APP下载

The memristive conductance studying of an Au/TinO2n-1 memristor device based on oxygen vacancy drifting

2016-09-29

科技视界 2016年20期

邱云生

【Abstract】With the insight of memristor researsh growing continuously, many semiconductor material has been manufactured to various kinds of memristive device. Yet the stabilization of memristive performance havent been fulfilled, the fathom of memristive-acting mechanism still not being generalized. To put a futher move on low-consuming and high-stable memristive conductance device, we built a high stable double-pair-electrode device, based on the fabrication of TiO2-x, which has been generally applied as a n-type semiconductor. Under the constant-repeating cyclic voltammerty; we nailed the memristive quality of our mental/semiconductor thin film device. Moreover, through multifarious analytical processes based on our doping, filming growing path, we build a rational model for our memristor‘s memristive conductance mechanism, which indicated the carrier motion and electron tunnel following the biasing voltage. Our work exhibited a new type of TiO2-x-based memristor, and emerged a new way to explicate the single-stage-switching memristive feature, which might initiate a new guiding ideology in semiconductor memristors studying.

1Introduction

Since 1971, Leon Chua put forward the concept of memristor, Memristive technology has always been a research hotspot in various fields, especially for Storage Media.

After 2008,hewlett-packard labs presented the pt-TiO2 / TiO2-x-pt memristive structure for the first time, the studying about nanoscale TiO2-x thin film memristor became one of the keynote research in memristor studying. Yet in the matter of fact, most of them consuming a high energy cost. To perform the memristive-switching conductance, various explanations have occurs on papers today: parallel conductive filament model, forward drifting of low-resistance region, solid-phase eletrolytic, phase transition, electron-deathnium centre and so on. These are the main process explanation.

Nowadays the understanding of memristive conducting is basically relating to the positive electrode material and the microstructures of memristive-device, but there are still no centralized guidance for the interpreting of memristive conduction appearance and the creating of long-notional memristor device. These main subjects are still an unbroken significance question to memristor studying.

To decline the energy wasage in memristor device fabrication, we adopt the hydro-thermal growing method to develop the nano-filamentary array of TiO2 which would been applied, as the film dielectric between two electrode. The basis layer which has been coated and fasted first on the FTO conductive glass would went through an oxygen-deprivation anneal first to bring into a limited quantity of oxygen vacanccies as the resistor-changing supporter under the consecutive voltage cycling. And the growing process also has a specfic temperature, envrioument regulation, to complete a necessary term for our devices idiosyncracy both in the structure and physical eigenstate, which fills the logic in our mechanism discussing.

The result shows by the cyclic voltammerty curves is promising to the vitality of memristive conductance peculiarity.

Review HP laboratories work in 2008 and Yuchao Yangs work in 2013, despite their difference in metal-base, both of their team build the memristor structure using nanoscale oxide devices, and the trial was under the theoretics that the V-O doped area could be transformed and budged under the voltage-charging electric field, this is one parlance that the memristive switching process is involved with the oxidizing resistance, which have complicated the voltages effect on the electron migration. Generally The V-O is migrated by the potentials drive, electrons obtain enough power to surmount Schottky potential barrier constantly. Despite the high energy-consume in the manufacturing process, the HV and HC problem still causing two much heat, and the memristive ability cannot be retained appropriated. In our work, the HV and HC has been highly bated, as it showen in the conductance imaging [Fig2.1-Fig2.2].

[Fig2.1]Our I-V test in negative bia from 1V to -2V.

Tested sample was annealed in Ar atmosphere and have passed the electro-forming gradation.

[Fig2.2]Our staid I-V image after 5000 times looping. Applied voltage deepening in negative bia from 1V to -3V, the circling loop was followed in every charging round. The on-off potential approached in a manageable scope, switching current a remain significant disparity.

The switching current along with the changing voltage has qualified with enough electricity difference within 0.1μAs restriction. Which has vigorously declined the heating problem in the switching process. This work attenuated the unsteadiness in memristor switching loop, and demonstrated us a new mechanism for the understanding of memristive conductance. In this resulting explication, the switching process is illustrated building on the physical eigenstate of loading carrier which is channeling in the contact surface of mental-semiconductor.

2Results and Discussion

2.1Vectored crystal growth and annealing atmosphere charged doping concentration.

TiO2-x has been a popular metal oxide material since it has been confirmed to have an RRAM function by Hp in 2008. By the way of radio frequency magnetron sputtering, a thin semiconductor film which combining two variable resistors is formed3. The semiconductor has a region with a high concentration of dopants.

Comprehended from our SEM test, TiO2 thin film is constituted by columnar crystal which are perpendicular to the FTO conductive glass altogether, and in the XRD component text, the nanoscale Ti-based mental oxide thin film exhibit a simplex diffraction peak.

Comparing to the Ti-based metal oxide thin file, which was formed and combined by PLD method, our thin film shows a simplex tropism as a single crystal and a stable semiconductor layer with a proper density. According to the Paulings law, the crystal growth in hydrotherm appears to the lowest surface energy. In this case, the growth direction would be the {001} crystallographic orientation.

[Fig.3] The Ti-O composition is different in different orientation, like it showed in pictures, it is more volatile in the (001) direction while compositions is connected in vertex (Paulings law).

Based on the density functional theory (DFT) Hua Gui Yang calculated the bonding energy for bulk anatase TiO2 on the clean surface of (001) and (101) are 0.93 J/m2 and 0.39 J/m2 respectively. This speaks the fact that our {001} crystallographic orientation layer has a strong application stability, which can shows in the figure I-V, the switching potential appears only minuscule changes after 5000 times periodic conductive tests.

In our methods, the doping process is achieved by annealing in inert gases at 510°C. This method guaranteed the original seed layer has a stable structure in which oxygen vacancy would hardly spread out. According to the XRD component analysis image, the main phase of our nanoscale layer is TinO2n-1 (n mostly equal to 4 or 5). Compared to HPs experiment, the doped region of our device is highly limited in the basic seed layer, which leads to a highly sensitive on switching voltage, a low-level switching point on our memristive device has appeared. After the annealing time is lengthen, V-O content in the base layer of semiconductor rises up.

With change in the study of nanoscale thin-films manufacturing-process, the proper represent of memristor switching process will be different. With a high doping ratio, the spirit of HPs switching mechanism performs about the forward-drifting of low-resistance region. The conductive character image shows a bipolar switching feature (the transform for the ON-potential and OFF-potential appears in different bias voltage). However, in our case, the oxygen vacancies only doping in the basic bottom layer, with a specific crystal orientation upon, the drifting of V-O is highly limited. As it showed in [Fig 2.1], under a small-applied voltage range (-1.5V-0Vor0.5V), the directional motion of V-O developed a finite amount of parallel conductive filament among semiconductor layer. This procedure changed the surface state of Schottky junction and partial voltage situation in semiconductor layer, eventually caused the unipolar feature (the transform for the ON-potential and OFF-potential appears in one bias voltage) of our memritor device. This situation is matching to Chen M-Cs theory.

2.2Electron concentration modeling.

As we discussed earlier, because of the small ratio of our memristive device, the directional motion of V-O can be described approximately as the stretching of a finite amount of parallel conductive filament, and this procedure was casted under a small-applied voltage range. This phenomenon shows two unique memristance character: First, In every circuit, different from Deok-Hwang Kwons research, whose V-A test progressed in a larger current (more than 10 mA in the ON-switching state). The OFF-switching state (the Ti4O7 filament break at this state) must been achieved at a different bias voltage and only leads to a 10 times decline. Our memristive device shows a unipolar switching feature, and the current gap ratio between ON-state and OFF-state comes to 100, as it appeared in [Fig2.1]; the ON-state current contains in micron scale and the switch repeat stably. Based on these differ results, we assume that the conductive filament in the film didnt penetrate the whole layer and bond upon the two terminal electrode, its just V-O drifts changed the voltage-dividing situation and the surface situation on M/S Schottky junction. Moreover, comparing to the discussion from Yangs team, the inducing of electric field didnt bank oxygen vacancy up to the mental/TinO2n-1 function in which case, a roughly debase on Schottky barrier would happen; our device remains nonvolatile in constantly repeating circuit, on the basis of these proposition, we considered that in the initial phase, the memristive conductance from our device is mainly produced by electron tunneling.

The drift current density of oxygen vacancy can be settled as:

In the state when the thickness of barrier comes to xc, the V-O concentration in the barrier surface of semiconductor NVo is large, which improves electro-tunneling probability sufficiently, the electro-tunneling current here will be siginificant.

Je∝P

Now comes the current density expression formula of our device:

J=J0+Jt+Je

Jt is the current of thermalemission, because the barrier height for electron on Aus side is invariable Jt basically stays abiding among the whole procedure.

JO and Je represent the main electric conductivity in the electro-forming process, in which the current of V-O drift decrease constantly. Je is the main conductive mechanism, in the final memristive-switching process, under a certain negative voltage applied, the tunneling current is sufficient to be the ON-state of our memristor device.

According to our conductance image, our memristive quality can be approximately classified to single-stage switching RRAM device which have no electric trap. In this condition, the current density should be

J=9∈iμVV2/8d2

is the dielectric constant of TiO2

The current density is proportional to the square of voltage, this mechanism happens when the oxygen vacancy were brought into semiconductor which only have a light doping. In our conductive mechanism, this equation can be used in high-potential testing region.

[Fig3.1]In a larger current-limiting(10μA) of the switching conductive test, the current circle shows a palpable increasing in the negative bia-voltages charge from 0.1V to 1.5V, from which shows the difference some of the oxygen vacancy made.

[Fig3.2]Bia-voltage testing contrast sample; current-limiting up to 10μA, the contrast changing from Fig3.1 is the samples annealing in an O2 atmosphere, in which the oxygen vacancy would be hardly draw into.The current loop shows a jarless tunneling current.

2.3The motion of oxygen vacancy

We consider that the oxygen vacaneys motion has caused our memristor devices conductive quality. In our case, as we discussed in last installment, the conductive filament among TiO2 layer is formed between two electrodes and enhanced in several times of switching loop. In this initial phase, which is usually defined as the electro-forming stage, currents here is mainly produced by V-O drifting and electron tunneling. Among the electro-forming stage, the voltage dividing on Schottky barrier is keep improving while semiconductor layers electro-resistance is declining. The eigenstate density at M/Ss contact region rises in the same time.

In the conductive switching phase we required, eigenstate density of M/S diffuse region has come to a stationary station approximately, width of the Schottky barrier is thin enough to make significant tunneling-currents. As it reflected in the mathematical formula, the switching current now will be turned on in a certain bias-voltage.

Our devices status transforming among the whole process can be all explicated as results of the vacancys motion vectored by bias voltage.

The length of unidimensional extent nanorods and the doping extent have done impinges on the memristive conductance. After Czochralski method-coating and inert gases-annealing, the basic layer which embodies oxygen vacancies is created (length of doped layer: 10-25nm). Based on our earlier experiments conclusion, the growth rate of TiO2 fiber rod, under 95℃ hydrotherm circumstance, is 1μm/24h. After optimized screening, our time-limit of the hydrothermal growth is settled on 4h (length of undoped layer: 150~200nm). According to Guo YuQuans research, the value evolution of “ROFF/RON” can be calculated in series of conductive changing, accompanying doped regions extending, “ROFF/RON” value descending constantly.

Oxygen vacancy is the principal carrier in the resistive changing under the charge of field-induction force, the differentia in V-O doping method and V-O quantity could catch out lots of characteristic variance in conductive switching mechanism. Qiangqiang Meng and his team studied the impact of oxygen vacancy on band structure of n-p co-doped anatase TiO2. As their research production, the oxygen vacancy prefers energetically to the neighbor

ing site of metal dopant downward, yet along with the original doping amounts reducing, this phenomenon appears to vanish, while the position of EF (Fermi energy) in energy band is also fade away from EC (conduction band bottom).

A fine state-migration model of our memristor can be built, the holistic procedure is found in the motion of oxygen vacancy.

As we discussed earlier, the voltage dividing of Schoktty barrier is charged by the formation of a finite amount of parallel conductive filament and the barrier potentials raise. For the conductive electron among the mental-semiconductor contact area, both of the height and width of the barrier potential is elevated.

The first procedure significantly reduced the nanaoscale layers resistance, the second is charged by the current carrier (oxygen vacancy)s gathering in the contact surface. With a negative voltage applied to Au electrode, attracting vacancies towards. The vacancy dopants drift in electric field through the lowest energy crystallographic plane. Free electron emerged with the oxygen vacancy toward inversly.

Combine these procedure above over, the memristive conductance switching mechanism can be expounded (We regulate the Fermi energy of based on these proposition).

Compared to others conductive filament, in which the carrier is mostly metal cation or actived transitional metals. The oxidize reduction reaction is involved, caused a significantly disparity in the stability of switching conductance.

In a bias voltage, our conductive circuit shows a directed loop shifting, which can be facilely linked to the V-Os drift. That resemble the conductance test run in Ting Changs research, a significant overlapping of neighboring hysteresis loop is appeared. He assumed this as potentiation process.

He assumes that the shifting and declining about the conducting changes in loop-repeating is caused by the drifting and forming of oxygen vacancy, which are constantly changing the internal state of his dielectric material.

Here, we use the activation of the easy doped conducting channel (oxygen vacancy drifting) and the potential state-changing of mental/TinO2n-1 boundary to formulation this process, which is a stabilize evolution Ting Changs device.

[Fig3.3]The electroforming is the initial and key step for the subsequent switching, which is caused by the electro-reduction and vacancy creation process at high electric fields and enhanced by electrical Joule heating. During the electroforming, oxygen vacancies are created and drift towards the cathode, forming located conducting channels in the oxide, and in a counter process O2- ions drift towards anode where they evolve O2 gas.

3Conclusion

A hydro-thermal growing method has already been successfully used in TinO2n-1 thin film memristive device, but in our research, the memristive transforming process shows a high reliable switching circuit. By the way of inert gases-annealing, we specifically doped a finite amount of oxygen vacancy on the bottom datum. In our {001} crystallographic orientation layer, the conductive filament shows a tough stabilization, with combining oxygen vacancy migration theoretics and pattern deducing of metal-semiconductor contact, our experimental theory was comprised. It explained the conductive switching character reasonably. Under our aborative coherent analysis, a detailed equation has been successfully applied to our devices memristive conducting system.