Progress In Electromagnetics Research, Vol.

151, 916, 2015

Realization of a Conductive Bridging RF Switch Integrated

2, * ,

Thais Vidal1 , Arnaud Vena3 , and Patrice Gonon4

AbstractThis paper presents a new approach for the realization of RF switches based on the
Conductive Bridging Random Access Memory technology (CBRAM). This promising approach allows
the realization of RF switches in an extremely simple manner at low cost. For the rst time, an RF
switch based on a MIM structure is realized with an insulator layer obtained from a commonly used
resin deposited by spin coating. The paper reports a RF switch based on CBRAM and demonstrates
a device integration onto plastic circuit board (PCB). The realized switch is validated by experimental
measurements for a frequency range up to 1.5 GHz with an activation voltage less than 1 V.

1. INTRODUCTION
With the tremendous development in means of communication in recent decades, the importance of
wireless communications has increased considerably, leading designers to new challenges such as the
development of versatile systems. To connect people to each other and across the Internet, RF designers
need to invent new multi-band and broadband RF circuits, particularly for telephony. Thus, RF switches
will play a key role not only in wireless communications, but also in sensors and actuators of tomorrow.
RF switches are basic components that allow reconguration of devices when they operate in order to
make them more agile. They are present in all communication systems and allow for instance signals
multiplexing, so as to recongure or dynamically control the system. The performances of such devices
can even impact the choice of the hardware architecture to be implemented in order to achieve the desired
functionality. This is why a broadband solution, based on a exible and low cost approach, where the
power consumption of the switch is reduced, is eagerly awaited. Todays technological solutions (solid
state and radiofrequency microelectromechanical (RF MEMS) based systems) need some improvements
to meet the challenges of the future. For examples, MEMS rest on heavy manufacturing technologies,
requiring an important number of masks. Even if RF MEMS fabricated using printed circuit processing
techniques has been proposed [1], most of the time, such components are expensive and associated with
an extremely rigid manufacturing process. The same complexity is observed in the design phase, where
mechanical and thermal considerations also have to be taken into account.
Current eorts focus on the future generations of memories [24]. The stakes are high, and the
development of new non-volatile memories, is expected to revolutionize the architectures of future
computer systems. Many emerging technologies are contending for supremacy, phase change (PC)
memory and resistance (switching) random access memory (ReRAM) among them. The common point
between these two technologies is that both have been evaluated for implementation in RF switches.
The specic dielectric used in PC materials can switch between amorphous (high resistance) phase
and crystalline (low resistance) phase upon application of thermic energy that can be in the form
Received 4 December 2014, Accepted 5 February 2015, Scheduled 11 March 2015
* Corresponding author: Etienne Perret (etienne.perret@LCIS.grenoble-inp.fr).
1 University Grenoble Alpes, LCIS, 50, rue de Laemas, BP 54, Valence 26902, France. 2 Institut Universitaire de France, 103
Boulevard St-Michel, Paris 75005, France. 3 Institut dElectronique du Sud, Place Eug`
ene Bataillon, Montpellier 34095, France.
4 LTM CNRS, CEA, Grenoble 38000, France.

10

Perret et al.

of voltage and current [3]. High performance RF switches based on PC materials have been recently
obtained [5, 6]. But the fabrication process remains complex and costly. Classical microfabrication cleanroom technology, high temperature processes, very specic dielectrics or even Joule heating simulations
are required. On the other hand, ReRAM which exhibits electrically switchable resistance, has also
attracted considerable attention in recent years [2]. Among these families of resistive memory cells,
the CBRAM technology rests on the electrochemical dissolution properties of an active electrode: the
metallic ions thus generated move through the solid electrolyte in order to form a conductive metal
bridge joining the other electrode, which is inert. The commutation of the device is thus obtained [7].
CBRAM are recognized by the International Technology Roadmap for Semiconductors (IRST) [8] as
suciently mature and powerful to be considered as very good candidates for the realization of future
memories. Based on this technology, which comes from microelectronics, it has been shown that the
operating principle of these memories can be used to realize RF switches. A new family of RF switches
using this principle and called Conductive Bridging RF Switch (CBRFS) recently appeared [911].
The feasibility of a RF switch operating between 1 to 6 GHz has been demonstrated [9, 10]. In this
frequency band, the switch has comparable performances to those that are currently obtained with
traditional approaches such as MEMS or solid state switches [10]. The CBRAM technology has shown
the potential to operate at lower energies and voltage (couple of volts), making it particularly interesting
for embedded applications. As compared to MEMS, CBRFS has lower switching time, dimensions
and cost. In [11], a multilayer structure based on CBRAM and realized without any photolithography
process has been proposed. This approach is simpler to implement, and therefore, potentially compatible
with printed electronic techniques. Moreover, integrating RF switches on a printed circuit board
(PCB) which is also a very challenging task seems possible with such approaches. In both
cases, the compatibility should be maintained with standard printed electronic and PCB manufacturing
technologies. For technological reasons, it is easier to stack dierent layers together (even for layers of
hundreds of nanometers) than to realize tens of micron gaps between electrodes made with dierent
metals. Unlike [9, 10] wherein a chalcogenide glass doped with silver ions is used, in [11], the solid
electrolyte is simply a SiO2 layer deposited by evaporation.
In order to be even closer to a completely printed RF switches solution, we describe a novel RF
switch made with a polymer resin, a common material that can be easily deposited by spin coating. No
complex technological, high temperature process such as photolithography or doping is used. Instead,
the approach is simple to implement and compatible with direct integration of the RF switches in
conventional RF PCB. The study that follows aims to demonstrate the implementation of these MIM
structures on a low cost and versatile substrate, massively used in the electronics manufacturing process
for signal switching applications.
2. CONCEPT OF THE RF SWITCH BASED ON CBRAM
2.1. Operating Principle
CBRFS are based on MIM solid state structures (Metal Insulator Metal), i.e., a stack of three layers
with no moving parts, as shown in Figure 1(a). By carefully choosing materials and their thicknesses,
it is possible to show that such a structure acts as a programmable resistor that keeps its value in the
absence of energy to maintain its state. This stack must be composed of an inert electrode (Al, Ni. . .)

Metal (1)
Insulator

R On /
R Off

Metal (2)

(a)

(b)

Figure 1. CBRAM Cell: (a) MIM structure, (b) electrical equivalent circuit.

Progress In Electromagnetics Research, Vol. 151, 2015

11

(a)

(b)

(c)

(d)

(e)

(f)

Figure 2. Operating principle of the CBRAM cell. (a) Initial O state. (b) Transition to the On state.
(c) Return to the O state. The polarization of the cell is indicated on the drawings. (d) Variation of
the bias voltage as a function of time, the typical tension variation value used here is 1 V/s. (e) Typical
changes in current/voltage across a CBRAM cell, transition to the On state (SET). (f) Transition to
the O state (RESET).
and the other must be chemically active (Cu, Ag. . .). An equivalent electric diagram of the structure is
shown in Figure 1(b). It comprises a variable resistance (relating to the On and O states) in parallel
with a capacitor, which depends on the geometry of the MIM structure and the nature of the dielectric
layer. Initially, the structure is not conductive (denoted O state in Figure 2(a)) because the electrolyte
is a good insulator. Under the action of a tension V set between the two electrodes, the insulator
allows the migration of Cu ions, which come from the active electrode, towards the inert electrode (see
Figure 2(b) [12]). The ions are then deposited on the inert electrode and are reduced with electrons
to obtain the Cu metal. Thus, a conductive lament grows until it touches the active electrode. As
we can see in Figure 2(e), the device then switches to a conducting (On) state (SET). To break the
lament and return to the O state (RESET), we have to simply reverse the voltage (see Figure 2(f)).
The nature of the dielectric plays an important role in the establishment of the conductive bridge. The
most widely used ones are chalcogenides [7], polymer [13] and oxides [14]. In this study, Poly(methyl
methacrylate) (PMMA) is used. This commonly used resin mixed with a solvent to form a liquid can be
simply deposited by spin coating. After a short curing (about 100 C during 1 min, which is the highest
temperature reached during the fabrication process.) we obtain a solid thin layer with few hundred
nanometers of thickness. Moreover, this resin has the advantage of being printable [15], which makes
this RF switches technology potentially compatible with standard printed electronics manufacturing
process.
2.2. CBRAM Approach for RF Applications
Currently, CBRAM are mainly designed to produce the next generation of non-volatile memories [8].
The objective here is to reuse their operating principle for the realization of RF switches. The expected
specications are very dierent from those memories and require a specic work using RF skills.
For memory applications, consumption, switching and memory access time are critical. In RF,
it is the frequency behavior, in particular, the ratio between the On and O state, which is decisive.
More precisely, the gure of merit of a switch is linked to the product of the On state resistance (ROn )
by the capacitance C. This product has to be minimized to obtain better performances. Given the

12

Perret et al.

specicities of each application, one of the main dierences between memories and the RF switches is
regarding the dimensions of the structures. Indeed, in RF, the miniaturization of the cell is not critical;
the switch must be sized to have a specic frequency behavior. It is thus necessary to ensure a capacitor
value C as low as possible on the largest frequency range. Thus, a conguration with a thick dielectric
layer (as compared to the classical thicknesses used for memory) appears as one of the most promising
candidates, thanks to the possibility of decreasing the RF capacitor between the two electrodes.
3. EXPERIMENTAL VALIDATION AND PRACTICAL ISSUES
Initial MIM structures were made like the one presented in [11], a simple stacking of three layers made
of dierent materials, without taking the RF accesses into consideration. These structures were used
to validate the switching nature of the component and to characterize some important values, such
as threshold voltages (Vset and Vreset ) that have to be used to set and reset the switch. We used
a conventional RF substrate (Rogers RO4003C), which acts as a board, but also provides the rst
metal electrode of the MIM structure, namely, the active electrode of copper (Cu) with a thickness
of 17 microns. Then, a PMMA dielectric layer of 90 nm was obtained using the spin coating process.
Finally, the aluminum (Al) electrodes were deposited by evaporation (100 nm thickness), using a metal
mask with a large number of circular holes. One of the structures produced is shown in Figure 3(a).
IV characterizations were performed with a programmable sourcemeter (Keithley 2635A) connected to
a probing station. The PMMA layer was removed from a corner of the sample to connect a tungsten tip
to the Cu bottom electrode. The second probe, a gold wire, was placed on the upper Al electrode. Thus,
while imposing a bias voltage, the current and the voltage were measured over time at the terminals
of the structure. Each programming cycle was composed of a positive triangle signal followed by a
negative triangle signal (see Figure 2(d)). Meanwhile, we imposed a limitation current, which could be
set unequally between the positive and negative sections of the cycle.
The rst ten programming cycles (set/reset) of the CBRAM cell are shown in Figure 4. The
switching behavior observed was also conrmed for other pads, and at each time several re-programming
cycles were possible. This shows that the desired switch function is compatible with the use of common
materials, such as standard RF substrates, PMMA as insulator and aluminum as inert electrode. As
expected, during the initial cycles, we observe that the SET voltages (Vset equal to 3 V5 V) and the
RESET voltages (Vreset equal to 2 V) are higher than for the other cycles. The programming current
used is always limited to 1 mA. After the process of bridge forming, the SET voltage stabilizes at around
0.5 V and the voltage RESET around 1 V. The values of the On and O state resistors, extracted from

(b)

(c)

(a)

Figure 3. (a) Picture of the Cu/PMMA/Al CBRAM cell, description of the dierent layers. (b) 50
microstrip line. (c) 50 microstrip line with the CBRFS at the center.

Progress In Electromagnetics Research, Vol. 151, 2015

13

Figure 4. Measured set/reset programming cycle of the Cu/PMMA/Al CBRAM cell. The PMMA
thickness is 90 nm. The programming current is limited to 1 mA.
the results are ROn = 50 and RO = 10 M. As expected, activation voltages are much lower than
those found in [11], where for a SiO2 layer of 160 nm, a threshold voltage switching was found to be
7 V. These values are of the same order of magnitude as those of [9, 10] (Vset = 1 V and Vreset = 1 V),
where a dielectric doped with Ag ions was used. We observe that the switch can operate more than ten
programming cycles (set/reset). However, after a few tens of cycles the component is blocked on keep
one of the two states. This behavior is sensitive to the manufacturing process; the results can be highly
variable from one plot to another. We believe that a specic work dedicated to the technology process
is expected to increase the switch life.
To validate the implementation of the CBRAM cell on a RF circuit, a 50 characteristic impedance
microstrip line has been used as shown in Figure 3(b), where a switch has been performed at the center
of the line (see Figure 3(c)). The CBRAM cell, i.e., the Cu/PMMA/Al stack of layers, has been made
by manually using a Kapton masking tape. PMMA has been deposited on the entire surface of the
sample, with the exception of a small area situated at one end of the microstrip line near the gap region,
i.e., the zone (1) described in Figure 3(b). At the end of the process, this area allows connecting the
aluminum electrode to the part of the copper line that will not include the CBRAM cell. The Kapton
tape is removed and then a new area is masked using Kapton: the entire sample is covered, except for
a small area situated at the center of the line (see the zone (2) in Figure 3 (b)). This area corresponds
to the precise location of the MIM structure (see Figure 3(c)) plus the connection between the upper
electrode and the second part of the microstrip line. To end the manufacturing process, a layer of
aluminum 100 nm thick is deposited by evaporation. We observe that the voltage to activate the switch
can be applied directly between the two parts of the microstrip line by using the inner conductor of the
SMA connectors.
Three dierent structures were realized, but only one was operational. This is likely due to the
non-uniformity of PMMA deposit caused by the copper layer roughness. For measuring S parameters,
this structure was connected to a VNA. A voltage generator was connected to the rear panel of the
VNA in order to add the DC component to the RF signal. In this way, a dierence of potential was
applied between the inner conductors of the two coaxial lines of the VNA, that is, between the two
metal electrodes of the CBRAM cell. Despite the presence of a sucient voltage to SET the switch, no
changes in the S parameters were observed. The structure kept the same behavior as a shorted line,
i.e., the structure in Figure 3(b): the RF line without the CBRAM cell. This behavior is caused by
the thickness of the weak aluminum electrodes. This eect could be easily eliminated by increasing Al

14

Perret et al.

thickness. But as it was not possible to use the previously employed evaporation system to increase
the thickness of the inert electrode signicantly, we chose to manually deposit a layer of silver lled
epoxy resin on the zone (2), as shown in Figure 3(c). As described in Figure 2(b), to set the switch, a
negative voltage is applied between the two electrodes. The measurement results obtained are shown
in Figure 5, where one can see the variations in frequency of the S-parameters for dierent negative
values of the voltage applied between two of the access lines. From 0 V to | 2 V|, the switch is in an
On state. An insertion loss (S21 ) of less than 4 dB is obtained up to 2.5 GHz. When the absolute value
of the voltage is increased, the switch gradually commutes from the On state to the O state. This
transition is clearly observable in Figure 5. Thus, for voltage values above |2 V|, the switch is in an O
state with an insertion loss (S21 ) less than 10 dB over the entire band 01.5 GHz. Thereafter, positive
voltages were applied to the switch; the results are shown in Figure 6.
Observations are quite similar to the one obtained with positive voltages: for low positive values
(below 1.5 V), the switch goes back to an On state; for higher voltages, it returns to the O state.
0

-1
-5
-2
0V
-0.5 V
-1 V
-1.5 V
-2 V
-2.5 V

-15

-3 V
-3.5 V

-20

-3
S11, S 22 (dB)

S11 , S 21 (dB)

-10

-4.5 V
-5 V

-7

-5.5 V
-6 V
-6.5 V
-7 V

-30

-5
-6

-4 V

-25

-4

0.5

1
1.5
Frequency (GHz)

-8
-9

2.5

0.5

1
1.5
Frequency (GHz)

(a)

2.5

(b)

Figure 5. S parameters measured for the CBRFS (Figure 3(c)) with respect to the tension (negative
variations) of polarization. (a) ( ) The return loss (S11 ), (+++) the insertion loss (S21 ). (b) ( ) The
return loss (S11 ), (+++) the return loss (S22 ).
0

0
-1

-5
-2
0V
0.5 V
1V
1.5 V
2V
2.5 V

-15

3V
3.5 V

-20

4.5 V
5V
5.5 V

-5

-7

6V
6.5 V
7V

-30

-4

-6

4V

-25

-3
S11, S 22 (dB)

S11, S 21 (dB)

-10

0.5

1.5

Frequency (GHz)

(a)

2.5

-8
-9

0.5

1.5

2.5

Frequency (GHz)

(b)

Figure 6. S parameters measured for the CBRFS (Figure 3(c)) with respect to the tension (positive
variations) of polarization. (a) ( ) The return loss (S11 ), (+++) the insertion loss (S21 ). (b) ( ) The
return loss (S11 ), (+++) the return loss (S22 ).

Progress In Electromagnetics Research, Vol. 151, 2015

15

Table 1. Positioning of the CBRFS presenting here in relation to others work. Comparison of principal
properties.
Present work

[11]

[10]

[6]

Frequency Range

DC to 1.5 GHz

DC to 0.15 GHz

DC to 6 GHz

DC to 20 GHz

Insertion Loss

4 dB

1.6 dB

0.5 dB

2.5 dB

Isolation

10 dB

20 dB

35 dB

15 dB

Actuation Voltage

1 V2 V

7V

1V

1V

Fabrication process

Simple/Common material

clean room

clean room

clean room

The observed behavior is clearly dierent from that of the MIM Cu/PMMA/Al structure previously
characterized with the DC probing system. This behavior can be explained by the presence of the
silver layer, which has been added to the aluminum electrode. Indeed, silver is known to be an active
electrode, and it has even smaller activation energy than copper for diusion in dielectric substrate.
This means that silver ions can diuse into the structure to create a lament when a positive voltage
is applied. The presence of the two active electrodes gives a symmetric behavior to the switch; in this
case, the Set and Reset voltages are similar with a value of around zero Volt [2, 3, 16]. Positioning of the
CBRFS presented here in relation to others work is done in Table 1. It is clear that the device reported
in this manuscript did not demonstrate improved performance in comparison to others works [6, 10].
However, the realization process is totally dierent, common materials are used and a clean room is not
required.
4. CONCLUSION
In this paper, we described the implementation of CBRFS with a technique commonly used in the
electronics manufacturing process. The approach has a great potential to be compatible with printed
electronics techniques. An RF switch having two distinct states over a frequency range of 01.5 GHz
was obtained from a conventional stack made of three materials, particularly by using a PMMA resin
as the insulator layer. No complex deposition or photolithography process has been used. This shows
the versatility and the huge potential of such an approach. Indeed, our approach is based on a simpler
manufacturing technique (fewer manufacturing steps, which do not require the use of etching processes,
based on common materials. . .) and could be compatible in the future with low cost printing techniques
conducted using a dielectric ink, or with common printed circuit techniques. In the latter case, RF
circuits including RF switches could be realized simultaneously by a classical PCB manufacturer. The
results are encouraging and can be used to solve the problem of CBRAM integration on to PCB.
The proposed technique is very promising for the realization of RF switches, but the reliability of
these devices is still weak and needs further work for improvement. Indeed there is still a lot of room
for improvement of the CBRFS performances. CBRFS should have a signicant impact in the future
thanks to its simple structure, low cost, non-volatility and low voltage activation, and should successfully
compete with current solutions that require much more complex processes for their realization.
ACKNOWLEDGMENT
The authors wish to thank the Up-line Technological Platform (PTA) for their technical support.
REFERENCES
1. Silva, M. W. B., S. E. Barbin, and L. C. Kretly, Fabrication and testing of RF-MEMS switches
using PCB techniques, 2009 SBMO/IEEE MTT-S International Microwave and Optoelectronics
Conference (IMOC), 96100, 2009.
2. Valov, I., R. Waser, J. R. Jameson, and M. N. Kozicki, Electrochemical metallization memories
Fundamentals, applications, prospects, Nanotechnology, Vol. 22, 254003, 2011.

16

Perret et al.

3. Derhacobian, N., S. C. Hollmer, N. Gilbert, and M. N. Kozicki, Power and energy perspectives of
nonvolatile memory technologies, Proceedings of the IEEE, Vol. 98, 283298, 2010.
4. Huang, G. M. and Y. Ho, Memristors for non-volatile memory and other applications, Advances
in Non-Volatile Memory and Storage Technology, Y. Nishi, Ed., Woodhead Publishing, 2014.
5. Shim, Y., G. Hummel, and M. Rais-Zadeh, RF switches using phase change materials, 2013 IEEE
26th International Conference on Micro Electro Mechanical Systems (MEMS), 237240, 2013.
6. Crunteanu, A., A. Mennai, C. Guines, D. Passerieux, and P. Blondy, Out-of-plane and inline
RF switches based on Ge2 Sb2 Te5 phase-change material, IEEE MTT-S International Microwave
Symposium (IMS), 14, Tampa Bay, Florida, US, 2014.
7. Kund, M., G. Beitel, C. U. Pinnow, T. Rohr, J. Schumann, R. Symanczyk, K. D. Ufert, and
G. Muller, Conductive bridging RAM (CBRAM): An emerging non-volatile memory technology
scalable to sub 20 nm, IEEE International Electron Devices Meeting, IEDM Technical Digest,
754757, 2005.
8. The international technology roadmap for semiconductors (ITRS), 2011, Available: http://www.
itrs.net/.
9. Nessel, J. A., R. Q. Lee, C. H. Mueller, M. N. Kozicki, M. Ren, and J. Morse, A novel nanoionicsbased switch for microwave applications, 2008 IEEE MTT-S International Microwave Symposium
Digest, 10511054, 2008.
10. Nessel, J. and R. Lee, Chalcogenide nanoionic-based radio frequency switch, Patent
No. US8410469 B2, USA, 2013.
11. Vena, A., E. Perret, S. Tedjini, C. Vallee, P. Gonon, and C. Mannequin, A fully passive RF switch
based on nanometric conductive bridge, IEEE MTT-S International Microwave Symposium (IMS),
13, Montreal, Canada, 2012.
12. Russo, U., D. Kamalanathan, D. Ielmini, A. L. Lacaita, and M. N. Kozicki, Study of multilevel
programming in programmable metallization cell (PMC) memory, IEEE Transactions on Electron
Devices, Vol. 56, 10401047, 2009.
13. Potember, R., T. Poehler, and D. Cowan, Electrical switching and memory phenomena in
CuTCNQ thin lms, Applied Physics Letters, Vol. 34, 405407, 1979.
14. Bernard, Y., V. Renard, P. Gonon, and V. Jousseaume, Back-end-of-line compatible conductive
bridging RAM based on Cu and SiO2 , Microelectronic Engineering, Vol. 88, No. 5, 814816, 2011.
15. De Gans, B.-J., L. Xue, U. S. Agarwal, and U. S. Schubert, Ink-jet printing of linear and star
polymers, Macromolecular Rapid Communications, Vol. 26, 310314, 2005.
16. Vianello, E., C. Cagli, G. Molas, E. Souchier, P. Blaise, C. Carabasse, G. Rodriguez, V. Jousseaume,
B. De Salvo, F. Longnos, F. Dahmani, P. Verrier, D. Bretegnier, and J. Liebault, On the impact
of Ag doping on performance and reliability of GeS2 -based conductive bridge memories, 2012
Proceedings of the European Solid-State Device Research Conference (ESSDERC), 278281, 2012.