www.advmat.de www.MaterialsViews.

com

COMMUNICATION

Local Doping of Silicon Using Nanoimprint Lithography and Molecular Monolayers

W. Pim Voorthuijzen, M. Deniz Yilmaz, Wouter J. M. Naber, Jurriaan Huskens,* and Wilfred G. van der Wiel*
A vast amount of semiconductor technology is focused on downscaling electronic components according to Moores Law.[1] Silicons semiconducting properties largely depend on the concentration of dopant impurities.[2] Therefore, novel methodologies of selectively introducing dopants to semiconductor materials at the nanometer scale could be benecial for the development of smaller integrated circuit components. Conventionally, introducing dopant impurities to silicon is achieved using ion implantation.[3] In this process, an ion beam is used to bombard the surface with highly energetic dopant ions, literally shooting the dopants into the top layers of the bulk silicon. True electrical activation is then achieved using a hightemperature annealing step. This fabrication step is required for repairing the crystal damage inicted by bombardment, followed by diffusion of (near) surface-located dopant atoms. Fabrication of ultrashallow, heavily doped junctions, including well-dened insulating areas, will be required for reliably manufacturing smaller scale devices.[1,4,5] In this respect, the ability for true deterministic positioning, i.e., control over the exact position of impurity atoms, in semiconductor structures will become essential. A possible methodology would involve control over lateral (xy) positioning, concomitant with considerate inuence on the penetration depth of dopant diffusion in the z-direction from the surface. In 2008, the group of Javey published a novel method for introducing dopants using molecular monolayers on oxide-free silicon.[6] Using this method, the authors showed it was possible to obtain high-level doping of silicon by application of hydrosilylation surface chemistry on SiH terminated silicon[7,8] with dopant-atom-containing organic monolayers. Rapid thermal annealing (RTA) ensures implantation of dopant atoms by diffusion. Using RTA, this method eliminates the relatively long annealing steps required to repair damage to the silicon crystal structure. Therefore, a rst requirement in deterministic positioning of dopant atoms is met, in the sense that using RTA dopant atoms mainly end up in the top 100 nm of the silicon substrate. Another benet of the aforementioned strategy is the use of a limited source, such as a molecular monolayer, which allows specic tuning of the dopant surface dose.[6] Moreover, molecular (self-)assembly is a cheap bottom-up nanofabrication method.[9] Introducing dopants by molecular monolayers in principle offers considerable advantages. However, actual device fabrication necessitates a combination of the approach of Javey et al. with the ability to selectively pattern, and therefore control, the lateral (xy) positioning of molecular monolayer structures. Recent progress aimed at a patterning procedure for organic monolayers on oxide-free silicon.[10] Monolayer patterns on oxide-free silicon from the nanometer- to micrometer-scale using top-down nanoimprint lithography were demonstrated. In this work, we combine the aforementioned patterning on oxide-free silicon with doping by molecular monolayer formation. Using a phosphorus-containing organic precursor, highly doped and patterned (micrometer-scale) regions in nearly intrinsic silicon were successfully fabricated. These patterned regions were characterized by time-of-ight secondary ion mass spectrometry (TOF-SIMS) and by Hall and sheet resistance measurements. The approach starts with a piranha-cleaned Si(100) sample. In the rst step (a) of the process (Scheme 1), the sample is fully stripped from its native oxide by NH4F, leaving a hydrogen-terminated (SiH) surface. In step (b), a full organic monolayer is formed by (thermal) hydrosilylation chemistry using an adsorbate with an alkene moiety for anchoring to the substrate and a dopant atom (P) for local doping of the underlying substrate in the subsequent steps. Next, imprint resist is spin-coated on the preformed monolayer (c). The sample is patterned using nanoimprint lithography (NIL) in step (d) (Figure S1, Supporting Information), creating micrometersized patterns suitable for inspection by SIMS imaging. At the same time, using NIL holds the promise for straightforward extension of the process to the sub-100 nm range by using high-resolution molds.[10] Reactive ion etching (RIE) using an O2 plasma is used to remove the residual layer and the monolayer underneath in step (e). Resist removal by acetone in step (f) removes residual polymer, which results in pattern transfer. Further sample processing (steps (g) and (h)) proceed according to the procedure reported by Javey et al.[6] The capping step (g) is used for prevention of loss of the organic monolayer during the subsequent RTA step. Diffusion of the dopant impurities is

W. P. Voorthuijzen, M. D. Yilmaz, Prof. J. Huskens Molecular Nanofabrication Group MESA+ Institute for Nanotechnology University of Twente P.O. Box 217, 7500 AE Enschede, The Netherlands E-mail: j.huskens@utwente.nl W. P. Voorthuijzen, Dr. W. J. M. Naber, Prof. W. G. van der Wiel NanoElectronics Group MESA+ Institute for Nanotechnology University of Twente P.O. Box 217, 7500 AE Enschede, The Netherlands E-mail: w.g.vanderWiel@utwente.nl

DOI: 10.1002/adma.201003625

1346

wileyonlinelibrary.com

2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Adv. Mater. 2011, 23, 13461350

www.advmat.de www.MaterialsViews.com

COMMUNICATION
Scheme 1. a) Native SiO2 is fully stripped by NH4F. b) Monolayer formation using an organic-dopant-atom-containing molecular precursor (1undecenyl-dimethylphosphonate). c) Spin-coating of the imprint resist. d) NIL. e) RIE of the residual layer and local removal of the monolayer using O2 plasma with concomitant oxidation of the exposed Si areas. f) Resist removal by ultrasonication in acetone. g) Deposition of a SiO2 capping layer by electron beam (e-beam) evaporation. h) Rapid thermal annealing (RTA) to distribute the dopant atoms.

achieved by RTA (step (h)) for 5 min at 1000 C. At this temperature the organic monolayer (1-undecenyl-dimethylphosphonate) fully disintegrates and the phosphorus dopant atoms diffuse into the silicon. The specic route[10] in Scheme 1 was selected to exclude formation of monolayers on SiO2-covered areas by reaction with the phosphonate capping group.[11] To analyze the distribution of P dopant atoms in our patterned sample, TOF-SIMS was used for imaging and depth proling (3D). In Figure 1, each image represents a 500 500 m2 area and qualitatively shows relative P concentrations. Integration of these images resulted in lateral line proles, which are shown in Figure 2. It should be noted that the P concentrations are given relative to Si and that the apparent background concentrations observed in the unpatterned regions arise from this comparison (most likely due to relatively high O concentrations and thus concomitantly lower Si concentrations in the samples, which correspond to depths close to the interface). We expect that the real P concentrations in the unpatterned regions are low, equal for all samples, and come from inherent impurities of the Si samples used, but we did not attempt to correct for this in the presentation of the data. An alternative explanation could be that a fraction of the dopant atoms spread homogeneously through the capping layer before diffusing evenly into the

Figure 1. Phosphor elemental images recorded with TOF-SIMS (normalized to Si, image sizes 500 500 m2) of a 1 1 cm2 intrinsic Si(100) sample containing 100-m-wide phosphorus-doped regions at 200-m period (RTA, 5 min at 1000 C) at successive depths: each image represents an interval of 10 nm (010 for image I, 1020 for image II, etc.). The coloring has been done articially and represents increasing doping concentrations from black to blue to green to yellow to red.

Adv. Mater. 2011, 23, 13461350

2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

wileyonlinelibrary.com

1347

www.advmat.de www.MaterialsViews.com

COMMUNICATION
1348

of the majority carrier, density is averaged over doped and undoped areas on the whole sample, while SIMS depth proling can be performed on specic locations. As the ratio of doped to undoped areas is 1:1, an estimation for the sheet density on the doped areas of the sample can be obtained by multiplication of the average value by a factor of 2 which results in ns,full = 3.8 1013 0.1 1013 cm2. At T = 150 K an average sheet density ns,av = 9.4 1012 0.1 1012 cm2 was determined from the curve in Figure S4b (Supporting Information), which results in ns,full = 1.9 1013 0.1 1013 cm2. The 300 K value (ns,full = 3.8 1013 0.1 13 10 cm2) agrees well with the local areal dose as determined by SIMS depth proling (N0,line = 5.6 1013 0.1 1013 P atoms cm2). Since the maximum amount of dopant Figure 2. Intensity proles in horizontal direction over the 10 images depicted in Figure 1, integrated in vertical direction. The top 30 and bottom 30 pixels are excluded. atoms that can diffuse into the silicon is determined by the self-limiting character of organic monolayers, scaling is also expected in terms of sheet silicon layer. This scenario is, however, highly unlikely since the resistance Rs. For a fully monolayer-doped, but unpatterned width of the unpatterned regions is 100 m and the apparent sample Rs = 7.6 102 0.1 102 1 was determined. The doping concentration seems distributed evenly over this large average sheet resistance of the patterned sample should natudistance. Overall, the sequence of images in Figure 1 and the rally be larger and Rs = 1.5 103 0.1 103 1 was found. trends shown in Figure 2 show a decrease in P content when Since the ratio of line width and spacing is 1:1, the sheet resistgoing deeper into the substrate and laterally homogeneous ances (by four-point probe measurements) are also expected to doping in the doped regions for all depths. scale accordingly. (Rs,unpatterned/Rs,patterned) is equal to 0.5, which In Figure 2, the upper (red) curve corresponds to integration corresponds to the expected scaling using a limited source conof the top-left image in Figure 1. The dopant surface dose on dition such as a molecular monolayer. For an estimated depth a doped area measured 2.3 1019 0.1 1019 P atoms cm3, of 100 nm for the electrically conductive layer this results in a corresponding to an areal dose on a doped area of N0,line = (bulk) resistivity = 1.5 102 0.1 102 cm for an unpat5.6 1013 0.1 1013 P atoms cm2 (Figure S2a, Supporting terned sample and = 7.6 103 0.1 103 cm for the patInformation). The exact location of the spot was determined by terned sample. optical interferometry (Figure S2b, Supporting Information). In comparison, on an unpatterned sample that was doped by a full monolayer (RTA, 5 min at 1000 C), the dopant surface dose after SIMS depth proling measured 5.4 1019 0.1 1019 P atoms cm3, which results in an areal dose N0,full sample = 1.1 1014 0.1 1014 P atoms cm2. Figure S3 (Supporting Information) shows the pattern from different axial directions, which conrms the directionality in 3D in one direction and the decrease in doping level with increasing depth. A clear resistance difference is observed (Figure 3) between the current measured perpendicular (black curve) and parallel (red curve) to the line orientation. As expected, the perpendicular orientation results in the lowest resistance, since the doped lines form a parallel resistor network in this case. With the contacts oriented parallel to the lines a signicantly larger resistance (resistors in series) is found. Hall measurements can provide useful information about the type (electrons or holes) and quantity of the (majority) charge carriers. Deduction of the charge carrier type (Figure S4a,b, Supporting Information) revealed electrons as the majority charge carriers, as expected for P implantation. In case of Figure 3. Currentvoltage (IV) measurements on a 1 1 cm2 intrinsic Si a patterned sample, an average sheet carrier density ns,av = (100) sample containing phosphorus n-doped line-shaped regions (width 1.9 1013 0.1 1013 cm2 was deduced from the Hall curve 100 m, spacing 100 m), using RTA, 5 min at 1000 C. For electrical charin Figure S4a (Supporting Information) at a temperature of T = acterization, silver epoxy contact pads were applied perpendicular (black curve, inset (a)) or parallel (red curve, inset (b)) to the line orientation. 300 K. It must be noted that in the electrical determination

wileyonlinelibrary.com

2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Adv. Mater. 2011, 23, 13461350

www.advmat.de www.MaterialsViews.com

COMMUNICATION

Figure 4. Sheet resistance Rs vs T by Van der Pauw measurements of a patterned Si sample with 100-m P-doped line patterns and 100-m spacing.

The calculated mobility of the majority carrier (m) at T = 300 K for the patterned sample was 2.3 102 0.1 102 cm2 V1 s1 and increased to 3.5 0.1 cm2 V1 s1 at T = 150 K. Figure 4 shows a plot of the sheet resistance Rs versus temperature for a patterned sample. According to the Drude model Rs = 1/(nse), where ns and are interdependent and both temperature dependent. At low temperatures, carriers are frozen out and ln (ns) 1/kBT, where kB is the Boltzmann constant. The mobility is dominated by scattering off ionized impurities in this regime, T3/2/NI, where NI is the ionized impurity density. At higher temperatures, ns saturates and is dominated by acoustic phonons, 1/T3/2.[12] For an estimation of the doping efciency, the maximum attainable surface concentration of a full monolayer as determined in the Supporting Information was used (2.2 1014 0.1 1014 P atoms cm2). The ratio of the SIMS areal dose to surface concentration times 100% is the doping efciency. The areal doses determined by SIMS for a doped line N0,line = 5.6 1013 0.1 1013 P atoms cm2 (Figure S2a,b, Supporting Information) and N0,full = 1.1 1014 0.1 1014 P atoms cm2 measured on a fully monolayer-doped, unpatterned sample were used in the calculation. This results in doping efciencies of 26% for a doped line and 50% on a full sample. In conclusion, this work shows that it is possible to fabricate highly doped regions in silicon using a combination of nanoimprint lithography and organic molecular monolayers. These doped regions are expected to be stable over time because at ambient temperatures diffusion coefcients of dopants in solids are negligible. On smaller (nano) scales, however, the effects of (lateral) diffusion to undesired areas can be expected to become more important. By demonstrating proof of principle of lateral control over the positioning of dopants using molecular monolayer formation, we imagine our work can contribute to fabrication of electronic circuitry and devices such as diodes, FETs, or nanowires.

purchased from University Wafer, Boston, MA, USA. Samples were cut to a 1 1 cm2 size using a dicing saw. Si substrates were cleaned by immersion in a Piranha solution for 30 min (concentrated H2SO4: 33% H2O2 = 3:1 v/v Caution: Piranha solutions should be handled with great care in open containers in a fume hood. Piranha is highly corrosive, toxic, and potentially explosive). Cleaned substrates were rinsed with Milli-Q water and blown dry in a stream of nitrogen. A hydrogenterminated Si(100) surface was generated by immersion of the substrate in an argon-sparged, 40 wt% aqueous NH4F solution[13] for 15 min, followed by extensive rinsing with Milli-Q water and drying in a stream of nitrogen. 1-Undecenyl-dimethylphosphonate was synthesized according to a literature procedure.[9] This specic precursor with a long alkyl chain was selected to reduce surface (re)oxidation effects.[14] Monolayer Formation: Monolayers were assembled on Si(100) using a literature procedure of Sieval et al.[15] with minor changes. Monolayers were formed using a diluted reagent as the molecular precursor in reuxing mesitylene (boiling point (b.p.) = 166 C) as the solvent. Glassware was predried with a heat gun at 200 C for 15 min. A small three-necked ask was equipped with a reux cooler. The other inlets were sealed with a rubber septum. A long needle, which served as the nitrogen inlet, was inserted through the septum in one of the inlets. Under vigorous N2 ow freshly hydrogen-terminated samples were inserted. Deoxygenated and argon-sparged reagent and solvent were taken up through a septum by a syringe and subsequently transferred to the reaction vessel after which the needle was placed just above the solution. The nitrogen ow was reduced to a minimum to reduce evaporation of liquids. A metal bath at 180 C was used to heat the ask containing the samples. After monolayer formation (2.5 h) the sample was rinsed with ethanol, acetone, and MilliQ water and ultrasonicated for 15 min in acetone to remove physisorbed materials. Nanoimprint Lithography: The silicon mold for NIL was prepared by using standard photolithographic methods followed by RIE using an Electrotech Twin system PF 340. The mold consisted of 100-m-wide trenches at a 200-m period with a height of 125 nm. Before use, the mold was coated with an antisticking layer by gas-phase deposition of 1H,1H,2H,2H-peruorodecyltrichlorosilane (PFDTS)[16] overnight followed by a 30 min bake at 120 C. Polymerized materials were removed from the silanized silicon molds by ultrasonication for 5 min in acetone. A resist layer of poly(methylmethacrylate) (PMMA, Mw = 350 kD) was applied by spin-coating on a preformed monolayer on Si (Scheme 1a). For imprinting, the mold and substrate were contacted at 40 bar pressure using a hand-press (Specac) at 180 C (PMMA). Residual layer removal was achieved by (anisotropic) RIE on a Electrotech Twin system PF 340 using O2 plasma. Silicon Doping: Doping of silicon by molecular monolayers was achieved as previously described.[6] A thin lm of SiO2 (100 nm) was deposited by electron gun evaporation of SiO2 pellets of 15 mm diameter (Kurt Lesker) using a Balzers BAK 600 apparatus operating at a base pressure of 2 106 bar. By RTA, the distribution of the dopant atoms was achieved using an Amtech Tempress Omega Junior 2-stack oven at 1000 C for 5 min in a N2-atmosphere. After cool down of the samples the template SiO2 layer was stripped off using buffered HF.

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

Experimental Section
Materials and Methods: Nearly intrinsic Si (100) (4 in., thickness 380 m, Float Zone (FZ), resistivity between 24.4 and 42.5 k cm1) was

Acknowledgements
MESA+ cleanroom staff is acknowledged for their help in the deduction of options for the various fabrication steps. Dr. J. G. M. van Berkum (MiPlaza Materials Analysis, Eindhoven, the Netherlands)

Adv. Mater. 2011, 23, 13461350

2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

wileyonlinelibrary.com

1349

www.advmat.de www.MaterialsViews.com

COMMUNICATION

is acknowledged for fruitful discussions and SIMS characterization. NWO-CW (M.D.Y., project 700.55.029) and W.G.v.d.W.s VIDI research program Organic materials for spintronic devices, supported by NWO and STW are acknowledged for nancial support. Received: October 4, 2010 Revised: November 12, 2010 Published online: February 9, 2011

[1] [2] [3] [4] [5] [6]

P. S. Peercy, Nature 2000, 406, 1023. M. Lundstrom, Science 2003, 299, 210. H. J. Queisser, E. E. Haller, Science 1998, 281, 945. D. H. Lee, J. W. Mayer, Proc. IEEE 1974, 62, 1241. A. Renau, Nucl. Instrum. Methods Phys. Res., Sect. B 2005, 237, 284. J. C. Ho, R. Yerushalmi, Z. A. Jacobson, Z. Fan, R. L. Alley, A. Javey, Nat. Mater. 2008, 7, 62. [7] J. M. Buriak, Chem. Commun. 1999, 1051.

[8] J. M. Buriak, Chem. Rev. 2002, 102, 1271. [9] R. T. Lee, R. I. Carey, H. A. Biebuyck, G. M. Whitesides, Langmuir 1994, 10, 741. [10] W. P. Voorthuijzen, M. D. Yilmaz, A. Gomez-Casado, P. Jonkheijm, W. G. Van Der Wiel, J. Huskens, Langmuir 2010, 26, 14210. [11] R. Yerushalmi, J. C. Ho, Z. Fan, A. Javey, Angew. Chem., Int. Ed. 2008, 47, 4440. [12] S. M. Sze, K. K. Ng, Physics of Semiconductor Devices, WileyInterscience, Hoboken, NJ, U.S.A. 2007. [13] a) G. S. Higashi, Y. J. Chabal, G. W. Trucks, K. Raghavachari, Appl. Phys. Lett. 1990, 56, 656; b) C. P. Wade, C. E. D Chidsey, Appl. Phys. Lett. 1997, 71, 1679. [14] A. B. Sieval, R. Linke, H. Zuilhof, E. J. R. Sudhlter, Adv. Mater. 2000, 12, 1457. [15] A. B. Sieval, V. Vleeming, H. Zuilhof, E. J. R. Sudhlter, Langmuir 1999, 15, 8288. [16] G. S. Ferguson, M. K. Chaudhury, H. A. Biebuyck, G. M. Whitesides, Macromolecules 1993, 26, 5870.

1350

wileyonlinelibrary.com

2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Adv. Mater. 2011, 23, 13461350