**1. Introduction**

Plasma electrolytic oxidation (PEO), often also known as micro-arc oxidation (MAO), is a novel environmentally friendly surface modification technique used for the production of ceramic coatings on a variety of light metals, such as Al, Mg, Ti and their alloys [1,2]. PEO is a complex and highly non-linear process due to the electrical, thermal, and plasma-chemical reactions in the electrolyte. Over the last decade, some prominent researches have been reported in literature related to the formation mechanism of PEO coatings. Various approaches have been reported in literature to identify the mechanism for the formation of the coating due to PEO, such as thickness measurement [3], element tracer [4,5], two-step oxidation method [6], and optical emission spectroscopy (OES) [7]. Nevertheless, it is still difficult to explain such a complex physical and chemical process during plasma discharge and significant work still needs to be undertaken to explain the formation mechanism of PEO coatings.

During PEO, dielectric breakdown of the growing oxide coatings at high voltage electrolysis results in a large number of short-lived micro-discharges [8]. Those discharge events mean not a single discharge but a cascade of smaller discharges that are bundled together in space and time. These distinct discharge events play an important role in the formation mechanism of the coating and strongly affect the microstructure and properties of the coating [9]. Due to the extreme non-linearity associated

with the plasma discharge, monitoring the evolution of an individual discharge event would aid in understanding the coating mechanism and allow optimization of the production process.

Recently, many methods, such as optical [10–12], spectral [12–16], electrical [16,17], frequency response [18,19], and acoustic [20], have been used to measure dynamic parameters of plasma discharge during PEO. Reported characteristic parameters, including duration, current level, apparent radii, event rate, temperature, and electron density, about PEO discharge events are summarized [10–17,21] as follows: Typical discharges are now known to occur in prolonged sequences ('cascades') at particular locations, and to have lifetimes of the order of a few tens to a few hundreds of microseconds, with 'incubation' periods between them of around a few hundred μs to a ms or two. Discharge currents are typically several tens of mA, discharge energies a few mJ and diameters of core discharge channels a few tens of μm. The frequency of plasma discharges per unit time and area was determined throughout processing, falling from initial values in the range 300–350 mm<sup>−</sup><sup>2</sup> s<sup>−</sup><sup>1</sup> to fewer than 50 mm<sup>−</sup><sup>2</sup> s<sup>−</sup><sup>1</sup> after 1000 s of processing. Plasma temperatures have been estimated via optical spectroscopy to range from about 4000 to 12,000 K, with some indications of a higher temperature core and a lower temperature surrounding region. Corresponding electron densities typically range from ~10<sup>15</sup> to 10<sup>18</sup> cm<sup>−</sup>3.

It can be seen that these monitoring techniques are e ffective in distinguishing certain basic characteristics of PEO discharge events. However, values of these characteristic parameters are spread across a wide range, sometimes di ffering by several orders of magnitude. For example, the lifetimes of discharges revealed by fast video imaging were in the range of 0.05–4 ms during AC PEO of magnesium alloys [10] and evaluations based on the spectroscopic method for the Al 1100 alloy show electron temperatures to be in the range of 3500–9000 K for a unipolar current mode and in the range of 3500–6000 K for a bipolar current mode [14]. This is probably the result of a low-resolution tool or inappropriate measurement procedure and data processing algorithm.

Special experimental methods, such as small area in-situ testing [16] and single pulse anodizing [22–24], have been designed to study discharge events and dielectric breakdown during PEO. The present work builds upon single pulse anodizing of aluminum micro-electrodes. Here, a partial aim is to produce a clear set of conclusions related to the discharge behaviors. In addition, an improved understanding of the e ffect of plasma discharge on subsequent oxide film formation is sought. Electrical characteristic monitoring and scanning electron microscopy (SEM) were used to study the e ffect of single pulse anodizing on aluminum micro-electrodes along with the e ffects of pulse parameters on surface morphology. Parameters such as current level, duration, diameter, and spatial distribution ratio were estimated from the current waveforms.

#### **2. Materials and Methods**

#### *2.1. Materials and Pre-Treatment*

The experimental platform used to study PEO is shown in Figure 1. A home-made 20 kW pulse power supply was used, which could provide a DC voltage ranging from 0 to 1000 V and pulses with the same magnitude and frequencies up to 20 kHz [25]. In this study, small area sample monitoring of currents in individual discharges was used to study the nature of these discharges and their distributions in time and position. In order to provide a su fficiently small area, micro-electrodes made of aluminum wires were chose as specimens. Pure aluminum wires (99.99%) with 0.6 mm diameter were anodized in electrolyte, which was composed of commercial distilled water, 5 g/<sup>L</sup> potassium hydroxide (KOH) and 10 g/<sup>L</sup> sodium silicate (Na2SiO3). Experiments were carried out at 25 ◦C in a 2-liter water-cooled stainless steel tank, which also served as the counter-electrode.

**Figure 1.** Schematic of the experimental platform for plasma electrolytic oxidation (PEO).

Aluminum wires were first anodized up to DC 550 V at a rate of 8 V/s, to cover the entire metal surface with a thick anodic oxide film. Next, these wires were embedded in an epoxy resin. To expose a clean and relatively flat surface, the embedded wires were sectioned with a slicing knife. All specimens were rinsed under running water to keep the tip surface clean before the next process.

#### *2.2. Single Pulse Anodizing*

In our experiments, every single pulse anodizing process was carried out with an individual specimen and the interval time between adjacent experiments was set for 5 min. All single pulse anodizing processes were carried out in two steps. Initially, pre-deposited oxide films on the tip surface were obtained by anodizing with a DC voltage sweep up to 300 V at a rate of 8 V/s. Then, a single pulse voltage with amplitude in the range of 325–525 V and a pulse width of 100–5000 μs was applied to cause dielectric breakdown.

In addition, a longer voltage pulse with a magnitude at 500 V and pulse duration of 50 ms was applied on these pre-deposited oxide films to estimate electrical characteristic parameters, such as current level, duration, diameter, and spatial distribution ratio.

#### *2.3. Data Monitoring and Post-Processing*

For pre-treatment and single pulse anodizing process, voltage and current data were recorded at a sampling rate of 2 MHz using a data acquisition card (National Instruments, PCI-6133, Austin, TX, USA) controlled by the LabVIEW software installed in PC (Thinkpad T580). Current signal was sampled by a hall current sensor (CSM002A) and voltage signal between anode and cathode electrodes was detected by probe of oscilloscope (DPO 3014) directly. Meanwhile, real-time waveforms of current and voltage were displayed on the screen of an oscilloscope. Surfaces of the specimens were observed using a LEO1530 VP scanning electron microscope.

Raw voltage and current data had considerable electrical noise due to mechanical vibration, electron avalanche, dielectric breakdown and high frequency electronic switching. As a result, it was necessary to carry out post-processing for data. The data were first filtered with a low pass filter with a cut-off frequency of 50 kHz. This allowed high frequency noise due to the electronic switching to be eliminated. Next, this data was smoothened using moving average filter. Finally, an appropriate scale function and wavelet basis function were adopted to extract random noise signal.
