasma after the ablation of a cold-rolled steelsample as a function of the laser fluence. This is a typicalablation curve. As can be seen, there is no plasma for-mation below 0.05 J/cm2 (see inset). Between 0.05 and0.30 J/cm2 time-integrated light emission grows linearlywith the laser fluence, and above 0.30 J/cm2 saturationmust be reached and a plateau appears between this valueand approximately 2 J/cm2. Above this last value, lightemission again has a linear dependence with the fluence.

Figure 3 shows the dependence of the time-integratedlight intensity emitted by the plasma as a function of thenumber of laser shots. Each curve corresponds to lasershots of the same fluence, impinging on a fixed point ofthe sample. It can be seen that for a fluence of 0.15 J/cm2, after a certain number of laser shots there is no moreplasma formation. At 0.9 J/cm2 the emission of light cor-responding to the first shot is higher than for 0.15 J/cm2

APPLIED SPECTROSCOPY 1477

FIG. 2. Time-integrated light intensity emitted by the plasma producedafter ablation as a function of the laser fluence. The inset shows the dirtablation fluence region in which there is no contribution of the substrate.

FIG. 3. Time-integrated light intensity emitted by the plasma producedafter ablation as a function of the number of laser shots that impingeon the same point of the sample. (m) Fluence 5 0.15 J/cm2; (V) Fluence5 0.9 J/cm2; and (m) Fluence 5 3.5 J/cm2.

FIG. 4. Spectrum emitted by the plasma generated by ablation of thesurface dirt. (a) Recorded by collecting the light without any spatialselection. (b) Recorded by screening the region of the plasma near thesurface. By screening regions of the plasma, the background radiationis substantially suppressed, improving the line-to-background ratio.

and after that, light intensity decreases shot to shot, butinstead of disappearing, after approximately 50 shots thelight intensity increases and stabilizes. At fluences of 3.5J/cm2, a higher intensity of emission is observed for thefirst shot, and after that, plasma intensity tends to rapidlyreach a constant value.

Microscopic inspection of the steel substrate after ab-lation shows that below 0.30 J/cm2 there is no damageor visible surface modification of the steel substrate. Onthe other hand, from this value on, microscopic inspec-tion clearly shows damage produced by ablation.

The above results can be interpreted as follows: at flu-ences below 0.05 J/cm2 no surface ablation takes place.Between 0.05 and 0.30 J/cm2 dirt ablation takes placewithout any effect or contribution of the substrate (Fig.3, filled squares) and the light intensity of the plasmagrows linearly as a function of the laser fluence (inset inFig. 2). Then, this corresponds to the dirt ablation regionand the fluence threshold for dirt ablation can be esti-mated as 0.050 6 0.015 J/cm2.

Above 0.30 J/cm2 the laser pulse not only ablates thedirt, but part of its energy also interacts with the sub-strate. Surface melting fluence thresholds of the order of0.7 J/cm2 were previously determined for steel by usingacoustic measurements.16 Then, results of Fig. 3 (opencircles) can be explained as follows: at 0.9 J/cm2, afterthe first shot, each new shot not only removes the dirtbut also melts part of the surface, changing its opticalproperties and lowering the fluence threshold for sub-strate ablation. Once the surface is completely cleaned,the fluence is enough to ablate the steel substrate due tothe melting effect. By increasing the fluence, lesser shotsare needed to clean the dirt, and substrate ablation startsto dominate the plasma emission. The plateau regionshown in Fig. 2 represents this behavior.

Above 1.2 J/cm2 substrate ablation is the main processthat contributes to plasma formation. In this region, thefirst shots produce dirt plus substrate ablation and thefollowing shots ablate the substrate (filled triangles, Fig.3). In this region again, the light intensity of the plasma

grows linearly as a function of the laser fluence, as canbe seen in Fig. 2.

We can conclude from this behavior that fluences be-tween 0.05 J/cm2 , F , 0.30 J/cm2 are the best suitedfor LIBS analysis of the composition of the dirt. In thisfluence region each shot of the laser ablates part of thesurface dirt and completely cleans the surface after a cer-tain number of laser shots.

Composition of the Surface Dirt. Figure 4 shows thespectra of the dirt in the region 475–575 nm obtainedimmediately after the plasma formation with no delaybetween the laser shot and the detection of the light emit-ted by the plasma. The spectrum shown in Fig. 4a cor-responds to the case in which all the light is collected.As is well known, the spectra are dominated by the back-ground continuum, which masks the observation of dis-crete lines. To maximize the contrast of these lines, it isnecessary to eliminate or at least minimize the emission

1478 Volume 58, Number 12, 2004

FIG. 5. Spectrum of the plasma produced after laser ablation of thesurface dirt of samples of cold-rolled steel plates corresponding to twodifferent stages of the production process. (a) 375–500 nm spectralregion; (b) 500–775 nm spectral region. T2 samples belong to the endstage of the production process and T1 samples belong to a previousintermediate stage. The Fe(I) line at 495.9 was selected as representativefor quantitative measurements of the presence of this element in thedirt.

of this continuum. Most LIBS works discriminate againstthis continuum emission by using time-gated detectorsystems to register the spectral lines after decay of thebackground continuum. This option is usually expensive,and some authors previously demonstrated17,18 that an al-ternative and effective method is to take in to accountthe spatial dependence of the plasma formed. Then, byscreening part of the plasma emission with the aid of aslit or a similar device the background continuum can beminimized. We used this method to obtain the spectrumshown in Fig. 4b, recorded by screening the region ofthe plasma near the surface and collecting only lightemitted by regions of the plasma far from the surface. Ascan be seen, screening the first part of the plasma sub-stantially improves the line-to-background ratio of thespectrum. Then, having in mind the development of asimple and cheap on-line identification method, we de-cided to use the screening procedure for the identificationof the main elements of the dirt.

For the analysis of the dirt spectrum, reference spectraof pure compounds or of well-known concentrations wereobtained previously under the same experimental condi-tions (i.e., iron, zinc, aluminum, sodium, calcium, oxy-gen, nitrogen, copper, carbon, lead, etc.).

Figure 5 shows the spectra of the laser-induced plasmaof the dirt in the region 375–775 nm (Fig. 5a from 375to 500 and Fig. 5b from 500 to 775 nm) correspondingto the two types of samples analyzed.

The spectral analysis shows lines of Fe(I), Na(I), Ca(I),and K(I) and structures of molecular band emission suchas CN and C2 that indicate the presence of carbon.

These results are in complete agreement with chemicallaboratory analysis provided by the factory, which showsthat the surface dirt generated during the production stag-es of cold-rolling steel is formed mainly from fine par-ticles of iron, oils, carbon residuals, and soap.

As can be seen from the spectra, Fe is present in bothtypes of samples. The fact that we found only lines cor-responding to the neutral atom Fe(I) is attributed to theconditions under which the spectra of the plasma wererecorded. It was previously demonstrated that for lines ofFe(I) the line-to-background ratio is improved substan-tially between 2 and 3 ms of the plasma lifetime.9 Underthe conditions of our experiments, the screening effectacts as a sort of temporal gating since there is a strongrelationship between the temporal and spatial evolutionof the plasma.

Calcium and K are present only in the T2 samples.Since plasmas of T1 and T2 samples where generatedunder the same experimental conditions, we can assumethat this fact is due to the presence of residuals of thecleaning treatment with soaps, which takes place only inT2 samples, during the final stages of the production pro-cess. The presence of Na in both types of samples is aconsequence of the primary stages of the production pro-cess in which oils and environmental contaminants areinvolved.

As can be seen, CN and C2 bands appear only in theT1 samples, indicating the possible presence of carbon inthe dirt, which probably recombines to form these mol-ecules as a consequence of the ablation processes. Thisbehavior should be expected taking into account that T1samples come from a stage in which carbon residuals

were produced by one of the specific treatments and arenot eliminated by the end treatments as in the case ofsample T2.

The obtained results show that the screening procedureseems to be adequate and is a simpler and cheaper meth-od for on-line qualitative determination of the main com-

APPLIED SPECTROSCOPY 1479

FIG. 6. Emission intensity time dependence of the l 5 495.9 nm Fe(I)line. (a) Intensity measured at l 5 495.9 nm (Fe(I) line plus backgroundcontinuum). (b) Intensity measured at l 5 494.5 nm or l 5 497.2 nm(background continuum only). (c) Background free intensity profile ofthe l 5 495.9 nm Fe(I) line, determined by subtracting (b) from (a).

FIG. 7. Emission intensity time dependence of the l 5 495.9 nm Fe(I)line measured by screening regions of the plasma produced after abla-tion of the surface dirt. (a) Screening 1.5 mm from the surface. (b)Screening 2.1 mm from the surface. (c) Screening 2.3 mm from thesurface. (d) Screening 3.3 mm from the surface.

ponents of the surface dirt produced during the differentstages involved in the production of cold-rolled steel.

Quantitative Determination of Fe Fine Particles inDirt. The amount of iron fine particles present in the dirtof cold-rolled steel plates is an important parameter forthe quality control of the different stages of the produc-tion process. Its determination is performed off the pro-duction line by means of a laboratory chemical methodbased on the following procedure: First, the surface dirtof a selected sample is cleaned with successive clothshumidified in methyl ethyl ketone (MEK). After that,these cloths are washed in nitric acid, the resulting so-lution is filtered, and the concentration is determined byusing a Hach DR 2000 spectrophotometer. This methodhas the following drawbacks: It has quality limitationsimposed by the procedure itself; it provides informationonly on a small part of the analyzed steel bobbin and notthe whole piece; and the determination must be made offthe production line and not in real time.

To resolve these problems we developed a new real-time on-line method for the determination of the relativeamount of Fe residuals contained in the surface dirt. Wecall this method MCLE (measurement of a characteristicline emission), and it is based on the measurement of theemission intensity time dependence of one characteristicline of Fe by using the experimental configuration Bshown in Fig. 1.19 For this purpose we choose the l 5495.9 nm line corresponding to Fe(I) and the nearbybackground regions at 494.5 and 497.2 nm, which allowsthe comparison of the behavior of an emission line withrespect to the continuum background. The selected lineof Fe has high emission intensity, appears in a region ofthe spectra with low intensity background radiation, andit is far from lines corresponding to other elements (seeFig. 5).

Figure 6a shows the time-dependent intensity of thelight emitted by the plasma at l 5 495.9 nm [Fe(I) lineplus continuum]. In the same figure the background (Fig.6b) is shown, measured under the same conditions, shift-ing the selected wavelength region to 494.5 nm or to497.2 nm. To avoid laser fluctuations as well as smallinhomogeneities of the surface dirt, each signal was ob-

tained by averaging sixteen different points located indifferent regions of the sample. Figure 6c shows the time-dependent background-free profile of the line of Fe(I)located at l 5 495.9 nm, obtained by subtracting Fig. 6bfrom Fig. 6a.

As was shown in Fig. 4, the signal-to-background ratiocan be improved by screening the light coming from cer-tain regions of the plasma. In order to find the best mea-surement conditions, the time dependence of the lightemitted at l 5 495.9 nm was measured as a function ofthe plasma region. These experiments were carried outby screening regions of the plasma located between 0 and3.5 mm out of the surface. Measurements were per-formed at constant excitation fluence. Figure 7 shows thisbehavior. Taking into account these results, we chose theregion between 2.1 to 2.3 mm for our measurements.

From the above experiments we used the quantity I/Fas a measurement of the relative amount of fine particlesof iron present in the dirt. I is the maximum of amplitudeof the time-dependent background free profile of the l 5495.9 nm Fe(I) line, measured by screening the region ofthe plasma between the surface and 2.2 mm, and F is thelaser fluence.

Figure 8 shows the correlation obtained between ourI/F measurements and the standard laboratory methodalready described for a collection of samples with differ-ent amounts of iron particles in the surface dirt. The insetshows an I/F measurement of one sample and representsa register of sixteen different points in a region 32 cmlong. The white point corresponds to the half value anddispersion of the measurement. As can be seen, a rea-sonable linear correlation can be obtained if one keeps inmind the limitations and uncertainties involved in the lab-oratory method and the possible effects of changes in theplasma emission positions, which our screening config-uration did not take in to account.

CONCLUSION

Laser-induced breakdown spectroscopy (LIBS) allowsthe characterization of the main components of the sur-face residual dirt produced in cold-rolled steel plates asa consequence of the manufacturing stages. Using flu-

1480 Volume 58, Number 12, 2004

FIG. 8. Correlation between LIBS (I/F measurements) and the stan-dard laboratory method used for the determination of iron particle con-centration (Fe) in the surface dirt. Uncertainty in the measurementsperformed with the standard method (not shown) is higher than 50%.Usually only a rough estimation of these uncertainties can be madebecause the procedure involves possible systematic errors that are dif-ficult to evaluate. The inset shows the results of a measurement of asample. The white point represents the half value and its dispersion.

ences between 0.05 J/cm2 , F , 0.30 J/cm2, we dem-onstrated that only dirt ablation takes place, without anyinfluence or effect of the substrate. We found the valueF 5 0.050 6 0.015 J/cm2 as the threshold fluence for dirtablation. For the analysis of the composition of the dirtwe used a plasma screening procedure that allows a sim-pler and cheaper way to optimize the line-to-backgroundratio without the necessity of using a gated delay system.

Results show that the main components of the dirt thatare present in cold-rolled steel plates are fine particles ofFe, mostly homogeneously distributed in a thin layer ofgrease and soaps. In the primary stages of the manufac-turing process, carbon residuals can also be found.

We developed a method for the determination of theconcentration of Fe particles present in the surface dirt,which uses the light intensity of emission of the l 5495.9 nm line of Fe(I), after laser ablation. Concentra-tions of Fe can be determined as relative values or inunits of mg/cm2 by calibration using the linear correlationcurves obtained between our method and the standardchemical laboratory procedure.

Real-time measurement of the elements that are present

in the surface dirt of manufactures represents for industrythe possibility of better quality control of the product; itmay be indicative of deviations from normal functioningin previous manufacturing stages, and it may impair theeffectiveness of downstream operations, such as paintingor galvanizing in the case of steel production. Then,based on the results obtained in this work we designedan on-line instrument for industrial applications. This in-strument identifies automatically and on line the com-ponents of the surface dirt; the amount (in mg/cm2) ofiron present in it; and its dispersion along the inspectedarea. One advantage of the low ablation threshold of sur-face dirt is the possibility of using a fiber optic for laserbeam delivery in a configuration suitable for operating inan industrial environment, allowing better safety condi-tions.

ACKNOWLEDGMENT

We thank Dr. Daniel Schinca for his useful comments on this man-uscript.

1. D. A. Cremers and L. J. Radziemski, ‘‘Laser Plasmas for ChemicalAnalysis’’, in Laser Spectroscopy and its Applications, L. J. Rad-ziemski, R. W. Solarz, and J. A. Paisner, Eds. (Marcel Dekker, NewYork, 1987).

2. D. A. Rusak, B. C. Castle, B. W. Smith, and J. D. Winefordner,Crit. Rev. Anal. Chem. 27, 257 (1997).

3. K. Song, Y. I. Lee, and J. Sneddon, Appl. Spectrosc. Rev. 32, 183(1997).

4. E. Tognoni, V. Palleschi, M. Corsi, and G. Cristoforetti, Spectro-chim. Acta, Part B 57, 1115 (2002).

5. J. A. Aguilera, C. Aragon, and J. Campos, Appl. Spectrosc. 46,1382 (1992).

6. C. Aragon, J. A. Aguilera, and J. Campos, Appl. Spectrosc. 47, 606(1993).

7. R. Noll, R. Sattmann, and V. Sturm, Proc. SPIE-Int. Soc. Opt. Eng.2248, 50 (1994).

8. L. Peter, V. Sturm, R. Noll, R. Hakala, J. Viirret, B. Overkamp, andP. Koke, Proc. SPIE-Int. Soc. Opt. Eng. 3823, 256 (1999).

9. V. Sturm, L. Peter, and R. Noll, Appl. Spectrosc. 54, 1275 (2000).10. J. Gruber, J. Heintz, H. Strasser, D. Bauerl, and N. Ramaseder,

Spectrochim. Acta, Part B 56, 685 (2001).11. L. Peter, V. Sturm, and R. Noll, Appl. Opt. 42, 6199 (2003).12. G. M. Bilmes, O. E. Martınez, P. Sere, D. J. O. Orzi, and A. Pig-

notti, Am. Inst. Phys. Conf. Proc. 557, 1944 (2001).13. G. M. Bilmes, O. E. Martınez, P. Sere, R. Musso, D. J. O. Orzi,

and A. Pignotti, Lat. Am. Appl. Res. 32, 263 (2002).14. G. M. Bilmes and O. E. Martınez, Patent 6.546.784- EEUU (2003).15. G. M. Bilmes and O. E. Martınez, Patent AR9906271, Argentine

(1999).16. M. Mesaros, O. E. Martınez, G. M. Bilmes, and J. O. Tocho, J.

Appl. Phys. 81, 1 (1997).17. C. Aragon and J. A. Aguilera, Appl. Spectrosc. 51, 1632 (1997).18. M. A. Khater, P. van Kampen, J. T. Costello, J.-P. Mosnier, and E.

T. Kennedy, J. Phys. D: Appl. Phys. 33, 2252 (2000).19. K. J. Grant, G. L. Paul, and J. A. O’Neill, Appl. Spectrosc. 44,

1711 (1990).