**5. Read Disturb**

It is known that high temperature can accelerate the speed of lateral charge migration in the storage layer and modulate the threshold voltage distributions of memory cells [8]. These factors can cause read disturb properties to become more complex at various temperatures. To understand the temperature impacts, we designed the following experiment: firstly, setting the work temperature of the chamber to the target temperature ranging from −30 to 70 ◦C; then, programming randomized data with subsequent 3000 times read cycling. For detail analysis, data were dumped and recorded every 100 times.

Measured read disturbs are summarized in Figure 5. Firstly, it was observed that RBER degradation turned out to be much more serious at cold temperatures; secondly, for high-temperature operations at 70 ◦C, a part of RBER can be recovered after serval times reading. Considering that the total RBER included two parts, down-shift errors from the charge loss and up-shift errors from charge accrual, we divided the total error bits to two groups for in-depth analysis: down-shift errors and up-shift errors. As shown in Figure 6, for read disturb-related RBER degradations, down-shift errors were the dominant part with clear temperature dependences, indicating that RBER changes mainly originated from the charge loss. Down-shift error degradation was much stronger at sub −25 ◦C, but it could be well suppressed at high temperatures, which can be explained by the narrower V*th* distributions at high temperatures [16]. The interesting phenomenon was that read disturbs from up-shift errors showed the opposite tendency while increasing the operating temperature. For read cycling at 70 ◦C, up-shift error bits can be partly recovered with read cycles. It was noticed that the decreasing error bits were mainly from cells with high program levels, like F-to-G errors in F-level cells. It should be noted that, as shown in Figure 6f, lower F-to-G error bits can be observed in the whole temperature range from −30 to 70 ◦C. However, G-to-F up-shift error bits are largely suppressed at 70 ◦C. Thus, with combined down-shift and up-shift errors, we observed abnormal "recovery" at 70 ◦C while performing read cycling.

**Figure 5.** Read disturb characterizations at different temperatures from −30 to 70 ◦C.

For further understandings, the word-line (WL) dependences of fail bit counts (FBCs) change at 70 ◦C were studied in detail. By comparing the data from the 1st and 3000th read cycles, it was observed that the dependence of the major state error decreased on the WL index. As shown in Figure 7, error bits from D-E, E-F, and F-G errors showed obvious decreasing trends in higher WL index, and each read cycle in this experiment followed the same observation. In other words, the WLs of the middle-to-low index were the dominant origins for the lower up-shift errors that were attributed to the observed error bits "recovery" at 70 ◦C.

**Figure 6.** Read disturb-related RBER changes were divided into (**a**) down-shift errors and (**d**) up-shift errors from −30 to 70 ◦C; (**b**,**c**) compares B-to-A errors and G-to-F down-shift errors, respectively, while (**e**,**f**) compares A-to-B errors and F-to-G up-shift errors, respectively, at −30, 25, and 70 ◦C.

**Figure 7.** Measured fail bit count (FBC) of different program levels: error bits from (**a**) D-to-E; (**b**) E-to-F; (**c**) F-to-G.

#### **6. Conclusions**

In this work, to achieve deep insights into the temperature impacts on the reliability properties of the 3D NAND flash, the TLC (3 bits/cell) 3D CT NAND flash memory chip was tested from −30 to 70 ◦C using the FPGA-based raw NAND chip tester together with the temperature-controllable chamber. With comprehensive characterizations, firstly, it was observed that program time had a clear dependence on both temperature and P/E cycles by which the TCPT model was proposed; secondly, it was found that RBER can be well suppressed at high temperatures and it degrades obviously at low temperature; then, by the designed cross-temperature measurements, it was found that thermal experience had negligible impacts on RBER degradation; finally, as for read disturbs, it was concluded that read disturbs cause more RBER degradations at cold temperatures while part of RBER can be recovered by read disturbs at high temperatures.

**Author Contributions:** The work presented here was completed in collaboration between all authors. Conceptualization, F.C. and B.C.; measurements and validation, F.C. and X.L.; simulations, F.C. and H.L.; formal analysis, F.C.; investigation, F.C., B.C., H.L. and Y.K.; writing—original draft preparation, F.C.; writing—review and editing, J.C.; supervision, J.C.; project administration, J.C.; funding acquisition, X.Z. and J.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (No. 62034006 and 91964105), the National Key Research and Development Program of China (No. 2016YFA0201802), and the Natural Science Foundation of Shandong Province (No. ZR2020JQ28 and ZR2019LZH009).

**Conflicts of Interest:** The authors declare that there is no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Micromachines* Editorial Office E-mail: micromachines@mdpi.com www.mdpi.com/journal/micromachines

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com