**Investigation of the Pd Nanoparticles-Assisted Chemical Etching of Silicon for Ethanol Solution Electrooxidation**

#### **Olga Volovlikova 1,\*, Gennady Silakov 1, Sergey Gavrilov 1, Tomasz Maniecki <sup>2</sup> and Alexander Dudin <sup>3</sup>**


Received: 7 November 2019; Accepted: 10 December 2019; Published: 12 December 2019 -

**Abstract:** The formation of porous silicon by Pd nanoparticles-assisted chemical etching of single-crystal Si with resistivity ρ = 0.01 Ω·cm at 25 ◦C, 50 ◦C and 75 ◦C in HF/H2O2/H2O solution was studied. Porous layers of silicon were studied by optical and scanning electron microscopy, and gravimetric analysis. It is shown that por-Si, formed by Pd nanoparticles-assisted chemical etching, has the property of ethanol electrooxidation. The chromatographic analysis of ethanol electrooxidation products on por-Si/Pd shows that the main products are CO2, CH4, H2, CO, O2, acetaldehyde (CHO)<sup>+</sup>, methanol and water vapor. The mass activity of the por-Si/Pd system was investigated by measuring the short-circuit current in ethanol solutions. The influence of the thickness of porous silicon and wafer on the mass activity and the charge measured during ethanol electrooxidation was established. Additionally, the mechanism of charge transport during ethanol electrooxidation was established.

**Keywords:** porous silicon; Pd nanoparticles-assisted chemical etching; etching rate; ethanol electrooxidation

#### **1. Introduction**

Technological innovation leads to an increase in energy and natural resources consumption, in particular, natural gas. Continuous consumption of non-renewable resources leads to their exhaustion. There is a need to develop and make use of alternative energy sources, based on environmentally-friendly technologies, because of natural resources limitations. Development of new alternative energy sources will allow us to receive scientific and technical results, with technologies which provide transition to resource-saving energy. One of the prospective directions in the resource-saving power area is the fuel element and portable electrochemical energy generators. They have many advantages: portability, high efficiency, small level of harmful emissions and quietness [1]. Nowadays, an interesting and prospective direction in the resource-saving power area is ethanol as fuel and a power source for the electric current generators [2].

Transformation to electric power is due to direct ethanol oxidation in the cell. It allows the simplification of the fuel supply system because of high specific energy of liquid alcohols, providing a short circuit in an environmentally-friendly cycle of transformation of energy in the natural scale due to a number of alcohols. The ethanol can be produced in biosystems in almost unlimited volumes [3]. The oxidation product of ethanol is CH4.

Recent studies for alternative fuels indicate a growing interest in the development of small fuel cells and energy generators based on porous silicon due to a few advantages: high specific surface; strong chemical loading ability of the surface; the possibility of changing the surface morphology of the porous layers at the nano- and micro levels; simplicity and low cost of manufacture and compatibility with silicon integrated technology [4]. The plasma chemical etching for porous silicon formation is widely used. This method is characterised by a high complexity of hardware provided [5]. Chemical etching in solutions of alkalis or acids is a cheaper method for porous Si formation. Porous Si is usually formed by anode etching in HF solutions [6–8]. However, this method means individual treatment of wafers.

In recent years, special development was received by the chemical etching induced by noble metals (Ag, Au, Pt, and Pd) [9–12]. This method is simple and enables carrying out group treatment of wafers that reduces the price of the technology of porous silicon formation. Besides, this method allows the production of porous silicon with a wide range of geometrical sizes by using a form and type of a metal mask. The thickness of the porous silicon is defined by the etching duration, electrolyte composition, and the metal amount. The MACE (metal-assisted chemical etching) method enables the production of a noble metal/porous silicon structure that combines the functions of both the anode and cathode of the generator's active element and fuel cell. This structure is called the Schottky junction (Schottky barrier). In work by Bin Zhu [13,14] a hydrogen fuel cell based on the Schottky barrier (metal, *p*-type semiconductor) was described.

Porous silicon functionalised with noble metals is of undoubted scientific and practical interest as an object for the production of new energy generators. Simplifying the design by switching to a single-layer functional structure will increase the productivity and reduce the cost of the finished device. Conducting studies of the electrocatalytic ethanol oxidation using the cathode/anode structure, based on Pd clusters in a porous layer, allows producing an environmentally-friendly and resource-saving energy sector. The efficiency of creating a structure is associated with a high specific surface area of the porous silicon, and size effects of a metal catalyst during the ethanol electrooxidation. It occurs on the local nano- and micro-regions of the anodes/cathodes inside of porous silicon. Understanding the electro-catalytic activities of the ethanol on the Pd/por-Si structure is very important for developing more active catalysts for the direct-ethanol generators. The purpose of the work is establishment of the influence of the porous layer thickness and porosity formed by Pd nanoparticles-assisted etching on the duration of gas evolution, and ethanol electrooxidation mass activity for different ethanol-based electrolytes.

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

Boron doped *p*-type silicon wafers with (1 0 0) orientation and resistivity of 0.01 Ω·cm were cleaned as follows: (i) dipping into H2SO4 (98%):H2O2 (30%) (in volume 1:1) solution, then (ii) into HF (40%):H2O (in volume 1:4) solution to remove native oxide and finally, (iii) into pure ethanol. The cleaned wafers were cut into pieces 3 <sup>×</sup> 3 cm2. The samples were placed in a Teflon cell. The samples were immersed into solution PdCl2:HCl (0.5 g/L PdCl2, 20 mL/L HCl) for 30 min at 25 ◦C for Pd film deposition. The por-Si formation was performed in solution HF (40%):H2O2 (30%):H2O (25:10:4 in volume) at *T* = 25, 50 and 75 ◦C. Porosity was calculated by gravimetric analysis. Samples with area *S* were weighed before etching (*m*1). Then non-polished side of the sample was covered by varnish and dried in the air. The etching duration was 30–120 min. Then, samples were cleaned in ethanol and dried at 65 ◦C. The varnish was removed and the samples were weighed (*m*2). Por-Si was dissolved in a water solution of NaOH and weighed (*m*3). Porosity was calculated by the equation:

$$P = \frac{m\_1 - m\_2}{m\_1 - m\_3} \times 100\% \text{.} \tag{1}$$

where *m*<sup>1</sup> and *m*<sup>2</sup> are the sample's masses before and after etching and *m*<sup>3</sup> is the sample mass after por-Si dissolution. Samples were etched in the same conditions because of the multi-sectional Teflon cell.

The duration of filling of the same gas volume (2.4 mL) during ethanol electrooxidation (EEO), with different C2H5OH concentrations, was fixed. The gas volume, as a result of the ethanol electrooxidation by 2.5 cm2 samples, was measured using a cell (Figure 1). The essence of the method is the displacement of water by the gas, under a glass cell.

**Figure 1.** Schematic presentation of cell for gas volume measurement: (**1**) sample, (**2**) ethanol solution, (**3**) gas, (**4**) flexible tube, (**5**) cell, (**6**) glass cell, (**7**) holes.

The sample (1) was placed into a glass cell with ethanol solution (2). Gas evolution (3) occurs. Then, gas follows on a flexible tube (4) in cell (5) located in glass cell (6) with water (level 0). Then the gas displaces water from cell through the holes (7) and the level changed. The displaced water volume equals to gas volume evolved during a certain time *V*water = *V*gas.

The cell was filled multiple times to calculate the gas evolution rate υ. It was calculated by the equation:

$$
\nu = \frac{V\_{\text{gas}}}{t} \tag{2}
$$

where *V*gas is the gas volume and *t* is the duration of the gas evolution.

Sample surface morphology was investigated by optical and scanning electron microscopy (Carl Zeiss Axiovert 40 MAT (Carl Zeiss Group, Oberkochen, Germany) and Helios NanoLab 650 (Thermo Fisher Scientific, Hillsboro, OR, USA)). Energy-dispersive X-ray (EDX) spectroscopy was performed on FEI Helios (FEI, Hillsboro, OR, USA) with an EDAX Octane Elite super EDS System (Octane Super, Mahwah, NJ, USA). The short-circuit current during etching in the galvanic cells and the mass activity as a function of EEO time were measured with a digital multimeter (UNI-T UT61C, UNI-T Group Ltd., Hong Kong). The value of the charge *Q*Excess Carrier was determined by numerical integration of current versus time, *Q* = *t* <sup>0</sup> *J*d*t*. The composition and conductivity of ethanol solutions for mass activity measurements are presented in Table 1. pH equals to 2 made by adding H2SO4 into ethanol solution.

**Table 1.** Solution properties.


The gas composition was measured using quadrupole mass selective detector Hiden Analytical HPR-20 (Hiden Analytical, Warrington, UK) in the *m*/*z* range 1–100. EEO was performed in glass reactor at *T* = 25 ◦C, *p* = 1 bar. To obtain information on types of chemical bonds presented in the porous layer, the samples were analyzed by infrared (IR) reflectance spectroscopy using Nicolet iS50 spectrometer (Thermo Fisher Scientific).

#### **3. Results**

Figure 2 shows scanning electron micrograph (SEM) images of Si (1 0 0) surface with palladium clusters deposited during immersion in PdCl2 solution for 30 min. The dimension of separate Pd particles varied in the range of 20 < *d* < 50 nm. The agglomerates from the Pd particles varied in the range of 0.1 < *d* < 2 μm.

**Figure 2.** Scanning electron micrograph (SEM) image of Pd particles deposited on Si (1 0 0) wafer by immersing in PdCl2:HCl solution for *t* = 30 min.

Figure 3 shows SEM images of the porous silicon surface after Pd nanoparticles-assisted chemical etching with different etching times.

**Figure 3.** Scanning electron micrograph of the cross section porous silicon film etched for (**a**) 2 min, (**b**) 15 min and (**c**) 120 min at 25 ◦C.

(**c**)

Figure 3 illustrates porous silicon formation during different etching times. The minimum etching duration was 2 min because of 1.25 μm porous layer is observed. The maximum etching duration was 120 min. The porous layer breaks after 120 min etching. The porous silicon has perpendicular macropores with diameters from 1 to 3 μm and mesopores tightly penetrating the walls of the porous matrix. Pd particles dissolve silicon, gradually plunging inwards and forming a pore. Parts of the particles were deposited on the pore walls during the etching process. This contributed to the formation of mesopores.

Each pore is characterised by the size and shape of the metal particle. The appearance of the inner pore cavity is a consequence of the etching. If the particle is agglomerated, the pore walls are like a sponge of mesoporous silicon. As the agglomerate moves deeper into the pore, individual particles with a diameter of 80 ± 5 nm are deposited on the pore wall, contributing to its dissolution. This phenomenon is described in [12]. As particles are deposited on the walls, the agglomerate decreases in size, forming a dimple of a smaller diameter, which ultimately leads to the formation of a conical pore. In this case, the conical pore is due to the gradual dissolution of the walls (Figure 4a). Etching segments can determine the particle size. The upper part of the pore promotes its expansion with the solution penetrating the portion, uniformly dissolving the walls of the pore by photoelectrochemical dissolution. In the case of the separate metal particles, the pores will be formed vertically (Figure 4b). The pore diameter will be equal to the particle diameter.

(**a**) (**b**) **Figure 4.** SEM image of cross section of sample after etching during 60 min: (**a**) the Pd particles are agglomerated, (**b**) the separate Pd particles.

It was found that Pd particles are present in the mesopores, which have access to a system por-Si/Pd. This system provides the ethanol electrooxidation (EEO). It can proceed according to one of a few schemes (Figure 5) [15]. The acetaldehyde, acetic acid (incomplete oxidation) and carbon dioxide (complete oxidation) are the products of electrooxidation of ethanol [16].

**Figure 5.** The scheme of the ethanol electrooxidation.

Figure 6a illustrates EEO. Intense gas evolution is observed from the functionalised metal-based silicon porous material immersed in ethanol solutions. Intense gas evolution gradually decreased. The gas evolution duration can be as long as several hours. The gas composition is shown in Figure 6b.

**Figure 6.** Electrooxidation (EEO): (**a**) sample photo during EEO, (**b**) the mass spectrum of a gas.

The mass spectrometry analysis of EEO products on por-Si/Pd shows that the main products are CO2, CH4, H2, CO, O2, acetaldehyde (CHO)<sup>+</sup> [17], methanol, ethanol and water vapor. The volumes of components relative to the total volume of EEO products are the following: H2—5%, CH4—7%, H2O—33%, CO—25%, O2—20%, CO2—10%. Figure 7a,b shows the influence of the sample porosity, etching duration and temperature on the duration of gas evolution. In this work, intensive gas evolution can be visually detected without using additional devices.

**Figure 7.** The dependence of the gas evolution on the (**a**) samples treatment duration, (**b**) por-Si porosity. Points are experimental results, lines are approximations.

It has been found that the maximum duration of intense gas evolution (32 min) is observed for samples formed at 50 ◦C for 120 min etching, and the minimum duration (2 min) is observed for samples formed at 75 ◦C for 60–120 min etching. Increase and reduction of the duration of gas evolution is caused by the increase and reduction of porous silicon thickness respectively. It was established that a linear increase in the thickness of the porous layer happened with an increase in the etching duration from 30 to 120 min for temperatures of 25 and 50 ◦C. The thickness of the porous layer was from 30 to 90 μm for 30–120 min and 25 ◦C etching; from 60 to 105 μm for 30–120 min and 50 ◦C etching; and from 100 to 45 μm for 30–120 min and 75 ◦C etching. It is due to porous layer dissolution. Reducing the porous silicon thickness leads to a decrease of the local metal/semiconductor and the EEO regions, respectively. The transfer of porous silicon at elevated processing temperatures is described in detail in [18]. The duration of the gas evolution is linearly dependent on the porosity of the layer. Therefore, the high porosity of the sample ensures access of the reactants to the surface of por-Si/Pd and removal of the reaction products.

Besides the duration of gas evolution, an important parameter for establishing the EEO mechanism is the gas evolution rate. It characterises the EEO reaction rate. Table 2 shows the results of the analysis of the rate of gas evolution.


**Table 2.** The rate of gas evolution for three ethanol solutions.

It has been found that the rate of gas evolution and the EEO is higher for solution 95/5. The high concentration of ethanol molecules promotes rapid adsorption. In addition, the rate of gas evolution is gradually reduced, which may be due to several factors:


The first two factors do not have an effect on the rate reduction. The porous layer destruction (SEM) and solution depletion has not been established. The addition of alcohol to the solution after the gas evolution stopped did not resume the process, while the as-prepared sample oxidized the spent solution. Treatment of the used sample in hydrofluoric acid contributed to the resumption of intense gas evolution in the spent solution.

Figure 8 shows SEM images of Pd/por-Si surface after EEO.

**Figure 8.** SEM images of porous silicon after EEO of solutions contained (**a**) 95/5, (**b**) 60/40 and (**c**) 30/70 ethanol/water at 25 ◦C during 30 min.

In Figure 8, it can be seen that the gradual reduction of the gas evolution rate was due to contamination of the surface with reaction products. The higher the concentration of ethanol in the solution, the denser the precipitates. The thickness of the precipitate layer covering the porous layer has a value of several micrometres. The element analysis (EDX method) of the porous surface after EEO (Figure 9) allows us to determine a non-uniform distribution of elements into the surface. The elements in porous silicon are Si, O, and C (Table 3). The element in porous silicon at Spots 1, 3 and 4 is silicon.

**Figure 9.** SEM image of porous silicon after electrooxidation and energy-dispersive X-ray (EDX) analysis spots.



Chemical bonds between the components of the porous layer were analyzed by infrared (IR) reflectance spectroscopy (Figure 10).

IR reflection spectra show the presence of bands typical for EEO by Pt and Pd catalysts. Table 4 presents wavenumbers corresponding to the bonds.

The CO3 <sup>2</sup><sup>−</sup> may be observed near 2846 cm<sup>−</sup>1. The acetate was displayed as two intense peaks at 1553 and 1410 cm−1. As the concentration of ethanol in solution increases, the acetate of CH3COO<sup>−</sup> band (1550 cm−1) and <sup>ν</sup> (C−H) of CH3CH2OH band (2900 cm−1) intensities also increase. As the concentration of ethanol in solution increases, the Si-H wag band intensities decrease. Decrease in intensity may be due to the increasing of the thickness of the precipitate layer covering the porous layer.

**Figure 10.** FTIR (Fourier-transform infrared) spectra of porous silicon after 120 min of electrooxidation of ethanol with different concentration.


**Table 4.** Surface bonding of porous silicon after electrooxidation (EEO).

#### **4. Discussion**

The studies of EEO products on platinum catalysts using various analytical methods show that the reaction predominantly involves ethanol oxidation to CO2 [22]:

Anodic reaction on Pd:

$$\text{C}\_2\text{H}\_5\text{OH} \xrightarrow{2\text{e}} \text{CH}\_3\text{OH} \xrightarrow{2\text{e}} \text{CH}\_X + \text{CO} \xrightarrow{8\text{e}} 2\text{CO}\_2\tag{3}$$

Cathodic reaction on Si:

$$\text{12H}^+ + \text{3O}\_2 + \text{12e}^- \rightarrow \text{6H}\_2\text{O} \tag{4}$$

O2 and H2 is a result of water splitting by porous silicon because of water solutions of ethanol [23]. Figure 11 shows the mass activity as a function of time on the Pd/porous silicon with a different thickness of por-Si for the solution No. 1–3. The mass activity characterises the amount of ethanol that was oxidized by the sample over a period of time *t* and the behavior of the process. The measurement was performed in a two-electrode cell, Pd/Si- anode, Pt- cathode. When the Si/Pd-system is dipped into the ethanol solution, EEO is occurred at the Pd/Si surface. The electrons' transport is going through the porous layer and silicon wafer. The electrons diffuse into the semiconductor and accumulated at the wafer's rear (unload) side. The current can be registered in the galvanic cell (Figure 10). The current flows through the electrolyte between the Pt-cathode and Si/Pd-anode. The Pt-electrode is arranged in the electrolyte in immediate proximity to the sample surface.

**Figure 11.** Current–time curves measured on Pd/por-Si for solutions: (**a**) No. 1 (10/90), (**b**) No. 2 (50/50) and (**c**) No. 3 (95/5).

The current density decreased with time for all porous samples and solutions. All catalysts showed the maximum current densities (*J*max) immediately after the step (I section) (Figure 12). Then, the current decreased with time (II section). After a few minutes, the current achieved a pseudo-steady state (III section). The curve type corresponds to current–time curves measured during electrooxidation of dimethyl ether on Pt/C and PtMe/C catalysts in sulphuric acid [24].

**Figure 12.** Typical *J*(*t*) sections.

Decrease of the current density is due to the formation of the contamination on the porous silicon surface. Table 5 shows the current–time curves analysis for different solutions.


**Table 5.** The current–time curves analysis for different solutions.

It was established that *J*max depends on the thickness of the porous silicon and concentration of ethanol. Such dependence is due to the EEO reactions yield. In this case, the thickness of the porous layer affects the amount of Pd particles in the porous layer, due to the sample preparation. The minimum value of *J*max and *J*steady state is observed for the case with a minimum porous layer thickness of 8 μm. The maximum values of *J*max equal to 77 and 79.6 μA/cm2, and *J*steady state equal to 7 and 5.5 μA/cm2, are observed for 38–40 μm thick porous layers. Ethanol concentration in solution does not affect *J*steady state, but affects *J*max. *J*max reaches a value between 17 to 23 μA/cm<sup>2</sup> for solution 10/90. *J*max reaches a value between 33 to 79.6 μA/cm<sup>2</sup> for solution 95/5. This may be due to the non-wettability of the surface of porous silicon formed Pd nanoparticles-assisted etching [25] by solution 95/5 and 50/50 (the contact angles are 140◦). The contact angle for solutions 10/90 is 140◦. The higher the ethanol concentration, the less the contact angle on porous surface [26]. Solution 10/90 showed the best activity for the electrooxidation in this study because of the high value of *J*steady state. Low surface contamination during electrooxidation (Figure 8c) facilitates ethanol access to the Pd surface, intensive mass transfer, and high *J*steady state value.

The corresponding charges from Figure 11 can be extracted and plotted versus a time scale to indicate the rate of formation of adsorbed species at this preparation potential of Pd/por-Si. The value of *Q*Excess Carrier was determined by numerical integration of the dependence of the current on time and presented in Figure 13. The *Q*Excess Carrier value is characterising excess charge carriers diffused into the substrate during EEO.

**Figure 13.** *Cont.*

**Figure 13.** The *Q*Excess Carrier versus duration of ethanol oxidation of porous silicon with different thickness and solution: (**a**) 10/90, (**b**) 50/50, (**c**) 95/5.

The charge of time is described by a polynomial of degree 2 with *R* = 0.99%. The charge passing through the substrate *Q*Excess Carrier depends on a few factors: concentration of charge carriers injected into silicon *Q*total, substrate thickness and specific resistivity (*Q*sub), and porous layer thickness (*Q*por-Si).

$$Q\_{\text{total}} = Q\_{\text{Excess Corrier}} + Q\_{\text{sub}} + Q\_{\text{por}} \tag{5}$$

*Q*total characterises the value of all charge carriers involved in the electrooxidation of ethanol. It the case of the present research work, *Q*total depends on Pd/por-Si contact area. The increase in charge carrier concentration on the surface of the hole, caused by injection, leads to the appearance of a diffusion electron flow directed along the x-axis perpendicular to the semiconductor surface, with the result that the carrier concentration increases not only on the surface but also in the depth of the semiconductor. In this case, injected carriers go deeper into the semiconductor at different distances, where they are recombined.

Figure 14 shows current–time curves measured on Pd/por-Si for different solutions at 25 ◦C. A decrease in the wafer thickness by 70 μm, with the same thickness of porous silicon 30 μm, increases the charge from 36 to 140 mC for solution 10/90, from 20 to 100 mC for solution 50/50, and from 14 to 72 mC for solution 95/5 at 3600 s oxidation.

**Figure 14.** Current–time curves measured on Pd/por-Si for solutions: 10/90, 50/50 and 95/5.

Having excluded the contribution of the thickness and resistivity of a single crystal silicon wafer, as well as the thickness of the porous layer because of equal value, we can obtain the equation:

$$Q\_{\text{total }n} - Q\_{\text{Excess Corrier }n} = Q\_{\text{total }m} - Q\_{\text{Excess Corrier }m} \tag{6}$$

where *n* and *m* are the solutions number, and *Q*Excess Carrier is measured by short-circuit current in the galvanic cells. We can calculate *Q*total for unknown solutions for any duration, using the same porous silicon samples and one test solution with *Q*total.

This approach can be used for porous silicon formation by Pd nanoparticles-assisted etching. The process of forming a porous layer is identical, with the only difference being that the thickness of the initial single crystal is 525 μm (Figure 15a) and 336 μm (Figure 15b), respectively.

**Figure 15.** Current–time curves measured during Pd nanoparticles-assisted etching of silicon with thickness: (**a**) 525 μm, (**b**) 336 μm.

The *J*(*t*) curves characterise the etching mechanism. Five characteristic regions can be identified:


*J*(*t*) reflects a change of the area (*S*) of the electrochemical reaction front. The changes of *J* with time (Figure 14) are related to the evolution of the Si morphology during pore nucleation. Pore formation takes place after the immersion of silicon into the solution containing HF and H2O2. The pore area depends on the duration of the treatment [27]. The increase in the surface area leads to the growth of the current density, the first region.

The current growth continues until a porous layer of critical thickness is formed on the Si surface. In this case, access of the solution to the surface of monocrystalline silicon becomes limited. The transport of holes through porous silicon is difficult due to the high specific resistance of the por-Si [27]. With an increase in the thickness of the porous layer, the concentration of excess holes that are not involved in the dissolution of Si becomes smaller, the second region. A further effect of the solution on the surface leads to the dissolution of the porous layer (the beginning of region III). Dissolution of por-Si reduces the thickness of the porous layer to less than the critical value, which increases the concentration of holes in Si, and, consequently, the current in region III. The current is constant in region IV due to the growth and etching of the porous silicon. The porous layer growth and current decrease in the region V.

The charge of the first cycle (single crystal is 525 μm) takes the value of 1.5 C, while the second (single crystal is 336 μm) 17.7 C. Thus, in a single crystal 189 μm thick with a specific resistance of 0.01 Ω·cm, carriers of charge of 16.2 C recombine. Removing silicon can increase the charge passing through the sample by 11.8 times. The currents of the second cycle have values exceeding the currents of the first cycle and allow a detailed study of the structural change during Pd nanoparticles-assisted chemical etching of silicon.

#### **5. Conclusions**

It is shown that por-Si, formed by Pd nanoparticles-assisted chemical etching, has the property of ethanol electrooxidation. Intense gas evolution is observed from the metal/porous silicon immersed in ethanol solutions. The chromatographic analysis of EEO products on por-Si/Pd shows that the main products are CO2, CH4, H2, CO, O2, methanol and water vapor. The duration of the gas evolution is linearly dependent on the porosity of the layer. Therefore, the high porosity of the sample ensures access of the reactants to the surface of por-Si/Pd and removal of the reaction products. The gradual reduction of the gas evolution rate was due to contamination of the surface with reaction products. The mass activity, as a function of time, was measured by the short-circuit current in the galvanic cells.

It was established that *J*max depends on the thickness of the porous silicon and concentration of ethanol. Such dependence is due to the EEO reactions yield. In this case, the thickness of the porous layer affects the amount of Pd particles in the porous layer, due to the sample preparation. The minimum value of *J*max and *J*steady state is observed for the case with a minimum porous layer thickness of 8 μm. The maximum value of *J*max equals to 77 and 79.6 μA/cm2, and for *J*steady state equals to 7 and 5.5 μA/cm2, is observed for 38–40 μm thick porous layers. Ethanol concentration in solution does not affect *J*steady state, but affects *J*max. *J*max reaches a value between 17 to 23 μA/cm<sup>2</sup> for solution 10/90. *J*max reaches a value between 33 to 79.6 μA/cm2 for solution 95/5.

A decrease in the wafer thickness by 70 μm, with the same thickness of porous silicon, increases the charge carriers, diffused into the substrate, from 36 to 140 mC for solution 10/90, from 20 to 100 mC for solution 50/50, and from 14 to 72 mC for solution 95/5 at 3600 s oxidation. Thus, the porous silicon thickness, porosity and solution composition are the main factors defined EEO efficiency.

**Author Contributions:** Conceptualization, O.V.; methodology, O.V. and S.G.; validation, O.V.; formal analysis, O.V.; investigation, G.S., T.M. and A.D.; resources, O.V., T.M. and A.D.; writing—original draft preparation, O.V.; writing—review and editing, S.G. and T.M.; visualization, O.V.; supervision, S.G.; funding acquisition, O.V.

**Funding:** This investigation was supported by the Russian Science Foundation (project No. 19-79-00205).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Microfabrication of X-ray Optics by Metal Assisted Chemical Etching: A Review**

**Lucia Romano 1,2,3,\* and Marco Stampanoni 1,2**


Received: 14 May 2020; Accepted: 10 June 2020; Published: 12 June 2020

**Abstract:** High-aspect-ratio silicon micro- and nanostructures are technologically relevant in several applications, such as microelectronics, microelectromechanical systems, sensors, thermoelectric materials, battery anodes, solar cells, photonic devices, and X-ray optics. Microfabrication is usually achieved by dry-etch with reactive ions and KOH based wet-etch, metal assisted chemical etching (MacEtch) is emerging as a new etching technique that allows huge aspect ratio for feature size in the nanoscale. To date, a specialized review of MacEtch that considers both the fundamentals and X-ray optics applications is missing in the literature. This review aims to provide a comprehensive summary including: (i) fundamental mechanism; (ii) basics and roles to perform uniform etching in direction perpendicular to the <100> Si substrate; (iii) several examples of X-ray optics fabricated by MacEtch such as line gratings, circular gratings array, Fresnel zone plates, and other X-ray lenses; (iv) materials and methods for a full fabrication of absorbing gratings and the application in X-ray grating based interferometry; and (v) future perspectives of X-ray optics fabrication. The review provides researchers and engineers with an extensive and updated understanding of the principles and applications of MacEtch as a new technology for X-ray optics fabrication.

**Keywords:** X-ray grating interferometry; catalyst; silicon; gold electroplating

#### **1. Introduction**

High resolution and high-efficiency diffractive optics have largely been unavailable for hard X-rays where many scientific, technological, and biomedical applications exist. This is due to the long-standing challenge of fabricating high aspect ratio high-resolution micro- and nano-structures.

Fabrication of high-aspect-ratio silicon micro- and nano-structures is a key process in many applications, such as microelectronics [1], microelectromechanical systems [2,3], sensors [4], thermoelectric materials [5], battery anodes [6], solar cells [7], photonic devices [8], and X-ray optics [9]. Microfabrication is usually achieved by reactive ion etching [10], which requires high investment in tools and maintenance. KOH-based wet etching [11,12] has been used for microfabrication in Si at micro- and nano-scale. However, the aspect ratio of etched trenches is limited by the etching rate ratio between different crystallographic orientations and only possible in simple geometries like linear gratings or crossed linear gratings defined by the direction of <111> crystallographic planes of Si. As an alternative approach for fabricating Si microstructures, metal assisted chemical etching [13] has attracted great interest [14] because of its simplicity, low fabrication costs, and ability to generate high aspect ratio nanostructures such as nanowires [15]. Several acronyms were reported for this process—MACE, MAE, MacEtch—since 2015 the community seemed to agree with the common acronym of "MacEtch", which was firstly introduced by X. Li [16] to distinguish the unique properties with respect of standard wet-etch and dry-etch techniques. Unlike KOH wet-etch [12], the MacEtch

process is almost independent of crystal orientation and may be used to create a wide variety of patterns, without suffering of microloading effects of dry-etch. An advantage of the method is the considerable reduction in fabrication costs and complexity with respect to the other techniques. MacEtch fabrication of nanoscale patterns has been successfully applied for synchrotron-based X-ray imaging methods [17,18]. For X-ray grating interferometry imaging, the fabrication of Si microgratings requires sharp vertical profiles, high aspect ratios, high accuracy of pitch size and duty cycle, uniformity over large area, and, finally, the possibility to fill up the Si template with a high X-ray absorbing material [19,20] such as gold [21,22]. These requirements are especially stringent for X-ray medical diagnostics for which extremely large field of view is necessary. Thus gratings require microfabrication on area of many squared centimeters [9], with aspect ratio and pitch size that depend on the used energy, specific design and performances (pitch size in the range of 1–20 μm, aspect ratio in the range of 10–100).

Since its discovery in 2000, by Li et al. [13], MacEtch of silicon has emerged as a new technique capable of fabricating 3D nano- and micro-structures of several shapes and applications [23]:—nano-porous film, nanowires [24], 3D objects [25], trenches, vias [26], micro-fins [27], nano-scale grooves, surface antireflection texturing [28], optoelectronic devices such as solar cells [29] and photodetectors [30], sensor devices [31], X-ray optics—in a few semiconductors substrates: Si [15], Ge [30], poly-Si [32], GaAs [33], β-Ga2O3 [27], SiC [34], etc.—and different catalysts: Ag, Au, Cu, Pt, and Pd [15]. MacEtch has been developed with a strong controlled vertical directionality with respect to the substrate and successfully applied for producing X-ray zone plates [17,23,35–38] and diffraction gratings [19,20,37,39–41]. In MacEtch, a catalyst layer (e.g., Au) is patterned onto the substrate (e.g., Si) to locally increase the dissolution rate of the substrate material in an etchant solution including a fluoride etchant such as hydrofluoric acid (HF) and an oxidizing agent such as hydrogen peroxide (H2O2).

To date, no comprehensive review of patterned microstructures by MacEtch exists in the literature. The existing reviews on MacEtch rarely focus on the aspects of X-ray gratings fabrication. This paper provides an extensive overview of the fundamentals and recent developments of MacEtch as well as addressing the research gaps in this field. After an overview about the MacEtch mechanism, we dedicated a particular attention to the conditions (catalyst, additives, and reaction temperature) to ensure the vertical etching of the (100) Si substrates. Then, we described the procedures for gratings fabrication, from pattern design to transfer in the silicon substrate and finally the template filling with a high X-ray absorbing material such as Au. In the last section we discussed the recent applications of Fresnel zone plates and X-ray interferometric gratings fabricated by MacEtch. In the concluding remarks we outlined the major challenges for large-scale MacEtch X-ray optics and the perspectives of MacEtch microfabrication.

MacEtch offers the possibility to fabricate high aspect ratio structures for hard X-ray diffractive optics and opens up new opportunities for high resolution imaging with compact X-ray sources and for synchrotrons and X-ray-free electron lasers with more complex wave front manipulation.

#### **2. MacEtch Mechanism**

The mechanism of MacEtch has been extensively debated in literature [15] even with controversial interpretations [42]. Etching occurs when the metal (catalyst) patterned Si substrate is immersed in a solution with an etchant (for example HF) and an oxidizer (for example H2O2). The solution-metal-silicon system constitutes a microscopic electrochemical cell that induces anodic silicon etch.

For the most commonly employed oxidants (H2O2), the proposed cathodic reactions provide free positive carriers to be transferred to the silicon, according to Equation (1). In the anode reaction, the silicon consumes the positive carriers and is solubilized through oxidation (Equation (2)). The

concentration of holes becomes higher in the region surrounding the metal catalyst, where silicon is readily oxidized by HF and forms silicon fluoride.

$$\text{H}\_2\text{O}\_2 + 2\text{H}^+ \to 2\text{H}\_2\text{O} + 2\text{h}^+ \tag{1}$$

$$\text{Si} + 4\text{h}^+ + 6\text{HF} \to \text{SiF}\_6^{2-} + 6\text{H}^+ \tag{2}$$

Common oxidizers for MacEtch and their associated cathode reactions are reviewed by Chiappini et al. [43]. A wide variety of metal salts can induce silicon porosification without the addition of any other oxidizer, as their electrochemical potential is sufficiently high to directly inject holes in the valence band of silicon [43]. Several other oxidizing agents have been studied [44], including oxygen [37,45] with the following reaction (Equation (3)):

$$\rm O\_2 + 4H^+ + 4e^- \rightarrow 2H\_2O.\tag{3}$$

Figure 1 shows Scanning Electron Microscopy (SEM) images of MacEtch after few seconds of etching. The removal of Si atoms occurs faster at the interface with the metal catalyst, where positive carriers have the maximum concentration. As the reaction proceeds, the catalyst sinks into the substrate and progressively the catalyst nanopattern is transferred to the substrate. The process continues as long as the etchants are present in the solution and the reaction byproducts diffuse out of the pattern. Despite of the simple mechanism, the full process has indeed a complex dynamic where several phenomena—mass transport of etchants and byproducts, charge carrier diffusion, catalyst stability, and gas release—interplay to determine the etching rate, the etching direction, and the quality of the etched structure.

**Figure 1.** MacEtch mechanism in solution of HF and H2O2. (**a**) The metal catalyst deposited on a Si <100> substrate decomposes H2O2 with consequent injection of holes (+) into the semiconductor. (**b**) Si consumes the positive carriers Si, it is readily oxidized by HF and forms silicon fluoride, the process continues and the catalyst progressively sinks into Si along the <100> direction, transferring the nanostructure pattern to the Si (formation of nanopillars in this case). Images are cross-section Scanning Electron Microscopy (SEM) of nano-patterned Pt on Si (**a**) after few seconds of MacEtch in solution of HF and H2O2.

The etching mechanism and the composition dependence have been extensively reported in literature [15,43,46].

Hydrogen peroxide is by far the most commonly employed oxidizer in MacEtch. Chartier et al. assume that the relative concentration of HF and H2O2 in a MacEtch etch solution plays a similar role to the current density Jps in anodic etch [46]. MacEtch solution is usually described in terms of concentrations ratio between HF and H2O2, according to Chartier's formula (Equation (4))

$$\rho = \frac{[\text{HF}]}{[\text{HF}] + [\text{H}\_2\text{O}\_2]} \tag{4}$$

where [HF] and [H2O2] are the molar concentration of HF and H2O2, respectively and Hildreth's [47] compact expression ρ[HF].

#### **3. Vertical Etching**

Huang et al. [48] demonstrated that MacEtch is intrinsically anisotropic along the preferred crystallographic <100> directions. Such an orientation dependence is related to the silicon lattice configuration at the reaction site. Removal of oxidized silicon by HF is associated with the cleavage of its back bonds, of which effective number density in different crystal planes increases with the order (100) < (110) < (111) [49]. Due to the different back-bond strength, the Si atom on the (100) surface plane is the most easily removed, and the etching occurs preferentially along the <100> directions. The anisotropy could be reduced or eliminated by varying the concentration of the etchants. In MacEtch of Si, the movement of the etching front (i.e., metal/Si interface) is a net consequence of the following two competing events: (1) injection of a positive charge carriers into bulk Si through the metal-Si interface and (2) removal of oxidized Si by HF from just underneath the catalyst metal. Since the generation of holes is related to the catalytic decomposition of H2O2 at the interface between the solution and the catalyst metal surface, the amount of holes injected into Si is proportional to the H2O2 concentration in the solution and the catalyst activity. In conditions of low H2O2 concentration, hole injection into Si atoms will be localized at the (100) plane, where there are the fewest Si back bonds to break, resulting in etching along the <100> direction. As the concentration of H2O2 increases sufficiently, removal of oxidized Si would be kinetically favored in the crystal planes with a higher density of silicon back bonds, resulting in etchings along non-<100> directions. The same argument can be played considering HF, in conditions of low HF concentration, the removal of oxidized silicon would control the reaction so the favorite etching direction is again the <100>. While, for high HF concentration, also the other directions are favored. A schematic is reported in Figure 2. It must be noted that H2O2 and HF are correlated in the solution, so the transition between one etching direction to the other should be determined as a function of the specific solution and the used catalyst. A complex ternary graph would result taking into account the water dilution, as reported by J. Kim et al. [50].

During the process of etching, a triggering event that produces unequal etch rates might occur. These events can change the effective forces on the catalyst and produce a resultant torque on the catalyst. [51]. Such triggering events are made more frequent in the case of higher etch rates as brought about by higher oxidant concentration or by etching at elevated temperature. To date, this represents one of the major challenge to optimize MacEtch as a reliable and controllable process for large area patterning of high aspect ratio structures. Hildreth et al. [25,52] demonstrated that controlled 3D motion of catalyst patterns during MacEtch can be achieved by locally pinning them with an electrically insulating material prior to etching. However, due to this movement, the aspect ratio achievable for features perpendicular to the substrate in an arbitrary dense pattern is limited.

Moreover, charge carriers are injected into Si and charge distribution affects the catalyst movement [53], so that parallel and elongated structures [17] are more difficult to etch than spaced cavities [26]. Several approaches have been attempted in literature to force the uniform etching along the <100> and to minimize the etching along the other directions [51] in order to realize high aspect ratio structures perpendicular to the (100) substrate (vertical etching). Electron-hole concentration balancing structures were used to achieve a vertical etch profile in X-ray zone plates [17]. Figure 3 shows some examples of balancing structures [36] used to define the vertical etching at the borders of the X-ray lens structures [41].

**Figure 3.** Balancing structures to control the vertical etching at the device border: (**a**) cross section SEM of etched linear grating (pitch 250 nm) with and without balancing structures, the figure was adapted with permission from C. Chang et al., 2014. [17]; (**b**) cross-section SEM of etched kinoform lens with outmost zone of pitch 150 nm, the figure was adapted with permission from M. Lebugle et al., 2018. [41]; (**c**) SEM of Au pattern of zone plates with balancing ring, the figure was adapted with permission from K. Li et al., 2017. [36]; and (**d**) detail of border with balancing ring in (**c**), the figure was adapted with permission from K. Li et al., 2017. [36].

Negative carbon mask [54,55], electrical bias [32,55], and magnetic catalyst [56] have been proposed to force the vertical etching and to improve the control of the catalyst movement [47,52,57]. MacEtch resulted to be very efficient for Si nanostructures, nanowires, and ordered nanopillars [15,29,58–61], but etching in the microscale regime is more critical [62], the etching rate is limited by the reactant diffusion through the metal mask. The effective transfer of reactants and their by-products would not be identical where the metal pattern size is nanometers or few micrometers. Therefore, two regimes can be distinguished in literature [2,14,63]: (i) nanoscale patterns, in which the etchant species diffuses through the pattern edges and (ii) microscale patterns with nano-porous films, in which the porosity of the film itself controls the diffusion length. In both regimes, the catalyst geometry significantly affects the etching performance. Catalyst optimization and etching conditions are here reviewed in order to address the vertical etching.

#### *3.1. Catalyst*

A wide range of transition metals can catalyze MacEtch. Noble metals are especially favored for the formation of nanowires as well as for nanostructures with defined cross sections since they better preserve their structure during the etch, as they do not dissolve in HF. Non-noble transition metals have been mostly used to form nano-pores, porous Si, and polished surfaces [43]. The most used MacEthc catalyst is Ag [15]. However, X-ray optics fabrication requires sophisticated patterning techniques such as electron beam lithography or UV photolithography and the catalyst film is usually deposited by thin film evaporation. Silver oxidation is quite difficult to prevent during thin film physical deposition, so Au is the most studied catalyst for thin film deposition. Here, we review the catalyst that have been used for X-ray optics fabrication, which are Au [17,39,40] and Pt [37]. Platinum is the metal with the highest catalytic activity so it allows to obtain the highest MacEtch rate [64]. The patterning of nanostructures requires high precision pattern transfer and high lateral resolution during etching, with MacEtch in liquid this corresponds to a condition of very high HF concentration [17]. Gold catalyst suffers of bad adhesion on silicon substrates, yet a detrimental pattern peel-off has been reported during MacEtch in conditions of high HF concentration [50,65]. On the other hand, uniform high aspect ratio has been reported for nanoporous Au catalyst in conditions of low HF and high H2O2 concentration [39,66]. In these conditions, the etching is more isotropic [63], the top of the trenches appear wider with respect to the bottom compromising the fidelity of the pattern transfer in the lateral dimension, so the process is not suitable for high aspect ratio structures.

Porous catalyst film is reported [2,63,66,67] to improve the etching performances of micro-scaled Si trenches structures with interconnected catalyst pattern. The porous morphology of the film allows the MacEtch reactants to pass through the catalyst spacing, significantly improving the mass transport and uniformity, which ensures a highly uniform etch rate over all the catalyst area. We recently applied the thermal de-wetting technique to carefully design the film porosity of Au and Pt catalyst and control the vertical etching in micrometer patterns of MacEtch for grating fabrication [21,37,39,40]. Thermal de-wetting is much more robust than evaporation rate to control the film morphology. De-wetting occurs when a thin metal film on a solid substrate is heated, inducing breaking and reassembling of the film [68,69]. The film morphology can be tuned as a function of film thickness and annealing temperature. Figure 4 reports an example of de-wetting for Au film and Pt film deposited on a Si substrate with a cleaned native oxide (oxygen terminated surface). The Pt de-wetting occurs in agreement with literature [69] with a progressive increase of film fractures density (250–350 ◦C) and finally the hole formation appeared (400–500 ◦C), followed by a coalescence process of holes expansion (550–600 ◦C). The thermal treatment in the case of Pt film has two different functions: it creates the porous structure in the metal coating and it forms a platinum silicide at the interface with the substrate that helps to stabilize the catalyst during etching [37].

**Figure 4.** (**a**) Au coverage in percentage of the surface area measured in SEM images in plan-view as a function of the annealing temperature. The Au film thickness was 10 nm and the annealing was performed in air for 30 min. Insets show SEM images of Au film annealed at 180 ◦C (left) and 230 ◦C (right), the scale marker is the same in both images. The figure was adapted with permission from L. Romano et al., 2017. [40] (**b**) Pt de-wetting (12 nm) on (100) Si substrate at temperature of 400, 550, and 600 ◦C. The figure was adapted with permission from L. Romano et al., 2020. [21].

#### *3.2. Alcohols Additives*

Ethanol [8] and isopropanol [39] alcohols have been largely used as surfactant in MacEtch solutions. Like in KOH aqueous solutions with addition of alcohol [70], also for MacEtch the alcohol does not take directly part in the etching process, but it strongly affects the etching. Both etch rate and roughness of the etched surface depend on the alcohol concentration in the etching solution, which is connected with the adsorption phenomena on the etched surface [40]. A common issue of MacEtch is the H2 gas release during the etching process. The H2 is produced as a by-product of reaction [15] and it can substantially affect the etching results since very large bubbles can be formed on the surface of the grating, dramatically preventing a uniform etching. This phenomenon appeared to be much more critical in patterned microstructures than mesh pattern for nanowires since the gas bubbles can be stabilized in the etched structure with liquid solution exhibiting the Cassie-Baxter wetting state [71,72]. The surfactant forms a layer physically covering the surface and prevents the formation of large H2 bubbles, reducing the amount and the size of etchant inhomogeneity in contact with the surface [40]. An example of grating fabricated with and without surfactant additive in the etching solution is showed in Figure 5.

**Figure 5.** SEM in cross-section of 4.8 μm pitch grating etched with regular MacEtch (**a**) and MacEtch solution with the addition of isopropanol alcohol (**b**). The arrows indicate the presence of a gas bubble preventing the uniform etching of the grating. The figure was adapted with permission from L. Romano et al., 2017. [40].

Formation of porous Si is a well-known phenomenon which has been observed in MacEtch. The porous morphology of Si using MacEtch has been attributed to the diffusion of holes outside of the metal-semiconductor interface and causing an additional but reduced extent of etching in the areas outside the metal mesh pattern [73]. Depending on etching conditions, pores with different density and thickness can be found at the catalyst/Si interface, along the sidewalls, and within the etched Si nanostructures. In general, a higher oxidant concentration or higher Si doping concentration results in higher levels of porosity. Once the oxidant is reduced on the surface of noble metal, holes are injected into the Si substrate. The holes diffuse from the Si under the noble metal to the off-metal areas that may be etched and form microporous Si. Balasundaram et al. [73] showed that porosity depends on Si doping, the dopant atoms are thermodynamically favorable sites for the formation of pores, and heavily doped Si in liquid MacEtch produces very porous structures even in conditions of very low H2O2 concentration. The thickness of the microporous Si can be additionally reduced by adding a small amount of alcohol to the etching solution [40]. Figure 6 reports a magnified SEM of the top Si lamellas in Pt-MacEtch with additional methanol, the microporous thickness is less than 50 nm. Methanol is less affecting the etching rate with respect of isopropanol and ethanol alcohols [74].

**Figure 6.** SEM in cross-section of Pt assisted chemical etching of silicon grating with 4.8 μm pitch (**a**,**c**), magnified view of the top (**b**) with etching solution ρ(HF) = 0.9920. Magnified view of top (**c**) with additional methanol in the etching solution. The figure was adapted with permission from L. Romano et al., 2020. [21].

However, a large amount of additive (methanol, ethanol, isopropanol, and acetonitrile in the HF–H2O2–H2O solution) can cause the changing of the etching direction, inducing the formation of curved or tilted structures [74]. An example is reported in Figure 7, where curved Si nanowires are produced in a solution with isopropanol and acetonitrile. H2O2 is more severely shielded from reaction sites by the additive than HF is, due to the higher surface tension of H2O2, effectively increasing the HF to H2O2 ratio locally.

**Figure 7.** SEM images of Si nanowires etched for 20 min in HF–H2O2–water–co-solvent (**a**) isopropanol 2:1:5:2 and (**b**) acetonitrile 2:1:5:2. All scale bars are 2 μm. The figure was adapted with permission from Y. Kim et al., 2013. [74].

#### *3.3. Temperature*

With a temperature in the range of 0 ◦C to 50 ◦C, Cheng et al. [75] observed a linear relationship between length of nanowire and etching time. The etching rate increased with increasing etching temperature with an activation energy of 0.36 eV for the formation of Si nanowires on a (100) Si substrate in AgNO3 and HF aqueous solution.

Temperature has also a strong effect on the etching direction. Figure 8 shows the effect of temperature on etching direction for (100) substrates, as the temperature increases (50–70 ◦C) the etching along the (110) instead of (100) is preferred. Figure 9 shows a comparison of etching at 30 ◦C and 8 ◦C, indicating the temperature reduction as a possible way to control the vertical etching.

**Figure 8.** (**a**) Cross-sectional SEM images of the as-etched Si produced with Au catalyst of *p*-type Si (100) substrate (0.005 Ωcm) in solution of 13.5 M HF and 0.16 M H2O2 at room temperature for 30 min and subsequently etched at the same condition at 60 ◦C for 1 min. Scale bar is 2 μm. The figure was adapted with permission from L. Kong et al., 2017. [51]. (**b**) Temperature dependence of etching rate for different H2O2 concentration, at high temperature the etching direction changes from (100) to (110). The figure was adapted with permission from J. Kim et al., 2011. [50].

**Figure 9.** Cross-sectional SEM images of the as-etched Si produced with a catalyst of 40/5 nm thick Au/Ti of n-type Si (100) substrate (0.9–1.1 Ωcm). The etching temperatures are 30 ◦C (**a**) and 8 ◦C (**b**), respectively. The etching duration and concentration of H2O2 are 12 h and 0.2 M, respectively. Scale bar is 20 μm in all figures. The figure was adapted with permission from J. Yan et al., 2016. [76].

The Si porosity is strongly affected by the etching temperature. Excess holes tend to diffuse laterally, resulting in lateral etching and the formation of pits on the sidewalls. Cold etching temperature is also highly advantageous to reduce the pits on the sidewalls [77]. K. Balasundaram et al. [73] noted

that when the etching temperature decreases, the porosity is reduced. R. Akan et al. [35] reported an increase in surface roughness and porosity at the etching temperature of 40 ◦C.

Another way to observe the effect of temperature is MacEtch in gas phase [37,78], where the HF is delivered in vapor phase and the oxidant in gas phase (oxygen from air) to the metal patterned Si substrate. It has been recently demonstrated that etching in the vapor phase avoids the issues related to wet-etching such as the nanostructures stiction due to capillary effects during liquid drying. Moreover, the pattern transfer from the metal mask to the silicon template is much more precise and defect-less due to the microporosity reduction and the extremely high concentration of HF, which are not accessible in wet-etching. By increasing the temperature in the range 35–40 ◦C, the etching rate increases in agreement with previous studies on MacEtch kinetics in liquid [75]. The etching rate has a maximum at 40 ◦C (see [37]), then it decreases as a function of temperature, indicating that the reaction rate is limited by the desorption of HF. Some examples of nano- and micro-structures of X-ray optics [37] are reported in Figure 10. The etching was realized by evaporating water diluted HF (50 wt.%) at room temperature and exposing the Pt-patterned Si substrate to the HF vapor and air, the gaseous O2 present in the air worked as oxidant for the MacEtch reaction. The Pt-patterned Si substrate is held at 55 ◦C during the etching in order to avoid the moisture's condense, so the MacEtch reaction temperature is 55 ◦C and the reaction is considered to happen with a solid–gas interface.

**Figure 10.** (**a**) Schematic of MacEtch in gas phase. SEM in cross section of structures by gas phase MacEtch at 55 ◦C, HF was evaporated from a water diluted HF solution and the oxidant is supplied by air: (**b**) Si nanowires; (**c**) linear grating with pitch of 4.8 μm; (**d**) circular grating with pitch 1 μm; (**e**) zone plate with outmost pitch of 200 nm. The figure was adapted with permission from L. Romano, 2020. [37].

The use of gas-MacEtch turned out to be very useful to improve the stability of free-standing Si nanostructures such as the nanowires and the zone plate in the outmost region. A totally interconnected catalyst design results in free-standing Si nanopillars. As noted by R. Akan et al. [35] for very high aspect ratios and smallest zone sizes, these pillars will become mechanically unstable. Chang et al. [17] and K. Li et al. [36] increased the number of Si interconnects to solve this issue, but this further reduces the active zone plate area and consequently the efficiency. The zone plate made by gas-MacEtch (Figure 10e) does not need Si interconnects since the stability of Si lamellas is not compromised by the liquid drying.

Moreover, nanowires can be used as diffractive optics in speckle based X-ray phase contrast imaging [79]. Nanowires are expected to improve the sensitivity by producing speckles of smaller size in comparison to sandpaper [80] or other membranes with feature size in the micrometer scale.

#### **4. Silicon Based Microfabrication of X-ray Optics**

Microstructures constitute the X-ray optical elements such as diffractive and refractive X-ray lenses for microfocusing applications at synchrotron beam lines [81] and diffractive gratings for interferometric systems [9,82]. For low energy applications (<10 keV) microfabrication can be realized in silicon but higher X-ray absorbing materials are necessary for hard X-rays. The most common microfabrication approach is based on creating a low X-ray absorbing template and filling it with highly absorbing metal. Historically, the templates are based either on polymer [83,84] or silicon [85].

Silicon microfabrication has been the key technology in manufacturing integrated circuits and microchips in the semiconductor industry. This gives the advantage of a well-assessed technology with a competitive mass production such as deep reactive ion etching [86] and KOH wet etching [87,88]. The combination of unconventional processing and the freedom from microelectronics constrains enrich the spectrum of capabilities and give a new life to the "old silicon material", with revolutionizing advancements in nanotechnology [37]. We recently reported about micro- and nano-fabrication processing for X-ray gratings, including lithography [89], dry [37,90,91] and wet etching methods [39], Au electroplating [22], Ir atomic layer deposition [92] and metal casting [19,20]. The use of MacEtch as a microfabrication process for X-ray optical devices was first reported in 2014 for Fresnel zone plate structures [17,18]. Some SEM images of Fresnel zone plate structures produced by MacEtch are reported in Figure 3; Figure 10. The X-ray nanofocusing effect of Fresnel zone plate optics fabricated by MacEtch and atomic layer deposition of Pt was recently reported by K. Li et al. [93]. In the following section we report an example of X-ray grating interferometry with gratings fabricated by MacEtch. The main challenge for X-ray grating interferometry is the fabrication of the absorption gratings [85], which are metal periodic microstructures, for high energy X-ray (>30 keV).

Deep X-ray lithography (also called LIGA) [83,84] is used to pattern the polymer template. This technology has the advantage that the polymer pattern can be created on whatever substrate, such as a metallic substrate that is used as a seed layer for the following Au electroplating process in order to create the final Au absorbing grating. The metal layer can be deposited on graphite that has the advantage of being flexible and allows to easily bend the Au grating structure. However, LIGA process is limited to relatively small area (10 <sup>×</sup> 10 cm2) and it is quite expensive since it requires a synchrotron facility. In the case of Si based technology, the Si etched structure can affect the quality of the Au electroplating filling, some distortions [22] or voiding inside the template, resulting in less X-ray diffraction efficiency. Atomic layer deposition of metallic coating has been implemented to create a metallization layer for Au electroplating with conventional damascene approaches [94,95] and Au bottom up superfilling processes [96]. MacEtch offers the possibility to benefit of the original catalyst layer as a seed for the Au electroplating filling.

#### *Gratings Fabrication for X-ray Phase-Contrast Imaging*

X-ray grating interferometry (GI) based imaging is a very promising, fast growing and competitive technique for medical, material science and security applications [1]. Contrast in X-ray imaging with GI can be boosted by exploiting refraction and scattering, in addition to conventional absorption. GI might have a large impact on the radiological approach to medical X-ray imaging because it will intrinsically enable the detection of subtle differences in the electron density of a material (like a lesion delineation) and the measurement of the effective integrated local small-angle scattering power generated by the microscopic structural fluctuations in the specimen (such as micro-calcifications in a breast tissue for instance). Similar enhancements are expected in homeland security or material science application, where structural properties such as orientation, degree of anisotropy, average structure size, and distribution of structural sizes can be inferred via omnidirectional tensor tomography [97].

The purpose of an X-ray interferometer is to encode propagation-induced phase changes in the beam wavefront—when passing through a specimen—into an intensity modulation measured by a (usually position sensitive) detector placed downstream. In its simplest configuration, called Talbot Interferometer, an X-ray interferometer consists of two gratings placed in a partially coherent beam. The latter is usually provided by a third/fourth generation synchrotron source or, with significantly less intensity, by a microfocus X-ray tube, see Figure 11. The first grating G1 (of period *p*1) is usually a phase-grating, i.e., it actually does not absorb the beam but it imposes a significant phase shift resulting in a controlled wavefront modulation at a specific distance downstream, usually where the second, absorbing grating G2 (of period *p*2) is placed. G1 essentially divides the incoming beam into the two first diffraction orders: being the grating pitch (*p*1~μm) much larger than the incoming wavelength (~Å), the resulting angle between both diffracted beams is so small that they almost fully overlap, resulting in a linear periodic interference fringe pattern downstream of G1, in planes perpendicular to the optical axis. This effect is known as the fractional Talbot effect [98]. A sample of interest is placed either in front or behind G1, and it usually absorbs, refracts and scatters the incoming beam. These interactions consequently affect the interference pattern: absorption leads to an average intensity reduction, refraction causes a lateral displacement of the fringes and scattering reduces the fringe amplitude. For a phase grating with a phase shift of π illuminated by a plane wave, the periodicity of the fringe pattern equals *p*1/2 [9]. The detector resolution might not be good enough to resolve the interference pattern and therefore a second, absorbing grating G2 (with the same periodicity as the fringes) is placed immediately in front of the detector at the position where the fringes form. This grating behaves as a transmission (analyzer) mask and converts local fringe positions into signal intensity variations. This is a crucial aspect of GI, as the analyzer gratings *de-facto* decouples the phase sensitivity of the system from its intrinsic spatial resolution, making GI suitable for operation on large samples and large field of views. In fact, when the source does not provide a sufficiently high spatial coherence, like in the case of a conventional X-ray tube, a third grating G0 of period *p*<sup>0</sup> can be introduced right after the source yielding to the so-called Talbot-Lau (Figure 11a) configuration. G0 is an absorbing structure that creates an array of individually coherent, but mutually incoherent sources. If the condition *p*<sup>0</sup> = *p*<sup>2</sup> × L01/L12 is fulfilled [85], where L01 is the distance between G0 and G1 and L12 is distance between G1 and G2, then the images created by each line source are superimposed in the image plane. This enables to carry out efficient phase contrast X-ray imaging on commercially, normally incoherent sources. Retrieval of the absorption, phase, and scattering signals has been done with various methods, with the phase-stepping [9] and the fringe scanning [99,100] being the most common used approaches.

**Figure 11.** (**a**) Sketch of an X-ray grating interferometer in the Talbot–Lau configuration. (**b**) Scheme of the phase stepping process.

The intensity modulation of the recorded fringe pattern is usually characterized by its visibility (Equation (5)):

$$V = \frac{I\_{\text{max}} - I\_{\text{min}}}{I\_{\text{max}} + I\_{\text{min}}} \tag{5}$$

where *Imax* and *Imin* are indicated in the phase step curve of Figure 11b. The visibility depends on the degree of spatial coherence of the illumination as well as its spectrum, on the system geometry and on grating's pitch and depth. The interference fringe visibility is a common figure of merit for the design of X-ray gratings interferometers [101]. This is because the formation of high-modulation fringe pattern is a prerequisite for robust grating interferometry.

The main challenge is the fabrication of the absorption gratings [85], which are metal periodic microstructures with high aspect ratio that are usually fabricated starting from templates produced by LIGA [84] or deep Si etching [22,39,85]. The period *p*<sup>2</sup> (*p*0) of the absorbing grating G2 (G0) is usually in the range of 1–20 μm (5–100 μm), while the height (h) depends on X-ray energy and absorption efficiency of the material [102]. A transmission of the structures of less than 25% is acceptable. It can be calculated [102] that 10 μm thickness is sufficient for photon energies below 20 keV, while for photon energies of 30 keV (60 keV), about 25 μm (160 μm) of gold is required. Very challenging, medically oriented projects, require the design of very sensitive interferometers extending on short geometries, imposing quite extreme aspect-ratios for G2 for which periods *p*<sup>2</sup> as small as 1 μm (or below) and gold height of 30 μm might be needed [103]. Such requirements pushed the research efforts toward MacEtch as a new technique that is able to provide aspect ratio structures with period ranging from tens of nanometers to tens of micrometers. Moreover, MacEtch offers the possibility to benefit of the original catalyst layer as a seed for the Au electroplating filling. Figure 12 summarizes the Si template fabrication by using MacEtch and the subsequent Au electroplating.

**Figure 12.** Schematic illustration of the grating fabrication process by MacEtch and subsequent Au electroplating: (**a**) growth of native silicon oxide on Si substrate, (**b**) pattern definition by means of a lithographic process, (**c**) Pt deposition by evaporation, (**d**) lift-off, (**e**) Pt de-wetting by thermal treatment, (**f**) MacEtch in a solution of HF and H2O2, (**g**) Si side wall oxidation in air, (**h**) electrical contact of the catalyst metal interconnected pattern to the electroplating electrode, (**i**) seeded Au growth by electroplating. The Pt pattern works as a catalyst for MacEtch and as a seed layer for Au electroplating. The figure was adapted with permission from L. Romano et al., 2020. [21].

Absorption gratings are usually fabricated by metal electroplating (typically of Au, which is one of the most efficient absorbing materials for X-rays), into high aspect ratio Si templates. The performance of the Au electroplated grating in terms of uniformity and quality of the filling can be well assessed with an X-ray interferometric set up. Figure 13 reports a typical example of grating characterization. The performance of an Au filled grating with a pitch of 6 μm was investigated with an X-ray interferometer [104] (design energy 20 keV) operated at the 3rd Talbot order with π/2 phase shifting G1 grating. The Pt-MacEtch grating (pitch 6 μm, height 39 μm) filled with Au up to 30 μm was used as G0 grating together with a G1 phase grating made of Si by deep reactive ion etching (DRIE) and a G2 absorbing grating fabricated by DRIE and Au electroplating [22]. The average X-ray fringe visibility of 17.5% (Figure 13a) is comparable to the values achieved in absorbing gratings fabricated by conventional DRIE followed by Au electroplating [22]. Figure 13b,c is examples of images obtained with the X-ray grating interferometer, phase contrast (Figure 13b) enhances the detection of low absorbing textures, while dark field (Figure 13c) highlights the presence of microstructures.

**Figure 13.** X-ray performance of an Au electroplated grating of 6 μm period used as a G0 absorbing grating. X-ray fringe visibility map and visibility histogram as insert (**a**), differential phase contrast (**b**), and scattering (**c**) images of a grain ear. The figure was adapted with permission from L. Romano et al., 2020. [21].

**(a) (b) (c)** 

#### **5. Conclusions and Perspectives**

MacEtch is a very powerful and promising technique that is competing the performances of more conventional etching technology, such as deep reactive ion etching and cryogenic processes [86]. Having a clear idea about the fundamentals and recent advances in this area allows researchers to have a better perspective. During the process of etching, a triggering event that produces unequal etch rates might occur. These events can change the effective forces on the catalyst and produce a resultant torque on the catalyst [51]. Such triggering events can deteriorate the etching uniformity and compromise the quality of the pattern transfer from the original lithography to the Si substrate. We discussed the role of the metal catalyst pattern in order to be able to extent the feature size of the etched structure from nanometers to micrometers and to control the etching uniformity. Metal de-wetting technique turned out to be a reliable method to create a porous catalyst layer that allows to etch very high aspect ratio structures. We discussed the critical role of etching solution by showing the effect of reactants depletion and the presence of alcohol additives. A small amount of alcohol can help to improve the etching uniformity and reduce the formation of the microporosity on the side wall trenches. MacEtch in gas phase showed new possibilities of etching nanostructures with exceptional aspect ratio up to 10,000:1. In the framework of X-ray optics, the fabrication of useful microstructures by MacEtch with successful story of implementations started to record new publications in the last few years, with continuously improving performances. In order to give an idea, we selected all the publications about MacEtch that are relevant to produce high aspect ratio structures in silicon (100) substrates for lithographic patterns such as ordered pillars arrays, electron beam lithographic patterns, gratings, Micro Electro Mechanical Systems (MEMS) microstructures, photonic crystals, etc. Figure 14 reports the number of publications as a function of the publication year to the best of the authors' knowledge. The total number of publications in the time range of 2008–2020 years is 81, including this special issue of Micromachines, the trend indicates a substantial increment in the last 10 years. The publications addressing the use of MacEtch for specific X-ray optics fabrication is 13 with the first reports in 2014.

**Figure 14.** Number of publications as a function of year regarding MacEtch for producing high aspect ratio (HAR) patterned structures [2,8,15,17,18,21,24–26,31,32,35–41,43,47,50,51,54,56–58,62,63,71,73,77, 78,93,105–153] and specific for X-ray optics fabrication [17,18,21,35–41,93,119,126,153].

X-ray optics with nanostructured features, such as kinoform lenses and Fresnel zone plates, showed to benefit from MacEtch fabrication. We can envisage new future applications of gas-phase MacEtch when high aspect ratio and feature size in the nanoscale are needed, such in X-ray microscopy with energy higher than 20 keV. A full process of gratings microfabrication for X-ray interferometry by using MacEtch and subsequent Au electroplating is demonstrated with performances similar to other fabrication methods. The Pt catalyst layer that sinks down into the Si substrate during MacEtch has been successfully used as seed layer for Au electroplating in order to fabricate a periodic structure with an high absorbing material for hard X-ray. The possibility to push forward the aspect ratio and the relative low fabrication costs on large area of MacEtch with respect to other technologies, such as reactive ion etching and LIGA, are motivating the investigation efforts. We envisage that MacEtch will become a new enabling technology to fulfill the requirements of grating's height for absorbing hard X-ray radiation and submicron grating's pitch for boosting the sensitivity in grating based high sensitive X-ray systems, as required for instance for early breast cancer detection.

The main bottleneck in grating based X-ray interferometry is the fabrication of high aspect ratio periodic structures, whose quality and homogeneity over large areas strongly affect the contrast of the generated images. There is the need to produce X-ray diffraction gratings with (i) very high aspect ratio (AR ≥ 50:1) in highly absorbing material such as gold; (ii) large area (mammography, e.g., asks for a field of view above 20 <sup>×</sup> 20 cm<sup>2</sup> [103,154]); (iii) good uniformity (no distortions and changes in the duty cycle and depth over the grating area); and (iv) bending capability in order to improve the field of view limitation of cone beam emission from the X-ray source. MacEtch technique has the clear advantage of silicon patterning with high aspect ratio at nanoscale, it is a relatively low cost technology since it is accessible even in labs with limited equipment (no vacuum or clean-room conditions). However, MacEtch is performed in solutions of hydrofluoric acid, we recommend to follow all the necessary safety protocols in order to handle heavily concentrated hydrofluoric acid. The bending capability of the silicon substrate depends on the wafer thickness, extreme bending has been

reported for silicon wafers with thickness below 50 μm but the bending of a thick absorbing grating can be challenging and we predict that dedicated processing needs to be developed to avoid cracks and distortions, such as eventually etching off the silicon template. Four inch wafer scale MacEtch gratings have been demonstrated with good uniformity and high control of trench profile and etching direction for aspect ratio up to 30:1 [39]. The process itself has no limits in terms of patterning area such as LIGA, photolithographic processes are available for Si based technology up to 12 inch wafer scale, periodic linear gratings with pitch size in the range of 100 nm can be patterned on wafer scale by interference lithography [89], nanoimprinting processes can be implemented to further increase the patterned area with nanoscale resolution [55]. High quality structures with high aspect ratio require an etching regime that is dominated by the diffusion of the reactive species. Stirring [26] and large volume of solutions [21] are used to homogenize the concentration of reactants during the etching but the effect on etching rate, aspect ratio and defects need to be systematically investigated in order to anticipate the commercialization of MacEtch as a grating fabrication technology. The fabrication throughput as a function of grating quality and performances needs to be assessed to start the technology transfer from the research laboratory level to an industrial R&D.

MacEtch as a technology is still at its infancy, its control and reproducibility over a large area are still not clear and n to be systematically studied. Further scientific efforts need to be made to take full advantage of high aspect ratio capability and exploit its application.

**Author Contributions:** Both authors contributed equally. All authors have read and agreed to the published version of the manuscript.

**Funding:** We acknowledge the support from: SNF Sinergia Grant CRSII5\_18356 "Clinical GI-BCT", EU Grant ERC-2012-StG 310005 "PhaseX", ERC-2016-PoC 727246 "Magic", Eurostar Grant E!1106 "INFORMAT" and NanoArgovia Grant 13.01 "NANOCREATE" (Swiss Nanoscience Institute) and Fondazione Gelu (TACI-C Project).

**Acknowledgments:** We would like to thank K. Jefimovs, J. Vila-Comamala, M. Kagias (PSI-TOMCAT) and C. David (PSI-LMN) for useful discussion.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Micromachines* Editorial Office E-mail: micromachines@mdpi.com www.mdpi.com/journal/micromachines

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18