Content uploaded by Weisi Guo
Author content
All content in this area was uploaded by Weisi Guo on Feb 06, 2018
Content may be subject to copyright.
1
SMIET: Simultaneous Molecular Information and
Energy Transfer
Weisi Guo1, Yansha Deng2, H. Birkan Yilmaz3, Nariman Farsad4, Maged Elkashlan5,
Chan-Byoung Chae3, Andrew Eckford6, Arumugam Nallanathan2
Abstract—The performance of communication systems is fun-
damentally limited by the loss of energy through propagation
and circuit inefficiencies. In this article, we show that it is
possible to achieve ultra low energy communications at the nano-
scale, if diffusive molecules are used for carrying data. Whilst
the energy of electromagnetic waves will inevitably decay as
a function of transmission distance and time, the energy in
individual molecules does not. Over time, the receiver has an
opportunity to recover some, if not all of the molecular energy
transmitted. The article demonstrates the potential of ultra-low
energy simultaneous molecular information and energy transfer
(SMIET) through the design of two different nano-relay systems,
and the discusses how molecular communications can benefit
more from crowd energy harvesting than traditional wave-based
systems.
Index Terms—biological energy harvesting, diffusion channel
modeling, energy efficiency, energy harvesting, molecular com-
munication, nano-scale machines
I. INTRODUCTION
Over the past decade, there is a growing focus on increasing
the energy efficiency of both mobile and fixed wireless sys-
tems. Whilst we have built up a good understanding of power
consumption mechanisms in terrestrial mobile networks, we
still lack understanding in how nano-machines1can com-
municate in an energy efficient manner. It is envisaged that
nano-scale communications will be critical to nano-machines
that seek to coordinate tasks such as in vivo drug delivery
and surgery [1]. Whilst many meso- and macro-scale in
vivo medical devices (i.e., pacemaker) are battery powered,
nano-batteries (50 microns [2]) are still significantly larger
or of the same dimension as their nano-machine counter-
parts. Therefore, charging batteries using externally generated
acoustic [3] and electromagnetic radiation is not always viable
and furthermore, nano-machines can be embedded in vivo
areas that are either sensitive to radiation or difficult for
radiation to penetrate. As such, energy harvesting from the
nano-machines’ locality is needed. Over the past decade, there
is increasing interest to harvest energy from communication
signals and achieve simultaneous wireless information and
power transfer (SWIPT). In this article, we draw on our
1W. Guo is with the University of Warwick, UK. 2Y. Deng and A. Nal-
lanathan are with King’s College London, UK. 3H. B. Yilmaz and C.-B. Chae
are with Yonsei University, Korea. 4N. Farsad is with Stanford University,
USA. 5M. Elkashlan is with Queen Mary University of London. 6A. Eckford is
with York University, Canada. Corresponding Author: yansha.deng@kcl.ac.uk
1Nano-machines are devices that are constructed using nanoscale technol-
ogy and typically have dimensions of 1-10 microns.
understanding of energy consumption and harvesting knowl-
edge in current wireless systems to better understand nano-
scale communications and exploit opportunities. In so doing,
we propose simultaneous molecular information and energy
transfer (SMIET).
A. Nano-Scale and Molecular Communications
The traditional practices of wireless planning with known
coverage areas and propagation environments starts to break-
down at the micro- and nano-scales. Communication systems
in complex biological environments must be: bio-compatible,
low power consumption, low complexity, small dimension,
and achieve reliable signalling in a fluid environment with
complex cell/tissue obstacles. Such constraints are challenging
for both EM-based THz systems and nano-acoustic systems.
Inspired by the abundant use of molecules in biological signal-
ing, Molecular Communication via Diffusion (MCvD) utilizes
molecular signal (i.e., a chemical pulse) as an alternative
carrier for information [4]. MCvD avoids the limitations of
wave generation and propagation, and allows the signal to
both persist and propagate to areas that are difficult to reach.
Indeed, this is perhaps why MCvD is prevalent in nature, both
at the inter-organism and at the inter-cell scales. In terms of
application, by allowing nano-scale devices to communicate
with each other, the potential of nano-robotics in medicine
maybe unlocked [1]. The economic potential is enormous: the
present market size is at $100 billion, growing at a projected
14% per year to $100 billion in 2016 (BCC research).
This article speculates on the potential of ultra-low energy
communications using molecular message carriers, especially
in the context of nano-communications. Nano-machines are
likely to require low data rate communications that lack the en-
ergy and complexity to perform coherent channel estimation,
synchronization, and range or location estimation. Therefore,
nano-machines need to perform basic mechanical tasks and
communicate to each other whilst expending a low amount of
energy. We set out the potential of energy harvesting using
molecular communications and design efficient and reliable
systems.
B. Biological Energy Harvesting Systems
In fact, energy harvesting using signaling molecules is
common in biology at the cellular level. One such biological
example that utilizes two types of molecules is depicted in
Fig. 1. Type A molecule generated by the source is absorbed
arXiv:1605.09474v1 [cs.ET] 31 May 2016
2
GABA
Transporter
Glutamine
Transporter
GABA
Glutamate
Glutamine
Glutamine
Glutamate
GABA
GABA
GABA
Synaptic Cleft
GABA-T
Glial Cell
Postsynaptic Cell
Postsynaptic
Receptor
Fig. 1. GABA reuptake mechanism at the synaptic cleft. Glial cell ab-
sorbs/harvests GABA molecules and converts them to Glutamine for utilizing
at the signaling mechanisms of presynaptic region.
and converted to type B molecule inside the receiving node to
be utilized in another subsequent signaling mechanism.
The γ-Aminobutyric acid (GABA) metabolism and uptake
is widely distributed across almost all regions of mammalian
brain. GABA is constructed by glutamate via enzymatic reac-
tion with glutamic acid decarboxylase (GAD) in the presynap-
tic neuron cell, which is then released as a neurotransmitter
for sending signal to both neighbor Glia cells and postsynaptic
neuron cells via GABA transporters (GATs). In this example,
the presynaptic neuron cell acts as the source to emit the
GABA as type A molecule, and Glia cell acts as the molecule
harvesting node, and emit glutamine as type B molecule in
response to the electrical charge polarization caused by GABA.
This example of molecule harvesting and manipulation will
be reverse engineered in this article to demonstrate how energy
harvesting in MCvD is in some ways similar to current sys-
tems, but with new opportunities for exploitation. This article
is organised as follows. In Section II, we discuss the general
power consumption model for the generation, propagation,
and reception of data. We compare MCvD and RF power
consumption models and offer a physical explanation on why
MCvD avoids the heavy propagation losses associated with
wave-based communications. In Section III, we demonstrate
how to achieve ultra-low energy communications via two
nano-relay systems: (1) single molecule type with interference
shielding, and (2) multiple molecule types with fragmenta-
tion and synthesis. In Section IV, we discuss how recent
advances in relay-based Simultaneous Wireless Power Transfer
(SWIPT) and crowd energy harvesting can be leveraged for
MCvD. In Section V, we conclude the findings and discuss
multi-disciplinary communication research opportunities going
forwards.
II. POW ER CONSUMPTION MO DE L
The power consumption model of a generic wireless system
can be approximately modularized into a number of con-
tributing components. In this paper, the authors focus on the
radio layer of consumption (including the data modulation,
amplification, antenna and propagation effects). The overhead
consumption due to signal processing and cooling elements
are left for future discussions. They are chiefly: i) the receiver
antenna gain with radius R, ii) the free-space propagation
loss λ, iii) the transmitter efficiency (i.e., power amplifier
efficiency µor chemical synthesis cost φ), iv) absorption loss
τ(also known as transmittance), and the v) the circuit power
consumption. In general, the received electromagnetic (EM)
power (PRx) or molecular number (NRx) is a small fraction of
the total power extracted by the transmitter PTotal:
PRx
PTotal ∝µ×τR2
dαfor EM
NRx
PTotal ∝1
φ(nTx −1) ×R
d+Rfor MCvD: t→ ∞,
(1)
assuming that the communication circuit power is relatively
small in a nano-machine. Like RF communications, there is
an energy efficiency factor for generating NTx molecules for
transmission. This is related to the number of basic chemical
components per molecule nTx and the energy cost to bind or
synthesize them φ[5], [6]. We now explain the reasoning and
details of the efficiency equations given in Eq. (1) in the rest
of the section.
A. Electromagnetic (EM) Wave Communications
We first examine the EM communications case. One can
see that the best achievable efficiency in Eq. (1) is limited
by d−αfor RF and limited by τd−αfor THz nano-scale
communications (where τ∝exp(−kf d)). The absorption loss
τis log-linear proportional to absorption coefficient k, which
depends on the chemical composition of the medium and is
typically 1×10−5for air and 1−3for water at f= 0−10 THz
[7]. To illustrate the aforementioned reasoning, we show an
illustration of radio wave versus molecular propagation in
Fig. 2. In the (a) subplot, an isotropic EM antenna transmits
to a receiver antenna with an effective area Aeff ∝(fR)2, and
the resulting received power after propagating a distance of d
is µEM R2
d−α, where the value of µEM is typically 10-30% for
RF systems [8].
B. Molecular Communication via Diffusion (MCvD)
MCvD on the other hand, relies on message bearing
molecules to freely diffuse from the transmitter to the receiver.
We do not consider the base energy cost of physical matter
(i.e., the molecules) as matter is not lost in the communication
process. We do consider the energy cost of creating specific
chemical compounds, as well as the energy benefits of restruc-
turing the compound.
In general, MCvD involves messenger molecules perform-
ing a random walk motion across the communication channel
through collision interaction and a diffusion gradient. For each
3
d
R
Point Source
Receiver Gain A, with
Effective Area Aeff
d
R
Point Source
Receiver Gain A,
with Radius R
(a) Radio Frequency (RF) Channel
(b) Molecular Communication via Diffusion (MCvD) Channel
PTotal
PTx
PRx
NTx
NRx
Power Amplifier
Efficiency µRF
PTotal
Synthesizing
Cost ϕ
Distance Dependent
Propagation Loss, λ(f,d)
Time and Distance
Dependent Propagation Loss,
λ(D,d,t)
f
D
Macro-molecule
synthesized from
basic components
Fig. 2. Illustration of power loss in transmitting signals in (a) Electromag-
netic (EM) Wave communications, and (b) Molecular Communications via
Diffusion (MCvD).
emitted molecule, there is a finite probability that it will reach
the intended receiver. The power in the system at any given
time instance is proportional to the number of molecules.
Whilst the stochastic process is intuitively unreliable and
requires a transmission time that is orders of magnitude longer
than wave propagation, these deficiencies can be mitigated
by communicating at very small distances (microns) or with
the aid of strong ambient flow. For a basic random walk
process, consider (as in Fig. 2b) a point emitter that transmits
NTx molecules. The full absorption receiver, will capture NRx
molecules given by [9]:
NRx =NTxhc, hc=R
d+Rd
√4πDt3e
−d2
4Dt ,(2)
where hcis known as the first passage time density distribu-
tion.
The resulting expected number of received molecules up
to time t=Tis NTxFc, where Fc=R
d+Rerfc(d
√4DT ). This
converges to NTx for 1-dimensional (1-D) space and NTx R
d+R
for 3-D space as t→ ∞. This means the full harvest of
all transmitted molecules is possible in certain conditions,
independent of the transmission distance. Naturally, the reality
is that molecules will have a half life and not all data bearing
molecules can be harvested. Reactions with other chemicals
(i.e., enzymes) in the environment can reduce the effectiveness
of energy harvesting in MCvD over time [10]. Yet, the poten-
tial to capture the vast majority of the transmitted molecules
due to the random walk nature of propagation demonstrates the
potential of MCvD over wave-based transmission. As with RF
communications, there is a cost to produce the NTx molecules
at the source in the first place. This can be shown to be [6]:
PTotal =φ(nTx −1)NTx, where φis the synthesising cost of
2 3 4 5 6 7 8 9 10
Distance (d)
10-6
10-5
10-4
10-3
10-2
10-1
Received Power PRX or Molecules N RX
PRX for α=2
PRX for α=4
NRX for T=12hours
NRX for T=1week
RF Received Power with varying
pathloss exponents, α
MCvD Received Molecules
with varying absorption periods, T
RF Free-Space
Decay (d-2)
RF Aggressive
Decay (d-4)
MCvD Asymptotic
Decay (d-1)
MCvD Aggressive
Decay (d-1erfc(d))
Fig. 3. Plot of received EM RF power (PRx) or MCvD molecules (NRx ) for
different different transmission distances (d). The results show that MCvD
can achieve asymptotic distance-dependent power decay at the rate of ∝d−1,
which is superior to all RF scenarios. Modeling parameters: mass diffusivity
(water molecules in air) D= 0.28 cm2/s, RF frequency 5GHz with parabolic
receiver antenna Aeff = 0.56πR2, and receiver radius of R= 10 cm.
bonding nTx amino acids per transmitted molecule.
Comparing MCvD with RF propagation at the macro-scale
to draw similar levels of performance (see Fig. 3), the received
RF power is ∝d−α, where αtypically varies between 2 to
4. On the other hand, the received molecules from MCvD
can asymptotically be ∝d−1, and at best independent of din
1-D space, provided the receiver is willing to wait for a long
time t→ ∞. Yet, the long waiting time is not as ridiculous
as it may appear for two reasons. Firstly, the rate of diffusion
is in reality accelerated by ambient air flow (i.e. convection
currents) or shortened to a few mili-seconds at the nano-scale.
Secondly, when one transmits a continuous stream of symbols,
the power emitted for the first symbol will be recovered by
the N-th symbol’s time (when Nis large). Hence, there are
no incurred delays to power or energy recovery in MCvD,
provided a long stream of symbols are transmitted.
III. DESIGNING NANO-REL AY SMIET
In order to utilize the energy harvesting concept, one needs
to have a system where the absorbed molecules are reused for
new transmissions. We propose two types of SMIET nano-
relays that can achieve low energy nano-scale communica-
tions. As illustrated in Fig. 4, we propose a 2-hop energy
harvesting relay system, where the relay is not only capable
of demodulating the information transmitted by the source at a
distance d1away, but also harvesting energy by collecting the
absorbed molecules. With the absorbed molecules at the relay,
the relay transmitter then re-emits the received molecules to
the intended destination at a distance d2away.
A. Single Molecule Type: Self Interference
We first consider the case where only one molecule type is
used on both links of the 2-hop relay. That is to say, every
receiver can absorb the molecule transmitted at the source
and retransmitted at the relay. Hence, in this scheme, the
4
R1
d1
Destination
Receiver
Source
Transmitter
(Synthesis)
(a) Single Molecule Type Relay with Interference Shielding
Energy Harvesting
Molecular Relay with Shielding
R2
d2
Transceiver
Link (Lossless)
(b) Multiple Molecule Relay with Chemical Converter
Shielding Options
against Self Interference
DA
Relay
Transmitter
R1
d1
Destination
Receiver
Source
Transmitter
(Synthesis)
Energy Harvesting
Molecular Relay with Converter
R2
d2
Transceiver
Link (Lossless)
DA
Relay
Transmitter
(Synthesis)
Chemical Bank of
basic components
(i.e., Amino Acids)
Molecule
Type A
Molecule
Type A
Molecule
Type B
Molecule
Type A
nB basic
components
nA basic
components
DB
DA
rB
rA
2H
rA
nB basic
components
rB
nA - nB additional
basic components
Synthesizing
Cost ϕB
Synthesizing
Cost ϕB-A
Multiple Molecule Type
Synthesis Mechanism
Self Interference
Straight Shielding Size
H
Relay
Receiver
Relay
Transmitter
h
Self Interference
Path must pass in
region from h=H
to beyond
Relay Receiver
(Fragment)
Relay Receiver
Fig. 4. Illustration of energy harvesting MCvD nano-relay systems: (a) single molecule type with self-interference shielding options, and (b) multiple molecule
types with chemical converter.
transmitter causes interference at the receiver in later symbol
slots due to longer propagation time. As illustrated in Fig. 4a,
this kind of system will also induce self-interference at the
relay’s receiver, as the relay transmitter’s emission will likely
to be immediately absorbed by the relay’s own receiver. Hence,
this type of system requires self-interference shielding at the
relay between the receiver and transmitter. In particular, we
propose the following: a) the relay transmitter needs to be at a
certain distance away from the receiver, and b) be separated by
a physical shield in order to avoid any emitted molecules being
immediately absorbed by the relay receiver. Two shielding
options are considered: (i) a spherical shield that partially
protects the relay receiver, so that molecules can only enter
from the source transmitter side, giving the relay receiver
directionality [11], and (ii) a straight shield that separates the
relay receiver and transmitter, similar to a knife-edge channel
[12]. The latter shield design can be relatively easily analyzed,
where any potential self-interference at the relay will need
to transverse via the region of h=Hto +∞(see Fig. 4a
right panel diagram). This transition probability can be found
to be: R+∞
H2f(t, h) dh, where 2fis the hitting distribution
applied to a hemisphere with the shield acting as a perfectly
reflecting boundary. The resulting self-interference probability
is therefore ∝erfc(H), which is intuitively minimized for a
large value of H. The specific design, especially in a nano-
machine sense will depend on the channel diffusivity Dand
other parameters beyond the scope of this article.
The binary concentration shift keying (CSK) with molecular
type A is considered as an example. In the first source-to-relay
(SR) link, the source emits NTx-S molecules to transmit bit-1
(a= 1), and emits 0 molecules to transmit a bit-0 (a= 0). The
relay simultaneously receives information and energy from
the SR link. The expected number of adsorbed molecules
(NRx-SR,a) at the relay during a bit interval NRx-SR[K], which
is contributed by the absorbed molecules due to the previous
K−1bits transmission at source. The molecules retrieved from
the SR link can be decoded and used directly as molecules
to carry data to the destination via the relay-to-destination
(RD) link. The total transmitted molecules from the relay
is under the power budget of NRx-SR[K]. Therefore, the RD
link cannot reliably encode by using the SR energy harvesting
process alone and appropriate safeguards need to be designed
to improve transmission reliability, such as introducing latency
[13]. We discuss these techniques in greater detail in Section
IV and leave detailed research in this area for the community.
B. Multiple Molecule Types: Fragmentation and Synthesis
We now consider the case where two molecule types are
used at the two links of the relay. In general, this can be
extrapolated to multiple hops with multiple molecule types.
Utilizing more than one molecule type is attractive for two
reasons. First of all, as mentioned in the previous section, we
are motivated to avoid self-interference at the relay. By using a
different molecule type (i.e., molecule type B), we can avoid
5
the necessity of a shield and significant separation distance
between the relay transmitter and receiver. Secondly, as this
section will explain, by changing the molecular composition,
the relay is able to control the rate of diffusion through the
relationship between molecule size and the rate of diffusion,
which has been recently explored in [5].
As illustrated in Fig. 4b, this kind of system requires all
the molecule types to be constructable from basic chemical
components such as amino acids (see GABA example in
Section I). Therefore, a chemical bank with the building
components is required at the relay, such that molecules of
type A can be transformed into type B, either by adding new
components or removing them. The cost of adding an amino
acid to the chain is independent of the type of the amino
acid. Therefore, the difference between type A and type B
molecules determines the energy efficiency for synthesizing
type B molecules from the harvested molecules at the relay
node. In this particular example, molecule type A is composed
of nAamino acids and molecule type B is composed of
nB6=nAamino acids. As mentioned in Section II, the cost to
producing a single molecule that comprises of namino acids
is [6]: (n−1)φ, where φis the synthesising cost and is given
as φ=202.88 zJ.
1) SR Link: In the first SR link, molecule type A is
used to carry data and as with the previous section. The
expected number of absorbed molecules (NRx-SR,a) at the
relay during the k-th bit can generally be expressed as:
NRx-SR[K] = NTx-S PK−1
k=0 ak[Fc((k+ 1)T)−Fc(kT )] for
K−1previous bits absorbed. If all bits are 1 (i.e., ak= 1
for all kfor some line-code), the energy harvested during bit
interval Tis equal to the total energy available to be harvested:
NRx-SR[K]→R1
R1+d1erfc(d
√4DKT )which asymptotically con-
verges to R1
R1+d1. Fig. 5 shows histogram plots of harvested
number of molecules at the relay receiver with varying line-
coding bit-1 probabilities: P(ak= 1) = [0.15,0.5,0.85].
The results show that the distribution of the number of
molecules harvested can be described by a Generalized ex-
treme value distribution (GEV) with varying parameters, i.e.,
NRx-SR ∼GEV(µ, σ, ζ). In general, the energy harvesting
potential in the SR link is quite high (approximately 12-
25% for P(ak= 1) >0.5) at a distance of 1mm. As the
technology for nano-machines shrink and the inter-distance
between machines also reduce, the potential to harvest energy
improves linearly (see Eq.(1)).
2) RD Link: In the second RD link, molecule type B is
used. Here, we have an opportunity to design type B such
that it is sufficiently dissimilar to type A to avoid chemical
interference, and to maximize the communication link perfor-
mance. It is well known that in the limit of low Reynolds
number (laminar flow), the mass diffusivity parameter Dis
inversely proportion to the dimension of the molecule through
the Stokes-Einstein relation: D∝r, where ris the radius of
the molecule which is related to the number of amino acids
required for each molecule [5]. By controlling the number of
amino acids in molecule type B in the RD link, the relay can
control the following:
•The total number of molecules (type B) available for
P(ak=1) = 0.15 P(ak=1) = 0.5 P(ak=1) = 0.85
Fig. 5. Histogram and distribution fit plots of harvested number of molecules
at the relay receiver with varying line-coding bit-1 probabilities P(ak= 1).
The distribution of molecules harvested can be modelled as a GEV and a
distribution fit is shown with modeling parameters D= 1 ×10−5cm2/s,
d= 1mm, and R= 0.2mm. In all cases, K= 50 bits were considered with
an 1000 iterations.
transmission, NTx-R[K] = NRx-SR [K]nA
nB. Reducing the
number of amino acids in type B will allow more
molecules to be available for transmission.
•The rate of diffusion D∝n−1/3
B(see Stokes-Einstein
relation). Increasing the diffusivity of type B molecules
will yield a higher peak signal response ∝D.
Therefore, with multiple molecule types permitted, there is a
real freedom to dynamically adjust the chemical composition
to suit communication needs. Three scenarios exist for modi-
fying the size of the molecule in the RD link (assuming same
constituent amino acid sub-chains):
•In the case of nA=nB, there is no change in the size of
the molecule and no synthesis power cost φat the relay.
Due to the fact the some molecules will be lost in the SR
link, the performance in the RD link will be weaker like
the scenario without any relay assistance.
•In the case of nA> nB, the type B molecules are smaller,
higher concentration, and diffuse faster, resulting in a
stronger RD link performance than the aforementioned
nA=nBscenario. The cost of synthesis is potentially
high, as all the amino-acids from type A molecules may
need to be disassembled and re-synthesized into type B
molecules, incurring a potentially high φpower cost.
•In the case of nA< nB, the type B molecules are
larger, fewer (lower concentration), and diffuse slower,
resulting in a weaker RD link performance than the
aforementioned nA=nBand nA> nBscenarios.
However, the propagation loss in the SR link is smaller
than that in the RD link, due to DA> DB, which enable
higher energy harvested from the SR link.
The detailed research on how to optimize the SR and RD link
with respect to the number of amino acids per molecule type,
the relative distance, and the effect of which has on the rate
of diffusion and communication performance is left for future
research. Nonetheless, it is worth mentioning that due to the
long tail distribution of the diffusion channel response, there
is a complex trade-off between the energy harvesting potential
and the high inter-symbol-interference (ISI) in MCvD systems.
Certainly, existing research on optimizing resource allocation
6
in relays can be adjusted to MCvD SMIET systems, and this
is discussed in greater detail in Section IV.
IV. LEA RN IN G FROM EXISTING SWIPT TECHNIQUES
A. Resource Optimisation
In RF SWIPT, the relay node receives information and
harvests energy from the source’s transmissions. If the re-
laying node cannot perform information decoding and energy
harvesting at the same time due to the hardware limitations,
then it is worthwhile to consider current SWIPT research
into time-switching (TS) and power-splitting (PS) schemes.
Furthermore, due to the unreliable nature of energy harvesting
from the MCvD (see Fig. 5 for the GEV distribution of energy
arrival in Section III), scheduling becomes very important,
as the energy may not arrive with the same quantity or at
all in any given time frame. To achieve efficient scheduling,
modeling the arrival distribution either as a Markov model or a
probabilistic model is useful for the construction of algorithms
such as advance-before-scheduling. Considerations to the finite
level of energy or molecular storage in a nano-machine, and
the impact it has on SWIPT as well as processing power
expenditure needs to be considered [14]. Existing research in
relay based SWIPT have largely focused on the optimisation
of the tune parameters for TS and PS schemes as a function
of the channel state information (CSI). The availability of
CSI allows for either online or offline computation of optimal
scheduling through complex matrix manipulations. However,
CSI (even statistical) is notoriously difficult to estimate in
MCvD channels due to the unpredictable and rapid changes
in the fluid conditions, as well as the movement of the nano-
machines themselves.
In terms of algorithm complexity, solutions that involve
matrix manipulations and multiple directional water-filling
computes [14] are complex and not suitable in the context
of simple low-energy nano-machines. Therefore, non-coherent
and low-complexity signal processing and scheduling algo-
rithms need to be devised to adapt SWIPT optimally.
B. Crowd Energy Harvesting
Crowd energy harvesting is a popular concept proposed
for battery powered wireless nodes (e.g. low-power sensors)
to harvest energy from high power transmissions from TV
broadcasts and cellular base stations. Despite the growing
density of RF transmissions across multiple spectrum bands,
the amount of energy available to harvest is dominated by the
closest high power link. The rapid loss in RF energy due to
transmission distance limits the potential for crowd harvesting,
and unless all the transmitters are spaced equal distant to
the receiver, crowd harvesting energy from Ntransmitters is
not significantly superior to receiving energy from the nearest
transmitter. For MCvD systems, as mentioned in Section II,
the energy of molecules do not obey the propagation laws
of waves. Instead of experiencing a hostile ∝d−αrate of
energy decay, molecular numbers (or energy) decays ∝d−1.
Therefore, the potential to harvest energy from a field of
transmitters is far greater for MCvD than for RF commu-
nications. If one assumes that the molecular transmitters are
(b) Scatter and Box Plot of Energy Harvested for Node Densities
MCvD
RF
Centre of Cluster (PPP)
Energy Harvesting
Node
Transmitter
(Gaussian)
(a) Transmitters Distributed in a Modified Thomas PCP
Fig. 6. Crowd harvesting energy from a formation of nodes distributed
according to a Thomas PCP. Subplot (a) shows an instant snap-shot of the
PCP formation of nodes with an energy harvesting receiver at the centre.
Subplot (b) shows a scatter and box plot of the percentage of energy harvested
for MCvD and RF transmissions as a function of node density (per m2).
Modeling parameters: a variable modeling area radius 0.1-1km, 20 clusters
each with 10 nodes 2D Gaussian distributed with s.d. 30m, mass diffusivity
D= 79.5µm2/s, pathloss exponent α= 2, transmit power PTx = 1W,
transmit molecule NTx = 1, and receiver radius of R= 1 m.
randomly and uniformly distributed according to a Poisson
Point Process (PPP) or Poisson Cluster Process (PCP), one
can leverage on existing stochastic geometry techniques [15]
to find the expected molecular energy. This typically involves
understanding the general distance distribution fD(d, n)from
the energy harvesting node to the n-th nearest transmitter node.
Fig. 6 shows a simulation of crowd harvesting energy from
a formation of nodes distributed according to a modified
Thomas PCP. Subplot (a) shows an instant snap-shot of the
PCP formation of nodes with an energy harvesting receiver
at the centre. Subplot (b) shows a scatter and box plot of the
percentage of energy harvested for MCvD and RF transmis-
sions. The results show that RF energy harvesting is far more
sensitive to the density of transmitter nodes than MCvD energy
harvesting. The random walk nature of molecular propagation
means that the distance distribution (i.e., fD(d, n)) is not a
key consideration in crowd harvesting, whereas it is for RF
7
systems. This demonstrates the potential for crowd harvesting
in molecular systems, which can achieve 2-5dB improvement
in harvesting efficiency compared to RF systems in a similar
setting, with the highest relative gain at low node densities.
V. CONCLUSIONS AND FUTURE WORK
The performance of communication systems is fundamen-
tally limited by the loss of energy through propagation and
circuit inefficiencies. In this article, we have shown that it is
possible to achieve ultra low energy communications at the
nano-scale. We show that whilst the energy of waves will
inevitably decay as a function of transmission distance and
time, the energy in molecules does not. In fact, over time,
the molecular receiver has an opportunity to recover some, if
not all of the molecular energy transmitted. Inspired by the
GABA metabolism system, which fragments and reassembles
molecules, we design two nano-relay systems that can achieve
extremely high energy harvesting efficiency (12-25%) in real-
istic nano-scale conditions. We further examine the potential of
crowd harvesting energy from a swarm of nano-machines and
demonstrate that molecular communications is significantly
less sensitive to the spatial distribution of nano-machines and
can achieve 2-5dB improvement in harvesting efficiency.
In terms algorithms, challenges remain in applying exist-
ing radio-frequency energy harvesting solutions to molecular
communications, namely the lack of channel state information
and the low complexity requirement of any algorithms that
operate in nano-machines with limited computational ability.
What is more interesting is the need to build functional energy
harvesting systems that involves a reservoir of molecules
which can be used for fresh transmissions and replenished
from received and decoded molecular signals. As mentioned
in this article, inspired by biological processes, there is oppor-
tunity to chemically manipulate the chemicals in the reservoir
to potentially improve the energy efficiency of information
relaying, which is something radio frequency communications
can never do. Yet, how to build such biological functionalists
into realistic systems remains beyond current engineering
capabilities. Nonetheless, the preliminary results in this article
indicate that the potential for simultaneous molecular infor-
mation and energy transfer (SMIET) is immense and further
research is needed to transform theory to reality.
REFERENCES
[1] I. F. Akyildiz, M. Pierobon, S. Balasubramaniam, and Y. Koucheryavy,
“The Internet of Bio-Nano things,” IEEE Communications Magazine,
vol. 53, no. 3, Mar. 2015.
[2] S. Gowda, A. Reddy, X. Zhan, and P. Ajayan, “Building energy storage
device on a single nanowire,” ACS Nano Letters, vol. 11, no. 8, Jul.
2011.
[3] M. Donohoe, S. Balasubramaniam, B. Jennings, and J. M. Jornet,
“Powering In-Body Nanosensors With Ultrasounds,” IEEE Transactions
on Nanotechnology, vol. 15, no. 2, Mar. 2016.
[4] N. Farsad, H. B. Yilmaz, A. Eckford, C.-B. Chae, and W. Guo, “A
Comprehensive Survey of Recent Advancements in Molecular Commu-
nication,” IEEE Communications Surveys and Tutorials, 2016.
[5] T. Furubayashi, T. Nakano, A. Eckford, Y. Okaie, and T. Yomo,
“Packet fragmentation and reassembly in molecular communication,”
IEEE Transactions on BioNanoscience, vol. 21, Feb. 2016.
[6] M. S. Kuran, H. B. Yilmaz, T. Tugcu, and B. O. Edis, “Energy model for
communication via diffusion in nanonetworks,” Nano Communication
Networks, vol. 1, no. 2, 2010.
[7] J. Jornet and I. Akyildiz, “Channel Modeling and Capacity Analysis for
Electromagnetic Wireless Nanonetworks in the Terahertz Band,” IEEE
Transactions on Wireless Communications, vol. 10, no. 10, pp. 3211–
3221, Aug. 2011.
[8] G. Auer, V. Giannini, C. Desset, I. Godor, P. Skillermark, M. Olsson,
M. Imran, D. Sabella, M. Gonzalez, O. Blume, and A. Fehske, “How
much energy is needed to run a wireless network?” IEEE Communica-
tions Magazine, vol. 18, no. 5, pp. 40–49, 2011.
[9] H. B. Yilmaz, A. C. Heren, T. Tugcu, and C.-B. Chae, “Three-
Dimensional Channel Characteristics for Molecular Communications
With an Absorbing Receiver,” IEEE Communications Letters, vol. 18,
no. 6, 2014.
[10] A. Noel, K. C. Cheung, and R. Schober, “Improving Receiver Perfor-
mance of Diffusive Molecular Communication with Enzymes,” IEEE
Transactions on NanoBioscience, vol. 13, 2014.
[11] L. Felicetti, M. Femminella, and G. Reali, “Smart antennas for diffusion-
based molecular communications,” in ACM Nanocom, 2015.
[12] W. Guo, C. Mias, N. Farsad, and J. L. Wu, “Molecular versus Electro-
magnetic Wave Propagation Loss in Macro-Scale Environments,” IEEE
Transactions on Molecular Biological and Multiscale Communications
(T-MBMC), vol. 1, no. 3, 2015.
[13] C. Huang, R. Zhang, and S. Cui, “Throughput maximization for the
gaussian relay channel with energy harvesting constraints,” IEEE Jour-
nal on Selected Areas in Communications (JSAC), vol. 31, pp. 1469–
1479, 2013.
[14] O. Ozel, K. Shahzad, and S. Ulukus, “Optimal energy allocation for
energy harvesting transmitters with hybrid energy storage and processing
cost,” IEEE Transactions on Signal Processing, vol. 62, no. 12, 2014.
[15] S. Akbar, Y. Deng, A. Nallanathan, and M. Elkashlan, “Downlink and
Uplink Transmission in K-tier Heterogeneous Cellular Network with
Simultaneous Wireless Information and Power Transfer,” in IEEE Global
Communications Conference (GLOBECOM), 2015.
