Building Software Defined Materials with Nanonetworks

ABSTRACTIn this paper, we present a class of programmable materials,whose electromagnetic properties can be controlled via soft-ware. These Software Defined Materials (SDMs) stem frommerging metamaterials with nanonetworks. Metamaterialsare artificial structures with properties that may not be foundin nature. They have inspired ground-breaking applicationsto a range of research topics, such as electromagnetic invis-ibility of objects (cloaking), radiation absorption, exquisitefiltering of light and sound, as well as efficient antennas forsensors and implantable communication devices. However,existing metamaterial structures are "rigid", i.e. they cannotbe restructured once constructed. This limits their fabrica-tion to a handful of well-equipped laboratories worldwide,slows down innovation, and, most importantly, restricts theirapplicability to static structures only. SDMs act as "plas-tic" (reconfigurable) metamaterials, whose properties can bechanged programmatically via a computer interface. Thiscontrol is achieved by a network of nanomachines, incorpo-rated into the very structure of the metamaterial. The nanoma-chines receive directives from a user and perform simple butgeometry-altering actions on the metamaterial structure, tun-ing its electromagnetic behavior. The present paper intro-duces SDMs, defining the concept and highlighting its promis-ing future aspects. Presently realizable implementation ap-proaches are given, alongside specifications of a suitable nano-networking model, its unique challenges and promising reso-lutions paths.

1. INTRODUCTIONPlasticity and programmability are important quali-

ties of computer systems. The �elds of Software De�nedNetworks, Storage and Radio have been formed to pur-sue plasticity in the respective research areas, acknowl-edging the general need for hardware reusability, cou-pled with constant system evolution [1, 2]. Up to date,programmability referred to Medium Access, Networkand Application layers and protocols. In the presentpaper we propose Software De�ned Materials (SDMs),which o�er programmatic control over their electromag-netic behavior, paving the way not just for plastic phys-

ical layers, but for a completely new class of network-ing applications as well. SDMs are made possible andare presently realizable by merging nanonetworking withmetamaterials, a class of arti�cially structured materialswith exquisite properties.Metamaterials consist of periodically repeating, µm-

structured, 2D or 3D patterns of conductive material.In other words, a pattern unit acts as a �building block�,creating the whole metamaterial via repetition [3]. Clas-sic 2D examples are shown in Fig. 1. In micro-scale,the laws of electromagnetism and optics are of courseretained. However, when observed from a higher scale,the metamaterial behaves as if these laws were over-ridden. The building blocks act as resonators, essen-tially bending, polarizing or slowing down laser rays orelectromagnetic waves, depending on the geometry ofthe building blocks [4]. Metamaterials exhibit a wideassortment of applications, such electromagnetic cloak-ing [5], antennas with augmented bandwidth, as well asversatile re�ection/di�raction of lasers and electromag-netic waves, such as custom beam redirection and per-fect re�ection [6]. However, metamaterials are presently�static�. The structure of their building blocks is �xed,and so are their electromagnetic properties. For exam-ple, making a cylindric object invisible to electromag-netic waves (cloaking) requires a diameter-dependent meta-material coating. A slight change in the dimensions ofthe cylinder requires new, highly complex calculationsand new metamaterial coating. Therefore, should therebe a mechanism to alter the geometry of the buildingblocks programmatically, one would have a material withexquisite, yet programmable physical properties.SDMs combine metamaterials and nanonetworking, re-

sulting into a new class of matter, with programmableelectromagnetic behavior. SDMs inherit the repeatedstructure of the metamaterials. However, their geome-try can be altered with the aid of nanonodes carefullyincorporated to the metamaterial structure. The nanon-odes act as programming agents who receive directivesfrom external devices (i.e. a PC), route them appropri-ately and translate them to geometry-altering actions.Nanonetworks refer to vast networks with extreme lim-itations in available node energy, wireless connectivity,

1

Keywords: Network Applications; Wireless; Theore-tial foundation/Modeling

C. Liaskos, A. Tsioliaridou, A. Pitsillides, N. Kantartzis, A. Lalas, X. Dimitropoulos, S. Ioannidis, M. Kafesaki, C. Soukoulis

CPU power and storage, rather than the actual spatialscale [7]. Due to these restrictions, some nanonetwork-ing approaches discard stack discrimination of proto-cols, packet queuing, error control and dynamic node ad-dressing capabilities, while focusing on multicast-basedparadigms [8]. These stringent operational conditions fa-vor node miniaturization and, subsequently, enable theirsuccessful incorporation to the physical structure of themetamaterials, at the required µm scale.Apart from introducing the SDM concept, the present

paper provides the following additional technical contri-butions: i) Two designs for SDMs are proposed. The �rstone relies on Micro-Electro-Mechanical Switches (MEMS)to alter the structure of a metamaterial. It is simpler andcheaper to implement but o�ers limited programmabil-ity. The second employs some exquisite properties ofgraphene to achieve extreme control at the cost of amore complex manufacturing process. Presently real-izable implementation plans for both cases are detailed.ii) SDM-tailored nanonetworking models for both casesare proposed, covering the communication among thenanonodes and the interfacing with the external pro-gramming environment. The networking challenges aredetailed and resolution approaches are outlined.Regarding their uses, SDMs can act as recon�gurable

metamaterials, extending their application pool to en-compass adaptivity and reusability. For example, ob-ject cloaking could be achieved with a universal SDMcoating and be programmatically switchable, partial orspectrum-selective. In addition, they could be used forcreating radars with no moving parts and, therefore,faster alignment. More importantly, SDMs constitute anovel application area for computer networking and serveas proof-of-principle of: i) The extension of the Internetto the level of materials and their physical properties,allowing for remote con�guration and control over thebehavior of matter. Automation follows, allowing for e.g.true, environment-adapting cloaking. ii) Cost-e�ective,accessible and reusable metamaterials, which by them-selves constitute a promising research �eld. The inter-disciplinary merge of computer networking and metama-terials is unique to the present paper.

2. BACKGROUNDPrior to detailing the SDM concept and the challenges

it entails, we shall provide a brief insight on the oper-ational principles of classic metamaterials. Such mediaare composed of periodically placed building blocks, inmost cases made of metallic elements on a silicon sub-strate, as in Fig. 1 (left). The achievable metamate-rial property depends on the geometry of these buildingblocks, for example the ring diameters and gap sizes inFig. 1 (top-left). In essence, one forms repeated patternson a substrate surface and the result as a whole yields therequired electromagnetic behavior. An example is given

Figure 1: Split-ring resonators (SRRs) andCrosses are common metamaterial patterns(left). Stacks of carefully designed SRRs can e.g.�cloak� an object (right).

in Fig. 1 (right). According to Snell's law of refraction,any material refracts an incident laser beam or electro-magnetic wave, altering its angle from Θ to Θ′. However,no natural material has ever been observed to yield neg-ative refraction angles. An SRR metamaterial can o�erany angle, either negative or positive, as a function of thegeometry of its building block. This exquisite propertyhas been exploited to construct electromagnetic invisi-bility cloaks, by carefully stacking metamaterials withdi�erent refraction angles [9]. A beam is bended aroundan object, as if it never interacted with it.Adjustable refraction angle is not the only property

that can be achieved by metamaterials. Di�erent build-ing blocks yield di�erent physical properties, such ashigh electromagnetic absorbance (bottom-right inset, Fig.1), near zero permittivity and permeability, peculiar aniso-tropic response (leading, e.g., to hyperbolic dispersionrelation), giant chirality, opposite phase and energy ve-locity, opposite Doppler e�ect, etc. [3]. Terminologyaside, these traits can be exploited in a variety of appli-cations, indicatively including wide-band telecommuni-cations, highly e�cient energy harvesting (photovoltaics,thermophotovoltaics), ultra high resolution imaging, sens-ing and military applications [10]. Nonetheless, a criti-cal limitation of metamaterials is their rigidness. Theirproperties stem directly from the geometry of their build-ings blocks, which obviously cannot be restructured onceconstructed. Thus, changing the radius of the cylindricalobject in Fig. 1 or the frequency of the incident waverequires the design of a completely new metamaterialcoating. The proposed SDMs address this application-restricting limitation.

3. SOFTWARE-DEFINED MATERIALS3.1 ConceptA Software De�ned Material is a 2D surface which

supports the programmatic creation of custom metama-terial building blocks (e.g. Fig. 1) and their o�set overthe SDM. An SDM comprises control modules and pas-sive material blocks. The control modules receive and

2

Figure 2: The architecture of the elementary unitof a Software-De�ned Material (left). A wirelessnanomachine receives commands and performsactions on passive material patches (right).

route programmatic commands and perform geometry-altering actions on the passive elements. These commu-nication and control duties are carried out by nanoma-chines, which act upon the passive segments of physicalmaterial (dubbed �patches�) as shown in Fig. 2.A nanomachine comprises a radio module (analog wire-

less communications), a rudimentary CPU, a Digital Sig-nal Processing unit and a front-end �actuator�. The wire-less approach is mandatory. Interconnecting or poweringa vast number of nanomachines with conductive wiringover the SDM surface can signi�cantly alter its electro-magnetic behavior. A harvesting-based, inductive powersystem provides the necessary energy [11].The actuator module of a nanonode can comprise a

physical switch in the form of trivial MEMS [12]. Switch-ing on and o� these physical connections acts as �stitch-ing� of the patches into a custom metamaterial pattern,as shown in Fig. 3. With MEMS, the stitching is per-sistent, meaning that the nanonodes need to be poweredjust as long as their programming requires. An actuatorcan operate without mechanical elements, providing bet-ter con�gurability and resilience at the cost of constantlypowered nodes. In this case, a node must continuouslyapply a voltage to a graphene-coated �patch�, controllingits transparency to the electromagnetic waves [13].

3.2 SDM Architectures and ImplementationThe capabilities of an SDM depend on the employed

implementation approach. Furthermore, each approachcan have di�erent nanomachine requirements and cande�ne a separate architecture. Two promising approachesare presented in the following subsections, alongside theirimplementation prospects.

3.2.1 MEMS-based ApproachFigure 3 presents an SDM approach based on pro-

grammable MEMS switches that alter the form of agiven metamaterial building block. The classic SRR isillustrated as a representative example in Fig. 3, butsimilar approaches can be followed for potentially anyother building block as well. Nanomachines can e�ec-

Figure 3: Nanomachines act as programmableswitches, "stitching" together copper rectangu-lar patches dynamically.

tively handle the connection between adjacent metal-lic patches, resulting in gaps with controllable width.Since the position of the gaps in conjunction with theangle of incidence are responsible for the overall SDMresponse, the proposed combination is able to alter theorientation of the device depending on the impingingwave. Nanomachines can dynamically decide the suit-able geometry for each case and implement it by prop-erly connecting or disconnecting the metal patches. Inthis manner, an SDM can yield a refraction angle cho-sen from a set of discrete, positive and negative values.Moreover, it is possible to set a di�erent refraction an-gle at each point of the SDM. Thus, one can achieve,for example, spatially varying electrical permittivity andmagnetic permeability response, ranging from positive tolarge negative values. At application level, these phys-ical attributes enable a variety of new devices, includ-ing controllable wave steering devices, electromagneticcloaks and arbitrary wavefront shapers.The key-characteristics of the MEMS approach are

discreteness and persistent actuation decisions: Not ev-ery refraction angle is possible (discrete set of choices),since it is not possible to freely alter the geometry ofthe building blocks. However, the �ipping of the MEMSis persistent, i.e. does not require constant node pow-ering. Furthermore, in this case the nanomachines areexpected to be sizeable (e.g. 5 × 5mm), comparable tothe metallic patches they interconnect, favoring ease ofimplementation further.

3.2.2 Graphene-based ApproachThe second approach for implementing SDMs adopts

not metallic, but graphene coated patches. An exquisiteproperty of graphene is that it can become transparent oropaque to electromagnetic waves by applying an externalbias (electrostatic �eld) [13]. Nanomachines can controlthis feature by applying the proper electric bias to theadjacent graphene segments (Fig. 2 - lower right inset).The design of a graphene-based SDM is given in Fig.

4. Essentially, the SDM surface becomes a nanomachine-controlled patch matrix, whereupon a programmer canform custom metamaterial building blocks at any point.Qualitatively, the concept is tantamount to drawing anycustom shape on a bitmap. Thus, the SDM can ex-

3

Figure 4: Graphene patches can be controlledto become transparent or opaque to electromag-netic waves, enabling the free formation of SDMpatterns on their surface.

hibit a diverse repertoire of spatially-varying electromag-netic behavior on-demand. However, in this case physicsrequires that the nanomachines be constantly powered,while they should also be considerably smaller than thecontrolled patches. Moreover, drawing patterns withgood resolution requires minimal patch size and a quicklyescalating number of nanomachines, accentuating manu-facturing cost and networking e�ciency considerations.

3.2.3 Justification of SDM realizabilityBoth of the proposed architectures are realizable with

presently available technology. The described, passivematerial patches can be constructed by subtractive struc-turing (etching) or laser ablation at 15µm− 50nm min-imal spacing and dimensions. Regarding nanomachines,Ultra-Low Power (ULP) wireless communications is thefocus of the IEEE 802.15.4q-ULP standard [14], andcommercial designs operate with 100µW inductive powersupplies [11]. MEMS switches can be constructed at ascale of a few µm with a good range of options (electro-static, thermal, piezoelectric or magnetic-based) [12]. Fi-nally, the complete SDM assembly, comprising nanoma-chines and patches in a grid topology, is straightforwardvia a Network-on-Chip approach, with companies o�er-ing commercial solutions already [15]. Furthermore, ananomachine with the architecture depicted in Fig. 2can be contained within a space of (50 → 200µm) ×(0.5 → 1mm) × (0.5 → 1mm) (width-height-depth, as-suming a top view). For the MEMS-based approach, apatch side must be 1 (min) to 5 (max) times the smallernanomachine dimension, i.e. 50−250µm at a minimum,which is also the minimal nanomachine spacing. For thegraphene-based approach, a patch side should be at least10 times larger than the smaller nanomachine dimensionwith, no upper bound.

4. SDM NANONETWORKING MODELThe previous sections provided necessary, low-level SDM

speci�cations. We proceed to contribute an appropriate,

high-level networking model for SDM nanonodes.

4.1 Abstracting the physics: higher-level con-nectivity, CPU and RAM modeling.

The topology of an SDM nanonetwork follows that ofthe controlled patches. In line with the presented imple-mentation approaches, the nanonodes can be assumedto form a square grid of several million vertices, com-prising {1 square patch-1 nanomachine} pairs. Fromthe SDM surface dimensions (input) and the nanonodespacing detailed in Section 3.2.3 (a given), one derivesthe node arrangement in rows and columns (N = n×m).The connectivity of each node in the grid can be con-

sidered as circular, as justi�ed in [8]. The connectivityradius is derived in a manner straightforward for wire-less communications [16]1. The operating frequency canrange within 1 − 100GHz, making for ≈ 10nsec-longdata packets carrying ∝ 101 bytes [17]. Related studieson nanonetworking have converged to simplistic modula-tion schemes (typically a direct representation of logical�1�s with very short pulses and silence for �0�s) and noerror correction capabilities [7, 17, 16, 8]. These choicesstem from power conservation and manufacturing costs.Given the vast number of nanonodes in an SDM, it is

obvious that each node should be as weak (and thereforecheap) as possible. In the present paper we aspire to pro-pose the most lightweight, yet SDM-capable, nanonet-working framework up to date. Speci�cally, we assume:i) the total absence of queueing capabilities in general,ii) a total memory capacity of ≈ 100 bytes and iii) alightweight, FPU-less CPU, able to perform integer cal-culations only2. Finally, all nodes are uniform.

4.2 Quasi-p2p communications paradigmNetworking a vast number of the outlined, extremely

weak nanonodes requires special approaches regardingtheir addressing, the type of relayed data and their reach-ability. We propose a quasi-p2p communications paradigmas a starting point. At �rst, we assume that the nodeshave an embedded, SDM-wide unique serial number bymanufacture. For example, in a 5×5 arrangement, node(1, 1) has a serial number of 1 and node (2, 1) is markedas the 6th. We proceed to propose a special data ex-change format. Each packet refers to any nanonode, butits translation depends on the serial number of the re-ceiver. Packets originate from an entity external to theSDM (programmer) and enter it at any point i, j. Twodata packet types are de�ned, depending on the natureof the information they carry. The con�g packet conveysdata on the nanonetwork topology, such as the number ofrows and columns, n×m. A node reads this information

1For the technically-adept reader: Tx power is 10nW −100µW . Signal to Interference plus Noise Ratio ≈ −10dB.The commonly assumed path loss model is free space, but atwo-ray-ground is also valid.2Namely, int counter increase/decrease, modulo operationsand evaluation of simple Boolean conditions

4

and derives its personal row-column indices, i, j, usingsimple, integer modulo arithmetic on its serial number.E.g. in the previous 5 × 5 example, the 6th node un-derstands that it is in position (2, 1). The setup packetcarries the directives for the formation of metamaterialpatterns on the SDM surface. A few bytes of informa-tion can describe the complete SDM form, exploitingthe repetitive nature of the metamaterial patterns andthe periodicity of the building blocks. An example isgiven in Fig. 5 (top inset). The depicted SRR build-ing block, comprises periodically repeating line segmentsand gaps. These are expressed in integer modulo-basedconditions, as functions of the i, j indices of the nanon-odes. Each nano-node receives these conditions in theform of a setup packet. It subsequently applies its per-sonal i, j indices and derives its intended, programmaticstatus. Moreover, notice that several line segments aresymmetric, compacting the packet size further, not re-quiring extra formulations. In essence, quasi-p2p pro-poses the exchange of programmatic functions, ratherthan plain raw data. A node receives a function, andevaluates it for its personal serial number. In our case,the function returns a simple Boolean value (true:actuateor false:do nothing). Thus, quasi-p2p combines the in-herent reachability and packet exchange minimizationbene�ts of broadcasting with the granularity of p2p com-mands, without explicit reference to any node. This re-sults in minimal-sized SDM directives, as shown in Fig.5. A sole setup packet of roughly 50 bytes describes thecomplete SDM.

4.3 Role-centric networkingIn classic communications, well-de�ned medium ac-

cess, routing and application layers receive, alter andencapsulate a packet separately. In this case queuing isa given, while extended, fast RAM and full-blown CPUsare there. Nano-networking protocols require out-of-the-box thinking. Why do we need a MAC layer? Thekey-answer is to limit interferences. Why do we needa routing layer? Primarily to ensure fast packet deliv-

ery via e.g. a short path. We therefore demonstrate alayer-less nanocommunications protocol that o�ers in-terference limitation and fast packet delivery, using thework of [8] as a promising starting point.We de�ne a role-based operation. A node starts of in

�neutral� role, during which it acts as packet retransmit-ter and also keeps separate count of packets it receivedsuccessfully or not. After a certain number of totalpacket receptions, a node assumes a �passive auditor� ora �blind retransmitter� role, depending on the aforemen-tioned counter values. Essentially, role-centric network-ing de�nes no MAC protocol and yet surpasses queuing-enabled handshake-based and �ood-based approaches insuccessful packet reception/transmission rate [8]. Fur-thermore, most nodes turn into �passive auditors�, con-serving energy. Moreover, �blind retransmitters� were

analytically proven to form predictable, well-de�ned, sym-metric patterns, as demonstrated in Fig. 5 (lower in-set). This symmetry ensures that packets travel overstraight lines, yielding shortest-path routing bene�ts aswell [8]. Finally, it has been analytically proven thatthe described traits are actually accentuated when thenumber of nodes increases [8].In the MEMS case, the role-centric approach can be

employed as-is. However, the graphene case introducesan interesting implication. A node cannot assume therole of �retransmitter� and apply voltage to a graphenepatch (�actuator� role) at the same time, due to powerrestrictions. In other words, the �actuator� and �retrans-mitter� roles are mutually exclusive. Thus, a node in-structed to �actuate� discards immediately the �retrans-mitter� role it may have assumed. An open issue iswhether the described properties of [8] continue to holdunder these circumstances. Simulation results provideinitial indications towards this direction:We assume a topology of 100 × 100 nodes, each with

a connectivity radius of 8 nodes (Fig. 5). Each nodeoperates as previously described3. SDM directives en-ter at the middle of the surface and propagate over it.We require the creation of a split-ring resonator SDM,as shown in the top inset of Fig. 5. We measure: i)the percentage of nodes that successfully receive (Rx)a given packet with directives, ii) the time required toreach this percentage (SDM programming speed), iii)the induced energy consumption rate measured by thenetwork-wide packet transmission rate (Tx). Further-more, we observe the formation of patterns of retrans-mitters. The SDM-tailored networking is compared to a�ood-based approach, which makes no di�erentiation be-tween �retransmitter� and �auditor� nodes (Fig. 6). Theresults refer to mean values over 1000 SDM setup trials.Both approaches yield almost 100% coverage. However,the SDM-tailored approach does so with ≈ 1/3 of the en-ergy expenditure of the �ood solution, as exhibited bythe respective packet transmission rates. Furthermore,it o�ers better SDM programming speed at the sametime. By assigning retransmitter roles to nodes withgood overall reception, the SDM-tailored approach limitsthe intereference events considerably, while retaining thesame rate of successful packet receptions network-wide.Finally, the retransmitters form symmetric patterns, de-spite the intervention of SDM-speci�c node roles.

4.4 Key-challengesThe envisioned key scienti�c challenges for SDMs are

the following. SDM topologies need to focus on min-imizing the number nanomachines, while still provid-ing a diverse set of programmable metamaterial pat-terns, reducing the manufacturing cost of SDMs. SDM-speci�c nanonetworking protocols are expected to an-3PHY parameter values: 100GHz frequency, 10nW trans-mission power, 10nsec packet duration.

5

Figure 5: Simulated creation of a split-ring res-onator (SRR) creation on an SDM.

tagonize over programming speed, i.e. the time requiredto recon�gure an SDM for new electromagnetic proper-ties. Notice that rapidly-programmable SDMs may un-lock new types of electromagnetic behavior stemmingfrom spatial and temporal variations. Reliabilty coverssecurity and error-resilience aspects. SDMs can connectmatter to the Internet, implying authorization concerns.A lightweight security mechanism to allow SDM pro-gramming to authorized users only can potentially relyon the quasi-random scrambling of node IDs during themanufacturing phase. Thus, SDM programming will bepossible only to programmers that know the setup of thenode IDs. However, more conditions will be required todescribe the SDM setup (Fig. 5), creating a trade-o�between packet size and security. Additionally, nanoma-chines are generally error-prone, due to the absence ofserious error correction schemes. Strict binary formatsfor packet directives and SDM tailored codebooks mayensure that faulty packets are detected and invalidatedwith high probabilty, prior to creating unplanned SDMgeometry alterations. Finally, measurement-based mech-anisms for detecting nanomachine failures can o�er amore programmer-friendly SDM interface, providing SDM�debugging� and real-time monitoring capabilities.

5. RELATED STUDIESResearch on metamaterials initially focused on proof-

of-concept implementations of materials that yield neg-ative refraction angles in the GHz frequency range [18],employing SRRs as building blocks (Fig. 1, 1). Subse-quent research gradually reached maturity, yielding so-lutions with minor power losses, operational at the THzregion [19]. The �rst magnetic metamaterial at opticalfrequencies was proposed in [20], which started a racetowards the visible, realized in 2006 [21]. Further en-

0

20

40

60

80

100

MeanCoverage

(%)

MeanCoverage

Time (nsec)

MeanTx rate

(pkts/nsec)

MeanRx rate

(pkts/nsec)

Mean packetcollision rate(pkts/nsec)

0

200

400

600

800

1000

FloodSDMNetworking

Figure 6: SDM-tailored networking yields highreachability, lower energy expenditure and im-proved SDM programming speed (coverage time)than a �ood-based approach.

hancements to boost the accessible operational frequencywere presented in [22]. However SRRs and refraction an-gle manipulation are just representative of the metama-terial potential. Further, indicative approaches can befound in [23]. It is worth noting that ongoing research onmetamaterials acknowledges the need for more �plastic�metamaterials [24]. First steps towards this direction re-fer to metamaterials that are still passive, but can e.g.react to temperature changes [24]. Yet, wide tunabilityand control (even non-point-to-point) are not possible.Miniaturized, wirelessly communicating nodes were orig-

inally outlined in [25]. Jornet et al. proposed a statisticalTHz-wireless channel model for nanomachines dispersedin gas mixtures [16]. Point-to-point node communicationis discussed at a higher level in [17, 7]. Arguing that thisparadigm may not be appropriate for the severely re-stricted nanoenvironment, Srinkath et al. proposed theclustering of nanonodes into groups, delegating commu-nication abilities only to more powerful cluster masters[26]. The nanonodes should still support an addressingprotocol, a timing system for duty-cycle operation, anda few powerful cluster heads dispersed throughout thecovered area. Authors in [8] addressed these limitations,as described in Section 4.3.

6. CONCLUSIONThe present paper introduced a novel application area

for networking, allowing for programmatic control overthe laws of electromagnetism. This is achieved via theproposed Software-De�ned Materials, an innovative com-bination of nanonetworks and metamaterials. The lat-ter are arti�cially designed materials, with unnatural,geometry-dependent electromagnetic properties. A net-work of nanomachines receives external, programmaticcommands and performs geometry-altering actions, yield-ing tunable electromagnetic behavior. The present paperabstracted a �tting networking model from the physicalunderpinnings of metamaterials and outlined its chal-lenges. Furthermore, it provided the foundations andrealizability clari�cations required to spur research onthe proposed, new �eld of research.

6

7. REFERENCES

[1] M. Kaminsky et al, �IOFlow: a software-de�nedstorage architecture,� in the 24th ACM

Symposium, pp. 182�196.[2] K. Tan et al, �Sora: high-performance software

radio using general-purpose multi-core processors,�Communications of the ACM, vol. 54, no. 1, p. 99,2011.

[3] B. Banerjee, An introduction to metamaterials and

waves in composites. Boca Raton and FL: CRCPress, 2011.

[4] M. Kafesaki, N. H. Shen, S. Tzortzakis, and C. M.Soukoulis, �Optically switchable and tunableterahertz metamaterials throughphotoconductivity,� Journal of Optics, vol. 14,no. 11, p. 114008, 2012.

[5] K. Iwaszczuk et al, �Flexible metamaterialabsorbers for stealth applications at terahertzfrequencies,� Optics Express, vol. 20, no. 1, p. 635,2012.

[6] B.-X. Wang et al, �Frequency Continuous TunableTerahertz Metamaterial Absorber,� Journal ofLightwave Technology, vol. 32, no. 6, pp.1183�1189, 2014.

[7] P. Wang et al, �Energy and spectrum-aware MACprotocol for perpetual wireless nanosensornetworks in the Terahertz Band,� Ad Hoc

Networks, 2013.[8] C. K. Liaskos and A. N. Tsioliaridou, �A Promise

of Realizable, Ultra-Scalable Communications atnano-Scale: A multi-Modal nano-MachineArchitecture,� IEEE Transactions on Computers,vol. PP, p. 1, 2014.

[9] D. L. Sounas and A. Alù, �Extinction symmetryfor reciprocal objects and its implications oncloaking and scattering manipulation,� Opt. Let.,vol. 39, no. 13, 2014.

[10] T. J. Cui, D. R. Smith, and R. Liu, Metamaterials:

Theory, design, and applications. New York:Springer, 2010.

[11] G. Dolmans, �imec range wireless projects,� HighTech Campus Eindhoven, 2013. [Online]. Available:http://www.hightechcampus.com/view�le.php/633

[12] H. Tao et al, �MEMS Based Structurally TunableMetamaterials at Terahertz Frequencies,� Journalof Infrared, Millimeter, and Terahertz Waves,vol. 32, no. 5, pp. 580�595, 2011.

[13] A. Vakil and N. Engheta, �Transformation OpticsUsing Graphene,� Science, vol. 332, no. 6035, pp.1291�1294, 2011.

[14] IEEE, �IEEE 802.15 Ultra Low Power WPAN:Task Group 15.4q,� Nov 2013. [Online]. Available:http://www.ieee802.org/15/pub/TG4q.html

[15] NetSpeed Systems, �Network-on-Chip Solutions,�2014. [Online]. Available:

http://www.netspeedsystems.com/[16] J. M. Jornet and I. F. Akyildiz, �Channel

Modeling and Capacity Analysis forElectromagnetic Wireless Nanonetworks in theTerahertz Band,� IEEE Trans. on Wir. Comm.,vol. 10, no. 10, pp. 3211�3221, 2011.

[17] J. M. Jornet, P. J. Capdevila, and P. J. Solé,�PHLAME: A Physical Layer Aware MACprotocol for Electromagnetic nanonetworks in theTerahertz Band,� Nano Comm. Networks, vol. 3,no. 1, pp. 74�81, 2012.

[18] P. Parimi et al, �Photonic crystals,� Nature, vol.426, no. 6965, p. 404, 2003.

[19] P. Tassin et al, �Graphene for TerahertzApplications,� Science, vol. 341, no. 6146, pp.620�621, 2013.

[20] C. Enkrich et al, �Magnetic Metamaterials atTelecommunication and Visible Frequencies,�Physical Review Letters, vol. 95, no. 20, 2005.

[21] M. W. Klein et al, �Single-slit split-ring resonatorsat optical frequencies: limits of size scaling,�Optics Letters, vol. 31, no. 9, p. 1259, 2006.

[22] K. Fan et al, �Optically Tunable TerahertzMetamaterials on Highly Flexible Substrates,�IEEE Trans. on Terahertz Sc. and Tech., vol. 3,no. 6, pp. 702�708, 2013.

[23] W. Zhu et al, �MicroelectromechanicalMaltese-cross metamaterial with tunable terahertzanisotropy,� Nature Communications, vol. 3, p.1274, 2012.

[24] A. Q. Liu, W. M. Zhu, D. P. Tsai, and N. I.Zheludev, �Micromachined tunable metamaterials:a review,� Journal of Optics, vol. 14, no. 11, p.114009, 2012.

[25] I. F. Akyildiz, F. Brunetti, and C. Blázquez,�Nanonetworks: A new communication paradigm,�Computer Networks, vol. 52, no. 12, pp.2260�2279, 2008.

[26] V. Srikanth et al, �Energy E�cient, Scalable andReliable MAC Protocol for ElectromagneticCommunication,� Int. Jrn. of Distr. and Parallel

Syst., vol. 3, no. 1, 2012.

7