SPECIALTY CHEMICALS AND ENGINEERED MATERIALS | WHITE PAPER Beam Performance of Fluoride Dopant Gases and Their Gas Mixtures Authors: Sharad Yedave, Ying Tang, Joseph Sweeney, Joseph Despres – Entegris

INTRODUCTION — Abstract – Fluorine co-implantation is used in advanced semiconductor device manufacturing Scaling down of microelectronic devices, as predicted by Moore’s law, continues to drive technological development and innovation. for defect engineering, shallow junction formation, As the device features continue to shrink, the reliability of device and material modification. A common setup for structures, defects/charge traps, and noise becomes more crucial. fluorine implantation includes feeding a fluoride There is growing interest in fluorine (F) implantation to resolve the dopant source gas into the ion source to provide related challenges, improve device performance and reliability, as fluorine (F) as one of the dissociation byproducts. well as improve device lifetimes1. Typical setup for fluorine implan- tation employs a gaseous fluoride source gas that feeds into the ion Use of fluorine containing feed gases creates a source to provide fluorine as one of the dissociation byproducts. cycle that can negatively impact ion Fluorine containing gas species – namely Boron Trifluoride (BF3), source performance, source life, implant tool pro- Tetrafluoride (SiF4), and (GeF4) ductivity, and cost of tool ownership. However, – are readily available and commonly used on implant tools for there has been a significant interest in improving p-type dopant and pre-amorphization implants (PAI). Ion source fluorine ion beam performance considering its operation with these fluorides tends to result in relatively poor source life and performance issues due to the halogen cycle. There broad potential applications. This paper presents is continued effort in the industry to overcome the challenges initial results of fluorine beam evaluation of com- posed with using fluoride dopants, and in this paper we present

mon dopant candidates (BF3, SiF4, and GeF4) and results on varying ways to address these issues. Included in this related hydrogen co-mixtures for added process discussion are results of fluorine beam evaluations including efficiency and tool productivity improvements. improvements with fluoride dopants and their hydrogen co- mixtures to reduce the predominance of the halogen cycle for added tool productivity gains2. Keywords – F; BF3; SiF4; GeF4; W; WFx; Fluorine implantation; Performance; Boron Trifluoride; Silicon Tetrafluoride; Germanium Tetrafluoride; Ion Implantation; Gas Mixture; H Co-mixture; EXPERIMENTAL SETUP 2 — H2 Co-flow; Source life; Optimization; Improve- ment; Process Efficiency A. Fluoride Dopants and H2 Co-flow Gas Mixture

For fluorine beam evaluation BF3, SiF4, and GeF4 were tested and

installed, along with H2, on separate gas cards. Different co-flow gas mixture compositions were accomplished by controlling the Mass Flow Controllers (MFC) and co-flowing the fluoride dopant

and H2 into the ion source at predefined flow rates. B. Implant Source Test Stand (STS) magnetic flux density and beam optics at each gas flow rate. During this test, the extraction voltage was All tests were performed on Entegris’ ion Source Test held constant at 20 kV and the source beam current Stand (STS), which is installed at our Danbury, CT facility. was also fixed at 20 mA. The STS tool is a fully functional ion implanter with the exception of wafer handling. Details of the facility can BF₃ BF + be found elsewhere3. Tests were conducted using an 5 2 (1) BF₃ Flow = 1.5 sccm Indirectly Heated Cathode (IHC) source with tungsten (2) Arc Power = 680 watt 4 (W) arc chamber, which represents a typical set-up used in the industry. Beam performance and dopant

nt (mA) + mass spectra were recorded to examine the effect of e 3 B + each test condition. F 2 Beam Curr

BF+ 1 FLOURINE BEAM EVALUATION — 0 51525354555 Fluorine (F) is the most electronegative element known, AMU Figure 1. BF Mass Spectrum tuned for fluorine beam. is extremely reactive, and exists as a diatomic gas (F2) 3 at room temperature, which is highly toxic. Due to its toxicity and safety concerns all fluoride containing SiF₄ dopant gases that were tested were packaged in 5 (1) SiF₄ Flow = 1.2 sccm ® ® 4 Entegris’ Safe Delivery Source (SDS ) cylinder . (2) Arc Power = 600 watt 4

A. Dissociation of Fluorides - Mass Spectra Si+

nt (mA) 3 + e SiF3 Dopant gas introduced into the arc chamber undergoes F+ electron impact ionization and for dopant fluorides 2 Beam Curr + represented by AFx (where A = B, Si or Ge), this results SiF + + + Si++ in dissociation into fragments A , F , and AFx-n (n=1, 2, …, 1 + SiF2 x-1) in the source plasma. For example, a BF3 molecule + + + + 0 will dissociate in its fragments - B , F , BF , and BF2 . The relative abundance of each ion in the plasma and 525456585 AMU extracted ion beam varies with the source conditions. Figure 2. SiF Mass Spectrum tuned for fluorine ion beam. Therefore, tuning process inputs, such as ion source 4 pressure via the gas flow rate and source power input, can result in an optimal beam condition. Mass spectra GeF₄ of BF3, SiF4, and GeF4 dopants obtained while optimiz- (1) GeF₄ Flow = 0.5 sccm 2 ing fluorine beam are presented in Figures 1, 2, and 3, (2) Arc Power = 750 watt respectively, which show typical dopant fragments F+ Ge+

with altered abundance. nt (mA) e

Improvement in F+ – fluorine beam – was observed 1 GeF+ under relatively lower dopant gas flow with increased Beam Curr Ge++ arc power (with Arc V @ 110 V) and lower source magnetic field condition.

0 B. F+ Performance Optimization – 525456585 AMU Fluoride Dopants Figure 3. GeF Mass Spectrum tuned for fluorine ion beam. Fluorine beam performance of selected fluoride dopants 4 tested at different values of arc voltage (90V, 110V) and dopant gas flow rates is presented in Figure 4. +F performance was evaluated by optimizing source

2 + C. Fluoride Dopant Interaction Byproducts F beam performance of H2 co- flow BF₃ (Arc = 110 V) mixtures of respective fluoride 2.5 dopants evaluated for the 2nd highest BF₃ (Arc = 90 V) points at 110 V with better stability Interaction of the tested corrosive fluorine containing SiF₄ (Arc = 110 V) feed gases with the tungsten arc chamber results in SiF₄ (Arc = 90 V) tungsten (W) and tungsten fluoride (WF , x=1-6) by- GeF₄ (Arc = 110 V) x ent (mA) 2 products. Tungsten fluorides, in the form of WF , act as GeF₄ (Arc = 90 V) x carriers and with typical ion source conditions promote

Beam Curr transport of tungsten within the arc chamber. This is + F 1.5 commonly referred to as the halogen cycle5. This can F+ Tuned Beam be seen below in Figure 5, which details the mass spectra in the AMU region where these compounds 1 are present. Figure 5 examines relative strength of the 01234 Gas Flow (sccm) byproducts of different fluorides that ultimately trans- lates to the degree of impact on source life. Due to Figure 4. Fluorine beam performance of BF3, SiF4 and GeF4 at different values of gas flow rate and arc voltage. high reactivity with tungsten, GeF4 generated relatively higher amounts of tungsten fluoride byproducts, As can be seen in the BF series in Figure 4, the highest 3 followed by SiF4 and subsequently BF3 in this test. fluorine beam current that was achieved in this test 0.6 was 2.3 mA. This occurred when the gas flow was at + GeF SiF BF the lowest tested value of 1.4 sccm and the arc voltage W 4 4 3 was set to 110 V. After BF3, SiF4 was able to produce the + next highest F beam current, which was approximately 0.4 5% lower than what could be obtained when using

BF3. For SiF4 the best performance was achieved with a gas flow rate of 1 sccm and an arc voltage setting of WF + 0.2 6 110 V. In comparison when using GeF4, the available free fluorine was found to be the lowest of the three Normalized Beam Current tested molecules. GeF4 provided approximately 25% lower beam current than BF3. It should be noted that in 0 + order to reach the maximum F beam current for GeF4 170 200 230 260 290 it required a significantly lower flow rate of 0.4 sccm AMU than the other tested molecules. In general, it was Figure 5. Tungsten byproducts (W +, WF +) resulting from application of observed that lowering the dopant gas flow rate 6 fluoride dopants (BF3, SiF4 and GeF4). and increasing the arc voltage under weaker source magnetic flux enables the optimum fluorine beam The effect of the halogen cycle is generally mitigated performance for the tested fluoride dopants. As seen by balancing the use of fluoride and hydride recipes in Figure 4, an increase in the fluoride dopant gas flow on the tool. However, this may not be as effective as rate results in increasing the arc chamber pressure needed and can also pose additional tool and sched- and under the tested given conditions, this results in uling issues. An alternative solution to minimizing the reducing the fluorine beam. Lower fluorine beam impact of the halogen effect and increasing ion source performance was seen at the lower arc voltage lifetime, is to co-flow hydrogen with the fluoride dopant. setting of 90 V for all flow rates and dopants tested. Entegris has shown the positive source life effect when utilizing a pre-mixed single cylinder solution for each of the tested fluoride dopants with hydrogen 2.

3 FLUORIDE DOPANT-H2 CO-FLOW EVALUATION BF₃ BF₃ + H₂ — BF₂+ 0.08 2 + + 0.06 W nt

A. F Performance – Fluoride Dopants e and H co-flow 0.04 2 WF + B+ 0.02 6 Effect of 2H co-flow was characterized for BF3, SiF4, and F+ 0.00 + 170210 250290 GeF4 dopants by evaluating F beam performance and 1 their mass spectra with different 2H co-flow rates. As presented in Figure 4, highest F+ beam performance Normalized Beam Curr BF+ can be noted at the lowest dopant gas flow however HF+ the beam conditions during the test were not very stable. 0 Beam instability noticed was due to the extreme arc 51525354555 chamber conditions as a result of the lower gas pressure AMU (lean gas flow) and higher source power at lower source magnetic flux. Hence, further +F beam performance SiF₄ SiF₄ + H₂ 0.16 W+ with H2 co-flow was evaluated at the second highest 2 0.12 +

F beam data point (presented in Figure 4 with arc @ nt + 0.08

e Si + WF6 110 V) for each dopant. The fluorine beam was tuned 0.04 by optimizing the source magnetic field and beam 0 170210 250290 optics for each of the co-flow mixtures. The following F+ 1 source conditions were held constant: (1) Fluoride SiF+

Dopant Flow – (a) BF Gas Flow = 1.5 sccm, (b) SiF Normalized Beam Curr 3 4 Si+ Gas Flow = 1.2 sccm, (c) GeF Gas Flow = 0.5 sccm, 4 HF+ (2) Arc Voltage = 110 V, (3) Extraction Voltage = 20 kV, (4) Source Beam Current = 20 mA. 0 10 20 30 40 50 60 AMU 1.00

GeF₄ GeF₄ + H₂ F+ Tuned Beam Ge+ 0.75 F+ 0.6 W+ 1

+ 0.4

nt HF

e + WF6 0.50 0.2

0 GeF+ BF₃-H₂ (Arc = 110 V) 170210 250290 Ge++ 0.25 SiF₄-H₂ (Arc = 110 V) 0.5 Normalized Beam Current- F+

GeF₄-H₂ (Arc = 110 V) Normalized Beam Curr

0.00 0 X2 X3 X7 X10 X H₂ Co-Flow (*A.U.) 0 *X on x-axis is a proprietary scale factor in Arbitrary Unit 15 35 55 75 95 AMU Figure 6. Effect of H2 on fluorine beam current for (a) BF3 /H2, (b) SiF4 / H2, and (c) GeF4 /H2 co-flow mixtures (X is a proprietary scaling factor). Figure 7. BF3, SiF4, and GeF4 mass spectra without and with H2 co-flow, + + + when optimizing F beam; insets show impact on W and WFx byproducts.

4 Figure 6 compares how the normalized F+ beam was 0.12 BF₃-H₂ (Arc = 110 V) 0.6 Normalized Beam Curr impacted by increasing the H co-flow dilution. There + SiF₄-H₂ (Arc = 110 V) 2 z + GeF₄-H₂ (Arc = 110 V) was an insignificant change in F beam for GeF4 over a nt – W e range of H2 co-flow values. For BF3 and SiF4 dopants, 0.08 0.4 there was minor impact on F+ beam below 2X range F+ Tuned Beam of H2 co-flow (X is proprietary scale factor 2H co-flow), e however, optimized fluorine beam appears to drop 0.04 0.2 nt – W slowly with further increase in H co-flow. Impact on

2 + Normalized Beam Curr + F beam was more for SiF4/H2 than BF3/H2 mixtures at higher H co-flow dilution. 0.00 0.0 2 0 X2 X3 X7 X10 X H₂ Co-Flow (*A.U.) Normalized mass spectra for BF3, SiF4, and GeF4 with *X on x-axis is a proprietary scale factor in Arbitrary Unit and without optimized H co-flow are compared in 2 Figure 8. Effect of H co-flow on Tungsten (W) byproduct, when using Figure 7. Each comparison for BF , SiF , and GeF shows 2 3 4 4 fluoride dopants – BF3, SiF4, and GeF4 (X is a proprietary scaling factor). expected overlapping of most of the dopant fragments except for the appearance of HF+ fragment with H 2 BF₃-H₂ (Arc = 110 V) co-flow. Presence of HF+ fragment at mass 20 AMU 0.03 SiF₄-H₂ (Arc = 110 V) 0.12 Normalized + GeF₄-H₂ (Arc = 110 V) in H2 co-flow mixtures indicates an interaction of 2H co-gas with reactive fluorine, which otherwise would nt – WF e fluorinate tungsten causing tungsten transport, thus, B 0.02 0.08 eam Curr interrupting halogen cycle in the arc chamber. Beam Curr + F Tuned Beam ent – WF Inset charts for BF3, SiF4, and GeF4 are shown in Figure 7 comparing mass spectra over higher AMU range rep- 0.01 z 0.04

+ Normalized resenting tungsten (W) and tungsten fluoride (WF6) fragments with and without H2 co-flow. Spectra for optimized fluoride – H co-flow mixtures show signifi- 0.00 0.00 2 0 X 2 X3 X 7 X 10 X cantly reduced W and WF6 byproducts, as a result of H₂ Co-Flow (*A.U.) reduced fluorination of arc chamber tungsten. *X on x-axis is a proprietary scale factor in Arbitrary Unit Figure 9. Effect of H co-flow on Tungsten Fluoride (WF) byproduct, Effect of H co-flow on +W and WF+ byproducts is 2 2 when using fluoride dopants – BF3, SiF4, and GeF4 (X is a proprietary shown in Figure 8 and Figure 9, respectively, when scaling factor). using BF3, SiF4 and GeF4 dopants with different 2H co-flow percentages. Note the difference in vertical scales on left (BF and SiF data) and right side (GeF 3 4 4 CONCLUSION data) in Figure 8 and Figure 9. Figure 8 shows slow — + continuous reduction in W byproduct for tested GeF4/ + Fluorine beam evaluations were conducted for perfor- H2 co-mixture dropping to ~ 54% level, whereas, W appears to drop quickly and eventually stabilizing to a mance improvement and characterization of source process conditions such as dopant flow, arc power, lower level of ~ 37% for SiF4/H2 and 20% for BF3/H2 co- ion source magnetic field, and hydrogen percentage. mixtures compared to original level without the H2. As shown in Figure 9, similar impact on WF+ can be seen Test results summarized in this paper show the as W+ displaying slow continuous reduction in WF+ following: byproduct declining to ~ 76% level for GeF4/H2, whereas, WF+ appears to drop lower and level off at ~ 27% for 1. Highest F+ beam with BF3 followed by SiF4 SiF4/H2 and 17% for BF3/H2 as compared to their original than GeF4. level without the H2. 2. Higher F+ beam at ~ 50% lower fluoride dopant + + Minimizing W and WF with H2 suggests reduced flow with lower source magnetic field at the tungsten transport and a positive impact on the ion expense of over ~ 200% higher arc power source, which results in a longer source lifetime. Thus, (@ higher Arc Voltage = 110 V) than the typical a fluoride dopant/H2 co-flow gas mixture of suitable dopant tuning conditions. composition can be selected to provide both optimized 3. Optimized H co-mixture of BF , SiF and GeF F+ beam and improved source lifetime performance. 2 3 4 4 can offer better solution with longer source life and beam current. 5 REFERENCES —

1 S.-K. Kwon et al., Effects of Fluorine on the NBTI Reliability and Low-Frequency Noise Characteristics of p-MOSFETs, IEEE J. Electron Devices Soc., Vol. 6, pp. 808-814, 2018.

2 (i) Y. Tang et al., Ion Implanter Performance Improvement for Boron

Doping by Using Boron Trifluoride (BF3) and Hydrogen (H2) Mixture Gases, IIT Proceeding, 2014, pp.361-364; (ii) S. Yedave, Y. Tang, O. Byl, J. Sweeney, Silicon Tetrafluoride Dopant Gas for Silicon Ion Implantation Optimization and Improvements, IIT Proceeding, 2014, pp.369-372; (ii) Barry Chambers et al., Germanium Ion Implantation Efficiency Improvement with Use of Germanium Tetrafluoride, IIT Proceeding, 2014, pp.399-402.

3 S. Yedave, J. Arnó, S. Bishop, F. DiMeo Jr., R. Kaim, L. Wang, ATMI’s Ion Implant Process Efficiency Research Laboratory (IIPERL), IIT Proceeding, 2006, pp. 489-492.

4 T. Romig, J. McManus, K. Olander, and R. Kirk, Advances in Ion Implanter Productivity and Safety, Solid State Technology, Vol. 39, pp.6974, Dec. 1996

5 O. Byl, S. Yedave, S. Sergi, J. Sweeney, S. Bishop, R. Kaim, D. Eldridge, Tungsten Transport in an Ion Source, IIT Proceedings, 2008, pp. 356-359.

FOR MORE INFORMATION Please call your Regional Customer Service Center today to learn what Entegris can do for you. Visit entegris.com and select the Contact Us link to find the customer service center nearest you.

TERMS AND CONDITIONS OF SALE All purchases are subject to Entegris’ Terms and Conditions of Sale. To view and print this information, visit entegris.com and select the Terms & Conditions link in the footer.

Corporate Headquarters Customer Service 129 Concord Road Tel +1 952 556 4181 Billerica, MA 01821 Fax +1 952 556 8022 USA Toll Free 800 394 4083

Entegris®, the Entegris Rings Design®, and other product names are trademarks of Entegris, Inc. as listed on entegris.com/trademarks. All third-party product names, logos, and company names are trademarks or registered trademarks of their respective owners. Use of them does not imply any affiliation, sponsorship, or endorsement by the trademark owner. ©2018 Entegris, Inc. | All rights reserved. | Printed in the USA | 9000-10263ENT-1118

www.entegris.com