DRIE of Silicon for MEMS Inkjet Heads
Janet Hopkins, Surface Technology Systems Ltd., Newport, UK -- Semiconductor International, 11/1/2004
|
Within the world of MEMS manufacturing, inkjet printer heads represent the largest and a substantial size of the total MEMS component market.1 The fabrication of inkjet heads is an area that to date has been dominated by established low-cost manufacturing methods; bead blasting and wet etching have been the predominant production methods for fabricating the holes in silicon-based inkjet heads. However, much like any other microelectronics manufacturing situation, with increased competition in the printer marketplace, manufacturers of MEMS inkjet heads are forced to address the sometimes-conflicting challenges of reducing production costs while improving product performance. In the case of inkjet heads, print resolution is one of the primary measures of product performance.
When printing with inkjet heads, smaller, more tightly spaced droplets of ink result in sharper print quality. However, this produces a smaller print area, resulting in an increased printing time. To optimize both print quality and speed, a printer head must deliver an increased number of smaller-sized droplets; this equates directly to an increase in the density of holes per inch in an inkjet head. The path to doing this lies with the adoption of deep reactive ion etching (DRIE) to replace blasting and wet etching.
Etching inkjet head MEMSBoth the current, predominant techniques used to create holes in an inkjet head (i.e., sand blasting and wet etching) are low-cost and high-throughput, which are highly desirable attributes. The main disadvantages of these techniques, however, are the limitations of accurately creating an increased density of holes. Wet etching yields a low-precision, isotropic etch, while sand blasting creates dust and has poor critical dimension (CD) control caused by the inaccuracy of the technique. Both techniques reduce production yield by contamination or mechanical damage.
Not unlike its application in IC fabrication, plasma dry etching for creating holes in inkjet heads has the advantage of high precision, and it is an inherently cleaner process than sand blasting. To date, however, the main issue with dry etching has been its lower etch rate, which is more problematic in the fabrication of inkjet heads than ICs. For dry etching to compete in the fabrication of the former, the adopted etch process must match the etch rates and throughputs that inkjet head manufacturers have become accustomed to.
DRIEDRIE of silicon for inkjet printer heads is best achieved using a sequence of short alternating passivation (C4F8 plasma) and etching (SF6 plasma) steps (i.e., the so-called Bosch Process2). Optimization of process parameters for the etch and passivation steps allows control of the etch profile, CD and etch rate. To maximize etch rate, the fluorine radicals reaching the silicon must be maximized while controlling the number of ions at the substrate. Typically, fluorine-radical density can be augmented by increasing coil power, gas flow and pressure using an SF6 plasma process (Fig. 1 ).
![]() |
| 1. The effect of gas flow (left) and coil power (right) on etch rate, using a non-switched SF6 plasma process (150 mm wafers with >95% silicon exposed). |
Figure 2 illustrates the increase in fluorine-ion density with coil power. Although an increase in ion density is preferable for increased etch rates, excess ions can cause damage to the etched silicon feature CD by removing the deposition protection too quickly, resulting in an imbalanced process. In addition, high ion bombardment is not necessary for DRIE processing of silicon, as the silicon is etched spontaneously when in contact with fluorine radicals. A small degree of ion bombardment at the wafer level is required for directional removal of the passivation from the base of the feature being etched, but beyond a critical level, this can cause detrimental effects such as sidewall breakdown and high mask etch rates.
To fully exploit plasma processing, specifically for etching inkjet heads, the chemically reactive species (i.e., fluorine radicals) and the ions reaching the wafer's surface must be controlled via the specific plasma etch system. For example, a decoupled plasma source that allows high-density plasma to be formed enables a high fluorine radical concentration remote from the substrate. Fluorine radicals have a long lifetime compared with ions, so this inherently reduces the number of ions at the wafer level. With use of an ion filter, an additional parameter in the process recipe, the number of ions reaching the wafer can be controlled further. Then, with parameter ramping — a method of changing process parameters with time throughout an active process — the profile or etch rate can be adjusted further.3
![]() |
| 3. Numerous parameters are ramped during the etch process. |
Figure 3 shows an example where several parameters, including gas flow and bias power, have been ramped during etch processing, resulting in a change in etch rate as shown in the change in gradient over time; at the start of the process, the gradient is less than towards the end of the process, indicating an increase in etch rate with the positively ramped process parameters.
Through detailed knowledge of what is happening in the plasma, high etch rates can be obtained without detriment to other parameters, such as profile control and selectivity. It is also advantageous to have the ability to set an individual pressure for both the etch and passivation steps. This increases the flexibility of the process and can aid etch rate.
Figure 4 shows 80 µm trenches etched >200 µm deep, with an etch rate of >22 µm/min. The uniformity across the wafers was less than ±2.3%. The roughness and profile of the etch channel in inkjet heads influences the way in which ink passes through a channel; this is an additional concern when optimizing an etch process for this application. However, the results in Figure 4 show an etched structure that meets even future requirements for high-performance inkjet head structures. This is proof that a controlled DRIE plasma etch process is capable of creating high-density precision channels that can improve the fabrication and resulting quality of inkjet heads.
| Author Information |
| Janet Hopkins studied at the University of Durham, gaining a B.S. in physics and chemistry of materials, followed by a Ph.D. in plasma chemistry. After completing her Ph.D., she joined Surface Technology Systems Ltd. as a process engineer in the silicon etch group. She later became a senior process engineer, ASE development team leader, and is now R&D process development team leader. She has authored several papers at the University of Durham and STS. |
| References |
|



