You are on page 1of 9

E-Machine

In describing the e-machine from the stress tensor. Symbols where generated from each
stress tensor components ( i.e. zz and zy) for each time frame of the trajectory.

GRAPH.
The symbols generated are base on an algorithm we deduced in analysing the pattern of
the stress tensors components as plotted in the graph over a period of discrete time step.
Its shown from the graph in figure 1 and 2 that the stress tensor fluctuates rapidly in
discrete time.

Figure 1: Fluctuation of the zz-component of the stress tensor

Figure 2: Fluctuation of the zy-component of the stress tensor


PLATEAU
In order to know it they might be an any existing behaviour of this random signal, if its purely
random or not. Figure 3 shows a basic two state symbols generator algorithm that generates
a set of strings for pattern analysis of the fluctuations of the stress tensor components. This
algorithm checks for the signal fluctuations from positive to negative or visa viz. This two state
is symbolized by 0 and 1. These symbols 0 and 1 means the system is at a tensile (negative)
state, and compressive (positive) state respectively.

ALGORITHM 1:

Function generatesymbolsbystateofstress(stcomponent)
s = ""
for i in range(0, len(stcomponent))
then
v = stcomponent[i]
if v >= 0:
s = s + "1"
if v < 0:
s = s + "0"
return s

Figure 3: A two state symbols generator

The symbols generate a finite length of 25000 strings. For example the generate string for the
zz-component is:
010000000100000010100110110000010000001100000000001011001011100100000001
11100011100000001110100111000100101000000000000000000000001000000000000001
00010000001001000001000001000111000001000010010000011100010101011000011111
00110101011010101100000001001000000010100000000000000010110010000000110101
0000101101010100001
From this string we generated the plateaus and e-machine at different time step as follows:

L 1 2 3 4 5 6 7 8 9 10

0.1000 2 4 8 16 32 64 128 256 512 1022
0.0656 2 4 8 16 32 64 128 256 512 1022
0.0430 2 4 8 16 32 64 128 256 512 1022
0.0282 2 4 8 16 32 64 128 256 512 1022
0.0185 2 4 8 16 32 64 128 256 512 1022
0.0122 2 4 8 16 32 64 128 256 512 1022
0.0080 2 4 8 16 32 64 128 256 512 1022
0.0052 2 4 8 16 32 64 128 256 512 1022
0.0034 2 4 8 16 32 64 128 256 512 1022
0.0023 2 4 8 16 32 64 128 256 512 1022
0.0015 2 4 8 16 32 64 128 256 512 1022
0.0010 2 4 8 16 32 64 128 256 512 1022
Figure 4: Plateau generated for 0.4ps of the zz-component, no converges.
L 1 2 3 4 5 6 7 8 9 10

0.1000 2 4 8 16 30 55 89 135 183 219
0.0656 2 4 8 16 31 55 91 135 178 216
0.0430 2 4 8 16 31 55 91 133 174 217
0.0282 2 4 8 16 31 55 90 134 175 215
0.0185 2 4 8 16 31 55 90 134 175 220
0.0122 2 4 8 15 31 55 90 133 176 220
0.0080 2 4 8 15 31 55 91 134 177 216
0.0052 2 4 8 15 31 56 90 134 177 222
0.0034 2 4 8 15 31 56 90 134 177 222
0.0023 2 4 8 15 31 56 90 135 176 216
0.0015 2 4 8 15 31 56 91 135 176 215
0.0010 2 4 8 15 31 56 91 135 176 215
Figure 5: Plateau generated for 24ps of the zz-component, no converges.

The systems starts showing a significant level of convergence from 89.6ps, with a total state
of 69 casual states.
L 1 2 3 4 5 6 7 8 9 10

0.1000 2 4 8 15 27 41 57 71 68 74
0.0656 2 4 8 15 27 40 56 70 68 74
0.0430 2 4 8 15 27 41 56 65 71 75
0.0282 2 4 8 14 27 41 54 69 72 75
0.0185 2 4 8 14 27 41 54 69 69 76
0.0122 2 4 8 14 27 41 54 69 69 76
0.0080 2 4 8 15 27 41 55 69 69 76
0.0052 2 4 8 15 27 41 55 69 69 76
0.0034 2 4 8 15 27 41 55 69 69 76
0.0023 2 4 8 15 27 41 55 69 69 76
0.0015 2 4 8 15 27 41 55 69 69 76
0.0010 2 4 8 15 27 41 55 69 69 76

Figure 6: Plateau generated for 89.6ps of the zz-component, converges to 69 states.


The diagram shows the e-machine with 69 states as depicted by the plateau.

In addition, we considered using a much complex algorithm to deduce the signals of the stress
tensor behaviour over time. This is done to know at the minimum any hidden pattern of the
stress tensor signals as the molecular trajectory evolved in time. We considered having an
initial five state based on the fluctuation of the signal through the compressive and tensile
behaviour of any stress component. In some case the stress component remains compressive,
before becoming tensile over time and visa viz. When the component is at a compressive
state, it either increases or reduces in compression, before transiting to a tensile state. This
also apply to the components at a tensile state. The algorithm presented generates the
symbols based on these behaviours of the stress components over the given discrete time
frame of the molecular trajectory.

ALGORITHM 2:

Function generatesymbols(stresscomponent):
s = ""
for i in range(1,len(stresscomponent))
then
x = stresscomponent[i-1]
y = stresscomponent[i]
# check +ve x and -ve y
if x > 0 and y < 0:
s = s + "a"
# check -ve x and +ve y
if x < 0 and y > 0:
s = s + "b"
# check for -ve x and -ve y
if x < 0 and y < 0:
if y >= x:
s = s + "c"
if y < x:
s = s + "d"
# check for +ve x and +ve y
if x > 0 and y > 0:
if y >= x:
s = s + "e"
if y < x:
s = s + "f"
end
return s

Figure 3: A Six state symbols generator

The symbols generate a finite length of 25000 strings. For example the generate string for
the zz-component is:
.badcdcdcddcdcccbacdbacdccdbadbadcddbadcddbacdbffacdcdbacdcbacbadcddbef
acdbabababfaccdbeefeacbfabababfabababeaddcddcbacbaccdcddbabadccdcddcdcddcdbabf
adbadccddcbeababacddbabeabababadccbacbabababacdbaddbabeacbabfeabacbfabefeabad
cbabefeababeadcbfadccbabababacbfacdcbacbeaddcbabfaccbeabadbabacdddcdcdcdcdbadc
dddcbefacbfeacbfeadccbacddcdccdcdbeababacbefeabadbeeacbabfacdbadcbadcbadbeabfac
bacdbacdddcdcbacbff.

From this sequence of strings different plateau where generated to depicts the convergence
of the system. The plateaus are discussed in the following sections.

L 1 2 3 4 5 6 7 8 9 10

0.1000 6 17 45 122 353 910 2139 4400 7665 11447
0.0656 6 17 44 129 378 1265 3057 6152 10187 14142
0.0430 6 17 43 124 379 1273 3054 6147 10188 14128
0.0282 6 17 43 124 376 1263 3052 6124 10172 14104
0.0185 6 17 43 123 374 1266 3044 6114 10164 14090
0.0122 6 16 44 122 378 1259 3030 6100 10132 14005
0.0080 6 16 42 121 402 1261 3016 6090 10129 14001
0.0052 6 16 43 120 404 1259 3012 6096 10107 13990
0.0034 6 15 43 121 403 1256 3012 6075 10089 14004
0.0023 6 15 43 117 403 1257 3011 6082 10081 13976
0.0015 6 15 40 118 404 1257 3005 6066 10069 13961
0.0010 6 15 41 119 432 1254 3001 6063 10069 13952

Figure 3: Plateau generated for 0.4ps of the zz-component, no convergence.

L 1 2 3 4 5 6 7 8 9 10

0.1000 6 35 211 957 2723 5187 9506 9603 9354 9252
0.0656 6 35 211 1143 3779 7368 9452 9577 9350 9251
0.0430 6 35 211 1142 3773 7336 9412 9580 9351 9221
0.0282 6 35 211 1144 3763 7309 9400 9576 9351 9218
0.0185 6 35 211 1145 3754 7275 9363 9540 9341 9210
0.0122 6 35 211 1145 3758 7242 9377 9515 9332 9209
0.0080 6 35 211 1145 3756 7255 9328 9511 9332 9209
0.0052 6 35 211 1146 3757 7224 9323 9510 9332 9209
0.0034 6 35 211 1144 3755 7218 9324 9508 9332 9209
0.0023 6 35 210 1144 3749 7220 9312 9507 9332 9209
0.0015 6 35 210 1146 3738 7224 9310 9507 9332 9209
0.0010 6 35 210 1145 3742 7230 9310 9507 9332 9209

Figure 4: Plateau generated for every 0.8ps of the zz-component, no convergence.


L 1 2 3 4 5 6 7 8 9 10

0.1000 6 32 204 740 1603 3169 3223 3112 3033 3032
0.0656 6 32 203 849 2163 3162 3224 3116 3033 3031
0.0430 6 32 203 846 2153 3157 3216 3100 3033 3031
0.0282 6 32 202 848 2163 3145 3224 3092 3034 3031
0.0185 6 32 201 849 2154 3131 3215 3093 3034 3031
0.0122 6 32 201 860 2153 3119 3207 3092 3032 3031
0.0080 6 32 201 860 2149 3115 3207 3092 3032 3031
0.0052 6 32 199 859 2146 3115 3209 3092 3032 3031
0.0034 6 32 201 857 2150 3111 3209 3092 3032 3031
0.0023 6 32 197 856 2136 3100 3209 3092 3032 3031
0.0015 6 32 200 859 2134 3087 3209 3092 3032 3031
0.0010 6 32 200 855 2135 3087 3209 3092 3032 3031

Figure 5: Plateau generated for every 2.4ps of the zz-component.

L 1 2 3 4 5 6 7 8 9 10

0.1000 6 32 167 586 1166 1340 1277 1218 1212 1216
0.0656 6 32 169 589 1165 1336 1274 1218 1212 1216
0.0430 6 32 167 586 1164 1333 1274 1218 1212 1216
0.0282 6 32 167 586 1148 1346 1270 1218 1212 1216
0.0185 6 32 167 594 1146 1338 1270 1218 1212 1216
0.0122 6 31 168 595 1147 1326 1268 1218 1212 1216
0.0080 6 31 168 594 1149 1326 1268 1218 1212 1216
0.0052 6 31 172 591 1140 1326 1268 1218 1212 1216
0.0034 6 31 173 591 1146 1326 1268 1218 1212 1216
0.0023 6 31 175 591 1144 1326 1268 1218 1212 1216
0.0015 6 31 175 589 1143 1326 1268 1218 1212 1216
0.0010 6 31 175 593 1142 1326 1268 1218 1212 1216

Figure 6: Plateau generated for every 6ps of the zz-component, no convergence.

Looking at the plateau from 0.4ps to 2.4ps it is clear that the system needs more data for
analysis and convergence. Though there are signs of reduction in states at different string
length (L) and significant ( ) as the time step increase from 0.4ps to 6ps. But this might
appear to be as a result of absence of much data or it has no true convergence. That is it is
purely a random noise at these time-step, and as such has no predictable pattern. Which can
only mean all the patterns read from the stochastic sequence are unique in themselves. That
is, the state of the system at each time-step is totally unique in itself.

The plateau starts showing a significant level of convergence as the time step increases to
over 11.8ps as depicted in Figure 7. There are two possible groups of causal states at different
L and , which the e-machines converges to. At L = 5 and = 0.0122 the e-machine converges
to 608 casual states; at L = 5 and = 0.0122 it converges to 608. While at L = 7 and =
0.0282 it converges to 581 casual states.

L 1 2 3 4 5 6 7 8 9 10

0.1000 6 33 150 345 611 622 583 581 573 575
0.0656 6 33 146 426 616 619 583 581 573 575
0.0430 6 32 145 418 620 611 583 581 573 575
0.0282 6 32 149 422 614 611 581 581 573 575
0.0185 6 32 149 425 610 609 581 581 573 575
0.0122 6 32 155 425 608 608 581 581 573 575
0.0080 6 32 155 428 605 608 581 581 573 575
0.0052 6 32 157 428 605 608 581 581 573 575
0.0034 6 31 155 426 604 608 581 581 573 575
0.0023 6 31 160 424 610 609 581 581 573 575
0.0015 6 31 162 422 610 609 581 581 573 575
0.0010 6 31 156 424 609 609 581 581 573 575

Figure 7: Plateau generated for every 12.4ps of the zz-component, no convergence.

Considering we have about n states of the system.

ABSTRACT
In this research, the stochastic microscopic stress tensor of liquid water is investigated. Three
methods are applied in this work. Molecular Dynamic Simulation(MD) and the Stress Tensor
Model are used to generated the required stochastic data used in the research. MD is used
to simulate the microscopic behaviour of liquid water and the generated microscopic
trajectory of the interactions of water atoms are analysed with the Stress Tensor Model at
each time-step of the atomistic trajectory. These trajectories is analysed using Computation
Mechanics by applying the Causal-State Splitting Reconstruction (CSSR) algorithm. Our result
shows that the microscopic stochastic stress tensor is not purely a random noise.

Future work.
This work can be extended in different ways. The stress tensor used to known to have errors
as state by the authors. This could affect the result of this research, though it cannot be
ascertained at this point. In addition, there is no unique local stress model, hence there is no
guarantee that our current result will be the same when compared to a different stress tensor
model.
Symbolic representation of highly noisy stochastic data is an issue. CSSR requires minimal
number of symbols, hence how can one choose the right number of symbols to use: knowing
CSSR complexity increases with the number of symbols. Therefore, instead of looking at the
data distribution, one could consider the meaning of the distribution (e.g. in our case the
tensile or compressive stresses) in order to discretised symbolically the complex time-series
data.
Also the GROMACs package used can be extend in such a way that the system will not shrink
the volume of the simulation box to the fit the total amount of substances in it.

By using the algorithm 2, the following where plateaus where generated of the zy and zz
components of the stress tensor

its clarity of the correlation in relationship to the stress tensor.

In the subsequent years of the completion of the project. The project will as outlined below:

In the left two years of the program, the following will be done.
- Intense evaluation of the stress tensor and its randomness in terms of different
volumes of the liquid water simulation. This is to ascertain the point of correlation of
the microscopic stress tensor. Two more methods will be evaluated as well at different
sizes of the simulation of box and at about 10 micro seconds of simulation time.
- A small peptide will be evaluated as well. This will give our findings more credibility in
terms of its evaluation with biomolecules like proteins.
- Finally, the write-up will be done

Summary
In this research the microscopic stress tensor was evaluated to know if it is truly a purely
random signal or not. But in order to carry out this task, a molecular dynamic simulation of
liquid water is being carried out and its trajectories where used to compute the microscopic
stress tensor data. The data produced by the stress calculation produces the nine components
of the stress signal. Two of this component where used as the data source to evaluate the
stress signal in terms of its randomness. From the computation mechanics analyses, we were
able to deduce that, the stress tensor is not purely a random signal, and as such, further
investigation will be done in this light to know at what point does the stress signal becomes
correlated in the molecular process.

I thank Professor Emmanuel for all the assistance and help towards me. In addition, I would
like to thank my supervisors Dr. Dmitry Nerukh for his advice and help towards a fruitful result
of this research. I am grateful to my Research Director of the School of Engineering and
Applied Science in Aston, Dr Generalis, Sotos C, the research director Sandra Mosley and all
the administrative staff for their patience and understanding all through this period.

This project is layout in this manner. In chapter one, we introduce the research and the
problem that was evaluated, chapter two explains the methods and procedure used in
generating. In chapter three, the method used in analysing the data is explained and in
chapter four, the results are given with explanations. Finally, in chapter five, the work is
concluded and recommendation are made and the outline of the rest of the thesis is stated.

You might also like