Novel Receiver for Superparamagnetic Iron Oxide Nanoparticles in a Molecular Communication Setting
Max Bartunik, Maximilian Lübke, Harald Unterweger, Christoph Alexiou, Sebastian Meyer, Doaa Ahmed, Georg Fischer, Wayan Wicke, Vahid Jamali Kooshkghazi, Robert Schober, Jens Kirchner
NNovel Receiver for Superparamagnetic Iron Oxide Nanoparticlesin a Molecular Communication Setting
Max Bartunik ∗ Maximilian Lübke ∗ [email protected]@fau.deInstitute for Electronics EngineeringFriedrich-Alexander-UniversityErlangen-Nuernberg (FAU)Erlangen, Germany Harald UnterwegerChristoph Alexiou
Section of Experimental Oncologyand Nanomedicine (SEON)University Hospital ErlangenErlangen, Germany
Sebastian MeyerDoaa AhmedGeorg Fischer
Institute for Electronics EngineeringFAU, Erlangen, Germany
Wayan WickeVahid Jamali KooshkghaziRobert Schober
Institute for Digital CommunicationFAU, Erlangen, Germany
Jens Kirchner [email protected] for Electronics EngineeringFAU, Erlangen, Germany
ABSTRACT
Superparamagnetic iron oxide nanoparticles (SPIONs) have recentlybeen introduced as information carriers in a testbed for molecu-lar communication (MC) in duct flow. Here, a new receiver forthis testbed is presented, based on the concept of a bridge circuit.The capability for a reliable transmission using the testbed anddetection of the proposed receiver was evaluated by sending a textmessage and a 80 bit random sequence at a bit rate of 1 / s, whichresulted in a bit error rate of 0 %. Furthermore, the sensitivity of thedevice was assessed by a dilution series, which gave a limit for thedetectability of peaks between 0 . . / mL. Compared to thecommercial susceptometer that was previously used as receiver, thenew detector provides an increased sampling rate of 100 samples / sand flexibility in the dimensions of the propagation channel. Fur-thermore, it allows to implement both single-ended and differentialsignaling in SPION-bases MC testbeds. KEYWORDS
Molecular communication, superparamagnetic iron oxide nanopar-ticles, SPION, differential signaling, receiver
ACM Reference Format:
Max Bartunik, Maximilian Lübke, Harald Unterweger, Christoph Alexiou,Sebastian Meyer, Doaa Ahmed, Georg Fischer, Wayan Wicke, Vahid JamaliKooshkghazi, Robert Schober, and Jens Kirchner. 2019. Novel Receiver for ∗ Both authors contributed equally to this research.Permission to make digital or hard copies of all or part of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full citationon the first page. Copyrights for components of this work owned by others than ACMmust be honored. Abstracting with credit is permitted. To copy otherwise, or republish,to post on servers or to redistribute to lists, requires prior specific permission and/or afee. Request permissions from [email protected].
ACM NanoCom 2019, September 25–27, 2019, Dublin, Ireland © 2019 Association for Computing Machinery.ACM ISBN 978-1-4503-9999-9/18/06...$15.00https://doi.org/10.1145/1122445.1122456
Superparamagnetic Iron Oxide Nanoparticles in a Molecular Communica-tion Setting. In
ACM NanoCom 2019: 6th ACM International Conference onNanoscale Computing and Communication, September 25–27, 2019, Dublin,Ireland.
ACM, New York, NY, USA, 6 pages. https://doi.org/10.1145/1122445.1122456
In the exploration of molecular communication strategies (see[6, 10] for an overview), testbeds play a central role as they al-low to evaluate communication theory, to identify new physicalaspects that have to be taken into account such as sources of inter-ference and to provide steps towards implementation of molecularcommunication for practical systems applications.For that purpose, testbeds have been proposed based on alcohol[3, 7] (see also [12] for spatial instead of temporal coding) andacids/bases [4] (refer to [5, 8] for corresponding transmitter andreceiver designs) as signaling molecules. A third type of informationcarrier was proposed by the authors in [11] for a testbed based onsuperparamagnetic iron oxide nanoparticles (SPIONs). In contrastto alcohol and acids/bases, these particles, which were originallydeveloped for magnetic drug delivery, are biocompatible and henceapplicable for use in humans. They can be fabricated relativelyeasily and are detected as a change in inductance of a measurementcoil wound around the propagation channel. Hence, detectors canbe operated noninvasively, i. e., they do not have to be inserted intothe propagation channel, which provides a major advantage forhuman use, e. g., with blood vessels as propagation channel.However, a major challenge when implementing testbeds for MCis the lack of appropriate detectors. In the present case, the receiverthat was used in [11], the commercial susceptometer Bartington ® MS2G, was not operated according to its original measurementpurpose, i. e., the characterization of material samples. It thereforeexhibited two considerable disadvantages: First, the detector coilhad a fixed width. This, on the one hand, posed restrictions on themaximum width of the propagation channel. On the other hand,for channel widths smaller than the detector width, enhancement a r X i v : . [ ee ss . SP ] J u l CM NanoCom 2019, September 25–27, 2019, Dublin, Ireland Bartunik and Lübke, et al. particleinjection pumpbackgroundflow pump detector coilssignal processingdata acquisition card waste bin
Figure 1: Testbed. From left to right are shown: Transmitterconsisting of water reservoir for background flow, syringe asreservoir for particle injection, two pumps for providing thebackground flow and the particle injection (upper and lowerpump, respectively) and Y-connector; propagation channelafter the Y-connetor; receiver consisting of the PCB’s for de-tector coils and signal processing (upper and lower PCB, re-spectively), data acquisition card and waste bin. of the measurement signal and thus of the detector sensitivity byreducing the width of the detector coil was not possible. Second, thesusceptometer, which is not intended for dynamic measurements,provided a maximum sample rate of only 10 samples / s. Hence, in or-der to achieve higher flexibility in the dimensions of the propagationchannel and to increase both the temporal and the measurementresolution, a new customized detector device for SPION-based MCis needed. In this paper, a first prototype of this device is presented.The proposed device offers yet another feature: It incorporatesa second coil to provide a reference signal. Hence, the detectorcan not only be operated in a single-ended signaling design withone propagation channel as in previous testbeds for molecularcommunication, but can also be used for a differential signalingdesign, where information is encoded in the difference of SPIONsconcentration between two tubes that serve as propagation channel.In the following sections, the testbed will be outlined (for moredetails, see [11]) and the design of the proposed detector will bedescribed. The device is evaluated by transmitting an exemplarytext message and by assessing its sensitivity for different concen-trations of SPION solutions. A discussion of the achievements andpotential future applications concludes the article. Figure 1 shows the complete testbed including the propagationchannel.
Information is encoded in the SPION concentration in aqueoussolution: A binary “1” is represented by the injection of SPIONparticles into the propagation channel, a binary “0” by no particlesbeing released. SPIONs originate from biomedical applications and thus havean established biocompatibility with ongoing studies on biosafeapplications [9]. The nanoparticles used as transmitter moleculesin this setup where synthesized by the Section of ExperimentalOncology and Nanomedicine (SEON) of the University HospitalErlangen. They have a hydrodynamic radius of 50 nm and a suscep-tibility of 8 . × − for 1 mg / mL. They are suspended in waterwith an estimated particle concentration of 5 × particles / mLfor a stock concentration of 10 mgFe / mL. The transmitter consists of the following components: a reservoirwith distilled water, which together with a first pump (Ismatec ® ISM831C) provides the channel medium in form of a backgroundflow with a continuous rate of 10 mL / min; a syringe as reservoirfor the SPION suspension; a second, computer controlled pump(Ismatec ® ISM596D) that injects a defined amount of this SPIONsuspension into the transmission channel. For that purpose, thetubes for background flow and particle injection are joint via aY-connector. The pump for SPION injection is controlled with aLabView application that allows to encode the desired binary se-quence in a series of particle releases, each with a volume of 104 µLand at a flow rate of 10 mL / min. A symbol duration of 1 s wasemployed. The propagation channel consists of a tube with a diameter of0 .
84 mm and a length of 5 cm. The channel begins after the Y-connector that joins background flow and particle injection andends at the detector coil of the receiver.
The bursts of SPIONs traveling through the propagation channelare measured with a detector coil as the SPIONs change the in-ductance value of the coil while they pass through it. This basicprinciple is exploited in magnetic susceptometers [1]. In this paper,the commercial susceptometer that was used in previous studies[11] (Bartington MS2G) is replaced by the device described in thefollowing section.
As in the previously used, commercial susceptometer, the nanopar-ticle are detected by a change of inductance of a coil that they passthrough. For the proposed design, two coils within a bridge circuitare utilized, one as reference and one for measurement [2, ch. 32].Combined with an appropriate capacitor each coil forms a resonatorcircuit that is driven by an amplified generator signal. SPIONs pass-ing through the measurement coil cause a change in inductance andas a consequence also a shift in resonance frequency. As a result thebridge circuit is unbalanced and the resulting differential voltage isdetected via an instrumentation amplifier and digital processing.
To amplify the generator signal and for use in the instrumentationamplifier, operational amplifiers that are driven by a differential ovel Receiver for SPIONs in an MC Setting ACM NanoCom 2019, September 25–27, 2019, Dublin, Ireland
10 MHz R C C L R C C L V diff Figure 2: Bridge circuit. The tunable capacitances C , C and tunable resistance R are used to balance the bridge dur-ing calibration. SPIONs that flow through the detector coilchange its inductance (either L or L ), which results in anon-zero voltage difference V diff . power source ( ± ±
15 V) were selected. To simplify the exter-nal power supply, all required voltages are produced by the devicefrom a single 5 V source.To supply the required ±
15 V voltage from the 5 V source, thesplit-rail converter TPS65131 (Texas Instruments) was used. The − −
15 V output via a negativevoltage regulator (LM79L05, Texas Instruments).
The resonator circuit is driven by a sine wave source with a fre-quency of 10 MHz, namely the ultra-low phase noise sine wavegenerator CVSS-945-10000 by Crystek. The voltage-tunable outputfrequency was set to 10 MHz by means of a 2 . Ω was connectedto the signal output, resulting in a peak-to-peak voltage of 3 V.As a higher voltage swing is desired, a current feedback amplifier(LT1223, Linear Technology) with a large output drive and a gainbandwidth of 100 MHz was used to amplify the generated signalto a peak-to-peak voltage of approximately 15 V. As the optimalamplification factor was determined during testing, a potentiometerwas fitted to adjust the resulting output voltage. A bridge circuit, typically consisting of four impedances, is imple-mented here with two resistors and the resonator circuits for thedetector coils as shown in Fig. 2. A potentiometer is used as resistor R and both branches contain adjustable capacitances ( C and C )to allow tuning of the resonance frequency. Simple switching ofdetector coils is enabled by use of SMA- (SubMiniature version A)headers.Ideally, if correctly tuned, the voltage V diff between the twobranches of the bridge circuit is zero as long as no nanoparticlespass through the detector coils. As soon as a difference in inductanceof one coil (either L or L ) occurs the corresponding resonatorcircuit will be de-tuned, the impedance changes, and the branchesof the bridge become unbalanced.By tuning the branches of the bridge circuit equally, a peak-to-peak difference voltage V diff below 40 mV could by achieved. This Figure 3: Detector coils mounted on PCB value hence constitutes the limit of tunability of the two branchesby use of the tunable resistor and capacitances.
To fit the requirements and geometry of the setup ideally, specialdetector coils were produced. They have an inner diameter thatallows a tube with a girth of up to 3 . To simplify the detection of voltage level changes in the bridgecircuit, an instrumentation amplifier was implemented using twocurrent feedback amplifiers of the type LT1395 from Linear Technol-ogy as input stage and the operation amplifier AD8033 from AnalogDevices as amplification stage. The chosen devices show low noiseover a sufficient bandwidth. The differential amplification of theinstrumentation amplifier can be adjusted with a potentiometer.The maximal amplification factor in the real-life applicationwas limited by the current the power source could provide. Aftertuning the branches of the bridge circuit, the amplification of theinstrumentation amplifier was adjusted to an output peak-to-peakvoltage of 1 V.
A simple digital measurement is made possible through an en-velope detector connected to the output of the instrumentationamplifier. The result is a DC-signal that varies in the range of 100 to500 mV. The output signal was captured with the data acquisitionmodule NI USB-6009 from National Instruments. The module hasa resolution of 11 bit with a programmable-gain amplifier and amaximal sample rate of 10 samples / s. The supplied LabView appli-cation “NI-DAQmx Base Data Logger” was used to digitally record CM NanoCom 2019, September 25–27, 2019, Dublin, Ireland Bartunik and Lübke, et al.
Figure 4: Signal acquisition circuitry. measured values. A sample rate of 100 samples / s was used for allmeasurements. The described circuitry was implemented on a four layered PCBconsisting of 1 .
55 mm thick FR-4 composite material with a squarelayout of 75 mm by 75 mm. Layers two and four, consisting solelyof ground planes, act as shielding. Layer 1 contains the signal com-ponents, while the power distribution is implemented in layer 3.All remaining space is covered with ground planes and connectedthrough a mesh of vias. Additionally, a via is placed closed to everysurface mounted device (SMD) with a ground connection. Figure 4shows the completely assembled detector device.Altium Designer 18.1.7 was used to layout the PCB, which wasproduced by the Multi Leiterplatten GmbH, Germany, and assem-bled with the own department facilities.
Balancing the bridge circuit is an important step for the function-ality of the device. To maximize the sensitivity as SPIONs passthrough the detector coil, the resonator circuits should ideally be inresonance at 10 MHz and identical to each other. Due to small dif-ferences in the inductance of the detector coils, a trade-off betweenideal resonance and identical resonance must be found.In the first step, the PCB was fitted with SMD capacitances ( C and C in Fig. 2) that shifted the resonance frequency into thedesired range. The tunable capacitances C and C were then ad-justed to maximize the signal output of each channel and thereforemaximize resonance. Using the potentiometer R , the peak-to-peakvoltage of both channels was matched. Finally, to compensate forthe influence of the used measurement tips, with a given capaci-tance of 16 pF, the tuning steps were repeated while measuring theoutput of the instrumentation amplifier, which in turn was adjustedto a peak-to-peak voltage of 1 V. All measurements were performedwith the oscilloscope WaveRunner LT262 (LeCroy). The proposed measurement device was evaluated in two respects:As a proof of concept, a simple text message was sent with thetestbed. Furthermore, the sensitivity of the device was assessedby comparing pulse series with different concentrations of SPIONsuspension.
A simple coding table was devised to transmit text messages com-posed of capital letters efficiently. Each codeword is headed by alogical “1” to allow receiver synchronization, followed by 5 bit rep-resenting the character to be coded. This latter code component isthe 5-bit binary representation of the number of the letter withinthe Latin alphabet, starting with zero for letter “A”. Hence,“A” ≡ “100000” , “B” ≡ “100001” , ... “F” ≡ “100101” , ... “U” ≡ “110100” , ... With this coding scheme, the sample sequence “FAU” with binaryrepresentation 100101 100000 110100 (1)was transmitted using a suspension with a particle concentrationof 7 . / mL.At the receiver, peak detection was performed by use of Matlab(MathWorks), before decoding the bit sequence. First noise in thereceived data was reduced by applying a moving average filter witha width of 21 samples. Next, a threshold to differentiate betweenthe logical values “1” and “0” was calculated from the received dataas the mean between the highest and the lowest value. Finally, therising edges were identified as threshold-crossings with positiveslope. Based on the time t of the first detected rising edge andsymbol interval T S , symbol intervals were defined by (for intervalno. k ) I k = [ t + kT S ; t + ( k + ) T S [ . (2)Bit no. k was set to “1” if the threshold was exceeded for at least30 % of samples in I k , and to “0” otherwise.A further test with a random sequence of 80 bit was performed:11001001 11111110 10110011 01101000 1000101001011001 01100011 11111000 11010101 00001000 (3) To characterize the quality of detection and allow for comparisonto other devices, the sensitivity of the receiver was determined. Tothis end, the bit sequence “10101” was transmitted using variousdilutions of the SPION suspension. Solutions with a particle con-centration of 10 mg / mL, 5 . / mL, 1 . / mL, 0 . / mL and0 . / mL were tested. Figure 5 gives a qualitative comparison of the measured receivercurves of the susceptometer that was used previously in [11] andof the proposed device (in red and blue, respectively). The figure ovel Receiver for SPIONs in an MC Setting ACM NanoCom 2019, September 25–27, 2019, Dublin, Ireland . . .
45 00 . . M e a s u r e d s i g n a l ( V ) · − S u s c e p t i b i l i t y ( S I ) Figure 5: Comparison of measured curves by use of the com-mercial susceptometer that was previously used (red) andthe proposed detector device (blue). clearly demonstrates the increased resolution of the signal with thenew detector.Compared to the susceptometer, the sampling rate of the pro-posed detector of 100 samples / s is already 10 times larger. It waschosen due to restrictions of the ready to use graphical user in-terface (GUI) that accompanied the data acquisition card. With acustomized data acquisition software, the sampling rate can furtherbe increased. The sample sequence “FAU” was successfully transmitted, as canbe seen in Fig. 6. All bits were detected correctly using the simplemethod described in the previous section.A second test with a random 80 bit sequence was performed,again with the optimal result of all peaks corresponding to a “1”being correctly detected (see Fig. 7).A characteristic waveform for each burst of particles with aquick rise time and a delayed return to zero, as the particles arewashed out of the detector coil by the background flow, can beobserved. Due to the slow decay of particles in the tube because oflaminar flow, successive bursts cause an accumulation of remainingparticles which increases the measured concentration and furtherdelays the return to zero, i. e., inter-symbol interference occurs.
The measured signals for the six tested SPION suspension dilutionsare shown in Fig. 8. The three peaks that were expected due to thetransmitted binary sequency with three equispaced ones can beidentified in all measurements except for the one with the lowestSPION concentration c = . / mL. The amplitudes scale withthe concentration and range from 0 . c =
10 mg / mL) to approx.0 .
02 V ( c = . / mL). For the considered setup the lowest con-centration that allows safe detection of a bit is therefore determinedas 0 . / mL. Successful communication at lower concentrationsmay be possible but would require more robust equalization andcoding schemes. 0 5 10 15 200 . . . . . M e a s u r e m e n t s i g n a l [ V ] Raw signal Filtered signalSymbol intervals Detected edges
Figure 6: Transmission test with text sequence “FAU” (bi-nary representation see (1)). Raw and filtered measurementsignals are plotted in black and blue color, respectively. Thesymbol intervals, as derived from the first detected risingedge, are indicated by dotted gray vertical lines, detectededges by solid red lines. Sufficient values above the thresh-old are found in each symbol interval that corresponds to abinary “1”. . . . . M e a s u r e m e n t s i g n a l [ V ] Raw signal Filtered signal
Figure 7: Transmission test with
80 bit random sequence (see(3)). Raw and filtered measurement signals are plotted inblack and blue color, respectively. Each peak correspondingto a binary “1” can be clearly identified.
An optimized receiver for detection of superparamagnetic iron ox-ide nanoparticles in a molecular communication (MC) testbed wasproposed. It was shown that the detector is capable of discriminat-ing different particle concentrations that pass through the propaga-tion channel, such that text messages and longer bit sequences canbe effectively transmitted.Compared to the previously used, commercial susceptometer,the proposed sensor device shows three advantages: First, it can beused with different detector coils, particularly custom made ones,
CM NanoCom 2019, September 25–27, 2019, Dublin, Ireland Bartunik and Lübke, et al. . . . M e a s u r e m e n t s i g n a l [ V ] (a) c = , . , . / mL (blue, green, red).Signals are offset corrected. . . . .
215 Time [s] M e a s u r e m e n t s i g n a l [ V ] (b) c = . / mL . . . . . .
170 Time [s] M e a s u r e m e n t s i g n a l [ V ] (c) c = . / mL . Figure 8: Sensitivity test. Measured series of three pulses for different particle concentrations c = , . , . , . , . / mL . which can be manufactured according to the dimensions of theMC testbed in use. Second, it provides an increased sample rate,which is currently limited by the software of the data acquisitioncard we used and which can be enhanced further by a custom madesoftware. And third, the reference, a second coil that the detuning ofthe measurement coil is compared to, is made available for the user.In this way, on the one hand, conventional single-ended signalingcan be implemented by fixing a tube with the channel medium(destilled water in our case) without particles as reference. On theother hand, differential signaling is now possible, too, by using twotubes as propagation channel: In this case, the transmitter encodesinformation in concentration differences of SPIONs between thetwo tubes rather than by particle injections into a single tube. Thiswould open up new possibilities for transmitter designs, e. g., bysteering particles into one tube or the other by magnetic fields,which would not suffer from the mechanical limitations of pumpsand, in combination with a closed-loop system, from the need of areservoir and a sink for the nanoparticles.Therefore, the next steps of research includes the realization ofa fully differential signaling testbed for molecular communication.Furthermore, the proposed sensor device will be evaluated andcompared to the commercial susceptometer as well as to alternativemeasurement approaches, particularly with respect to bit error ratesfor different measurement setups. Finally, an optimized receiver willallow to improve the testbed design and to investigate appropriateencoding and decoding strategies in a more detailed manner thanit has been possible so far. ACKNOWLEDGMENTS
This work was supported in part by the Emerging Fields Initiative(EFI) of the Friedrich-Alexander-Universitat Erlangen-Nürnberg(FAU), the STAEDTLER-Stiftung, and the German Federal Ministryof Eduction and Research (BMBF), project MAMOKO.
REFERENCES [1] Nilangshu K. Das, Sounak Dey, Sarbajit Pal, and Parthasarathi Barat. 2013. Inno-vative Instrumentation to Measure Magnetic Susceptibility.
IEEE Transactions on Magnetics
49, 9 (Sep. 2013), 4965–4969. https://doi.org/10.1109/TMAG.2013.2259247[2] Bob Dobkin and Jim Williams (Eds.). 2011.
Analog Circuit and System Design: ATutorial Guide to Applications and Solutions . Elsevier. https://doi.org/10.1016/B987-0-12-385185-7.00032-9[3] Nariman Farsad, Weisi Guo, and Andrew W. Eckford. 2013. Tabletop MolecularCommunication: Text Messages through Chemical Signals.
PloS ONE
8, 12 (Dec.2013), 1–13. https://doi.org/10.1371/journal.pone.0082935[4] Nariman Farsad, David Pan, and Andrea Goldsmith. 2017. A Novel Experi-mental Platform for In-Vessel Multi-Chemical Molecular Communications. In
GLOBECOM 2017 - 2017 IEEE Global Communications Conference . 1–6. https://doi.org/10.1109/GLOCOM.2017.8255058[5] Laura Grebenstein, Jens Kirchner, Renata S. Peixoto, Wiebke Zimmermann,Florian Irnstorfer, Wayan Wicke, Arman Ahmadzadeh, Vahid Jamali, GeorgFischer, Robert Weigel, Andreas Burkovski, and Robert Schober. 2019. BiologicalOptical-to-Chemical Signal Conversion Interface: A Small-Scale Modulator forMolecular Communications.
IEEE Transactions on NanoBioscience
18, 1 (Jan 2019),31–42. https://doi.org/10.1109/TNB.2018.2870910[6] Vahid Jamali, Arman Ahmadzadeh, Wayan Wicke, Adam Noel, and RobertSchober. 2019. Channel modeling for diffusive molecular communication – Atutorial review. https://arxiv.org/abs/1812.05492 submitted for publication.[7] Bon-Hong Koo, Changmin Lee, H. Birkan Yilmaz, Nariman Farsad, Andrew W.Eckford, and Chan-Byoung Chae. 2016. Molecular MIMO: From Theory toPrototype.
IEEE Journal on Selected Areas in Communications
34, 3 (March 2016),600–614. https://doi.org/10.1109/JSAC.2016.2525538[8] Bhuvana Krishnaswamy, Caitlin M. Austin, J. Patrick Bardill, Daniel Russakow,Gregory L. Holst, Brian K. Hammer, Craig R. Forest, and Raghupathy Sivakumar.2013. Time-Elapse Communication: Bacterial Communication on a MicrofluidicChip.
IEEE Transactions on Communications
61, 12 (December 2013), 5139–5151.https://doi.org/10.1109/TCOMM.2013.111013.130314[9] Antje Lindemann, Ralph Pries, Kerstin LÃijdtke-Buzug, and Barbara Wollenberg.2015. Biological Properties of Superparamagnetic Iron Oxide Nanoparticles.
IEEETransactions on Magnetics
51, 2 (Feb 2015), 1–4. https://doi.org/10.1109/TMAG.2014.2358257[10] Tadashi Nakano, Andrew W. Eckford, and Tokuko Haraguchi. 2013.
MolecularCommunication . Cambridge University Press, Cambridge. https://doi.org/10.1017/CBO9781139149693[11] Harald Unterweger, Jens Kirchner, Wayan Wicke, Arman Ahmadzadeh, DoaaAhmed, Vahid Jamali, Christoph Alexiou, Georg Fischer, and Robert Schober.2018. Experimental Molecular Communication Testbed Based on MagneticNanoparticles in Duct Flow. In . 1–5. https://doi.org/10.1109/SPAWC.2018.8446011[12] Linchen Wang, Nariman Farsad, Weisi Guo, Sebastian Magierowski, and An-drew W. Eckford. 2015. Molecular barcodes: Information transmission via per-sistent chemical tags. In2015 IEEE International Conference on Communications(ICC)