Application of Femtosecond Laser Ablation Time-of-Flight Mass

Jun 5, 2003 - Vanja Margetic, Kay Niemax, and Roland Hergenro1der*. Institute for Spectrochemistry and Applied Spectroscopy, Bunsen-Kirchhoff Strasse ...
0 downloads 0 Views 96KB Size
Anal. Chem. 2003, 75, 3435-3439

Application of Femtosecond Laser Ablation Time-of-Flight Mass Spectrometry to In-Depth Multilayer Analysis Vanja Margetic, Kay Niemax, and Roland Hergenro 1 der*

Institute for Spectrochemistry and Applied Spectroscopy, Bunsen-Kirchhoff Strasse 11, 44139 Dortmund, Germany

A femtosecond laser system was used in combination with a time-of-flight mass spectrometer (TOF-MS) for in-depth profiling of semiconductor and metal samples. The semiconductor sample was a Co-implanted (1017ions/cm3) silicon wafer that had been carefully characterized by other established techniques. The total depth of the shallow implanted layer was 150 nm. As a second sample, a thin film metal standard had been used (NIST 2135c). This standard consisted of a silicon wafer with nine alternating Cr and Ni layers, each having a thickness of 56 and 57 nm, respectively. An orthogonal TOF-MS setup was implemented. This configuration was optimized until a sufficient mass resolution of 300 (m/∆m) and sensitivity was achieved. The experiments revealed that femtosecond-laser ablation TOF-MS is capable of resolving the depth profiles of these demanding samples. The poor precision of the measurements is discussed, and it is shown that this is due to pulse-to-pulse stability of the current laser system. Femtosecond-laser ablation TOFMS is shown to be a promising technique for rapid indepth profiling with a good lateral resolution of various multilayer thin film samples. Multilayer coatings of different compositions and thickness are widely used in material science and in the production of hightechnology materials.1,2 Important characteristics of the materials can be significantly improved by the use of single- or multicomponent thin layers. Appropriate methods for in-depth profiling in the nanometer to micrometer range are required to improve the technology of such advanced materials. Today, typical methods for characterizing such thin layers are secondary ion mass spectrometry (SIMS), dc- or rf- glow discharge (GD-) optical emission (OES) or mass spectrometry (MS). SIMS is used for nanometer-thin layers and cannot be used for layers thicker than a few micrometers. GD OES/MS has a depth resolution of ∼10 nm and can also be used for thicker layers; however, it is limited by a poor lateral resolution, at best, a few millimeters, and specific requirements on the form and dimension of the sample. For a detailed discussion on the figures of merit of these and related * To whom correspondence should be addressed. E-mail: hergenroeder@ isas-dortmund.de. (1) Grasserbauer, M.; Werner, H. Analysis of Microelectronic Materials and Devices; Wiley: New York, 1991. (2) Ives, M.; Lewis, D.; Lehmberg, C. Surf. Interface Anal. 1997, 25, 191-201. 10.1021/ac020791i CCC: $25.00 Published on Web 06/05/2003

© 2003 American Chemical Society

surface techniques, see Bubert et al.3 It is clear that new, alternative techniques would be of great value. Coupling a time-of-flight mass spectrometer (TOF-MS) with a laser-induced mass removal process is an attractive combination for the chemical characterization of solid material. A TOF mass analyzer is inherently multielement-capable and is well-suited to transient ion signals, such as those produced by a laser ablation (LA) event. The capability of TOF-MS to analyze charged particles from a laser plasma has been discussed in several articles.4-6 The advantages of laser ablation are well-known: no sample preparation is needed; conducting and nonconducting samples of arbitrary structure can be analyzed directly; spatial resolution down to a few micrometers can be obtained; and a rapid, simultaneous multielement analysis is possible. Over the last three decades, most of the efforts were invested in the development of LA as a technique for bulk analysis. Only recently it has been demonstrated that LA in combination with an inductively coupled plasma (ICP),7,8 direct optical emission spectroscopy of the laser produced plasma (LIBS),9-12 or LA combined with a TOF-MS13,14 is also suitable for in-depth profiling of solid samples. Typically, Nd:YAG or excimer lasers with laser wavelengths in the near-UV or near-infrared and laser pulse durations of a few nanoseconds are used in LA. Femtosecond-lasers (fs-lasers) have only been used in the past few years in analytical spectroscopy,15-19 because of the previous unavailability of reliable, high-power femtosecond lasers. Since (3) Bubert, H., Jenett, H., Eds; Surface and Thin Film Analysis; Wiley- VCH Verlag: Weinheim, 2002. (4) Vaeck, L. V.; Struyf, H.; van Roy, W.; Adams, F. Mass Spectrom. Rev. 1994, 13, 189-208. (5) Sysoev, A. A.; Sysoev, A. A. Eur. J. Mass Spectrom. 2002, 8 (3), 213-232. (6) Ledingham, K. W. D.; Singhal, R. P. Int. J. Mass. Spectrom. Ion Process. 1997, 163, 149-168. (7) Plotnikov, A.; Vogt, C.; Hoffmann, V.; Ta¨schner, C.; Wetzig, K. J. Anal. At. Spectrom. 2001, 16, 1290-1295. (8) Bleiner, D.; Plotnikov, A.; Vogt, C.; Wetzig, K.; Gu ¨ nther, D. Fresenius’ J. Anal. Chem. 2000, 368, 221-226. (9) Vadillo, J. M.; Garcia, C. C.; Palanco, S.; Laserna, J. J. J. Anal. At. Spectrom. 1998, 13, 793-797. (10) Anderson, D. R.; McLeod, C. W.; English, T.; Smith, A. T. Appl. Spectrosc. 1995, 49 (6), 691-701. (11) Garcia, C. C.; Corral, M.; Vadillo, J. M.; Laserna, J. J. Appl. Spectrosc. 2000, 54 (7), 1027-1031. (12) St. Onge, L.; Sabsabi, M. Spectrochim. Acta, Part B 2000, 55, 299-308. (13) Vadillo, J. M.; Laserna, J. J. J. Anal. At. Spectrom. 1997, 12, 859-862. (14) Kim, M. K.; Takao, T.; Oki, Y.; Maeda, M. Jpn. J. Appl. Phys. 2000, 39, Part 1, No. 11, 6277-6280. (15) Margetic, V.; Bolshov, M.; Stockhaus, A.; Niemax, K.; Hergenro ¨der, R. J. Anal. At. Spectrom. 2001, 16, 616-621.

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003 3435

an increasing number of commercial application of fs-lasers are found, for example, surgery20 and micromachining,21 the situation has considerably changed. Nowadays, fs-lasers are fully selfcontained, diode-pumped, and solid-state technology. There is a fundamental difference between the ablation process of ultrashort (10 ps) pulses, which results in different mechanisms of energy dissipation in the illuminated sample. Whereas in the case of ultrashort laser pulses, at the end of the laser pulse, only a very hot electron gas and a practically undisturbed lattice is found, for longer pulses, it is typical that during the laser pulse, the material undergoes changes in the thermodynamic state from solid, through liquid, into a plasma state. The time taken to transfer the laser energy from the excited electron gas to the lattice via collisions and to start the removal of the material is ∼10 ps.22,23 This is not long enough for significant thermal diffusion to take place into the surrounding material, even in materials such as metals with a high thermal diffusivity (e.g., Au: 0.93 cm2/s (1000 K)). In the case of the ultrashort pulse, the zone affected by the laser around the laser spot is less than several tens of nanometers. This is different for longer pulses. In the case of a 10-ns pulse, the heated zone would be around 1 µm. Although it is easy to focus a short-pulse laser down to several micrometers, the lateral resolution is restricted to the area of collateral damage, which can be much larger than the optical resolution. The same has to be expected for the depth resolution. An ablation rate of several tens of nanometers can be achieved with a short-pulse laser, but one has to expect thermal mixing and distortion of the boundary layers and a degradation of the depth resolution, which clearly has to be distinguished from the ablation rate.7,24 In such a situation requiring high lateral resolution with subsequent laser shots in close spatial proximity or medium to high depth resolution, ultrashort pulse lasers offer new possibilities because of the strong confinement of the delivered laser energy. The goal of our work is to test these expectations with well- characterized metallic and semiconductor samples. THEORETICAL DISCUSSION The interaction of femtosecond laser pulses with metals and metal-like materials has been intensively studied over the past few years, and a detailed model, namely the one-dimensional, twotemperature model, has been developed to describe the lightmaterial interaction.22,23 The electron subsystem and the lattice can be described by the evolution of two different temperatures (16) Margetic, V.; Pakulev, A.; Stockhaus, A.; Bolshov, M.; Niemax, K.; Hergenro ¨der, R. Spectrochim. Acta, Part B 2000, 55, 1771-1785. (17) Margetic, V.; Niemax, K.; Hergenro ¨der, R. Spectrochim. Acta, Part B 2001, 56, 1003-1010. (18) Russo, R. E.; Mao, X.; Gonzalez, J. J.; Mao, S. S. J. Anal. At. Spectrom. 2002, 17, 1072-1075. (19) Ye, M. Q.; Grigoropoulos, C. P. J. Appl. Phys. 2001, 89, 5183-5190. (20) Loesel, F. H.; Fischer, J. P.; Go ¨tz, M. H.; Horvarth, C.; Juhasz, T.; Noack, F.; Suhm, N.; Bille, F. J. Appl. Phys. B 1998, 66, 121-128. (21) Kautek, W.; Kru ¨ ger, J. In Laser Materials Processing: Industrial and Microelectronics Applications; Beyer, E., et al., Eds.; SPIE: Bellingham, WA, 1994; Vol. 2207. (22) Nolte, S.; Momma, C.; Jacobs, H.; Tu ¨ nnermann, A.; Chichkov, B. N.; Wellegehausen, B.; Welling, H. J. Opt. Soc. Am. B 1997, 14 (10), 27162722. (23) Chichkov, B. N.; Momma, C.; Nolte, S.; von Alvensleben, F.; Tu ¨ nnermann, A. Appl. Phys. A 1996, 63, 109-115. (24) Wetzig, K.; Baunack, S.; Hoffmann, V.; Oswald, S.; Pra¨ssler, F. Fresenius’ J. Anal. Chem. 1997, 358, 25-31.

3436

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

after the laser pulse: Te (electrons) and Ti (lattice). Thermal diffusion of the hot electrons into the bulk takes place simultaneously with the transfer of energy to the lattice (electron-phonon coupling) by collisions. The heat diffusion of the electron subsystem is much faster than the lattice-mediated thermal diffusion.25 Therefore, the thermal conductivity of the lattice can be neglected. The temporal and spatial evolution of the electrons and lattice, temperatures, Te and Ti, respectively, can be described by the following coupled one-dimensional differential equations:

∂Te ∂Q(z) ) - γ(Te - Ti) + S ∂t ∂z

Ce

Ci

∂Ti ) γ(Te - Ti) ∂t ∂Te ∂z

Q(z) ) -ke

S ) I(t)R exp(-Rz) where z is chosen perpendicular to the sample surface. Q(z) is the heat flux; I(t), the laser intensity measured in [W/m2]; R, the material absorption coefficient, including the surface absorptivity; and Ce and Ci are the electron and lattice heat capacities per unit volume, respectively. γ is the electron-phonon coupling coefficient, and ke, the electron thermal conductivity.26 Assuming an Arrhenius-type of evaporation, solving the coupled equations for ultrashort and using the criteria that CiTi > FΩ, where F is the density and Ω, the specific heat per unit mass for the occurrence of a significant removal of material, the ablation depth per laser pulse can be calculated. Two cases have to be distinguished: • For low fluence, the number density of hot electrons is considered to be low enough that the energy transfer between electrons and the lattice occurs only within the volume characterized by the optical penetration depth, δ. • For high fluences, the volume of the lattice heated by the electron gas is given by the electron diffusion length l, defined by l ) (Dτa)1/2, with D being the electron diffusivity constant and τa quantifying the typical time necessary for the energy transfer from the electron to the lattice. In the low fluence case, the ablation depth per pulse, LδR, is calculated to be

LδR ) δ ln

( )

F , Fδth ) FΩδ, F ≡ absorbed fluence δ Fth

whereas the high fluence case is described by

LlR ) l ln

( )

F , Flth ) FΩl Flth

F δth and F lth are the ablation thresholds for the low and high fluence regimes, respectively. A typical energy dependency measurement of the ablation rate is shown in Figure 1. In this (25) Wellershoff, S.-S.; Hohlfeld, J.; Gu ¨ dde, J.; Matthias, E. Appl. Phys. A 1999, 69 [Suppl.], 99-107. (26) Furusawa, K.; Takahashi, K.; Kumagai, H.; Midorikawa, K.; Obara, M. Appl. Phys. A 1999, 69 [Suppl.], 359-366.

Figure 1. The LA-ICP-MS signals as functions of the laser fluence on a logarithmic scale. Sample, brass; circles, Cu; triangles, Zn.

case, an ICP signal that is proportional to the ablated mass is plotted as a function of the laser fluence. Clearly, both regimes can be distinguished. This measurement was performed for copper, but the same principal curves can be measured for silica or other metals, such as chromium or nickel. The only difference is the thresholds. Obviously, for in-depth profiling with a high depth resolution, the ablation rate has to be chosen to be as low as possible. The following experiments are all conducted in the low fluence regime with the energy as close as possible to the threshold. Therefore, an ablation rate in the order of the optical penetration depth, which is for most metals ∼10 nm, can be expected.27 EXPERIMENTAL SECTION Fs-Laser System. The commercial fs-laser CPA-10 (ClarkMXR Inc., MI) was used for ablation. The seed laser of the system is an erbium-doped fiber ring laser.28 Typical parameters of the seed pulses are pulse energy, ∼0.15 nJ; mean power, ∼6 mW; and pulse duration, 150 fs. The seed pulses are amplified in a Ti/ sapphire laser working under the principle of chirped pulse amplification CPA.29 The parameters after amplification were central wavelength of the pulse spectrum, 775 nm; spectral bandwidth, ∼5 nm; pulse energy, ∼0.5 mJ; pulse duration, 170200 fs; and repetition rate from single pulse up to 10 Hz. The relative standard deviation of the pulse energy was 5%. Important features in a fs-laser ablation experiment are post- and prepulses, which are connected to the process of amplification. The ratios of the post- and prepulse intensities to the main pulse were found to be better than 100:1 and 500:1, respectively. The laser was focused onto the sample by a single lens with a focus length of 200 mm. The lateral resolution currently achievable with this setup is ∼30-40 µm. Time-of-Flight Mass Spectrometer. A schematic diagram of the time-of-flight mass spectrometer is shown in Figure 2. The TOF-MS was designed and built at ISAS (ISAS, Berlin). It is a 0.5-m linear TOF-MS with a microchannel plate detector that can (27) Ba¨uerle, D. Laser Processing and Chemistry, 3rd ed.; Springer-Verlag: Berlin, Heidelberg, New York, 2000; p 699. (28) Tamura, K.; Ippen, E. P.; Haus, H. A.; Nelson, L. E. Opt. Lett. 1993, 18, 1080-1082. (29) Strickland, D.; Mourou, G. Opt. Commun. 1985, 56, 219-221.

Figure 2. Experimental arrangement of the fs-laser ablation TOFMS. (Left) 3D view of the lower part of the instrument: the sample holder (movable in two directions), the horizontal slit, the repeller plate, the grids with accelerating voltage (Vdrift), the deflection plates at the beginning of the drift tube, the incident laser beam, and the ion trajectories. (Right) Electrical connections.

Figure 3. Fs-laser ablation TOF-MS spectrum of the Co-implanted silicon sample. Insert: saturated silicon signal.

be operated in analogue and pulse-counting mode. The residual pressure was kept below 10-6 mbar. The sample is placed vertically at the side of the entrance to the drift tube. The distance between the repeller plate and the first drift tube grid is 27 mm. The ablation takes place 5 mm above the repeller. A narrow horizontal slit (1 mm) in front of the sample reduces the dispersion of the ion beam in the vertical direction of the drift tube. The typical voltage of a rectangular repeller pulse is 300-400 V, and its duration 4-20 µs. Sample Preparation and Characterization. A silicon wafer with a thin and shallow implanted layer of Co ions (1017 ions/ cm2) was prepared (Johannes Kepler University of Linz). The sample was sliced into small pieces (1.2 × 1.1 cm) and characterized by Rutherford backscattering spectrometry (RSB) and total reflection X-ray fluorescence (TXRF) in combination with a recently developed wet chemical etching technique, both wellestablished methods.30 As a second sample, a thin film metal standard was used (NIST 2135c). This standard consists of a silicon wafer with nine (30) Klockenka¨mper, R.; von Bohlen, A.; Becker, H. W.; Palmetshofer, L. Surf. Interface Anal. 1999, 27, 1003-1008.

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3437

Figure 4. Circles, LA-TOF-MS intensity-pulse number profile for the Co-implanted silicon sample; full line, fast Fourier transform smoothed LA-TOF-MS intensity. Dotted line, TXRF of etching HF solutions concentration-depth profile; squares, RBS of the original nonetched wafer (“as implanted”).30

alternating Cr and Ni layers, each having a thickness of 56 and 57 nm, respectively. RESULTS AND DISCUSSION In-Depth Profiling of Silicon. As already stated, a necessary condition for a high depth resolution is a low ablation rate. However, the ion yield and kinetic energy of the ions, which both influence the sensitivity, as well as the mass resolution are also functions of the applied laser fluence. Therefore, initial measurements on a pure silicon sample were conducted to optimize these parameters by establishing the optimal fluence: The fluence should be low enough that the MCP detector is not saturated and the mass resolution of different elements is possible. The optimal conditions were achieved with fluences of 200-400 mJ/cm2, (Fth (Si) ) 200 mJ/cm2).14 The optimization of other instrumental parameters (delay and voltage of the repeller pulse, the drift tube voltage, the deflection plates voltage) were performed until a satisfactory mass resolution and signal intensity range were obtained. The sensitivity of the MCP can be varied slightly, and the software enables up to 10× amplification of the signal. When the signal is accumulated from several laser pulses, the average value is stored. The Co implanted sample was analyzed using the optimal parameters. A typical mass spectrum can be seen in Figure 3. Because of shunt capacities in the connectors, the mass peaks show a slight overshoot. An internal standardization to the silicon signal was not possible because the mass resolution changes during the ablation and the dynamic range of the MCP is limited (2.5 orders of magnitude). The result is shown together with the previous RBS and TXRF measurements in Figure 4. A Fourier-smoothed curve was fitted to the measured points. The fs-LA measurement clearly follows the expected concentration depth profile. The RBS and TXRF measurement revealed a total implantation depth of 150 nm, with a half width at full maximum of 60 nm. An ablation rate of 4.5 nm per pulse can be estimated from the comparison of these measurements. 3438 Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

Figure 5. Fs-LA-TOF-MS intensity-pulse number in-depth profile of the NIST standard 2135c. Squares, Cr; circles, Ni.

In-Depth Profiling of NIST 2135c. A similar procedure for optimizing the mass-resolution, ablation rate, and ion yield was applied. In comparison to the previous sample, only the major components Cr and Ni had to be measured. A further complication could come from the fact that both metal thin films have different ablation thresholds, and a mean value has to be found. However, no significant difference could be found within the errors of our measurements. The value chosen as optimal was 170 mJ/cm2. Each mass spectrum was stored separately, and zero results, which are measurements without an ion current due to energy fluctuations, were neglected. The result of the analysis of the standard is shown in Figure 5. The signals are normalized to the sum. A signal variation between 0 and 1 indicates a clean ablation without contamination of subsequent layers by previous layers. It can be seen that the first six to seven layers could be resolved, and an ablation rate of ∼2 nm can be estimated. The resolution decreases for the deeper layers. This could be caused by the inhomogeneity in the laser beam profile.15 A resolution of the interface is not possible and cannot be expected with this technique. Auger sputter depth profiling or SIMS would be

necessary to achieve this. However, even then the interface is hardly to be seen.31 It should be kept in mind that the task of the experiment was to confirm the theoretically predicted ablation rate and to investigate where a fs-laser ablation method could find its niche. Precision. In both cases, the extremely thin and shallow depth structures could be resolved to a certain degree; however, the precision is poor in both cases. To achieve the required depth resolution of a few nanometers, it is necessary to work close to the threshold of the given material. Small fluence variations have a large impact on the ablation rate as a result of the exponential, Arrhenius-type of ablation. This can be seen from the relative change of ablation rate (∆L/L) as a function of the relative fluence fluctuation (∆F/F), with Fth being the threshold fluence.

∆L ) L

1 ∆F F F ln Fth

If the applied fluence is close to Fth, the logarithmic in the denominator tends toward 0, which describes the large fluctuation in the ablation rate. Therefore, the stability in the laser energy is a crucial parameter in this kind of application. The fs-laser in this experiment has a mean energy stability of (5%, which is poor when compared to state-of-the-art laser systems, in which an energy stability around 1% is realized. In this case, it could be expected that the same resolution can be achieved with an improved precision. Further problems arise from the mechanical stability of the beam delivery system. In our current system, the laser beam path (31) National Institute of Standards and Technology, Certificate Standard Reference Material 2135c, 1999.

is several meters. A size reduction would improve the stability of the overall system. CONCLUSION The principal feasibility of a femtosecond laser ionization timeof-flight mass spectrometer for in-depth characterization of layered material has been demonstrated. It could be proven that femtosecond laser ablation is a tool that even on high thermalconductivity material offers the possibility to achieve ablation rates around 10 nm and even better without significant collateral damage in the surrounding material. This is necessary to reach high lateral and depth resolution. It has been shown that the selected orthogonal TOF configuration achieves a mass resolution of the order of 300, which is sufficient for the current applications. For the future, it is easy to envision an improved system with a reflectron. The full capabilities of a femtosecond laser ablation instrument could be achieved if the material is removed under nonhighvacuum conditions or even under ambient atmosphere. The instrument would combine a micrometer lateral resolution with a depth resolution of the order of 10 nm and, therefore, may find a niche between GD-OES/MS, which has a much lower lateral resolution, and SIMS or Auger sputter depth profiling, which has a much better resolution but needs ultrahigh-vacuum conditions. ACKNOWLEDGMENT The financial support from the Deutsche Forschungsgemeinschaft (DFG) contract no. He 1941-2/2 is gratefully acknowledged. Received for review December 30, 2002. Accepted April 8, 2003. AC020791I

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3439