Study on the strain in a silicon microchannelplate by micro-Raman analysis

Shaohui Xu1,2, Jiabing Fang1, Dajun Wu1, Chi Zhang1, Yiping Zhu1,Dayuan Xiong1, Lianwei Wang1, Pingxiong Yang1 and Paul K Chu2

1Key Laboratory of Polar Materials and Devices, Ministry of Education, Department of ElectronicEngineering, East China Normal University, 500 Dongchuan Road, Minhang District, Shanghai, 200241,People’s Republic of China2Department of Physics and Materials Science, City University of Hong Kong, Tat Chee Avenue,Kowloon, Hong Kong, People’s Republic of China

E-mail: shxu@ee.ecnu.edu.cn

Received 22 December 2015, revised 17 February 2016Accepted for publication 26 February 2016Published 29 March 2016

AbstractMicro-Raman analysis was used to identify the oxidation of a silicon microchannel plate(SiMCP) and it indicated that the bend phenomenon of the SiMCP was related to the release ofstress and the volume expending of the silicon wall during the oxidation process.

Keywords: silicon microchannel plates, oxidation, stain, micro-Raman

(Some figures may appear in colour only in the online journal)

1. Introduction

A typical microchannel plate (MCP) consists of about10 000 000 closely-packed, oriented parallel channels of acommon diameter, typically, 10–100 μm. The channel matrixcan allow charge replenishment from an external voltagesource and each channel acts as an independent, continuousdynode photomultiplier (PMT). MCPs have direct sensitivityto charged particles and energetic photons, and they havebeen used in a wide range of particle and photon detectionsystems due to their high electron multiplication factors andhigh time spatial resolutions, such as electromagnetic wavesincluding x-ray, visible light and ultraviolet light, high energyparticle detection in astronomy, e-beam fusion studies andnuclear science [1–8].

A conventional MCP is formed by drawing, etching orfiring a lead glass matrix in hydrogen [1]. However a fewbasic problems exist with a glass MCP, such as operation andfabrication at relatively low temperatures, evolution ofhydrogen ions from the surface (potential photocathodedegradation), gain degradation and short gain-lifetime at highincident electron fluxes, and relatively low pore areal packingfractions, etc. In this case, the MCP based on silicon hasinspired many interesting topics and attracted much attentiondue to its low noise, high-temperature compatible process,long stability and potential for there to be much more space

for the improvement of gain [2–7]. A silicon MCP (SiMCP) isfabricated by the etching of silicon, which is compatible withthe use of micromachining processes. Beetz et al [2] fromNanoscience Co. Ltd developed an electrochemical processand successfully fabricated a silicon-based MCP. The devicehad a high secondary electron yield and better durability.However, the conventional backside thinning process ofgrinding and polishing the backside of the substrate is a costlyand time-consuming process. Moreover, the lapping processmay cause a block of the channels due to the particles gen-erated during the lapping and polishing process. However,through controlling the backside illumination intensity andsuitable bias simultaneously in the photo-assisted electro-chemical etching process, the SiMCP can be separated fromthe substrate when the desired depth is reached (self-undercut)[9–11]. In this case, a SiMCP can be fabricated easily bytypical microelectromechanical system techniques involvingphoto-assisted electrochemical etching and delaminating fromthe Si substrate by a modified electrochemical procedure. Dueto its larger surface area and three-dimensional (3D) archi-tecture, the obtained SiMCP can be employed to fabricate 3Delectrochemical energy storage devices, such as Li-ion bat-teries and supercapacitors [12–16]. In order to decreaseexcessive conductivity and change the surface properties ofSiMCP, which may be an application in a PMT or some kindof 3D electrochemical energy storage devices, the high-

Semiconductor Science and Technology

Semicond. Sci. Technol. 31 (2016) 055010 (8pp) doi:10.1088/0268-1242/31/5/055010

0268-1242/16/055010+08$33.00 © 2016 IOP Publishing Ltd Printed in the UK1

temperature thermal oxidation process is needed to convertthe SiMCP entirely or partially to a silicon-oxide structure(amorphous quartz-qMCP). However the bendy SiMCPalways appears after the thermal oxidation processing andmay influence the later process to fabricate the devices. It isdifficult to calibrate the bendy SiMCP, in which the macro-scopical plate (always centimeter size) is a composite ofmicrometer-size channels. The bendy morphology of aSiMCP is always a macroscopic view and can be observed byeye or an optical microscope. However, the smooth structurecan always be observed in micrometer size by a scanningelectron microscope (SEM). It is difficult to calibrate thedetailed bend curve of oxidated SiMCP in the microscopicview. In this work, the micro-Raman analysis is used toidentify the bendy oxidated SiMCP and the mechanism isproposed.

2. Experimental details

The starting materials were 100 mm, 5–15Ωcm, single-side-polished p-type (100) silicon wafers with a thickness of525 μm; the standard microelectronics fabrication steps wereimplemented to produce the SiMCP. The steps includedthermal oxidation to produce a masking layer and3 μm×3 μm squares were patterned by lithography and wetetching. The patterned wafer was pre-etched in a tetramethylammonium hydroxide (TMAH) solution (25% at 85 °C). Thepatterned sample is etched in 2 mol L−1 Hydrofluoric acid(HF) under 15 V anodic bias at room temperature with theetching current density about 20 mA cm−2. The electrolyteswere prepared by mixing HF (40% aqueous solution), N,N-Dimethylformamide (DMF, 99.5%), and deionized water atthe volume ratio 1: 4: 5. A non-ionic surfactant Triton X-100was used to remove bubbles and HCl was added to adjust thepH value by about 2 to passivate the pore wall. The set-upconsisted of a double cell and Light Emitting Diode (LED)arrays with a wavelength of 850 nm. A platinum grid (9 cm indiameter) was used as the counter electrode and anotherplatinum grid with the same size was used as a workingelectrode. The current was adjusted by the light intensity. Thewhole system was controlled by a computer system. Moredetails about the process can be found in [9–11]. Finally, thestandard SiMCP circle samples in 16 mm diameter sizes wereobtained through a 200W CO2 laser-cut machine (as shownin figure 1(a)). The oxidation step of the free-standing SiMCPsamples was performed at 1080 °C, and the oxidation pro-cesses including that of the former 20 min dry oxidation and3 h wet-oxygen oxidation and the later 20 min dry oxidation.The phosphorous-diffusion step of the free-standing SiMCPsamples was performed at 1030 °C for 30 min in a nitrogenatmosphere using POCl3 as the diffusion source. The sampleswere designated as SiMCP, SiMCPO (after the oxidationprocess) and NSiMCP (after phosphorous diffusion).

The morphology and cross section of the SiMCP sampleswere examined by a field-emission SEM (FE-SEM, HitachiS-4800, Japan). Raman scattering was carried out on a HoribaJobin-Yvon LabRam Aramis micro-Raman spectrometer with

the 785 nm laser line as an exciting source. The laser powerwas 5 mW and the spot size on the sample was about 2 μm.The Raman system was recorded in a backscattering config-uration with mapping capabilities with a 400 mm mono-chromator and Charge-Coupled Device (CCD) detector; thegrating 600 g mm−1 and microscope objective lens 100 timesare selected to obtain the Raman data. All the experimentswere performed at room temperature, and the thermal effectdue to the heating of the sample by laser was negligible. TheSiMCP in the macroscopic view can also be obtained throughan optical microscope.

3. Results and discussion

After electrochemical etching, the SiMCP is well aligned andthe channel has square holes 5×5 μm in size, pore walls1 μm and period 6 μm, as shown in figure 1(b). The edgeview of the cross section of a dense array of channels with themicromachining process halted at 200 μm, and chevroning upto ∼7 degrees to the surface plane has been demonstrated, asshown in figure 1(c). After the oxidation process, the almost0.7 μm thickness silicon dioxide is shown at each side of the0.3 μm silicon wall in cross-sectional views of the thin-wallSiMCP substrate converted to amorphous quartz, as shown infigure 1(d). However it must be pointed out that the bendcurve cannot be identified by a SEM, which can used to showthe microscopical view in micrometer size. The opticalmicroscope can be used to identify a typical bendy SiMCPO,as shown in the schematic diagram in figure 2(a). In order toaid observation, the x axis is used to identify the direction,which the two-edge sides of SiMCPO can touch with thebaseplate. The y axis is vertical with the X direction and isused to identify the direction, which the two-edge sides ofSiMCPO upturn. The schematic diagram can be identified bythe optical picture as shown in figures 2(b) and (c). In order toshow the full bendy curves of SiMCPO, the splicing of twosegments is used. Along the X direction, two-edge sides ofSiMCPO are upturning about 0.45 mm from the baseplate.Considering the bendy curve is equal to the diameter of theSiMCP circle (16 mm), it can estimate that the radius ofcurvature is 74.8 mm. Along the Y direction, two-edge sidesof SiMCPO are touched with the baseplate and the middle-point of the bendy curves raise about 0.4 mm; the radius ofcurvature is 78.7 mm. It indicates that the compression stressmakes the middle of the SiMCP upsweep along the Xdirection, and the tensile stress makes the two-edge sides ofSiMCP upsweep along the Y direction during the oxidationprocess; the schematic diagram is shown in figure 2(a). Thealmost identical radius of curvatures along the X and Ydirections show the similar bend tendencies. Also, themicroscopy picture (figure 2(d)) shows that the SiMCPO iscurved along the diagonal line.

In order to give detailed information about the bendcondition of the SiMCPO, the mapping function of the micro-Raman spectrometer can be used to identify the micro-structure of the channel. In all test processes, the suitable linesare selected in macroscopic view to obtain the Raman map in

2

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

Figure 1. (a) Photograph of SiMCP in size 16 mm formed by photo-assisted electrochemical etching. (b) Morphology of the SiMCP withsquare holes 5×5 μm in size and pore walls 1 μm. (c) Cross-sectional overall view of the SiMCP with chevroning up to ∼7 degrees. (d)Cross-sectional views of the thin-wall SiMCP substrate converted to amorphous quartz after the oxidation process.

-30

-25

-20

-15

-10

-5

0

5

10

15

20

25

30-20 -10 0 10 20

Length (µm)

Wid

th (µ

m)

30

5µm

-30

Along the Y direction

Along the X direction

Y axis

X axis

1

2

3

a b

c

d

Figure 2. (a) Schematic diagram of bendy SiMCP after the oxidation process (SiMCPO). The optical microscope of bendy curves ofSiMCPO along the X (b) and Y (c) directions. (d)Microscopy surface image of SiMCPO. The solid white lines are the direction to obtain theRaman map.

3

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

a wavenumber region. At first, the silicon wafer is used to geta Raman mapline along vertical directions (X and Y) in awavenumber region between 500 to 550 cm−1 as shown in the3D pictures in figures 3(a) and (b). It shows that the Ramanpeak of silicon appears at wavenumber 520 cm−1 and theRaman intensity of the peak is uniform along the X and Ydirections in the field view about 20 μm. For the SiMCP, thevertical lines are selected along the silicon wall (thickness∼1 μm) in the macroscopic view to obtain the Raman map asshown in the 3D pictures of figures 3(c) and (d). It can benoticed that the Raman peaks of silicon appear at lowerwavenumbers than 520 cm−1 and the Raman intensityupgrades and downgrades periodically along the X and Ydirections. It shows that the Raman intensity is sensitive to the

space distribution of silicon and the Raman intensity formspeaks along the spatial directions due to the regular squarechannel structure in the SiMCP. Then the integrated intensityof Raman spectra along the spatial also show the periodicalpeaks for the X and Y directions (figure 3(e)). It indicates thateach peak corresponds to the crosspoint of the silicon wall,which the larger silicon area can obtain the maximum Ramanintensity. In this case the positions of the Raman peaks infigure 3(e) can be read and show the linear relation with thepeaks’ distance, as shown in figure 3(f). The slopes of linesare about 5.97 μm and 5.96 μm for the X and Y directions,respectively, which are consistent with the spatial size of theSiMCP (about 6 μm, as shown in figure 1(b)). Note that thenon-uniformity or variations of Raman peaks are shown in

Figure 3.Raman map in 3D format of silicon wafer (a) and (b) and SiMCP (c) and (d) along the X and Y directions. (e) Integrated intensity ofRaman spectra along the X and Y directions for the SiMCP. (f) Positions of Raman peaks with peaks’ distance along the X and Y directionsfor the SiMCP.

4

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

figures 3(a)–(d); this may be due to the defects or con-tamination on the surface of Si or the SiMCP. However thenon-uniformity or variations can be neglected after the inte-gration process of the Si Raman peak between the wave-number 500 to 550 cm−1.

It shows that the mapping function of the micro-Ramanspectrometer can be used to identify the size of the micro-channel. Then the SiMCP after the oxidation process(SiMCPO) can also be identified by the micro-Raman spectrawith vertical lines mapping as shown in figure 2(a). In orderto show the different bend condition of the SiMCPO, the threepoints are selected. Point 1 is the middle of the circle, point 2stays at the front side of the circle and point 3 stays at theright side of the circle. The integrated intensity of Ramanspectra along the spatial also show the periodical peaks for theX and Y directions in figures 4(a), (c) and (e) for points 1, 2and 3 respectively, and the corresponding positions of theRaman peaks can be read and show the linear relation withthe peaks’ distance, as shown in figures 4(b), (d) and (f),respectively. Considering the bendy condition of theSiMCPO, the lines of the Raman map should be the diagonalline, as shown by the solid white lines in the microscopypicture in figure 2(d). That is to say, the standard distance ofthe silicon crosspoint (or distances of Raman peaks) should be8.48 μm. In this case, it shows that the square structure of thechannel at point 2 (front side of the circle) is not influenced bythe bendy SiMCPO, for which the distance of the Ramanpeaks are 8.49 and 8.48 μm along the x and y axes, respec-tively. However, for point 1 (middle of the circle), the squarestructure of the channel changes to a rectangle, and the siliconwall is lengthened to 9.27 μm along the x axis and shortenedalong the y axis (8.36 μm). For point 3 (right side of thecircle), the square structure of the channel changes to a rec-tangle, and the silicon wall is lengthened to 9.01 μm along they axis, and 8.59 μm along the x axis. It indicates the differentchanges of the square structure of the channel at differentpoints. The upward and downward bending of the SiMCPOalong the x and y axes, respectively, makes the X side walllengthen, and the Y side wall shorten for the middle of thecircle (point 1). However the square structures cannot bechanged due to the bendy structure along the x axis for point 2(front side of circle). For point 3, the raising-up structuremakes the silicon wall lengthen both for x and y axes.

In order to give detailed information about the strainparameters of the bending condition, the Raman spectra ofdifferent samples can be obtained at the middle of the circle;the corresponding Raman spectra are shown in figure 5(a),which are marked as SiMCP, NSiMCP and SiMCPO,respectively. The Raman spectrum for a monocrystalline Siwafer was also shown for comparison (marked as Si). In orderto fit easily, all the experimental data are normalized. Toquantitatively explain the experimental results, the phononconfinement model just described is employed [17–19]. TheSiMCP is modeled as an assembly of thin films with thick-nesses of about 1.0 μm. The first-order Raman spectrum isthus given by a weighted Lorentzian contribution over the

whole Brillouin zone

òww w

=- + G /

IC q q

q

d

21

BZ

2 3

20

2( ) ∣ ( )∣

| ( ) | ( )( )

where a0 is the lattice constant of the bulk Si crystal and Γ0 isthe Raman intrinsic linewidth—full width at half maximum ofthe unperturbed line shape determined from the bulk Si. Thephonon dispersion relation ω(q) is taken according to:

w w= -⎛⎝⎜

⎞⎠⎟q

q

q120 20

0

2

( ) ( )

where ω0=520 cm−1 and q0=2π/a0 gives the scatteringprobability of phonons with the different wave vector q.

p= - - /

⎛⎝⎜

⎞⎠⎟C Lq 1 erf

iqL

32exp q 8 32

22 2| ( ) | ( ) ( )

L denotes the average size of the thin film. By usingequations (1)–(3), the first-order Raman spectra can be cal-culated to fit the experimental Raman spectrum (dots) throughadjusting the peak shift and the peak width. The calculatedresults are also shown (solid lines) in figure 5(a). For Si, thesymmetrical Raman band stays at 520 cm−1 with intrinsiclinewidth-full width Γ0 3.5 cm−1 (including instrumentationbroadening). After the oxidation process (bendy SiMCPO),the Raman band shifts to 519 cm−1 with the almost samelinewidth-full width Γ0. However, it can be noticed that theRaman band broadens asymmetrically on the low energy sideand shifts to lower frequencies are observed for the SiMCP; itis clear that the phonon confinement model does not totallycorrelate to the experimental results. It has shown that thelattice expansion will cause an additional red shift of theRaman peak for Si or porous silicon [20–23]. It suggests thatthe difference between the experimental data and the theorycurve can be attributed to the effect of strain, and the differentrelationship Δω/ω0 can be used to give an estimation of thestrain by measuring the difference of peak shifts (Δω) at aspecific linewidth. It shows that the asymmetrical Ramanband at 518.5 cm−1 with the Raman intrinsic linewidth-fullwidth Γ0 7.3 cm

−1 can be obtained by theory calculation (dotsline); it further reveals that the data can be consistent withexperimental data with a low frequency shift 2.8 cm−1, andthe strain Δa/a 0.2% can be estimated [19].

It indicates that the strain (lattice expansion) appears inthe SiMCP structure during the photo-assisted electro-chemical etching process. The origin of the lattice expansionis believed to be related to the hydrogen–silicon bonds at theinner surface of the SiMCP layer [21, 22]. Based on thedissolution chemistries of silicon [24–26] during electro-polishing at the pore tips, etching will occur in the pore tipsdue to the geometric field enhancement and all four Si elec-trons are electrochemically active, and the Si hydride bondspassivate the Si surface unless a hole is available. The stressin the SiMCP deduced by strain cannot be released until theoxidation process is introduced, and the bendy SiMCPOappears to release the stress during the conversion process ofsilicon to silicon dioxide. The hypothesis can also be

5

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

supported by the experimental observation of the phosphor-ous-diffusion process (NSiMCP). After the phosphorous-dif-fusion process at high temperature (1030 ◦C), the plate is notbendy. The Raman band broadens asymmetrically on the lowenergy side and shifts to lower frequencies (517.0 cm−1) witha Raman intrinsic linewidth-full width Γ0 7.0 cm

−1 observed

for NSiMCP. However, the distorted channel structure can befound at the microcopy image, as shown in figure 5(b). That isto say, the phosphorous-diffusion process of SiMCP releasesthe stress in the inner structure and distorts the silicon wall ofSiMCP, and the macroscopical structure of NSiMCP is notbendy.

Figure 4. Integrated intensity of Raman spectra and positions of Raman peaks with the peaks’ distance along the X and Y directions foroxidated SiMCP (SiMCPO) for point 1 (a) and (b), point 2 (c) and (d) and point 3 (e) and (f).

6

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

4. Conclusion

Micro-Raman analysis, especially the Raman mapping func-tion, has been used to identify the macroscopical silicon plate(mm size) composite of micrometer-size channels. The rea-sonable agreement of the experimental and theoretical Ramanresults indicates that strain exists in the SiMCP during theelectrochemical etching process. The bend phenomenon ofSiMCP during the oxidation process is related to the releaseof stress and the volume expending of Si changes as silicondioxide. The results can be used to explain the mechanism ofcurving SiMCP after the oxidation process and make theoxidated SiMCP easier to apply in a PMT or energy-storingdevices.

Acknowledgments

This work was jointly supported by the Shanghai PujiangProgram (No. 14PJ1403600), the Shanghai Fundamental KeyProject (No. 11JC1403700), the National Natural ScienceFoundation of China (No. 61176108), PCSIRT, the ResearchInnovation Foundation of ECNU (No. 78210245), and theCity University of Hong Kong Applied Research Grant(ARG) No. 9667104.

References

[1] Wiza J L 1979 Microchannel plate detectors Nucl. Instrum.Methods 162 587

[2] Beetz C P, Boerstler R, Steinbeck J, Lemieux B and Winn D R2000 Silicon-micromachined microchannel plates Nucl.Instrum. Methods Phys. Res. A 442 443

[3] Siegmund O H W, Tremsin A S, Vallerga J V and Hull J 2001Cross strip imaging anodes for microchannel plate detectorsIEEE Trans. Nucl. Sci. 48 430

[4] Siegmund O H W 2004 High-performance microchannel platedetectors for UV/visible astronomy Nucl. Instrum. MethodsPhys. Res. A 525 12

[5] Pagonis D N and Nassiopoulou A G 2006 Free-standingmacroporous silicon membranes over a large cavity forfiltering and lab-on-chip applications Microelectron. Eng.83 1421

[6] Winn D R 2007 High gain photodetectors formed by nano/micromachining and nanofabrication IEEE Nucl. Sci. Symp.Conf. Record 1 65–72

[7] Franco A, Riesen Y, Wyrsch N, Dunand S, Powolny F,Jarron P and Ballif C 2012 Amorphous silicon-basedmicrochannel plates Nucl. Instrum. Methods Phys. Res. A695 74

[8] Franco A, Geissbühler J, Wyrsch N and Ballif C 2014Fabrication and characterization of monolithically integratedmicrochannel plates based on amorphous silicon Sci. Rep.4 101

[9] Chen X, Lin J, Yuan D, Ci P, Xin P, Xu S and Wang L 2008Obtaining of high area-ratio free-standing siliconmicrochannel plate via modified electrochemical procedureJ. Micromach. Microeng 18 037003

[10] Yuan D, Ci P, Tian F, Shi J, Xu S, Xin P, Wang L and Chu P K2009 Large size P-type silicon microchannel plates preparedby photo-electro chemical etching J. Micro/Nanolithography, MEMS, and MOEMS (JM3) 8 033012

[11] Peng B, Wang F, Liu T, Yang Z, Wang L, Fu R K Y andChu P K 2012 Novel method of separating macroporousarrays from p-type silicon substrate J. Semicond. 33 043004

[12] Liu T, Xu S, Wang L, Chu J, Wang Q, Zhu X, Bing N andChu P K 2011 Miniature supercapacitors composed ofnickel/cobalt hydroxide on nickel-coated siliconmicrochannel plates J. Mater. Chem. 21 19093

[13] Wang F, Xu S, Zhu S, Peng H, Huang R, Wang L, Xie X andChu P K 2013 Ni-coated Si microchannel plate electrodes inthree-dimensional lithium-ion battery anodes Electrochim.Acta 87 250

[14] Li M, Xu S, Liu T, Wang F, Yang P, Wang L I and Chu P K2013 Electrochemically-deposited nanostructured Co(OH)2flakes on three-dimensional ordered nickel/siliconmicrochannel plates for miniature supercapacitors J. Mater.Chem. A 1 532

[15] Li M et al 2015 Hierarchical 3-dimensional CoMoO4

nanoflakes on a macroporous electrically conductivenetwork with superior electrochemical performanceJ. Mater. Chem. A 3 13776

[16] Wu D et al 2015 Hybrid MnO2/C nano-composites on amacroporous electrically conductive network forsupercapacitor electrodes J. Mater. Chem. A 3 16695

Figure 5. (a) Raman spectra of the Si wafer (Si), the SiMCP and the SiMCP after oxidation (SiMCPO) and phosphorous-diffusion (NSiMCP)processes. (b) Microcopy image of NSiMCP.

7

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al

[17] Campbell H and Fauchet P M 1986 The effects of microcrystalsize and shape on the one phonon Raman spectra ofcrystalline semiconductors Solid State Commun. 58 739

[18] Fauchet P M and Campbell I H 1988 Raman spectroscopy oflow dimensional semiconductors Crit. Rev. Solid StateMater. Sci. 14 79

[19] Yang M, Huang D, Hao P, Zhang F, Hou X and Wang X 1994Study of the Raman peak shift and the linewidth of light-emitting porous silicon J. Appl. Phys. 75 651

[20] Lehmann V and Föll H 1990 Formation mechanism andproperties of electrochemically etched trenches in n‐typesilicon J. Electrochem. Soc. 137 653

[21] Lehmann V and Gosele U 1991 Porous silicon formation—aquantum wire effect Appl. Phys. Lett. 58 856

[22] Bellet D and Dolino G 1996 X-ray diffraction studies of poroussilicon Thin Solid Films 276 1

[23] Xu S H and Wang L W 2009 Porous silicon microtube structuresinduced by anisotropic strain J. Appl. Phys. 106 073516

[24] Lehmann V and Rönnebeck S 1999 The physics of macroporeformation in low-doped p-type silicon J. Electrochem. Soc.146 2968

[25] Bisi O, Ossicini S and Pavesi L 2000 Porous silicon: a quantumsponge structure for silicon based optoelectronics Surf. Sci.Rep. 38 1

[26] Christophersen M, Carstensen J, Rönnebeck S, Jäger C,Jäger W and Föll H 2001 Crystal orientation dependenceand anisotropic properties of macropore formation of p- andn-type silicon J. Electrochem. Soc. 148 E267

8

Semicond. Sci. Technol. 31 (2016) 055010 S Xu et al