Released-Node Calculations

We describe here the simplified version of the released-node method with Green s function sampling. This version allows large step sizes without step size error, eliminates conditional sampling, and eliminates the use of a guide function. Importance sampling is incorporated by use of variable sample weighting. 

The method is based on GFQMC as described above, and the procedure is based on the repeated sampling sequence 

For importance sampling to be most effective, the individual values of the products ill summations of Eq. [40] must be approximately equal. 

This may be achieved by dividing walkers to produce weights inversely proportional to the values of their associated trial wavefunctions. 

The procedure is begun with an initial distribution of positive walkers at positions (X ) in the region where the trial function is positive. The weight of each walker is multiplied by [V(X )/ ] and the walker is moved to a new position (X ) obtained by sampling Go(X,X ). The wavefunction yo( ) e 

---

Not to leave the reader with the impression that transient estimate and release node calculations are not useful, let us mention briefly some results obtained with these methods. In 1980 a good convergence was obtained on the electron gas with up to 54 electrons [6j. Some of these calculations have recently been redone by Kwon [99] using better wavefunctions and DMC algorithms. Ceperley and Alder [9] also studied some small molecules (LiH, Hj). Recently, Caffarel and Ceperley [67]... 

A few alternatives to the FN constraint are available within QMC. The released node method starts from a FN-DMC calculation [56, 57]. When the FN constraint is relaxed (released node QMC) an estimate of the exact ground state energy may be obtained by incorporating a factor of —1 for each walker that crosses the nodal surface ... 

He ), the node problem can be overcome by exact cancellation methods (described below), and exact solutions can be obtained. For systems of as many as 10 electrons,released-node or transient estimate methods (also described below) can provide excellent approximate solutions. But, in general, the method of choice for systems of more than about 10 electrons is the fixed-node method. Although the fixed-node method is variational and does not yield exact results, it is the only choice available for quantum Monte Carlo calculations on many larger systems. The fixed-node method is remarkably accurate and generally yields energies well below those of the best available analytic variational calculations. 

B. Chen and J. B. Anderson,/. Chem. Phys., 102,4491 (1995). A Simplified Released-Node Quantum Monte Carlo Calculation of the Ground State of LiH. 

The current situation with the exact Monte Carlo methods, cancellation and released-node, is that, despite early and recent successes, these methods have yet to progress beyond LiH. However, these earlier calculations were performed with simple trial functions, and functions of much higher quality are now available. In addition, recently developed projection methods might extend the applicability of released-node Monte Carlo. 

The term on the left side of the equation represents the flow of internal energy in and out of the system, where m is determined by the kinetic Equations (1-9), Ac is determined by the total number of nodes generated, N, and the feedstock residence time in the reactor, which can be calculated by equations 4-7. The first term on the right side represents the heat transfer from the bottom heating plates TVs is the temperature of the heating plate and Ohai is the heat transfer coefficient which is determined by the heat transfer equations (Eq. 1-3), The second term is the radiation heat transfer contribution from the reactor wall. The last term represents the kinetic energy released during the pyrolysis reaction, which is assumed to be proportional to die rate of pyrolysis reaction (Eq.8-9). 

In the second computer run, a node is released extending the crack length to a length of a -I- Aa. This allows calculation of the new nodal positions of the same nodes that were... 

Sensitivity studies checking for time step and noding effects, and for machine dependencies, were done. The major problem identified was a machine dependency associated with exponentials and very small numbers it resulted in significantly different releases being predicted on different machines for refractory species. Other problems associated with differences in roundoff of small numbers were also found. All these problems were corrected immediately, and no machine dependencies were found in our final calculations. 

---