Laser induced plasmas for analytical spectroscopy

Spectrochimica Acta, Vol. 26B, pp. 311to 332. PergamonPress1970.Printedin NortheinIreland

Laser induced plasmas for analytical spectroscopy R. National

Physical

Research

H.

SCOTT

and A.

STRASHEIM

Laboratory, Council for Scientific Pretoria, South Africa (Received

12 December

and Industrial

Research,

1969)

Abstract-A series of high speed photographs are presented which show the crater and plasma formation when pulsed laser light is focussed onto the surface of a metal. Some features of the evaporation process are discussed. A spatially and time resolved spectral study of various laser plasmas was undertaken. We distinguish between three different modes of laser operation, i.e. normal mode (free running), semi-&-switched mode and giant pulse mode. Various laser plumes were investigated as spectrochemicrtl sources for quantitative analysis. For comparison, a series of aluminium alloy standards was analysed for copper content. The detection system employed a. time resolution gating technique.

with the pulsed laser as a source for spectrochemical analysis is a lack of understanding of the processes of evaporation by laser light. As precision and sensitivity are dependent on the evaporation process it was necessary to undertake a fundamental study of the laser beam-surface interaction in order to fully investigate the source. Such studies have been undertaken by a number of authors [l-20]. It is well-known that when a pulse of high energy laser light is focussed onto the surface of a metal, considerable heat is generated within the confined area of the focal spot [l-4]. The absorbing material undergoes a change of state. This is visibly A PROBLEM

[l] G. M. RUBANOVA and A. 0. SOKOLOV, Soviet Phys.-Tech. Phys. 12, 1226 (1968). [2] V. P. VEIKO, YA. A. Ims, A. N. KORORA and M. N. LIBENSON, Soviet Phys.-Tech. Phys. 12, 1410 (1968). [3] P. I. ULY~OV, Soviet Phys. JETP 25, 537 (1967). [4] Yu. V. A~ANASYEV, 0. N. KROKHIN and G. V. SKLIZEOV, IEEE QE-2,483 (1966). [5] C. DAVID, P. V. AVIZONIS, H. WEICHEL, C. BRUCE and K. D. PYATT, IEEE QE-2, 493 (1966). [6] A. W. EHLER, J. Appl. Phys. 37, 4962 (1966). [7] J. M. DAWSON, Phys. Fluids 7, 981 (1964). [S] E. H. PIEPMEIER and H. V. MALMSTADT, Anal. Chem. 41, 700 (1969). [9] H. KLOCBIE, Spectrochim. Acta 24B, 263 (1969). [lo] D. LICIITMAN and J. F. READY, Phys. Rev. Lett. 10, 342 (1963). [ll] L. G. PITTAWAY, J. SMITH, E. D. FLETCHER and B. W. NICHOLLS, Brit. J. AppZ. Phys. (J. Phys. D). 1, 711 (1968). [12] P. LANGER, G. TONON, F. FLOUX and A. DUCAUZE, IEEE &E-2,499 (1966). [13] L. I. GRECHIKHIN and L. YA. MIN’KO, Soviet Phys.-Tech. Phys. 12, 846 (1967). [14] T. J. HARRIS, IBM J. Res. Develop. 7, 342 (1963). [15] W. B~GERSHAUSEN and R. VESPER, Spectrochim. Acta 24B, 103 (1969). [IS] E. ARCHBOLD, D. W. HARPER and T. P. HUGHES, Brit. J. AppZ. Phys. 15, 1321 (1964). [17] J. F. READY, J. Opt. Sot. Am. 53, 514 (1963). [18] J. F. READY, Appl. Phys. Lett. 3, 11 (1963). [19] J. F. READY, J. Appl. Phys. 36, 462 (1965). [20] J. M. BALDWIN, Bibliography of Laser Publications of Interest to Emission Spectroscopists. U.S.A.E.C. Repts. IN-1219, IN-1262 (1968). 1

311

312

I-t. H. SCOTT

and A.

STRASHEXN

evident in the form of a luminous vapour plume which erupts at a high velocity [5-S]. A crater is formed in the surface of the sample [9]. The crater characteristics are dependent on the mode of laser radiation. Xon-Qswitched or normal mode radiation produces deep molten-edged craters. The depth of the craters depends on the thermal properties of the sample and the energy density of the laser light. Craters produced by Q-switched radiation show less dependence on the properties of the sample and there is a notable absence of molten material. The luminous plume of vapour above the surface is accompanied by thermionic electrons and ions which are emitted at high current densities f 10-121. Peak electron currents of 193amp per em2 for one joule normal mode radiation have been observed in this laboratory. Calculations of the surface temperature using the thermionic emission equation indicated peak temperatures of the order of the boiling point of the metal. Present knowledge indicates that laser energy is absorbed in the initial stages of the laser beam acting on a surface. Later a plume appears, consisting i~t~a~y of atmospheric species which are then replaced by sample material. The excited spectrum is affected by the absorption of the laser beam in the plume. Special attention is given in this study to the plume and surface phenomena. By using a high speed framing camera, clear photographs of the reaction of ruby laser light with the sample and of subsequent phenomena were obtained. This paper presents a series of these high speed photo~aphs from which some interesting observations are made. A rotating sample technique was also used to investigate the transient behaviour of the laser interaction. Spectra of the plumes produced by normal mode, semi Q-switched mode and giant pulse laser radiation were recorded photographically at various heights above the sample surface. A double trace oscilloscope adjoining a direct reading speetrometer was also used to obtain temporal information of value for the analytical use of the different modes of laser radiation, Finally, the laser plumes were used as spectrochemical light sources and evaluated by a special direct reading technique. ~~STRUME~TATI~~ Laser

The laser source was comprised of a ruby rod 76 mm long, 6 mm dia. and a flashlamp mounted in an air-cooled cylindrical ellipsoid pump cavity. The flashlamp discharge circuit consisted of a single mesh L-C network. The laser output mirror had 58% reffeetivity at 6943 8. Table 1 gives the parameters of the laser radiation for different operational modes. In normal mode the laser was operated with a 99.9% reflective back mirror. A maximum output energy of 3 J in approximately 850 psec was obtained at a pump energy of 1000 J. When operated in semi-Q-switched mode, the back mirror was rotated at 3000 revlmin. A maximum output energy of O-4J was obtained in a pulse train consisting of 15-20 spikes for a duration of approximately 30 psec. For double or single giant pulse operation, a spinning prism was used as the back reflector of the laser resonator. The prism was driven by an air turbine at up to

Laser induced plasmas for analytical spectroscopy

313

40,000 rev/min. The output consisted of 1 or more pulses of radiation (depending on the input energy and Q-switch speed) each having an energy of 0.1 J and &half-width of less than 50 ns8c. The laser beam was monitored by incorporat,~g a beam-splitt8r to reflect & sm&ll percentage of the laser light into an RCA type 922 photodiode. This was coupled to & Tektronix type 555 oscilloscope equipped with an oscilloscope camera. Table 1. Parameters of laser radiation Mode

Normal mode

Semi-Q-switched mode

Giant pulse mode

35 m-850wee

Synchronous slectric motor; 3000 rev/min; rotating mirror 0.4 5 25-30 jisec

Air turbine 540,000 rev/min; rotating prism 0.4 J < 1 psec

O-1-0.3 psec -20 NO.02 J -100 kW 14kW

t50 nsec l-3 0.1 J -2iaW 2MW

Q switch

Max. energy Total duration Individual spike duration Number of spikes Energy per spike Power per spike Me&n power

0.1-0.3 #Usec -550 NO.005 J -25 kW 3.5 kW

Except where otherwise stated, a 27 mm objective lens w&sused to focus the laser radiation onto the surface of the sample. The lens front surface was protected from glowing p&rticlesejected from the sample surface by a microscope cover slide. High speed camera

To photo~&ph the laser beam-surface inter&ction, & Barr and Stroud ultra high speed framing earner&w&sused having & m&ximum framing rate of 4 x IO6pictures p8r S8C. The 35 mm Kodak type 4 X film was developed in Promicrol ultra fin8 grain developer. The surface interaction was illuminated by reflecting onto it a percentage of the light given off by the laser flashlamp through the partially transmissive pump cavity of the laser. Rotating sample apparatw? For further studies of the transient bshaviour of the laser interaction, the surf&ce on which the radiation w&s focussed was rotated at ELhigh speed in order to time resolve the reaction of individual laser spik8s. At the s&metime, the laser radiation w&smonitored and the osc~loscope display photo~aph8d. A comp&rative study was later m&de of the surface damage and the corresponding oscillo~aph of the radiation. The sample disc was aluminium with &diameter of 7 cm and &thickness of 1 mm. It was attaohed to the shaft of an air turbine. A maximum speed of 40,000 rev/min could be obtained, giving a circumferential resolution of 7 ysec per mm.

Time integrated, spatially resolved spectra were recorded using & fiilger Medium Spectrograph. The entrance slit width was 20 p with a length of 3 mm. The plumes

R. H. SCOTTand A. STRASHEIM

314

formed at the surface of the sample by the focussed laser beam were imaged onto the entrance slit in a 1: 1 ratio. Spectra of single laser shots were recorded on Ilford Selochrome spectrographic plates, which were developed in type D19B developer. Temporal studies were made by imaging the laser plumes onto the entrance slit of a Hilger Medium Spectrograph equipped with a direct reading attachment. The spectrometer photomultipliers were equipped with load resistors of 1 KR and the photomultiplier dynode chains fitted with capacitor decoupling. To record the time profiles of the spectral lines, a Tektronix type 555 dual beam oscilloscope with an oscilloscope Polaroid camera was used. The two sweeps were triggered simultaneously by a pulse obtained from the laser beam monitor. The upper sweep displayed the laser radiation and the lower sweep displayed the spectral emission of the plasma cloud. Catilzg detection

system

For direct reading quantitative analyses, a gating device* was designed to reduce the contribution to the integrated signal by photomultiplier dark current and also by the spectral background in the case of the single and double pulse modes. In this system, the photomultiplier tubes were gated at the collectors using field effect transistors as the switching elements. The gating pulses were triggered by the laser monitor. When no time resolution was employed, the gates were opened at the outset of laser action (triggered on the leading edge of the initial laser spike) and held open for the total duration of the laser plume emission. This technique reduced the contribution of photomultiplier dark current to the integrated signal. To reduce the spectral background contribution, the spectra were time-resolved by delaying the integration for the period of laser irradiation in the case of the single and double pulse modes. The use of a gating technique for integrating the spectral signal from a single laser plume was advantageous as the integrated voltage could be substantially increased by using small integrating capacitors of 0.2 uF. The voltages on the capacitors were read by a digital voltmeter with an input impedance of 101* CL RESULTS AND DI~~TJ~~I~N 1. High speed photography

We photographed the surface interaction which occurs when various laser power and energy modes are focussed onto the surface of an aluminium sample. In particular, a detailed study of the surface phenomena relating to the normal mode interaction was undertaken. The high speed camera optics is glass and no wavelength filter was used. The photographs therefore show total visible light emitted by the laser plasmas. Table 2 gives the various laser energies that were used, the focal length of the laser focussing lens, the mean energy and power density on the sample and the mean crater size in each case. * To be published shortly.

Laser induced plasmas for analytical specttroscopy

315

(a) The normal mode inter&ion High speed photographs showing details of the evaporation process using a mean power density of 7.6 ~~7~e~n~ are presented. The evaporation process for lower power densities is discussed later. First, a series of photographs of the complete interaction was recorded. The framing interval, i.e. the time between the peak intensities of two successive frames, Table 2. Laser radiettion parameters used for high speed photography surface intemction

(J)

(cm)

Mean energy density* (kJ/cm2)

0.75 1-5 2-25 3.0 1.5 3.0

2.7 2.7 2.7 2.7 7.5 7.5

1.6 3.2 4.8 64 0.49 0.98

Energy Mode

Sormal

mod0

Semi-Q-

Lens focal length

Eean power density (MW/cm2) 1.9 3.8 5=7 7-6 O-58 1.2

0.1

2.7

0.21

switched mode

0.4

2.7

0.84

31

mode

0.1

2-7

0.21

4200

7.8

of the

Xean crater diameter (cc) 150 240 300 340 500 750 650 x 450t

300

* The energy was corrected by the lens transmission factor end the spot size oaloulated from the divergenoe of the laser beam (-8 m rad.). $ Elliptically shaped crater. 22 psac, From the time resolved ~h~tographs~ the following information became apparent : When irradiation commences, a small plasma forms on the surface of the sample. The plasma rapidly develops into a stream of luminous vapour with an ejection velocity of approximately 8300 om/sec (measured at the front edge of the luminous cloud). A late& expansion ocours as the stream travels upwards, and a pulsation in va,pour ejection is noticed. Ejected particles are then detected. The vapour cloud emission decreases and a stage is reached when no emission is observed even though laser action continues. Particle ejection is predominent at this stage, An explanation for this observation is given later. To investigate the evaporation process in greater dotail, we photographed the surface phenomena to about 2 mm above the sample at a higher magnification than that used previously. (The scafe on each figure gives the relative magni~~~tio~~.~ The framing interval was also reduced to S i~ec. Under these conditions it was ~mFossibIe to photograph the complete laser cycle. Figure 1 shows the first 416 ,usec of the interaction. In frame 7, a heated spot on the sample surface is seen. A small cloud of plasma appears in frame 8. The expansion velocity of this cloud is approximately 9000 cm/see in this initial stage. In frame 10, the first sign of a crater is seen A fine spray of particles is observed in frames f&-18, The velocity of these particles is approximately 4000 omfsee. The solid angle of ejection becomes less as, presumably, the crater deepens in the sample (frames 11-18).

wm

316

R. H.

SCOTT

and A. STFCASHEIM

Striate layers are clearly seen in the vapour cloud in frames 3 l-41. The horizontal thread-like layers are regions in the plasma having a high radiative intensity. These striations have an average velocity in the vertical direction of 2200 cm/see. They are not seen lower than 0.6 mm above the crater in the sample surface. Commencing in frame 47 another phenomenon is observed. What appears to be a sheath of molten material rises out of the crater. The upper portion seems to disintegrate and the mixture of plasma and molten material flows turbulently upwards in the vapour stream from within the crater.

2

t

I 0

I

I

,

I

I

I

I

I

I

2

3

4

5

6

7

6

POWER DENSITY

ihlW/cm?

Fig. 4. Graph showing the relationship between the laser radiation power density and the plasma ejection velocity, when normal mode laser radiation onto an duminium sample.

is focussed

More detail of the sheath formation is shown in Fig. 2. This sequence of photographs commences at 300 psec after the interaction began. (Frame 1 in Fig. 2 corresponds approximately to frame 43 in Fig. 1.) The sheath of molten metal appears in frame 3 and rises at a velocity of 1200 cm/set. As it rises, the upper portion contracts, and at a height of l-6 mm above the surface, when it is approximately O-2 mm in dia., the contracting sheath interrupts the incident laser beam and excitation occurs as a result of the absorption of the laser radiation. This excitation is seen in frames 11-14. The following frames show a general disintegration of the sheath into a plasma and particles in chaotic motion. A new sheath of material seems to form in frames 20-27 and to contract in frames 28-30. The column of molten material that is formed then separates near the surface and spirals upwards. The crater which remains in the surface after the interaction is completed is seen in frames 57-59. The spiral behaviour of the molten material was observed in most laser shots.

_

_

Wig. 1. High

spwd

photographs,

taken with a framing interval of 8 ,rscc. show det,ails of the first, 416 /csec of tjhcsint,eractim of 7.6 MW/crn2 laser radiation with an nlumininm salnplc.

u 0 0

w c

Laser induced plasmas for analyticrtlspectroscopy

317

Figure 3 shows another sequence of photographs commencing at 300 psec after the start of laser irradiation. The sheath contraction, the elevated vaporization and excitation in the path of the laser beam and the turbulent nature of the system is clearly seen. When the power density was decreased, the heating time (before vaporization commenced) increased. This was visible as a spot on the sample surface. The longest heating time measured was approximately 48 psec for a mean power density of 0.58 MW/cm2. However, this heating time was not reproducible as it depended on the intensity of the initial laser spikes. The presence of a large spike early in the discharge could cause vaporization in less than 8 ,usec. This was investigated more fully in Section 2. A decrease in power density also caused a decrease in the plasma ejection velocity. This is shown graphically in Fig. 4. Vortexing was not in evidence at lower power densities. Furthermore, the fine spray of particles which was observed in the initial stage of the interaction was not detected for mean power densities of 0.58 and l-16 MW/cm2. (b) The semi-&-switched mode interaction Photography of the semi-Q-switched (and giant pulse) plumes was more difficult than the normal mode plumes as the rotating mirror of the camera could not be synchronized with the rotating back reflector of the laser cavity. However, sufficient observations were made on the grounds that the chance of the event being partly recorded without any synchronization was 1: 3. A complete sequence of the plasma formation for a mean power density of 7.8 MW/cma is shown in Fig. 5. The framing interval is 1 ,usec. In frame 3 a small plasma is seen. The plasma expands and assumes the shape of an oblate spheroid, as the initial horizontal expansion occurs at a higher rate than the vertical expansion. The horizontal and vertical expansion rates are approximately 5 x 10” and 3 x 10” In frames 12-21 the plasma extends over and down the right cm/set, respectively. hand side of the sample. A tapering core of vapour which appears to radiate more intensely than the surrounding cloud is seen in frames 12-17. Later frames (22-42) show the cloud growing less emissive and dividing into a surface component and an elevated component. In these frames no bright cores is seen and the cloud emission We note that the plasma is still clearly visible in appears rather inhomogeneous. frame 42, i.e. at least 40 ,usec after laser irradiation commenced. Another feature is that no particle ejection occurred throughout the sequence. Time resolved photographs showing details of the interaction at a mean power In this case we decreased the framing density of 31 MW/cma were also obtained. interval to 250 nsec and increased the magnification to approximately 20 times that of Fig. 5. A comparison between the monitored oscillogram of the laser radiation and the time resolved photographs showed that each spike of the discharge could be associated with a bright spot at the sample surface and an ejected plasmoid. Plasmoids were ejected at a rate of more than 3.2 x lo6 cm/set during these periods of irradiation (0.1-0.3 psec) . They then cooled by expansion at an average rate of 3 x lo* cm/set, as illustrated by the graphs in Fig. 6 which shows the expansion rates of a number of different plasmoids.

R. H.

0

0.25

SCOTT and A.

0.5

STRASHEIM

0.75 TIME

I.0

I.25

(.Li see)

Fig. 6. Graphs showing the expansion rates of plasmoids ejected by various laser pulses of a 31 MW/cm2 semi-Q-switched discharge.

(c) The giant pulse mode interaction Photographs of the formation of the Q-switched plasma showed the formation of a small spherically shaped plasma which expanded and decreased in intensity. The exact time of the laser spike could not be categorically determined. It was also not possible to accurately measure the expansion rate during the evaporation process due to the limited resolution which could be obtained in the framing mode. However, the expansion rate in this initial stage exceeded lo6 cm/set. The emission of radiation from the plasma decreased rapidly, and after a period of approximately 20 ,usec the plasma was no longer visible. During this period it remained relatively motionless, resembling a nebula suspended above the surface of the sample. NO particle ejection was observed. 2. INVESTIGATION OF TRANSIENT SURFACE PHENOMENA BY A ROTATING SAMPLE TECHNIQUE In this investigation the sample surface on which the laser radiation was focussed, was rapidly rotated in order to present a new surface area to each laser spike in the sequence that constitutes a normal mode or semi-Q-switched mode discharge. By monitoring the laser radiation oscillographically, comparisons could then be made of the vaporising ability of laser spikes of differing intensities. Using semi-Q-switched radiation having an approximate power per spike of 100 kW, we found that each laser spike of the sequence could be associated with a crater on the sample surface. Furthermore some craters were displaced off the true

Laser induced plasmas for analytical spectroscopy

319

arc, indicating the existence of off-axis modes caused by the rotation of the back mirror of the laser cavity. A direct comparison of normal mode spikes (-25 kW) and crater formations could not be made due to the large number of spikes present. However, it was possible to compare the intensities of monitored semi-&-switched spikes with normal mode spikes by resetting the laser to operate in the normal mode. We found that spikes of relatively low intensity (approximately 10% of the intensity of the largest spikes in the normal mode discharge) were capable of causing a small amount of vaporization. This leads to the conclusion that all but the very low intensity spikes of the normal mode laser discharge will vaporize a small quantity of the sample. Spikes having an intensity of about 10% or less of the intensity of the largest spikes of the 3 J delocalized discharge predominently caused heating with little vaporization of material. 3. SPECTRAL STUDIES OF LASER PLUMES In this study, a comparison was made of the spectral characteristics of various laser plumes. The three types of plumes which were investigated were formed by normal mode, semi-Q-switched mode and giant pulse laser radiation, having a mean power of 3.5 kW, 14 kW and 2 MW, respectively. A 27 mm lens was used to focus the radiation onto the sample surface in each case. SPECTROMETER

ENTRANCE

SLIT

I

SAMPLE

PiUME

Fig. 7. Schematic diagram showing the setup used for sampling the laser plumes in cross section.

1. Sfpatial resolution

The laser plumes were imaged onto the entrance slit of the spectrograph in a ratio. The sample, an NBS 603 aluminium alloy standard, was placed vertically so that the spectrometer entrance slit was parallel to the image of the sample surface at the slit. The laser beam was focussed onto the surface in the horizontal direction, as shown in Fig. 7. Spectra could thus be recorded at various heights above the sample surface. Only “single shot” spectra were investigated. Figure 8 shows a section of typical spectra which were obtained. The lower spectrum is for normal mode, the centre for semi-Q-switched mode, and the upper for the giant pulse mode plasma at a height of 1 mm above the sample surface. We immediately notice the increase in background for increasing laser power. Furthermore, the lines in the upper and centre spectrum are considerably more Self-absorption, on the other hand, was broadened than in the lower spectrum. found to be more in evidence in the centre and lower spectrum. This was especially 1: 1

noticeable

in the case of the Al I 3082 and 3092 A lines.

R. H.

320

SCOTT and

A. STRASHEIM

Figure 9 shows densitometer tracings of the copper lines 3247 and 3274 A for the three plasmas; (A) normal mode, (B) semi-Q-switched mode, and (C) giant pulse mode. We note the high line-to-background ratio in (A) and the low line-to-background ratio in (C). The findings illustrated in Figs. 8 and 9 will be further discussed in the following paragraphs.

I 25 0’ -

3247

3274

3247

3274

3247

I

Fig. 9. Densitometer plasma: A: normal

3274

I

tracings of the copper lines 3247 and 3274 s for the three mode, B: semi-Q-switched mode and C: giant pulse mode.

To investigate the time integrated spatial structure of the three plasmas, spectra were recorded at intervals of 1 mm, starting from 1 mm above the sample surface. The following observations were made : (a) Normal mode plasmu We found that the normal mode spectra photographed at 1 mm above the sample surface differed from those at 10 mm above the surface. The spectra of the upper region of the plume were relatively weak compared to the lower region and the former contained molecular bands which were radiated almost uniformly across the sampled cross section with a slight increase in intensity in the plasma core. The bands were identified as those of the molecule aluminium oxide (see Table 3). Generally, these aluminium oxide bands are very sensitive to temperature, and their intensity is high at plasma temperatures of 3000~4000°K. Higher temperatures cause dissociation of the molecule. In the lower region of the plume these bands were not clearly visible due to the background, and the only bandheads which were detected were at 4842.1 h and 4648.2 A. The spectral background (continuum) was low compared to the other types of laser plumes. The background at 1 mm above the sample surface was more intense than that of the upper region and was localized in the core of the plasma column. The line spectra of the lower region of the plume were found to be more intense than the upper region in which the only visible lines (apart from the molecular bands)

45

50

55

I I I I I I I I I I I I I I I I I I I I I I I I1111111111111

Fig. 8. A section of spectra obtained at a height of 1 mm above the sample surface. The lower spectrum is that of the plurnc: of normal mode radiation, the centre spectrum of the semi-Q-switched mode plume and the llpper spectrum of t,he giant pr~lse laser plume.

iLr

321

Laser induced plasmas for analytical spectroscopy

were the aluminium raies-ultimes. In the lower region, the aluminium III lines 5722.6, 5696.5, 3601.9 and 3612.3 were detected. The excitation potentials of these lines are all 17.8 eV. These high temperature lines were of low intensity, diffuse and confined to the core of the plasma column. The aluminium II lines 5593.2 (15.4 eV), 3586.5 (15.3 eV), 4663-O (13.2 eV) and 3900.7 A (10.6 eV) (excitation potentials given in parenthesis) were all visible in the plasma core. The 2816.2 A aluminium II line (Il.8 eV) radiated strongly and uniformly across the sampled cross section. Table 3. Band heads of aluminium oxide (degraded to the red) 4470.5 A 4494.0 4516.3 4537*6* 4557.5 4648.2 4672.0 4694.6 4715.5

4735.5 4842.1* 4866.1* 4888.4 5079.3 6102.1 5123.3 5142.9 5160.8

* Intense.

Near complete self-reversal occurred for all the aluminium resonant lines. In all cases the self-reversal was asymmetrical with the intensity peak of the emission line on the higher wavelength side of the absorption line. Or

f3 z 07 i z 2 k

25

I

2568

-

2566

2575

2575

502575

Li g

2568

75.I:;:_;:;,I:,;I:

:

100

A

c

B

Fig. 10. Densitometer tracings of the duminium resonant lines 2668 and 2575 A for the three plasmas: A: normal mode, B: semi-Q-switched mode and C: giant pulse mode. Note the asymmetrical self-absorption and extreme broadening of the lines. Densitometer Fig.

10(A).

tracings

The explanation

of the

aluminium

2568

for this asymmetrical

and

2575 A lines

self-absorption

are shown

in

lies in the asym-

metrical broadening and shift of the emission lines at high vapour pressures. The average of the measured wavelength separations between the emission and absorption peaks of the lines 2568, 2575, 2652 and 2660 A was 0.8 A.

322

R. H. SCOTT and A. STRASHEIM

(b) Semi-Q-switched

mode plasma

The spectra of the plasma produced by the semi-Q-switched mode laser radiation was photographed in steps of 1 mm up to 10 mm above the sample surface. At a height of 8 mm only the aluminium lines 3952 and 3962 A were detected. All the prominent aluminium oxide molecular bands were detected. Peculiarly, they were most intense at heights of l-2 mm and 6-7 mm, and less intense at 4 mm. This may be due to absorption of laser energy in the upper portion of the plume.

O25 L

Imm

3mm

2mm HEIGHT

ABOVE

SAMPLE

4mm

5mm

SURFACE

Fig. 11. Cross sectional background scans in the region 4100-4125 A for various heights above the sample surface for the semi-Q-switched mode plasma.

The intensity of the background continuum was high near the sample surface and became lower for increasing height above the surface. In the lower region of the plume, the background was most intense in the plasma core. Figure 11 shows densitometer tracings of the cross-section background profiles at heights of 1, 2, 3, 4, and 5 mm in the wavelength interval 4100-4125 A. Although the background decreased with increasing height, atomic line emission remained fairly constant up to a height of approximately 4 mm, and then decreased. This is shown in Fig. 12 for the copper line 3274 A. The highest line-to-background ratio for lines of low excitation potential (approx. 3 eV) was found at a height of approximately 3-4 mm above the sample surface. This could be an important result in the analytical application of this plasma. The aluminium III lines in the spectra were confined to the core of the plasma and were detected up to a height of 3 mm. The lines with the highest excitation potentials i.e. 5163.9 (259 eV), 5150.8 (25.9 eV), 2907-O (25-O eV), 2762.8 (25-O eV), 4701.6 (23.4 eV) and 3713.1 A (21.2 eV) were not detected. Most other aluminium III lines were visible but localized in the plasma core. Aluminium II lines were found up to a height of 5 mm (e.g. 5593.3, 3586.5, 4663.0 A). They were not as localized as the aluminium III lines. Self-reversal and extreme broadening of the resonant lines were characteristic

323

Laser induced plasmas for analytical spectroscopy

Cu 3274

“A

4

1

\ \

M-

Imm

2mn

HEIGHT

irn

Irn

ABOVE

SAMPLE

-I

5mm

SURFACE

Fig. 12. Densitometer tracings of the copper 3274 A line at various heights above the sample surface for the semi-Q-switched mode plasma. 0

r-

25-

2660

lOOL--(

Fig. 13. Densitometer tracings of the aluminium resonant lines 3082, 3092, 2568, 2575, 2652 and 2660 A for the semi-Q-switched mode plasma, superimposed with the resonant lines of an aluminium hollow cathode lamp to show the laser plasma line shifts. of these

spectra

(see Fig.

10(B)).

Line

broadening

was at a maximum

region near the sample surface in the plasma core. Because self-absorption, no measure could be made of line broadening. iated

with

measured

pressure by

effects

superimposing

and

seen

the

spectrum

as asymmetrically of an aluminium

in the lower

of the high degree of The line shifts assoc-

self-reversed hollow

lines, cathode

were lamp

(~6 torr) onto the existing spectrum without altering the position of the spectrographic plateholder, entrance slit, or external optics. In this way a red shift of both

324

2%.H. SCOTT

and A.

STRASHEIM

the absorption lines and emission lines was recorded. Figure 13 shows densitometer tracings of the superimposed lines. The emission line shift for the lines 2652, 2660, 2568,2575,3082and 3092 A was approximately l-6 L%to the red. The absorption line shift was approxima~Iy 0.3 A to the red. This absorption shift to longer wavelengths may be due to the existence of ground state atoms at an elevated pressure around the plasma core. These results indicate a high pressure gradient in the laser plume. (c) Giant PZse laser Plasma No intense aIum~um oxide moiecuIar bands were present in any of the recorded spectra. Only one weak bandhead at 4842.1 A was detected in the lower region of the plasma. The background radiation was more intense than that of the former plasma types. Background intensity was very high throughout the lower region. A very narrow band of continuum was seen in the core of the plasma (see Fig. 8 upper spectrum). This corresponds to the path of the laser pulse through the plasma. The absorption of the tail of the laser pulse in the plasma which is ejected could give rise to this continuum. High temperature lines which were only faintly seen in the spectra were the aluminium III lines 5163.9 (25.9 eV), 4479.9 (23.5 eV), 3713.1 (21~2 eV), 3702-L (21.2 eV), 4512-5(20,5 eV) and 4529-2(20-5eV). These Iines were broadened as a result of the high pressure associated with the plasma. Lines which were detected in this plasma but in neither of the former plasmas were afuminium III 5163.9 A and aluminium III 5000*9A, with excitation potentials of 25.9 and 17.9 eV, respectively. From the densitometer tracings of the aluminium resonant lines (Eig. 10(C)), a difference of 1.2 k was found between the absorption and emission peaks. 2. Temporal resolution Various laser plumes were further investigated by direct reading spectroscopy. Each laser plume was viewed in cross-section as described previously. Intensity-time profiles of various spectral lines were recorded oscillographically.*

For normal mode radiation, we found that all the spectral lines investigated were radiated after a delay of approximately 30 psec after lasing commenced, and that spectral emission continued in a pulsatory nature for approximately 400 psec. No further emission occurred, even though lasing continued. Figure 14(A) shows an os~illo~am of the normal mode laser radiation (upper beam) and the emission of the aluminiuIn 3092 A line (lower beam). The delay during which the sample heats up, and the time after which no further * We found that in order not to saturate the photomultipliers, the plumes produced by the single pulse, double pulse and semi-Q-switched mode had to be defocussed on the spectromater entrance slit. This seems to show that for direct re~ingspectroe~lemic~I~n~lysis, a fast spectrometer is not necessarily an essential requir~~n~nt. Although the integrated intensity of a spectral line radiated by one laser plume is low, the intensity peak is high, and therefore photomultiplier seturation csn easily occur.

Laser induced plasmas for analytical spectroscopy

325

vaporization occurs, corresponds to the results obtained with the high speed photographs of crater formation and eventual particle ejection. Thus the progress of vaporization is arrested by the presence of particles of the material which scatter the laser radiation. This assumption was made by Bagershausen and Vesper [15] who referred to a “Funkenspriihen” or “spark spray” effect. The background radiation was also of a pulsatory nature. This background, however, was much less intense than the spectral line emission. Not all laser spikes caused spectral emission from the plume. Even fairly similar laser radiation pulses cause very variable spectral radiation. This is shown in Figs. 14(B) and (C). This phenomena is thought to be due to inhomogeneous partial absorption of the laser pulses in the plume. The temporal study of the normal mode plasma has shown that spectral emission occurs during individual laser spikes and decays to zero in most cases during the intervals between spikes. The vapour cloud therefore cools to a relatively low temperature during these intervals. The vapour seems to be constantly present and can be assumed to be a quasi-stationary collection of elementary vapours which are ejected in a pulsatory manner. Consider an individual laser pulse which travels through the vapour plume towards the crater. Part of its energy is absorbed by the plume and the remainder is converted to heat within the crater. Because of the short duration of the pulse (0.1-0.3 psec) the vapour cannot undergo complete isothermal expansion and its temperature increases. Spectral line emission occurs, the radiation being readily self-absorbed in the cooler vapour surrounding the heated core. After the laser pulse has ended, the spectral emission rapidly decreases due essentially to adiabatic expansion. This process continues, and the plume grows larger due to the additional vapour which is injected. As the vaporization process gives way to the melting flushing process, and as, presumably, the density of the plume decreases, the pulsating spectral emission diminishes and finally particle ejection becomes the predominant feature of the interaction. As the spectral background is excited with every laser spike it cannot be time-resolved from the line emission, and it can be stated that a time resolution technique will not improve analytical conditions if this mode of laser excitation is used for analytical purposes. (b) Semi-Q-switched mode plasma All the spectral lines which were investigated were radiated after delay of approximately 20 ,usec after laser irradiation commenced, even when the image of the sample surface was set as near to the spectrometer entrance slit as possible (see Fig. 7). This delay may be due to the spatial non-stationarity of the laser discharge, coupled with the rather low pulse frequency of the mode, which is dependent on factors such as laser mirror alignment, ruby temperature and flashlamp pump power. The neutral lines Al I 3082, 3092 and 2660 A and the ionic lines Al II 2816 A and Al III 5695 A were all modulated with intensity peaks corresponding to the laser spikes. An example is given in Fig. 15 which shows the semi-Q-switched radiation in the upper sweep and the aluminium II 2816 A line emission in the lower sweep. (Each sweep was triggered simultaneously by the initial laser spike.)

326

R. H. SCOTT and A. STRASHEIM

The intense background radiation showed similar characteristics to the line emission. The laser spikes caused intensity peaks in the background radiation. Thus we concluded that it would be impractical to separate, by time resolution the background signal from the line emission for analytical purposes, when the semi-Q-switched laser is used. (c) Q-switched mode p1asm.a (i) Single p&se. The radiation of the plasma produced by a single giant puIse was measured with the spectrometer entra.nce slit at a height of ~6.5 mm above the image of the sample surface. The cooling plasma sadiated spectral lines having intensity maxima at different times according to their excitation potentials. As an example, Fig. 16 shows the three lines Al I 3961 (3.14 eV), Al II 2816 (11.8 eV) and AI III 5696 A (17.8 eV) in time profile. The duration of the laser pulse (upper sweep on the oscillographs) is actually < 100 nsec. The longer decay shown on the sweep is as a result of the relatively long time constant of the laser monitor. The characteristics of the Q-switched aluminium plasma have been investigated by Piepmeier and Malmstadt [8], and their results indicated that a.luminium ions are ejected into the plume a.head of most of the atoms and, furthermore, unlike the high voltage spark, that recombination of ions and electrons plays a smaller role in the production of atomic species than the major source, which appeared to be the later arrival of atoms from the sample. We are inclined to believe that the recombination mechanism is favourable in the This is clearly evident from the time profile studies of later stages of the interaction. the spectral lines, which indicated that ion species were predominant in the early stage, followed by maximum atomic line emission a.fter the laser pulse had ended, due to the decreasing plasma temperature. Measurements of plasma expansion velocities (e.g. Fig. 6) indicated an initial ejection of material to form a highly active plasma which then cooled and recombined. The background emission of the plasma is unresolved in Fig. 16. However, by monitoring the background signal alone, it was found that it existed mainly during the laser pulse. This confirms the results obtained by ARCHBOLD et at. [16], and makes possible the separation of line and back~oulld for analytical purposes. (ii) Double pulse. We have also investigated the laser plume produced by a double giant pulse. In this case the second pulse (also ~2 MW) occurred 500 nsec after the initial pulse. We found that the integrated spectral line intensities increased by a factor of approximately 2. Because the background radiation occurred mainly during the laser pulses, it was possible, as with the single pulse plasma., to separate the spectral line emission from the background, using a gating technique, and thus improve anaIytica1 conditions if the plasma is used as a spectrochemical source. 4. FURTHER Discussion

ON TEE KINETICS OF THE NORMAL MODE

INTEEACTION Our detailed studies of the interaction of ruby laser radiation with aIum~ium in the range of power densities 106-10’ W/cm2 have revealed a more complex system than that generally described for metsls.

LOWER

UPPER

LASER PULSES TRACE:At3092

TRACE:

MODE

Fig. 14. Oscillograms of the normal mode laser pulse train (upper sweep) and the corresporldil~g plasma emission of the line Al 3092 A (lower sweep). The difference betweell traces A, B and C is in the sweep speed of the oscilloscope.

tnosec/cm 150 psec)

2psec/cm (Delay 150,usec)

NORMAL

Laser induced plmmas for analytical spectroscopy

327

In 1963 HARRIS [14] published high speed photographs showing the formation of metal plasmas (including aluminium) for peak power densities of approximately 2 MW/cm2. He was able to detect a delay in the vaporization process following the onset of irradiation. A thermal explosion was said to occur. HARRIS’S observations seemed to support speculations at that time [lo, 17, 18) that the absorption of intense laser radiation (of the order used) would heat the material to some depth below the surface to its vaporization temperature before the material had absorbed its latent heat of vaporization. This was thought to induce a pulse of high pressure in the underlying material and to result in the emission of the vaporized material similar to a chemical explosion (thermal explosion). In 1965, however, READY published a paper [19] in which the effects of non-Qswitched and Q-switched laser radiation were treated differently. Vaporization by normal mode lasers (peak powers of 104-10” W) was described by the conventional description of boiling, whilst that by Q-switched lasers ( 107-10s W) was treated phenomenologically as a superheating of underlying material due to the high recoil pressure of vaporized material. In 1967, GRECHIKHINand MIN’KO [13], using normal mode neodymium radiation in the power density range lo’--lOa W/cm 2, found that plasmoids were ejected discontinuously and that a time correlation existed between the ejection of plasmoids and the laser spikes. Furthermore, they showed, by streak photography, that the breakdown products of the material moved along tracks at various angles with the time axis, and that they occurred at the end of the plasma jet emission for metals. One reason for the ejection of such particles was given as the mechanical effect of the plasma during the emission period. They also reported the existence of density jumps in the ejected plasmoids, which indicated that the propagation velocities were supersonic when ‘incomplete expansion’ occurred. B~QERSHAUSENand VESPER, in a paper [15] on the production and heating of laser plasmas (1969), explained that the “spark spray” phenomenon (ejection of glowing particles) was due to liquid material being ejected by high back pressure out The single spikes of the normal laser pulse were of the crater during evaporation. said to follow each other with very much varying intensity so that spikes with smaller then average power have just sufficient energy to heat up the sample above the In this way deep melted zones are melting point without effective evaporation. formed which are evaporated by spikes of high radiation density and particles are ejected by the accompanying back pressure. To avoid “spark spray” it was necessa’ry to delay the arrival of successive spikes so that the temperature of the sample decreased below the melting point during the spiking intervals. A consequence of the ejection of liquid sample material is that the laser light is strongly scattered by the droplets and contributes, therefore, in a lesser way to the evaporation of the sample. We found that the initial heating of the sample was dependent on the mean This delay in the breakdown process (which power density of the laser radiation. was observed by HARRIS) is probably due to the fact that at the outset of laser action, especially at low mean power densities, the metal is not heated to vaporization temperature by the individual laser spikes, and the time interval between spikes may be large enough to warrant sufficient cooling of the metal so that vaporization by successive low intensity spikes is minimal. As the mean heat flux increases, vaporization 2

328

R. H. SGOTTand A. STRASEEIM

The flux density of individual laser pulses increases, becomes more predominant. even becoming sufficient to cause vaporization of small amounts of material originally at approximately %Y’C. The space-time nonstatiorla~ty of the heat flux results in a vapour cloud of ~homogeneous density due to the ejection of discr%teamounts of vapour. The vapour ejection is accompanied by a fine spray of particles, which increases for increasing power density of the incident laser radiation. This particle ejection is probably caused by the back pressure associated with non-uniform elementary acts of erosion. As the breakdown proceeds and the crater depth increases, the flow of vapour is accompanied by larger quantities of molten material. This is due to a melting and flushing mechanism which is more inertial than the relatively pure vaporization process. This quasi-continuous action is responsible for the removal of the bulk of molten material from within the crater. The fiushing mechanism may be described as the action of vapour, at a high pressure within the crater, on the melt-similar to a boiling process. The molten material is flushed up the walls of the deepening crater and protrudes above the surface of the metal, forming a cylindrically-shaped hollow sheath which contains the ilow of escaping vapour from the base of the crater, as The upper portion of the ejecting well as the converging beam of laser radiation. sheath ~ontinua~y breaks up into fragments which tend to scatter the incident laser radiation. The observed sheath contraction is considered to be a hydrodynamic phenomenon which may be explained by Bernoulli’s theorem. This consideration is based on the assumption that a substantial vapour flow exists during the period of contraction. ft is unrealistic to vigorously apply hydrodynamic equations to describe the phenomenon as the vapour flow is neither isothermal nor adiabatic. We do suggest, however, that a low pressure exists within the sheath. This low pressure is associated with a high flow-velocity of the vapour in this region, and could be the primary cause of the sheath contraction. When the contracting sheath interrupts the laser beam, a violent disintegration and vaporization process occurs, and particles are blown randomly in all directions. Further scattering of the incident laser beam undoubtedly occurs, nevertheless the quasi-continuous flushing process continues and a new sheath is formed which contracts in a similar manner. At this stage the laser radiation power is diminishing as the end of the discharge is neared, and further interaction is limited. The mixture of vapour and fragments of materia1 exhibits turb$ence and, in some cases, a twisting action. This vortex-like a~angement appears to spiral in towards a central column where the vapour rushes upwards. The mechanism is not fully understood, but it appears that the rapidly rising current and associated low pressure in the centre of the lower region is the primary cause of this phenomenon. 5. QUANTITATIVE DIRECT READINCJANALYSIS Usma

GATED INTEGRATION

The various laser plumes were investiga~d as spectrochemical sources for quantitative analysis. For comparison, a series of ten aluminium alloy standards was analysed for copper content.

Laser induced plasmas for analytical spectroscopy

329

Although a number of papers on laser microprobe analyses have been published [20], relatively few are concerned with direct reading quantitative analyses. A comparison of the various laser plumes as spectrochemical sources was therefore made using a direct reading detection system. Table 4 gives the experimental conditions that applied. Note that when no time resolution was used, the gates on the photomultipliers were set to integrate the total line emission from the plume. Time resolution was used in the case of the single and Table 4. Experimental Spectrometer entrance slit height above sample surface*

conditions for direct reading analysis

Plume type

(mm)

Time resolution

Normal mode Semi-Q-switched mode Semi-Q-switched mode Double pulse Double pulse

1

No

0

800

1

No

0

150

3.6 1 1 1

No No Yes No

0 0 1 0

150 20 19

1

Yes

1

19

Single pulse Single pulse

Gate delayt

Gating time

Met)

(Wet)

20

* See Fig. 10. t Triggered on leading edge of 1st laser pulse.

double pulse plumes to reduce the spectral background contribution. As found in Section 3, these are the only modes in which the spectral background can be successfully reduced by a gating technique. Integrated intensity readings of the copper 3274 and aluminium 2660 A lines were obtained for each individual laser shot. Each sample was presented 12 times and The laser was fired at 1 min intervals. each time a new surface area was vaporized. The data obtained was then processed by computer. For each laser shot the The averages of the copper and intensity ratio I CU 3274/I Al 2660 was calculated. aluminium line intensities, the intensity ratios and the standard deviations and variation coefficients were calculated for each set of readings. Linear calibration curves were obtained when the copper concentrations were plotted against the intensity ratios on log.log. graph paper. Figures 17 and 18 show the calibration curves obtained for four experimental conditions. The normal mode and double pulse mode gave similar calibration curves. As can be seen from the graphs, time resolution of the single and double pulse modes increased the gradients and thus the sensitivity. Also the gradient of the semi-Q-switched mode calibration curve is greater at 3.5 mm sampling height than at 1 mm. This result was expected as the line-to-background ratio in the upper region of the plume was less than in lower regions. However, increased self-absorption is evident in the higher concentration range, indicating a mantle of cool atoms around the high temperature plasma. Table 5 gives the values of the gradients of the calibration curves, the mean

R. 33. SCOT~C and A. %FRASEEIM

330 IO

1

I-

I Cu 3274 1 At 2660 .I*

1 Cu 3274

I At

2660 II -

-01 i “001

I

.Ol

I

,

I

I

Of.

SINGLE Fig.

X$3. Wibr&on

improvement

‘7

IO

cu PULSE

curves obtained for the single giant pulse mode, showing aSforded by the g&kg technique described in tha text.

the

variation coeflkienfs of the line intensities and the copper concentration for all modes. Mean variation coefficients are given for comparison purposes only. The best reproducibility of intensity ratios was obtained in the double and single pulse modes (5-8%), but the calibration curve gradients are poor and therefore the reprodu~ibilit~es in actual eon~ntration readings are poor. The best gradient obtained was for some-~-s~~itched mode at a height of 35 mm above the sample surfaces” This gave the best reproducibility of concentration values (17%), followed by time-resolved single pulse and time-resolved double pulse modes. These resuIts indicate that for quantitative analyses of alloys using laser radiation, time-resolved techniques are of value. Furthermore, for satisfactory re~~oducibiIity, an integration of multipIe shots is necessary. The fact that the low gra,dient amdysis curve for the double giant pulse mode gives improved results to that of the single pulse mode stresses the importance of multiple samplings,

Laser induced plasmas for analytical spectroscopy

331

Table 5 Mean variation coefficients

-

Plume type Kormal mode Semi-Q-switched mode (1 mm) Semi-Q-switched mode (3.5 mm) Double pulse Double pulse (time resolved) Single pulse Single pulse (time resolved)

Gradient of calibration curve

Copper intensity

Aluminium intensity

Concentration (from calibration curves)

0.4 0.3

26.9 % 10.0 %

24.3 % 14.7 %

36.5 % 39.0 %

0.5 0.25

7.3% 8.6%

9.0% 8.3%

17.1% 25.2 %

0.3 0.25

9.6% 12.8 %

8.1% 13.4%

17.4% 33.9 %

0.4

11.5%

9.2%

CONCLUSION Detailed high speed photographs of the interaction of normal mode laser radiation with an aluminium sample for power densities of 10s10’ W/cm2 have revealed a complex hydrodynamic system. After a heating period which is dependent on the power density, vaporization plays a major role, accompanied by fine particle ejection for the higher power densities. As more heat enters the system, the vaporization process gives way to a melting-flushing mechanism which is responsible for the removal of the bulk of the material from the crater. This flushing out of the melt is the cause of the formation of a sheath of material which protrudes above the sample surface. The sheath contracts, and elevated vaporization and disintegration occurs as a result of the interaction of the incident laser radiation with the upper portion of the contracting sheath. The resulting mass of molten material, fragments and vapour forms a vortex-like system in some cases which spirals upwards. We conclude that the interaction is essentially hydrodynamic in nature and that pure vaporization is not responsible for the bulk removal of material from the crater. Furthermore, the melting-flushing process is the result of heating by the incident laser radiation which penetrates the vapour cloud. This radiation is partially adsorbed in the v&pour, resulting in the pulsatory spectral emission of the vapour cloud. In the semi-Q-switched mode, vaporization plays the major role throughout the interaction, and the melting-flushing process, although in evidence on examination of the crater, is negligible. Detailed studies of the semi-Q-switchedinteraction showed that vaporization progresses in a pulsatory manner. The spectral emission of the plume occurs after a relatively long delay after the initial laser pulse due to the spacetime non-stationarity of the multipulse discharge. Comparative studies of the normal mode, semi-Q-switched mode and giant pulse mode plasmas by spatial and temporal resolution showed that higher temperatures occur in the giant pulse plasma and semi-Q-switched plasma than in the normal mode

332

R. H.

SCOTT

and A.

STRASEEIM

plasma. Self-absorption line shifts were recorded as well as emission line shifts indicating high pressure gradients. Finally, the results of comparative analytical studies indicate that significant improvement in the slopes of the analytical curves can be obtained by spatially selecting an area of the plume for analysis, as well as by using time resolution. Time resolution can only be used with any degree of success when the laser source is operated in the giant pulse mode (a maximum of three giant pulses). The technique allows the gating of spectral background which occurs mainly during the laser pulses, and the integration of the signal from the cooling, radiating atoms or ions. The reproducibilities which were obtained are single shot variations for the various modes, and it is easy to calculate the expected reproducibility of results for multiple integrations. An integration of multiple shots is necessary if the laser source is to be successfully used in quantitative analyses. Acknowledgements-The authors are grateful to W. IV. SCHROEDER and J. J. VAN NIEKERK their assistance with the gating device used for the time resolved analyses. The assistance constructive comments of F. BLU?~ is also appreciated.

for and