2
Background and Related Work

2.1 Introduction

S ending information from one point to another is a well-established concept. The use of mobile phones and the internet have shown the importance of information transmission, and this is currently achieved via the utilisation of electromagnetic (EM) waves. Even though the achievements are numerous, there are shortcomings, for which alternative methods of communications can be utilised.

In nature the transmission of information can happen at the macro-scale, where communication can be achieved by using both chemical and EM , and at micro-scale, where communication is mostly done chemically. Commercial communication technologies rely on EM waves ranging from radio to optical bands. However, there are problems that conventional EM communications face (electromagnetic-interference (EMI), crosstalk etc.), where alternative approaches could be utilised, which do not suffer from the mentioned hindrances. An example of EM -harsh environments could be the use of EM waves in closed environments (e.g., tunnels, pipelines) or in aqueous environments, where the signal attenuation can be very high [5], [6], [7]. In water, EM waves do not travel as effectively as they would do in air. The signal experiences a high propagation delay and if the travelling medium has additional chemicals present (salt, minerals etc.), the conductivity of the medium is changed. This can add to the already high attenuation and increase the complexity of modelling the system [7]. There have been alternative techniques studied to overcome this problem. One of them being acoustic communications (AC): sending information using sound waves. The fundamental difference between AC and EM is the former uses pressure waves as the information carrier and the latter uses EM waves. AC is utilised in many ways in nature, such as in invertebrates and fish [8], [9]. It is primarily used in underwater communication where EM communications suffer from high diffraction loss; however, compared to EM , AC has low data rates and has problems such as multi-path propagation, time variations of the channel, small available bandwidth and high signal attenuation over long distances [10]. An alternative example would be the use of EM communication in nano-scale applications. At these scales, it is possible to produce nano-scale antennas for high-frequency communications. However, the complexity of the transceiver design, and the integration of the antennas to nano-machines, still remain substantial engineering problems. Even if the integration were possible, the output power of the transceiver would not be powerful enough to establish a bidirectional communication channel [11], [12]. In addition, the ratio of antenna size to the EM wave also poses a major problem [1], [13]. Optical communication would also not be efficient, since it either requires a guided medium (i.e., an optical fiber) or a line-of-sight (i.e., free-space optical communications [14]). The handicaps of using EM communication can be summed up as [6]:

Ratio of the antenna size to the wavelength
High absorption losses of EM waves
Propagation difficulties without using waveguides
Limitations caused by spectrum availability

Figure 2.1: Diagram of the scale of molecular communications. (A) At the smaller scales of communication DNA is used to store information to pass it on to the next generation. (B) In blood vessels hormones and other signal chemicals are transmitted to maintain a living organism. (C) Large creatures use pheromones to locate mating partners across long distances (D) Mass transport can also be seen on a global scale such as the Gulf stream.

In order to overcome these problems an alternative communication paradigm can be used. Instead of using EM waves to send information, chemicals signals can be employed. This type of transmission is called molecular communication [5]. In harsh environments, where the path loss is very high for EM waves, chemical signals may prove more reliable [6], [15]. In molecular communication, a transmitter releases information particles (e.g., proteins, molecules) into an aqueous or a gaseous medium, where the particles can propagate using either passive (diffusion) or active transport (advection). The receiver then decodes the information particles based on predefined communication protocols [16].

In nature, the use of molecular communication can be seen in various events. At ranges of micro-scale (nm to $μ$ m), chemical signals are used for inter-cellular communications [17]. Whereas in macro-scale (cm to m), pheromones are used for long-range communications between the same species [18]. However, in nature the transmission of pheromones is normally used to send limited determined messages. The engineering prospect of sending reliable data by using molecules is the main challenge of this communication paradigm [19]. In Table ( 2.1 ), page li , the three types of communication mentioned in the introduction are compared. A scale which molecular communication can be achieved can be seen in Figure 2.1 . Since molecular communication relies on particles (i.e., molecules, pheromones) to transmit information, there are scales where this is already achieved; from small scales such as DNA replication (i.e., semiconservative replication) to grand scales such as migratory patterns [20].

2.1.1 Research History

Figure 2.2: The significant points of molecular communications development.

While the concept of chemicals as a means of communication in the biological domain existed in some form for over a billion years, the use of molecular communication in engineering terms was first described by T. Nakano, et. al in 2005 [21]. This work proposed a method for nanomachines to communicate using signalling network (i.e., calcium signalling) in an aqueous environment and designed a simple network with an encoder and a decoder. As it mostly focused on a biological environment, molecular communication evolved and improved with the concept of biocompability in mind. Following this work, different aspects of molecular communications were studied. One of the first aspects studied were the channel itself. In a biological environment, the channel is generally an aqueous environment. A. W. Eckford in 2007 presented a channel capacity analysis with Brownian motion [19]. While this study was done in an idealised environment, following works started to fill in the spaces.

These works mentioned previously have all been done under micro-scale, where the communication range happens from nm to $μ$ m. However, as there are animals that can communicate in distance up to kms, possibility exist where communication can be established using chemicals in engineering terms. One of the first papers to create a definition of macro-scale molecular communication was presented by I. F. Akyildiz et. al in 2008 [12]. While this paper focused mostly on nano communications a substantial amount of work was describing a system of pheromonal molecular communication.

Along with the channel analysis, modulation analysis was carried out for micro-scale molecular communication by M. U. Mahfuz et. al. in 2010 [15], where on-off keying was introduced and studied. As more and more aspects of molecular communications were being scrutinised, simulation analysis was starting to appear with a study conducted by E. Gul et. al. being one of the firsts [22]. The study implemented an already established network simulator NS-2 for use in micro-scale molecular communications. Error correction analysis soon followed with a study implementing already established principles for molecular communications with a work published by M. S. Leeson et. al. in 2012 [23].

Aforementioned works have all been theoretical. As micro-scale would entail the implementation of sensor and detector relative to the size of a red-blood cell, work has been carried out theoretically. However, macro-scale work in human scale (cm-km) and therefore sensors and detectors already present can be used to design an experimental platform to study the communication paradigm. The first "proof-of-concept" experimental study of molecular communication in macro-scale was carried out by N. Farsad et. al. in 2013. The proposed system uses a modified spray and an off-the-counter MQ3 spray to initiate a transmission. Following these culmination of the previous work a comprehensive survey was published by N. Farsad in 2016 [16]. This body of work presented an overview of the work carried on molecular communication with more emphasis on the micro-scale communication with respect to theory and applications. All the experimental research up-to now has been focused on a single input single output system and in 2016 the first experimental study in multi-input multi output system is conducted [24]. It has shown that molecular communication performance can be expanded via the use of MIMO. As experimental approach to study macro-scale gained tracktion, different receivers were tested. One being mass spectrometer where in 2017 it was used to test long range molecular communication by transmitting different chemicals as information concurrently [25]. Mass spectrometers, unlike a MQ3 sensors, are capable of distinguishing numerous chemicals concurrently and shown to be viable receivers for use in macro-scale applications. As more experiments starts to surface, the need to standardise the procedures also becomes a problem which was addressed in [26]. In this work a procedure was described to increase readability across many scientists by creating a standard for experimentation.

These receivers mentioned shown to be viable for use in large scales but one of the big push in studying molecular communication is its bio-compability and its possible use in in-vivo applications. A study conducted in 2018 shown a system where the bacteria is used in molecular communication[ 27 ]. By giving specific light pulses the chemical concentration generated by the bacterica can be changed and this can be received by a pH sensor. While this system is a proof-of-concept, it has shown the possibility for using organisms as a means of information transfer. A recent study was done in 2019 to analyse both experimentally and theoretically the modulation properties of macro-scale molecular communications [4].

2.1.2 Micro-Scale Molecular Communication

The use of molecules to transmit information on the micro-scale has been employed, being perfected in each iteration, by evolutionary processes such as protein creation and DNA replication [28] in a cell. In addition, being bio-compatible makes it a perfect choice for use in medical applications, and if transmission is solely by diffusion, the system does not require any energy from an external source. The transmission can also be made using active methods such as advection-diffusion (i.e., convection) [29], which will be discussed in section 2.5.2 in detail. A diagram of the biological application, DNA , can be seen in Figure 2.3 and a paper published in 2018 has developed a protocol where information is encoded with DNA strands for use in nanomachines [30].

Figure 2.3: Deoxyribonucleic acid ( DNA ) is used in biological systems to store genetic information. The prospect of storing and accessing information on a small and dense scale shows the possibility of implementing information transmission and storage on small scales [ 31 ]. An engineering approach to use DNA in molecular communication is to use it as a protocol between nanomachines [ 30 ].

2.1.3 Macro-Scale Molecular Communication

The possibility of molecular communications at the macro-scale was first described in [32], where macro-scale is defined as a communication system with a transmission distance of a “ few cm and more ” [32]. Similar to the micro-scale, nature has been utilising the macro-scale for a long time. The application of pheromones [18] can be given as an example and a representation of this application can be seen in Figure 2.4 . In addition, the use of chemicals to deter predators, to define locations can be given as further examples. Even though in principle the mechanics of the macro-scale are similar to the micro-scale, the problems that have to be overcome are different.

Figure 2.4: Molecular communication can also be seen on the macro-scale. In moths, calling females attract their mates late at night with intermittent release of a species-specific sex-pheromone blend [ 33 ]. [Photo: Male Gypsy Moth, Credit: Ilona L/Flickr]

2.2 Review Structure

The structure of the review is as follows. In Section 2.1 , the topic is introduced. The historical development of molecular communication is presented in Section 2.2. Section 2.3 focuses on the channel theory of molecular communications. Propagation is further analysed based on how it is accomplished: diffusion or advection-diffusion in Section 2.5 . In Section 2.6 , types of modulation that were developed for the field are discussed along with the inter-symbol interference (ISI) and error-correction methods in Section 2.7 and Section 2.8 , respectively. Section 2.10 discusses the reception process in molecular communications based on the scale of the communication. In Section 2.11 , various experiments done in molecular communication are reviewed. This is then followed by the application of this new communication paradigm in Section 2.12 . Section 2.13 is the review of numerous simulation platforms used in the study of molecular communication and Section 2.14 focuses on the standardisation effort of the communication method. The review concludes with Section 2.15 .

Table 2.1: Comparison of communication systems.

Property	Electromagnetic Communication	Macro Molecular Communication	Micro Molecular Communication
Communication type	Wave	Particle
Transmission distance	mm - km	cm - m [32]	nm - $μ$ m [5]
Transmission medium	Air ( EM Radiation) and Cable	Liquid and Gas
Transmission speed	up to 3 $\cdot 1 0^{8}$ m/s (speed of light)	Passive Flow (Diffusion): $\leq$ 1 $c m^{2} ∕ s$ Active flow (Advection): $≪$ 3 $\cdot 1 0^{8}$ m/s
Propagation	Line-of-Sight	Diffusion [34] and Advection [35]
Methods	Ground Wave, Sky Wave	Convection, Turbulent Flow [36] [29]	Gap Junctions [37], Molecular Motors [38]
Transmitter	Antennae ( $T X$ / $R X$ )	Sprays ( $T X$ ) [39] & Arduino Uno ( $T X$ / $R X$ )	Genetically modified cells ¹ ( $T X$ / $R X$ ) [40], [41], [42]
&	Laser Diodes ( $T X$ )	Electronic Noses ¹ ( $R X$ ) [43] & pH Sensors ( $R X$ )	Artificial Cells ¹ ( $T X$ / $R X$ ) [44], [45]
Receiver	Photodetector ( $R X$ )	Mass Spectrometers (MS) ( $R X$ ) [2], [25], [46], [47]	Synthesizing Receptors ¹ ( $R X$ )[ 48 ], [49]
		Biomimicry Sensors ¹ ( $R X$ ) [50], [51]	Novel Materials ¹ ( $T X$ / $R X$ ) [52]
		Peristaltic Pump ( $T X$ ) & Susceptometer coil ( $R X$ )
		Odour Generator ( $T X$ )[ 2 ], [25], [46], [47]
Information element	Electrical Signals, EM Waves + Light	Odours [2], [25], [46], [47] , Scent [43]	DNA , Proteins [53] [54]
Power Source	Electrical (external)	Chemical, Thermal (internal/external) & Flow (external)
Energy Consumption	High	Low [55]
Used in Nature	Platypus [51]	Hormone (TSH) [56]	Cell-to-Cell communication [57]
	Eel [51]	Plant-to-Plant communication [58]	Intra-cell communication [59]
		Pheromone (Fish, Ants, etc.) [18], [60], [61]
Predictability	Accurate, Linear	Stochastic, non-linear
Applications	Cellular communication	Infastructure Monitoring [62]	Medicinal [52], [63]
	Radio	Underground Communication [64]	Targeted drug delivery [65]
	Television	Transmission through pipes [6]	Nanorobots [66]
	Internet	Studying animal behaviour [18]	Disease diagnosis and treatement [63], [67]
	Computer	Odour tracking robots [68], [69], [70], [71], [72], [73], [74], [75], [76]	Nanorobot communication[ 77 ], [78]
		Pheromone based communication [50], [79], [80], [81]

¹ Theorized

2.3 Channel Theory

Information theory provides an invaluable insight into the workings of a communication system [82] and this also holds true with molecular communication, where the application of information theory is relatively new and has recently become a research interest [83]. One of the most important aspects of this field is the application of channel modelling. A generalised block diagram that represents a channel can be seen in Figure 2.5 .

Figure 2.5: Diagram of a generalised communication system.

In a physical sense, a communication channel can be described as the environment in which the signal travels (i.e., air, space, water, etc.).

In a mathematical sense, the channel is a function that bridges the information definition between input and output. In theory, the transmitted signal ( $X$ ) and the received signal ( $Y$ ) can be the same. This would be called a noiseless channel. In a practical sense though, there are uncertainties that arise from the environmental noise, interferences from other devices (e.g., EMI ) or worse, interference from itself (i.e., ISI ). In molecular communication, the noise can be caused by the detector, diffusion of the messenger chemicals, particle collisions etc. This can cause a problem when estimating the channel for a reliable communication [84]. As any communication needs an environment to send information, no transmission method is immune from disturbances [85] and these limitations impose a tight upper bound on the amount of information that can be reliably transmitted. This is also known as the channel capacity ( $C$ ). This limitation is shown in Shannon’s channel capacity of a communication system with additive white Gaussian noise (AWGN) [85]:

C = \log_{2} (1 + \frac{S}{N_{0}}),

(2.1)

where $S$ is the signal power ( $W$ ) and the $N_{0}$ is the power of the noise ( $W$ ). However, in molecular communication, there are ways in which the effect of noise can be diminished as some propagation methods produce less variance (e.g., advection) compared to others (e.g., diffusion).

In a noisy system, the channel capacity plays a prominent role in modelling the communication [82]. For a discrete memory-less channel, where the received signal is independent of all the previous transmitted symbols, the achievable mutual information rate can be calculated from $P_{X Y} (x, y)$ , $P_{X} (x)$ and $P_{Y} (y)$ in Eq. ( 2.2 ) [83], [85] :

I (X; Y) = \sum_{x, y} P_{X Y} (x, y) \log [\frac{P_{X Y} (x, y)}{P_{X} (x) P_{Y} (y)}],

(2.2)

where $I (X; Y)$ is the mutual information, $P_{X Y} (x, y)$ is the joint probability distribution and $X$ and $Y$ are two discrete variables. If the transmission possesses no memory effect, the capacity of the channel can be written as [82], [85] :

C ≜ \max_{p (x)} I (X; Y) .

(2.3)

However, if memory begins to play an effective role, the mathematical expression for the channel capacity of a block size of $n$ with memory is given as [82], [86] :

C ≜ \lim_{x \to \infty} s u p \frac{1}{n} I (X^{n}; Y^{n}) .

(2.4)

For a discrete memory-less channel the channel capacity can be calculated using special algorithms if $p$ ( $y | x$ ) is known beforehand. [82], [87], [88]. However, as will be mentioned in the ISI section, molecular communication shows memory effects, and this changes the calculation method considerably [89].

The channel characteristics of molecular communication have been studied for various properties. These include channel modelling [90], [91], [92], [93], [94], [95], [96], channel capacity [97], [98], [99] and noise [100], [101].

Table 2.2: Speed comparison examples between micro-scale and macro-scale molecular communication methods.

Macro-Scale Molecular Communication

Micro-Scale Molecular Communication


Example	Chemical	Speed	Reference

Morphogen	EGF	$1 \cdot 1 0^{- 7}$ $c m^{2} ∕ s$	[102]

Airstream	Bombykol	0.6 m/s	[103]

Hormonal Signalling	TSH	5 cm/s	[56]

Neural Signalling	Active Potential	10-100 m/s	[17]
Southern Ocean	Particles	$1 \cdot 1 0^{4}$ $m^{2} ∕ s$	[104]
Cyclone Olivia	Particles	408 km/h	[105]
The Great Red Spot	Particles	618 km/h	[106]

Example	Chemical	Speed	Reference

Vesicular trafficking	Proteins	1 $μ m ∕ s$	[53], [107], [108]

Bacterial Migration	DNA	10-20 $μ m ∕ s$	[54], [109]

Calcium Signalling	$C a^{+ 2}$ Wave	10-30 $μ m ∕ s$	[110], [111]
Chemotactic Signalling	Protein	1-10 $μ m^{2} ∕ s$	[112]
Membrane Diffusion	DPPE (lipid)	11.58 $m^{2} ∕ s$	[113]
Molecular Motors	Kinesin	6400 Å/s	[114]

2.4 Propagation Speed

The speed with which the symbols are transmitted from one point to another is an important factor in any communication system. In a traditional EM -based communication the speed at which a signal can propagate can reach up to the speed of light ( $c_{l}$ = 299,792,458 m/s) in a perfect vacuum.

The speed of an EM wave is governed by the refraction index of the medium. For example, an EM wave travelling in air has a refractive index of 1.0003 so the speed of propagation is $\sim$ 100 %; however, in an aqueous environment such as water the index is 1.33, slowing the speed of propagation to 75 % of the speed of light. Communication can also be made with conductive materials such as copper or optical fibre, in which the speed can range from 50 % up to 99 % of $c_{l}$ .

Molecular communication, however, is based on particles instead on waves and therefore, due to the principals of special relativity, a transmission speed of $c_{l}$ is not possible. To transmit particles, the speed can be generated by two methods of propagation. It can be generated from internal forces (diffusion) or from external forces (advection). A combination of both these forces can also be utilised in propagation (advection + diffusion). These methods are discussed in detail in the following section.

2.5 Methods of Propagation

2.5.1 Diffusion

Also known as Brownian motion, diffusion is a process of random particle motion caused by collisions of other fast-moving atoms or particles in a gaseous or aqueous medium [34], [115]. Even though this process is random by nature, information can be transmitted from one point to another solely by diffusion. The main advantage of this technique is that the energy required for propagation is sourced from the thermal energy of the environment, and can be tapped into without the need for any external energy source. The application of this method can be observed in numerous biological processes. An example would be the exchange of $O_{2}$ and $C O_{2}$ in the lungs. While inhaling, the alveoli are flooded with $O_{2}$ -rich air and are only separated from the $C O_{2}$ -rich bloodstream by a 1-2 $μ$ m of membrane. The close proximity of the two different concentrations causes the diffusion process. There are other examples of diffusion in a biological system, such as DNA replication [116], protein production [117], etc.

Diffusion is a characteristic property of a chemical in a specific environment. As the temperature increases, the diffusivity also increases, and in an aqueous environment the diffusion propagation is also slower compared to the gaseous environment. The diffusivity coefficient in water can be estimated using the Stokes-Einstein relation [118] :

D = \frac{k_{B} T_{E}}{6 π η r},

(2.5)

where $k_{B}$ is Boltzmann’s constant ( $J ∕ K$ ), $T_{E}$ is the temperature (K), $η$ is the dynamic viscosity ( $N \cdot s$ / $m^{2}$ ) and $r$ is the radius of the sphere ( $m$ ). For the gas phase the diffusion coefficient can be expressed as [119] :

D = \frac{2}{3} \sqrt{\frac{k_{B}^{3}}{π^{3} m}} \frac{T_{E}^{3 ∕ 2}}{P_{E} d^{2}},

(2.6)

where $m$ is the mass of the gas (kg), $d$ is the diameter of the gas molecule (m) and $P_{E}$ is the pressure (Pa). For two different gases (A and B) their diffusivity relative to each other can be expressed as :

D_{A B} = \frac{2}{3} \sqrt{\frac{k_{B}^{3}}{π^{3}}} \sqrt{\frac{1}{2 m_{A}} + \frac{1}{2 m_{B}}} \frac{4 T_{E}^{\frac{3}{2}}}{P_{E} {(d_{A} + d_{B})}^{2}},

(2.7)

where $m_{A}$ and $m_{B}$ are the molecular masses of gas A and B ( $k g$ ), $d_{A}$ and $d_{B}$ are the molecular diameters of gas A and B ( $m$ ), respectively. However, defining diffusion as one thing that can be applied to all circumstances may not be correct. There are other types of diffusion that explain this propagation in different circumstances, which are governed by different mathematical concepts and models. A major aspect of diffusion is the environment it is initiated. For example, diffusion happening through a biological membrane behaves differently to diffusion happening in a turbulent environment. These affect the mathematics used to model the diffusion. Examples of different types of diffusion can be seen below.

Anomalous Diffusion [120]: Diffusive behaviour with a non-linear relation to time. Can be seen in protein diffusion in cells and porous media
Eddy Diffusion (i.e., Turbulent Diffusion) [121]: Diffusive behaviour caused by eddy motion. Seen in turbulent environments.
Facilitated Diffusion [122]: Diffusion caused by spontaneous passive transport. Can be seen in the transportation of molecules across a biological membrane
Knudsen Diffusion [123]: Diffusion occurring in porous media where the pore length is comparable or smaller than the mean free path. This effect is observed in porous membrane environments.

The speed range of molecular communication can range from $μ m ∕ s$ to m/s. In addition, since propagation can involve flows, the speed can theoretically be increased further. A comparison of various speeds of molecular communication in nature can be seen in Table 2.2 .

2.5.1.1 Stochastic Approach

Diffusion is based on random movements and therefore can be modelled based on its probabilistic behaviour. There have been studies to model this effect, such as using Markov Chains [124], [125] and Monte Carlo [126]. The diffusion process can be modelled mathematically using the following model [16], [127]:

[\begin{matrix} Δ x_{i} \\ Δ y_{i} \\ Δ z_{i} \end{matrix}] = [\begin{matrix} x_{i - 1} \\ y_{i - 1} \\ z_{i - 1} \end{matrix}] + Δ r \cdot [\begin{matrix} \cos 𝜃_{i} \sin ϕ_{i} \\ \sin 𝜃_{i} \sin ϕ_{i} \\ \cos ϕ_{i} \end{matrix}],

(2.8)

where $Δ r$ is the displacement ( $m$ ) , both $𝜃_{i}$ and $ϕ_{i}$ represent the 3D angular position and $x, y, z$ represent the spatial coordinates ( $m$ ). However, there are other methods in which diffusion is modelled in the literature.

One way of approaching the problem of random propagation is to model the process as a stochastic continuous-time problem. This process, known in literature as the Wiener process, has been used in modelling Brownian motion and, as a reflection, has been utilised in modelling the propagation for use in molecular communication [101]. Let $ς_{1}, ς_{2}, \dots$ be independent and identically distributed random variables with an expected value of 0 and a variance of 1. For each $n$ , the continuous time stochastic process can be defined as :

W_{n} (t) = \frac{1}{\sqrt{n}} \sum_{1 \leq k \leq ⌊ n t ⌋} ς_{k}, t \in [0, 1],

(2.9)

The above equation represents a random step function with each increment of $W_{n}$ being independent of one another. For large values of $n$ the stochastic process approaches a normal distribution of $N (0, t)$ due to the central limit theorem.

Another approach is to apply the Langevin equation [128], which describes the time evolution of a subset of the degrees of freedom. Based on the equation, the position ( $p_{υ}^{n} (t)$ ) of a particle $n$ with a mass of $m$ at time $t$ along any of the dimensions ( $υ$ ) obeys the following interpretation of the equation :

m \frac{\partial^{2} p_{υ}^{n} (t)}{\partial t^{2}} = - 6 π η r \frac{\partial p_{υ}^{n} (t)}{\partial t} + f_{N} (t),

(2.10)

where $f_{N} (t)$ is the noise present in the environment (i.e., particles present in the environment). Stochastic approaches of diffusion in molecular communication can be seen in many studies; such as modelling the complete system of molecular communications [129], noise analysis in ligand-binding [130] and channel capacity in a fluid medium [131]. Finally, point process theory was also studied for use in molecular communications [132].

2.5.1.2 Analytical Approach

The second approach to modelling the propagation is to model the system as a partial differential equation (PDE). The diffusion-only equation, also known as Fick’s $2^{n d}$ law, which is described in [133] can be expressed as;

\begin{align} \frac{\partial c}{\partial t} & = D (\frac{\partial^{2} c}{\partial x^{2}}), & (2.11a) \\ \frac{\partial c}{\partial t} & = D (\frac{\partial^{2} c}{\partial x^{2}} + \frac{\partial^{2} c}{\partial y^{2}}), & (2.11b) \\ \frac{\partial c}{\partial t} & = D (\frac{\partial^{2} c}{\partial x^{2}} + \frac{\partial^{2} c}{\partial y^{2}} + \frac{\partial^{2} c}{\partial z^{2}}), & (2.11c) \end{align}

where $D$ is the diffusion coefficient ( $c m^{2} ∕ s$ ) and $c$ is the concentration at a given point in space ( $k g ∕ m^{3}$ ). Therefore, the concentration of the molecules is dependent on the spatial coordinates as well as the time.

The PDE given in Eq. ( 2.11 ) can have numerous solutions, based on the boundaries of the system. One of the minimalist approaches of solving Eq. ( 2.11 ) is to solve it at the point of release of the chemicals ( $t_{0}$ ). If $M_{0}$ is the initial number of molecules, or mass, released from the transmitter at the time $t_{0}$ then the initial conditions for the molecular concentration $c (x, t)$ for 1D, also known as the “ thin film solution” in the literature, diffusion process can be expressed as:

\begin{align} c (| x | > 0, t_{0}) & = 0, & (2.12a) \\ c (x = 0, t_{0}) & = M_{0} δ (x), & (2.12b) \\ c (| x | \to \infty, t) & = 0 . & (2.12c) \end{align}

In these equations $δ (x)$ represents the continuous Dirac delta function for a given spatial dimension ( $x, y, z$ ) and is defined as [134]:

\int_{- \infty}^{+ \infty} δ (x) d x = 1 .

(2.13)

By implementing the boundary conditions given in Eq. ( 2.12 ) to Eq. ( 2.11 ), the PDE can be solved and the solutions for the each dimension are:

\begin{align} c (x, t) & = \frac{M_{0}}{\sqrt{4 π D t}} \exp (- \frac{x^{2}}{4 D t}), & (2.14a) \\ c (x, y, t) & = \frac{M_{0}}{\sqrt{{(4 π D t)}^{2}}} \exp (- \frac{x^{2} + y^{2}}{4 D t}), & (2.14b) \\ c (x, y, z, t) & = \frac{M_{0}}{\sqrt{{(4 π D t)}^{3}}} \exp (- \frac{x^{2} + y^{2} + z^{2}}{4 D t}) . & (2.14c) \end{align}

The derivation of this solution can be seen in Appendix 8.2.5 . There are other ways the diffusion equation can be solved with different initial conditions [117]. The channel response can be obtained by changing the conditions of the boundary if there is an absorbing receiver [135]. The capture function of the system can be derived by the integration of the channel response from the differential equation with a boundary [136]. By calculating the impulse response differentiation with respect to time, it is possible to calculate the peak concentration at the peak time [136], [137].

2.5.2 Advection

Advection is defined as the transportation of material or heat by using the flow of another fluid [35]. In molecular communications, introducing a velocity element to the propagation can be mathematically described with two approaches. The former being a stochastic approach and latter being an analytical approach.

2.5.2.1 Stochastic Approach

An approach to modelling the diffusion with an advection element is to introduce a velocity vector to random-walk simulations. This process can be modelled by using the Monte Carlo method [126] by introducing a velocity component ( $u$ ) to the diffusion process presented in Eq. ( 2.15 ) [16]:

[\begin{matrix} Δ x_{i} \\ Δ y_{i} \\ Δ z_{i} \end{matrix}] = [\begin{matrix} u_{x, i - 1} \\ u_{y, i - 1} \\ u_{z, i - 1} \end{matrix}] \cdot [\begin{matrix} x_{i - 1} y_{i - 1} z_{i - 1} \end{matrix}] + Δ r \cdot [\begin{matrix} \cos 𝜃_{i} \sin ϕ_{i} \\ \sin 𝜃_{i} \sin ϕ_{i} \\ \cos ϕ_{i} \end{matrix}] .

(2.15)

As mentioned in the diffusion stochastic process, stochastic differential equations (SDEs) can also be implemented to model the advective element. A different approach to modelling is to use the Fokker-Planck (aka. Smoluchowski) equation [138]. This equation describes the time evolution of the probability density function of the velocity of a particle under the influence of drag forces ( $μ_{D} (x, t)$ ) (i.e., drift) and random forces ( $D_{V} (x, t)$ ) (i.e., diffusion) in space and time. The mathematical equation for the Fokker-Planck is given in Eq. ( 2.16 ) :

\frac{\partial p}{\partial t} (x, t) = - μ_{D} (x, t) \frac{\partial p}{\partial x} (x, t) + D_{V} (x, t) \frac{\partial^{2} p}{\partial x^{2}} (x, t) .

(2.16)

To solve this ordinary partial differential equation (OPDE), different types of boundaries have been proposed that can produce different solutions: such as infinite environment [139], [140], infinite source [23], [141] or long-term capture [142]. This approach was used in [142] to analyse macro-scale communications and the aspect of the sensor cleaning from a theoretical point of view.

2.5.2.2 Analytical Approach

In this approach the system is modelled deterministically. However, to model the propagation the advection term must be mathematically defined. In this definition of propagation, some aspects of the transported substances are conserved, such as energy. An example would be the pollutants in a river or the gas flow through pipelines. The mathematical expression of advective flow can be seen as [143] :

u \cdot \nabla = u_{x} \frac{\partial}{\partial x} + u_{y} \frac{\partial}{\partial y} + u_{z} \frac{\partial}{\partial z},

(2.17)

where u = ( $u_{x}, u_{y}, u_{z}$ ) is the velocity field and the $\nabla$ is the del operator. The advection equation for a conserved quantity for a scalar field of $ψ$ is expressed by the continuity equation [144]:

\frac{\partial ψ}{\partial t} + \nabla \cdot (ψ u) = 0 .

(2.18)

With the advection defined, the final approach would be to reintroduce diffusion to this propagation. Convection is the sum action of both diffusion and advection and mathematically this phenomenon is modelled using the advection-diffusion equation. This equation can be derived from the continuity equation given below:

\frac{\partial c}{\partial t} + \nabla \cdot \vec{j} = R,

(2.19)

where $\vec{j}$ is the total flux and the $R$ is the source or the sink for the function $c$ . In this system, there are two flux sources. The former being the diffusive flux from the diffusion of particles and the latter being the advective flux due to the motion created by the flow:

\begin{matrix} \begin{aligned} {\vec{j}}_{D} & = - D \nabla c, \\ {\vec{j}}_{A} & = u c . \end{aligned} \end{matrix}

(2.20)

By inserting these equations into the continuity equation the advection-diffusion equation can be obtained. In the literature this is also known as the convection-diffusion equation and drift-diffusion equation [145], which is expressed as [146]:

\frac{\partial c}{\partial t} = \underset{C o n v e c t i o n}{\underset{⏟}{\overset{D i f f u s i o n}{\overset{︷}{\nabla \cdot (D \nabla c)}} - \overset{A d v e c t i o n}{\overset{︷}{\nabla \cdot (u c)}}}} + R,

(2.21)

where;

$c$ is the concentration of mass transfer ( $k g ∕ m^{3}$ ).
$D$ is the diffusivity coefficient (i.e., mass diffusivity for particle motion) ( $m^{2} ∕ s$ ).
u is the velocity field that the quantity is moving with ( $m ∕ s$ ). For example, if $c$ is the salt concentration in a river, u is the velocity of the water flow.
$R$ is the “ sink/source” of the quantity $c$ . In chemical processes, $R > 0$ means the system is creating more chemicals (i.e., sources) and $R < 0$ means the system is destroying more chemicals (i.e., a sink).

An approach to solve the equation is to change the frame of reference from stationary coordinates ( $x, t$ ) to moving coordinates ( $λ_{x}, t$ ). This reference frame can be mathematically expressed as [147]:

λ_{x} = (x - u_{x} t) .

(2.22)

If this frame is inserted into Eq. ( 2.21 )

\begin{align} \frac{\partial c}{\partial t} (λ_{x}, t) & = \frac{\partial c}{\partial λ_{x}} \frac{\partial λ_{x}}{\partial t} + \frac{\partial c}{\partial t} \frac{\partial t}{\partial t} = - u_{x} \frac{\partial c}{\partial λ_{x}} + \frac{\partial c}{\partial_{x} t}, & (2.23a) \\ \frac{\partial c}{\partial x} (λ_{x}, t) & = \frac{\partial c}{\partial λ_{x}} \frac{\partial λ_{x}}{\partial t} + \frac{\partial c}{\partial t} \frac{\partial t}{\partial x} = \frac{\partial c}{\partial λ_{x}}, & (2.23b) \\ \frac{\partial^{2} c}{\partial x^{2}} (λ_{x}, t) & = \frac{\partial^{2} c}{\partial λ_{x}^{2}} . & (2.23c) \end{align}

By substituting from Eq. ( 2.23 ) for $\partial c ∕ \partial t$ , $\partial c ∕ \partial x$ and $\partial^{2} c ∕ \partial x^{2}$ to Eq ( 2.21 ) becomes :

\frac{\partial c}{\partial t} = D \frac{\partial^{2} c}{\partial λ_{x}^{2}} .

(2.24)

As can be observed Eq. ( 2.24 ) this equation is Fick’s $2^{n d}$ law, with the only difference being having a moving reference frame $λ_{x}$ . By using the same boundary conditions described in Eq. ( 2.12 ), the solution for the equations becomes:

c (x, t) = \frac{M_{0}}{\sqrt{4 π D t}} \exp (- \frac{λ_{x}^{2}}{4 D t}) .

(2.25)

Converting back to the stationary frame of reference, the final equation is obtained and can be seen in Eq. ( 2.26a ). The 2D and 3D solutions can be seen in Eq. ( 2.26b ) and ( 2.26c ), respectively.

\begin{align} c (x, t) & = \frac{M_{0}}{\sqrt{4 π D t}} \exp (- \frac{{(x - u_{x} t)}^{2}}{4 D t}), & (2.26a) \\ c (x, y, t) & = \frac{M_{0}}{\sqrt{{(4 π D t)}^{2}}} \exp (- \frac{{(x - u_{x} t)}^{2}}{4 D t}) \exp (- \frac{{(y - u_{y} t)}^{2}}{4 D t}), & (2.26b) \\ \begin{aligned} c (x, y, z, t) & = \frac{M_{0}}{\sqrt{{(4 π D t)}^{3}}} \exp (- \frac{{(x - u_{x} t)}^{2}}{4 D t}) \\ \times \exp (- \frac{{(y - u_{y} t)}^{2}}{4 D t}) \exp (- \frac{{(z - u_{z} t)}^{2}}{4 D t}) . \end{aligned} & (2.26c) \end{align}

Eq. ( 2.11c ) represents the diffusion of a chemical in a 3D unbounded space. However, on a macro-scale, the propagation of the chemicals can be made in a bounded domain. This has benefits over the boundary-less methods. The most important being that by using a closed system, more chemicals will arrive at the receiver and make it possible to increase the distance of communication.

For using a bounded communication a cylindrical shape has a few advantages. The most important one being that because it does not have any corners to act as stress concentrators, and it helps the fluid flow to be smoother compared to other geometric shapes. The cylindrical coordinate representation of the advection-diffusion equation (ADE) can be seen in Eq. ( 2.27 ) [148]:

\frac{\partial c}{\partial t} = D_{L} \frac{\partial^{2} c}{\partial z^{2}} - u \frac{\partial c}{\partial z} + \frac{D_{R}}{r} (r \frac{\partial c}{\partial r}),

(2.27)

where $r$ , $𝜃$ and $z$ are the cylindrical coordinates, $D_{L}$ is the longitudinal diffusivity ( $c m^{2} ∕ s$ ) and $D_{R}$ is the radial diffusivity ( $c m^{2} ∕ s$ ). $𝜃$ is omitted since it is assumed that the system has angular symmetry.

The solution for this kind of equation with initial and boundary conditions can be obtained by implementing the Hankel [149] and Laplace Transforms [148]. One such solution is shown below [148]:

\begin{matrix} c (r, z, t) = \frac{M_{0}}{ϕ ρ_{r}^{2} \sqrt{D_{L} π^{3} t}} \sum_{n = 0}^{\infty} \frac{1}{λ_{n}} {\exp [- \frac{{(u t - z)}^{2}}{4 D_{L} t} - D_{R} λ_{n}^{2} t] \\ - \frac{u}{2 D_{L}} \exp [- \frac{u z}{D_{L}} - D_{L} λ_{n}^{2} t] e r f c (\frac{z + u t}{\sqrt{4 D_{L} t}})} (\frac{ρ_{r}^{2}}{r^{2}} + \frac{2 ρ_{r} J_{1} (λ_{n} ρ_{r}) J_{0} (λ_{n} r)}{r^{2} | J_{0} (λ_{n} r) |^{2}}), & (2.28) \end{matrix}

where $r$ is the radius of the cylindrical system (m), $ρ_{r}$ is the radius of the inner injected zone, $ϕ$ is the porosity of the material and $λ_{n}$ is the Hankel transform parameter determined by the following transcendental equation:

\frac{J_{0} (λ_{n} r)}{d r} = 0 .

(2.29)

As can be seen, defining boundaries unto the propagation greatly increases the complexity of the analytical solution, relying on different methods for solving the PDE, such as separation of variables [150] or change of variables [151]. Studies in boundary medium transmission in molecular communication has seen interest both experimentally [25], [47] and theoretically [152].

2.5.3 Turbulent Flows

Turbulent flows are caused by the random fluctuations in a flow caused by the chaotic motions of particles [36]. These movements of particles in the flow are called eddy diffusion, because of their similarity to molecular diffusion. In molecular diffusion, the random motion of molecules themselves from the thermal energy from the environment cause the diffusion, whereas in turbulent diffusion it is the motion of the fluid that causes the propagation [153], [154], [155].

Turbulence is heavily dependent on the bulk flow, as the turbulence only happens if the flow is above a critical number. This critical number is defined as Reynolds number and is calculated by the ratio of the inertial forces to the viscous forces of the fluid in question [156] :

R e = \frac{ρ u L_{D}}{η},

(2.30)

where $ρ$ is the density of the fluid ( $k g ∕ m^{3}$ ), $u$ is the velocity of the fluid ( $k g ∕ m$ ), $L_{D}$ is the characteristic linear dimension of the environment ( $m$ ) and $η$ is the dynamic viscosity of the fluid ( $P a \cdot s$ ). Distribution of particles in a turbulent environment is analogous to heat diffusion in a solid [157]. The equations which explain this phenomenon are similar to Eq. ( 2.11 ) and can be written for spherical coordinates. Modelling the system in spherical coordinates has the advantage of possessing spherical symmetry. In addition, in nature, communication using molecules has been done using spherical puffs of particles (i.e. plant-to-plant [58]). In light of this, Eq. ( 2.11 ) can be converted to spherical coordinates with spherical symmetry:

\frac{\partial c}{\partial t} = \frac{K}{r^{2}} \frac{\partial}{\partial r} (r^{2} \frac{\partial c}{\partial r}),

(2.31)

where $K$ is the coefficient of eddy-diffusion ( $m^{2} ∕ s$ ), $c$ is the concentration function ( $k g ∕ m^{3}$ ) and $r$ is the radius of the transmission (m) with the following identity:

r^{2} = x^{2} + y^{2} + z^{2} .

(2.32)

By applying the boundary conditions defined in [158] a solution can be derived:

c (x, t) = \frac{M_{0}}{\sqrt{{(4 π K t)}^{3}}} \exp (- \frac{r^{2}}{4 K t}),

(2.33)

Turbulence in molecular communications is a new field of study with experimental work only recently gaining traction to analyse this kind of propagation in macro-scale aqueous propagation [159], [160].

2.6 Modulation

Figure 2.6: Types of modulation methods developed for use in molecular communications: Particle Quantity, Type and/or Structure of Particle and Time of Release.

Modulation is the process of changing the property of a periodic waveform with a modulating signal that typically contains information to be transmitted. The symbols are encoded into the carrier signal properties. In an EM -based communication system, the carrier waves are sinusoids and can be expressed mathematically as;

s (t) = A_{w} \sin (2 π f t + φ),

(2.34)

where $A_{w}$ is the wave amplitude, $f$ is the wave frequency (Hz) and $φ$ is the wave phase (rad). The amplitude is the peak-to-peak height of the signal, $f$ affects the number of oscillations that the signal makes in a given second and the phase is the shift the signal makes from the origin. Based on this, information can be encoded onto a sine wave by using any of the three properties mentioned above: Amplitude-Shift Keying (ASK), Frequency-Shift Keying (FSK) and Phase-Shift Keying (PSK). However this is not limited to using only one modulation scheme and hybrid modulations can be utilised.

Compared to EM -based communication systems, in molecular communication, the information carriers are particles (i.e. molecules) that are very tiny (range of Å). By using the EM modulation methods as a basis, the information can be encoded into molecules based on its following properties.

In this section 2.4, the main modulation methods developed for use in molecular communication will be discussed in detail. In further sections, the ISI and error-correction properties of different modulation methods will be analysed. Finally, a comparison of all the modulation methods described in the section can be seen in Table ( 2.3 ).

In this review, modulation methods are categorised into four distinct sections.

Those that primarily utilise particle quantity.
Those that exploit particle differences for modulation.
Those that modulate information based on the release times of particles.
Those that utilise more than a single property for modulation (aka., hybrid).

2.6.1 Particle Quantity

By implementing the quantity of the molecules as the method of modulations, the information can be encoded into the quantity of molecules used for transmission. An alternative approach would be to use a concentration based approach ( $m o l ∕ m^{3}$ ). This could be a better approach for use in macro-scale because of the scale of communication.

One of the first works carried out on molecular modulation can be traced back to [15]. This paper proposed two different modulation methods for use in diffusion-based propagation channels. The former is analogous to On-Off Keying (OOK) of EM communication systems where the bit 0 was defined as the concentration level of 0 ( $Q_{0} = 0$ ) and the bit 1 was defined as the concentration level of Q ( $Q_{1} = Q$ ).

In [161] and [162] two novel modulation methods were discussed for use in diffusion based propagation channels. The first method was based on the number of molecules arriving at the receiver below or above a certain threshold ( $τ_{s}$ ). This method was given the term Concentration Shift Keying (CSK) and is similar to ASK in an EM communication system; and the number of bits per transmitted symbol can be increased (Binary CSK 2-bits, Quadruple CSK 4-bits, $\dots$ ). The CSK modulation scheme mentioned is designed for a transmitter and a receiver with fixed positions. In [163], the CSK modulation/demodulation is anaylsed when the transmitter and/or the receiver becomes mobile.

2.6.2 Type and/or the structure of Particles

By using the chemical characteristic of the molecules, chemical formulas can be used as individual symbols for transmission. An example would be the use of acetone ( ${(C H_{3})}_{2} C O$ ) to encode the symbol 00 and n-hexane ( $C_{6} H_{14}$ ) to encode 11. The first proposed modulation discussed in [161] and [162] relies on the chemical difference of messenger molecules passing a certain threshold ( $τ_{s}$ ). This has been named Molecular Shift Keying (MoSK) and hydrofluorocarbon based chemicals were given as information carrying molecule examples (1-florobutane as 00, 1,3 - difluorobutane as 10 etc.). This method can also increase its transmitted bits in a similar fashion to CSK (Binary MoSK 2-bits, Quadruple MoSK 4-bits, $\dots$ )

There has also been research done on modulation methods for use in in-body molecular communication. In [164], it has been proposed to use aldohexose isomers as information carriers, called isomer based molecular shift keying (I-MoSK). The authors have also stressed that for the molecular communication to be bio-compatible, harmless compounds should be chosen as information carriers unlike the hydrofluorocarbons used in [162]. The advantages of using isomers in a molecular communication system is that isomers have the same type of atoms, which decreases the complexity of synthesis. Based on isomers, the authors have considered using isomers in CSK , I-CSK and MoSK , I-MoSK . Another approach would be to use isotopes which were studied in [165]. In [166] a modulation technique called depleted MoSK (D-MoSK) was proposed, which requires reduced number of the types of molecules for encoding compared to MoSk to reduce the complexity on the nano-machines.

A novel approach is to have two chemicals representing (+) and (-) signals which was studied in [167] where two chemicals are transmitted A and B, respectively, defined as type based sign (TS) modulation. In addition a variation of this modulation was proposed to implement it to utilise chemical reactions [168].

Table 2.3: Comparison of modulation methods.

Modulation Type	Modulation Method	Acronym	Principle			Reference

			Mass	Type	Time
Particle Quantity	On-Off Keying	OOK	$\times$	$\cdot$	$\cdot$	[15]
	Concentration-Shift Keying	n - CSK	$\times$	$\cdot$	$\cdot$	[161], [162], [163]

Particle Type	Molecular-Shift Keying	n - MoSK	$\cdot$	$\times$	$\cdot$	[161], [162]
	Isomer based Molecular-Shift Keying	I - MoSK	$\cdot$	$\times$	$\cdot$	[161], [162]
	Isomer based Concentration-Shift Keying	I - CSK	$\cdot$	$\times$	$\cdot$	[161], [162]
	Isotropic based Molecular-Shift Keying	n - MoSK	$\cdot$	$\times$	$\cdot$	[165]
	Depleted Molecular Shift Keying	D - MoSK	$\cdot$	$\times$	$\cdot$	[166]
	Type Based Sign	TS	$\cdot$	$\times$	$\cdot$	[168]

Time of Release	Pulse Amplitude Modulation	PAM	$\cdot$	$\cdot$	$\times$	[169]
	Molecular Frequency Shift Keying	——	$\cdot$	$\cdot$	$\times$	[15]
	Time-elapse Communication	TEC	$\cdot$	$\cdot$	$\times$	[170]
	SMART Time-elapse Communication	SMART - TEC	$\cdot$	$\cdot$	$\times$	[170]
	Pulse Position Modulation	PPM	$\cdot$	$\cdot$	$\times$	[169]
	Concentration through Silence	CtS	$\cdot$	$\cdot$	$\times$	[169]
	Rate Modulation	RM	$\cdot$	$\cdot$	$\times$	[169]

Hybrid Modulation	Multilevel Amplitude Modulation	M-AM	$\times$	$\times$	$\cdot$	[15]
	Isomer-Based Ratio-Shift Keying	IRSK	$\times$	$\times$	$\cdot$	[164]
	Molecular ARray-based COmmunication	MARCO	$\cdot$	$\times$	$\times$	[171]
	Molecular Transition-Shift Keying	MTSK	$\cdot$	$\times$	$\times$	[172]
	Power-Adjusted MTSK	MTSK - PA	$\cdot$	$\times$	$\times$	[172]
	Hybrid Molecular Scheme 1	HMS D1	$\times$	$\times$	$\cdot$	[173]
	Hybrid Molecular Scheme 2	HMS D2	$\times$	$\times$	$\cdot$	[173]
	Molecular Space Shift Keying	MSSK	$\times$	$\times$	$\cdot$	[174]
	Quadrature Molecular Space Shift Keying	QMSSK	$\times$	$\times$	$\cdot$	[174]
	Molecular Spatial Modulation	MSM	$\times$	$\times$	$\cdot$	[174]

2.6.3 Time of Release

Information can be encoded to the time in which the chemicals are either released or received by the transmitter [19]. An example would be to send Acetone at $t = 0 s$ to encode 00 and at $t = 5 s$ to encode 11. However, encoding information into timing can be a challenging task since the propagation of chemicals is random for diffusive (i.e., Brownian motion) channels in micro-scale and therefore this type of modulation may be more suited for application on macro-scale.

A variation on OOK is studied in [169]. The study was conducted on Pulse Amplitude Modulation (PAM). In this scheme, the bit-1 is encoded as a spike in the beginning of the time frame and the bit-0 is encoded as total silence in the time frame.

In [15], the concentration levels of the chemicals is varied in accordance to the sinusoidal signal in a given frequency similar to FSK in EM communications. This would add up to the already available concentration in the environment ( $Q_{a v g}$ ) an additional element is sent to the environment ( $Q_{a m p} \sin (2 π f)$ ) to make the chemical amplitude behave like a wave. With this the modulation is analogous to that of FSK .

Bacteria have the ability to behave as transceivers, known as quorum sensing [175], which enables them to interact with one another. A method that utilises this is the time-elapse communication (TEC) discussed in [170] and proposed for use in biological applications (i.e., bacteria) to increase the natural data rate of $1 0^{- 5}$ $b i t ∕ s$ . This is for very slow networks (i.e., on-chip bacterial communication). In this method, information is encoded into the time intervals between pulses. This method was tested on genetically engineered Escherichia coli (E. coli) bacteria. These bacteria would produce fluorescence light proportional to the information particle concentration however, the method experiences bit errors in transmission. In order to overcome that, the authors have proposed an improved TEC called SMART- TEC [170].

In [169] a modulation method based on the time property of the chemical emission is proposed. Pulse position modulation (PPM), relies on the emission of particles in an ordered manner to define the symbol. In a two bit system, the bit - 1 would be defined as a short pulse then silence for the remainder of the timeframe, whereas bit - 0 would be defined as a short pulse in the middle of the time frame. In [176], the concept of PPM is further developed and higher order ( $M_{C} > 2$ ) values were shown to outperform CSK .

A second modulation method proposed in [169] is named Communication through Silence (CtS), where the symbol is encoded between the start pulse and the stop pulse. The final emission-based modulation scheme proposed in [169] is Rate Modulation (RM). This modulation relies on the number of pulses in a given time frame. An example would be, encoding “9” with 9 pulses in a time frame $t$ and “12” with 12 pulses in the same time frame $t$ .

2.6.4 Hybrid Modulation Methods

In this review, hybrid modulation describes a method that utilises two or more types of modulation methods. In [15] a combination of OOK and molecular FSK was implemented to increase the throughput of the communication named multilevel amplitude modulation (MAM). This is achieved by modulation $Q_{a v g}$ in conjunction with $Q_{a m p} \sin (2 π f)$ to achieve a higher rate of encoding. In [164] the authors have proposed using isomer ratios to encode information, which has been named IRSK . This method relies on the ratios of the quantity of isomers the detector absorbs. The authors have stressed that this kind of modulation would be more robust, since the ratios are the encoding element rather than the quantity. In [177], a method based on the combination of type and the time of release was proposed. This method, being asynchronous, could achieve higher information rates compared to time based methods. To define the model, the method was presented as an event-driven system. In addition a hybrid modulation scheme based on CSK and MoSK , named run-length aware hybrid modulation have been proposed that reduces the ISI of the communication [173].

Another work on modulation in continuous diffusion was [178], a pulse based modulation scheme was proposed. This modulation scheme, however is used to send simple commands to adjoining nano-machines (i.e., detection of unwanted particles) and based on fast change of concentration in a given environment. In [179] a modulation was proposed combining the properties of concentration shift keying (CSK) and molecular-shift keying (MoSK) into a single modulation method for use in high data rates. Another modulation method that combines CSK and MoSK was proposed in [173]; these are, modulation method hybrid modulation scheme design 1 (HMS-D1) and 2 (HMS-D2). In [174] novel modulation methods for use in MIMO application was developed. Fist, named molecular space shift keying (MSSK) uses the antennae indices as the information source. Second, quadrature MSSK (QMSSK) uses two types of chemical and finally a molecular spatial modulation which is a combination of MoSK and MSSK .

2.7 Inter-symbol Interference (ISI)

In molecular communications the noise of the environment behaves differently compared to an EM communication. In EM , the environmental noise ( $N_{0}$ ) generally stays stable, however molecular communication is based on particles and because of that, particles may remain in the propagation medium and may affect future transmissions and may or may not arrive at an unintended time slot. This causes a decrease in molecular communication channel capacity and creates a memory effect [16].

The memory effect can be a major problem if there is a continuous stream of transmission, which makes ISI a big problem in molecular communication since continuing transmission would eventually decrease the channel capacity to such extent that information transmission would become infeasible. Because of this, reduction of ISI should be a primary concern when designing a modulation method. Due to this hindrance, the property of ISI has been studied extensively [46], [162], [180], [181], [182], [183], [184].

There are methods in which ISI can be reduced. Based on the literature and the research done in this problem, ISI reduction can be classified into two types:

Single type of messenger chemical.
Multiple types of messenger chemicals.

In the following sections ( 2.7.1 and 2.7.2 ), both approaches will be discussed in detail.

2.7.1 Single Messenger Chemical

The approach of utilising only a single chemical has the limit of only having a single messenger molecule type to work with. Increasing the time slot of each symbol in the communication can be given as an example. However this would sacrifice the throughput of the channel for transmission reliability. This method, however, would make the already slow molecular communication even slower, which has forced new methods to mitigate ISI in an efficient way.

Another approach of ISI mitigation is to remove the leftover chemicals in the communication channel. A method to accomplish this is to use enzymes or inhibitor chemicals [185], [186], [187], [188], [189]. In this method, the enzymes that are present in the channel degrade the information particles as they are transmitted, thereby reducing the ISI and the receiver error probability. Alternative approaches are also present in the literature such as; ISI mitigation by calculating of the optimum reception delay [190], adaptive detection [191] and photolysis reactions [192].

2.7.2 Multiple Messenger Chemicals

Utilisation of multiple chemicals for messenger particles has advantages over a single chemical messenger system. One being that since chemicals are independent of each other, more symbol time can be given to the chemical without hindering the throughput.

In [172] an approach to reduce the ISI was proposed by combining CSK and MoSK and calling it molecular transition shift keying (MTSK). In its binary form two types of molecules were implemented, defined as type-A and type-B. The modulator then decided which molecule to send based on the previously sent bit and the current bit. If type-A is sent the system sends type-B and vice versa. In the same paper, a power efficient version (MTSK-PA) is also proposed where the residual molecules from the previous symbols are utilised to reduce the energy consumption.

An alternative ISI reduction method is to utilise the order of released information particles [171]. This method, called MARCO by the authors, defines bit-0 as releasing first particle $A$ and then later $B$ . For bit - 0 the release is vice versa. This method was shown to reduce the ISI. In [193], a similar idea was implemented in which alternating molecules were used to reduce ISI (e.g., type-a molecule for even time slots and type-b molecules for odd time slots). By using more chemicals, more rotations can be made between time slots and this can help reduce ISI greatly.

2.8 Error Correction

By definition, error correction is used to diminish the effects of noise on the transmitted signal that could be caused by the environment or sensors and correct the errors by introducing redundancy [16]. This redundancy can later be detected to check whether information was changed or not. For example, in a repetition code, bit - 0 is encoded as 00000 and bit - 1 as 11111 when this code is received as 00101 and 10111, it can easily be identified and decoded correctly [194].

Figure 2.7: A natural process in error correction occurs in DNA replications. The error correction in DNA is a collection of processes by which a cell identifies and corrects damage to the DNA molecules that encode its genome [ 195 ].

One of the first attempts to use error correcting codes was to implement one based from EM systems. A Hamming code [23] was applied to On-Off Keying (OOK) with diffusion based molecular communications. It was shown that by implementing such codes and by using a large number of molecules the system outperformed communication without redundancy. However, if fewer particles were used the inverse became true. This happens because of the extra ISI added by the parity bits. The authors also expressed that it is not an energy efficient solution if the separation distance between the transmitter and the receiver is small. Based on this energy issue work was done to introduce energy efficient codes for use in molecular communication [196] and in [197] energy comparison of Self-Orthogonal Convolutional Codes (SOCCs) was conducted.

Another attempt was made in [198], [199] where an alternative to the already established Hamming distance is proposed, named Molecular Coding Distance Function (MCDF). In this function, the coding distance is not defined as a static value, but based on a transition probability. It was shown that this method shows better performance than Hamming coding in molecular communications. Finally applications of already existing error-correction code methods have been reported: High-order Hamming Code [196], Minimum Energy Codes [200], Reed-Muller Codes [201] and Reed-Solomon Codes [202].

In [196], a new family of channel codes have been proposed called ISI - free codes that takes the limited decoding ability of the sensors in mind. These coding schemes have been simulated in a diffusion environment by the use of Brownian motion. In [201] a comparison study of the recently proposed channel coding schemes were reviewed.

A problem of modelling the channel is the inconsistency across a given distance. This is caused by numerous properties of the environment (temperature, molecule collision etc.). To overcome this, channel state information is calculated. This method helps to describe how the signal propagates. The channel state information (CSI) can be classified based on which knowledge is known; either the current condition instantaneous CSI or the statistical knowledge statistical CSI . In [84], it was shown that in a CSK modulation the strongly constant weight code (SCW) can be used in molecular communication to improve transmission. In [203] constant-composition codes were used to mitigate the use of channel state information (CSI). In [204], error performance analysis of diffusion based molecular communication with OOK was done. In [205] zero error codes were implemented for molecular communications. In [206] error-correction via-amplitude width encoding was done. In [207] the performance of Bose-Chaudhuri-Hocquenghem [208] and Reed-Solomon codes [209] for molecular communication with diffusion are evaluated by simulation and results are analysed.

2.9 Information Security

An important aspect of any communication system is the security it gives to its users. Information transmission and encryption is a heavily studied topic in EM communication; however, in molecular communication, it has yet to gain the interest of researchers. A study was made in [210] to analyse the implementation of a secure channel for molecular communications and in [211] security and privacy were discussed for use in molecular communication.

2.10 Receivers

In order to decode the message, a receiver must be used. The receiver can be any device that is able to detect chemical concentrations and be able to convert them into electrical impulses. For use in macro-scale molecular communication, an electronic nose (section 2.10.2.1 ) or a mass spectrometer (MS) (section 2.10.2.2 ), which are already common place in many industrial applications, can be utilised. In micro-scale the sensor and applications are still a new concept and are a current research interest.

2.10.1 Micro-Scale Molecular Communication

The use of sensors in the micro-scale is still a relatively new concept. The use of nanoscale sensor networks (NSN) was first proposed in [12], where sensors are defined as nodes at nano-scale sizes. These small scale devices can be used as a network and can be used in applications such as medicine or military [212]. However the conceptualisation and the production of these kind of sensors is hard to produce to the same performance standard compared to their macro-scale sensor counterparts because of their dimensions. In [213] networking schemes for these kind of sensors were discussed. Recently there have been developments of nano-sensors that can be utilised for micro-scale [214], [215], one of the ideas being the use of biological circuits, based on chemical reaction networks and DNA transcription processes [216], [217].

An alternative way of implementing sensors for use in micro-scale can be derived from organic systems. In [218] the use of artificial neural networks (ANN) was proposed for use in micro-scale. These ANNs could be interfaced to a nano-device and achieve communication in a small size. In [170], quorum sensing in bacteria were utilised to transmit information, which can further be used as sensors in future applications.

In literature, sensors are generally categorised as either passive or active [219]. A passive receiver is defined as a receiver that does not interfere with the environment such that particles that are detected are not removed from the environment [220], [221], [222], [223], [224], [225]. An active receiver, on the other hand, directly interferes with the particles in the environment such that the particles are absorbed and removed from the system [136], [226], [227], [228].

An important part of any communication is synchronisation and arrival time [229], [230]. A communication that has synchronisation stops slips from occurring and maintains performance. Based on this there have been studies done on the possibility of synchronisation in molecular communications with diffusion [231], [232] and with advection [233].

Deep Learning has been gaining traction as an impressive tool in optimisation and therefore has been used in numerous fields on real-life applications. Recently, molecular communication has also seen deep learning in decoding transmitted signals [234], [235] and receivers [236].

In [237], a detection network was devised for molecular communications. In [238] a new pre-coding algorithm was introduced at transmitter side to mitigate the ISI for adaptive threshold detectors was shown and applied. In [239] signal detection based on derivation for high-data rate molecular communication is done. In [240] the effect of spatial partitioning on the variance of the output signal is investigated and in [241] improving the receiver performance for molecular communication is improved by use of an adaptive weighted algorithm.

2.10.2 Macro-Scale Molecular Communication

An advantage of macro-scale over micro-scale is that there are technologies already available that can be implemented as detectors. The information particle of a macro-scale can utilise gas particles and there are various types of detectors in the market that can be used. An important group of gases are volatile organic compounds (VOCs). These types of gasses are used in nature by plants mostly for communication purposes [58]. These chemicals are mostly in their gas phase at room temperature and they can be detected using electronic noses or the use of mass spectrometers (MS) which are discussed below in detail.

2.10.2.1 Electronic Nose

An electronic nose is defined as a sensory tool used to detect chemical odour/scent. The first conceptualisation of this device was made by Persaud and Dodd in 1982 [242]. From [243] an electronic nose has the following:

sensor matrix to simulate a human olfactory system.
pattern recognition system that recognises the olfactory pattern data processing unit which can perform similar function as an olfactory bulb.

The ability of discriminating and recognising a variety of gases and odours using small numbers of sensors led the electronic nose to be used in various areas, such as medical and diagnostics [244], [245], [246], food [247], [248], [249] and cosmetics sectors [250]. A block diagram of an electronic nose can be seen in Figure 2.8 .

Figure 2.8: A block diagram of an electronic nose.

The basic prospect of an electric nose is to convert the chemical footprint into a detectable electrical signal. This could be achieved using different properties of a type of material (change of resistance, temperature etc.). However, some measurement of characteristics are suited better for certain applications. In [43] the use of electronic noses was deemed to be the best candidate for use in an odour communication system due to their portability compared to mass spectrometers. However, as it will be mentioned in the next part, mass spectrometers provide specific qualities that can make the MS a de facto standard for use in macro-scale molecular communication experimentation.

2.10.2.2 Mass Spectrometer

Mass Spectrometers (MSs) have played a key part in many phases of drug discovery and the identification of proteins. This is all possible because of the MS ’s sensitivity and speed of analysis. Aside from its importance in the drug industry, MS is also a vital part in many fields. Examples would be identifying and monitoring bio-markers in physiological fluids, rapid screening of drug-target binding and other biochemical analysis. As a result of its vast usage in most fields and for its versatility, MS has became a routinely used technique. MS have the ability to detect numerous chemicals concurrently and with high resolution making it a valuable option for use in molecular communications as a receiver. A detailed description of a type of MS, quadrupole mass analyser (QMA) is given in Chapter 3.

Figure 2.9: The principle diagram of the quadrupole mass analyser used in chemical detection: (1) Once the samples are introduced into the detector via the membrane inlet the chemicals are ionised and passed through a focus lens. (2) The analyser is made up from four hyperbolic rods with applied potentials. Based on the electric potential applied to the rods, which the connection diagram can be seen in (5) , the ions can be separated based on their mass-to-charge values. Ions with stable trajectories, such as (4) will travel through the electric field and will arrive at the detector shown in (6) , whereas ions with unstable trajectories (non-resonant ion), e.g. (3) , will collide with the quadrupole and will be filtered out from the detection. Detected ions are amplified and presented visually as mass chromatogram shown in (7) .

The name “ Mass Spectrometry” gives an unclear definition of the name. The measurement is not done on the mass of the sample (kg) but on the mass-to-charge (m/z) ratio. A mass-spectrum is defined as the relative ion abundance ( $%$ ) vs. m/z. The spectrum is using the units of Dalton (Da) per unit charge ( $q$ ) [251]. In a MS analysis, all information comes from gas-phase ions. There are three primary components of a mass spectrometer: an ionisation source, a mass analyser and a detector.

The use of MS in molecular communication is relatively new with proof-of-concept transmission first reported 2017 in [25]. To utilise the experiments, a membrane inlet mass spectrometer (MIMS) with a QMA is used. In [47] the study was further developed and the modulation methods of OOK and CSK were investigated. In [46] an experimental message was sent using a MIMS and [2] an extensive analysis was done on the parameters of the macro-scale. In [3], the noise was analysed to be additive white Gaussian noise (AWGN) and the open-air transmission of macro-scale was analysed.

2.10.3 MIMO Applications

Figure 2.10: A representative diagram of the implementation of MIMO for use in molecular communications.

In radio applications, multiple input multiple output (MIMO) is a method to increase the capacity of a radio link using multiple antennas and receivers [252]. This technique is widely used in traditional communication such as IEEE 802.11n (Wi-Fi), IEEE 802.11ac (Wi-Fi) etc.

There have been studies to implement this method for use in molecular communications. In a study conducted in [140], one of the first instances of MIMO use in molecular communication was conducted. However, in this study, the effects of ISI were not considered. The first study that considered both MIMO and ISI was in [24].The main drive of the utilisation of MIMO is to increase the already slow communication of molecular communication. Based on this premise various studies were realised on the properties of molecular MIMO , such as detection [253], [254], [255], [256], modulation analysis [174], channel analysis [257], [258], [259] and cooperative relaying [260].

2.11 Experimental Analysis

Figure 2.11: Experimental test beds used in literature

(A): The diagram of the odour generator-MS experimental setup [ 2 ], [ 3 ], [ 4 ], [ 46 ], [ 47 ]: (1) $N_{2}$ gas; (2) modulation information; (3) automation platform; (4) MFC for the carrier flow; (5) MFC for the signal flow; (5) evaporation chamber (EC); (7) mixing chamber; (8) propagation medium; (9) MS inlet; (10) electronics control unit; (11) pressure gauge; (12) controller and the regulator cables; (13) data acquisition and analysis.
(B): Bench-top experimental setup [ 27 ]: (1) transmitter; (2) voltage source; (3) incubator; (4) arduino; (5) LED; (6) pH sensor; (7) magnetic stirrer; (8) bacteria in glass tube; (9) pH meter; (10) Receiver.
(C): Tabletop Molecular Communication [ 39 ]: (1) Transmitter; (2) Arduino; (3) Adafruit LCD shield kit; (4) DuroBlast electrical spray; (5) MQ-3 sensor; (6) Arduino; (7) receiver.

The study of molecular communication has mostly been done on the theoretical aspects. However, macro-scale molecular communications have seen the use of practical experimental test-beds to study the aspects of the methods of communication.

One of the first test beds to study the effect of molecular communications is shown in [39]. In this study, as a transmitter an electrical spray controlled by an Arduino is used and for detector a MQ-3, MQ303A and MR513 sensors were utilised. The transmitted chemical was chosen as isopropyl alcohol ( $C_{3} H_{8} O$ ). In [261], an experimentally validated end-to-end channel model for molecular communication systems based on the setup in [39] was developed and experimented.

Another approach uses the characteristics of the molecules to convey information which was accomplished in [262]. In this research the pH values were used. The carriers of information are designated to be hydrogen ions and the detection of these particles is achieved by use of a pH sensor. The transmitter is made up from a series of peristaltic pumps which compresses the tube to create a flow. The propagation channel is a silicon tube and the receiver is a pH meter interacting with an Arduino Uno. From experimental results it was demonstrated that transmission speeds of 4 bps could be achieved.

In [263] an alternative approach was considered, where instead of propagating the particles in a gaseous environment an aqueous environment was chosen and the transmitter chemical was chosen to be lauric acid coated SPION s (superparamagnetic iron oxide nanoparticles), which were employed in drug delivery applications [264]. Propagation and transmission is accomplished by a peristaltic pump and the detection is done by susceptometer coil which generates an electrical signal when there are magnetic materials in close proximity.

A short study was conducted in [160] where the information rate of molecular communication is studied in laminar and turbulent flow.

A recent study was done in [27] where a biological approach was considered. The setup utilised a colony of bacteria ( E. Coli ) expressing bacteriorhodopsin in their cell membrane for easy-to-control optical signal conversion. Based on the input from the photo sensor, bacteria change the pH of the environment and the transmission is detected via a pH sensor.

In [265] a similar experimental setup to [39] was used to study and analyse anti- ISI demodulation schemes for diffusion based molecular communications.

A study was done in [266] where a proof-of-concept multi-input multi output (MIMO) molecular communication was conducted. This MIMO application was further analysed in [24] and in [267] the application was studied with an additional advection element.

In [268] the first controlled information transfer through an in-vivo nervous system by modulating digital data from macro-scale devices onto the nervous system of common earthworms and conducting successful transmissions was demonstrated.

While the experimental studies conducted rely on different principles of operation (electronic sprays, bacteria colonies etc.) most of the experiments conducted are short distance transmission (< 1 m). Long distance ( > 1 m) transmission is only studied by either using an Arduino or a MS as a receiver. While the approach using Arduino allows transmission over long distances, it relies on external fans to propagate and utilises liquid chemicals to send its information. This approach may not be suitable for systems where the transmission needs to be conducted in a boundary (i.e., pipe). This is caused by the gravitational forces acting on the liquid chemical.

The last approach, MS , overcomes this problem by utilising gas-phase-only transmission with pressurised $N_{2}$ gas to create propulsion. This approach allows transmission in closed environments over long distances (> 1 m).

Table 2.4: Experimental test beds used to study molecular communications.

Transmitter	Propagation Medium	Messenger Particles	Detector	Distance	bps	Reference
Odour Generator	Gaseous	VOC s	Mass Spectrometer	4 m	0.05	[2], [25], [46], [47]
Arduino with			MQ-3
Electrical Spray	Gaseous	Water Droplets	with Arduino	4 m	0.5	[39], [261]
LED with Arduino	Aqueous	E. Coli Bacteria	pH Sensor	< 1 m	0.016	[27]
		Superparamagnetic
		Iron Oxide
Peristaltic Pump	Aqueous	Nanoparticles	Susceptometer coil	5 cm	0.25	[263]
Peristaltic Pump	Gaseous	Ions	pH Sensor	< 1m	4	[262]

2.12 Applications

2.12.1 Micro-Scale Molecular Communication

Micro-scale molecular communications can be used in fields such as drug delivery systems/medicine [11], [269], [270], [271], [272]. In addition, studying micro-scale molecular communications gives insight to the communication done in the nm to $μ$ m range. This knowledge could further be used to develop bio-computers [273], [274].

An important application of micro-scale is the manipulation of biological processes from an engineering perspective. This can be done by using quorum sensing [175]. By mimicking this stimuli system, self-organising collective sensors (SECOAS) can be improved to be used in environmental monitoring system. This is a swarm made by individual nodes, which chooses a node to transmit in order to save on limited energy [275]. This molecular application can also be applied to ad-hoc networks [276]. There are biological approaches in which micro-scale molecular communication can be applied such as ${C a}^{+ 2}$ signalling [277], ligand-receptor binding mechanism [278], [279], synthetically engineered bacteria [280] and electroencephalogram (EEG) waves [281].

Molecular communications can be used to analyse in-vessel transport. It was proposed in [32] that vessel-like structures provide a boundary to the transportation and guides propagation to the desired destination. This aspect of communications has been studied in literature on its various properties, such as decoding method [282], channel geometry [283], [284] and channel model [285].

2.12.2 Macro-Scale Molecular Communication

The current applications of macro-scale are limited since most of the conventional large scale communications are dominated by traditional communication methods (EM) that have been well developed and established. However, there are areas where EM communication is not desirable and an alternative might prove to be a better choice. A study has shown that EM based sensors have limited reliability in infrastructure monitoring [62]. In addition, environments where EM communication is limited because of obstacles [64] (i.e. caves, mines) macro-scale can be a good alternative and can also be used in sending information using pipes. In [6], it was shown that under specific conditions molecular communications outperforms electromagnetic communication in terms of signal quality in a pipe where EM has higher attenuation per distance in a copper pipe.

In addition to being a tool for communication, macro-scale molecular communication can also be used for studying biological processes such as pheromone communication in animals [18]; ants using chemical trails for navigation [60] or moths using chemical signalling over several $k m$ [51]. Based on these biological processes, much research has been done to implement this biological application and make it applicable as an engineering approach [50], [286], [287], [288].

Figure 2.12: Applications of molecular communication can be implemented both in micro-scale and macro-scale. At smaller scales, understanding of particle propagation and reception can be used to control bacteria and be used in implementing drug delivery systems. At larger scales, molecular communication can be used in studying animal behaviour.

Another application of macro-scale is in robotics. By utilising pheromone communication and olfaction [68], [79], [289] communication between robots can be established. The research field for this application can be divided based on two distinct approaches:

Chemical (odour) tracking robots [68], [69], [70], [71], [72], [73], [74], [75], [76] .
Pheromone based communication [79], [80], [81], [290] .

A different approach to the use of macro-scale molecular communication would be odour transmission using digital medium [43]. The topic has been known since the 1950s and deals with sending and receiving scent in digital medium; it has become more important with the mass availability of virtual reality (VR) hardware [291]. These hardware are impressive in simulating the reality for both audio and video but the sensory input for odour is still a problem to emulate. The transmission of audio and video are well understood, however the transmission of odour from transmitter to the receiver is still a topic of research [292], [293]. The knowledge gained on studying macro-scale chemical propagation can be further implemented to this topic which can be used to create realistic scent streams to further improve the illusion of reality of a VR environment.

2.13 Simulations

Simulation is the imitation of an operation of a real world process [294]. As it plays a very important part in many fields of engineering, molecular communication has also seen a fair share of simulation platforms designed to simulate the process of propagation and detection. One of the limitations of micro-scale over macro-scale is the scarcity of experimental analysis. This is due to size of the scale and lack of equipment able to carry out the experiments. Based on these drawbacks simulations play an important role in the analysis and the study of numerous parameters of the communications.

As mentioned there are a variety of simulation models designed to analyse specific properties of this novel communication paradigm. In this section each simulation platform will be briefly discussed and their comparison can be seen in Table 2.5 .

Table 2.5: Simulation platforms for use in molecular communications.

Simulator Name	Customisation	License	Function	User Interface	Implementation	References
BiNS2	Customisable Domains	Open	Diffusion based	Command Line	Java	[295], [296]
CalComSim	via Python	Open	Calcium Signalling in tissues	Command Line	Python	[297]
N3Sim	via Config File	Open	Diffusion	Command Line	Java	[298], [299]
Comsol	Script based module	Commercial	Multipurpose	Graphical	N/A	[284]
NS-3	via C++ & Python	Open	Network Simulator	Command Line	C++ & Python	[297], [300]
NanoNS	via C ++ & OTcl	Open	Diffusion Based	Command Line	C++	[22]
NanoNS3	via C ++ & OTcl	Open	Diffusion Based	Command Line	C++	[301]
BNSim	via Java	Open	Bacteria Networks	Command Line	Java	[302]
NCSim	via C++ & Python	Open	Flagellated Bacteria Networks	Command Line	C++	[298]
HLA	via C++ & Python	Open	Flagellated Bacteria Networks	Command Line	C++	[303]
AcCoRD	via C++ & Python	Open	Flagellated Bacteria Networks	Command Line	C++	[192], [219], [304], [305], [306], [307]
MUCIN	via MatLAB	Commercial	Diffusion	Graphical	MatLAB	[308]

2.13.1 BiNS

Biological and nano-scale communication simulator (BiNS) is a multi-threaded simulation package for molecular communication systems developed by the researchers from the University of Perugia [295]. Its customisable design provides a set of tools for generating objects and allows modelling the behaviour of biological entities, including the collision handling and the modelling of diffusion with advective flow propagation onto both constrained and open space environments.

2.13.2 N3Sim

N3Sim is a java-based simulation framework that used diffusion only molecular communications. This platform allows the analysis of molecular networks possessing several transmitters and receivers [298]. The diffusion process through the medium is modelled as Brownian motion, which takes account of particles’ inertia and collisions among particles as they travel through the medium. The simulation is based on a three-layer architecture:

User interface layer.
Data layer.
Domain layer.

The simulation parameters are defined by the text configuration file.

2.13.3 COMSOL

COMSOL multiphysics is a commercial multi-purpose platform designed for simulating and analysing physics-based problems through a unified work flow for electrical, mechanical, fluid, and chemical applications [309]. It implements finite element analysis (FEA), to study different physics and engineering applications. An example of the use of COMSOL multiphysics for simulating a molecular communication drug delivery system is presented in [309] and is also used in simulating micro-fluidic environments [284].

2.13.4 NS2 and NS3

NS2 and NS3 are discrete-event network simulators. Even though they were not originally developed for molecular communications, their flexible structure has allowed implementing some basic elements of molecular communications.

NanoNS [22] is an NS-2 based simulator for diffusion based molecular communication in aqueous environments, with continuous thermal motion of molecules.

NS-3 is a simulation tool that has been developed in the framework of the IEEE P1906.1 working group [310]. User programs can be written in C++ or Python programming languages.

2.13.5 BNSim

BNSim is a multi-threaded Java-based simulation platform for analysing bacteria networks [302]. These networks interconnect engineered bacteria that communicate at the nano-scale.

BNSim functions by integrating three simulation methods:

Gillespie stochastic simulation algorithm [311].
Stochastic differential equations (SDEs), used to model large-scale chemical system with a controlled level of approximation.
Hybrid algorithm which integrates the above methods.

2.13.6 NCSim

NCSim is a simulation platform for molecular communications that utilises flagellated bacteria for transmission of information [312]. The simulations primary objective is on the analysis of different message encoding techniques. It can simulate several simultaneous links between the nano-machines. NCSim incorporates the stochastic model for bacteria mobility, and the plasmid/chromosome transfer between bacteria through the conjugation process. NCSim consists of three modules:

Physical layer of bacterial nano-networks.
Scenario generator and simulation monitor.
Plot generation.

2.13.7 HLA

High level architecture (HLA), which is standardised under IEEE 1516, is used to design and develop a distributed simulation tool for molecular communication and in [303], the authors introduced a simulator design based on the HLA model focusing on scalability. It is used to design a distributed simulation tool for molecular communications, so that different scalability options can be used to include additional processing power to reduce the execution time. This model allows the design of large systems, which could be difficult to do otherwise.

2.13.8 AcCoRD Simulator

AcCoRD (actor-based communication via reaction-diffusion) is a molecular communication simulator and designed as a generic reaction-diffusion solver for flexible system configuration. Actors are placed as sources (i.e., transmitters) or observers (i.e., receivers) of molecules. Environments can be defined with a combination of microscopic and mesoscopic regions [192], [219], [304], [305], [306], [307].

2.13.9 Other Works

Aside from simulation platform there are other studies which aid in the simulation studies of molecular communication, such as; algorithms [313], [314], [315], [316], multithreaded CPU and GPU implementation [317], parallel processing [318], analytical analysis [319], [320], [321], modelling [322], [323] and frameworks [324].

2.14 Standardisation

Standardisation is defined as the process and the implementation of technical standards based on the consensus of different parties [325]. Molecular communication is a relatively new field of study compared to EM where most of the technologies are standardised. However, there are attempts at standardising aspects of molecular communications.

One of these attempts is the unification of the research of molecular communications into a single architecture, named molecular communication markup language (MolComML) [26]. This standardisation is similar and influenced by the systems biology markup language (SBML) which is the de-facto standard for representing biological processes through computational models in systems biology [326] and in [327] a new metric was proposed for the performance evaluation of molecular signal within the context of molecular communications.

Aside from the research standardisation, there are also works on implementing standards for the communication aspect of the system. In [212] the communication system was analysed based on the open systems interconnection (OSI) model and each layer was discussed in detail and in [328], transmission control protocol (TCP) was implemented and studied for molecular communications.

Another approach is to standardise how information is transmitted. This is done in EM communications as packets where information is encoded to the package to define its numerous properties; beginning, end, where it is headed to etc. This approach was also recently studied in [329] where data frames in molecular communication are envisioned.

2.15 Conclusion

In this Chapter, a review into the molecular communications is undertaken. Molecular communications is still novel and still being experimented upon, and there are aspects of the field where studies need to be conducted. One of these fields is information security. Information security is a pivotal aspect of modern communication and becomes ever more important as time passes. However, aside from a few preliminary studies, research is required. Studies have been done into modulation methods, ISI mitigation and mathematical modelling of the propagation. These properties are mostly based on already established principles in EM. There are still aspects of molecular communication yet to be discovered that is intrinsic to its particle properties that cannot be implemented to a wave-based communication system. Experimental studies have recently gained traction and have shown physical proof of the validity of molecular communication both biological and mechanical. Simulation platforms were developed to increase the speed of progress of this field. Applications exist where molecular communications is a better alternative to already existing communication methods and in the future, molecular communications shows promise in improving our ability to communicate in areas where the current methods (i.e., EM) struggle by fundamentally changing the method of communication.

2 Background and Related Work