Title: Insights into CO2/N2 Selectivity in Porous Carbons from Deep Learning
Authors: Song Wang, Zihao Zhang, Sheng Dai and De-en Jiang
Fully formatted reference
[1] .Wang S, Zhang Z, Dai S, et al. Insights into CO2/N2 Selectivity in Porous Carbons from Deep Learning[J]. ACS Materials Letters, 2019, 1(5): 558-563.
Keywords: Deep Learning, CO2/N2 selectivity, Porous Carbon, CO2 capture
Brief summary of the reference
• Aims/objectives/background
含N2的CO2吸附很重要。虽然porous carbon 的CO2吸附模型已经可以用函数表示:M=0.095+2.10Vmicro+3.51Vmicro*Vmeso
但是含N2的CO2 选择性经常很难测量准确。深度学习又是一个很有用的工具,所以就用深度学习去做这件事情。(废话说完)
• Data used
作者从文献中报导的纯碳的多孔碳中总计收集了1138个CO2 uptake的数据, 314个N2 uptake的数据。使用BET比表面积、微孔体积、介孔体积、温度、压力作为输入张量。 Gas uptake作为输出张量。 输入和输出张量都归一化0-1。
1138个CO2 uptake数据来源(6篇文献):
[25] Zhang, Z. H.; Schott, J. A.; Liu, M. M.; Chen, H.; Lu, X. Y.; Sumpter, B. G.; Fu, J.; Dai, S. Prediction of Carbon Dioxide
Adsorption Via Deep Learning. Angew. Chem., Int. Ed. 2019, 58, 259−263
[34]Aleghafouri, A.; Mohsen-Nia, M.; Mohajeri, A.; Mahdyarfar,M.; Asghari, M. Micropore Size Analysis of Activated Carbons Using Nitrogen, Carbon Dioxide and Methane Adsorption Isotherms:Experimental and Theoretical Studies. Adsorpt. Sci. Technol. 2012,
30, 307−316.
[35] Zhuo, H.; Hu, Y. J.; Tong, X.; Zhong, L. X.; Peng, X. W.; Sun,R. C. Sustainable Hierarchical Porous Carbon Aerogel from Cellulose for High-Performance Supercapacitor and CO2 Capture. Ind. Crops Prod. 2016, 87, 229−235
[36] Xu, C.; Ruan, C. Q.; Li, Y. X.; Lindh, J.; Stromme, M. High Performance Activated Carbons Synthesized from Nanocellulose for
CO2 Capture and Extremely Selective Removal of Volatile Organic Compounds. Adv. Sustain. Syst. 2018, 2, 1700147
[37] Chang, B.; Sun, L.; Shi, W.; Zhang, S.; Yang, B. Cost-Efficient Strategy for Sustainable Cross-Linked Microporous Carbon Bead with Satisfactory CO2 Capture Capacity. ACS Omega 2018, 3, 5563−5573
[38] Heo, Y.-J.; Park, S.-J. A Role of Steam Activation on CO2 Capture and Separation of Narrow Microporous Carbons Produced
from Cellulose Fibers. Energy 2015, 91, 142−150
314个N2 uptake 数据
[12] Hao, G. P.; Li, W. C.; Qian, D.; Lu, A. H. Rapid Synthesis of Nitrogen-Doped Porous Carbon Monolith for CO2 Capture. Adv.
Mater. 2010, 22, 853−857
[24] Dura, G.; Budarin, V. L.; Castro-Osma, J. A.; Shuttleworth, P. S.; Quek, S. C. Z.; Clark, J. H.; North, M. Importance of MicroporeMesopore Interfaces in Carbon Dioxide Capture by Carbon-Based Materials. Angew. Chem., Int. Ed. 2016, 55, 9173−9177
[37] Chang, B.; Sun, L.; Shi, W.; Zhang, S.; Yang, B. Cost-Efficient Strategy for Sustainable Cross-Linked Microporous Carbon Bead with Satisfactory CO2 Capture Capacity. ACS Omega 2018, 3, 5563−5573.
[38] Heo, Y.-J.; Park, S.-J. A Role of Steam Activation on CO2 Capture and Separation of Narrow Microporous Carbons Produced
from Cellulose Fibers. Energy 2015, 91, 142−150
[39] Hong, S.-M.; Choi, S. W.; Kim, S. H.; Lee, K. B. Porous Carbon Based on Polyvinylidene Fluoride: Enhancement of CO2 Adsorption by Physical Activation. Carbon 2016, 99, 354−360
[40] Hao, G.-P.; Li, W.-C.; Qian, D.; Wang, G.-H.; Zhang, W.-P.;Zhang, T.; Wang, A.-Q.; Schüth, F.; Bongard, H.-J.; Lu, A.-H.
Structurally Designed Synthesis of Mechanically Stable Poly-(benzoxazine-co-resol)-Based Porous Carbon Monoliths and Their
Application as High-Performance CO2 Capture Sorbents. J. Am. Chem. Soc. 2011, 133, 11378−11388.
[41] Balahmar, N.; Al-Jumialy, A. S.; Mokaya, R. Biomass to Porous Carbon in One Step: Directly Activated Biomass for High
Performance CO2 Storage. J. Mater. Chem. A 2017, 5, 12330−12339.
[42] Mahurin, S. M.; Gorka, J.; Nelson, K. M.; Mayes, R. T.; Dai, S. ́ Enhanced CO2/N2 Selectivity in Amidoxime-Modified Porous
Carbon. Carbon 2014, 67, 457−464.
[43]Yuan, B.; Wu, X.; Chen, Y.; Huang, J.; Luo, H.; Deng, S. Adsorption of CO2, CH4, and N2 on Ordered Mesoporous Carbon: Approach for Greenhouse Gases Capture and Biogas Upgrading. Environ. Sci. Technol. 2013, 47, 5474−5480.
[44] Sun, N.; Sun, C.; Liu, J.; Liu, H.; Snape, C. E.; Li, K.; Wei, W.; Sun, Y. Surface-Modified Spherical Activated Carbon Materials for
Pre-Combustion Carbon Dioxide Capture. RSC Adv. 2015, 5, 33681− 33690.
[45] Ghosh, S.; Sarathi, R.; Ramaprabhu, S. Magnesium Oxide Modified Nitrogen-Doped Porous Carbon Composite as an Efficient
Candidate for High Pressure Carbon Dioxide Capture and Methane Storage. J. Colloid Interface Sci. 2019, 539, 245−256.
• Methodology/Techniques used
本文使用的深度学习框架为DNN。该DNN包括一层输入层、两个隐藏层和一个输出层。其中输入层包括ET比表面积、微孔体积、介孔体积、温度、压力5个输入神经元,每一隐藏层包含7个神经元,输出层1个张量。
• Results/discussions
略
• Likely relevance to my project
- 从文献中收集数据对于我目前的研究十分有用。如果是做实验去获得数据的话,根本不现实。对PDF格式的文献进行数据挖掘,通过文本和图片提取自己想要的数据,如果能够做这一步就可以做很多事情,非常有利于后续的研究。可以学习数据挖掘,同时要注意copyright的提交,避免侵权问题。
- Zhang, Z. H.; Schott, J. A.; Liu, M. M.; Chen, H.; Lu, X. Y.; Sumpter, B. G.; Fu, J.; Dai, S. Prediction of Carbon Dioxide
Adsorption Via Deep Learning. Angew. Chem., Int. Ed. 2019, 58, 259−263, 从这篇文献中可以学习到化学材料领域中的机器学习数据集的制作方法和数据来源问题。
import pdfplumber
import pandas as pd
with pdfplumber.open(r'C:\Users\ydf\Desktop\Insights into CO2 N2 Selectivity in Porous.pdf') as pdf:
first_page = pdf.pages[0]
#获取文本,直接得到字符串,包括换行符
print(first_page.extract_text())
#获取本页全部表格,也可以使用extract_table()获得单个表格
p0 = pdf.pages[0]
for table in p0.extract_tables():
#得到的table是嵌套list类型,转化为DataFrame更加方便查看和分析
df = pd.DataFrame(table[1:], columns=table[0])
print(df)
结果:
L
CiteThis:ACSMaterialsLett.2019,1,558−563 www.acsmaterialslett.org
e
t
Insights into CO /N Selectivity in Porous t
2 2
e
Carbons from Deep Learning
r
Song Wang,† Zihao Zhang,‡ Sheng Dai,‡,§ and De-en Jiang*,†
†
Department of Chemistry, University of California, Riverside, California 92521, United States
‡
Department of Chemistry, The University of Tennessee, Knoxville, Tennessee 37996, United States
§
Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States
C).
Ts.
17 (Uarticle
4:d
2e
9 at 00:publish
11, 201y share
ber atel
mm
O on Noveow to legiti
Ch
Sn
NCIns o ABSTRACT: Porous carbons are an important class of porous material for carbon capture. The textural properties of
Ao porouscarbonsgreatlyinfluencetheirCO adsorptioncapacities.Butitisstillunclearwhatfeaturesaremostconductive
A SAN FRnes for opti tutohpetanackhueisseevidnintpghoheriogmuhosCdceaOlr2bt/ooNns2scsrbeeaelsenecdtpivooinrtoyt.ueHsxtecurare2rab,lowfneeasttruaanriednsetodofdmpereiecpdroincpetouCrreaOlv2noaeluntwmdoeNr,k2msuefrpsootampkoetrshe,evaoesxluwpmeerlelim,aasennCdtaOBlE2d/TaNtas2uosrefflaCecceOtia2vriaetnya.d.WWNee2
NIeli foundthatthehighestCO /N selectivitycanbeachievednotattheregionsofhighestCO uptakebutattheregionsof
ORuid lowest N uptake where m2esop2ores disrupt N adsorption. This insight will help guide exp2eriments to synthesize better
CALIFharingg porous ca2rbons for post-combustion CO2 cap2ture.
NIV OF acs.org/s Post-combustion CO2 capture from flue gas (composed mathematicalmodelbasedonsampledata.25AsapopularML
Us. mainlyofN )isanimportantapproachtomitigatethe method based on artificial neural networks (ANNs), deep
oaded via https://pub ocarpgatunriceCfrOhaam2veeemwboiserskeiosn2n(sMd.1e−Ov3eFlAso)pr,5ea−dn7,gecionovcfalpuleodnritno−guosrzmgeaoantleiictreifsar,la4smfomerweCtoarOlk−s2 lideteaecrapnninlegbauerinmldinpgaloycinsomacnphleAemxNiNnstorynwliihtnhaevamerumblteiopednleeli.hniTcdrhdeeeansainplagpyleiqcrsua,ticisokonlysthoiantf
wnlSee (COFs),8−10 porous organic polymers,11 porous carbons,12−14 recent years.25−29
Do ionic liquids,15,16 advanced solvents,17−19 and mem- One application of machine learning is to rapidly screen
branes.20−22 materials,28 predict properties,25 and suggest promising
Porous carbons are versatile because of their tunable pore candidates for synthesis.30 Recently, deep learning was for
size, pore shape, and electronic conductivity.23 An empirical the first time applied to predict CO adsorption in porous
2
relationforCO adsorptioncapacityasafunctionofmesopore carbons.25Itwasfoundthat,inadditiontomicroporevolume,
2
(V )andmicropore(V )volumeshasbeenproposedfor mesopore volume and BET surface area (S ) are also
meso micro BET
porouscarbons:MaCdOs2=0.095+2.10×Vmicro+3.51×Vmicro× importantindictatingCO2uptake.25Inspiredbythisworkand
V .24AlthoughthisrelationshipcanbeusedtopredictCO moving beyond CO uptake, herein, we use deep learning to
meso 2 2
uptake, it will have difficulty in predicting CO /N selectivity, evaluate another important factor in using porous carbons for
2 2
sinceitdoesnotapplytoothergasessuchasN2.Inaddition,it post-combustion carbon capture, namely, CO2/N2 selectivity.
isoftenverydifficulttomeasureaccuratelythelowN uptakes To achieve our goal, we built two deep neural networks
2
in materials with high CO2/N2 selectivity, so it is desirable to (DNNs)foradsorptioncapacitiesofCO2andN2,respectively.
predict N uptakes to guide synthesis.
2
To establish the complex relationship between the textural Received: September10, 2019
properties of porous carbons and the desired CO /N Accepted: October16,2019
2 2
selectivity, machine learning is a powerful tool to build a Published: October16,2019
©2019AmericanChemicalSociety 558 DOI:10.1021/acsmaterialslett.9b00374
ACSMaterialsLett.2019,1,558−563
NaN NaN NaN
0 None None None
1 CiteThis:ACSMaterialsLett.2019,1,558−563
2 None None None
ABSTRACT: Porous carbons are an important class of porous material for carbon capture. The textural properties of\nporouscarbonsgreatlyinfluencetheirCO adsorptioncapacities.Butitisstillunclearwhatfeaturesaremostconductive\n2\ntoachievinghigh CO /N selectivity.Here, wetrained deep neuralnetworks fromthe experimentaldataof CO and N\n2 2 2 2\nuptakesinporouscarbonsbasedontexturalfeaturesofmicroporevolume,mesoporevolume,andBETsurfacearea.We\nthen used the model to screen porous carbons and to predict CO and N uptakes, as well as CO /N selectivity. We\n2 2 2 2\nfoundthatthehighestCO /N selectivitycanbeachievednotattheregionsofhighestCO uptakebutattheregionsof\n2 2 2\nlowest N uptake where mesopores disrupt N adsorption. This insight will help guide experiments to synthesize better\n2 2\nporous carbons for post-combustion CO capture.\n2
0