Astrochemistry: Discovery of Novel Forms of Water in Uranus and Neptune

1
Astrochemistry Discovery of Novel Forms of Water
in Uranus and Neptune

• Nir Goldman
• Lawrence Livermore National Laboratory
• March 8th, 2006
• This work was performed under the auspices of the
  U. S. Department of Energy by the University of
  California Lawrence Livermore National Laboratory
  under contract No. W-7405-Eng-48.

Experiments Alex Goncharov, Jonathan Crowhurst,
Joe Zaug Theory Chris Mundy, Will Kuo, Larry
Fried (PI)
2
Uranus and Neptune have similar properties
3
Voyager II data provides indirect insight into
planetary interiors

• Voyager II spacecraft data shows Uranus and
  Neptune have strong magnetic fields
• Due to unique forms of water in interior? Novel
  lattice phases?
• Estimate of interior composition is based on the
  density profile, and assumed chemistry and
  Equation of State models

Equation of State models provide P-T profiles,
and possible states of water
4
H2O dissociation could yield high magnetic field

• Conductivity of matter inside planet crudely
  characterizes the flow that produces the
  planetary magnetic field
• Water inside Uranus and Neptune could have high
  conductivity
• Maybe caused by complete molecular ionization
• H2O 2H O2-
• Exotic phase Superionic water?

Predictions about planetary interiors rely upon
accurate Equation of State modeling (EOS)
5
Equation of state models yield very diverse
results for H2O at extreme conditions

• EOS models relate pressure and temperature to
  chemical composition
• Accuracy is heavily dependent on initial guesses
  of chemical products
• Requires inputs from theory and experiment

Great need for description of interior chemistry
in order to derive models consistent with
observational data
6
Chemistry under extreme thermodynamic conditions
is not well understood

• Major Issues
• Rapid bond dissociation
• molecular to non-molecular transition
• Covalent vs. ionic bonding???
• Novel states of matter
• Metallization of H2, N2
• Chau et al., PRL (2003)
• Galli et al., Nature (2005)

Experimental/theoretical challenges involve
attaining/modeling this extreme P-T regime
7
Gas guns and Diamond Anvil Cells are used to
achieve extreme conditions
Diamond Anvil Cell
Gas gun
Description
Advantages
Disadvantages
We have to rely on computations to determine the
atomic structure and dynamics
8
Molecular Dynamics simulations (MD) can provide
key answers

• Calculate molecular trajectories via Newtonian
  mechanics
• MD recreates system on computer as close to
  nature as possible
• Underlying physics is very simple. However
• Computationally, MD can be very difficult
• Real challenge is in coming up with decent
  Potential Energy Surface model V(rN)

Tools from Statistical Mechanics allow us to
connect simulation to experiments
9
Example of MD simulation of ambient liquid water

• Historically, MD simulations could not accurately
  depict bond breaking
• Ab initio modeling explicit modeling of
  electronic ground state required
• Computers were not fast enough for high levels of
  theory

Faster computers and more efficient theory will
allow issue of superionic water to be resolved
for the first time
10
Ab initio MD provides structural and dynamic info
about extreme water

• Car-Parrinello Molecular Dynamics (CPMD) ab
  initio MD software
• Explicit modeling of electrons and nuclei (few
  empirical equations)
• Density Functional Theory (DFT) based MD, using a
  plane-wave basis set
• Use larger system size and much larger basis set
• 54 H2O, 120 Ry (vs. 32 H2O, 70 Ry)

We will provide chemical insight into experiments
on hot, compressed water
11
CPMD computational details

• CPMD is about 150,000 lines of F90.
• The computational engine is the 3-D FFT
  parallelized using both MPI and OpenMP directives
  to take advantage of SMP nodes.
• CPMD achieves 65 parallel speed up for 1,920
  CPUs (960 nodes)

• LLNLs Thunder
• Linux cluster, Itanium2 processors (1.4 GHz)
• 1024 nodes, 4 procs/node
• Peak performance 22.9 TFlops/s
• Currently 11 on Top500 list
• (1 once upon a time)

12
Even small systems (100s of atoms) require
LLNLs supercomputers

• 2.0 g/cc, 34 GPa, 2000K
• Real space mesh 126 processors needed
• UV (Power5) 16,000 Hours
• Thunder (Itanium2/Linux) 32,000
• MCR (Xeon/Linux) 40,000
• Need 6 8 densities for each temperature
• 500,000 CPU hours total

An entire supercomputer is filled with many
smaller jobs instead of a single gigantic one
13
Does superionic water exist and what is it
exactly?

• Validate theory via experiments
• Calculate diffusion constants of oxygen and
  hydrogen and vibrational spectra
• Create a simple chemical picture of superionic
  water
• Observe structure via radial distribution
  functions
• Calculate species concentrations and lifetimes

14
Our starting point calculated H2O phase diagram

• Constant pressure-temperature simulations with
  Carr-Parrinello molecular dynamics (CPMD)
• Small system size 32 H2O
• P30-300 GPa, T 300-7000K
• Superionic phase has oxygen bcc sublattice,
  mobile protons
• DH (2000K, 30 GPa) 1.8 x 10-3 cm2/s

• Uranus, Neptune 56 H2O, 36 CH4, 8 NH3
• hot ice mixture contributes to magnetic field
  measurements by Voyager 2 spacecraft
• Due to high ionic conductivity from completely
  ionized H2O

Cavazzoni, et al., Science, 283, 44, 1999.
15
Structure of Superionic water

• Superionic Solids exhibit exceptionally high
  ionic conductivity
• partial melting one ion diffuses through
  crystalline lattice of remaining types
• some famous examples PbF2, AgI.
• Originally thought to be uncommon in
  hydrogen-bonding compounds

16
Somewhat contradictory pieces of data about
superionic water
1. HIGH IONIC CONDUCTIVITY
Schwegler et al., Phys. Rev. Lett., 87, 265501
(2000)
vs.
23 GPa 1390 K

• CPMD results with 54 H2O do not show mobile
  protons or oxygen lattice.
• Absence of lattice confirmed by X-ray data Frank
  et al., Geochim. et Cosmochim. Acta, 68, 2781,
  2003.

17
Water at High Pressure and Temperature

• Our simulations much bigger than before.
• - 1000K 2000K
• - 1.5 g/cc to 3.0 g/cc
• - Pressures from 15 to 115 GPa

Compressing a liquid configuration
Heating an Ice VII configuration
Cavazzoni et al., Science, 283, 44, 1999.

• We have determined a more accurate phase boundary
  of superionic water
• We have devised a simple chemical picture for
  this phase.
• Fundamental question
• -How can we define a molecule at these conditions?

18
Simulations show dramatic changes in the
structure of water with increasing pressure
19
The diffusion constant is calculated from the
Einstein-Smoluchowski relation
20
Oxygen and hydrogen diffuse on two different time
scales
2000 K
Hydrogen
Oxygen
We determine the superionic phase boundary from
the oxygen freezing point as a function of
temperature
21
Abrupt changes in the Vibrational spectra allow
us to determine phase boundaries
Melting curve at high pressure and temperature
was determined via the changing phonon mode
Goncharov, Goldman, Fried, Crowhurst, Mundy, Kuo,
Zaug, PRL, 2005
22
Experiment and simulations show weakening of the
O-H bond in liquid water
23
We have redefined the phase diagram of water at
extreme conditions
Phase diagram of water

• Melting line is in agreement with externally
  heated DAC data

• Triple point at 47 GPa and 1020 K, significantly
  higher than Parrinello (25 GPa)

• Transition to a superionic phase is inferred
  from a combination of experiments and simulations

Goncharov, Goldman, Fried, Crowhurst, Mundy, Kuo,
Zaug, PRL, 2005
Goldman, Fried, Kuo, Mundy PRL (2005)
Simulations can also provide a chemical picture
of superionic water
24
Statistical Mechanical Analysis

• Validate theory via experiments
• Calculate diffusion constants of oxygen and
  hydrogen
• Create a simple chemical picture of superionic
  water
• Focus on results at 2000 K (particularly unique)
• Observe structure via radial distribution
  functions
• Calculate species concentrations and lifetimes

25
Radial distribution function (RDF) yields
structure, g(R)
Probability of finding any two particles in the
config. (r1,r2)
Investigate pairs of OO, OH, HH
26
Oxygen-Oxygen structure (RDF), 2000 K
Average out vibrations bcc lattice, like ice
VII, ice X
3.0 g/cc, 115 GPa
2.6 g/cc, 75 GPa
2.0 g/cc, 34 GPa
27
Oxygen-Hydrogen RDF
Intra-molecular
Inter-molecular
28
We use the ROH free energy surface to define
molecules
2000 K
115 GPa
The O-H free energy barrier decreases
dramatically with pressure.
75 GPa
34 GPa
29
We have a simple picture for proton mobility
1-D free energy surface shows pronounced drop in
dissociation barrier
30
We use the ROH free energy surface to define
molecules
2000 K
115 GPa
75 GPa
34 GPa
31
Concentrations and lifetimes at 2000 K(lt10 fs
non-molecular)
H2O, H3O, OH-
polymer

• 34 GPa (2.0 g/cc)
• H2O lifetime 40 fs
• H3O, OH- lt 10 fs

• 75 GPa (2.6 g/cc)
• All species lifetimes 10 fs or less

Neutral and ionic (H2O)2 (H2O)6
The polymer species consists of very
short-lived networks of bonds
32
Network solid is partially covalent at 95 115
GPa

• Non-molecular (based on lifetimes)
• At 2000K, 115 GPa, 50 covalent bonding
• Tetrahedrally coordinated oxygen
• Analog to ice X symmetric H-bonding

Goldman, et al., Phys. Rev. Lett., 94, 217801
(2005).
33
Network solid is partially covalent at 95 115
GPa
Goldman, et al., Phys. Rev. Lett., 94, 217801
(2005).
34

• Liquid, highly reactive H2O (34 58 GPa)

2. Superionic phase with asymmetric H-bonding (75
GPa)
3. Superionic phase with symmetric H-bonding (95
115 GPa)
35
Discussion

• Hydrogen diffusion rates can be extremely rapid
  over disordered, mobile oxygen phase
• Superionic phase occurs at higher pressure than
  previously predicted
• 75 GPa at 2000K (Cavazzoni et al. 30 GPa)
• Superionic water is best understood as transient
  partially covalent bonds which form networks
• Ensemble of transition states

• Goncharov, Goldman, Fried, Crowhurst, Mundy, Kuo,
  Zaug, Phys. Rev. Lett., 94, 125508 (2005).
• Goldman, Fried, Kuo, Mundy, Phys. Rev. Lett., 94,
  217801 (2005).

36
Discussion

• Planetary implications
• High water ionic conductivity can happen in
  absence of superionic phase
• Water could be the source of the large magnetic
  fields in Uranus and Neptune
• How does water behave in presence of CH4, NH3?
• What simple rules govern superionic behavior?

• Goncharov, Goldman, Fried, Crowhurst, Mundy, Kuo,
  Zaug, Phys. Rev. Lett., 94, 125508 (2005).
• Goldman, Fried, Kuo, Mundy, Phys. Rev. Lett., 94,
  217801 (2005).

Field of Extreme Chemistry has many exciting
research opportunities
37
Prediction of Superionic Hydrogen Fluoride (HF)

• At 66 GPa and 900 K, have stable F bcc lattice
• symmetric H-bonding
• Model superionic system more easily achievable
  with Diamond Anvil Cell
• Further study will allow us to develop simple
  rules for this system

Possible superionic hydrogen diffusion mechanism
1.
2.
3.
38
Shocked molecular simulations of soft condensed
matter
Advances in tera-scale computing and a novel
Multi-scale simulation technique allow for
accurate shock simulations for the first time
We observe graphite forming diamond at shock
velocities of 12 km/s
Novel phases and reaction pathways can be
elucidated through our simulations
39
Acknowledgments

• Larry Fried
• Experiments Alex Goncharov, Jonathan Crowhurst
  and Joe Zaug
• Chris Mundy and Will Kuo

40
Molecular simulation is the foundation for
understanding extreme chemistry
Models of Uranus and Neptune rely on Equation of
State predictions
EOS models require inputs from experiment and
theory
Molecular simulation is needed in order to
provide simple chemical pictures for experiments
We have used experiments and theory to resolve
controversy regarding superionic water
41
Experiments have difficulty describing chemical
composition
Nitrogen has metallized Chau et al., PRL (2003)
Atomic nitrogen Radousky et al., PRL (1986)
What is made when we shock N2 ? Atoms ? Chains ?
Metal ?
We have to rely on computations to determine the
atomic structure and dynamics
42
Diffusion constants, 1000 2000K
D 10-4 cm2/s
D 10-5 cm2/s to zero
Hydrogen
Oxygen
We determine the superionic phase boundary from
the oxygen freezing point as a function of
temperature
43
H2O lifetimes, 1200 2000K
1200 K
1500 K
2000 K
onset of superionic phase
Non-molecular lifetime of all species is less
than 10 fs (one O-H vibrational period)
Molecular to non-nolecular transition occurs at
densities greater than superionic transition (2nd
phase transition)
44
Hydrogen-Hydrogen RDF
45
Hot ice interior contains small molecules at
extremely high pressures and temperatures

• Gravitational moments and atmospheric composition
  could provide insight into chemistry of physics
  of the interior
• Data provides constraints for equation of state
  of candidate materials

Uranus and its moons, from Voyager II
Water at high P-T conditions of the interior
could have unique chemistry which affect
planetary processes
46
Diffusion constant and vibrational spectral
results

• Superionic diffusion of hydrogens occurs in
  presence of disordered oxygen phase
• At 2000K, oxygen freezing occurs at ca. 2.6 g/cc
  (75 GPa)
• Experimental Raman spectra validate theory
• Diamond Anvil Cell experiments (DAC) are
  currently technologically limited
• Limits P lt 50 GPa, T lt 1500 K
• Missing interesting features along 2000K isotherm

47
Structural analysis

• O-O exhibits stable bcc lattice at higher
  densities
• Confirms earlier thoughts about superionic water
• H-H and O-H shows structure as well (ice X-like)
• Lattices are very transient (lt 10 fs lifetimes)
• Shift in first minimum in g(ROH)

48
Future Work Shocked Materials

• High P-T conditions can be achieved
  experimentally by shocking materials
• Presents very difficult simulation challenges
• High level of theory required to accurately model
  chemical bond dissociation
• Traditionally, shocked simulations require very
  large system sizes
• Subsequently, we must use very low levels of
  theory (no ab initio MD)

49
Future Work Predicting new Hydrogen-bonding
Superionic solids

• H-bond symmetrization is a unique phase of
  superionic H2O
• Cannot be observed with current Diamond Anvil
  Cell technology
• Halogen hydrides show promise as model systems
  (HF, HCl, HBr)
• Evidence of non-superionic but symmetric
  H-bonding at lower P and T

50
Advances in theory and tera-scale computing allow
for ab initio simulations
Overlaps between the timescales of molecular
simulations and experiments are becoming
possible.
51
Water is nonmolecular at 3 g/cc and 2000 K

• Free H and O2- have negligible lifetimes

Non-molecular lifetime of all species is less
than 10 fs (one O-H vibrational period)
It forms a covalent transient network structure
52
Water phase diagram is largely unchartered at
extreme conditions

• Predictions are based on equation of state (EOS)
  models
• Accuracy is heavily dependent on description of
  chemical products at those thermodynamic
  conditions