# Dielectric properties of porous silicon for use as a substrate for the on-chip integration of millimeter-wave devices in the frequency range 140 to 210 GHz

- Panagiotis Sarafis
^{1}and - Androula Galiouna Nassiopoulou
^{1}Email author

**9**:418

**DOI: **10.1186/1556-276X-9-418

© Sarafis and Nassiopoulou; licensee Springer. 2014

**Received: **9 May 2014

**Accepted: **15 July 2014

**Published: **21 August 2014

## Abstract

In this work, the dielectric properties of porous Si for its use as a local substrate material for the integration on the Si wafer of millimeter-wave devices were investigated in the frequency range 140 to 210 GHz. Broadband electrical characterization of coplanar waveguide transmission lines (CPW TLines), formed on the porous Si layer, was used in this respect. It was shown that the dielectric parameters of porous Si (dielectric permittivity and loss tangent) in the above frequency range have values similar to those obtained at lower frequencies (1 to 40 GHz). More specifically, for the samples used, the obtained values were approximately 3.12 ± 0.05 and 0.023 ± 0.005, respectively. Finally, a comparison was made between the performance of the CPW TLines on a 150-μm-thick porous Si layer and on three other radiofrequency (RF) substrates, namely, on trap-rich high-resistivity Si (trap-rich HR Si), on a standard complementary metal-oxide-semiconductor (CMOS) Si wafer (p-type, resistivity 1 to 10 Ω.cm) and on quartz.

### PACS

84.40.-x; 77.22.Ch; 81.05.Rm

### Keywords

Porous Si Nanoscale semiconductors RF characterization Dielectric permittivity Millimeter-wave passive devices## Background

The co-integration of radiofrequency (RF) and millimeter-wave passive devices with complementary metal-oxide-semiconductor (CMOS) circuitry is quite challenging due to the low resistivity of the CMOS Si substrate that introduces important losses during electromagnetic wave propagation (eddy currents into the substrate). Another drawback towards this integration is the high permittivity of Si (*ε*_{
r,Si
} = 11.7) that causes an increase in crosstalk between lines, a decrease in antenna efficiency, and a reduction of the frequency of operation of the inductors. A viable solution recently investigated towards this integration is the formation of a local substrate with the appropriate dielectric properties on the Si wafer, on which the RF and millimeter-wave devices will be integrated. Such a substrate is a thick porous Si layer with high porosity, which can be optimized for best device performance by choosing the appropriate layer thickness, in order to minimize electromagnetic propagation losses into Si, and the appropriate low values of the dielectric permittivity, *ε*_{
r
}, and loss tangent. These last values are tunable by changing the material structure and morphology [1–6].

Porous Si structure (pore size, inter-pore distance) and morphology affect all its macroscopic properties (electrical, mechanical, optical, etc.) [7]. An intensive effort was made in the literature to correlate the electrical properties of the material with its structural parameters [8–12]. In view of the application of porous Si for the on-chip integration of RF and millimeter-wave devices, its dielectric properties (dielectric permittivity and loss tangent) as a function of frequency should be known, in order to be used by the device designer for an accurate prediction of device operation. In addition, since the dielectric properties of the material depend strongly on its structure and morphology [13], it is desirable to have an experimental method to extract the dielectric parameters of the specific material used in each application.

In this work, we will first discuss the existing models that correlate the structural properties of porous Si (porosity and morphology) with its dielectric properties and we will compare them with results obtained by a broadband extraction method, based on the measurement of the S-parameters of coplanar waveguide transmission lines (CPW TLines) integrated on the porous Si substrate. By combining these measurements with electromagnetic simulations, the dielectric permittivity and loss tangent of the substrate (porous Si) can be obtained. This method has been previously used by the authors to extract the dielectric parameters of porous Si in the frequency range 1 to 40 GHz [13, 14]. In this work, measurements are extended to the frequency range 140 to 210 GHz. Finally, by comparing the performance of CPW TLines on porous Si and three other substrates used in RF, namely, a trap-rich high-resistivity (HR) Si substrate [15–17], a standard CMOS Si wafer (p-type, resistivity 1 to 10 Ω.cm), and a quartz substrate, we demonstrate the superiority of porous Si as a local substrate for RF and millimeter-wave on-chip device integration.

## Methods

### Fabrication of CPW TLines on porous Si layers

*ρ*= 1 to 5 mΩ.cm). The electrolytic solution was a mixture of HF and ethanol (3 EtOH(99.9%)/2 HF(50%) v.v.) and the anodization current density was

*J*= 20 mA/cm

^{2}. The resulting layer had a porosity of 76% and a dendritic structure as presented in Figure 1. The porous Si layer was capped with 500 nm SiO

_{2}in order to stabilize it over time and achieve better planarization of the porous Si surface for further processing. On top of PSi, covered by SiO

_{2}, a set of coplanar waveguide transmission lines (CPW TLines), made of 1-μm-thick patterned Al, was integrated (see Figure 2).

For comparison, identical CPW TLines were also fabricated on three other substrates, as follows: the first was the state-of-the-art trap-rich high-resistivity (HR) Si RF substrate [15]. This substrate was an n-type HR-Si wafer with nominal resistivity higher than 10 kΩ.cm, covered by a bilayer of a 500-nm-thick trap-rich poly-Si layer, deposited by low-pressure chemical vapor deposition (LPCVD) at 625°C, and a-500 nm-thick TEOS SiO_{2} layer. The trap-rich layer is used to minimize the parasitic surface conduction within the Si layer underneath the silicon oxide by trapping the parasitic charges and thus restoring the initial high resistivity of the Si substrate [17]. The second substrate was a 380-μm-thick standard Si wafer used in CMOS-integrated circuits (ICs) (p-type, resistivity 1 to 10 Ω.cm). Finally, the last substrate was a 500-μm-thick quartz substrate, which is one of the off-chip RF substrates with almost negligible losses. This last substrate was used for comparison with the three other Si-based substrates.

### RF measurements and de-embedding

The S-parameters of the CPW TLines were measured in the 140-to-210-GHz range with an HP 8510B vector network analyzer (VNA) from Agilent (Santa Clara, CA, USA), combined with a millimeter-wave VNA extension module by Oleson Microwave Labs (Morgan Hill, CA, USA). All the measurements were calibrated using the Line-Reflect-Reflect-Match (LRRM) algorithm of the WinCal software from Cascade Microtech (Beaverton, OR, USA). A de-embedding procedure is always necessary in order to decouple the device response from the parasitics due to the contacts and pads. The method followed was the two-line method, using the measured S-parameters of two lines with different length (8 mm and 500 μm) [18]. The characteristic impedance, the propagation constant, and the effective permittivity of the lines were extracted and used to derive the dielectric parameters of the PSi substrate [13, 14].

### Microscopic models for determining PSi dielectric properties

Porous Si structure and morphology depend on the electrochemical conditions used for its formation as well as on the starting wafer resistivity. Its dielectric properties are highly dependent on its structure and morphology. There are several works in the literature that correlate the material structure with its dielectric properties. According to [9, 21, 22], the ac electrical transport of porous Si follows two mechanisms. The first is limited by the length of the carrier random walk through the fractal structure of the material and is valid in the very low frequency range, while at higher frequencies, the random path is shorter and the hopping length stops to be the critical factor. In that case, conduction is mainly determined by the distance between inhomogeneous areas [22]. The dielectric permittivity of porous Si (*ε*_{
PSi
}) describes the polarization of the atoms and the impurities inside the material. As it is shown in [22], *ε*_{
PSi
} depends on frequency only for frequencies <100 Hz. For higher frequencies, its value is saturated and remains constant up to at least 100 kHz. This value is also independent of temperature.

### Effective medium approximation

Instead of considering carrier transport at the nanoscale, it is easier to correlate the dielectric properties of the material with macroscopic parameters, as for example the material porosity. An effective medium approximation (EMA) is used in this respect. This approximation is valid when the wavelength of the signal is much larger than the typical dimensions of pores and nanostructures composing the material. The most common models used in the literature to correlate the permittivity of non-oxidized porous Si with its porosity are the following:

*Vegard’s approximation*[1]

where *ε*_{
PSi
} is the permittivity of porous Si, *ε*_{
air
} is the permittivity of air, and *ε*_{
Si
} is the permittivity of Si and P is the porosity.

*Maxwell-Garnett’s approximation*[23]

*f*(where

*f*= 1 -

*P*) of the porous material is far smaller than the porosity (

*P*) [23]. The following expression is obtained:

*Bruggeman’s approximation*[23].

*Bergman’s approximation*[24]

*g(n,P)*to take into account the nanotopology of the material. The following expression is obtained:

From all the above models, Vegard’s approximation is the simplest one. The most commonly used model is the Bruggeman’s model [11, 25]. Both the Vegard’s model for non-oxidized Si and the Maxwell-Garnett’s model have been proven to be insufficient to explain the results of several experiments [13, 24, 26]. An improved version of the Vegard’s model incorporates also the SiO_{2} native oxide surrounding the Si nanostructures composing the material [27]. Better agreement between the model and experimental results is obtained in this case. The oxidation of the Si skeleton leads to a decreased permittivity of the material [11, 27]. This is because the oxidation not only changes material composition, but also leads to reduction of material porosity. Finally, the Bergman’s approximation predicts quite well the dielectric behavior of PSi in the optical frequencies. The spectral density function *g(n,f)* that describes the micro-topology of the material has to be extracted in this respect [12].

### Dielectric parameter extraction using broadband electrical measurements

The models of Vegard, Maxwell-Garnett, and Bruggeman, as presented above, relate *ε*_{
PSi
} with material porosity. However, they were insufficient to explain the experimental results of several groups [13, 26, 28]. This can be attributed to the complexity of the PSi structure and morphology, which differs from one sample to another, even if the macroscopic porosity is the same. It is also quite difficult to find a representative *g(n,f)* function that accurately describes the specific porous Si structure and morphology in each case, making the Bergman’s model difficult to use. It is thus interesting to have a method to characterize the specific PSi material used in each experiment by simply measuring the electromagnetic (EM) response of a device integrated on it. This is the approach we use in this work. By integrating CPW TLines on top of porous Si and measuring their S-parameters, we extract porous Si dielectric parameters by combining the experimental results with electromagnetic simulations and conformal mapping calculations. This method has been described in detail in [13, 14], and the results have been proven to be in very good agreement with full-wave EM simulations [14].

^{+}-type Si wafer with resistivity 1 to 5 mΩ.cm and had a surface area of 4 cm

^{2}. Identical transmission lines were integrated on all three samples (see Figure 2b). The obtained results were compared with those obtained using Vegard’s, Maxwell-Garnett’s and Bruggeman’s models for PSi by applying formulas (1) to (3) given above. From Figure 4, it can be seen that the values of the extracted permittivity using broadband electrical measurements of the specific CPW TLines are between those obtained with the Bruggeman’s and Vegard’s models for non-oxidized PSi. On the other hand, by using the more elaborated Vegard’s law described in [27], which takes into account the presence of a native oxide shell surrounding the Si nanostructures (in our case, we considered a native oxide thickness of 1.5 nm and a Si skeleton thickness of 10 nm), better agreement is achieved between our experimental results and the calculated ones.

## Results and discussion

### Porous Si dielectric parameters in the frequency range 140 to 210 GHz

*ε*

_{ PSi }from 3.19 to 3.12 and it then stabilizes around this value (Figure 5a). In the high-frequency range (140 to 210 GHz),

*ε*

_{ PSi }oscillates around the values of 3.1 and 3.2, within a maximum deviation of 0.1. Similarly, the value of the loss tangent is between 0.031 and 0.023 in the range 5 to 40 GHz (see Figure 5b), while it stays constant at 0.023 in the range 140 to 210 GHz, with a maximum deviation of 0.005. In the 1-to-5-GHz range, the results are not reliable due to the contact resistance between the RF probes and the pads and were omitted. This effect becomes negligible at higher frequencies.

In overall, from the above, we can deduce that the dielectric permittivity of porous Si is almost constant in the studied frequency ranges. It also shows a continuity of the two curves, suggesting the same constant value in the frequency range 40 to 140 GHz. The loss tangent shows a slight decrease with frequency, while again there is continuity between the low- and high-frequency curves.

### Comparison of PSi with other RF and millimeter-wave substrates

The observed reduction of signal attenuation *a* and the increase of the quality factor *Q* of the CPW TLine on PSi versus bulk Si is attributed to the reduction of the material loss tangent and dielectric permittivity through nanostructuring. As shown previously by the authors, the achieved low permittivity of porous Si at high porosities shows advantages in many RF and millimeter-wave devices, namely, high-characteristic impedance of the CPW TLines [5], inductors operating at higher frequencies [29, 30] and antennas with reduced surface waves induced into the substrate can be obtained.

## Conclusions

The dielectric properties of a 150-μm-thick PSi layer with 76% porosity were presented for the first time for the frequency range 140 to 210 GHz, using broadband electrical measurements combined with electromagnetic simulations. It was found that the dielectric permittivity is almost constant in the above frequency range, having approximately the same value as at lower frequencies. The loss tangent is also almost constant with frequency. Finally, a comparison between the performance of CPW TLines on PSi, trap-rich HR Si, quartz, and standard low-resistivity CMOS Si was made in the above frequency range. An almost equal performance was obtained between the trap-rich HR Si, PSi, and quartz. At 210 GHz, porous Si showed an attenuation as low as 1 dB/mm and the quality factor was ~30. This performance is added to the other advantages of PSi compared to other Si-based substrates, e.g., its compatibility with the low-resistivity CMOS substrate (permitting co-integration of CMOS logic with RF and millimeter-wave devices on the same substrate) and its low achievable permittivity). All the above make PSi an excellent local substrate on the Si wafer for RF and millimeter-wave device integration on the Si chip, paving the way towards the digital/RF analog system-on-chip (SoC) of the future.

## Declarations

### Acknowledgements

The trap-rich high-resistivity Si wafers were provided by UCL Belgium (Jean-Pierre Raskin), while measurements in the frequency range 140 to 210 GHz of the CPW TLines were conducted in the facilities of VTT, Helsinki, Finland (arranged by A. Markus) during a visit of P. Sarafis to VTT. This work was supported by the EU Network of Excellence ‘Nanofunction’ through the EU 7th Framework Programme for Research under Contract 257375.

## Authors’ Affiliations

## References

- Kim H-S, Xie Y-H, DeVincentis M, Itoh T, Jenkins K a: Unoxidized porous Si as an isolation material for mixed-signal integrated circuit applications.
*J Appl Phys*2003, 93: 4226. 10.1063/1.1555700View ArticleGoogle Scholar - Welty R, Park S, Asbeck PM, Dancil K-PS, Sailor MJ: Porous silicon technology for RF integrated circuit applications. In
*1998 Top. Meet. Silicon Monolith. Integr. Circuits RF Syst. Dig. Pap. (Cat. No.98EX271)*. IEEE; 1998:160–163.View ArticleGoogle Scholar - Gautier G, Leduc P: Porous silicon for electrical isolation in radio frequency devices: a review.
*Appl Phys Rev*2014, 1: 011101. 10.1063/1.4833575View ArticleGoogle Scholar - Capelle M, Billoué J, Poveda P, Gautier G: RF performances of inductors integrated on localized p + -type porous silicon regions.
*Nanoscale Res Lett*2012, 7: 523. 10.1186/1556-276X-7-523View ArticleGoogle Scholar - Issa H, Ferrari P, Hourdakis E, Nassiopoulou AG: On-chip high-performance millimeter-wave transmission lines on locally grown porous silicon areas.
*IEEE Trans Electron Devices*2011, 58: 3720–3724.View ArticleGoogle Scholar - Capelle M, Billoué J, Poveda P, Gautier G: Study of porous silicon substrates for the monolithic integration of radiofrequency circuits.
*Int J Microw Wirel Technol*2013, 6: 39–43.View ArticleGoogle Scholar *"Properties of porous silicon" Emis Datareviews. Ser. No18, IEE, an INSPEC Publ.UK, edited by L.T.Canham. 1997*. 1997.Google Scholar- Di Francia G, La Ferrara V, Maddalena P, Ninno D, Odierna LP, Cataudella V: AC conductivity of porous silicon: a fractal and surface transport mechanism.
*Il Nuovo Cimento D*1996, 18: 1187–1196. 10.1007/BF02464696View ArticleGoogle Scholar - Lehmann V, Hofmann F, Möller F, Grüning U: Resistivity of porous silicon: a surface effect.
*Thin Solid Films*1995, 255: 20–22. 10.1016/0040-6090(94)05624-MView ArticleGoogle Scholar - Boarino L, Borini S, Amato G: Electrical properties of mesoporous silicon: from a surface effect to coulomb blockade and more.
*J Electrochem Soc*2009, 156: K223. 10.1149/1.3232202View ArticleGoogle Scholar - Astrova E, Tolmachev V: Effective refractive index and composition of oxidized porous silicon films.
*Mater Sci Eng B*2000, 69–70: 142–148.View ArticleGoogle Scholar - Theiβ W: The dielectric function of porous silicon - how to obtain it and how to use it.
*Thin Solid Films*1996, 276: 7–12. 10.1016/0040-6090(95)08036-8View ArticleGoogle Scholar - Sarafis P, Hourdakis E, Nassiopoulou AG: Dielectric permittivity of Porous Si for use as substrate material in Si-integrated RF devices.
*IEEE Trans Electron Devices*2013, 60: 1436–1443.View ArticleGoogle Scholar - Sarafis P, Hourdakis E, Nassiopoulou AG: Porous Si dielectric parameter extraction for use in RF passive device integration : Measurements and simulations.
*43rd Eur. Solid-State Device Res. Conf. Bucharest*2013, 99–102.Google Scholar - Neve CR, Ben Alia K, Malaquin C, Allibert F, Desbonnets E, Bertrand I, Van Den Daele W, Raskin J-P: RF and linear performance of commercial 200 mm trap-rich HR-SOI wafers for SoC applications. In
*2013 IEEE 13th Top. Meet. Silicon Monolith. Integr. Circuits RF Syst*. Austin, TX, USA: IEEE; 2013:15–17.View ArticleGoogle Scholar - Ben AK, Neve CR, Gharsallah A, Raskin J: RF performance of SOI CMOS technology on commercial 200-mm enhanced signal integrity high resistivity SOI substrate.
*IEEE Trans Electron Devices*2014, 61: 722–728.View ArticleGoogle Scholar - Lederer D, Raskin J-P: Effective resistivity of fully-processed SOI substrates.
*Solid State Electron*2005, 49: 491–496. 10.1016/j.sse.2004.12.003View ArticleGoogle Scholar - Mangan AM, Voinigescu SP, Tazlauanu M: De-embedding transmission line measurements for accurate modeling of IC designs.
*IEEE Trans Electron Devices*2006, 53: 235–241.View ArticleGoogle Scholar - Lamb JW: Miscellaneous data on materials for millimetre and submillimetre optics.
*Int J Infrared Millimeter Waves*1996, 17: 1997–2034. 10.1007/BF02069487View ArticleGoogle Scholar - Cheng H-J, Whitaker JF, Weller TM, Katehi LPB: Terahertz-bandwidth characteristics of coplanar transmission lines on low permittivity substrates.
*IEEE Trans Microw Theory Tech*1994, 42: 2399–2406. 10.1109/22.339773View ArticleGoogle Scholar - Ben-Chorin M, Möller F, Koch F: Nonlinear electrical transport in porous silicon.
*Phys Rev B*1994, 49: 2981–2984. 10.1103/PhysRevB.49.2981View ArticleGoogle Scholar - Ben-Chorin M, Möller F, Koch F, Schirmacher W, Eberhard M: Hopping transport on a fractal: ac conductivity of porous silicon.
*Phys Rev B*1995, 51: 2199–2213.View ArticleGoogle Scholar - Campos AM, Torres J, Giraldo JJ: Porous silicon dielectric function modeling from effective medium theories.
*Surf Rev Lett*2002, 09: 1631–1635. 10.1142/S0218625X02004116View ArticleGoogle Scholar - Theiβ W, Henkel S, Arntzen M: Connecting microscopic and macroscopic properties of porous media: choosing appropriate effective medium concepts.
*Thin Solid Films*1995, 255: 177–180. 10.1016/0040-6090(94)05649-XView ArticleGoogle Scholar - Khardani M, Bouaïcha M, Bessaïs B: Bruggeman effective medium approach for modelling optical properties of porous silicon: comparison with experiment.
*Phys Status Solidi*2007, 4: 1986–1990. 10.1002/pssc.200674420View ArticleGoogle Scholar - Ramani S, Cheville A, Escorcia Garcia J, Agarwal V: Conductivity of free-standing porous silicon layers using Terahertz differential time-domain spectroscopy.
*Phys Status Solidi*2007, 4: 2111–2115. 10.1002/pssc.200674393View ArticleGoogle Scholar - Theodoropoulou M, Pagonis DN, Nassiopoulou AG, Krontiras CA, Georga SN: Dielectric characterization of macroporous thick silicon films in the frequency range 1 Hz-1 MHz.
*Phys Status Solidi*2008, 5: 3597–3600. 10.1002/pssc.200780153View ArticleGoogle Scholar - Menard S, Fevre A, Capelle M, Defforge T, Billoue J, Gautier G: Dielectric behaviour of porous silicon grown from p-type substrates. In
*Int. Conf. Porous Semicond. - Sci. Technol, 0*. Benidorm-Alicante; 2014:122–123.Google Scholar - Sarafis P, Hourdakis E, Nassiopoulou AG, Roda Neve C, Ben Ali K, Raskin J-P: Advanced Si-based substrates for RF passive integration: comparison between local porous Si layer technology and trap-rich high resistivity Si.
*Solid State Electron*2013, 87: 27–33.View ArticleGoogle Scholar - Capelle M, Billoue J, Poveda P, Gautier G: N-type porous silicon substrates for integrated RF inductors.
*IEEE Trans Electron Devices*2011, 58: 4111–4114.View ArticleGoogle Scholar

## Copyright

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.