Investigation of Boron Distribution at the SiO2/Si Interface of Monolayer Doping.

Monolayer doping is a possible method for achieving complex-geometry structures with different semiconductors. Understanding the dopant diffusion behavior of monolayer doping, especially under different heating sources, is essential for further improvement. We examine and compare the doping profile and dopant activation with two different heating sources (rapid thermal annealing and microwave annealing), especially focused on SiO2/Si interface. These heating sources are used for junction diode fabrication, to realize current switching behavior. Direct observations of monolayer doping profiles, especially inside the capping oxide, are discussed to provide quantitative information for dopant concentration. This can provide significant information for better tuning of surface chemistries and process protocols applied in monolayer doping methodologies.

Introduction

Monolayer doping is a potential doping technique to tackle the challenges in the formation of sub-10-nm ultrashallow junctions. It is suitable for doping in complex-geometry structures, such as nanopillar arrays or fins in junctionless devices. It has advantages for mass production and is applicable to semiconductors like n class="Chemical">Si, Ge, Inpan>As, GaAs, etc.

First, n class="Chemical">dopant-carryinpan>g molecules are covalenpan>tly immobilized on the semiconductor surface via a surface reaction with anpan> anpan>chorinpan>g group. Due to the self-limitinpan>g properties on the surface, the dose of pan> class="Chemical">dopant molecules can be manipulated by optimizing the anchoring group reaction conditions and molecular size. Subsequently, this monolayer doping technique can create an ultrashallow junction. This is followed by an annealing step for the incorporation and diffusion of dopants into substrates.[1,2]

n class="Chemical">Boron dopants inpan>troduced by monolayer dopinpan>g inpan>deed draw much attenpan>tion, ranpan>ginpan>g from molecule selection[3] anpan>d monolayer composition[4] to capping oxide consideration.[5] Ye and co-workers designed a new molecule with carborane cluster having 10 boron atoms per molecule to enhance doping levels in silicon.[3] Besides, they also performed a detailed study on the relation between the mixing ratio of molecules in solution and the doping dose inside the silicon after annealing.[4]

A few monolayer doping methods were developed to enhance doping efficiency. Briefly, the key difference is whether capping n class="Chemical">oxide is applied. Previous studies used a layer of pan> class="Chemical">oxide over the monolayer before thermal annealing and stated that this decreases the escape of organic dopant molecules from the surface into the ambient environment.[1] Recent publication demonstrates the effect of capping oxide for noncovalent monolayers of phenylboronic acid (PBA), where the capping oxide not only prevents boron outdiffusion but also acts as a thin solid source for monolayer doping methodologies.[5] Gao et al. noted that oxygen from the SiO2 capping layer can diffuse into the silicon substrate and bond with boron dopants that slightly deactivate boron dopants (but less than 1%).[6] Apparently, oxide capping layer plays a significant role in the monolayer doping process. However, there is no direct observation of the boron dopant distribution inside capping oxide layer after thermal treatment. Therefore, quantitative analysis is desired to provide significant information for better tuning of surface chemistries and process protocols applied in monolayer doping methodologies.

Precisely controlling the n class="Chemical">dopant distribution inpan> nanpan>oscale tranpan>n class="Chemical">sistors is a challenge. Therefore, dopant diffusion is often confined using rapid thermal annealing (RTA) with a low thermal budget. Previous works have confirmed that microwave annealing (MWA) is a good candidate to replace RTA during the activation process for nanoscale devices and three-dimensional (3D) integrated circuits (ICs).[7−9] This is due to its ability to provide a lower process temperature for satisfactory activation, restraining the dopant diffusion. With such low-temperature characteristics of MWA, an extended annealing time is required to provide an equal amount of thermal budget for dopant diffusion and activation, compared to RTA.

This work focuses on understanding the n class="Chemical">dopant distribution inpan>n class="Chemical">side Si and the capping oxide layer, especially with different thermal treatments by RTA and MWA. First, boron-containing molecules (vinylboronic acid dibutyl ester) modify the Si substrate, followed by oxide capping. After advanced annealing by RTA and MWA, monolayer molecules thermally decompose and dopant atoms coming from a segment of the monolayer diffuse into the target substrates (Figure ). X-ray photoelectron spectroscopy (XPS) and point-by-point correction of secondary-ion mass spectrometry (PCOR-SIMS) are utilized to understand the type of dopant diffusion and the precise dopant distribution with dopant loss minimization. Further, this study also demonstrates ultrashallow diode junction formation with monolayer doping by RTA and MWA activation, which is applicable for device fabrications.

Figure 1

Schematic representation of boron monolayer doping with rapid thermal annealing and microwave annealing.

Schematic representation of n class="Chemical">boron monpan>olayer dopinpan>g with rapid thermal anpan>nealinpan>g anpan>d microwave anpan>nealinpan>g.

Results and Discussion

Formation of Boron-Containing SAM (B SAM)

The exposure of the Si surface to a solutionpan> of aqueous (HF) (5 wt %) for 1 minpan> results inpan> the removal of native pan> class="Chemical">oxide, with hydrogen termination (water contact angle ≈80°, Figure , H-Si). After hydrosilylation with vinylboronic acid dibutyl ester, a clear change in the polarity of the surface is observed, as witnessed by a drop in the contact angle to 60 ± 1°, indicating the formation of the boron-containing SAM (B SAM).

Further evidence for the formation of the B SAM is the peak at 185 eV in the XPS B 1s spectrum, which is characteristic of the O–B–O n class="Chemical">boron of the B SAM (Figure a). The B SAM was capped with a layer of pan> class="Chemical">SiO2 (15 nm) to prevent the outdiffusion of boron dopant during thermal annealing. This oxide capping is removed before the XPS measurement. To get clear information on the diffusion of dopant atoms after thermal annealing, angle-resolved analysis (ARXPS) was used to detect the chemical bonding states of boron inside the silicon lattice, focusing on the surface layer. The B 1s spectra can be individually decomposed into two spectra having binding energies of 186 and 188.8 eV (Figure b–d). The binding energy at 186 eV is assigned to electrically activated boron, and the binding energy at 188.8 eV is attributed to boron clusters (such clusters cannot contribute electrical conduction).[10] The B SAM treated by RTA at 1000 °C for 60 s (Figure b) has a higher ratio of activated boron, relative to annealing by MWA at 2800 W for 150 s (Figure c). Microwave annealing is a relatively low-temperature process compared with RTA.[7] Therefore, boron atoms from monolayer doped into silicon substrate by MWA cannot achieve full activation by silicon atom replacement and some boron remains as deactivated clusters. The spectra of BF2 implantation (20 keV with a dose of 5 × 1015 cm–2) reveal that boron clusters are the dominant component on the Si surface without annealing, and they require a descent amount of thermal treatment for boron activation along with Si recrystallization. In general, full recrystallization of Si lattice is hard to achieve without junction depth compromised (Figure d).[9,11]

Figure 2

X-ray photoelectron spectra of B 1s of (a) B SAM, followed by (b) rapid thermal annealing or (c) microwave annealing; (d) sample prepared by traditional BF2 implantation as reference. The two orange lines indicate the position of activated boron (186 eV) and boron clusters (188.8 eV) individually.[10]

X-ray photoelectron spectra of B 1s of (a) B SAM, followed by (b) rapid thermal annealing or (c) microwave annealing; (d) sample prepared by traditional n class="Gene">BF2 implanpan>tation as referenpan>ce. The two oranpan>ge linpan>es inpan>dicate the position of activated boron (186 eV) and boron clusters (188.8 eV) individually.[10]

Secondary-ion mass spectroscopy measurements were performed to quantify the doping profile. PCOR-n class="Chemical">SIMS (EAG Laboratories) was used to clearly definpan>e the pan> class="Chemical">boron distribution along the matrix (SiO2/Si). To get complete information on the dopant distribution after annealing, the capping oxide layer was kept for measurements, to observe the boron diffusion inside the SiO2 and Si layers (individually). The interface of SiO2/Si was defined at 10% oxygen intensity, which is decided by comparing the oxygen concentration before and after oxide removal, to reflect the exact interface position. As shown in Figure , the peak concentration is inside the oxide capping layer above the interface at about 1–2 nm for both annealing methods. This further confirms the successful anchoring of B SAM on the Si surface, acting like a self-limiting doping source. The standard definition for junction depth is the distance from the surface at which the doping concentration drops below 1018 atoms/cc. The junction depth is 5.1 nm for boron dopants by MWA at 2800 W for 150 s, while the depth is slightly increased to 7.1 nm for boron doping with RTA at 1000 °C for 60 s. Dopant profile appears to have unsymmetric distribution that nearly 80% of boron remains inside SiO2. Furthermore, the difference in total thermal budget by MWA and by RTA appears to impact dopant distribution. A longer annealing time by MWA allows boron having pronounced diffusion inside SiO2 by thermal conduction resulted from microwave energy absorption. Limited boron diffusion into Si was observed by considering MWA as low-temperature treatment, while a relatively higher temperature by RTA allows dopant diffusion inside SiO2 and into Si uniformly from monolayer location.

Figure 3

(A) Schematic drawing of dopant distribution at SiO2/Si interface by RTA and by MWA. (B) SIMS profile of B SAM-modified Si substrate by RTA at 1000 °C and by MWA at 2800 W. The boundary of SiO2/Si was defined at 10% O intensity. The junction depth is about 7.1 nm for doping with RTA and about 5.1 nm for microwave annealing. Note that the maximum peak for both methods is in the same position at 1–2 nm from the SiO2/Si interface, which exactly corresponds to the monolayer location.

(A) Schematic drawing of n class="Chemical">dopant distribution at pan> class="Chemical">SiO2/Si interface by RTA and by MWA. (B) SIMS profile of B SAM-modified Si substrate by RTA at 1000 °C and by MWA at 2800 W. The boundary of SiO2/Si was defined at 10% O intensity. The junction depth is about 7.1 nm for doping with RTA and about 5.1 nm for microwave annealing. Note that the maximum peak for both methods is in the same position at 1–2 nm from the SiO2/Si interface, which exactly corresponds to the monolayer location.

By integrating the surface area of the n class="Chemical">dopant profile (see the Supportinpan>g Inpan>formation), anpan>nealinpan>g by pan> class="Chemical">RTA shows a 2 times higher dopant concentration inside the Si substrate than by MWA. The total concentration inside Si and SiO2 was shown to be a similar level and consistent with the estimated monolayer coverage. Monolayer doping occurs without outdiffusion under the oxide capping layer protection. However, less than 20% of the dopants successfully diffuse inside Si. This is different from the traditional ion implantation, leading to a higher junction resistance for device applications.

Four-point probe measurements (Figure A) were used to obtain sheet resistanpan>ce values (Rs) of monolayer dopinpan>g on pan> class="Chemical">Si after RTA at 900 and 1000 °C, and by MWA at different powers from 1200 to 2800 W individually. For RTA, the Rs values show a sharp decrease with annealing time, at both 900 and 1000 °C. This can be attributed to the efficient activation of dopants in a short period of time. The enhancement of the diffusivity constant of B at elevated temperatures can explain that the doping profile at 1000 °C shows a consistently lower resistance than at 900 °C after the same annealing time. More boron atoms diffuse into the SiO2 and Si substrates individually from the B SAM at a higher temperature in the same period. However, they eventually have similar resistances due to self-limited source characteristics. For MWA, the Rs values only show a slight decrease after annealing from 1200 to 2300 W. When the microwave power increases to 2800 W, Rs drops about 1 order of magnitude but is still higher than with RTA (Figure A). Temperature correlation of microwave power shows that microwaving with 2800 W for 150 s is less than 800 °C[8] and does not provide as sufficient a thermal budget as RTA at 900 and 1000 °C for boron dopant activation, which is shown in the XPS spectrum (Figure a). The SIMS profile presents higher concentrations of boron atoms inside SiO2 for MWA treatment. The insufficient thermal budget and less dopants inside Si contribute to the higher resistance of monolayer doping by MWA than by RTA.

Figure 4

(A) Sheet resistance with annealing time for boron monolayer doping with RTA at 900 °C, at 1000 °C, and MWA with different powers. The highlighted circles are the conditions used for the XPS and SIMS analyses. (B) Sheet resistance as a function of junction depth (X at 1 × 1018/cm3) for boron-doped Si by RTA and MWA, which is compared with boron monolayer doping from Ho et al.,[1] Ye et al.,[3,4] and traditional implantation techniques.[12−16]

(A) Sheet resistanpan>ce with anpan>nealinpan>g time for pan> class="Chemical">boron monolayer doping with RTA at 900 °C, at 1000 °C, and MWA with different powers. The highlighted circles are the conditions used for the XPS and SIMS analyses. (B) Sheet resistance as a function of junction depth (X at 1 × 1018/cm3) for boron-doped Si by RTA and MWA, which is compared with boron monolayer doping from Ho et al.,[1] Ye et al.,[3,4] and traditional implantation techniques.[12−16]

Figure B compares the current work of monolayer doping by n class="Chemical">MWA anpan>d pan> class="Chemical">RTA with other implantation results. Junction depth (Xj) is defined at a concentration of 1 × 1018/cm3, and the shallow junction (Xj) was observed in this study by MWA, as shown by PCOR-SIMS (Figure B). The current work with monolayer doping reveals junction requirement, while a higher Rs can be further decreased by multilayer doping of boron-containing molecules. Such a shallow junction was successfully demonstrated for its superior properties as conformal shell doping of a junctionless FinFET device.[17,18]

The p–n junction formation shows rectification for modern electronic applications and is widely used in MOSFETs to prevent the source and drain region from leaking current. The starting material is (100) n-type n class="Chemical">silicon wafer that is patterned with the B SAM, capped with pan> class="Chemical">SiO2, and followed by two different annealing methods: RTA and MWA. Figure shows the I–V characteristics of the monolayer doping with RTA at 1000 °C for 60 s and MWA at 2800 W for 150 s. The control diode was fabricated by directly patterning a metal pad on the Si surface, without monolayer doping. For boron monolayer doping with MWA, the I–V curve shows rectification behavior as a control sample. This may result from current tunneling with such an ultrashallow junction, incomplete junction formation due to an insufficient dopant concentration, and activation level inside the Si layer. For boron monolayer doping with RTA, a lower leakage current was observed that contributed to a higher activation rate of dopants with RTA at 1000 °C for 60 s and relatively deeper junction formation. To improve the junction for monolayer doping with MWA, B SAM modification followed by MWA was repeated twice. Nearly 2 times the current enhancement was observed for the forward current. The leakage current was further suppressed, indicating better junction formation, as shown in the Supporting Information.

Figure 5

(A) I–V characteristics of a diode junction with monolayer doping by MWA at 2800 W for 150 s and by RTA at 1000 °C for 60 s. (B) Schematic showing the diode structure without and with monolayer doping.

(A) I–V characteristics of a diode junction with monolayer doping by MWA at 2800 W for 150 s anpan>d by pan> class="Chemical">RTA at 1000 °C for 60 s. (B) Schematic showing the diode structure without and with monolayer doping.

Conclusions

Direct observation of the monolayer doping profile, especially inn class="Chemical">side cappinpan>g pan> class="Chemical">oxide, is for the first time being discussed comprehensively. Such a dopant diffusion behavior shows significant influences by heating source manipulation. Boron dopants diffuse symmetrically into Si and SiO2 with uniform heating by RTA, while showing an unsymmetrical distribution profile by MWA. Therefore, the distribution of dopants can be manipulated by controlling the heating source types with the self-limited characteristics of monolayer doping.

Monolayer doping provides a shallow n class="Chemical">dopant distribution that canpan> be obtainpan>ed eapan> class="Chemical">sily, especially without lattice damage. Traditional implantation damage cannot be fully recovered. This damage-free monolayer doping with MWA has already been successfully demonstrated for junctionless device fabrication, showing active leakage current suppression.[17,18] However, monolayer doping has some remaining challenges. The actual dopant concentration inside Si is much less than that in traditional ion implantation due to its self-limiting source, and significant diffusion loss occurs inside the capping oxide, as demonstrated from this work. Further improvement needs to be focused on the improvement of the self-limiting source, such as using an organic multilayer structure or polymer and optimizing the capping materials and thickness to reduce the dopant loss.

The microwave-assisted dopinpan>g method was extended to additionpan>al pan> class="Chemical">dopants such as phosphorus using suitable molecular precursors.[17] This method can be further applied to other substrates such as Ge and two-dimensional (2D) materials by appropriate surface chemistry, for improving the fine control of nanomaterials.[19]

Methods

Substrate and Monolayer Preparation

n class="Chemical">Single-n class="Chemical">side-polished silicon substrates (100) and p-doped and n-doped silicon substrates were used for monolayer preparation. A hydrogen-terminated silicon surface was generated by immersion of the substrates in an HF solution (5 wt %) for 1 min, followed by extensive rinsing with water. After drying with a nitrogen stream, the substrates were used immediately for providing freshly hydrogen-terminated surfaces to form boron-containing monolayers (hydrosilylation by vinylboronic acid dibutyl ester).[20] Monolayers were formed in all experiments using diluted reagents as the molecular precursor, in refluxing mesitylene (bp = 166 °C) as the solvent. A long needle, serving as the nitrogen inlet, was inserted through the septum in one of the inlets. Under N2 flow, freshly hydrogen-terminated samples were inserted. The reactions were carried out for 4 h at 166 °C. The samples were then rinsed with mesitylene and acetone to remove physisorbed materials.

Contact Angle Measurements

Contact angles were measured with Millipore n class="Chemical">water (18.2 MΩ·cm) onpan> a Krüss GH-100 conpan>tact anpan>gle measurinpan>g inpan>strument equipped with a CCD camera. Static anpan>gles (θstatic) were determinpan>ed automatically by a drop shape anpan>alyn class="Chemical">sis.

XPS

XPS spectra were recorded on a Thermo Fisher Scientific theta probe equipped with a monochromatic Al Kα X-ray source operated at 1486.6 eV. Angle-resolved analysis (ARXPS) (takeoff anpan>gle, 30°) was used to probe subsurface chemical states without sputterinpan>g (which would destroy the surface). The spectra were referenpan>ced to the mainpan> C 1s peak set at 284.0 eV. The X-ray beam pan> class="Chemical">size was varied from 15 to 400 μm. The data were collected from a surface area of 100 μm × 300 μm with a pass energy of 224 eV, with a step energy of 0.8 eV for survey scans and 0.4 eV for high-resolution scans. For quantitative analysis, the sensitivity factors used to correct the number of counts under each peak were C 1s, 1.00; N 1s, 1.59. The measurement was made after 25 cycles of scanning.

PCOR-SIMS

PCOR-SIMS (EAG Laboratories) was used to measure the continpan>uous pan> class="Chemical">dopant distribution based on matrix composition. This is also called point-by-point-corrected SIMS or PCOR-SIMS. The SIMS measurements were carried out using a Physical Electronics ADEPT-1010 quadrupole setup. 10B, 11B, and 30Si were monitored under O2+ bombardment, with an impact energy of 650 eV incident at 45°. The analysis chamber was backfilled with a partial pressure of O2 to decrease ion yield variations at the surface and improve quantification. Secondary ions were collected from the center 10% of a 450 × 450 μ raster area. Stylus profilometry was used to determine the depth of sputtered craters and calibrate the depth scale, assuming a constant sputter rate for the entire profile. Concentrations of 10B and 11B in the Si were calculated, using the relative sensitivity factor determined from a standard sample. The B profiles were normalized on a point-to-point basis to the 30Si profile before the relative sensitivity factor was applied.

Microwave System

A custom-made microwave system consisted of a vertical heatinpan>g chamber (5.57 × 106 cm3) with pan> class="Chemical">six heating powers. It can be operated to adjust the input power from 100 to 600 s. Microwaves were generated by magnetrons, and the frequency is 5.8 GHz. The power amplitude of each magnetron is about 500–600 W. A nitrogen purge was performed before starting the microwave, and the N2 flow was maintained throughout the process. The microwave power and time settings were obtained from a previous correlation study with RTA.

Thermal Processor

Rapid thermal annealing was performed un class="Chemical">sing anpan> AG Associates Heatpulse 610 rapid thermal processor. The semiconpan>ductor-grade quartz process chamber canpan> use 21 tunpan>gsten halogen lamps inpan> anpan> upper anpan>d lower array with anpan> extended-ranpan>ge pyrometer (400–1300 °C). The heatinpan>g rate is user-conpan>trollable from 1 to 200 °C/s.

Motorized Four-Point Probe

Sheet resistanpan>ce was measured by a motorized four-poinpan>t probe setup (Swinpan>g, MFP6). All samples were sealed inpan> a pan> class="Chemical">nitrogen flow chamber immediately after HF dipping and measured by a tungsten-steel probe.