# Granular Experiments

###### tags: `experimental results` `SoftQC` `experimental setup` `granular`

---

---

### Owners (the only one with the permission to edit the main test)

EF, AP, FS, GF

---

---

## Background

## Plans

## To Do

### Plan for end of December 2022 - start January 2023
- Joblib to fasten tracking
- Try to measure structure factor applying a mask to avoid boundary effects (which impose specific directions)
- QC8 try state3 xs=0.65 phi=0.849
- QC8 try state2 with lower packing fraction
- S1: do fracS1 vs phi to observe jump. Four of five packing fraction between 0.83 and 0.86 three scquisition for each pack
- Try to form S3 in experiments?

### Plan for SA September/October 2022

- **Exp:** Change state point with large stup (L=20 cm). $\Delta \phi \pm$ 0.005 $x_S \pm$ 0.025. New Driving: $f=120$ $A_o=110 mV_{pp}$.Total of four experiments. Do final acquisition at high rate to study mobility. TOTAL: 4 exp runs.
- **Exp:** Change setup size (small setup L=10 cm). Try both new driving $f=120$ $A_o=110$ and old driving $f=350$ $A_o=450 mV_{pp}$ . TOTAL: 2 exp runs.
- **Exp:** Check acceleration of the plate with the accelerometer screwed in different place to check the modes.
- **Data analysis:** Do a script for local observables. qs, local mobility, local packing fraction. local concentration of small/big particles.
- **Methods:** Start use hdf files. Use graphical library to crop.

Starting from 01/06 we'll go again in the exp room. Here are some things to do before doing experiments again:
:::success
- Check sync VS. no-sync with accelerometer
We see the difference between Sync in the absorbed state and completely non sync in the active state. Nevertheless it is difficult to understand if it would be possible to distiguish the two kinds of absorbed states (fully sync and partially sync).
- Test buffering with personal laptop. Do we still have double equal frame?
With laptop no problem of acquisition. Still not clear which is the problem of the lab's computer. WE'LL BUY A NEW ONE!
:::

- Start using python to control func. gen. (Strange things with computer lab Try with laptop)
:::success
- check if for modes with salt (f=350 VS f=120)
No clear cimatic pattern but Confirmation that you have less mobility in the center.
:::

### On the setup
- Calibration with grains -> Be sure of the real amplitude of vibrations
For this issue it could be useful to use the lock-in techique. Look (https://doi.org/10.1103/PhysRevE.61.5600)
:::success
- Script that takes and saves pictures with pypylon or equivalent
pylon does it
:::
- improve a little bit the light for imaging
:::success
- how to properly fix the glass plate
:::
:::success
green is for finished tasks
:::

### On the measurements

- check for gravitational gradients

PC https://www.dell.com/fr-fr/shop/notre-gamme-pro/vostro-3710/spd/vostro-3710-desktop/n6542_qlcvdt3710emea01
### IDEAS on absorption
how to distinguish continuous and discountinuous in the experiments? Difficult to see jump in the fract active VS phi.

## EXPERIMENTAL ADSORBTION
### GIF
$\phi < \phi_c$

$\phi \sim \phi_c$

$\phi > \phi_c$

### Fraction of active particles

#### $f=53$ Hz $A_o=170$ $mV_{pp}$ **varying phi**
This analisys can be surely refined but it seems to work. I'll add some details later.

There are some artifact that I will explain. But they don't affrct the final results.

#### $f=53$ Hz $\phi=0.12$ **varying $A_o$**

## EXPERIMENTAL SELF ASSEMBLY

### Experimental debug (Feb 2023)
Really strange results during acq0602 and acq0702. We need to debug the exp.

**check calibration with pentacol**
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. Exp time 5575. Acquisition at 0.25 fps.
Before acq0702

After acq0702.

It changed but not dramatically... Considering that at the very beginning of acq0702 I silghtly adjusted by hand the horizontal calibration

Same thing at 0.1 fps for two hours done with d=2.381mm used for QC8 exp
It is coherent with what observed before.

**Check for magnetic-induced pattern (d=2.381 used for QC8 exp)**
$L=0.1$ m, $N=196$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. Exp time 5575. Acquisition at 0.1 fps for 2 hours.
We don't see any pattern

**Cleaning of beads and plate with ethanol**
After cleaning the plate I saw some black residuals on the tissue signaling aging of the anodized coating.

After on night things seem to be ok. By eye we clearly see that there is not a lot of etherogeneity. Here a (static) comparison between the two runs after 1000 minutes.

Dynamic comparison of the two exp

0702
https://www.dropbox.com/s/m842sh1a5pa8p60/acq0702_GIF.gif?dl=0
0902
https://www.dropbox.com/s/6qa1nquxq26pbks/acq0902_GIF.gif?dl=0

To sum up the 09/02 I did the following "debug" procedure:

- Unload the system from the sample

- Calibration test with dilute gas of usual beads (d=2.5 mm). Calibration was good (slighlty tilted but ok).

- Calibration test with dilute gas of beads coming from the sample (d=2.381 mm). Calibration was good (slighlty tilted but ok).

- Magnetic test with beads coming from the sample (d=2.381 mm). No magnetic pattern observed.

- Cleaning of the sample and the aluminum plate with pure ethanol. I found some residuals of black powder (probably coming from the anodized coating) on the "Kimtech" tissue I used.

- Reload the washed sample and restart the exp with the same driving with no changes in the calibration.

The result is a quite homogeneous dynamics and no formation of hexagonal cluster

### Pentacol Experiment (September 2022)
Experiment suggested by Puglisi to check if our beads feel the magnetic field of the shaker. We also use it to check for horizontal angle good or not.Low density only big beads: $N=295$, $\phi=0.036$, d=2.5 mm. Exp Time=5575. $f=120$ $(A_o=110$ $mV_{pp}$). We use anisotropies of density field to have an idea of the inclination of the setup.
#### Density field after calibration with the levels
This is the number density map mediated over the entire run (2 hours)
This is the result of superimposing snapshots

Unfortunately also after many calibration runs the anisotropies are still there (see below). I would like to find a more efficient/faster protocol to improve the horizontal calibration after the one done with the levels.

This is the number density map mediated over the entire run (1 hours). Obtained after 6 experiments with calibration in bwtween.
This is the result of superimposing snapshots

This is a comparison between the avg density field of the first row (the most asymmetric one) before and after 6 calibrations

### Pentacol Experiment (October 2022, small system)
Here I used the reducers (L=0.1) so the available area is divided by 4. This makes the exploration of of the phase space faster. I also considered a slighly more dense system $N=185$, $\phi=?$. I runned a 3 hours exp after the calibration with the levels. It resulted to be unbalanced but I checked that such unbalances are stationary with time (i.e. averaging over the 1st,2nd and 3rd hour I obtain the same result).

Here I plot the density field on the main diagonal (from bottom left to upper right) averaged on different 1 hour time interval during the experiment. We clearly have stationarity.
I do the same thing on 20 minutes intervals at the start, in the middle and at the end of the exp. We see that 20 min are enough make the signal emerging from fluctuations. This makes the calibration procedure much more faster with respect the larger system with a more dilute granular fluid!

After two calibrations by hand and analysing the density field after 20 min averages I finally obtained this resul which has a muche better symmetry.

The darker region is not perfectly centered but I''l start the experiment in this condition: with the current screws the probability of producing a worst calibration is very high.
I also plot the density field of the diagonal and column 4 which are the most unbalanced to check the the relative change is not that much ($\sim$ 13%).

### Pentacol Experiment (Novembre 29th 2022, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$. Acquisition at 0.25 fps for 20 minutes for each density map.
After calibration with levels:


After 6 adjustments:


However, when I loaded the system with state 2 I clearly observed non homogeneous mobility. The following is calculated over 230 frames with lowrate acquisition. It is representative of the non-homo you can check by eye.

before adj (LOWRATE):
There is clearly higher mobility in the bottom left corner

After trying adjustments b eye I obtained the following mobility map averaged on the same number of frame (colors are comparable).

after adj (LOWRATE):
It is less unbalanced but still not really homogeneous. Looking the video at high spees one can actually check that the top right corner is more mobile.

I also did the high rate mobility before and after the adjustment they show almost the same things even if the mob map before adjustment is unbalanced in a slightly different way (but it is ok because it feels more fluctuations).

before adj (HIGHRATE):

after adj (HIGHRATE):

FINALLY AFTER 2 MORE ADJUSTMENTS IT SEEMS TO BE VERY HOMOGENEUS
after 3 adj (LOWRATE)

### Pentacol Experiment (December 5th 2022, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$. Acquisition at 0.25 fps
before adjustments

after 4 adjustments

### Pentacol Experiment (Feb 06 2023, small system) WITH CEILING!
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. Exo time 5575. Acquisition at 0.25 fps.
before adjustments

after 5 adjustments

This calibration was not good! The only difference with respect previous time was the presence of the ceiling. It probably induces some effects because particles sometimes hit the ceiling at that driving... **DON'T PUT CEILING DURING CALIBRATION**

### Pentacol Experiment (Feb 07 2023, small system) WITHOUT CEILING!
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
before adjustments

after 4 adjustments

### Pentacol Experiment (Feb 13 2023, small system) WITHOUT CEILING!
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. Exp time 5575. Acquisition at 0.25 fps.
before adjustments

after 4 adjustments

### Pentacol Experiment (Feb 20 2023, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.

before adjustments before cleaning (right after a one-week long experiments, witouth cleaning the plate)
Kimtech after cleaning

before adjustements after cleaning
after 3 adjustments

### Pentacol Experiment (Mar 1 2023, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
After the qc12 experiments (20 and 21 of Feb) the plate was very dirty (Kimtech)
before adjustments after cleaning

after 5 adjustments


### Pentacol Experiment (Mar 13 2023, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
Cleaning was ok, not too much black residual.
before adjustments after cleaning

after 5 adjustments

### Pentacol Experiment (Mar 21 2023, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
Cleaning was ok, not too much black residual.
before adjustments after cleaning
after 5 adjustments

### Pentacol Experiment (May 2 2023, small system)
$L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
Cleaning was ok, not too much black residual.
before adjustments after cleaning

after 7 adjustments

### Pentacol Experiment (May 11 2023, small system)
Valid for acq1105 and acq1505. $L=0.1$ m, $N=187$, $f=120$ Hz, $A_o=110$ $mV_{pp}$ SINU. ExP time 5575. Acquisition at 0.25 fps.
Cleaning was ok, not too much black residual.
before adjustments after cleaning

after 1 adjustments

### QC12

| $x_S$ $\downarrow$ --- $\phi$ $\rightarrow$ | 0.85 |
| ------------------------------------------- | ---- |
| 0.4 | State1 |

### QC12 State1 RAMP SMALL: $\phi=0.85(?) \pm $, $x_S=0.4 (?) \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 20/20/22)
Start conf

Final conf

Boops

Really bad: almost Nothing is evolving, more 12-fold symm at the beginning than after. We can improve changing the driving and checking for the correct target composition. The sample was made with an unusual scale.

### First attempts QC8 from December 2022

#### Sample

We first try with the best state point we found in lammps simulations namely: $q=0.5$ $\phi=0.85$ $x_S=0.7$. This is realized with small beads $\sigma_S=1.2$ mm and large beads $\sigma_L=2.381$ mm.

### QC8 State1 RAMP SMALL (short exp): $\phi=0.848 \pm $, $x_S=0.699 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 05/12/22)

Start Conf

Final Conf

Structure factor after $\sim 500$ frames

Mobility Map mediated over 500 frames

### QC8 State1 SINU SMALL (short exp): $\phi=0.848 \pm $, $x_S=0.699 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 12/12/22)

Start Conf

Final Conf

Structure factor after $\sim 500$ frames

Mobility Map mediated over 500 frames

### Official QC8 runs. RAMP

| $x_S$ $\downarrow$ --- $\phi$ $\rightarrow$ | 0.848 |
| ------------------------------------------- | ---- |
| 0.675 | State2 |
| 0.699 | State1 |

### QC8 State2 RAMP SMALL (long exp): $\phi=0.848 \pm $, $x_S=0.699 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 13/12/22)

Start Conf

Final Conf

Comment: threshold per bond 34

Structure factor GIF

Structure factor start
Structure factor after 2100 (17.5 hours) frames -> Best peaks
Structure factor after 4767 (39 hours) frames

COMMENT on structure factor: Emerging 8 fold simmetry. There is maximum brightness of peaks around 2100 frames then they are more blurred untile the last framed where they are bright again. Probably this comes from multiple qc8 domain.

First attempt tile analysis

Mobility Map mediated over half of the experiment
Is a bit unbalanced

### QC8 State2 RAMP SMALL (long exp): $\phi=0.848 \pm $, $x_S=0.675 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 14/12/22)

Start Conf

Final Conf

GIF structure factor
Start
Final

First attempt tile analysis
FRAME 6937

**CommentI*:** there are a lot of "wrong" tiles like equilateral triangles or pseudo squares (half small half large squares). Moreover, there are some technical issues when using this script on experimental data (see technical paragraph at the end of the page). However, these results should be quite representative of what's going on. The nice thing of this realization is that we have a sinlge qc8 withing the tolerance of Etienne's script. Indeed, the orientation of the tiles in different region of the sample is compatible with a unique qc8.

**CommentII**: During the discussion of 11/01 we realized that some tiles have not the "dominating" orientational order imposed by the boundaries see here a patch of non aligned triangles.
At the same time the squares (big and small) seems to be fairly aligned and one can indeed check that traingles near the square are consequently with the right orientational order. THM could be isolated patches of traingles spontaneusly nucleates with any orientation but "complete" patches (squares and triangle are correctly aligned).

Here also the gif tho check evolution of intermediate (i.e. wrong) bonds
Wrong blue peaks seem to fluctuate around a costant value while good ones are raising... could be a kind fuel?

#### Refined analysis with Good and ambiguous tiling as a function of time

We now classify better the amibguous triangles. We impose that if a tracked triangle does not match (with a given threshold) with the overall orientational order then it is classified as ambiguous. For triangles this method works better (i.e. it is more strict) than the original method based on just "vote".

**Auto-imposed overall order**
FRAME 6900
Yellow tiles are the ambiguous ones

This is the evolution of the number of aligned and ambiguous tiles
And the aligned over aligned+ambiguous
The total number aligned+ambiguous

From this analysis I would say that the QC is still growing (very slowly) at the end of the experiment. The number of good tiles grows much faster than the number of ambiguous one. We can immagine different processes at play: liquid <-> amb. tile,liquid <-> align. tile,amb. tile <-> align. tile. From the figures we see that the overall number of tiles is growing but sometime we see some bumps: this could be an evidence of the fact that some part of the crystal need to melt to be reorganized and contribute again to the crystal.

**Overall order imposed by hand (long bonds aligned with hard boundaries)**
FRAME 6900


These curves are cleaner because the overall orientation is fixed (i.e. it does not fluctuate as a function of the system's state). We can draw the same conclusions but here we have a more evident change of regime around frame 2000. There is a rapid flow from aligned tiles to ambiguous. INTUITIVE GUESS: QC can grow independently from the boundaries but after a certain length it must be reorganized to unite the four sub QCs.

### QC8 State3 RAMP SMALL (long exp): $\phi=0.848 \pm $, $x_S=0.65 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 09/01/23)
Not well calibrated

Start Conf
Final Conf

Structure Factor
start
final

Tiles

over time

### QC8 State2 RAMP SMALL (strange exp): $\phi=0.848 \pm $, $x_S=0.675 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 07/02/23)

Start Conf (Snap 1)
Snap 2507
over time

**IT DID CRAZY THINGS DURING THE NIGHT LIKE CONVECTIVE MOTIONS**
https://www.dropbox.com/s/m842sh1a5pa8p60/acq0902_GIF.gif?dl=0

### QC8 State2 RAMP SMALL (after cleaning): $\phi=0.848 \pm $, $x_S=0.675 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 13/02/23)
Start Conf.

Final Conf.

Over time
Structure Factor smoothed over the last 100 snapshots

Tiles
Very Long

### QC8 State3 RAMP SMALL: $\phi=0.848 \pm $, $x_S=0.65 \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 01/03/23)
Start Conf

Final Conf ($\sim$ 26000 frames)

over time

### QC8 State2 SINU SMALL: $\phi=0.848 \pm $, $x_S=0.675 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 13/03/23)
Start Conf

Final Conf

Over Time

### QC8 State3 SINU SMALL: $\phi=0.848 \pm $, $x_S=0.65 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 21/03/23)
Start Conf
Final Conf

Over Time

### QC8 State4 SINU SMALL: $\phi=0.843 \pm $, $x_S=0.675 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 03/05/23)

Final Conf

Over Time

### QC8 State5 SINU SMALL: $\phi=0.843 \pm $, $x_S=0.65 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 11/05/23)
Acquisition stopped during the exp. We miss 1.5 days of data days

Final Conf

Over Time

### QC8 State5 SINU SMALL: $\phi=0.843 \pm $, $x_S=0.65 \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 15/05/23)
Restarted same state point of 11/05
Final Conf

Over Time

### Comparison between long experiments QC8

### Experiments for S1 formation (October 2022)
We tried different state points ($\phi$, $x_S$) and driving. For each experiment we show initial and final configuration, evolution of $q_4$ and $q_6$, maps of different observables (mobility, $q_4$, $q_6$, Number of small beads, $x_S$, $\phi$). The maps are averaged on 50 frames in three temporal region of the experiment: start, middle final.
In the best experiments we also perform the analysis of S1 clusters.

| $x_S$ $\downarrow$ --- $\phi$ $\rightarrow$ | 0.85 | 0.855 | 0.86 |
| ------------------------------------------- | ---- | ----- | ---- |
| 0.496 | State1 | State2 | |
| 0.524 | State4 | State3 | |
| 0.55 | | State5 | State6 |

### S1 State 1 bad dr.: $\phi=0.8490\pm0.0003$, $x_S=0.4970\pm0.0005$. Driving: $f=120$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 03/10/22 - Shorter than the others)
Start Conf
Final Conf (139 minutes)
BOOPS
Maps

### S1 State 2 bad dr.: $\phi=0.8529\pm0.0008$, $x_S=0.495\pm0.001$. Driving: $f=120$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 04/10/22)
Start Conf
Final Conf
BOOPS
Maps

### S1 State 3 bad dr.: $\phi=0.852\pm0.001$, $x_S=0.520\pm0.002$. Driving: $f=120$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 05/10/22)
Start Conf
Final Conf
BOOPS
Maps

### S1 State 4 bad dr.: $\phi=0.8496\pm0.0001$, $x_S=0.5240\pm0.0003$. Driving: $f=120$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 10/10/22)
Start Conf
Final Conf
BOOPS
Maps

### S1 State 4 good dr.: $\phi=0.8496\pm0.0001$, $x_S=0.5240\pm0.0003$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 07/10/22)
Start Conf
Final Conf
BOOPS
Maps
Clusters: number and average size of clusters made of particles with $q_4>0.8$ with more than 10 elements.
Animation of cluster formation: gif with only $q_4>0.8$ particles showed.

### S1 State 3 good dr: $\phi=0.8545\pm0.0001$, $x_S=0.5240\pm0.0002$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 14/10/22)
Start Conf
Final Conf
Maps
BOOPS
Clusters: number and average size of clusters made of particles with $q_4>0.8$ with more than 10 elements.
Animation of cluster formation: gif with only $q_4>0.8$ particles showed.

### S1 State 2 good dr: $\phi=0.8546\pm0.0001$, $x_S=0.4955\pm0.0003$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 17/10/22)
Start Conf
Final Conf
Maps
BOOPS
Clusters: number and average size of clusters made of particles with $q_4>0.8$ with more than 10 elements.
Animation of cluster formation: gif with only $q_4>0.8$ particles showed.

### S1 State 1 good dr: $\phi=0.8491\pm0.0002$, $x_S=0.4956\pm0.0003$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 18/10/22)
Start Conf
Final Conf
Maps
BOOPS

At frame 435 (half of the exp) we see qc12 blurred peaks

### S1 State 5 good dr: $\phi=0.8543\pm0.0002$, $x_S=0.5516\pm0.0003$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 20/10/22)
Start Conf
Final Conf
Maps
BOOPS
Clusters: number and average size of clusters made of particles with $q_4>0.8$ with more than 10 elements.

### S1 State 6 good dr: $\phi=0.8593\pm0.0002$, $x_S=0.5516\pm0.0004$. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 21/10/22 - Short because of electricity cut)
Start Conf
Final Conf
Maps
BOOPS
Clusters: number and average size of clusters made of particles with $q_4>0.8$ with more than 10 elements.

### Some conclusions (Large system)

Up to 18/10, the best condition for SA is State3 (\phi=0.8545\pm0.0001$, $x_S=0.5240\pm0.0002$) which is the bottom right corner of 4-point phase diagram that we have.
Nice to look at q4 and q6 all togheter
We realize that up tp 21/10 the best conditions in terms of q4 are the ones with even more higher concentration of small particles ($x_s=0.55$). Actually I have to check if it is an artifact because looking at final configurations I would say that state3 and stat4 are better than state5 and state6.

Now we compare the **FRACTION OF PARTICLES BELONGING TO S1 ENVIROMENT** (wrong xlabel). I compute it as fraction of particles that have at least

This observable gives more the feeling of what you see in the final configurations. We see that state 3 is the best.

### S1 State 3 good dr SMALL: $\phi= \pm $, $x_S= \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 27/10/22)

STATE POINT acq2710dumpLammps_SmallBig_WithMask.csv
Nsmallmean=1507.9984214680346 Nsmallstd=34.925741461112516
Nbigmean=1389.4483030781373 Nbigstd=0.6427077754497112
xSmean=0.5203866341494805 xSstd=0.005912795825025821
phimean=0.8525944443086172 phistd=0.003919704425629896

Starting configuration
After $\sim$ 2 hrs and 30 min
Final Configuration (after $\sim$ 12 hrs)
BOOPS
Average number of clusters
Average size of the 10 biggest clusters. We see a high peak aroun 12 hours.!
Fraction of particles belonging to S1 env (Still Naive method)

MAPS
(Look real conf: python3 doFractionOfS1part_refined_livePlot.py boops_acq2710_dumpLammps_THR30.csv 30)

### S1 State 3 good dr SMALL (Long exp): $\phi= \pm $, $x_S= \pm $. Driving: $f=120$ $Hz$, $A_o=95$ $mV_{pp}$. (acq 03/11/22)

STATE POINT acq0311dumpLammps_SmallBig_WithMask.csv
Nsmallmean=1532.732894255397 Nsmallstd=3.3289586811248637
Nbigmean=1389.286864251738 Nbigstd=0.5291013121795197
xSmean=0.5245450929836672 xSstd=0.000593089047422409
phimean=0.8553126011779912 phistd=0.0003412185544715966

Very long acquisition performed over the night.
Starting configuration
After $\sim 10$ hrs (Big s1 cluster)
After $\sim 12$ hrs (Big s1 cluster)
After $\sim 22$ hrs (Hexagonals again!)

BOOPS
This seems an unlucky case cause the q4 start decreasing earlier with respect the other day.
Average number of clusters
Average size of the 10 biggest clusters. We see a high peak aroun 12 hours.
Fraction of particles belonging to S1 env (Still Naive method)
(Look real conf: python3 doFractionOfS1part_refined_livePlot.py boops_acq0311_dumpLammps_THR30.csv 30)
MAPS

### S1 State3 RAMP SMALL (Long exp): $\phi= \pm $, $x_S= \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 09/11/22)

STATE POINT acq0911dumpLammps_SmallBig_WithMask.csv
Nsmallmean=1535.1564267602887 Nsmallstd=5.632364640969984
Nbigmean=1389.272088940901 Nbigstd=0.5261701880495238
xSmean=0.5249406312855577 xSstd=0.0009383897327548833
phimean=0.8555794434331159 phistd=0.0006173957537805303

Very long acquisition performed for more than 40 hours with ramp driving. This kind of driving exhibits a more homogeneous mobility perhaps it reduces the effect of the vibrational modes.
COMPARISON OF MOBILITY MAP DONE WITH HIGH ACQUISITION (I still have to upload it)

Starting configuration

Final configuration

BOOPS

Average number of clusters

Average size of the 10 biggest clusters.

Fraction of particles belonging to S1 env (Still Naive method)
**This observable has a growing trend!**
(Look real conf: python3 doFractionOfS1part_refined_livePlot.py boops_acq0911_dumpLammps_THR30.csv 30)

In addition to this growing trend, the good thing of this experiment is that we really don't see hexagonal growing for more than 40 hours. Then, the cluster contributing to S1 are the same for the last 20 hours. I will quantify this the next days. In other words, contrary to what we see with sinusoidal shaking, here we see smaller cluster but more persistent in time.

MAPS

### S1 State3 RAMP SMALL (Long exp): $\phi= \pm $, $x_S= \pm $. Driving: $f=110$ $Hz$, $A_o=110$ $mV_{pp}$. (acq 15/11/22 - reproducibility)

Very long acquisition performed for more than 40 hours with ramp driving. It is a kind of reproducibility test for the 09/11 experiment. We don't have stop and go experimental run in this case.

Starting configuration

Final configuration

BOOPS

Average number of clusters

Average size of the biggest clusters.
probably the trend
Fraction of particles belonging to S1 env (Still Naive method)
(Look real conf: python3 doFractionOfS1part_refined_livePlot.py boops_acq1511_dumpLammps_THR30.csv 30)

MAPS
Here we have inhomogeneities of mobility. Probably after many hours of experiments the system moves a bit or the floor get deform. We should do calibration every X hours of experiments where X has to be dermined.

### S1 - Before Summer 2022
If no specified we assume in the following: $f=350$ $A=9\mu m$ $(A_o=450$ $mV_{pp})$.
If no specified we assume in the following $\phi=0.85$, $x_L=0.5$ ($N_s=N_L=5630$). However, I still have to do some checks on the accuracy of this estimate.

#### Pentacol Experiment (spring)
Experiment suggested by Puglisi to check if our beads feel the magnetic field of the shaker. $N=1120$, $\phi=0.137$. I adjusted by eye the horizontal angle waiting for a spatially homogeneous state. No notes about this exp n the logbook but checking some files in the laptop I could say that it has been done around the 10th of May at 0.2 fps for 1 hous (720 frames total)

##### Snapshot
Single frame
Note for tracking: Very good at low density, it recognize also the faster blurred grains.

Superimposed frames
With Magnetic beads the Puglisi and Gnoli found strange patters with this check. In our case it seems to be ok, with just a bit of inhomogeneity.

#### acq0605_HigherCeil_HomoStart_Fr120_A120
Experimental run that starts after an homogeneisation of the sample performed with a low freq/high amp shaking. Tracking in cropped images (I ignore particles near the boundaries that are covered by shadow). Tracking method FOURTH WAY see below.
Here $f=120$ and $A_o=120$ $mV_{pp}$. I did this short experiment to see if playing with the driving we can obtain a higher mobility in the center.

> [name=gfoffi]Perhaps this is a good candidate for the dynamical transition using the cell??

##### GIF

It seems that this driving allows more mobility in the bulk. However we have the same Hex growing at the boundary. Moreover, $f=120$ Hz is transmitted more from our room to the neighbours. Anyway it is fine to see that playing with the driving can do something.

Note for tracking: I used the same sample of acq0505 and they work for this tracking.

#### acq0505_HigherCeil_HomoStart
Experimental run that starts after an homogeneisation of the sample performed with a low freq/high amp shaking. Tracking in cropped images (I ignore particles near the boundaries that are covered by shadow). Tracking method FOURTH WAY see below.
##### Snapshot comparison
Last Image
Last Sanpshot
##### GIF
Positions

Bonds

From these videos is even more clear that the center is freezed and that the left side is hotter than the right one.

Positions Big and small particles
Sometime few small particles disappear especially when they pile up.
##### BOOP

Here we see that on large timescales HexL is favourite but at short time we see that q4 is growing.
zoom at short times. Here is clear that for the first five minuted of the experiment q4 is the most favourite boop.
#### Packing fraction
Packing fraction calculated in the cropped area of the images.
I'm confident on the value of the big particles while i loose some of the small in the tracking. So this is a lower bound of the real local packing fraction. We are really near to the target value 0.85 (considering also the missing small particles). In any case sooner or later I will take pictures of small and big alone to be sure of the extimate.

## TILES ANALYSIS (technical part)

The Etienne's code works well on experimental data but it can be improved. For some reason the right threshold to correctly detect both short and long bonds is not 1.7\sigma as for EDMD but we need >= 2.\sigma. With this threshold there is a problem in detecting small squares: since they are slightly deformed, a single diagonal bond is detected but is not removed because it is not a crossing bond


Increasing the threshold to 2.4, also the second diagonal is detected so both of them are removed since they are crossing bonds. However, with this larger threshold another problem is occurring: there are some wrong triangles which are confused with right ones.
Here you can see a right-angle triangle which is detected as an isosceles one. Question: why the code is not removing these cross bonds (CHEK)?


I would like to work with a lower threshold so I'll try to fix this by imposing a more severe criterion to remove single diagonals of small triangles.

Otherwise I can work with a larger threshold remove only the longer left bonds

**PARTIALLY SOLVED** with better elimination of crossed bonds: https://stackoverflow.com/questions/563198/how-do-you-detect-where-two-line-segments-intersect

## TRACKING
### Preliminary attempts
* Conda enviroments + trackpy and related modules
* A Very nice module for image manipulation is skimage (Francois suggestion)
* Tre ways of tracking an image with a dense configuration. Always starting from che cross correlation map between a sample and an image.

SAMPLE
FIRST WAY: locate features with tp.locate in the correlation map
Very rare fake features but miss particles near walls

SECOND WAY: use intensity of correlation map to identify features (Francois Script)
Very rare fake features, find all large particels

THIRD WAY: Treshold the image before doing the cc map then use intensity of correlation map to identify features
Very rare fake features, find all large particels

FOURTH WAY:
Big particles $\to$ use intensity of correlation map to identify features (Francois Script)
Small particles $\to$ Treshold the image and erase big particles before doing the cc map then use intensity of correlation map to identify features
Find all big particles (except the ones divided by the cropping), miss some small ones when they pile up.
Tested on acquisition_0505_HigherCeil_HomoStart, it gives the a value of packing fraction near the one expected. See above.

### Tracking for qc8 state point
UPDATE!

* Comment: the second and the third ways are clearly more efficient. The method of tresholding can be useful in future to track also the small particles.

* next steps on tracking: reconstruct trajectories with tp.link

## DIARY
### From 02/05 to 06/05
Many experimental news (more bad than good).
* Our beads seems to be magnetic. Fred is discussing with the builders how to overcome this limitations. They said that they sell us steel 316 but it is actually 304. In Rome they use 316. Differences discussed here: https://www.greenwoodmagnetics.com/resource/what-is-the-difference-between-304-and-316-stainless-steel/
* The new bubble spirit level is extremely precise and very difficult to calibrate. Probably is too precise for the precision of the calibration that is currently allowed by screwing the feet.
* Electronic level works fine but after using that you need to use dilute vibrated grains for the final calibration.
* Adsorbtion exp doesn't work. Particels are stuck with the lower glass. Try with the higher ones in very dilute conditions.
* Many experiments performed for SA
* Pentacol experiment performed. It has been suggested by A. Puglisi to chech if the magnetic field of the shaker influences the particle motion. To be analized.

### 19/04/2022
* Chat with Andrea Puglisi (first) and Andrea Gnoli (after). Regarding horizontal calibration they both said that observing the homogeneity of the granular fluid is the final test to check the correct orientation of the plate. A spirit level with high accuracy helps a lot because it allows to start the final calibratio by eyes from an almost perfetly horizontal condition. IMPORTANT: place the level at the center of the plate because its weight can tilt the plate. Gnoli explained how to calibrate the level. Gnoli said that a direct measure of the displacement can be done with optical tools (laser on vibrating surcae + detector that measure difference of light path). Gnoli suggests some experiment to measure the coefficient of restitution of the grains, we don't need now to do that but can be useful in future. Gnoli said that in future we may need to centralize the control/acquisition of the experiments with a card. For what we want now from the experiments it is still not a priority. For them it was because they have a motor to drive the intruder and an angular encoder that need to be interfaced.
* I have some small samples of glass and plastic beads to do some tests.

### 14/04/2022
* Experiment with salt revealed the presence of a gravitational gradient that cannot be measured with the little bubble we have. Buy a new and better one

* The horizontal mobility of the big particle is extremely hindered by the glass plate. It is trapped in a fast and "stochastically unperturbable" series of vertical bounces. Once removed the glass plate big grains move horizonontally fair better. We checked that with 6mm of free veritcal space, the horizontal motion is good too (now the chamber height is 3mm). We need to obtain higher boundaries for the plate. Etienne past simulations show that without roof (and consequently with an higher one) self assembly may still occurs. These conditions are just more delicate because a small particel can jump over a big one to form a galilean cannon. Will our glass plate survive to this dramatic shock? Stay tuned.
* We did many news acquisitions (ssh -X user@129.175.83.43:/home/user/Desktop/acquisition1404 in the dilute regime. Start playing with trackpy and that.

### 08/04/2022
* Found the pylon command for the automatic acquisition: https://docs.baslerweb.com/recording.
* Experimental run in the morning 4 hours. One frame every 10 minutes. Find the results connecting to ssh -X user@129.175.83.43. The frames are in /home/user/Desktop/S1_EXP/acquisition_20220408a/ you can visualize them with gthumb. Looking those images my feeling is that the S1 islands grows and shrinks alternatively.
* Look for a way to control the signal generator from the PC and in turn also remotely. It seems to be feasible:
- https://github.com/python-ivi/python-usbtmc
- https://stackoverflow.com/questions/41810965/load-and-read-signal-into-a-wave-generator-using-python-and-usbtmc-in-linux
- https://gist.github.com/pklaus/2597049
* Signature of gravitational gradient: at very low amplitude one can see that just the particles on a particular side of the box start vibrating. In any case we have to remembre to put the bubble not on the glass but on the aluminium plate when it is unloaded.
* Unloaded the setup. It takes a lot of time and the risk of loosing small particle is high. Think about a smart way to do it
* Rubber on the side of the aluminium plate to have a more homogeneous pressure on the glass plate when you screw it. Now we can really screw at maximum
* Recalibration of the setup with the glass plate on it. Found the equivalent amplitude to have 9 micrometers of displacement when the glass is on. Unforntunately the signal on the oscilloscope is not a perfect sinusoidal but is better then the last week. Nevertheless it seems that there are modes related to the glass plate. Maybe we can put some sand on it to see their effect.

### OPEN QUESTIONS

* Do we see S1 growing?
*