Influence of surface properties on the electrical conductivity of silicon nanomembranes
© Zhao et al; licensee Springer. 2011
Received: 9 March 2011
Accepted: 31 May 2011
Published: 31 May 2011
Because of the large surface-to-volume ratio, the conductivity of semiconductor nanostructures is very sensitive to surface chemical and structural conditions. Two surface modifications, vacuum hydrogenation (VH) and hydrofluoric acid (HF) cleaning, of silicon nanomembranes (SiNMs) that nominally have the same effect, the hydrogen termination of the surface, are compared. The sheet resistance of the SiNMs, measured by the van der Pauw method, shows that HF etching produces at least an order of magnitude larger drop in sheet resistance than that caused by VH treatment, relative to the very high sheet resistance of samples terminated with native oxide. Re-oxidation rates after these treatments also differ. X-ray photoelectron spectroscopy measurements are consistent with the electrical-conductivity results. We pinpoint the likely cause of the differences.
PACS: 73.63.-b, 62.23.Kn, 73.40.Ty
Semiconductor nanomembranes (NMs), ultrathin layers of single-crystal semiconductor, can, because of the high surface-to-volume ratio, have electronic transport properties that are extremely sensitive to surface and interface conditions [1–3]. This surface sensitivity has a potential, so far not fully realized, for diverse applications, among others biological and chemical sensors [4–6], chemically gated transistors, or light-gated switches . This sensitivity highlights the need for a greater understanding of the influence of chemical modifiers, even quite mundane ones, on the near-surface electronic properties and thus the charge transport properties of the semiconductor.
We consider here a model system, silicon nanomembranes (SiNMs). SiNMs can be either freestanding or bonded to a host substrate. In the simplest form, they are bonded to a SiO2 layer on a bulk Si substrate, so-called silicon on insulator (SOI). Very thin SiNMs under ambient conditions, i.e., terminated with oxide on all sides, are highly resistive because interface states trap most of the free charges. Consider, for example, a 10-nm thick crystalline sheet of Si with a nominal doping density of 1015 cm-3. The sheet density of dopants in this thin membrane (109 cm-2) is much less than the typical trap density of even a high-quality Si/SiO2 interface (1011 cm-2 eV-1); thus, carrier depletion due to interface charge trapping renders the membrane effectively intrinsic. Consequently, any surface modification that significantly alters the trap density or provides free carriers by some other means (such as transfer doping via clean-surface states ) has a profound impact on the conductivity.
As an example of the extreme sensitivity of the conductivity of SiNMs to surface chemical condition, we describe in this paper the replacement of the surface oxide with a hydrogen termination [1, 6, 8–11] in two different ways. In one case, H-terminated SiNMs are prepared using surfaces cleaned in ultrahigh vacuum (UHV) that are dosed with pure hydrogen. The other approach uses the conventional HF wet etching of the oxide, which leaves the surface "H-terminated". Electrical transport properties are measured via the van der Pauw method in dry air (relative humidity <5%) and correlated with the surface chemical condition, determined from X-ray photoelectron spectroscopy (XPS). As suggested by prior work, the two treatments produce different surface chemistry [12–16]. The transport data demonstrate an extreme sensitivity of the conductivity to even trace surface chemical differences, a phenomenon totally absent in the bulk. The data lead us a step closer to an atomistic understanding of the complicated problem of the influence of surfaces on charge transport in semiconductor nanosystems.
The influence of the surface condition on silicon nanomembrane conductivity has been dramatically illustrated with scanning tunneling microscopy (STM) on Si(001). STM on clean bulk Si(001) produces high-quality images. Several attempts were made to image the surface of thin SOI(001) with STM, without success [17, 18]. Considering the above estimate of charge carrier density when interface states are present, STM should not, in fact, be possible because STM requires a reasonably conducting sample. When the Si surface is carefully cleaned in ultrahigh vacuum (UHV), however, revealing the signature Si(001) 2 × 1 reconstruction and its associated surface state bands, STM images can be readily obtained with a quality similar to that of bulk samples. When the sample is not so carefully cleaned, and the 2 × 1 reconstruction is weak or absent, imaging is not possible. The mechanism for conduction for the clean, reconstructed surface is ascribed to an interaction between the surface bands produced by the 2 × 1 reconstruction and the "bulk bands" in the membrane. A thermally activated charge transfer between the bulk and surface bands, called "surface transfer doping," produces enough carriers to create high conductivity .
If the clean-surface bands are disrupted, the STM image quality rapidly degrades. So it is conjectured that incomplete removal of the oxide in UHV (oxide removal without destroying the template Si layer is quite a difficult problem for SOI), or any other kind of surface disorder, prevents formation of the 2 × 1 surface bands . Additionally, the STM image degrades quickly if the clean surface of thin SOI (001) is exposed to H in UHV , while bulk Si(001) continues to be easily imaged . Because the dimer-row-created surface band structure on clean Si is rapidly destroyed by H , the mechanism for conduction in thin SiNMs is eliminated.
The observation of H dosing of thin SOI(001) under UHV conditions producing a rapid drop in STM image quality (and therefore a presumed large reduction in conductivity) is, however, in contrast to recent electrical measurements of HF-treated membranes, which feature a dramatic increase in conductivity . It is well known that HF etching of the oxide leaves a nominally H-terminated Si(001)surface, but with some residual F [12, 13].
The STM measurements are only an indirect measure of conductivity; one can calculate for a given STM and given operating conditions what the limits on sheet conductance are to obtain a good image . Whereas we are not able to duplicate here the exact conditions in UHV, because we at this stage cannot perform van der Pauw measurements in UHV, we prepare the vacuum H termination samples in a manner that permits dosing a nominally clean surface with clean H, as an intermediate step to the ultimate. We perform ex situ sheet resistance measurements and compare to H termination via HF etching for the same membrane thicknesses and sheet resistance measurement conditions.
P-type SOI(001) samples with nominal doping levels of 1015 cm-3 were patterned into 4 × 4 mm squares using photolithography and reactive-ion etching. Two different thicknesses of Si template layers (220 and 28 nm) were subjected to HF and vacuum H-dosing (VH) treatments. The choice of thicknesses is dictated by the earlier realization that 220 nm SiNMs behave close to bulk, while 28 nm SiMNs exhibit great surface sensitivity because the total dopant number is an order of magnitude smaller. The 28-nm samples were prepared with cycles of thermal oxidation and buffered oxide etching from the 220-nm commercial SOI wafers (Soitec, Peabody, MA, USA). From X-ray diffraction measurements of several samples diced from the same region of a wafer after thinning, we estimate that the thickness variation within each sample is less than 1 nm and the surface roughness, determined with atomic force microscopy, is typically 0.3 nm . For HF treatment, samples were pre-cleaned with acetone and isopropyl alcohol, followed by a deionized (DI) water rinse, then dipped into 49% HF solution for 1 min, followed by a 1-min DI water rinse, and finally dried with flowing N2. For the VH treatment, samples were cleaned with Piranha (H2SO4 + H2O2), AHP (H2O+NH4OH + H2O2), and dipped in 10% HF immediately prior to loading into the vacuum chamber. Once under vacuum (10-9 torr), the sample was flash heated to 900°C to remove the ex situ H termination and obtain the 2 × 1 reconstruction (as determined by reflection high-energy electron diffraction). Silane in a H2 carrier gas was then flowed across the sample for 30 s while it was maintained at 580°C and for an additional 2 min after turning off the heat source. It is known that this procedure produces a H-terminated surface , and indeed, the samples were hydrophobic after removing from vacuum.
Indium point contacts were soldered to the sample immediately after terminating with hydrogen in both preparations. The sheet resistances were measured by the van der Pauw method  with a semiconductor parameter analyzer (Agilent 4156C, Agilent Technologies, Palo Alto, CA, USA) in dry air (relative humidity, RH < 5%) to restrict the influence of humidity because water vapor plays an important role in the conductivity of low-doped thin SOI . The first sheet resistance measurement was always performed within 20 min of forming the H termination.
XPS measurements were performed with a PHI 5400 ESCA system (Physical Electronics, Division of Ulvac-PHI Chanhassen, MN USA) using Mg Kα X rays (1,254 eV). A larger take-off angle (45°) was chosen for survey spectra to obtain a stronger overall signal. A smaller take-off angle (15°), which brings more information from the surface, was selected for high-energy-resolution spectra to observe the change of the Si 2p core level peak in SiO2 to reveal the oxidation rate.
Results and discussion
Figure 1a shows the evolution of the sheet resistance of 220-nm Si membranes in dry air for the two surface modifications. The data points are averages of several samples. The first data points were measured no more than 20 min after the surface modifications, as mentioned above. The HF treatment reduces the sheet resistance by more than two orders of magnitude compared to samples with the native-oxide termination (gray band). The sheet resistance increases slowly with time as the surface oxidizes gradually to form electrically active interface traps that deplete free carriers. This behavior is consistent with the results in Ref .
Figure 1a also shows that the behavior of the sheet resistance after treating with pure hydrogen under vacuum (VH) is quite different from the HF-treated case. The sheet resistance rapidly reaches a value only slightly below the sheet resistance of samples with a native-oxide surface. At early times, the sheet resistance may be as much as an order of magnitude lower than the oxide-termination values, but these measurements are less reliable. Such changes in conductivity would not be observable with bulk samples because the surface-to-volume ratio is much smaller for bulk samples, and the conducting paths through the bulk would drown out any changes due to surface modification.
Although the conductivity of SiNMs is increased, relative to the oxide termination, for both H surface modifications, the differences observed for the two suggest structurally or chemically different surfaces. In fact, HF treatment results in a sheet resistance even lower than the value calculated for a NM with bulk doping (1015 cm-3) in the complete absence of interface traps, as has been shown earlier . It has been suggested that the cause is residual species such as F and OH  and their chemical action with time [12, 14, 23].
Similar results are obtained for the 28-nm SiNMs, as shown in Figure 1b. Comparison of Figure 1a,1b shows that thinner Si membranes are more sensitive to the surface modifications (a much higher sheet resistance for the oxidized surface, as expected from discussions earlier, and a dramatic drop in sheet resistance with H termination). HF treatment causes about three orders of magnitude drop in the sheet resistance, while VH treatment produces over two orders of magnitude drop, relative to the oxidized NM surface. Interestingly, the post-treatment values are not as different from each other for the two NM thicknesses as are the values for the oxidized surface. The implication is that both H treatments reduce or eliminate the oxide-induced interface states and instead introduce somewhat differing states that do not trap as much charge.
It is attractive to regard the change in sheet resistance in terms of re-oxidation of the surface, i.e., the re-formation of oxide interface states from the condition of the surface produced by H termination. The simplest situation for the H-terminated surface would be complete passivation of surface states by H (including the "poisoning" of the clean-surface bands produced by the 2 × 1 reconstruction) [2, 3]. If one adopts this model, then the evolution of sheet resistance plots the re-oxidation rate, and differences in re-oxidation rates of the NMs can be very sensitively inferred from the change of sheet resistance. There will then be a crossover point at 8 h where the degrees of oxidation of the two surface terminations switch. The VH termination appears then to be much more resistive to re-oxidation, i.e., it provides a much better surface chemical passivation than HF etching.
Semiconductor nanomembranes offer an excellent platform to investigate small changes in surface chemical and structural differences, if appropriate electronic transport measurements are made. The reason is that the density of mobile charge in thin sheets or ribbons, and all Si nanowires, is so low at conventional doping levels that any small modification in the charge density, no matter from what source, significantly influences the conductivity. Such changes in conductivity would not be observable with bulk samples because the surface-to-volume ratio is much smaller for bulk samples and the conducting paths through the bulk would overwhelm any changes due to surface modification. Especially the oxidized surface, because of interface states, has a very high resistivity at conventional doping levels, and for that reason, studies employing Si nanowires invariably use close to degeneratively doped material.
To demonstrate this surface sensitivity, we investigated two surface modifications on thin SiNMs that are nominally identical but differ in trace amounts of adsorbed species, H termination via HF etching and H termination via exposure of a clean surface to H in an UHV system. The difference in these surfaces appears to be a trace adsorption of F, although we cannot be sure that H occupies the same surface sites in both situations. The results confirm the extreme sensitivity of the conductance of thin semiconductor sheets to chemical changes and demonstrate how thinness plays a role. In addition, we investigated the oxidation rate via sheet resistance measurements as a function of time. Within the assumption that oxide-produced interface states replace the H passivation, the sheet resistance sensitively tracks the coverage of oxygen on the surface, to the extent that one observes up to four orders of magnitude change in sheet resistance (depending on NM thickness, NM bulk doping, and nature of the surface modification) for the replacement of one monolayer of atoms. It is expected that this sensitivity can be exploited in future sensor applications using nanomembranes. It is also clear that nanomembranes provide an excellent vehicle for investigating interface states and defects that control the conductivity. In particular, electrical measurements in UHV (currently underway) will enable a more detailed understanding of the mechanisms of transport modulation via surface and interface modification in very thin semiconductors. That field of inquiry is still in its very early stages.
This work was supported by DOE, grant no. DE-FG02-03ER46028. XFZ was supported by the China Scholarship Council (CSC). Facilities support from NSF/MRSEC is acknowledged.
- Scott SA, Peng WN, Kiefer AM, Jiang HQ, Knezevic I, Savage DE, Eriksson MA, Lagally MG: Influence of surface chemical modification on charge transport properties in ultrathin silicon membranes. ACS Nano 1999, 3: 1683–1692.View ArticleGoogle Scholar
- Zhang PP, Tevaarwerk E, Park BN, Savage DE, Celler GK, Knezevic I, Evans PG, Eriksson MA, Lagally MG: Electronic transport in nanometer-scale silicon-on-insulator membranes. Nature 2006, 439: 703–706. 10.1038/nature04501View ArticleGoogle Scholar
- Zhang PP, Nordberg E, Park BN, Knezevic I, Evans PG, Eriksson MA, Lagally MG: Electrical conductivity in silicon nanomembranes. N J Phys 2006, 8: 200. 10.1088/1367-2630/8/9/200View ArticleGoogle Scholar
- Cui Y, Wei Q, Park H, Lieber CM: Nanowire nanosensors for highly sensitive and selective detection of biological and chemical species. Science 2001, 293: 1289–1292. 10.1126/science.1062711View ArticleGoogle Scholar
- Nikolaides MG, Rauschenbach S, Bausch AR: Characterization of a silicon-on-insulator based thin film resistor in electrolyte solutions for sensor applications. J Appl Phys 2004, 95: 3811–3815. 10.1063/1.1650880View ArticleGoogle Scholar
- Ha S: Electrical properties of surface-modified silicon membranes. PhD thesis. University of Wisconsin-Madison, Physics; 2008.Google Scholar
- Werner R, Zimmermann C, Kalz A: Light dependence of partially depleted SOI-MOSFETs using SIMOX substrates. IEEE Trans Electron Devices 1995, 42: 1653–1656. 10.1109/16.405280View ArticleGoogle Scholar
- Dubey G, Lopinski GP, Rosei F: Influence of physisorbed water on the conductivity of hydrogen terminated silicon-on-insulator surfaces. Appl Phys Lett 2007, 91: 232111. 10.1063/1.2822417View ArticleGoogle Scholar
- Watanabe D, En A, Nakamura S, Suhara M, Okumura T: Anomalously large band-bending for HF-treated p-Si surfaces. Appl Surf Sci 2003, 216: 24–29. 10.1016/S0169-4332(03)00486-0View ArticleGoogle Scholar
- Schlaf R, Hinogami R, Fujitani M, Yae S, Nakato Y: Fermi level pinning on HF etched silicon surfaces investigated by photoelectron spectroscopy. J Vac Sci Technol A 1999, 17: 164–169. 10.1116/1.581568View ArticleGoogle Scholar
- Angermann H, Henrion W, Rebien M, Röseler A: Wet-chemical preparation and spectroscopic characterization of Si interfaces. Appl Surf Sci 2004, 235: 322–339. 10.1016/j.apsusc.2004.05.105View ArticleGoogle Scholar
- Takahagi T, Ishitani A, Kuroda H, Nagasawa Y: Fluorine-containing species on the hydrofluoric acid etched silicon single-crystal surface. J Appl Phys 1991, 69: 803–807. 10.1063/1.347367View ArticleGoogle Scholar
- Lee CH, Ritz CS, Huang MH, Ziwisky M, Blise R, Lagally MG: Wafer-scale integrated freestanding single-crystal silicon nanowires: conductivity and surface treatment. Nanotechnology 2011, 22: 055704–055710. 10.1088/0957-4484/22/5/055704View ArticleGoogle Scholar
- Gräf M, Grundner M, Schulz R: Reactions of water with hydrofluoric acid treated silicon (111) and (100) surfaces. J Vac Sci Technol A 1989, 7: 808–813. 10.1116/1.575845View ArticleGoogle Scholar
- Grundner M, Schulz R: The surface state of Si (100) and (111) wafers after treatment with hydrofluoric acid. AIP Conf Proc 1987, 167: 329–337.View ArticleGoogle Scholar
- Takahagi T, Ishitani A, Kuroda H, Nagasawa Y, Ito T, Wakao S: Fluorine-containing species on the hydrofluoric acid etched silicon single-crystal surface. J Appl Phys 1990, 68: 2187–2191. 10.1063/1.346521View ArticleGoogle Scholar
- Lin KC, Holland OW, Feldman LC, Weitering HH: Surface characterization of silicon on insulator material. Appl Phys Lett 1998, 72: 2313–2315. 10.1063/1.121346View ArticleGoogle Scholar
- Sutter P, Ernst W, Sutter E: Scanning tunneling microscopy on ultrathin silicon on insulator (100). Appl Phys Lett 2004, 85: 3148–3150. 10.1063/1.1806539View ArticleGoogle Scholar
- Zhang PP: Structure and electronic properties of nanometer-scale silicon membranes in silicon-on-insulator. PhD thesis. University of Wisconsin-Madison, Physics; 2006.Google Scholar
- Boland JJ: Structure of the H-saturated Si(001) surface. Phys Rev Lett 1990, 65: 3325–3328. 10.1103/PhysRevLett.65.3325View ArticleGoogle Scholar
- Liehr M, Greenlief CM, Offenberg M, Kasi SR: Equilibrium surface hydrogen coverage during silicon epitaxy using SiH4. J Vac Sci & Technol A 1990, 8: 2960–2964. 10.1116/1.576613View ArticleGoogle Scholar
- van der Pauw LJ: A method of measuring specific resistivity and Hall effect of discs of arbitrary shape. Philips Res Repts 1958, 13: 1–9.Google Scholar
- Ling L, Kuwabara S, Abe T, Shimura F: Multiple internal reflection infrared spectroscopy of silicon surface structure and oxidation process at room temperature. J Appl Phys 1993, 73: 3018–3022. 10.1063/1.353007View ArticleGoogle Scholar
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/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.