AVS2014 Session PS-WeA: Plasma Diagnostics, Sensors, and Control

Absorption spectroscopy is a powerful diagnostic for reactive plasmas, providing absolute density measurements of numerous atoms, molecules and free radicals in ground and various excited states. The sensitivity is determined in practice by the characteristics of the light source used: spectral range, stability and intensity. Previously Xe arc lamps have been used but they suffer from spatiotemporal fluctuations, limiting the sensitivity to about 10^-3 in absorption. More recently UV light-emitting diodes have been used, giving greatly increased stability, but these have very limited spectral ranges (a few 10’s of nm), necessitating the use of specific diodes for each species detected.

We have constructed a new absorption bench that overcomes these difficulties. The light from a broad-band (200-1000) nm laser-induced plasma source (Energetiq-LDLS EQ-99) is collimated and steered with reflective (achromatic) optics. After passing through the reactor the beam is analyzed with an aberration-corrected spectrograph (Acton SCT-320) equipped with a 1024 element photodiode array detector. Three gratings allow spectral ranges of 32, 63 or 250 nm to be analyzed at one time. This setup gives spectra in minutes with random noise and baseline stability better than 10^-4, allowing detection of species which only show weak absorption bands over wide spectral ranges.

The gases (O₂, Cl₂ and mixtures) are flowed through a cylindrical aluminum chamber (55 cm diameter, 10 cm height). The plasma is excited at 13.56 MHz by a 4-turn planar coil through a dielectric window. We observed molecular absorption bands from ground state Cl₂, vibrationally excited O₂ and of various O_xCl_y reaction products, allowing the densities, vibrational and rotation temperatures to be determined. As far as we know this is the first time oxychlorides densities have been measured in plasmas. This data is complemented by measurements of absolute atom densities (Cl and O) by TALIF and hairpin probe measurements of electron density. The interpretation of this data set will be discussed.

Hydrogen plasmas have been used for decades in the microelectronic industry with applications in the fields of deposition (PECVD, Plasma Enhanced Chemical Vapor Deposition) [1], etching [2] and surface treatment (reduction of the photoresist roughness in the lithography step [3]). However hydrogen is a very peculiar element due to his low mass and his electronegative character, and the mechanisms at stake in low pressure hydrogen plasma aren’t well understood yet. A complete fundamental study with reliable diagnostics would be highly valuable for many applications [4]. Moreover, hydrogen plasmas present a great potential interest for the treatment of new materials such as graphene [5] or carbon nanotubes [6]. To modify the surface of such ultrathin layers without damaging the material, very low ion energy bombardment is required (conditions such as those obtained in pulsed ICP reactor [7]). By contrast, for other applications such as etching of several nanometer thick layers, the ion energy must be very high to get a significant etch rate. To assist the development of innovative processes in H₂ plasmas, we have thus analyzed systematically CW and pulsed H₂ plasmas both with and without RF bias power. In particular, we carry out time-resolved ion flux, and time-averaged ion energy measurements in different pulsing configurations. A large variety of ion energies and shapes of IVDF are reported depending on pulsing parameters. The IVDF are typically very broad (due to the low ion transit time of low mass ion through the sheath) and either bi or tri-modal (H⁺, H₂⁺ and H₃⁺ contributions). The time variations of the ion flux in pulsed plasmas also presents peculiar features that will be discussed.

References

1. W. A. Lanford, and M. J. Rand, Journal of Applied Physics, 49, 2473-2477 (1978)

2. A. Efremov, N. K. Min, J. Jeong, Y. Kim, and K. H. Kwon, Plasma Sources Science and Technology, 19 (2010)

3. E. Pargon, L. Azarnouche, M. Fouchier, K. Menguelti, R. Tiron, C. Sourd, and O. Joubert, Plasma Processes and Polymers, 8, 1184-1195 (2011)

4. M. Sode, T. Schwarz-Selinger, and W. Jacob, Journal of Applied Physics, 113 (2013)

5. E. Despiau-Pujo, A. Davydova, G. Cunge, L. Delfour, L. Magaud, and D. B. Graves, Journal of Applied Physics, 113 (2013)

6. A. Hassanien, M. Tokumoto, P. Umek, D. Vrbanic, M. Mozetic, D. Mihailovic, P. Venturini, and S. Pejovnik, Nanotechnology, 16, 278 (2005)

7. C. Petit-Etienne, M. Darnon, P. Bodart, M. Fouchier, G. Cunge, E. Pargon, L. Vallier, O. Joubert, and S. Banna, Journal of Vacuum Science & Technology B, 31 (2013)

Over the last decade, semiconductor industry growth has been driven chiefly by the demand for consumer electronics: the move to mobile smart devices such as phones and tablet PC’s. It is now common place for hand-held mobile devices to have 32 Gb of memory and processor speeds of over 1 GHz, a truly remarkable feat that would have been unthinkable 10 years ago. This capability has been enabled by the continuation of IC scaling to smaller and smaller features sizes with the present technology being mass produced by 28 nm node technology and smaller nodes down to 10 nm currently being developed by IC manufacturers. The limitations of lithography to keep up with the decrease in dimensions required for these smaller nodes has resulted in new challenges for plasma etch to enable patterning at these small feature sizes. Device performance requirements also drive critical dimension (CD) non-uniformity to less than one nanometer across the entire 300 mm wafer for sub-20 nm features and yield requirements extend this pattering region to within 1.5 mm of the wafer edge. Wafer fabrication production also relies on plasma etch solutions to be stable at these levels across long periods of time and capable of flexibility in multiple applications. The realization of all of these goals has been greatly facilitated by a much better understanding of the basic chemical, physical, and electromagnetic processes that occur during the plasma etch of semiconductor devices.

In this paper we will discuss the crucial role diagnostics play in achieving this understanding and in the development of state of the art plasma etch chamber technology that allow the continuation of Moore’s Law. Diagnostics are essential not only to understand etch mechanisms and chamber characteristics but to also accelerate hardware development in order to meet customer time critical needs. We will review the different types of diagnostics commonly used in plasma etch chamber development with reference to findings from literature and augment this with diagnostic work undertaken at Lam Research. Finally, we will discuss the suitability of diagnostics in main stream production and give some thoughts on future diagnostics that may be required for production enhancement and also angstrom level etching.

In microelectronics fabrication the angular distribution of the bombarding ions can impact the process outcome. The ion energy distribution as a function of ion angle at specific locations on the substrate or wafer surface need to be controlled in certain anisotropic etching and conformal deposition plasma processes. We report a novel method for the measurement of ion energy distributions as a function of ion angle, at the substrate location, using a planar retarding field analyser.

Planar retarding field analyzers are commonly used to measure ion energy distributions but provide no information about the angular distribution of ions bombarding the substrate surface. Here, we report on a novel planar retarding field analyser design capable of resolving the angular distribution of the energetic ions. The design has three active grids, a collector plate and an aperture with variable aspect ratio (height / diameter) to control the angular spread of the ions allowed through the device for detection. First, the potential of the ion energy discriminating grid is modulated to select ions with a specified energy resolution for analysis. Then, the aspect ratio of the aperture is varied from large acceptance angle to narrow acceptance angle with specified angle resolution - predetermined from the aperture geometry. The ion current is recorded for each acceptance angle to give an integral form of the ion angular distribution at a given ion energy where the angular distribution can be recovered by taking a first derivative. The procedure is then repeated for each ion energy. Once the angular distribution is determined as a function of ion energy, the energy distribution as a function of ion angle is easily calculated.

The analytical theory used to define ion current as a function of incident ion angle, ion energy and aperture aspect ratio is presented. The method used to vary the aspect ratio of the additional aperture is also discussed. This novel method allows ion angular distributions to be determined using a compact planar retarding field analyser.

With the modified retarding field analyzer design and advanced analytical technique, ion angular distributions with angle resolution as low as 3^o have been measured and resolution up to 1^o can be achieved. This technique adds important functionality to the retarding field analyser technology - which has become one of the most important technologies in the field of plasma diagnostics in recent years.

Electrostatic (Langmuir) probes are powerful instruments for diagnostics of non-equilibrium plasmas in experimental and industrial plasma reactors. There are three levels of the probe diagnostics comprising different equipment complexity and having different accuracy of obtained the plasma parameters. These three approaches are based on inferring the plasma parameters from: i – the ion part of the probe characteristic, ii – the electron part of the probe characteristic (classical Langmuir probe method), and iii - by differentiation of the probe characteristic to obtain an Electron Energy Distribution Function, EEDF. Then, the electron temperature T_e and plasma density N, as well as, the rates of collisional processes and transport coefficients are found as corresponding integrals of the measured EEDF. Methods i and ii assume Maxwellian EEDF which is not valid for most cases of non-equilibrium plasmas. This and many others questionable assumptions in methods i and ii make actual EEDF measurements the only reliable contemporary probe diagnostics. Langmuir probes in plasma processing reactors are often subjected to the probe surface contamination and high level of rf and low frequency noise. Reliable EEDF measurements require the probe system capable of efficient mitigation of these environmental distortions. The presentation main subject is comparison of EEDF measurement results obtained with different commercial probe systems. It is shown that the measurement accuracy of the plasma parameters in many commercial probe systems is compromised by heavily distorted EEDF in low and high energy region. Low energy distortions make impossible to detect low energy electrons (comprising the majority of electron population), thus leading to underestimation of the plasma density, while high energy distortions make impossible detection of fast electrons producing excitation and ionization. Therefore, some commercial probe systems yielding distorted EEDFs are unable to reveal any additional valid information to that obtained with classical Langmuir procedure.

Silicon nitride thin films have shown many useful applications in microelectronic and optoelectronic industries. Fabrication, of these films at low temperature, is done typically by PECVD using a mixture of silane (SiH₄) and ammonia (NH₃). Recent trend shows that the most practical deposition of low-hydrogen-content-silicon-nitride-films (SiNx:H) is to use N₂ instead of NH₃ as the main nitrogen source. However, N₂ has an inherently much higher bonding energy than NH₃, which makes N₂ more difficult to dissociate into free nitrogen active species, thus nitrogen deposition rate is significantly reduced. But if low hydrogen nitride films can be obtained, which may give better device performance, the deposition rate may not be an important factor and PECVD of silicon nitride by little addition of SiH₄ and NH₃ to the N₂ is still an attractive process. Moreover, the important deposition parameters for any PECVD process are RF power, working pressure, substrate temperature, the gas flow ratio of the reactant gases and the electrode spacing (for CCD or parallel plate system). All these parameters have significant role on the deposition and etch rates along with other physical and optical properties of film depending on device applications.

Electron beam generated plasmas are characterized by high plasma densities and very low electron temperatures, making them well-suited for next-generation processing techniques where high fluxes of low energy ions are desirable. In this work, we focus on plasma generation in fluorine containing gas backgrounds due to their relevance to a number of industrial plasma applications. In particular, we focus on the effect of fluorine-containing gas dilution on the plasma properties of electron beam generated plasmas including electron density, total plasma density, electronegativity, and electron temperature. These parameters are measured through a combination of Langmuir probe, and RF impedance spectroscopy techniques.

Atomic hydrogen recombination rate is measured for a range of materials. Toroidal plasma sources are then built with these materials as the plasma facing wall and are characterized for their discharge parameters and atomic hydrogen output. The discussion will include a description of the diagnostic tools used in this study; a comparison of modeled source discharge parameters running in Ar and H2 with experiment along with a brief comparison of the impact of volumetric and surface recombination of radical species is presented.

This work was specifically targeted for the use of radicals produced by a toroidal remote plasma sources for semiconductor applications but is generalizable to discharges in hydrogen independent of excitation or application.

Surface recombination of the hydrogen radical is shown to be a dominant mechanism in determining process parameters for semiconductor applications.