Table 5 Properties of ML models compared to experiments and other popular water models

From: Machine learning coarse grained models for water

 

Exp

ML-BOP

ML-BOP/dih

ML-mW

mW

TIP4P/2005

MB-pol

iAMOEBA

SPC/E

TIP3P

Neighbors (3.3 Å cutoff)

4.51a

4.66

4.67

4.58

4.49

4.44

4.58

4.46

4.41

4.55

D298 K (×10−5 cm2 s−1)

2.3b

3.0

3.0

4.7

6.4

2.1

2.2

2.5

2.5

5.2

ρ298K,1atm (kg m−3)b

997.0

995.6

996.5

997.0

997.3

993

1007

997

994

982

ρmax (kg m−3)c

999.9

998.3

999.0

998.5

1003.8

1001

1014

999.9

1012

1038

TMD (K)c

277

276

278

279

251

278

258

277

241

182

ΔHvap (kcal mol−1)

10.52

10.01

10.01

10.30

10.66

11.98

10.1

10.94

11.69

10.49

Tm (K)

273

273

273

289

273d

252

264

261

215

146

TMD—Tm (K)

4

3

5

−10

−24

26

−6

16

26

36

ΔHmelt (kcal mol−1)c

1.44

1.23

1.23

1.40

1.26

1.16

1.19

0.74

0.30

ΔSmelt (cal mol−1K−1)

5.27

4.52

4.52

4.84

4.60

4.6

4.56

3.44

2.06

ρliq at Tm (kg m−3)c

999.8

997.95

998.0

998.5

1001.0

993

1013

999

1011

1017

ρIh at Tm (kg m−3)c

917

929

930

928

978

921

920

929

950

947

ΔVmelt (cm3 mol−1)

−1.61e

−1.35

−1.39

−1.38

−0.42

−1.42

−1.80

−1.36

−1.14

−1.31

(dp/dT)melt (bar K−1)

−137f

−141

−136

−146

−463

−135

−141

−126

−66

  1. aref. 23
  2. bref. 20
  3. cProperties that are included in the training of ML models
  4. dref. 64
  5. eref. 63
  6. fref. 59
  7. Properties comparison from experiments55, popular polarizable8, 19 and non-polarizable models26