*3.3. Effect of Different Electrodes on P-E Ferroelectric Hysteresis*

As shown in Figure 6a,b, AlScN has typical ferroelectric properties. The maximum applied drive voltage ranges from 50 V to 82 V in steps of 4 V. The maximum voltage of the Pt sample was only up to 78 V. The sample was broken down after the voltage was increased to 84 V, whereas the Mo sample was broken down at 78 V. As can be seen from the hysteresis loop, the coercive field is approximately 3 MV/cm. However, the samples with two different bottom electrodes show completely different leakage behavior. The *Al*0.7*Sc*0.3*N* films grown on Mo have a large leakage current during the negative polarization. We speculate that this is due to the poor crystal quality and the huge amount of abnormal orientation grains, as the SEM image shows. Comparatively, the Pt sample also exhibits current asymmetry, but very weakly. The remanent polarizations of the two samples obtained from the hysteresis line test were approximately 100 µC/cm<sup>2</sup> and 350 µC/cm<sup>2</sup> , respectively.

**Figure 6.** Hysteresis lines of *Al*0.7*Sc*0.3*N* on two electrodes at different voltages. (**a**) Pt samples, with maximum remanent polarization, were approximately 100 µC/cm<sup>2</sup> and (**b**) Mo samples were approximately 350 µC/cm<sup>2</sup> .

In order to further analyze this leakage behavior, we measured the hysteresis loops at different frequencies and set the drive voltage just beyond the coercive field. For the Pt sample, ignoring the "gap" caused by electrode asymmetry, the P-E curves in the range of 1 K to 10 K show near standard ferroelectricity, as shown in Figure 7. When the frequency reaches 10 K, there is almost symmetry. However, the P-E curve of the Mo sample is not so good. Although the frequency is increased to a relatively high level so that the polarization does not switch repeatedly, it still exhibits significant asymmetry.

**Figure 7.** The hysteresis P-E loops with different frequencies of the *Al*0.7*Sc*0.3*N* on (**a**) Pt—as the frequency increases, the polarization flips incompletely and gradually becomes symmetrical—and (**b**) Mo—always asymmetrical.

Since the Mo sample exhibited asymmetry, we swapped the drive and sense terminals and obtained the electrical response as shown in Figure 8a. The maximum voltage applied was 50 V, which does not exceed the coercivity field, so the resulting current contains only the leakage current component and not the polarization current. The results show that there is a large leakage current of 1 mA in the negative direction only. Such a phenomenon could be attributed to different electrode materials, as well as asymmetric polarization hysteresis. Once the driving voltage exceeds the coercive field, the large leakage current on the Mo sample causes the hysteresis loop to completely deform, as illustrated in Figure 7b.

In Figure 8b, the peak currents of negative polarization at different frequencies are compared by extracting at 60 V. The measured current of the Mo sample is around 8 mA at a 1 kHz frequency, while showing currents over two times that of the Pt at all frequencies. Such a large leakage current will make the Mo electrode sample easier to break down when its polarization is reversed. On the other hand, the large leakage current makes it possible to output a stronger signal during polarization reversal, which greatly reduces the possibility of a loss of reading.

**Figure 8.** (**a**) Switching current with top and bottom drive under maximum 50 V triangle driving voltage. (**b**) The maximum voltage current as a function of driving frequency with the maximum voltage set to 60 V.

### *3.4. PUND Test to Obtain the Remanent Polarization*

A PUND test was performed to further analyze the ferroelectric properties of *Al*0.7*Sc*0.3*N* on both metals. As mentioned earlier, the main conditions that can be changed in the PUND test are *Vmax*, pulse width and pulse delay. *Vmax* just needs to be large enough to ensure that the polarization can switch. Therefore, we only changed the pulse width and pulse delay to see how the remanent polarization of the device changes.

First, *Vmax* was set to 60 V, pulse width to 0.5 ms, pulse delay to 10 ms and each test was subjected to 20 repetitions of the experiment, as shown in Figure 9a,b. An interesting phenomenon appears here: the remanent polarization in the negative direction of the Pt sample is larger for the first few times of the power-up test after resting at one end of the time, and then gradually decreases and stabilizes. This may be due to some parasitic parameters, which are subject to further analysis. In addition, the remanent polarization of the Pt sample is around 200 µC/cm<sup>2</sup> in the positive direction and 260 µC/cm<sup>2</sup> in the negative direction, a difference brought about by the Schottky contact between Pt and AlScN. The polarization of the Mo sample is very large, more than twice that of the Pt sample in both the positive and negative directions, which is in agreement with the polarization current pattern recorded earlier.

**Figure 9.** PUND measurements under different conditions. Repeat test at fixed *Vmax* = 60, V, pulse width = 0.5 ms and pulse delay = 10 ms: (**a**) Pt and (**b**) Mo. *Vmax* and pulse width were kept constant: 60 V and 0.5 ms, pulse delay set to 1, 10, 100, 1000 and 10,000 ms. (**c**) Pt and (**d**) Mo. Pulse width was taken as 0.5, 0.25, 0.125, 0.625 and 0.05 ms. In addition, *Vmax* = 60 V, pulse delay = 1 ms. (**e**) Pt and (**f**) Mo. Distribution of PUND results for different samples at pulse wide was 0.5 ms, pulse delay changed from 1 ms to 10,000 ms, (**g**) Pt and (**h**) Mo.

Then, the test was performed with different pulse widths. The *Vmax* and pulse delay were set to 60 V and 1 ms, respectively, and the pulse width was taken as 0.5, 0.25, 0.125, 0.625 and 0.05 ms. It can be seen in Figure 9c,d that the remanent polarization increases with increasing pulse width for both samples. At small pulse widths, it is not enough to support a complete flip of polarization, resulting in a decrease in the remanent polarization value. Therefore, devices utilizing the ferroelectricity of AlScN thin films require a special design when setting the operating frequency. The remanent polarization, leakage currents and breakdown voltage, as well as the retention of the ferroelectricity, should be taken into account during the device and architecture design.

Finally, the *Vmax* and pulse width were kept constant and tested at pulse delays of 1, 10, 100, 1000 and 10,000 ms, respectively. It can be seen that the Pt sample results are smooth with no significant change, which is basically the same as the previous test. The remaining polarization value in the positive direction of the Mo sample also has no significant change, whereas the value in the negative direction gradually increases, as shown in Figure 9e,f. This means that the Mo sample has a larger polarization loss in the negative direction, which also corresponds to a larger leakage current in the negative direction. Therefore, AOGs on the film surface perpendicular to the c-axis can greatly compromise the ferroelectric properties. Moreover, multiple samples were tested to observe whether there is good consistency, as shown in Figure 9g,h. It can be seen that the negative polarization fluctuation of the Mo sample is slightly larger, and other points float in a small range.

Therefore, the abnormal grain orientation may seriously affect the ferroelectric properties of the films. It is possible that the abnormal grain orientation changes the original wurtzite structure near the interface between the metal and dielectric, and then affects the polarization properties of the film. From the test results, this effect is unidirectional and will greatly change the polarization characteristics in one direction. It can be reasonably speculated that, in addition to Mo, other metal materials may also bring different effects, which is worthy of further experimental verification.
