Journal of ELECTRONIC MATERIALS, Vol. 43, No. 11, 2014

DOI: 10.1007/s11664-014-3336-6 Ó 2014 The Minerals, Metals & Materials Society

Study of Shallow Backside Junctions for Backside Illumination of CMOS Image Sensors CHUNG SEOK CHOI,1 SANG CHUL YEO,2 DOHWAN KIM,1 JONGCHAE KIM,1 KYUNG DONG YOO,1 and HYUCK MO LEE2,3 1.—CIS Process Integration Team, Image Development Group, System IC Division, 215 Daesuinro, Heungdeok-gu, Cheongju-si, Chungcheongbuk-do 361-725, Republic of Korea. 2.—Department of Materials Science and Engineering, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 305-701, Republic of Korea. 3.—e-mail: [email protected]

Backside illumination complementary metal oxide semiconductor image sensors (BSI CISs) represent an advanced technology that produces high-quality image sensors. However, BSI CISs are limited by high dark signals and noise signals on the backside. To address these problems, backside junctions are commonly used. High-dose backside junctions effectively reduce dark signals and noise signals. The depth of the implantation profile is a key factor in determining the junction depth. A laser thermal annealing process is conducted only near the surface to the activation, and thus broader doping profiles are limitations to be activation of dopants. Changing the dopant from B to BF2 can decrease the implant projected range. However, there are abnormal activation rates for BF2 in applications involving laser thermal annealing processes for shallow junctions. Although the need for BF2 is increasing, a mechanism for its slow activation and low activation rates has not yet been confirmed. Here, we identify the mechanism by which BF2 undergoes low activation after a melting threshold temperature and explain why this phenomenon occurs. In addition, we confirm a condition that provides high activation rates of BF2 and show the reduction of dark signals and noise signals at the high density BSI CISs. Key words: CMOS image sensor, backside illumination, DFT, backside junction, BF2

INTRODUCTION Complementary metal oxide semiconductor (CMOS) technology can produce features down to a few tens of nanometers. For the development of CMOS technology, shallow junctions are essential. In the 1980s, laser thermal annealing (LTA) was investigated by a number of researchers.1,2 Since then, LTA has emerged as a promising technique for fabricating shallow junctions.3,4 LTA is used in both scientific research and industry. Backside illumination CMOS image sensors (BSI CISs) and thick film transistors in light-emitting diodes (LEDs) are common in the LTA industry. BSI CISs technologies (Received March 18, 2014; accepted July 18, 2014; published online August 14, 2014)

can be improved by shaping the backside junctions using LTA, which reduces dark signals and noise signals. More importantly, LTA protects buried structures from thermal budget effects during thermal processing. Implantations make end of range (EOR) defects near implantation tails, and low LTA energy has limited areas where crystal defects were cured. To address these problems, backside junctions must be formed by shallow P-type doping using high activation concentrations and abrupt doping profiles. Buffering layers such as silicon oxide are commonly used to reduce the projected range during the implantation process to fabricate shallow backside junctions. However, different melting temperatures between Si and SiO2 leads to materials defects such as dislocation.5 In addition, low energy implantation of 11B does not 3933

3934

Choi, Yeo, D. Kim, J. Kim, Yoo, and Lee

EXPERIMENTAL PROCEDURES Slow BF2 Activation

Analysis of the Mechanism of Slow BF2 Activation We performed calculations using the Vienna ab initio simulation package (VASP), a plane wave code that employs a projected augmented wave (PAW).9,10 The spin-polarized density–functional theory (DFT) calculation was performed using the generalized gradient approximation (GGA) for the exchange–correlation functional according to the Perdew, Burke and Ernzerhof (PBE) method.11 For the bulk Si, we employed a (2 9 2 9 2) supercell containing 64 atoms. The bulk Si was then allowed to relax. An energy cutoff of 280 eV and a (5 9 5 9 5) Monkhorst–Pack grid were used for k-point sampling. Our calculation yielded a Si lat˚ . To perform an electronic tice parameter of 5.46 A structure analysis, we used Bader charge analysis between each species. To examine the effect of surface coverage, we used a pre-covered nitrogen atom on bulk Si. Condition for Fast BF2 Activation To confirm the increase in laser energy density, we implanted 49BF2 30 keV 3E12 atoms/cm2, 49BF2 30 keV 4E13 atoms/cm2, 49BF2 30 keV 4E14 atoms/cm2, 49BF2 30 keV 3E15 atoms/cm2,

Table I. Test conditions of dark signals and row noise signals

Integration time Temperature Digital offset Frame rate Analog gain Digital gain Capture frame

Dark signals

Row noise signals

0.2 ms 60°C 0 code 60 frames/s 98 92 1

125 ms 60°C 64 code 60 frames/s 98 91 100

24 22 20 18 16 14 12 10 8 6 4 2 0

11B 5keV 4e13(RMS) 49BF2 20keV 4e13(RMS) 11B 5keV 4e13(Resistivity) 49BF2 20keV 4e13(Resistivity)

10000

1000

Resistivity Ω/square

We implanted 11B 5 keV 4E13 atoms/cm2 and 49BF2 20 keV 4E13 atoms/cm2 on (100) P-type Si substrates and pulsed a 306-nm excimer laser for 160 ns (1.5–3.0 J/cm2). The roughness mean square (RMS) was measured using atomic force microscopy ˚) (AFM; Dimension Icon). The maximum RMS (A corresponds to near the threshold melting energy density (TMED).8 To confirm B activation, fourpoint probes were used for measuring sheet resistivity.

and 49BF2 30 keV 3E16 atoms/cm2 on (100) P-type Si substrates and pulsed a 306-nm excimer laser for 160 ns (1.7–3.0 J/cm2). We used transmission electron microscopy (TEM; FEI Titan G2 80–300 kV) to confirm the pre-melting of the Si and to identify any physical defects. X-ray photoelectron spectroscopy (XPS; AXIS/NOVA) was conducted to define the chemical bonding after laser annealing.

RMS( )

provide stable conditions. The minimum implantation energy is at least 3 keV. Altering the P-type dopant from B to BF2 is ideal for reducing the implantation tail and forming stable implantation energy conditions. Fluorine can also reduce the transient enhanced diffusion, which is a reason for broader doping profiles and low activation rate.6 However, the activation of BF2 requires a higher LTA energy density than does the activation of B, which is considered as the slow activation rate of BF2 in this paper. This slow activation property of BF2 leads to a broader doping profile and thermal budget effects on pre-fabricated metal lines.7 Therefore, the use of a high LTA energy density limits the usefulness of BF2 implantation. In this paper, we describe the principles underlying slow BF2 activations, and demonstrate the fabrication of a shallow backside junction using BF2, and the reduction of dark signals and noise signals.

100 1.8

2.0

2.2

2.4

2.6

2.8

3.0

LTA(J/cm2)

˚ ) and sheet resistivity for BF2 and B Fig. 1. Comparison of RMS (A implantation for increasing LTA energy density.

Table II. Total energy, formation energy, and bonding length for various bonding types of Si, B, and F Total energy (eV) SiI-B SiI-F SiV-B SiV-F SiI-F-B-SiI SiV-F-B-SiV Si bulk Si bulk-1vac

350.07 350.26 347.54 342.92 354.52 348.68 347.19 338.12

Formation energy (eV) 2.60 2.54 9.14 4.27 6.51 9.74 – –

Bonding ˚) Length (A 2.19 2.19 2.09 2.39 1.44 2.33 2.37 2.37

Study of Shallow Backside Junctions for Backside Illumination of CMOS Image Sensors

3935

Fig. 2. Structures of (a) SiI-B, (b) SiI-F, (c) SiV-B, (d) SiV-F, (e) SiI-F-B-SiI, and (f) SiV-F-B-SiV.

Table III. Sheet resistivity of BF2 implantation conditions with increasing LTA energy

BF2 BF2 BF2 BF2 BF2

30 30 30 30 30

keV keV keV keV keV

3E12 4E13 4E14 3E15 3E16

1.7 J/cm2

1.8 J/cm2

1.9 J/cm2

2.2 J/cm2

2.4 J/cm2

2.6 J/cm2

3.0 J/cm2

85,071 83,570 8200 228 28

52,431 45,116 4086 165 25

17,217 15,763 748 75 17

8488 5819 436 60 11

4863 4081 368 66 10

4197 156 371 50 9

166 154 391 40 8

Analysis of White Clusters and Shallow BF2 Backside Junctions

Fabrication of High Density Backside Illumination CMOS Image Sensors

The material structures and elements were analyzed by high-resolution TEM (HRTEM) and energy filtered TEM (EF-TEM; FEI Titan G2 80–300 kV). Doping profiles were measured by secondary ion mass spectroscopy (SIMS; CAMERCA IMS6F).

The backside junction was fabricated in the 1.4lm high-density (HD) pixels BSI CISs at 200 mm wafer. We conducted two-step implantation for building backside junctions. The implantation dose for the backside junction were N2 10 keV 9E13

3936

Choi, Yeo, D. Kim, J. Kim, Yoo, and Lee

(a)

(b)

20nm

20nm

(d)

(e)

20nm

20nm

(c)

20nm

(f)

5nm

2

Fig. 3. TEM images after LTA at 1.7 J/cm for (a) BF2 4E13, (b) BF2 3E12, (c) BF2 4E14, (d) BF2 3E15, and (e) BF2 3E16. (f) HRTEM image of BF2 3E16.

atoms/cm2 and 49BF2 8 keV 1E14 atoms/cm2, respectively. The projected range (Rp) of implantation was calculated by the stopping and range of ions in matter (SRIM)-2006. In addition, LTA at 2.4 J/cm2 and H2N2 annealing 400°C for 150 min were followed. The dark signals and row noise signals were measured by a probe test (IP750EP test system). The dark signal is the amount of black level with respect to integration time at 0 lux. The row noise signal for an image camera is a measure of the temporal noise present in row averages, which manifests itself as flickering rows when imaging in low light conditions. The dark signal and row noise signal were measured by the standard mobile imaging architecture (SMIA) characterization specification, which is one of internal standards of mobile imaging architecture.12 Table I shows the test conditions of dark signals and row noise signals. RESULTS AND DISCUSSION Occurrence of Slow BF2 Activation To confirm the occurrence of slow BF2 activation, we implanted 11B 5 keV 4E13 atoms/cm2 and 49BF2 20 keV 4E13 atoms/cm2 on (100) P-type Si substrates and pulsed a 306-nm excimer laser for 160 ns (1.5– ˚ ) and sheet 3.0 J/cm2). Figure 1 displays the RMS (A ˚ resistivity. In each case, the RMS (A) reached its maximum value at TMED when initiative Si melting occurred, due to the non-uniform LTA energy distribution. In the case of BF2 doping, high sheet resistivity was observed, despite the fact that the energy density was higher than the TMED. In the case of BF2

˚ ) reached its maximum 4E13 atoms/cm2, the RMS (A 2 value at 1.8 J/cm , indicating that this point was the TMED. The TMED point is meaningful because dopants show 100% activation after completing the melting–recrystallization process.7,13 However, BF2 showed high sheet resistivity until it reached an energy density well over its TMED (2.5 J/cm2). Slow BF2 activation requires a high energy density. Consequently, BF2 does not fulfill the requirements for shaping a backside junction. To overcome this unfavorable property, we examined the principles determining slow activation. Principles of Slow BF2 Activation We conducted VASP and PAW simulation by using a (2 9 2 9 2) super-cell containing 64 atoms in order to study why BF2 shows slow activation. Table II shows the total energy, formation energy, and bonding length for the various structures according to their bonding types. The total energy values for bulk Si and bulk Si-1vac were 347.19 eV and 338.12 eV, respectively. In Fig. 2c, SiV-B shows B activation with the formation energy of 9.14 eV. However, in Fig. 2f, SiV-F-B-SiV shows the lowest formation energy ( 9.74 eV), which implies that SiV-F-B-SiV is the most stable structure for BF2 implantation. The SiV-F-B-SiV structure contained B and F in the vacancy sites of the Si. Consequently, the F-B bonding at the vacancy sites prevented B from bonding to the activation sites. This result indicates that additional energy would be required to break the SiV-F-B-SiV bonds to activate B. This additional energy must be

Study of Shallow Backside Junctions for Backside Illumination of CMOS Image Sensors

transferred only to SiV-F-B-SiV without thermal scattering, molecular photons, etc. As a result, supplemental energy is required to activate B after melting due to interstitial BF bonding.

3937

activation to be activated by additional energy after TMED. We implanted 49BF2 30 keV 3E12 atoms/ cm2, 49BF2 30 keV 4E13 atoms/cm2, 49BF2 30 keV 4E14 atoms/cm2, 49BF2 30 keV 3E15 atoms/cm2, and 49BF2 30 keV 3E16 atoms/cm2 on (100) P-type Si substrates and pulsed a 306-nm excimer laser for 160 ns (1.7–3.0 J/cm2) to confirm the effects of additional energy after TMED, because those conditions reduced Si melting temperature (about

Conditions for Fast BF2 Activation According to the DFT simulation, a greater laser energy annealing density is required for BF2

(a)

(b)

(c)

(d)

(100)

5nm

(e)

(f) Intensity (arb. units)

2100

Si-O-F

2000

687.3

1900 1800 1700 1600 1500 680

682

684

686

688

690

692

694

Bonding Energy(eV)

Fig. 4. (a) HTREM, (b) boron mapping, (c) oxygen mapping, (d) silicon mapping, (e) fluorine mapping, and (f) the XPS F1s peak for BF2 30 keV 3E16.

3938

Choi, Yeo, D. Kim, J. Kim, Yoo, and Lee

200°C) by transforming crystal Si to amorphous Si.14 Table III shows the changes in sheet resistivity that occurred for increasing concentrations of fluorine. For low concentrations of BF2 (3E12 atoms/ cm2 and 4E13 atoms/cm2), sheet resistivity drop point occurred at an energy density higher than the TMED. However, at a concentration of 4E14 atoms/ cm2, sheet resistivity dropped immediately after Si melting. Concentrations of 3E15 atoms/cm2 and 3E16 atoms/cm2 showed low sheet resistivity (228 X/square and 28 X/square, respectively, at 1.7 J/cm2). To observe the effect of high implant concentrations and implant energy, we analyzed the crystal structures by TEM. Figure 3 shows TEM images of the BF2 implantation conditions at an LTA of 1.7 J/cm2. Figure 3a and b shows that single-crystalline Si resulted in the case of low BF2 concentrations; however, Fig. 3c and d shows partial melting of the remaining amorphous Si and crystalline Si at the surface. Figure 3e shows that amorphous Si completely transformed to crystalline Si at a concentration of 3E16 atoms/cm2. The decrease in sheet resistivity drop point occurred above concentrations of 4E14 atoms/cm2, well matched with degrees of the melting and recrystallization process. By increasing the concentration of BF2 and the implantation energy, we transformed the Si from a crystalline to an amorphous state. Consequently, by reducing the Si melting temperature, we achieved an increase in

Figure 4 shows HRTEM and element mapping images obtained from the EF-TEM of the BF2 30 keV 3E16 atoms/cm2 sample. Figure 4a indicates a continuous (100) lattice plane between the white clusters and the Si substrate, with an absence of plane defects such as grain boundaries. B mapping in Fig. 4b shows that B aggregated at the center of the white clusters. Figure 4c–e demonstrates that the B was surrounded by Si, O, and F, indicating that Si, O, and F were located at the edges, with B situated in the center. In the XPS profile of F1s in Fig. 4f, a 687.3-eV peak can be observed at the edge of a cluster with Si-O-F bonding.15 Si-O-F clusters may promote B activation by allowing F-B bonds to

F1 Intensity (arb. units)

Intensity (arb. units)

390Å

530

White Clusters and Their Properties

(a)

O1s

525

the laser energy density for the transformation of SiV-F-B-SiV to SiV-B due to a decrease in the TMED. Therefore, at concentrations of 4E14 atoms/cm2, 3E15 atoms/cm2, and 3E16 atoms/cm2 (Table III), decreases in sheet resistivity drop point at low LTA energies resulted from the addition of sufficient energy to convert SiV-F-B-SiV to SiV-B. Although the BF2 30 keV 3E16 atoms/cm2 condition showed the lowest sheet resistivity (28 X/square) at the LTA energy, unintended white clusters appeared at the junction, as shown in Fig. 3f. Thus, we investigated whether the clusters negatively impacted the performance of the device.

535

540

680

(b) 390Å

687.3

685

687.6

690

695

Bonding Energy(eV)

Bonding Energy(eV)

O1s 530

(d) Intensity(arb. units)

Intensity (arb. units)

(c)

525

700

535

Bonding Energy(eV)

540

680

685

690

695

700

Bonding Energy(eV)

Fig. 5. Depth profiles for (a) BF2 30 keV 3E16 XPS O1s and (b) XPS F1s; (c) BF2 30 keV 4E13 XPS O1s and (d) XPS F1s.

Study of Shallow Backside Junctions for Backside Illumination of CMOS Image Sensors

3939

˚ ) at N2 55 keV 4E14 and 11B 5 keV 4E13 Fig. 6. Implantation of N2 55 keV 4E14, (a) no LTA, (b) LTA 1.4 J/cm2, (c) LTA 1.6 J/cm2, (d) RMS (A by laser annealing energy density.

be broken. Figure 5b shows XPS depth images for BF2 30 keV 3E16 atoms/cm2 and BF2 30 keV 4E13 atoms/cm2. As the depth increased, the F1s bonding energy shifted to the right and left of 687.3 eV due to Si-O-F precipitation.16 Figure 5a and b shows that the F1s peaks appear at the same ˚ ), indicating that O depth as the O1s peaks (390 A was influential in the appearance of the F clusters. Atmospheric O may react with F in BF2.16 As a result, the white clusters were primarily formed from B surrounding by O, F and Si. Importantly, the surrounding Si-O-F bonding may be acting as another noise source or impedance of hole mobility. However, low concentration implantation conditions (BF2 30 keV 4E14 atoms/cm2) did not produce white clusters. Therefore, we should use low concentration implantation conditions with a pre-amorphization process. Figure 6a shows amorphous Si after we implanted N2 55 keV 4E14 atoms/cm2. The amorphous Si was partially recrystallized at only

1.4 J/cm2 and totally recrystallized at 1.6 J/cm2, which indicated the reduction of TMED in Fig. 6b ˚ ) was increased at 1.4 J/cm2 and and c. The RMS (A maximized at 1.8 J/cm2 as shown in Fig. 6d. The 1.4 J/cm2 resulted from selective melting of the Si amorphous regions, and crystal Si melting led to ˚ ) at 1.8 J/cm2. Therefore, maximization of RMS (A we conducted a two-step implantation processes to build shallow and high activation backside junctions for preventing the formation of white clusters. The backside junction was fabricated in the HD BSI CISs at 200 mm wafer. The two-step implantations were N2 10 keV 9E13 atoms/cm2 and 49BF2 8 keV 1E14 atoms/cm2, respectively, and LTA 2.4 J/cm2 and H2N2 annealing 400°C for 150 min were followed. The additional low thermal budget annealing was conducted to reduce the residual crystalline defects that could not be annihilated by laser annealing.17 The N2 10 keV implantation energy was selected to approximately cover the

3940

Choi, Yeo, D. Kim, J. Kim, Yoo, and Lee

Fig. 7. (a) SIMS profile of 11B or 49BF2 implantation at photo diode of HD BSI CISs, (b) diffusion after H2N2 annealing at 400°C for 150 min, (c) dark signals at 11B 3 keV 1E14 and 49BF2 8 keV 1E14 implantation, (d) row noise signals at 11B 3 keV 1E14 and 49BF2 8 keV 1E14 implantation.

800Å

1E20

1100Å 2

B Concentration(atoms/cm2)

11B 3K 1E14+LTA 2.4J/cm

1E19 1E18 1E17 1E16 1E15 0

500

1000

1500

2000

2500

Fig. 9. Reproduced sample image with a back side junction made up of 49BF2 8 keV 1E14.

Depth(Å) Fig. 8. Melting depth with implantation 11B 3 keV 1E14 at LTA 2.4 J/cm2.

calculated Rp of 11B 3 keV implantation instead of 55 keV implantation energy. Figure 7a indicates boron doping depths on photodiode of HD BSI CISs after LTA at 2.4 J/cm2 and H2N2 annealing at 400°C for 150 min. The implantation tail at 49BF2 8 keV ˚ , but 11B 3 keV 1E14 atoms/cm2 was about 1500 A 2 ˚ . We defined 1E14 atoms/cm showed about 1700 A

the tail as a dose of 1E17 atoms/cm2. Figure 7b shows the boron diffusion with H2N2 annealing. ˚ after Boron was diffused as much as about 400 A H2N2 annealing. In the case of as-implanted 49BF2 8 keV 1E14 atoms/cm2, the implantation doping tail ˚ , while it was 1300–1500 A ˚ at was about 600–800 A 11B 3 keV 1E14 atoms/cm2. However, LTA curing ˚ depth at 2.4 J/cm2 was approximately 800–1100 A (Fig. 8). This means that the implantation doping tail at 11B 3 keV 1E14 atoms/cm2 could not be cured by doing LTA at 2.4 J/cm2, while it could be

Study of Shallow Backside Junctions for Backside Illumination of CMOS Image Sensors

Table IV. General description of HD BSI CISs HD BSI CISs Optical format Pixel size Active array format Frame rate Output format Operating temp. Supply voltage Digital I/O Digital core Analog & pixel

1/9-in. 1.4 lm 9 1.4 lm 1280H 9 720 V 60-fps @ full resolution RAW 8bit, 10bit 20°C to 60°C

3941

white clusters, which may act as another noise source and reasons for the reduction of hole mobility. Therefore, we conducted a pre-amorphization process for the reduction of TMED and showed a fast activation rate of BF2. As a result, the implantation doping tail was more cured at 49BF2 8 keV 1E14 atoms/cm2, and the backside junction made up of BF2 implantation led to reduction of 10% dark signals and 6% noise signals in HD BSI CISs. ACKNOWLEDGEMENT

1.7–3.0 V (1.8 V/2.8 V) 1.2 V/2.8 V 2.7–3.0 V (2.8 V)

This work was supported in part by National Research Foundation of Korea (NRF) grant funded by the Korean government (MEST) (2011-0028612) NRL. REFERENCES

cured at 49BF2 8 keV 1E14 atoms/cm2. Figure 7c shows the dark signals of a BSI photo diode (PD), an explanation for why BF2 implantation is required. In case of 49BF2 8 keV 1E14 atoms/cm2, the dark signals were reduced by 10% compared to 11B 3 keV 1E14 atoms/cm2. More importantly, the 6% reduction of row noise signals was accompanied by the reduction of dark sources (Fig. 7d). The decrease in dark signals and row noise signals might result from the reduction of dark sources near the implantation tail because dark sources could cause an increase in the dark signals and noise signals due to dangling bonds and defect sites on the surface that easily produce electrons despite dark light conditions.18 Furthermore, 8 keV implantation energy is quite stable compared to 3 keV. So it is useful to fabricate a backside junction made up of BF2 implantation for BSI CISs. In addition, Fig. 9c shows the sampled image of HD BSI CISs with the backside junction made up of 49BF2 8 keV 1E14 atoms/cm2. Table IV shows a general description of HD BSI CISs.

1. C.W. White, W.H. Christie, B.R. Appleton, S.R. Wilson, P.P. Pronko, and C.W. Magee, Appl. Phys. Lett. 33, 662 (1978). 2. J. Narayan, O.W. Holland, C.W. White, and R.T. Young, J. Appl. Phys. 55, 1125 (1984). 3. Z.A.F. Ali-Guerry, D. Dutartre, R. Beneyton, P. Normandon, and G.N. Lu, Sens. Lett. 9, 2137 (2011). 4. W.W. Luo, S.Z. Yang, P. Clancy, and M.O. Thompson, J. Appl. Phys. 90, 2262 (2001). 5. M.L. Geyselaers, J. Haisma, F.P. Widdershoven, T.M. Michielsen, and A.H. Reader, Appl. Phys. Lett. 54, 1311 (1989). 6. J.H. Park, Y.J. Huh, and H.S. Hwang, Appl. Phys. Lett. 74, 1248 (1999). 7. E.V. Monakhov, B.G. Svensson, M.K. Linnarsson, A. La Magna, M. Italia, V. Privitera, G. Fortunato, M. Cuscuna, and L. Mariucci, Mater. Sci. Eng. B: Solid State Mater. Adv. Technol. 124, 232 (2005). 8. K. Huet, C. Boniface, J. Venturini, Z.A.F. Ali-Guerry, R. Beneyton, M. Marty, D. Dutartre, and F. Roy, 18th International Conference on Advanced Thermal Processing of Semiconductor (RTP), ISSN 1944-0251 (2010), pp. 50–52. 9. D. Hobbs, G. Kresse, and J. Hafner, Phys. Rev. B 62, 11556 (2000). 10. G. Kresse and J. Furthmuller, Phys. Rev. B 54, 11169 (1996). 11. J. Paier, R. Hirschl, M. Marsman, and G. Kresse, J. Chem. Phys. 122, 234102 (2005). 12. SMIA, SMIA 1.0 Part 5: Carmera Characterization Specification Rev A (2004). 13. H.C. Cheng, F.S. Wang, Y.F. Huang, C.Y. Huang, and M.J. Tsai, J. Electrochem. Soc. 142, 3574 (1995). 14. M.O. Thompson, G.J. Galvin, and J.W. Mayer, Phys. Rev. Lett. 52, 2360 (1984). 15. K. Hanamoto, H. Yoshimoto, T. Hosono, A. Hirai, Y. Kido, Y. Nakayama, and R. Kaigawa, Nucl. Instrum. Methods Phys. Res. Sect. B: Beam Interact. Mater. Atoms 140, 124 (1998). 16. J. Wu, Y.L. Wang, C.P. Liu, S.C. Chang, C.T. Kuo, and C. Ay, Thin Solid Films 447, 599 (2004). 17. K.K. Ong, K.L. Pey, P.S. Lee, A.T.S. Wee, X.C. Wang, C.H. Tung, L.J. Tang, and Y.F. Chong, Appl. Phys. Lett. 89, 122113 (2006). 18. R.C. Westhoff, B.E. Burke, H.R. Clack, A.H. Loomis, D.J. Young, J.A. Gregory, and R.K. Reich, Proceedings of the SPIE, Sensors, Cameras, and Systems for Industrial/Scientific Application X Conference (San Jose, CA: Convention Center, 2009), vol. 7249, pp. 724900J1–724900J11.

CONCLUSIONS We analyzed the mechanisms by which BF2 doping produces slow activation and built a shallow backside junction in a BSI CISs. Slow BF2 activation was found to result from SiV-F-B-SiV bonding. To activate B 4E14 atoms/cm2, additional energy should be injected to transform SiV-F-B-SiV to SiV-B without energy loss. To increase the effective energy density transferred to SiV-F-B-SiV, the melting threshold temperature can be reduced instead than increasing the energy density. This strategy is favorable because an increased energy density can lead to broad doping profiles and the thermal budget effect on buried structures. High-dose BF2 leads to