Sputtering in radio frequency powered discharges Control of the sputtering energy

One of the interesting features of an RF discharge is that it allows more control of the effective DC bias potential at the sample and thus to some extent separate control of the energy and flux of the sputtering ions. This control allows sputtering to be done with very low-energy ions if desired – conditions that have found use both in the cleaning of samples prior to analysis and in the preparation of damage free surfaces for inspection by high resolution electron microscopy and atomic beam methods (K. Shimizu and T Mitani, 2007).

How is this control possible?
In a DC discharge, most of the applied potential occurs across the cathode sheath – the thin region with a positive space charge in which ions from the negative glow are accelerated towards the surface to cause sputtering of neutral atoms and electrons. It is these secondary electrons that allow the discharge to be self-sustaining, and the bias voltage in a DC glow normally depends on the secondary electron yield, bias voltage increasing as the secondary electron yield decreases (see secondary_electron_yield.htm). The actual energy of the ions arriving at the cathode (i.e. the sample) depends on the number of collisions with neutral atoms that an ion makes in the sheath, but can be comparable to the sheath voltage. The energy of ions incident upon the sample in an RF discharge can be lower than for a similar DC discharge for two reasons which are explained below.
First, in an RF discharge there is an additional mechanism available to accelerate electrons to energies where electron impact ionization can occur. The electrode sheaths oscillate at the applied RF frequency, moving towards and away from the electrodes as each electrode changes between acting as an anode or a cathode; this motion can impart energy to electrons as they are reflected by the sheath potential in a process known as ‘wave-riding’ or stochastic heating (e.g. Belenguer and Boeuf, 1990).  At low discharge powers the plasma electrons are predominately heated by the wave-riding mechanism (the alpha regime) whereas at high discharge powers it is the secondary electrons emitted from the electrodes by ion bombardment that do most of the work, having been accelerated within the sheaths. This transition between wave-riding or secondary emission being the dominant electron production process is also marked by a decrease in the mean electron energy. An RF plasma in the secondary electron regime must have a root mean square voltage comparable to that of a similar DC discharge, but the voltage in the wave-riding regime can be considerably lower.
Second, in an RF discharge with asymmetric electrodes, as is the case for an analytical glow discharge, there will be at steady-state a constant self-bias voltage between the electrodes on which the RF voltage is superimposed (e.g. Lieberman and Lichtenberg, 1994). It is this bias voltage which will determine the energy of ions which strike the sample (acting most of the time as a cathode because it is the electrode with the smaller effective surface area). The DC bias voltage is determined by the electrode geometry and the discharge conditions but will generally be of the order of half the peak-to-peak RF voltage.
So, the bias voltage in an RF discharge is considerably less than the peak-to-peak applied RF voltage and the voltage necessary to sustain an RF discharge is anyway lower than for a DC discharge. Thus the energies of the ions incident upon the sample may also be lower allowing greatly reduced or no sputtering when the discharge is used at low powers.

References:

• P Belenguer and J P Boeuf “Transition between different regimes of rf glow discharges”, Phys. Rev. A 41(8);1990; pp4447-4459.
• K Shimizu and T Mitani (2007) “A novel use of rf-GD sputtering for sample surface preparation for SEM: its impact on microscopy and surface analysis” presented at the European Working Group on Glow Discharge Spectroscopy meeting, Brussels, September 2007.
• M A Lieberman and A J Lichtenberg “Capacitive Discharges” chapter 11 in “Principles of plasma discharges and materials processing”, Wiley-Interscience 1994 (1st edition) ISBN 0-471-00577-0

Additional background material:

• B. Chapman, Glow Discharge Processes; Wiley, New York, 1988
• Bogaerts A., Neyts E., Gijbels R., van der Mullen J., Gas discharge plasmas and their applications, Spectrochim. Acta: Part B 57; 2002; 609.

First published on the web: 15.02.2008

Author: James Whitby. The text is based on a lecture given by Philippe Belenguer (LAPLACE Toulouse), at the first GLADNET training course in Antwerp Sept. 2007.

TOP

Sputtering rate

In 1972, Paul Boumatns introduced the following equation for the sputtering rate, q, in GD-OES⁽¹⁾

where q is in mass/s, C_Q is a sputtering constant which may vary with the plasma gas species but not with current, potential or pressure in the source, i_g is the current, U_g the voltage and U₀ a turn-on voltage, typically 300 V in DC operation.

In 1994, Payling suggested a modification to this equation when he found sputtering rates were not quite linear with voltage⁽²⁾

where N = 0.74. With the modified equation, U₀ is typically about 400 V in DC operation, so that when the two equations are plotted together they appear very similar. As a result many continue to use the original Boumans' equation.

Blue points are original equation, red points are the modified equation

The sputtering rate can also be expressed as

where P_g is the power and P₀ is a constant. Since P₀ is generally quite small, perhaps 2-3 W, the sputtering rate is nearly proportional to power. But also depends on the carrier gas pressure in the discharge source.

References:
 (1) P W J M Boumans, Anal. Chem. 44, 1219 (1972).
 (2) R Payling, Surf. Interface Anal. 21, 791 (1994).

First published on the web: 15 May 2000.

Authors: Richard Payling & Thomas Nelis

TOP

Profilometers

Sputtering rates are normally measured by recording the depth of a crater after a several minutes of sputtering. The sputtering rate per unit area (g/s/m²) is then equal to the depth of the crater (um) divided by the time of sputtering (s) and multiplied by the density of the material (g/cm³).

Several types of instruments are used to measure crater depths. Amongst these are profilometers, also called surface roughness instruments, and interferometers. The most common profilometers use either a contact diamond stylus, optical focus, or laser. Some profilometers are capable of recording the whole crater but many are capable of only a single trace across the crater.

Line Scan vs Area Scan

A sputter crater formed on an iron sample, recorded with a scanning laser profilometer. The peaks on the edge of the crater are an artifact caused by flaring of the laser at the sharp edges of the crater.

To determine the average depth of the crater it is first necessary to level the slope of the sample and then to identify the areas inside (blue) and outside (yellow) the crater. The difference in height of these two regions is then given here as 16.8 mm.

 

A single laser scan across the crater is shown below, the rough bottom is caused by the differential sputtering of different grains in the iron sample:

Again it is a matter of identifying the regions inside and outside the crater:

The difference in height of these two regions from the single trace is 16.9 mm.

Here the crater depths estimated by scanning the whole crater or by a single scan across the crater give almost the same answer. This is usually the case when the crater bottom is nearly flat, as here. But will not generally be the case if the crater bottom is not flat.

TOP

Pure Materials
Element	CQrel Calc.	CQrel Meas.	qrel Meas.	Ref.
Al	0.36	0.40	0.34	(3),(4)
Ag	4.79		9.3
Au	7.57		11.1
Co	1.00		2.4	(4)
Cr	0.89	0.76	0.77	(3)
Cu	1.57	3.2	3.54	(5)
Fe	1.00	1.00	1.00
Mo	1.12		1.40
Nb	0.94		0.71	(6)
Ni	0.97		1.52, 1.49±0.02	(4),(7)
Pb	21.8		17.1
Pd	3.07		0.64	(6)
Si	0.24		0.17	(4)
Sn	5.05	5.2	6.54	(3),(4)
Ta	2.26		3.43
Ti	0.60		0.43
V	0.59		0.50
W	2.05		3.33
Zn	7.06	7.1	8.60	(3),(4)
Zr	1.17		1.03

Sputtering Rate Constant

The sputtering rate constant depends principally on two factors: 

• how efficient is the transfer of energy from the incident Argon ion to the target atoms (this will vary with their relative masses)
• how difficult is it to break atomic bonds in the target to free sputtered atoms (given by the sublimation energy)

The sputtering rate constant C_Q of a pure material, expressed as a ratio to some reference material, eg pure iron, is given approximately by⁽¹⁾

where m, m_ref, and m_Ar are the atomic masses of the material, the reference material, and Argon (the sputtering gas), and U_S the sublimation energy. Hence the relative sputtering rate constant tends to increase slowly with the mass of the material and to decrease with increasing sublimation energy.

This theoretical equation involves many assumptions and works well for higher atomic mass targets (above Ni) but does not work particularly well for low atomic mass targets. So I used the following empirical version of this equation, which works better at lower masses:

where in a first study, a ~ 2.4, b ~ 1.8, and c ~ 1.5.

Relative Sputtering Rates

When several elements are present in the target, the surface will become preferentially enriched with the slower sputtering elements, and these will dominate the sputtering rate constant.

The relative sputtering rate constant for a material composed of different elements is therefore given approximately by⁽¹⁾

where c_i is the mass fraction of element i and C_Qreli is the value of C_Qrel for the pure material.

Experiment

Strictly speaking, to determine C_Qrel it is necessary first to measure U₀. Sometimes this is done, but mostly it is assumed U₀ is constant and relative sputtering rates are used instead. Values for U_S are from ref. (2).

Additional sputtering rate data are available at the website of TAZ GmbH, by clicking here. Note: they express relative sputtering rates as the inverse.

References:
 (1) R Payling, in R Payling, D G Jones and A Bengtson (Eds), Glow Discharge Optical Emission Spectrometry, John Wiley, Chichester (1997), pp 260, 267.
 (2) E A Brandes and G B Brook (Eds), Smithells Metals Reference Book, Butterworth-Heinemann (1992), pp 8-1 to 8-3.
 (3) A Bengtson and L Danielsson, Thin Solid Films, 124 (1985) 231.
 (4) T Nelis, Private communication (1999).
 (5) Z Weiss, Lecture at GD-OES Seminar, Vito, Mol, Belgium, (1999).
 (6) TAZ GmbH, Website www.tazgmbh.de (2000).
 (7) M Köster, Private communication (2001).
 (8) L Ohannessian, PhD Thesis, Université Claude Bernard, Lyon, France, pp 88 (1986).

First published on the web: 1 June 2000.

Author: Richard Payling

TOP

Improved Relative Sputtering Rate

As described in Pt 1, the relative sputtering rate for a material composed of different elements is given approximately by⁽¹⁾

where c_i is the mass fraction of element i and q_i is relative sputtering rate of the pure material.

However, when values calculated with this equation are compared with experiment, the calculated values are often more extreme (ie too dominated by the slower sputtering species) than measured values.

The equation assumes that each species behaves independently, and the discrepancy with experiment suggests that this is not entirely valid. A simple way to include some form of general interaction is to use an equation of the form

where d  is a fitted parameter. When d  = 0, the original equation is obtained. The following graph shows how the relative sputtering rate might vary in a binary alloy of elements a and b for different values of d, from -0.3 to 0.6, as a function of the content of a, where the relative sputtering rate of a is 7.06 (Zn) and b is 0.36 (Al).

The following graph shows experimental results^(2,3) for Zn-Fe with d = 0 and 0.6.

Reference:
 (1) R Payling, in R Payling, D G Jones and A Bengtson (Eds), Glow Discharge Optical Emission Spectrometry, John Wiley, Chichester (1997), pp 260, 267.
 (2) S Miyake, M Koda and T Yoshii, Nippon Steel Eng. Rept. 65, 51 (1992).
 (3) Y Matsumoto, N Fujino and S Tsuchiya, Transactions ISIJ 27, 891 (1987).

First published on the web: 19 June 2000.

Author: Richard Payling

TOP