highly porous silicon membranes fabricated from...

1© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com

Highly Porous Silicon Membranes Fabricated from Silicon Nitride/Silicon Stacks Chengzhu Qi , Christopher C. Striemer , Thomas R. Gaborski , James L. McGrath , and Philippe M. Fauchet *

1. Introduction

Ultrathin porous nanocrystalline silicon (pnc-Si) membranes

fabricated by crystallizing amorphous silicon thin fi lms

deposited by rf sputtering were fi rst published by Striemer

et al. in 2007. [ 1 ] The advantages this new nanoporous mem-

brane offers over other nanoporous membranes were clear.

First, the ultrathin nature of this membrane, only a few tens

of nanometers thick, is comparable to the size of many bio-

molecules. As a result, mass transport through this membrane

is greatly enhanced as compared to traditional tortuous path

polymer membranes. [ 2 ] Another advantage is the simple

and inexpensive fabrication process, involving a bottom-

up approach, rather than a top-down lithographic approach

that other porous nano-membranes rely on. [ 3 ] Nanopores

are spontaneously formed in silicon during rapid thermal

annealing of ultrathin (≈15 nm) amorphous Si fi lms sand-

wiched between nm-thick SiO 2 layers deposited on a Si

wafer, a novel phenomenon that had never reported before.

A lot of research has been conducted to utilise this unique

porous membrane. [ 4–9 ] Applications range from biological

and gas separations to use as sensors (pressure, tempera-

ture, biological) or as a platform for electron imaging and

spectroscopy. [ 4–9 ] DOI: 10.1002/smll.201303447

Nanopore formation in silicon fi lms has previously been demonstrated using rapid thermal crystallization of ultrathin (15 nm) amorphous Si fi lms sandwiched between nm-thick SiO 2 layers. In this work, the silicon dioxide barrier layers were replaced with silicon nitride, resulting in nanoporous silicon fi lms with unprecedented pore density and novel morphology. Four different thin fi lm stack systems including silicon nitride/silicon/silicon nitride (NSN), silicon dioxide/silicon/silicon nitride (OSN), silicon nitride/silicon/silicon dioxide (NSO), and silicon dioxide/silicon/silicon dioxide (OSO) were tested under different annealing temperatures. Generally the pore size, pore density and porosity positively correlate with the annealing temperature for all four systems. The NSN system yields substantially higher porosity and pore density than the OSO system, with the OSN and NSO stack characteristics fallings between these extremes. The higher porosity of the Si membrane in the NSN stack is primarily due to the pore formation enhancement in the Si fi lm. We hypothesize that this could result from the interfacial energy difference between the silicon/silicon nitride and silicon/silicon dioxide, which infl uences the Si crystallization process.

Porous Silicon Membranes

Dr. C. Qi, Prof. P. M. Fauchet Department of Electrical Engineering and Computer ScienceVanderbilt University Nashville , TN 37235 , United States E-mail: [email protected]

Dr. C. Qi Materials Science ProgramUniversity of Rochester Rochester , NY 14627 , United States

Dr. C. C. Striemer, Prof. P. M. Fauchet Department of Electrical and Computer EngineeringUniversity of Rochester Rochester , NY 14627 , United States

Prof. T. R. Gaborski Department of Chemical and Biomedical EngineeringRochester Institute of Technology Rochester , NY 14623 , United States

Dr. C. C. Striemer, Prof. T. R. Gaborski SiMPore Inc., West Henrietta New York 14586 , United States

Prof. J. L. McGrath Department of Biomedical EngineeringUniversity of Rochester Rochester , NY 14627 , United States

small 2014, DOI: 10.1002/smll.201303447

http://doi.wiley.com/10.1002/smll.201303447

C. Qi et al.

2 www.small-journal.com © 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

full papers Our previous results have shown that one of the prereq-

uisites for the nanopore formation in the silicon fi lm is the

presence of both top and bottom SiO 2 layers. [ 10 ] The top

SiO 2 layer prevents the silicon fi lm from agglomerating [ 11 ]

while the bottom SiO 2 layer works as a barrier layer to pre-

vent homoepitaxy [ 12,13 ] from the Si wafer during annealing.

Recent results by Balachandra Achar et al. [ 14,15 ] confi rm our

results even though their deposition technique was different

and they used much thicker SiO 2 layers. Experimenting with

barrier layers other than silicon dioxide could be a fi rst step

toward generalizing the phenomenon to other material pairs.

Silicon nitride, another dielectric material with favorable

optical properties, high thermal stability, chemical inert-

ness, and good dielectric properties, has been widely used in

optoelectronic devices [ 16,17 ] and integrated circuit fabrication

processes such as gate dielectrics, diffusion barriers, isolation

materials and fi nal passivation layers. [ 18–20 ]

In this work, we incorporate silicon nitride as an

alternative barrier layer in stack geometries including

nitride/silicon/nitride, oxide/silicon/nitride and nitride/sil-

icon/oxide (NSN, OSN and NSO), and demonstrate for the

fi rst time that silicon nanopores can also be spontaneously

formed by these thin fi lm stacks. The porosity, pore density,

and average pore diameter are compared in these four mate-

rials systems.

2. Results and Discussion

Four different stack systems including OSO, NSN, OSN and

NSO have been deposited on silicon wafers. The thickness of

the silicon fi lm was kept at 25 nm and that of the oxide and

nitride barrier layers at 30 nm. These four stacks were then

annealed at different temperatures using a rapid thermal pro-

cessor (fabrication details see Experimental Section).

In the OSO system, nanopores and silicon nanocrystals

are formed during rapid thermal annealing and the annealing

tempeature is a key factor that controls the pore size and

porosity. Figure 1 shows bright fi eld transmission electron

microscopy (TEM) images of the OSO system annealed at

four different temperatures from 800 °C to 1100 °C. In the

bright fi eld TEM images, the pores appear as bright spots

while the silicon crystals are in grey or black depending on

their crystallographic orientation. Only a few small pores

are formed in the silicon fi lm when the annealing tempera-

ture is below 1000 °C. The size of these nanopores is around

10 nm in diameter. With increasing annealing temperature,

the pores become larger, and reach 30 nm at 1100 °C. Both

pore size and porosity increase with the annealing tempera-

ture, in agreement with our previous results. [ 21 ]

By replacing the top and/or bottom silicon dioxide with

silicon nitride, three different stack systems including NSN,

small 2014, DOI: 10.1002/smll.201303447

Figure 1. TEM images of the OSO system after RTP annealing at a) 800 °C, b) 900 °C, c) 1000 °C and d) 1100 °C.

Highly Porous Silicon Membranes Fabricated from Silicon Nitride/Silicon Stacks

3www.small-journal.com© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

OSN and NSO are explored. Figure 2 shows TEM images

of the annealed NSN system at four different temperatures

from 800 °C to 1100 °C. This is the fi rst time that pores are

reported in ultrathin Si membranes fabricated using another

system besides the OSO system. As with the OSO system, the

pore size and density in the NSN system increase with the

annealing temperature. However, direct comparison of the

TEM images shows that the pores are bigger and the pore

density is much higher in the NSN system.

Since nanopores can be formed in both the NSN and

OSO systems, it is likely that nanopores would be formed

in the mixed OSN and NSO systems. In the OSN system the

top oxide layer is replaced with nitride while in the NSO

system the bottom oxide layer is replaced with nitride. The

TEM images of Figure 3 and 4 confi rm that nanopores are

formed in silicon fi lms after annealing in these mixed systems.

Both systems produce a high density of nanopores when the

annealing temperature is greater than 1000 °C. They show

similarities in the pore evolution with annealing temperature

among the four systems. However, the choice of the insu-

lating layers and their deposition order affect the pore char-

acteristics, such as the pore size and pore density.

In order to fully compare the pore characteristics of the

four different systems, customized pore recognition soft-

ware was developed to quantify the pore density, pore size

and porosity from the TEM images (available for down-

load at nanomembranes.org/ resources/software, Supporting

Information).

Pore density is an important parameter and Figure 5 pre-

sents a pore density comparison among the four systems. The

pore density in the NSN system, which stays the highest among

the four systems for all annealing temperatures, is already

high at 800 °C, increases gradually with temperature and

appears to saturate at 1000 °C. The pore density in the OSO

system, which remains the lowest among the four systems for

all annealing temperatures, is very low at 800 °C, changes little

until the annealing temperature reaches 1000 °C and increases

slightly at 1100 °C. Even at 1100 °C the pore density of the

OSO system is 5 to 6 times lower than for the other three

systems. The OSN and NSO systems also start with very few

pores at 800 °C, however, the pore density increases rapidly as

the annealing temperature increases. The pore density of the

OSN system exceeds that of the NSO system at 1000 °C and

reaches that of the NSN system at 1100 °C. This comparison

shows that high density pores can be formed in Si fi lms in the

OSN, NSO and NSN systems at high annealing temperature.

The pore formation in the Si fi lm is greatly enhanced by the

existence of even a single silicon nitride capping layer.

Pore size is another important pore characteristics.

Figure 6 compares the average pore diameter (APD) of

small 2014, DOI: 10.1002/smll.201303447

Figure 2. TEM images of the NSN system after RTP annealing at a) 800 °C, b) 900 °C, c) 1000 °C and d) 1100 °C.

C. Qi et al.


full papers

the four systems. The differences in the APD among these

four system are not large as for the pore density. The NSN

system yields slightly larger APD among four systems when

the annealing temperature is below 1000 °C. The APDs

in the OSN and NSO systems are smaller than in the NSN

system for all temperatures. The APD for the OSO system

is the smallest at 800 °C but it increases rapidly, becoming

the largest at 1100 °C. This can be understood by plotting the

pore distribution. Figure 7 shows the pore distributions of the

four systems at 1100 °C. In the OSO system, the pore den-

sity is not high but the majority of the pores are from 24nm

to 38nm in diameter. In the NSN system, the pore density is

much higher than that of the OSO system but the majority of

the pores are from 20nm to 30nm in diameter. The pore dis-

tributions of the OSN and NSO systems yield slightly smaller

pores compared to the NSN system.

Pore distribution plots can also be used to illustrate the

pore evolution with annealing teamperature. Figure 8 com-

pares the pore distribution of the OSO and NSN systems at

four different annealing temperatures. In the OSO system,

as the annealing temperature increases, the pore size dis-

tribution quickly shifts to larger diameters without signifi -

cantly increasing the total pore density. This indicates that

pore growth rather than new pore formation donimates the

pore evolution process in the OSO system. The NSN system,

however, shows a different pore evolution. With increasing

annealing tempreature, the pore distribution slowly shifts to

the large pores. At the same time, the density of relatively

small pores (<20 nm) remains high. This indicates that new

pore formation plays a big role in the pore evolution of the

NSN system at higher annealing temperatures.

In addition to the pore density and pore size, porosity

is another important characteristics. Figure 9 compares the

porosity among the four systems. The NSN system yields the

highest porosities at all temperatures. The porosity increases

rapidly with the annealing temperature and reaches 20%

when the NSN system is annealed at 1100 °C. In contrast,

the porosity in the OSO system stays the lowest among the

four systems at all annealing temperatures. It increases very

slowly when the annealing temperature is below 1000 °C.

Even at 1100 °C, the porosity is only 3.8%. The porosities in

the OSN and NSO systems start low at 800 °C and increase

quickly with annealing temperature. They stay between those

of the NSN and OSO systems at all annealing temperatures.

Consistent with our previous observation, the large differ-

ence in porosity between the NSN and OSO systems is due

to the pore density. At higher annealing temperatures, the

NSN system produces not only larger but denser pores while

only a limited number of larger pores are formed in the OSO

system. In other words, pore formation is enhanced in silicon

small 2014, DOI: 10.1002/smll.201303447

Figure 3. TEM images of the OSN system after RTP annealing at a) 800 °C, b) 900 °C, c) 1000 °C and d) 1100 °C.


5www.small-journal.com© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheimsmall 2014, DOI: 10.1002/smll.201303447

fi lms when sandwiched by silicon nitride layers compared to

silicon dioxide layers, yielding the highest pore density and

porosity. This highly porous Si membrane from the NSN

stack is very promising for fi ltration applications since a large

number of relatively big pores would enhance the fl uid fl ow

through the membrane.

Figure 4. TEM images of the NSO system after RTP annealing at a) 800 °C, b) 900 °C, c) 1000 °C and d) 1100 °C.

800 900 1000 11000

50

100

150

200

250

300

350

NSNNSOOSNOSO

Por

e de

nsity

(µ m

-2)

Temperature (°C)

Figure 5. Pore density for the four systems as a function of the annealing temperature. Each number is the average of three results from three different TEM images and the error bar represents the standard deviation.

800 900 1000 11000

5

10

15

20

25

30

35

NSNNSOOSNOSO

Ave

rage

por

e di

amet

er (

nm)

Temperature (°C)

Figure 6. Average pore diameter (APD) for the four systems as a function of the annealing temperature. The vertical bars represent the middle 50% of the pore size distribution in the OSO system at each temperature. The width of the distributions increases with temperature for all four systems.

C. Qi et al.


full papers

small 2014, DOI: 10.1002/smll.201303447

In all four systems the pore density, size and porosity

increase with the annealing temperature, indicating that

nanopore formation is thermally driven and closely related

to crystallization in the silicon fi lm. It has been reported

that the type of interface would infl uence the crystallization

behavior of amorphous silicon. [ 22,23 ] As the NSN and OSO

systems have different interfaces, it is likely that the crystal-

lization of the amorphous silicon fi lms would be affected by

different interfaces in these two systems. A possible explana-

tion for the pore formation enhancement in the NSN system

might be found in the interface energy difference between

silicon/silicon nitride and silicon/silicon dioxide. [ 23–26 ] Indeed,

our work has constantly shown that pore formation is asso-

ciated with Si nanocrystal formation. Further investigation

is needed to fully understand the pore formation process in

these two different systems.

3. Conclusion

This work reported for the fi rst time that nanopores can

be formed in an amorphous silicon fi lm sandwiched by

two silicon nitride fi lms during a rapid thermal annealing.

This discovery introduces a new control variable for pnc-Si

membrane fabrication, demonstrating new properties in the

expanded NSN materials system. Four different systems

including OSO, NSN, OSN and NSO have been tested. It was

found that the NSN system produces the highest porosity and

pore density after annealing, while the OSO system produces

the lowest. The porosity and pore density of the mixed OSN

and NSO systems fall between those with matched barrier

0 10 20 30 40 50 600

10

20

300

10

20

30

0

10

20

30

40

500

2

4

6

0 10 20 30 40 50 60D

ensi

ty (

µm-2)

Diameter (nm)

NSN

Den

sity

(µ m

-2) NSO

Den

sity

(µm

-2) OSN

Den

sity

(µm

-2) OSO

Figure 7. Pore distribution comparison of the NSN, NSO, OSN and OSO systems after annealing at 1100 °C. Note that the vertical scales on the four panels are different.

0

2

4

0

1

20

1

2

3

40

2

4

6

Den

sity

(µ m

-2)

Den

sity

(µm

-2)

Den

sity

(µ m

-2)

Den

sity

(µ m

-2)

0 10 20 30 40 50 600

10

20

30

400

10

20

30

40

0

10

20

30

400

10

20

30

0 10 20 30 40 50 60

Den

sity

(µm

-2)

Diameter (nm)

800°C

Den

sity

(µ m

-2) 900°C

Den

sity

(µm

-2) 1000°C

Den

sity

(µm

-2) 1100°C

Figure 8. Pore distribution comparison for the OSO and NSN system at four different annealing temperatures; the blue color represents the NSN system and the red color represents the OSO system. Note that the left and right vertical scales on the four panels are different.

800 900 1000 11000

3

6

9

12

15

18

21NSNNSOOSNOSO

Por

osity

(%

)

Temperature (°C)

Figure 9. Porosity for the four systems as a function of annealing temperature. Each number is the average of three results from three different TEM images and the error bar represents the standard deviation.


7www.small-journal.com© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheimsmall 2014, DOI: 10.1002/smll.201303447

layers. These differences may be due to the difference in

interfacial energy between silicon/silicon dioxide and silicon/

silicon nitride.

4. Experimental Section

Fabrication of pnc-Si membranes involves a bottom-up approach. First, the backside of (100) N-typc silicon wafers with 100 nm thick thermal oxide are patterned using standard photolithography tech-nique. Then four different stacks, including the NSN, OSN, NSO and OSO, are deposited on the front side of the patterned wafers using rf-magnetron sputtering (ATC-2000 V, AJA International, Inc., North Scituate, MA) respectively. Each stack is a three-layer struc-ture. The thickness of the silicon fi lm is kept at 25 nm and that of the nitride and oxide fi lms at 30 nm. The deposition temperature is 150 °C for all fi lms and the deposition pressure is 3mTorr for the amorphous Si and 5 mTorr for the oxide and nitride. Amorphous silicon is deposited under a 5 W bias while silicon dioxide and silicon nitride are deposited under a 50W bias.

After deposition the wafers are etched from the patterned side using a customized single-side Si etch cell with a wet anisotropic etchant, ethylene diamine pyrocatechol with pyrazine (EDP-300F, Transene Company, Inc., Danvers, MA) at 110 °C. The bottom sil-icon nitride or silicon dioxide works as an etch stop because EDP has very high silicon-to-oxide and silicon-to-nitride etch ratios. [ 27 ] A rapid thermal processor (RTP) is then used to induce solid phase crystallization (SPC) of the amorphous silicon layer (Solaris 150, Surface Science Integration, El Mirage, AZ). Samples are annealed at different temperatures ranging from 800 °C to 900 °C, 1000 °C, and 1100 °C respectively. The annealing process involves a fast ramp up rate of 50°C/s and a soaking time of 1 minute at the set temperature. Samples are placed inside a silicon carbide coated graphite susceptor to improve heating uniformity. We did not explore higher annealing temperatures because delamination of the fi lms was observed after annealing at 1200 °C. The protective silicon nitride and silicon dioxide fi lm are removed using HF after annealing. Figure 10 illustrates the process for the NSN system.

After removal of the silicon substrate and the protective layers, the free-standing pnc-Si membranes were characterized using a Hitachi 7650 transmission electron microscope (TEM) operating at 100 kV. The pnc-Si membranes were patterned to be compat-ible with the TEM specimen holder. Because the thickness of the membranes was only 25 nm, no additional sample preparation was required. Images were acquired with an Olympus Cantega 11 megapixel digital camera.

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements

The authors would like to thank D. Fang for fabrication assistance, J. DesOrmeaux and C. Chan of SiMPore for ellipsometry assis-tance; B. McIntyre and K. Bentley for microscopy assistance; M. Bindschadler’s work on the pore processing program (available for download at http://nanomembranes.org/resources/software); Silicon membrane fabrication processing was conducted at the University of Rochester Integrated Nanosystem Center (URnano) and the Semiconductor and Microsystems Fabrication Laboratory (SMFL) at the Rochester Institute of Technology. This work was sup-ported by the National Science Foundation (NSF CBET 1159579). As founders of SiMPore Inc., CCS, TRG, JLM, and PMF declare a competing fi nancial interest in this work.)

(100) Si waferpattern

sputter Si3N4/a-Si/Si3N4

etch bulk silicon

rapid thermal anneal strip nitride and oxide

free standing pnc-Simembrane

SiO2 single crystal Si amorphous Si

nanocrystal SiSi3N4

Figure 10. Schematic of the pnc-Si membrane fabrication process from the NSN system.

[1] C. C. Striemer , T. R. Gaborski , J. L. McGrath , P. M. Fauchet , Nature 2007 , 445 , 749 .

[2] A. van den Berg , M. Wessling , Nature 2007 , 445 , 726 . [3] H. D. Tong , H. V. Jansen , V. J. Gadgil , C. G. Bostan , E. Berenschot ,

C. J. M. van Rijn , M. Elwenspoek , Nano Lett. 2004 , 4 , 283 . [4] J. L. Snyder , A. Clark , D. Z. Fang , T. R. Gaborski , C. C. Striemer ,

P. M. Fauchet , J. L. McGrath , J. Membr. Sci. 2011 , 369 , 119 . [5] M. N. Kavalenka , C. C. Striemer , D. Z. Fang , T. R. Gaborski ,

J. L. McGrath , P. M. Fauchet , Nanotechnology 2012 , 23 , 145706 . [6] R. Ishimatsu , J. Kim , P. Jing , C. C. Striemer , D. Z. Fang ,

P. M. Fauchet , J. L. McGrath , S. Amemiya , Anal. Chem. 2010 , 82 , 7127 .

[7] T. R. Gaborski , J. L. Snyder , C. C. Striemer , D. Z. Fang , M. Hoffman , P. M. Fauchet , J. L. McGrath , ACS Nano 2010 , 4 , 6973 .

[8] A. A. Agrawal , B. J. Nehilla , K. V. Reisig , T. R. Gaborski , D. Z. Fang , C. C. Striemer , P. M. Fauchet , J. L. McGrath , Biomaterials 2010 , 31 , 5408 .

[9] M. N. Kavalenka , C. C. Striemer , J. P. S. DesOrmeaux , J. L. McGrath , P. M. Fauchet , Sens. Actuators, B 2012 , 162 , 22 .

[10] C. Qi , MRS Spring Meeting & Exhibit , San Francisco , April 2012 . [11] D. T. Danielson , D. K. Sparacin , J. Michel , L. C. Kimerling , J. Appl.

Phys. 2006 , 100 , 083507 . [12] S. Clarke , M. R. Wilby , D. D. Vvedensky , Surf. Sci. 1991 , 255 , 91 . [13] B. E. Weir , B. S. Freer , R. L. Headrick , D. J. Eaglesham , G. H. Gilmer ,

J. Bevk , L. C. Feldman , Appl. Phys. Lett. 1991 , 59 , 204 . [14] H. V. Balachandra Achar , S. Sengupta , E. Bhattacharya , IEEE Trans.

Nanotechnol. 2013 , 12 , 583 .

C. Qi et al.


full papers

Received: November 5, 2013 Revised: February 25, 2014Published online:

small 2014, DOI: 10.1002/smll.201303447

[15] H. V. Balachandra Achar , E. Bhattacharya , Int. J. Nanosci. 2011 , 10 , 793 .

[16] L. Shi , C. A. M. Steenbergen , A. H. deVreede , M. K. Smit , T. L. M. Scholtes , F. H. Groen , J. W. Pedersen , J. Vac. Sci. Technol. A 1996 , 14 , 471 .

[17] M. Modreanu , M. Gartner , J. Mol. Struct. 2001 , 565 , 519 . [18] S. Isomae , S. Yamamoto , S. Aoki , A. Yajima , IEEE Electron Device

Lett. 1986 , 7 , 368 . [19] W. Ting , H. Hwang , J. Lee , D. L. Kwong , Appl. Phys. Lett. 1990 , 57 ,

2808 . [20] M. D. Dange , J. Y. Lee , K. Sooriakumar , Microelectr. J. 2003 , 22 ,

19 . [21] D. Z. Fang , C. C. Striemer , T. R. Gaborski , J. L. McGrath ,

P. M. Fauchet , J. Phys. Condens. Matter 2010 , 22 , 454134 .

[22] O. Nast , A. J. Hartmann , J. Appl. Phys. 2000 , 88 , 716 . [23] M. Zacharias , P. Streitenberger , Phys. Rev. B 2000 , 62 , 8391 . [24] Y. Tu , J. Tersoff , Phys. Rev. Lett. 2000 , 84 , 4393 . [25] V. Misra , H. Lazar , M. Kulkarni , Z. Wang , G. Lucovsky , J. R. Hauser ,

Proc. Mat. Res. Soc. Symp. 1999 , 567 , 89 . [26] M. Nerding , S. Christiansen , R. Dassow , K. Taretto , J. R. Köhler ,

H. P. Strunk , Solid State Phenom. 2003 , 93 , 173 . [27] M. C. Elwenspoek , H. V. Jansen , Silicon Micromachining ,

Cambridge University Press , Cambridge , 1998

highly porous silicon membranes fabricated from...

Documents