High power impulse magnetron sputtering : Current-voltage-time characteristics indicate the onset of sustained self-sputtering

TL;DR: In this article, it was shown that the HIPIMS discharge typically exhibits an initial pressure dependent current peak followed by a second phase that is power and material dependent, which suggests that the initial phase is dominated by gas ions, whereas the later phase has a strong contribution from self-sputtering.

Abstract: The commonly used current-voltage characteristics are found inadequate for describing the pulsed nature of the high power impulse magnetron sputtering (HIPIMS) discharge; rather, the description needs to be expanded to current-voltage-time characteristics for each initial gas pressure. Using different target materials (Cu, Ti, Nb, C, W, Al, and Cr) and a pulsed constant-voltage supply, it is shown that the HIPIMS discharges typically exhibit an initial pressure dependent current peak followed by a second phase that is power and material dependent. This suggests that the initial phase of a HIPIMS discharge pulse is dominated by gas ions, whereas the later phase has a strong contribution from self-sputtering. For some materials, the discharge switches into a mode of sustained self-sputtering. The very large differences between materials cannot be ascribed to the different sputter yields but they indicate that generation and trapping of secondary electrons play a major role for current-voltage-time characteristics. In particular, it is argued that the sustained self-sputtering phase is associated with the generation of multiply charged ions because only they can cause potential emission of secondary electrons, whereas the yield caused by singly charged metal ions is negligibly small.

Summary

Introduction

• The commonly used current-voltage characteristics are found inadequate for describing the pulsed nature of the high power impulse magnetron sputtering discharge, rather, the description needs to be expanded to current-voltage-time characteristics for each initial gas pressure.
• The authors adopt the term HIPIMS, as opposed to high power pulse magnetron sputtering , because the latter is inconsistently used for either HIPIMS-like systems or for medium frequency pulsed sputtering with very large area targets.
• Argon, or a similar gas, is only needed to get the process started and may well be shut off afterwards.
• The condition for sustained self-sputtering reads 1SSαβγ ≥ (1) where α is the ionization probability, β is the probability that a sputtered and ionized atom will return to the target, and SSγ is the self-sputter yield.
• The authors will show that the current-voltage characteristic cannot be reduced to a single curve representing current-voltage pairs for given conditions (pressure, geometry, etc.) but rather one needs to map to the currentvoltage-time space for each of those conditions.

II. EXPERIMENTAL

• The experiments were carried out using a 2-inch (5 cm) planar, balanced magnetron.
• The authors intentionally used such a small magnetron because it allowed us to achieve very high power density.
• The short pulse limit is given by the pulser to 5 μs, and the long pulse limit by the capacitively stored energy, which practically means several milliseconds (one would see a large voltage droop, especially at high current).
• The total pressure was monitored by an MKS Baratron® gauge.
• The total ion current was recorded using an ion collector of about 100 cm2 area placed at 20 cm distance from the target; the collector was biased to -50 V with respect to ground.

D. Carbon (Graphite)

• The other extreme, compared to niobium, was the behavior of graphite (Fig.9).
• Here, the current curves are essentially characterized by the initial peak each exhibits.
• At later times and higher voltages, one can see a slight increase in current but it remains at a relatively low level.
• Unfortunately, it was not possible to utilize the full voltage capability of the power supply because the discharge tended to arc when the voltage was set to 800 V or higher.

F. Aluminum

• Aluminum generally shows a similar behavior, with the transition to a high level of current starting at about 550 V (Fig. 11).
• The current at later times exceed the initial peak for voltages greater than about 700 Volt.
• As with copper, the aluminum discharge pulses are characterized by a new equilibrium, as indicated by the constant current later in the pulse.

G. Chromium

• The chromium current was low, reaching maximum values of only 7 A at 1000 V, the maximum voltage of the power supply.
• This was surprising given the relatively high self-sputter yield of chromium.
• The curves showed a couple of interesting features (Fig. 12).
• This strange feature in the set of current curves was repeatedly reproduced several times by increasing and decreasing the voltage level.
• Measurements of the ion flux using the HIDEN EQP spectrometer indicated the presence of singly and doubly charged chromium and argon ions (Fig. 7, bottom).

A. Some basic physics of the HIPIMS discharge

• In the following discussion, the authors start with describing some general processes of the HIPIMS discharge, followed by a more specific discussion of the results described in section III.
• The sheath is dynamic such that it is greatly dependent on the sheath voltage and plasma density at the sheath edge, and to a lesser degree on the electron temperature.
• While self-sputtering is associated with a feedback mechanism that leads to amplification of sputtering and ionization, there is also an increase in “losses” of sputtered atoms from the target zone.
• One can also see that the ion current at 20 cm distance tends to increase throughout the pulse even as the discharge parameters have found new steady-state values.
• The authors do not have a convincing explanation for the minimum of the initial (argon ion) peak observed at a voltage of 550 V, and for the differences to other metals like tungsten, for example, given that work function, ionization energies, and sputter yields are not very different.

ACKNOWLEDGMENTS

• The authors thank Günter Mark, David Horwat and Joe Wallig for technical help, and Thomas Schenkel for helpful discussions.
• A.E. and J.A. acknowledge support from EPSRC grant EP/D049202/1 and the Wenner-Gren Foundations, respectively.
• This work was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Building Technology, of the U.S. Department of Energy, under Contract No. DE-AC02-05CH11231.

Full text

High power impulse magnetron sputtering: Current-voltage-time characteristics
indicate the onset of sustained self-sputtering
André Anders
a)
, Joakim Andersson
Lawrence Berkeley National Laboratory, University of California, 1 Cyclotron Road,
Berkeley, California 94720
Arutiun Ehiasarian
Materials and Engineering Research Institute, Sheffield Hallam University, Howard St.,
Sheffield S1 1WB, UK
Abstract
The commonly used current-voltage characteristics are found inadequate for
describing the pulsed nature of the high power impulse magnetron sputtering (HIPIMS)
discharge, rather, the description needs to be expanded to current-voltage-time
characteristics for each initial gas pressure. Using different target materials (Cu, Ti, Nb,
C, W, Al, Cr) and a pulsed constant-voltage supply it is shown that the HIPIMS
discharges typically exhibit an initial pressure dependent current peak followed by a
second phase that is power and material dependent. This suggests that the initial phase of
a HIPIMS discharge pulse is dominated by gas ions whereas the later phase has a strong
contribution from self-sputtering. For some materials the discharge switches into a mode
of sustained self-sputtering. The very large differences between materials cannot be
ascribed to the different sputter yields but they indicate that generation and trapping of
secondary electrons plays a major role for current-voltage-time characteristics. In
particular, it is argued that the sustained self-sputtering phase is associated with the
1

generation of multiply charged ions because only they can cause potential emission of
secondary electrons whereas the yield caused by singly charged metal ions is negligibly
small.
a)
Corresponding author; Electronic mail aanders@lbl.gov
2

I. INTRODUCTION
High Power Impulse Magnetron Sputtering (HIPIMS) is an emerging physical
vapor deposition (PVD) technology. It is characterized by a high pulse power density at
the sputtering target, which is typically two orders of magnitude greater than the average
power density. We adopt the term HIPIMS, as opposed to high power pulse magnetron
sputtering (HPPMS), because the latter is inconsistently used for either HIPIMS-like
systems or for medium frequency pulsed sputtering with very large area targets. In the
latter case, even as the power is high, the pulse power density does not much exceed the
average power density.
With greatly enhanced power density, ionization of sputtered atoms can occur to a
much larger extent. The presence of ionized sputtered material is of great importance to
some PVD applications such as the filling of trenches and vias of semiconductor
microprocessors. Additionally, HIPIMS may be used for applications that are
traditionally served by cathodic arc plasma processing such as substrate etching
1
and film
deposition.
2
Ionization of the sputtered material enables the processes of self-ion-assisted
deposition and energetic condensation
3-5
.
The roots of ionized sputtering may be traced back to research in Japan some 30
years ago when Hosokawa and coworkers noticed a discrepancy between the measured
and calculated copper and aluminum deposition rates.
6
They estimated that, in the case
of aluminum, about 18% of the ion current to the target are due to ionized sputtered
atoms, which cause self-sputtering. A few years later, they published a condition for
sustained self-sputtering, i.e., a condition for sputtering that exclusively relies on self-
sputtering. Argon, or a similar gas, is only needed to get the process started and may well
3

be shut off afterwards.
7
The sustained sputter process can operate with ionized metal
only when the power density is very high
8
(typically several 100 W/cm
2
). The condition
for sustained self-sputtering reads
1
SS
α
βγ
≥ (1)
where
α
is the ionization probability,
β
is the probability that a sputtered and ionized
atom will return to the target, and
SS
γ
is the self-sputter yield. Sinc 1e
α
< an 1d
β
<
,
the conditi 1on
SS
γ
> is necessary but not sufficient for sustained self-sputtering.
Sustained self-sputtering has been demonstrated by several researchers but only
for a very limited number of materials. For example, Posadowski and Radsimski
9
showed that the principle works for copper and silver, which are metals of very high self-
sputter yield. The experiments showed that sustaining the self-sputtering without
processing gas required operation with a high current density on the target. This
suggested to go to even higher current density (or, equivalently, power density), which is
only possible by using pulses in order to not exceed the average power rating of the
magnetron.
At a 1996 symposium in Berkeley, Sergey Bugaev and coworkers
10
reported
about pulsing a filament-assisted, hollow cathode magnetron to high power, with a pulse
voltage up to 800 V and a peak current of 450 A, leading to a deposition rate of 11
μm/min for copper. In 1999, Kouznetsov and coworkers
11
published their much-cited
work in which they explicitly outline the possibility to operate a
planar magnetron at
very high power density, with peak power soon to approach the 1 MW level, leading to
deposition of the target material from the plasma phase, as opposed to from the neutral
vapor phase of sputtered atoms. Although the number of papers in the field of HIPIMS
4

has grown since then, many aspects of the complicated physics have not been fully
understood.
In the following we report on experiments using a constant voltage HIPIMS
power supply and focus on measuring and interpreting the current-voltage characteristics
for a number of different target materials. We will show that the current-voltage
characteristic cannot be reduced to a single curve representing current-voltage pairs for
given conditions (pressure, geometry, etc.) but rather one needs to map to the current-
voltage-time space for each of those conditions. This can be accomplished by providing
a set of time-dependent current curves, each taken at a fixed voltage.
II. EXPERIMENTAL
The experiments were carried out using a 2-inch (5 cm) planar, balanced
magnetron. The targets were ¼ inch (6.25 mm) thick; the magnetic field induction at the
center of the target surface was 64 mT and lower elsewhere; the diameter of the circular
racetrack was 25 mm. We intentionally used such a small magnetron because it allowed
us to achieve very high power density. The peak power density was up to 5 kW/cm
2
averaged over the whole target area, and even exceeded 10 kW/cm
2
when considering the
more physically relevant, effective racetrack area of about 10 cm
2
or less.
The power was supplied by a slightly modified SPIK2000A pulse power supply
(Melec GmbH) operating in the unipolar negative mode at constant voltage. A great
feature of this pulser is the ability to freely select the pulse length. The short pulse limit
is given by the pulser to 5 μs, and the long pulse limit by the capacitively stored energy,
which practically means several milliseconds (one would see a large voltage droop,
5

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "High power impulse magnetron sputtering: current-voltage-time characteristics indicate the onset of sustained self-sputtering" ?

Using different target materials ( Cu, Ti, Nb, C, W, Al, Cr ) and a pulsed constant-voltage supply it is shown that the HIPIMS discharges typically exhibit an initial pressure dependent current peak followed by a second phase that is power and material dependent. This suggests that the initial phase of a HIPIMS discharge pulse is dominated by gas ions whereas the later phase has a strong contribution from self-sputtering. 

Q2. What is the effect of a low-power HIPIMS pulse on the shea?

A low-power “keeping” discharge or operation at high duty cycles ensures thatthe applied HIPIMS pulse can immediately lead to a strong rise in discharge current because there are enough ions near the target available to be accelerated. 

Q3. What is the main reason for the self-sputtering phase?

The yield of secondary electrons depends strongly on the potential energy of the primary ions, and therefore the authors have reason to believe that the generation of multiply charged metal ions is critical for the onset and maintenance of the self-sputter-dominated phase. 

Q4. What is the effect of the sheath on the electron flux?

The sheath is dynamic such that it is greatly dependenton the sheath voltage and plasma density at the sheath edge, and to a lesser degree on the electron temperature. 

Q5. Why is the peak of argon much wider than with other targets?

Due to the lower ionization, the atom’s low mass and the low flux of sputtered atoms, the rarefaction of argon is much slower: the initial peak is much wider than with other target materials. 

Q6. What is the main process of ion impacting the target surface?

Ions impacting the target surface cause two main secondary processes: (i)emission of secondary electrons and (ii) sputtering of atoms. 

Q7. What is the effect of the magnetic field on the sheath?

SEs emitted far from the racetrack center will experience a significantly tiltedmagnetic field which allows them to readily leave the sheath and to become energetic electrons (up to the full sheath voltage). 

Q8. Why was copper selected as a material of primary interest?

Copper was selected as a material of primary interest because of its relevance forsemiconductor metallization and because it has one of the highest sputter yields of all metals (only exceeded by silver). 

Q9. What is the effect of the negative voltage pulse on the sheath?

As a negative voltage pulse is applied to the target, the development of the sheathdepends strongly whether or not plasma is present. 

Q10. What is the mean free path of atoms at a typical pressure?

At a typical pressure of, say, 1 Pa (7.5 mTorr), the mean free path of atoms islarger than the sheath thickness, and only a small fraction agf s λ≈ of sputtered atomswill be slowed by a collision in the sheath. 

Q11. How does the transition to the self-sputter-dominated mode work?

The transition to the self-sputter-dominated mode is observed at about 700 V of applied voltage, although the absolute current value does not reach or exceed the initial (argon ion) peak value. 

Q12. Why are the authors interested in pulses longer than 100 s?

The authors are especially interested in pulses longer than 100 μsbecause this allows the discharge to evolve into the metal discharge phase, as the authors will discuss. 

Q13. What is the current in the discharge pulses of aluminum?

As with copper, the aluminum discharge pulses are characterized by a new equilibrium, as indicated by the constant current later in the pulse. 

Q14. What is the transitional phase of the current curves?

The transitional phase, where the high current level starts to appear, is around 700 V, and the curves at about 720 V vary from pulse to pulse. 

Q15. How much voltage did the authors use to reach the metal-dominated phase?

Base on experiments with chromium using another, higher voltage system at the Sheffield Hallam University, the authors believe the authors would have reached the metal-dominated-phase in this experiment if the voltage was greater than 1 kV. 

Q16. What was the power supplied by the pulser?

The power was supplied by a slightly modified SPIK2000A pulse power supply(Melec GmbH) operating in the unipolar negative mode at constant voltage. 

Q17. What is the description of the self-sputter phase?

Other materials show the self-sputter-dominated phase in a more-or-less pronounced manner, the details of which call for time-dependent modeling of the discharge taking into account at least two spatial dimensions. 

Q18. What is the ionization energy to obtain doubly charged ions?

Another possible factor is that the ionization energy to obtain doubly charged ionsis relatively low, hence the concentration of doubly charged ions is expected to be high. 

Q19. What is the role of the ion in the self-sputter-dominated phase?

the authors suggest that multiply charged metal ions play a critical role for the onset and maintenance of the self-sputter-dominated phase. 

Q20. Why was it not possible to use the full voltage capability of the power supply?

it was not possible to utilize the full voltage capability of the power supply because the discharge tended to arc when the voltage was set to 800 V or higher. 

Q21. What is the reason why carbon did not go into a second phase?

Carbon did not go into a second high-current, selfsputter-dominated phase, which is not a surprise given the low sputter yield and high ionization energy.