Summary

We investigate boron enhanced diffusion phenomenon in SONOS memories.

Higher boron doping device has large on current, less leakage current, small variation, non-significant subthreshold swing degradation and faster program/erase speed. The better performances due to boron reduce 𝑁𝑖𝑆𝑖₂ formation energy and mismatch between 𝑁𝑖𝑆𝑖₂ and c-Si interface causing faster MILC rate and larger grain size. In order to further check boron enhanced diffuse effect, we extract source/drain resistance, overlap length and universal mobility and measure GIDL current to confirm this phenomenon again. The reliabilities are also analyzed by retention and endurance in both devices. The high boron doping device show longer retention time at high temperature than low doping device at the same operation voltage and sever endurance degradation for high overdrive voltage. In this thesis, we demonstrate that boron enhanced diffusion effect occurs in SONOS memories which improves device electrical characteristics.

39

5E14 1E15

𝑽_𝒕𝒉 (V) -4.3283 -3.3127

S.S.(mV/dec) 515 437

on/off ratio 𝟏𝟎^𝟒 𝟕 ∗ 𝟏𝟎^𝟓

Table 3.1. The SONOS memories characteristics were presented with MILC crystallization process.

𝑺𝒍𝒐𝒑𝒆 (𝒅𝒎 𝒅𝑳⁄ ) 𝝁_𝟎(universal) 𝐑_𝐒𝐃 (Ώ) ∆𝐋 (𝝁𝒎)

5E14 𝟔. 𝟔𝟐 ∗ 𝟏𝟎^𝟒 151 𝒄𝒎^𝟐/V.s 660 0.54

1E15 𝟓. 𝟐 ∗ 𝟏𝟎^𝟒 192 𝒄𝒎^𝟐/V.s 566 0.9

Table 3.2. The slope, universal mobility, source/drain resistance and overlap length for high and low boron doping devices.

40 𝑬_𝒂= 𝝏𝒍𝒏(𝒕_𝑹)

𝝏(𝟏/𝒌𝑻) ∆𝐕_𝐭 = 𝟎. 𝟎𝟓𝐕 ∆𝐕_𝐭 = 𝟎. 𝟏𝐕

5E14 0.496eV 0.545eV

1E15 0.704eV 0.709eV

Table 3.3. The of window losses in 0.05V and 0.1V are about 0.496eV, 0.545eV for boron doping concentration 5 ∗ 10¹⁴𝑐𝑚⁻² and 0.704eV, 0.709eV for boron doping concentration 1 ∗ 10¹⁵𝑐𝑚⁻².

Mechanism Activation

Energy (eV)

Tunnel oxide breakdown ≈0.3

ONO ≈0.35

Oxide defects ≈0.6

Cycling induced charge loss ≈1.1

Ionic contamination ≈1.2

Intrinsic charge loss ≈1.4

Table 3.4. Both high and low doping devices values approximate 0.6eV which charge loss mechanism are dominated by oxide defects [3.14].

41

Crystallize Size (Å )

(H K L) 5E14 1E15

(1 1 1) 479 1171

(2 2 0) 321 867

(3 1 1) 291 563

average 363.67 867

Table 3.5. The XRD result of boron doping concentrtion 5 ∗ 10¹⁴𝑐𝑚⁻² and 1 ∗ 10¹⁵𝑐𝑚⁻².

Fig. 3.1. The AFM graph of Si-NCs deposited on Si₃N₄ layer [3.4].

42

Fig. 3.2. The TEM graph of Si-NCs deposited on Si₃N₄ layer [3.4].

30 40 50 60 70

(311) (220)

(111) (111)

(220)

(311)

Intensity (a.u.)

*degree)

5E14 1E15

Fig. 3.3. X-ray diffraction (XRD) results of poly-silicon channel with different boron doping concentrations.

43

Fig. 3.4(a). The SONOS memories 𝐼_𝑑− 𝑉_𝑔 characteristics were compared between boron doping 5 ∗ 10¹⁴𝑐𝑚⁻² and 1 ∗ 10¹⁵𝑐𝑚⁻² devices with MILC crystallization process.

Fig. 3.4(b). The 𝐺_𝑚 is higher in doping 1 ∗ 10¹⁵𝑐𝑚⁻² device.

44

Fig. 3.5. The pelgrom plot shows devices variation. The high boron doping with slop 0.57 V/μm is smaller than low boron doping slop 1.1 V/μm.

Fig. 3.6. The subthreshold swing versus. channel length change for different boron doping devices. The high boron doping device shows better subthreshold leakage suppression as channel length beyond μm,.

45

Fig. 3.7(a). The memory window are 0.48, 1.31V and 1.84V as programming voltage 18V, 20V and 22V in boron doping 5 ∗ 10¹⁴𝑐𝑚⁻² device.

Fig. 3.7(b). The memory window are 0.5, 1.6V and 2V as programming voltage 18V, 20V and 22V in boron doping 1 ∗ 10¹⁵𝑐𝑚⁻² device.

46

Fig. 3.8(a). The memory window are 0.12 and 1.26V as erasing voltage 18V and 20V in boron doping 5 ∗ 10¹⁴𝑐𝑚⁻² device.

Fig. 3.8(b). The memory window are 1.1 and 2V as erasing voltage 18V and 20V in boron doping 1 ∗ 10¹⁵𝑐𝑚⁻² device.

47

Fig. 3.9(a). The programming speed curves show threshold voltage values ranging from 18V to 22V with filled and open symbols presenting low and high boron doping devices respectively.

Fig. 3.9(b). The erasing speed curves show threshold voltage values ranging from -18V to -20V with filled and open symbols presenting low and high boron doping devices respectively.

48

Fig. 3.10. The schematic of channel conduct in bottom gate device.

Fig. 3.11(a). The 𝐼_𝑑 − 𝑉_𝑔 curves of initial state, programmed state, forward read and reverse read in boron doping 5 ∗ 10¹⁴𝑐𝑚⁻² device.

49

Fig. 3.11(b). The 𝐼_𝑑 − 𝑉_𝑔 curves of initial state, programmed state, forward read and reverse read in boron doping 1 ∗ 10¹⁵𝑐𝑚⁻² device.

Fig. 3.12. The 𝑅_𝑆𝐷 and ∆L are gate bias dependent parameters which decrease as V_𝑔 increasing.

50

Fig. 3.13. The La Moneda method extracting R_SD and ∆L procedures involve three plotting which are discussed on channel length 1μm, 5μm and 10μm devices. The first procedure is generating the plots of 𝑹_𝒎 verse (𝑉_𝐺𝑆− 𝑉_𝑡)⁻¹. The boron dosage (a) 5 ∗ 10¹⁴𝑐𝑚⁻² and (b) 1 ∗ 10¹⁵𝑐𝑚⁻² devices.

51

Fig. 3.14. The second procedure is differential of R_m and m, so we can get R_SD values. The boron dosage (a) 5 ∗ 10¹⁴𝑐𝑚⁻² and (b) 1 ∗ 10¹⁵𝑐𝑚⁻² devices.

52

Fig. 3.15. The final procedure is differential the m and L, the intercept point on x-axis is overlap length. The boron dosage (a) 5 ∗ 10¹⁴𝑐𝑚⁻² and (b) 1 ∗ 10¹⁵𝑐𝑚⁻² devices.

53

-16 -12 -8 -4 0

0 40 80 120 160

W/L=10m/10m

Gate Voltage (V) Fieled Effect Mobility (cm

/V.s )

5E14 1E15

Fig. 3.16. Higher doping concentration device has bigger grain size and fewer scattering than lower boron concentration device leading to large mobility value.

54

Fig. 3.17. The top graph is energy band diagram for PMOS device. The bottom graph is electric field versus distance between Si/SiO₂ interface and drain [3.12]. The tunneling distance 𝑥_𝑡𝑢𝑛 is much less than the total depletion length.

Fig. 3.18. The electrons tunneling from the valence band subsequently leave via the substrate and the generated holes enter the drain creating the leakage current.

Gate

hole

55

Fig. 3.19. The boron doping (a) 5 ∗ 10¹⁴𝑐𝑚⁻² and (b) 1 ∗ 10¹⁵𝑐𝑚⁻² devices shows source and drain sides GIDL current. The filled simples are current from drain to gate and the open simples are current from source to gate.

56

Fig. 3.20. The average GIDL current for source and drain sides of two different boron doping (5 ∗ 10¹⁴𝑐𝑚⁻², 1 ∗ 10¹⁵𝑐𝑚⁻²) concentration before MILC process samples and one MILC process before boron doping 5 ∗ 10¹⁵𝑐𝑚⁻² concentration sample, respectively.

Fig. 3.21. T. Ma and M. Wing have reported MILC rate is found to increase at the highest dose of 3 ∗ 10¹⁵𝑐𝑚⁻² [3.1].

57

Fig. 3.22. The high and low doping devices have small charge loss percentages of 3.5% and 3.3% as time of stress increasing to 10⁴ seconds.

Fig. 3.23. The high and low boron doping devices have similar trends as constant temperature before 10² seconds and the loss percentages increase to 24% and 23.3%

as 10⁴ seconds.

58

Fig. 3.24. For temperature at 125℃, the high and low doping devices charge loss values are 44.3% and 62.6% at 10⁴ seconds.

Fig. 3.25. The Arrhenius plots of retention time characteristics for window dropping 0.05V and 0.1V.

59

Fig. 3.26. The device were programmed at 20V for 0.8s and erased at -20V for 1.6s. A large dropping in programmed threshold voltage and a slightly decrease in erased threshold voltage are observed in both (a) high and (b) low doping devices after 10³ P/E cycles.

60

Fig. 3.27. The ∆S. S. and ∆𝑉_𝑡ℎ versus the cycling numbers graphs. Meanwhile,

∆S. S. shows a similar trend to ∆𝑉_𝑡ℎ with an approximate relation of (a) ∆S. S.∗

10~∆𝑉_𝑡ℎ and (b) ∆S. S.∗ 2~∆𝑉_𝑡ℎ for high and low boron doping device. .

61

Chapter 4

Conclusion and Future Work