Phosphorus ionization in silicon doped by self-assembled macromolecularmonolayersHaigang Wu, Ke Li, Xuejiao Gao, and Yaping Dan

Citation: AIP Advances 7, 105310 (2017);View online: https://doi.org/10.1063/1.4999232View Table of Contents: http://aip.scitation.org/toc/adv/7/10Published by the American Institute of Physics

Articles you may be interested inPhotovoltage responses of graphene-Au heterojunctionsAIP Advances 7, 105001 (2017); 10.1063/1.5001771

Competition between the inter-valley scattering and the intra-valley scattering on magnetoconductivityinduced by screened Coulomb disorder in Weyl semimetalsAIP Advances 7, 105003 (2017); 10.1063/1.4998395

Strain compensation in InGaN-based multiple quantum wells using AlGaN interlayersAIP Advances 7, 105312 (2017); 10.1063/1.5000519

High performance metamaterial device with enhanced electromagnetic energy harvesting efficiencyAIP Advances 7, 105204 (2017); 10.1063/1.5002165

A quantum chemical analysis of Zn and Sb doping and co-doping in SnO2AIP Advances 7, 105010 (2017); 10.1063/1.4999014

Interactions between the first mode and the second Bragg gap in a cylindrical waveguide with undulatedwallsAIP Advances 7, 105011 (2017); 10.1063/1.4993478

AIP ADVANCES 7, 105310 (2017)

Phosphorus ionization in silicon doped by self-assembledmacromolecular monolayers

Haigang Wu,1,2 Ke Li,2 Xuejiao Gao,2 and Yaping Dan2,a1School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China2University of Michigan-Shanghai Jiao Tong University Joint Institute,Shanghai Jiao Tong University, Shanghai 200240, China

(Received 7 August 2017; accepted 4 October 2017; published online 13 October 2017)

Individual dopant atoms can be potentially controlled at large scale by the self-assembly of macromolecular dopant carriers. However, low concentration phosphorusdopants often suffer from a low ionization rate due to defects and impurities introducedby the carrier molecules. In this work, we demonstrated a nitrogen-free macro-molecule doping technique and investigated the phosphorus ionization process bylow temperature Hall effect measurements. It was found that the phosphorus dopantsdiffused into the silicon bulk are in nearly full ionization. However, the electronsionized from the phosphorus dopants are mostly trapped by deep level defects thatare likely carbon interstitials. © 2017 Author(s). All article content, except whereotherwise noted, is licensed under a Creative Commons Attribution (CC BY) license(http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/1.4999232

The precise control of individual dopants at arbitrary location is the key to developing atomic scaledevices. Hydrogen lithography by scanning tunneling microscopy (STM)1 and single ion implantationtechnique2 have been demonstrated to manipulate single dopant atoms. However, these techniquesare time-consuming serial processes and difficult to control atoms at large scale. Previously, weproposed to utilize the self-assembly of macromolecular dopant carriers to control single dopantatoms at large scale.3 However, the success of this process is hindered by the low activation rateof phosphorus dopants.3–5 Without a high activation rate, individual dopants, although controlled atlarge scale, will not function properly in electronic devices. Nitrogen is one of the main sources thatlower the activation rate of phosphorus dopants.6 We have shown that nitrogen introduced by thecarrier molecules or coupling reagents will significantly deactivate the P dopants in silicon.3,5

In this work, we demonstrated a nitrogen-free macromolecule doping technique and investigatedthe phosphorus ionization process. Hyperbranched polyglycerols (hbPGs) that do not contain nitrogenwere first synthesized. Each hbPG macromolecule carries one phosphorus atom. The hbPGs moleculeswere then grafted onto H-terminated silicon surfaces without using N-containing coupling reagents.Secondary ion mass spectroscopy and low temperature Hall effect measurements were employedto analyze the sample. It was found that the phosphorus dopants diffused into the silicon bulk arein nearly full ionization. However, the electrons ionized from the P dopants are mostly trapped bydeep level defects that are likely carbon interstitials, resulting in a low nominal ionization rate ofphosphorus dopants.

Hyperbranched polyglycerols (hbPGs) were synthesized by anionic ring-opening multibranchingpolymerization7 of glycerols (J&K Scientific) with diphenyl-phosphinyl hydroquinone (TCI (Shang-hai) Development Co., Ltd) as the core to initiate the reaction. The monomer/core molar ratio is setat ∼210 to produce high molecular weight hbPGs. After purified by dialysis bag (molecular weightcut-off > 50 kDa) in DI water, hbPGs were characterized by nuclear magnetic resonance (NMR)techniques (1H NMR and 13C NMR) and dynamic light scattering (see the results in our previouswork3). The results indicate that the number average molecular weight of hbPGs is approximately84000 g/mol and the diameter of globular hbPGs is ∼11 nm.

aTo whom correspondence should be addressed: yaping.dan@sjtu.edu.cn.

2158-3226/2017/7(10)/105310/6 7, 105310-1 © Author(s) 2017

105310-2 Wu et al. AIP Advances 7, 105310 (2017)

FIG. 1. Doping protocol for self-assembled macromolecule monolayers.

The doping protocol for the self-assembled macromolecule monolayers is depicted in Fig. 1. Itstarts with a silicon wafer cleaned by Piranha solution and is then followed by HF etching to obtainH-terminated silicon surfaces. To graft functional molecular groups on the H-terminated silicon sur-faces, coupling reagents that contain nitrogen (e.g. dicyclohexylcarbodiimide) are often used.8–11

From this potential source, nitrogen contamination might be introduced into the silicon samples,resulting in significant electrical deactivation of P dopants as we observed previously.3,6 Here, weadopted a nitrogen-free coupling strategy in which hbPGs molecules (dissolved in methanol) weredirectly grafted onto an intrinsic silicon wafer (10000 Ω·cm, Institute of Electronic Materials Tech-nology) in Ar ambient. After grafting, X-ray photoelectron spectroscopy (XPS) was employed toanalyze the sample surface. The results show that a major carbon peak in the C1s scan (Fig. 2a) at286.6 eV is observed, which is attributed to the etheric carbon.12 This observation confirms that thehbPGs were successfully immobilized to the silicon surface. To analyze the coverage of hbPGs on thesilicon surface, atomic force microscopy (AFM) was applied to investigate the surface morphology(Fig. 2b). The AFM image indicates that a dense hbPGs film is formed on the silicon surface.

After the analysis, the hbPGs-modified silicon surface was then coated with spin-on-glass (SOGIC1-200, Futurrex Inc.) to form a 200 nm thick SiO2 capping layer. The purpose of the cappinglayer is to prevent possible external contamination in the rapid thermal annealing (RTA) process(1050 C, 30 s), during which phosphorus dopants were diffused into the silicon wafer. In theend, the SiO2 capping layer was removed in buffered oxide etch (BOE) to obtain oxide-free Sisurfaces for secondary ion mass spectroscopic (SIMS) analysis and low temperature Hall effectmeasurements.

To quantify the P dopants in silicon, SIMS was performed to profile the distribution of P dopants,as shown in Fig. 2c. The first few points near the surface are generally not accurate due to the inherenttechnical issue of SIMS. The P concentration starts from ∼8×1016 cm-3 a few nanometers below thesurface and exponentially drops to ∼ 1015 cm-3 within ∼ 80 nm below the surface. By integrating allthe P dopants in this thin layer, we found that the area concentration of phosphorus is 1.97×1011 cm-2.We will discuss the carbon profiles shown in Fig. 2d later.

To analyze the electrical activities of the phosphorus dopants in the sample, low temperatureHall measurements were conducted in a physical property measurement system (Quantum DesignEverCool-II). As shown in Fig. 3a, the Hall resistance is linearly dependent on the magnetic fieldat room temperature but the dependence becomes increasingly nonlinear at lower temperature. Thenonlinearity is caused by the magnetoresistance13,14 of the sample. The actual Hall resistance wasextracted following the method that we previously reported.6 The electron concentration per unit areaobtained from Hall measurements is plotted as a function of temperature in Fig. 3b.

105310-3 Wu et al. AIP Advances 7, 105310 (2017)

FIG. 2. a) XPS narrow scan of C1s spectra for hbPGs-modified silicon surfaces. The main peak at 286.6 eV is attributedto etheric carbons (C-(O)), and the side shoulder at 285.0 eV is assigned to exogenous carbon contamination (C-C) fromatmosphere. b) AFM image for the surface morphology of hbPGs-modified silicon surfaces. c) Phosphorus profile in siliconby performing SIMS twice (black and red). Inset: Phosphorus profile in log scale. d) Carbon profile in silicon by SIMS fordoped (black) and blank (red) sample that are cleaned by oxygen plasma right before SIMS profiling.

In Fig. 3b, the electron concentration of the doped silicon sample decreases from∼5.25×1010 cm-2

at 300 K to ∼2.48×109 cm-2 at 200K, approximately a decrease of 21 folds in electron concentration.The P dopants in silicon have an energy level of 45 meV below the conduction band. The ionizationrate of P dopants will decline at most 2.4 times when the temperature is lowered from 300K to 200K.Clearly the ionization process in the doped silicon is dominated by some deep energy level dopantsor defects. The chemicals used in the synthesis are of CMOS grade. A blank intrinsic silicon waferdoes not show any significant change in electrical conductivity after going through the whole processexcept the surface modification of hbPGs monolayers. Therefore, any detectable concentration ofextrinsic dopants is not possible. Using deep level transient spectroscopy, we recently detected a fairlyhigh concentration of carbon-related defects introduced into silicon by the self-assembled molecularmonolayer.15 These carbon-related defects not only capture electrons ionized from the P dopants

FIG. 3. a) Hall resistance dependence on magnetic field at different temperature (from 200K to 300K). b) Charge carrierconcentration per unit area as a function of temperature in kT. Red curve is the fitting curve of eq. (3). Inset: Diagram showingionization and trapping process.

105310-4 Wu et al. AIP Advances 7, 105310 (2017)

but also electrically deactivate part of P dopants by forming interstitial carbon and substitutionalP (Ci-Ps) pairs. Indeed, we observed a relatively high concentration of carbon impurities introducedby the molecular monolayer as shown in Fig. 2d. The average areal concentration of carbon impuritiesis ∼7×1012 cm-2 in the first 80nm thick layer where most phosphorus dopants were located. As aresult, the ionization process in Fig. 3b is dominated by deep level carbon-related defects in additionto the electrically active P dopants.

If we assume that only deep level donor-type dopants (Ci-Ps pairs for example) dominantlycontribute to the electrical activity observed in Fig. 3b, the extracted concentration for this typeof dopants is significantly larger than the total concentration of phosphorus profiled by SIMS inFig. 2c. As a result, it is not possible that Ci-Ps pairs as donor-type defects play a dominant role here.It is more likely that deep level acceptor-type defects (carbon interstitials) and shallow donor-typedopants (P) are playing the main role. In this case, the electron concentration is governed by thefollowing equation (1).

ne =Nd

1 + 2exp( EF−EdkT )

−Nt

1 + 14 exp( Et−EF

kT )(1)

where ne is the electron concentration in the conduction band, Nd the electrically active dopantconcentration, Nt the defect concentration, EF the Fermi energy level, Ed the donor energy level,Et the defect energy level, k the Boltzmann constant and T the absolute temperature.

The above analysis on the electron concentration as a function of temperature in Fig. 3b indicatesthat the deep level defects dominate the ionization process between 200K and 300K. Therefore, thetemperature dependence of the P ionization plays a negligible role here. For simplicity, we can assumethat the P dopants always remain complete ionization. The temperature dependent term exp

(Et−EF

kT

)can be rewritten as exp

(−

EF−ECkT

)* exp(−EC−Et

kT ) in which exp(−

EF−ECkT

)=

Ncne

and (EC Et) is definedas the activation energy∆Et of the defects. In the end, eq.(1) can be rewritten as eq.(2) in the followingform.

ne =Nd −Nt

1 + Nc4ne∗ exp

(−

∆EtkT

) (2)

where NC is the effective density of states associated with the conduction band (Nc ≈ w ∗ kT3/2, andw is the corresponding constant multiplied with the integration length which is 80 nm in our case).Eq. (2) is a quadratic equation, the real positive solution of which is as following:

ne =−18

[Nc exp

(−∆Et

kT

)+ 4Nt − 4Nd

]

+18

√[Nc exp

(−∆Et

kT

)+ 4Nt − 4Nd

]2

+ 16NcNd exp

(−∆Et

kT

)(3)

By fitting Eq. (3) into the data in Fig. 3b, we find the defect concentration, defect activation energyand the concentration of electrically active P dopants as listed in Table I.

The SIMS data indicate that the P concentration in the bulk is 1.97×1011 cm-3, which is sur-prisingly close to the extracted concentration for the electrically active P dopants 1.98×1011 cm-3

(Table I). This is unlikely a coincidence. It implies that the P dopants diffused into the silicon bulkare almost all electrically active. This is possible, because the P dopants in this work is ∼1016cm-3 orlower, meaning that the average distance between P atoms is 46 nm or more. Fig. 2d clearly showsthat the carbon concentration within the diffusion distance of P dopants is ∼ 1018 cm-3. The averagedistance between carbon atoms is ∼10 nm. The chances that carbon binds with phosphorus or carbon

TABLE I. Extracted parameters by fitting eq. (3) to Fig. 3b. Note: R-square is equal to 0.999.

Parameter Fitting results Standard error

Nd/cm-2 1.98× 1011 6.02× 1010

Nt/cm-2 1.05× 1012 9.17× 1011

∆Et/meV 133.56 14.30

105310-5 Wu et al. AIP Advances 7, 105310 (2017)

are low, meaning that carbon atoms are most likely in substitutional and interstitial forms. A certainportion of phosphorus may bind with carbon, forming deep level donor-type Ci-Ps defects. But thisportion must be small because the fitting results do not support a dominant role of Ci-Ps defects.What’s more, we previously observed15 that only 20% of all P dopants can form Ci-Ps pairs with theinterstitial carbon in a circumstance that the concentration of P dopants and C impurities are similarlyhigh at ∼1018 cm-3. A lower concentration of P dopants in this work shall result in at most the sameor likely a smaller probability for P to bind with C.

At room temperature, the electron concentration is ∼ 5.25×1010 cm-2. If compared to the phos-phorus SIMS data, it appears that the nominal activation rate is only 26.7% (0.525/1.97). But theabove Hall effect analysis indicates that the P dopants are in fact nearly completely ionized. Thisseemingly contradiction is due to the fact that the electrons ionized from the P dopants are mostlytrapped by the acceptor-type of defects that has a concentration of 1.05×1012 cm-2 (Table I). Thedefect energy level is located at 133.56 meV below the conduction band. Although it is a singleenergy level, it is more likely a combination effect of multiple defect energy levels. The defects areprobably standing alone carbon interstitials16 and unlikely the Ci-Ps pairs that all have defect energylevels larger than 250 meV.17 This is in line with the above analysis on the average atom distance andthe finding that P dopants in the silicon bulk are in nearly full ionization. Fig. 2d shows that carbonimpurities have diffused much farther into the bulk (500 nm) than phosphorus dopants (80 nm). Thecarbon interstitials at deep locations will not get involved in the electron trapping because the built-inelectric field due to the separation of electrons and phosphorus ions will confine most electrons nearthe P dopants. As a result, it is the carbon impurities within the first ∼80 nm layer (∼7×1012 cm-2)that contribute to the electrically active carbon interstitials (∼1.05×1012 cm-2 from Table I). It meansthat ∼15% of carbon impurities are interstitial and the rest ∼85% are substitutional.

The carbon interstitials will distort the electronic activity of phosphorus dopants by trappingelectrons. As a result, electronic devices based single dopants will not function properly unless theconcentration of carbon impurities introduced by the molecular monolayers is significantly reduced.One possible approach is to anneal the sample in high-purity oxygen at appropriate temperature(500 C for instance) that is high enough to oxidize carbon into gaseous CO2 and CO, but not toohigh to diffuse carbon impurities into the silicon substrate. After carbon is removed, rapid thermalannealing is then applied to diffuse the desired dopants into the substrate. Without carbon impurities,devices based on single dopants are expected to function properly.

In conclusion, we developed a nitrogen-free monolayer doping technique by self-assembledhbPGs molecules. From the SIMS analysis and low temperature Hall effect measurements, we foundthat the P dopants are in nearly full ionization. Most of the electrons excited from the P dopants aretrapped by acceptor-type of defects that are likely carbon interstitials. Removing the carbon-relateddefects is the key to the development of a defect-free doping technique by self-assembled molecularmonolayers.

ACKNOWLEDGMENTS

The work was supported by the National Science Foundation of China (21503135) and the MajorResearch Plan, Shanghai Science and Technology Commission (16JC1400405). The XPS and lowtemperature Hall effect measurements were conducted at the Instrumental Analysis Center (IAC) ofShanghai Jiao Tong University. SIMS was performed at the Evans Analytical Group (Shanghai).

1 M. Fuechsle, J. A. Miwa, S. Mahapatra, H. Ryu, S. Lee, O. Warschkow, L. C. Hollenberg, G. Klimeck, and M. Y. Simmons,Nature Nanotechnology 7(4), 242 (2012).

2 T. Shinada, S. Okamoto, T. Kobayashi, and I. Ohdomari, Nature 437(7062), 1128 (2005).3 H. G. Wu, B. Guan, Y. R. Sun, Y. P. Zhu, and Y. P. Dan, Scientific Reports 7, 41299 (2017).4 J. O’Connell, S. Biswas, R. Duffy, and J. D. Holmes, Nanotechnology 27(34), 342002 (2016).5 L. Ye, S. P. Pujari, H. Zuilhof, T. Kudernac, M. P. de Jong, W. G. Van Der Wiel, and J. Huskens, ACS Applied Materials

& Interfaces 7(5), 3231 (2015).6 B. Guan, H. Siampour, Z. Fan, S. Wang, X.-Y. Kong, A. Mesli, J. Zhang, and Y. P. Dan, Scientific Reports 5, 12641 (2015).7 R. K. Kainthan, E. B. Muliawan, S. G. Hatzikiriakos, and D. E. Brooks, Macromolecules 39(22), 7708 (2006).8 C. M. Duan and M. E. Meyerhoff, Microchimica Acta 117(3-4), 195 (1995).9 V. Chechik, R. M. Crooks, and C. J. M. Stirling, Advanced Materials 12(16), 1161 (2000).

10 B. L. Frey and R. M. Corn, Analytical Chemistry 68(18), 3187 (1996).

105310-6 Wu et al. AIP Advances 7, 105310 (2017)

11 E. Moore, B. Delalat, R. Vasani, G. Mcphee, H. Thissen, and N. H. Voelcker, ACS Applied Materials & Interfaces 6(17),15243 (2014).

12 https://xpssimplified.com/elements/carbon.php13 P. G. Le Comber, D. I. Jones, and W. E. Spear, Philosophical Magazine A 35(5), 1173 (1977).14 K. W. Edmonds, R. P. Campion, K. Y. Wang, A. C. Neumann, B. L. Gallagher, C. T. Foxon, and P. C. Main, Journal of

Applied Physics 93(10), 6787 (2003).15 X. J. Gao, B. Guan, A. Mesli, K. X. Chen, and Y. P. Dan, “Deep level transient spectroscopic investigation of phosphorus-

doped silicon by self-assembled molecular monolayers,” (under review).16 L. W. Song and G. D. Watkins, Physical Review B 42(9), 5759 (1990).17 X. D. Zhan and G. D. Watkins, Physical Review B 47(11), 6363 (1993).