Page 1
Mater. Res. Soc. Symp. Proc. Vol. 1729 © 2015 Materials Research SocietyDOI: 10.1557/opl.2015.
Influence of Graphene Interlayers on Electrode-Electrolyte Interfaces in Resistive Random
Accesses Memory Cells
Michael Lübben1, Panagiotis Karakolis2, Anja Wedig1, Vassilios Ioannou2, Pascal Normand2,
Panagiotis Dimitrakis2, Ilia Valov1 1Peter Gruenberg Institut, Forschungszentrum Jülich GmbH, Jülich, Germany 2 Institute of Nanoscience and Nanotechnology, NCSR Demkritos, Athens, Greece
ABSTRACT The behavior of the redox-based resistive switching memories is influenced by chemical
interactions between the electrode and the solid electrolyte, as well as by local environment. The
existence of different chemical potential gradients is resulting in nanobattery effect lowering the
stability of the devices. In order to minimize these effects we introduce a graphene layer at the
active electrode – solid electrolyte interface. We observe that graphene is acting as an effective
diffusion barrier in the SiO2-based electrochemical metallization cells and acts catalytically on
the electrochemical processes prior to resistive switching.
INTRODUCTION
Redox-based resistive switching memory (ReRAM) cells are the most emerging memory
devices for technology nodes <16 nm according to the latest edition of the international
technology roadmap for semiconductors (ITRS) [1]. Mainly studied are ReRAMs showing either
the vacancy change (VCM) or electrochemical metallization mechanism (ECM) [2]. It is well
known that the resistance switching in these devices is governed by the electrochemical
interaction between the terminal electrodes and the solid-state electrolyte material [3]. The
chemical instability and the presence of (electro)-chemical potential gradients at the
electrode/electrolyte interface are the main issues for ReRAM cells that affect both performance
and reliability [4,5]. We aim to control the electrochemical interaction at the interfaces by
inserting an intermediate layer between the electrode and electrolyte as shown for other systems
in order to serve as diffusion barrier and prevent intermixing and chemical interactions [6] as
shown for e.g. Ag/GeSx based devices [7]. Graphene was tested as such intermediate layer as the
thinnest electronically conductive film with sufficient electronic conductivity in plane and
mostly insulating out of plane, possessing good mechanical properties and the ability to
homogeneously cover the active electrode. Graphene monolayers (GML) have unique properties
in terms of conductivity, mechanical strength, diffusion barrier, transparency, heat conductivity
etc. [8] and thus graphene has been used as electrode material in many applications [9].
Recently, graphene (G) has been used as an intermediate layer between the electrode and
the metal oxide material in two-terminal ReRAM devices. It has been demonstrated that in VCM
ReRAM [10] device the GML offers higher low-resistance-state (LRS) and better high-
resistance-state (HRS) uniformity, which are both important for ReRAM characteristics. The
built-in series resistance due to the GML substantially raises the LRS resistance, effectively
reduces current overshoot, and significantly lowers the programming power. The GML may
serve as an oxygen capture reservoir in addition to being a passive barrier. Finally, it was found
[11] that by increasing the number of graphene layers from one to three, the series resistance
213
Page 2
between the active electrode and the metal oxide increases, which allows the easy integration of
a selector device to realize one-selector one-resistor (1S1R) structures.
In this work, results on the performance of graphene-modified G-ReRAM cells
Ag/G/SiO2/Pt and Cu/G/SiO2/Pt ECM cells where GML is deposited at the active
electrode/metal oxide interface are shown. The GML transfer process was characterized and
graphene layers were tested as an effective diffusion barrier by secondary ion mass spectrometry.
The G-ReRAM cells were characterized by cyclic voltammetry, prior to the resistive switching
event, subsequently being compared to the reference cells without GML.
EXPERIMENT
Graphene Monolayer (GML) transfer process
Several techniques have been developed in order to transfer GML on various substrates.
The technique used is similar to the technique of etch-back of a copper foil as described in
reference [12]. The GML on copper foil used in this work was commercially available. In this
technique, the Cu/GML/PMMA foil is put in a copper etching solution (see Fig. 1(a)). When the
copper is etched off, the GML/PMMA film is floating on the liquid etchant surface (see Fig.
1(b)). Next, we put the target substrate in the solution and the film is attached on substrate’s
surface by Van der Waals forces (see Fig. 1(c)).
(a) (b) (c) (d)
Figure 1 The three steps of the etch-back process for GML transfer on a substrate: (a) the
GML on Cu is covered by PMMA and is put in a copper etching solution, (b) the Cu is etched
and the GML/PMMA film is floating, (c) the floating film is transferred on the substrate and
(d) transfer is completed.
Iron (III) chloride (FeCl3) was used as an etchant of the copper foil in the process. It must be
noted that in order to: a) eliminate any contamination of the graphene/PMMA film and the
substrate with iron (Fe) or chloride (Cl) residues and b) avoid any interaction of the substrate
with chloride ions, a two-step cleaning procedure was performed. Ammonium persulfate (APS)
was tested as well. The complete process is summarized below:
Step A: Cu Etch
1. PMMA / Graphene sample on Cu foil (~25 μm) was cut in size of a small sample
2. 25 ml of FeCl3 solution (17.4 g FeCl3 / 100 ml DI water) for Cu etching
3. Sufficient time to etch Cu (~50 min) Step B: Sample cleaning – Removal of any metallic residues
1. 10 ml of DI water was rinsed on the sample to clean Cu etchant. Wait for 5 minutes (repeat 6-
8 times)
2. HCL: DI solution (1:2) was rinsed to remove metal particles
Page 3
3. Rinse with DI water
4. Repeat steps 2 & 3 (at least twice)
5. Rinse with DI water (repeat at least 8 times)
6. Transfer PMMA / Graphene on the substrate
7. Samples left to dry overnight
Step C: PMMA removal
1. Samples cleaned overnight in acetone and IPA respectively
Fabrication of ReRAM cells
ReRAM cells were prepared on oxidized silicon substrates covered with 5 nm TiO2 as adhesion
layer and 30 nm Pt as bottom electrode. Both layers were in situ deposited with DC sputtering.
For the devices with tantalum and its corresponding oxide RF sputtering was used. Here, 15 nm
metal was deposited. One sample was covered with graphene as described above. Another
tantalum sample was left in ambient air, to have similar storage conditions. Afterwards, 15 nm of
tantalum pentoxide was deposited, prior to a photolithography step for the 30 nm thick platinum
top electrodes. The SiO2 based samples were prepared by coverage of the platinized substrates
with 30 nm silicon dioxide films, which were deposited via electron beam evaporation at 0.01
nm/s. On the oxide, graphene was transferred before the formation via photolithography and
deposition (metal lift-off) of 30 nm Cu electrodes and subsequently capped by a 30 nm Pt thin
layer using the same e-beam evaporation system.
Figures 2 and 3 are typical images of samples with GML transferred on SiO2 (left figure)
and Ta2O5 (right figure). Images were taken just after the transfer and the subsequent drying
overnight (Step B-7), i.e. PMMA film is not removed in order to have a clear picture of the
sample’s area covered by GML.
Figure 2 Graphene on SiO2substrate. Figure 3 Graphene on Ta2O5 substrate.
RESULTS
Graphene Monolayer (GML) characterization by Raman Spectroscopy
Following the transfer procedure described in previous section, GMLs were transferred
on 280 nm thick SiO2 layers, thermally grown on Si wafers. In order to characterize the quality
of the transferred GMLs, Raman microscopy was used extensively. Measurements were
performed at room temperature with a Renishaw spectrometer with a laser at 514.5 nm. The
resulted spectra for one of the samples are presented in figure 4. The D band (indication of
presene of defects) is minimized while the G and 2D bands are very sharp. The sharpness of the
GML
Page 4
2D bands indicates that a GML monolayer is measured which is expected and verifies the
uniform transfer of our graphene layer. Also, as calculated, the ratio of 2D - D peak intensity is
very high, which also indicates that the transfer process results in high quality graphene
monolayer.
(a)
(b)
Figure 4 (a) Optical microscope image for sample shown at figure 2 (named P2). Cross-marks
denote the points where the Raman spectra shown in (b) were measured.
SIMS results:
The effectiveness of the graphene as a diffusion barrier has been studied by secondary ion
mass spectrometry (SIMS). Samples with and without graphene were annealed at 200 °C for 2
h. The depth profiles presented in figure 5 show that for the samples with graphene on the Ag
electrode there is no significant intermixing, whereas for the reference sample without
graphene silver atoms diffused into the SiO2. This reveals the capability of graphene to
function as a diffusion barrier in ReRAM devices.
Electrical characterization
Cyclic voltammetry was performed on stacks consisting of 30 nm Pt/30 nm SiO2/graphene/30
nm Cu. The measurements were performed at different sweep rates and compared to results
from previous measurements without graphene between the SiO2 electrolyte and the copper
electrode. The obtained cyclic voltamograms are shown in figure 6. The intermediate graphene
layer has a significant influence on the redox-behavior of the cells prior to the forming or
switching event. Obviously the oxidation peaks shift to smaller voltage values while the
current increases. In comparison to the standard cell in Fig. 6 the graphene seems to have
catalytic impact on the oxidation of the active copper electrode while the reduction is slightly
depressed. This leads to an ON-switching after few cycles. Without graphene the switching
processes are shifted to higher voltages.
Page 5
(a) (b)
Figure 5 SIMS profiles. (a) Ag/SiO2 stack without graphene (b) Ag/SiO2 stack with
graphene at the interface.
(a) (b)
(c)
(d)
Figure 6 (a)-(b) Cyclic voltammetry with switching events for the Cu/G/SiO2/Pt samples at of 129 mV/s
and 4.1mV/s sweep rates respectively. (c) Cyclic voltammetry on the reference ReRAM cell without
graphene as intermediate layer [13]. (d) Schematic stack of the tested devices.
Page 6
CONCLUSIONS
In this paper we have shown that graphene acts as a diffusion barrier for Ag and stops the
intermixing of the active electrode and the solid electrolyte. It is also increasing the catalytic
activity of the active electrodes as shown in cyclic voltammetry experiments. Memory cells with
graphene layer are switching at lower voltages compared to cells without graphene. We also
discuss the technological specifics and challenges in preparing and characterizing these type
devices.
ACKNOWLEDGMENTS
This research has been financially supported by the Greek-German bilateral joint research project
“G-ReRAM” supervised by the GSRT project Nr. GER_2316 and BMBF project Nr. 03X0140.
Dr A. Kontos from INN-NCSR “Demokritos” is highly acknowledged for the measurements of
Raman spectra.
REFERENCES
1. The International Technology Roadmap for Semiconductors-ITRS 2011 Edition. (2011).
2. R. Waser and M. Aono, “Nanoionics-based resistive switching memories”, Nature Materials,
6, 833 (2007)
3. I. Valov, “Redox-Based Resistive Switching Memories (ReRAMs):Electrochemical Systems
at the Atomic Scale”, ChemElectroChem, 1, 26–36 (2014)
[4] I. Valov, E. Linn, S. Tappertzhofen, S. Schmelzer, J. van den Hurk, F. Lentz & R. Waser,
“Nanobatteries in redox-based resistive switches require extension of memristor theory”, Nat.
Commun. 4, 1771 (2013)
[5] S. Tappertzhofen, E. Linn, U. Bottger, R. Waser, I. Valov, “Nanobattery Effect in RRAMs -
Implications on Device Stability and Endurance”, IEEE Electr. Dev. Lett. 35, 208 (2014)
[6] J. van den Hurk, A.C. Dippel, D.Y. Cho, J. Straquadine, U. Breuer, P. Walter, R. Waser, I.
Valov, “Physical origins and suppression of Ag dissolution in GeSx-based ECM cells”, Phys
Chem Chem Phys, 16, 18217 (2014)
[7] D.-Y. Cho et al. “Direct Observation of Charge Transfer in Solid Electrolyte for
Electrochemical Metallization Memory”, Adv. Mater. 24, 4552 (2012)
8. R. S. Ruoff, C. W. Bielawski, and D. R. Dreyer, “From Conception to Realization: An
Historial Account of Graphene and Some Perspectives for Its Future”, Angew. Chem. Int. Ed.
(2010), 49, 9336 – 9345
9. G. N. Dash, S. R. Pattanaik, S. Behera, “Graphene for Electron Devices: The Panorama of a
Decade”, J. Electron Device Society 2, 77-104 (2014)
10. Hong-Yu Chen, He Tian, Bin Gao, Shimeng Yu, Jiale Liang, Jinfeng Kang, Yuegang Zhang,
Tian-Ling Ren, H.-S. Philip Wong, “Electrode/Oxide Interface Engineering by Inserting Single-
Layer Graphene:Application for HfOx–Based Resistive Random Access Memory”, IEDM 2012
Techn. Digest, 489-492
11. Yuchao Yang , Jihang Lee , Seunghyun Lee , Che-Hung Liu , Zhaohui Zhong , and Wei Lu,
“Oxide Resistive Memory with Functionalized Graphene as Built-in Selector Element”, Adv.
Mater., 26, 3693–3699 (2014)
12. J. W. Suk, A. Kitt, C.W. Magnuson, Y. Hao, S. Ahmed, J. An, et al., “Transfer of CVD-
Grown Monolayer Graphene onto Arbitrary Substrates” ACS Nano,5, 6916–6924 (2011)
13. S. Tappertzhofen, S. Menzel, I. Valov, R. Waser, “Redox Processes in Silicon Dioxide Thin
Films using Copper Microelectrodes”, Appl. Phys. Lett., 99, 203103 (2011)