In a HiPIMS discharge, the electron density is still significant long into the afterglow, for example ne ≈ 3×1014 m−3 at t = 3 ms for pO2/ptotal = 0.1. Hence, negative ions
generated in the plasma volume are generally confined by the potentials contained in the sheaths at the chamber walls during this time and therefore ambipolar diffusion of O− to the walls can also be neglected for this case. Only when the plasma degrades into an ion-ion state does the electric field change significantly, allowing negative ions to be transported to the walls [180]. Neglecting wall losses and using the reaction sets listed in table 6.2 and 6.3, balance equations governing the rate of change of OM
2 and O
− in the
discharge can be written as
dnMO2 dt =kexnenO2 −nOM 2 k M atne+kdissne+k1Ddissne+kcr1nO2 +kcr2nO2+kArnAr (6.7)
6. NEGATIVE ION DENSITIES IN REACTIVE HIPIMS dnO− dt =k M atnenOM 2 −nO− kedne+k O+2 mn1nO+2 +k O+2 mn2nO+2 +k O+ mnnO+ +kadnO+kAr + mn nAr+ (6.8)
Given that ne as measured during the active phase was found to be almost iden-
tical for the three oxygen partial pressures, it is suggested that a greater number of OM
2 are generated for increasing oxygen content in the discharge via electron-impact
excitation (kexnenO2). During the afterglow, further generation of O
M
2 is unlikely due
to the sharp drop in Te and the associated drop in kex. The rate coefficients of the
dominant loss terms in the afterglow are kcr1, kcr2 and kdiss1D, so a characteristic decay
time of metastable oxygen can be estimated by averaging over the afterglow and from τOM
2 ≈ kcr1nO2 +kcr2nO2 +k
1D dissne
−1
∼ 1600µs, 525 µs and 315 µs for increasing oxy- gen partial pressure. The τOM
2 values are much shorter than the calculated diffusive loss
times and exhibit shorter times for increasing oxygen content, consistent with the results presented by You et al. [137]. Using the calculated decay times and the rate balance equation above, the estimated densities of oxygen metastables arenOM
2 = 1.3×10
18 m−3,
1.5×1019 m−3 and 1.8×1019 m−3 for p
O2/ptotal = 0.1, 0.3 and 0.5, respectively, which
are approximately an order of magnitude higher than ne during the on-time.
As argued above, the higher the O2 content of the discharge, the greater the amount
of OM
2 available at the beginning of the afterglow and therefore one would expect a higher
negative ion density n− and a greater rate of increase in electronegativity as electrons
undergo dissociative attachment to OM2 . Indeed, from figures 6.8, this is what is observed as the electronegativity increases for greater oxygen partial pressure. Furthermore, an associated and more pronounced decay inneis also observed for increasing oxygen content.
In particular, the second decay phase of ne, τn2, is found to be significantly shorter;
τn2 = 326µs, 293µs and 92µs across the range ofpO2/ptotal for increasing oxygen content
and substantially shorter than for the Ar-only discharge where τn2 = 550µs. This could
be evidence of increased negative ion formation for higher oxygen availability as electrons are lost in dissociative attachment to metastable oxygen. From figure 6.9, it is observed that the average value of negative ion density during the afterglow is somewhat higher for pO2/ptotal = 0.3 and 0.5 (hn−i ≈7×10
15 m−3) when compared with p
O2/ptotal = 0.1
(hn−i ≈ 3×1015 m−3). This agrees with a global model developed by Gudmundsson
and Thorsteinsson [176] which demonstrated that for increasing Ar content in an Ar/O2
discharge one can expect a decrease in n−.
The oxygen negative ion density is observed to decrease in the afterglow with char- acteristic times of τ− = 1.4 ms, 2.1 ms and 2.5 ms for pO2/ptotal = 0.1, 0.3 and 0.5,
respectively. Considering the dominant loss terms (mutual neutralisation) for O− ions and averaged plasma density values during the afterglow, effective characteristic decay
6. NEGATIVE ION DENSITIES IN REACTIVE HIPIMS
times can be estimated by
τ− =− katM ne nO− nOM 2 − kO + 2 mn1nO+2 +k O+ mnnO+ +kAr + mn nAr+ −1 (6.9)
with the positive ion species fractions estimated from energy-resolved mass spectrometry measurements as discussed above. Using equation 6.9, the expected characteristic decay times of negative ions are found to be 750 µs, 3.7 ms and 3.1 ms for increasing pO2/ptotal.
This is in approximate agreement with the experimentally determined values, to within a factor of 2, and also agrees insofar as that for low oxygen content (i.e. pO2/ptotal = 0.1) the
decay times are shorter than for higher oxygen content (i.e. pO2/ptotal = 0.3 and 0.5). For
increasing oxygen partial pressure, it is expected that larger amount of OM
2 are generated
in the active phase which results in an increased rate of negative ion formation (kM
atnenMO2)
in the afterglow. However, it is speculated that increasing oxygen beyond a critical point will lead to a decreased formation rate of negative ions despite the increase in nM
O2, as
ne is observed to decrease for reduced Ar partial pressure in an Ar/O2 discharge [181].
It is possible that this critical point exists for pO2/pAr > 1, as the characteristic decay
time of negative ions is observed to be a monotonically increasing function of oxygen partial pressure in the results presented here. Furthermore, the results presented here show that, for reactive HiPIMS, the decay of negative ion species in the afterglow is approximately an order of magnitude longer than those measured in reactive pulsed DC magnetron sputtering [137]. This could be a consequence of a longer electron density decay time in HiPIMS when compared with pulsed DC magnetron sputtering, which would facilitate negative ion formation via dissociative electron attachment for longer periods in the afterglow.
Although negative ions are assumed to be confined to the bulk plasma by the ambipolar electric field during the on-time and for afterglow investigated here, the long off-times in HiPIMS will generally result in the eventual collapse of this potential and a subsequent flux of negative ions to the chamber walls as well as any grounded substrates or processing surfaces. This can be a very important consideration for many plasma processing methods which rely upon the flux of reactive species to surfaces, such as reactive ion etching. Assuming the negative ions follow the Boltzmann relation, akin to electrons, the negative ion flux density to a surface is given by
Γ− =n− kBT− 2πM− 1/2 exp − eφ kBT− . (6.10)
For even small sheath potential values, φ, the negative ion flux to the walls is strongly reduced, however, when the plasma enters an ion-ion state and the potential at the cham- ber walls collapses (i.e. φ →0), negative ions are free to diffuse to the walls. Therefore, by assuming that the oxygen negative ions possess a temperature close to that of the
6. NEGATIVE ION DENSITIES IN REACTIVE HIPIMS
background gas (i.e. T− ≈Tgas) and takingn− ≈1015 m−3, the flux of O− negative ions
to the chamber walls, in the absence of a potential, can be approximated using equation
6.10 and is found to be of the order of 1017 −1018 m−2 s−1. This is a non-negligible
flux (for comparison, this is about 1% of the positive ion flux during the on-time is Γi =nics ∼1019−1020 m−2 s−1) and may influence the properties of deposited or etched
materials in plasma processing. Indeed, negative chlorine ions (Cl−) generated in the off-time of an inductively couple plasma source have been associated with the suppres- sion of ‘notch’ formation during the etching of poly-Si [173]. Further investigation into the temporal evolution of metastable oxygen molecules during reactive HiPIMS would greatly facilitate in the validation (or otherwise) of the simple chemical model employed here and help to better understand the oxygen negative ion dynamics during the discharge afterglow.
Figure 6.10. The HiPIMS discharge current,Id(t), and target potential, Vd(t), wave-
forms for different oxygen partial pressures;pO2/ptotal = 0.1, 0.3 and 0.5.
The pulse width was 100 µs with a repetition rate of 100 Hz and and average discharge power of 200 W.
6. NEGATIVE ION DENSITIES IN REACTIVE HIPIMS
6.3.3
Laser-aided photodetachment
As outlined above, laser photodetachment measurements were made using a different experimental set-up than the one used for the Langmuir probe investigation. The HiPIMS discharge current and applied target potential waveforms are presented in figure 6.10 for the different oxygen partial pressures: pO2/ptotal = 0.1, 0.3 and 0.5. The time-averaged
discharge power was maintained at 200 W. The value of the peak power density varied between 0.21 and 0.24 kW cm−2, corresponding to peak current density values of 0.46−
0.56 A cm−2 across the range of oxygen partial pressures.
The delay time of the discharge current rise is slightly longer forpO2/ptotal = 0.1 than
for pO2/ptotal = 0.3 and 0.5 with values of ∼ 13 µs and ∼ 8 µs, respectively. As the
delay in the current on-set is longer for lower oxygen partial pressures, the peak current is necessarily higher for a constant average discharge power. As the applied potential does not vary significantly between the conditions, the disparity in the current delay times may be due to reduced poisoning of the target for lower oxygen partial pressures, which would alter the secondary electron emission coefficient.