Charge Transport during High-Temperature Baking

After discussing the charge transport during ±FN operations and FN cycling endurance test, we next investigate the charge transport during short-term (< 1 day) and long-term (> 1 day) high-temperature baking. In addition to charge loss, the intra-nitride transport behaviors are also introduced.

5.3.1 Baking Characteristics of SONOS and BE-SONOS

The 200^oC and 250^oC baking characteristics of SONOS (54/70/90Å) and BE-SONOS (15/20/25/70/90Å) are shown in Fig. 5.7(a) and (b), respectively. The devices are first programmed by +20V 0.26sec before high-temperature baking. For SONOS, the CS capacitor shows slightly VFB gain while GS capacitor shows slightly VFB loss within 1 day baking. The extracted Q and x are shown in Fig. 5.7(c) and (d). For SONOS, there is almost no real charge loss after baking for ~ 1 day (Q is unchanged), but there is a significant decrease in the charge centroid (x). Therefore, the V_FB shift must be caused by trapped electrons move to lower portion of nitride. This is an evidence of intra-nitride transport. However, after long-term baking (> 1 day), significant real charge loss is observed (Q is decreased) while x shifts higher.

This indicates that electrons mainly de-trap from the bottom portion of nitride after long-term baking. Since the B.O. of SONOS is thick enough to block the trap to band tunneling and the amount of charge loss increases as the rising temperature, we speculate that this charge loss comes from thermal emission [5.8]. The higher the baking temperature, the more significant charge loss and x shift (temperature dependent de-trapping rate). BE-SONOS also shows similar behaviors as SONOS. However, in the first day of baking, we observe both intra-nitride transport and charge loss simultaneously. This may due to BE-SONOS have more staring stored charges that build a higher internal e-field and cause more charges loss.

It is interesting to note that the x-Q plots in Fig. 5.7(d) are very similar for different baking temperatures, implying that for electrons at location x the tunneling probability is

independent of temperature. Moreover, de-trapping is most efficient near the bottom portion of nitride since the x increases during de-trapping.

We also investigated the post-cycling baking characteristics of BE-SONOS (15/20/25/70/

90Å), as shown in Fig. 5.8(a). The devices are first programmed by +19V after different P/E cycles (Fig. 5.6(a)), and then we measure the retention data at 250^oC. From Fig. 5.8(a), BE-SONOS shows similar retention behaviors for post-cycled and fresh devices. Within 1 day baking, the devices after P/E cycling have larger initial VFB loss. However, after long-term baking, the V_FB loss is almost the same as the fresh state. Therefore, we conclude that the P/E cycling stress only causes more initial charge loss, but it has no effect on the long-term retention characteristics.

The extracted Q-t, x-t, and x-Q are shown in Fig. 5.8(b) and (c). At long-term baking, the Q-t and x-t of different cycle times merge together. On the other hand, the x-Q plots are very similar for different cycling times. This suggests that P/E cycling does not affect electron de-trapping trajectory at high-temperature baking.

5.3.2 Baking Characteristics of SONOS with a Thicker Nitride

In order to enhance the signal of intra-nitride transport, we have investigated the high-temperature baking retention characteristics (250^oC) of SONOS with a thicker nitride (O/N/O = 70/95/75Å), as shown in Fig. 5.9(a). Within 1 day baking, CS device shows obvious V_FB gain while GS device shows V_FB loss. This opposite shift of CS and GS devices clearly implies that there is intra-nitride transport inside nitride.

The extracted Q and x are shown in Fig. 5.9(b) and (c). Q is almost unchanged at t <

10⁴sec while x is significantly decreased. This implies that the total charge inside nitride is unchanged, but the trapped electrons move to lower portion of nitride, i.e. intra-nitride transport. The thicker nitride shows a slightly more significant change in x compared to the

thinner nitride. The exact mechanism still needs to be examined. It should be mentioned that this intra-nitride transport is observable only at very high-temperature baking (> 200^oC). At lower baking temperature (150^oC), the intra-nitride transport behavior is very minor. Fig.

5.10(a) shows that there is no significant V_FB shift for both CS and GS capacitors after long-term baking. The extracted Q and x in Fig. 5.10(b) and (c) also indicates that the change is minor.

We also investigated the room temperature intra-nitride transport under low bias voltage (±5V) and long time (1×10⁵sec) stress. We first programmed the devices to high V_FB states (+21V, 0.26sec), and then applied ±5V stress. The extracted Q and x are shown in Fig. 5.11.

The results show that under +5V stressing (< 5MV/cm in B.O.), the Q and x are almost unchanged within 1×10⁵sec. Moreover, under –5V stressing (> 5MV/cm in B.O.), the x is slightly decreased after 1×10³sec stressing, indicating that charges inside nitride are very stable. Therefore, we conclude that charge spreading inside nitride is not the dominant retention mechanism for SONOS-type devices at lower baking temperature (< 150^oC) or under moderate internal e-field. Hence, nitride trap provides a very reliable charge storage material for non-volatile memory applications.

5.4 Summary

In this chapter, we could directly investigate the charge transport and intra-nitride behaviors of SONOS-type devices by using GSCS method. Using this novel method, we could monitor the charge centroid (x) and charge density (Q) during various programming/erasing and reliability tests.

Our results clearly indicate that for the electron injection (+FN program), the electron centroid migrates from the bottom interface toward the center of nitride. For the hole injection (–FN erase) in SONOS with a thin B.O. or BE-SONOS, holes first recombine with the bottom electrons and then gradually move upward. We also proved that electron and hole injection

centroids have vertical mismatch after erasing. However, this mismatch of the electron and hole centroids should be viewed as a snapshot of erased state, and should not be interpreted as two separate pockets of charges co-inhabiting the nitride. During P/E cycling, most trapped electrons/holes are neutralized by the erasing holes/programming electrons. Therefore, SONOS-type devices can still possess excellent P/E cycling endurance.

On the other hand, for the electron de-trapping processes under –V_G stressing (SONOS with a thicker B.O.), the trapped electrons de-trap first from the bottom portion of nitride.

For the high-temperature retention, after short-term baking, the trapped electrons move to lower portion of nitride, and this intra-nitride transport becomes more significant for a thicker nitride. On the other hand, after long-term baking, the charge loss mainly comes from the bottom portion of nitride. It should be mentioned that the intra-nitride transport is significant only at very high-temperature baking (> 200^oC). The charge vertical or lateral spreading may not be the dominant retention mechanism at lower storage temperature.

In summary, our GSCS method provides numerous new observations of intra-nitride charge-trapping behaviors, which were never discovered before.

+FN program

8x10¹² 10x10¹² 12x10¹²

x (Angs trom )

electron centroid gradually migrates toward the center of nitride. After –VG stressing, the electron density decreases while the electron centroid moves upward. It indicates that electrons mainly de-trap from the bottom portion of nitride.

x-Q

Q (e/cm ² )

6x10

¹²

9x10

¹²

12x10

¹²

x (An g s tro m )

0 10 20 30 40 50 60 70

54A: -12V 54A: -14V 70A: -15V 70A: -16V 90A: -16V 90A: -17V

Nitride/B.O. Interface Nitride/T.O. Interface

Diff B.O.

N/T.O.=70/90A

Fig. 5.2 X-Q plots of various SONOS devices (B.O. = 54Å, 70Å, and 90Å) during –VG

stressing. Thicker B.O. shows less x-variation at first. At long-term stress, the x variations are becoming identical for all samples.

Time (sec)

erasing injected holes first recombine with the bottom electrons and then gradually move upward, thus causing the upward motion of the charge centroid.

Time (sec)

Fig. 5.4 (a) –FN hole injection characteristics of BE-SONOS (13/20/25/70/90Å). (b) Q-t and (c) x-t/x-Q plots. The x-Q plot is transformed by Q-t and x-t plots. The injected holes first recombine with the bottom electrons.

Q_h-t

Time (sec)

10^-6 10^-5 10^-4 10^-3 10^-2 10^-1 10⁰ Q h (hole-cm-2 )

0 2x10¹² 4x10¹² 6x10¹² 8x10¹² 10x10¹²

-13 V -14 V -15 V

x_h-Q_h

Q_h (hole-cm^-2)

0 5x10¹² 10x10¹²

x h (Angstrom)

0 10 20 30 40 50 60 70

-13 V -14 V -15 V Nitride/T.O. Interface

Nitride/B.O. Interface

(a) (b)

Fig. 5.5 (a) Qh-t and (b) xh-Qh plots of BE-SONOS (13/20/25/70/90Å) during –FN erasing by using Eqs. (E-5, E-6). Hole centroid also starts at the bottom interface. After longer injection, it gradually migrates upward. However, its centroid is much lower than that of trapped electrons.

P/E Cycles

P/E

10⁰ 10¹ 10² 10³ 10⁴

∆∆∆∆ V

FB

(V)

1 2 3 4 5

GS: +20V 5 msec CS: +20V 5 msec

GS: -16V 20 msec CS: -16V 20 msec

(a)

P/E Cycles

P/E

10

⁰

10

¹

10

²

10

³

10

⁴

x ( A n g st ro m )

20 30 40 50 60 70

Q ( e/cm

2

)

4x10¹² 6x10¹² 8x10¹² 10x10¹²

Q: +20V 5 msec x: +20V 5 msec Q: -16V 20 msec x: -16V 20 msec

(b)

Fig. 5.6 (a) 10K P/E cycling endurance of BE-SONOS (15/20/25/70/90Å). (b) Extracted Q and x. After +FN programming, the Q (electrons) increases, and x is close to the nitride center. After –FN hole injection, Q (still more electrons since Q > 0) decreases, and x shifts higher.

200

^o

C Baking

different baking temperatures are similar.

250^oC Post-cycle Baking of BE-SONOS

The devices are programmed by +19V before baking. (b) Q-t and (c) x-t/x-Q plots.

The x-Q plot is transformed from Q-t and x-t plots. P/E cycling stress only causes more initial charge loss, but it has no effect on the long-term retention.

250^oC Baking of SONOS (70/95/75) programmed by +20V 0.26sec or +21V 0.26sec before baking. Within 1 day baking, CS device shows obvious VFB gain while GS device shows VFB loss. (b) Q-t and (c) x-t plots. During 1 day baking, Q is almost unchanged while x is significantly decreased. This indicates that electron moves from the top portion toward the bottom portion.

150^oC Baking of SONOS (70/95/75) is much smaller than that in Fig. 5.9(c).

Q-t

Time (sec)

10^-1 10⁰ 10¹ 10² 10³ 10⁴ 10⁵

Q (e /c m

2

)

9x10¹² 10x10¹² 11x10¹² 12x10¹² 13x10¹² 14x10¹²

- 5 V + 5 V

x-t

Time (sec)

10^-1 10⁰ 10¹ 10² 10³ 10⁴ 10⁵

x ( A n g st ro m )

0 10 20 30 40 50 60 70 80 90

Nitride/B.O. Interface Nitride/T.O. Interface

- 5 V + 5 V

(a) (b)

Fig. 5.11 (a) Q-t and (b) x-t plots of SONOS (ONO = 70/95/75Å) during ±5V gate stressing at 25^oC. The Q and x are very steady at room temperature.

在文檔中 以閘極感應與通道感應方法與脈衝電流電壓技術分析SONOS類型元件中捕捉電荷之特性 (頁 136-151)