Improved Ion Source Stability Using H2 Co-gas for Fluoride Based Dopants

by Tseh-Jen Hsieh and Neil Colvin
Axcelis Technologies, Inc., Beverly, MA

Introduction

It has been shown [1] that the common fluoride compounds such as GeF4 and SiF4 used for pre-amorphization (PAI) species and BF3 as a source gas for B11 and BF2 P-type dopant create problems not only for ion source operation but also impact device performance.

In recent years significant advances have been made in the development of high mass molecular (HMM) beam sources for dopant implantations into silicon. The driver for the development of these sources has been the need for very low energy implants where energy is partitioned between the atoms of a molecule in direct proportion to their mass. BF2+ is the most the widely used molecular ion with an atomic mass of 49 and has a single boron atomic mass of 11 that result in equivalent implant energy of 49/11[2].  This allows higher energy implants with improved beam transport efficiencies and also suppresses boron diffusion during the anneal step [3].

At the gate edge where BF2 is used for several implants such as Source-Drain Extension (SDE), Source-Drain (SD), PAI, co-implanted species, halos and pockets, small variations in any one of these can lead to unacceptable fluctuations in device performance. The precision required to perform these implants therefore becomes critical. As an example, diffusion in SDE and halo implant is strongly influenced by damage generated during SD implantation, the understanding of defect generation as a function of beam characteristics of the ion implanter is critical in the process integration of advanced devices 65nm and beyond.

This study investigates the characteristics of BF3 when it is ionized and the positive effects of using hydrogen as a co-gas. The quantity of hydrogen fluoride (HFx) formed during the recipe optimization is aided by monitoring the AMU spectra and required dopant beam current level. The production and extraction of these F- radicals from the ion source and through the graphite extraction optics forms CFx vapors which will encompass the entire optics assembly attacking/coating critical high voltage insulators causing beam instabilities.

The production of HF is the result of the H+ and F- reaction in the ion source plasma which reduces the probability of free fluorine radicals attacking the materials from which the ion source/extraction electrode is constructed (tungsten, graphite, molybdenum, Al2O3). Due to the formation of these unwanted fluorine based radicals (WF6, CFx, AlFx) system design may not always prevent damage to critical source components and the mass resolution of the bending magnet with a resolving aperture may not have a resolution of 1 amu to exclude contaminants such as CF2. (amu 50). These radical gas compounds may also lead to increased wafer contamination over time as they transport and deposit down the beam line.  Formation of WF6 in a high temperature vacuum environment (arc chamber) permits the halogen cycle to promulgate which further shortens source life and stability. Controlling the amount of free F- can translate into reduced glitching, wafer repaints, improved source life and metal contamination.

Experimental

The design of experiments was conducted in two parts using the same tool where a baseline test without the co-gas was performed and after rebuilding the source and electrode the only variable changed was the addition of a previously optimized hydrogen co-gas flow. The GSD 200 E2 was used to conduct the tests using a customer’s high volume production BF2 recipe. The beam recipe was tuned without the hydrogen co-gas, and then the H2 flow was optimized with regards to the amount of HFx formed and effect on the required beam current as shown in Figure 1.  The beam AMU spectra experiment indicates that the amount of HFx will reach a maximum and then saturate for a given H2 flow at a given set of beam tuning parameters.

Results and Discussion

The fluoride based dopant gases generate free fluorine radicals during ionization in the arc chamber cavity. The high chemical volatility of refractory fluorides at the operating temperatures has proven not only to be detrimental by etching ion source refractory metals, electrode graphite but also causing the erosion of strike plates for off energy species located in the AMU beam guide (downstream of the ion source). 

Regardless of what refractory metals are used to construct the arc chamber liners, thermal evaporation, sputtering, plasma and chemical etchings are several methods of converting refractory metals to the vapor phase. In the confines of the plasma charge neutrality is preserved and the specific density of fluorine radicals will react with tungsten ions to form WF6 with a melting point of 2°C[4], as it is volatile it will decompose on heated surfaces as part of the halogen cycle.  Due to the difference in ionization efficiency of GeF4 and BF3, the electro-negativity of Ge is 1.8 and F is 4.0 (delta = 2.2) and B is 2.0 (delta = 2.0), respectively [5].  The greater the difference in electronegativity, the higher the ionization efficiency is expected.  This is explains why the halogen cycle for BF3 is less severe than GeF4. One would think that the first step in reducing available W is to develop a recipe where the arc voltage is as low as possible for a given recipe as the sputter rate is proportional to the voltage, as shown in Figure 2, but it is a double edged sword as optimal beam current requirement may not be met. The reduced sputter rate of the cathode may exasperate the effects of the halogen cycle requiring higher H2 flows, but then again the formation of HFx may be at it saturation and increased H2 flow will have no effect.

Figure 1
Figure 1: Beam Optimization with and without Hydrogen Co-gas.

To deal with this problem, the key is to reduce the available free fluorine radicals that lead to the halogen cycle.  Under the ionization plasma condition there are equal amounts of positive ions and negative ions in the free space within the arc chamber.  The positive ions are B+, BF2+, Si+, Ge+, etc. and the negative ions are F-.  By introducing a pre-determined amount of hydrogen gas into the ionization chamber the highly reactive hydrogen ions (H+) will combine with fluorine ions forming the chemical compound hydrogen-fluoride (HFx). This reduction in WF6 will not only prolong the life of the refractory metal components, but also reduce amount of doubly charged WF6++ that can be transported down to the beam line to the wafer surface. Secondly, the trace amount of HFx will chemically etch residuals within the ion chamber inner surfaces and as it is passes through the extraction electrode apertures it will chemically etch those surfaces as well as any adjacent insulators (known as in-situ chemical cleaning). 

GeF4 Reaction Model: (with/without the presence of H2 co-gas)

GeF4(g) + 2H2(g)  → Ge+(s) + 4HF-(g) (1)  

 Or:              

GeF4(g)  → Ge+(s) + 4F- (g) (2)

 Or:

WF6(g)  ↔ W+(s) + 6F- (g)          (3)

As noted throughout this paper these free fluorine radicals

Figure 2
Figure 2: Cathode Erosion vs. Voltage Curves

attack tungsten surfaces to form WF6.  H2 works out to be a terrific choice as a co-gas to capture free fluorine radicals once they form in the arc chamber plasma.  Hydrogen is relatively inexpensive, non-toxic, highly reactive but relatively safe to handle.   When one GeF4 molecule decomposes, two hydrogen molecules (forms four hydrogen ions) are needed to chemically react with the four fluorine ions, as shown in eq.(1).   Whether tying up free fluorine when the dopant is cracked or if the halogen cycle releases fluorine back in to the plasma the H2 co-gas will minimize the quantity of tungsten consumed.

The approach has shown good beam stability, source lifetimes and minimal de-lamination of coatings in the arc chamber. Stable glitch free beams typically translate into lower overall particle performance.

The beam spectrum scan is a built in feature for most ion implanters and is used to demonstrate how much of the ionized material at selected m/e was extracted and transported to the beam Faraday and to verify composition of the extracted ion beam by displaying off energy ions. By introducing the hydrogen as a co-gas to the BF3 plasma, the chemistry changes completely because hydrogen is a strong reduction agent that reacts with the fluorine radicals, as indicated in equations (3) and (4) and (5).

4BF3(g) + 3H2(g)  → B + F + BF + 2BF2 + 6HF(g)           (4)

WF6(g) + 3H2(g) ® W+(s) + 6HF- (g)                      (5)

(With the presence of co-gas)   

To form WF6 it requires six free fluorine radicals which will then later de-compose back into tungsten onto heated surfaces.  The introduction of hydrogen as a reducing gas will form hydrofluoric acid (HFx).  The hazardous nature of chemicals like hydrofluoric acid indirectly helped improve process performance but as it is formed in trace amounts it is unlikely to be a device contamination issue or a safety concern during machine maintenance.

The benefits of H2 co-gas incorporation [6] can be summarized as follows:

3.1 Stable Source Operation: It has been demonstrated repeatedly in production environments that trace amounts of HFx produced as a by-product of tying up the F- radicals can chemically etch away the coating on the electrical isolation ceramic insulators so that breakdowns will be minimized and or eliminated. The high chemical volatility of refractory fluorides at ion source operating temperatures has been proven to be detrimental to the arc chamber components.  Any gas leakage through and away from the arc chamber tends to coat the insulators between the cathode and arc chamber potential leading to premature failure. Included in the design are patented seals to further minimize the leakage of these corrosive/conductive gases [7]. Best tuning methods and practices must be employed to minimize extraction electrode suppression and ground currents as that is the signature of the beam striking the extraction apertures. The formation of CFx from the graphite electrode apertures forms a conductive

Figure 3
Figure 3: Trace amount of WF1 (202), WF2 (221), WF3 (237), WF4 (257), and WF5 (282) can be observed with and without co-gas.

coating onto the high voltage insulators where the micro-discharges occur under high electrical field stress.  This discharging which is commonly referred to as glitching has been a problem since the introduction of the ion implant product. The perpetuation of these glitches is further magnified as device nodes shrink as they can be correlated to particles which can increase yield loss and lost productivity through repaints of the wafer. Test A with no H2 co-gas average number of glitches and high flyers was ~50/hour where Test B with H2 co-gas was ~1/hour.

3.2 Metal Reductions [8]: When the scattered light reaches a depleted silicon region of an image sensor any defect due to refractory metal contamination shows up as a bright spot. In terms of CMOS image sensors, the production of sensor wafers is similar to those used for devices such as NAND-flash memories and DRAM [9]. During the no co-gas ion source operation, the trace of tungsten chemical compounds: WF1 (202), WF2 (221), WF3 (237), WF4 (257), and WF5 (282)

Figure 4
Figure 4: The glitches count over numerous of implants where Test A is the baseline no co-gas and Test B is with co-gas.

Figure 5
Figure 5: Repeller with and without co-gas after 100 hours of beam operation with and without H2 co-gas.

can be observed, as shown in Figure 3. Even though it can be filtered by the bending magnet it has the tendency to move down the beam line possibly pushed along by the beam or by pressure differentials and it then may deposit onto the wafer as surface contamination.  The introduction of the hydrogen reducing co-gas not only stabilizes beam performance (reduced glitching) by the factor of ~50 glitches over hundreds of hours from Test A and B, as shown in Fig. 4 where glitches with co-gas stays fairly stable over a long period of wafer counts.  Also, the trace amount of refractory tungsten metal is reduced by a factor of 5 when the co-gas is introduced.  Figure 5 Test A shows the interior of arc chamber where the repeller has heavy deposits of tungsten indicating excessive halogen cycle (high formation of WF6) where Test B with H2 co-gas shows that the repeller has a slightly eroded surface indication minimal formation of WF6. Both tests had equal run times with all the same settings except H2 co-gas.

References

[1]    T. J. Hsieh, N. Colvin, S. Kondratenko, Enhanced life ion source for germanium and carbon ion implantation, AIP conf. Proc. 1496, Ion Implantation Technology 2012 372-375 (2012)

[2]    Bill Chang and Michael Ameen,  Axcelis Technologies, Inc., Beverly, USA; “High Mass Molecular Ion Implantation”; http://cdn.intechopen.com/pdfs-wm/17725.pdf

[3]    M. Y. L. Jung, R. Gunawan, R. D. Braatz, and E. G. Seebauerz, A Simplified Picture for Transient Enhanced Diffusion of Boron in Silicon, Journal of The Electrochemical Society, 151 (2004)

[4]    http://www.webelements.com/compounds/tungsten/tungsten_hexafluoride.html;  tungsten fluorides Physical properties.

[5]    Atsushi et al, “Ion generation device, ion irradiation device, and method of manufacturing a semiconductor device”, US Patent 5656820, Aug.12, 1997

[6]    Implementation of CO-Gases for Germanium and Boron Ion Implants, US patent 20120119113 A1, Patent Pending.

[7]    Special Seals; Patent Pending. 

[8]    ”Implementation reduction of refractory metals contamination on ion                              implantation”,  Patent Pending,

[9]    Jon Titus, Image-sensor defects can ruin an image, http://www.edn.com/electronics-news/4381021/Image-sensor-defects-can-ruin-an-image,  EDN Network, June 01, 2007.

High dose p-type 49BF2 and 11Boron can be a productivity challenge for conventional ion implanters used for semiconductor device applications.  As device geometries continuously shrink, there is a trend to deliver higher beam currents at lower energies, with a corresponding reduction in particles and metal contamination requirements. From this study, we have demonstrated there are many positive attributes for using H2 as a co-gas. The results have shown 50 x reductions in glitches for a BF2 beam throughout source life, reduced W metals contaminant by 5x, and improved ion source lifetimes from 200 hrs to 300 hrs for dedicated BF2.