Etch stop

Electrochemical etching is one way of controlling the etch rate and determine a clear etch stop layer when bulk micromachining Silicon. In this case, the wafer is used as anode in an HF-Electrolyte. Sufficiently high currents lead to oxidation of the silicon. The resulting oxide which is dissolved by the HF-solution. Since lowly doped silicon material is not exhibiting a notable etch rate, it can be used as an etch stop. 

Silicon-based pressure sensors are amongst the most common devices making use of this process. A thin low-n-doped epitaxial layer on the wafer determines an etch stop depth and thus the thickness of e.g. the pressure sensor membrane. 

Aqueous electrolytes of high pH etch silicon even at open circuit potential (OCP) conditions. The etch rate can be enhanced or decreased by application of anodic or cathodic potentials respectively, as discussed in Section 4.5. The use of electrolytes of high pH in electrochemical applications is limited and mainly in the field of etch-stop techniques. At low pH silicon is quite inert because under anodic potentials a thin passivating oxide film is formed. This oxide film can only be dissolved if HF is present. The dissolution rate of bulk Si in HF at OCP, however, is negligible and an anodic bias is required for dissolution. These special properties of HF account for its prominent position among all electrolytes for silicon. Because most of the electrochemistry reported in the following chapters refers to HF electrolytes, they will be discussed in detail. 

As shown in Fig. 3.3, the I-V curve in this regime shows a cathodic potential shift and a slight change of slope, if the doping density is increased. Compared to p-type substrates the I-V curve of p+ is shifted cathodically by about 0.1 V and that of n+ by about 0.2 V [Gal, Zh5[. This shift can be exploited for etch stops and selective formation of PS, as discussed in Section 4.5. 

If the relevant literature is surveyed for the keywords etch stops and silicon, a confusing multiplicity of methods is found using different electrolytes, different bias and differently doped silicon substrates. This section does not aim to be a comprehensive review of all these techniques [Co2], but an introduction to the basic principles of electrochemical etch stops, which will be illustrated by a few typical examples. 

An etch stop is a method that allows for selective removal of a specific material. Selectivity, therefore, is the most important property of an etch stop. It is defined as the etch rate of the faster etching material divided by the etch rate of the slower etching material. 

All electrochemical etch stops of silicon are based on the dissolution behavior discussed above and a method to have different parts of the electrode interface under different potentials. 

The easiest way to have different parts of the electrode surface under different bias is to disconnect them by an insulator. This method is elucidated by an experiment in which an electrochemical etch-stop technique has been used to localize defects in an array of trench capacitors. In a perfect capacitor the polysilicon in the trench is insulated from the substrate whereas it is connected in a defect capacitor, as shown in Fig. 4.15 a. If an anodic bias is applied the bulk silicon and the polysilicon in the defect trench will be etched, while the other trenches are not etched if an aqueous HF electrolyte is used, as shown in Fig. 4.15b. The reverse is true for a KOH electrolyte, because the only polysilicon electrode in the defect trench is passivated by an anodic oxide, as shown in Fig. 4.15 c. 

Fig. 4.15 The use of etch stops to visualize defects in a dynamic random access memory (DRAM) structure for subsequent electron microscopy inspection. If a bias is applied to the silicon substrate, as shown in (a), the polysili-...

Small leakage currents or a transistor-like action of the junction are sufficient to generate a small current that may cause undesired passivation. This can be circumvented by application of an additional potential to the etching layer, shown by the broken line in Fig. 4.16 a. This electrochemical etch-stop technique is favorable compared to the conventional chemical p+ etch stop in alkaline solutions, because it does not require high doping densities. This etch stop has mainly been apphed for manufacturing thin silicon membranes [Ge5, Pa7, Kll] used for example in pressure sensors [Hil]. 

For doping-dependent anodic etch stops in HF, a general hierarchy of dissolution is observed [La5] illuminated n-doped and n+-doped areas are most easily dissolved, followed by p+-doped areas. Next likely to be dissolved are p-type areas. Moderately n-type doped areas kept in the dark are least likely to be etched. This hierarchy corresponds to the potential shift of the I-V curve in the regime of PS formation [Gal, Zh5]. 

But even in a homogeneously doped material an etch stop layer can be generated by an inhomogeneous charge carrier distribution. If a positive bias is applied to the metal electrode of an MOS structure, an inversion layer is formed in the p-type semiconductor. The inversion layer passivates in alkaline solutions if it is kept at the PP using a second bias [Sm5], as shown in Fig. 4.16b. This method is used to reduce the thickness variations of SOI wafers [Og2]. Illuminated regions... 

A prerequisite for all etch-stop techniques discussed so far is an electrical connection to an external power supply. However, if the potential required for passivation in alkaline solutions is below 1 V, it can be generated by an internal galvanic cell, for example by a gold-silicon element [As4, Xil]. An internal galvanic cell can also be realized by a p-n junction illuminated in the etchant, as discussed in the next section. Internal cells eliminate the need for external contacts and make this technique suitable for simple batch fabrication. 

An initially flat silicon electrode surface will develop a surface topography if the photocurrent varies locally. This variation can be caused by a lateral variation of the recombination rate or by a lateral variation of the illumination intensity. The photoelectrochemical etching of a silicon electrode is related to the etch-stop techniques discussed in Chapter 3. While different etching rates for different areas of the electrode may be obtained by electrical insulation or by a different doping density, the etch rate may also be altered by a difference in illumination intensity. Basically four photoelectrochemical etching modes are possible for homogeneously doped substrates ... 

Sensors for measurements of physical parameters such as pressure, rotation or acceleration are commonly based on elongation or vibration of membranes, cantilevers or other proof masses. The electrochemical processes used to achieve these micromechanical structures are commonly etch-stop techniques, as discussed in Section 4.5, or sacrificial layer techniques, discussed in Section 10.7. 

In the manufacture of micromechanical devices electrochemistry is commonly used to realize etch stop structures or to form porous layers. The first of these is discussed in Section 4.5. In the latter case, the use of PS as a preserved layer or as a sacrificial layer can be distinguished. In the first case PS is an integral part of the ready device, as discussed in Sections 10.4 to 10.6, while in the latter case the PS serves as a sacrificial layer and is removed during the manufacturing process. 

The membranes of the microhotplates were released by anisotropic, wet-chemical etching in KOH. In order to fabricate defined Si-islands that serve as heat spreaders of the microhotplate, an electrochemical etch stop (ECE) technique using a 4-electrode configuration was applied [109]. ECE on fully processed CMOS wafers requires, that aU reticles on the wafers are electrically interconnected to provide distributed biasing to the n-well regions and the substrate from two contact pads [1 lOj. The formation of the contact pads and the reticle interconnection requires a special photolithographic process flow in the CMOS process, but no additional non-standard processes. 

The electrochemical etch-stop technology that produces the silicon island is rather complex, so that an etch stop directly on the dielectric layer would simplify the sensor fabrication (Sect. 4.1.2). The second device as presented in Fig. 4.6 was derived from the circular microhotplate design and features the same layout parameters of heaters and electrodes. It does, however, not feature any sihcon island. Due to the missing heat spreader, significant temperature gradients across the heated area are to be expected. Therefore, an array of temperature sensors was integrated on the hotplate to assess the temperature distribution. The temperature sensors (nominal resistance of 1 kfl) were placed in characteristic locations on the microhotplate, which were numbered Ti to T4. 

Another possibihty to improve the temperature homogeneity is to introduce an additional polysiHcon plate in the membrane center. The thermal conductivity of polysilicon is lower than that of crystalline siHcon but much higher than the thermal conductivity of the dielectric layers, so that the heat conduction across the heated area is increased. Such an additional plate constitutes a heat spreader that can be realized without the use of an electrochemical etch stop technique. Although this device was not fabricated, simulations were performed in order to quantify the possible improvement of the temperature homogeneity. The simulation results of such a microhotplate are plotted in Fig. 4.9. The abbreviations Si to S4 denote the simulated temperatures at the characteristic locations of the temperature sensors. At the location T2, the simulated relative temperature difference is 5%, which corresponds to a temperature gradient of 0.15 °C/pm at 300 °C. 

B. Kloeck, S.D. CoUins, N.F. de Rooij, and R.L. Smith. Study of electrochemical etch-stop for high-precision thickness control of silicon membranes , IEEE Transactions on Electron Devices 36 (1989), 663-669. 

The steam reformer is a serpentine channel with a channel width of 1000 fim and depth of 230 fim (Figure 15). Four reformers were fabricated per single 100 mm silicon wafer polished on both sides. In the procedure employed to fabricate the reactors, plasma enhanced chemical vapor deposition (PECVD) was used to deposit silicon nitride, an etch stop for a silicon wet etch later in the process, on both sides of the wafer. Next, the desired pattern was transferred to the back of the wafer using photolithography, and the silicon nitride was plasma etched. Potassium hydroxide was then used to etch the exposed silicon to the desired depth. Copper, approximately 33 nm thick, which was used as the reforming catalyst, was then deposited by sputter deposition. The reactor inlet was made by etching a 1 mm hole into the end... 

For systems in which is not zero, the shape of the etched profile can be controlled to a certain extent by adjusting the stoichiometry of the discharge. This is illustrated in Fig. 3-7 using an idealized Si sample. As hydrogen is added to a CF. discharge the etch rate of Si will decrease as we have seen earlier in Fig. 3.2. However, the etch rate will stop on surfaces not subjected to ion bombardment (point A in Fig. 3.7) before etching stops on surfaces which are exposed to energetic ion bombardment. This means that the lateral etch rate has been eliminated and features with vertical sidewalls can be etched if an etch gas mixture of CF. — 10% is used in this example. 

---