can Verified User
  • Member since Jan 20th 2017

Posts by can

    Leif Holmlid uploaded:

    Response to Comment on Ultradense protium p(0) and deuterium D(0) and their relation to ordinary Rydberg matter: a review [Physica Scripta 94 (2019) 075005]


    In this answer to the Comment by Hansen and Engelen it is shown, that if there is any violation of the baryon number conservation law in H(0) nuclear reactions, it is not at all of the form that the authors believe. Their belief is disproved by cited well-known scientific results from other groups. It is further shown that quantum mechanics in H(0) molecules is different than these authors believe, not formulated in kinetic energy terms but defined by angular momentum quantization. Repetition of experiments is required, not pondering by non-specialists.

    Although what they're doing exactly is not clear since they hardly publish anymore detailed information about their models, they are most likely not continuously indexing web pages. The data for training the language models is usually assembled in advance, then the models trained for weeks/months on that data.

    For example, for LLaMA (which I linked earlier) the authors used a dataset composed like this:

    The "tokens" OpenAI lists on their pricing page are for the input/output generated through their service/API. You can check in practice how tokens work here:

    Paper on LLaMA from the authors:

    LLaMA: Open and Efficient Foundation Language Models
    We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it…

    LLaMA: Open and Efficient Foundation Language Models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, Guillaume Lample

    We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community.

    Here are a few examples on how LLaMA-33B (one of Facebook/Meta's recently released language models, arguably among the best available for local use) thinks a powder-based LENR reactor could be made (don't try these up).

    Putting jokes aside, a problem is that the quality of the responses strictly depends on what data was used for the training. At their core, large language models (even ChatGPT-4) are for the most part "text predictors" and don't have real reasoning or logical skills, so they shouldn't be blindly trusted even if they can give the semblance of intelligence. They are also prone to "hallucinating", i.e. make stuff up.

    I think It would be interesting to see a "LENRGPT", i.e. a language model fine-tuned on material from the LENR community. One of the already available "pre-trained" models could be used as a base (e.g. Meta's recently released LLaMA, or other models with a more liberal license), then further source data (papers, documents, etc) from the LENR community used for the fine-tune. Many of these models are already trained on a large amount of papers from arXiv, but not specifically on LENR papers.

    While this would not be something exceedingly out of reach for small groups or universities, even just the fine-tune step takes significant amounts of compute power (i.e. money) for anything remotely smart, though. On the other hand, the fine-tuned models could be sized such that for inference (querying the language model) consumer-grade GPUs may be sufficient.

    [...] To achieve excess heat tendency of the pressure must be declining. Excess heat - at least in this experiment - will not show up until fuel will "decide" to load the hydrogen back, at least in a limited way. If the gas inside the cell is really hydrogen then tendency is strongly rising. But it should rather continue in what we saw immediately after filing. If that trend would continue we would see excess heat.

    Does this imply that the copper-coated meshes will be more reliable in showing excess heat? It feels like there is a competing negative effect due to palladium forming a hydride/deuteride.


    Utah’s Cold Fusion Moment
    In 1989, The University of Utah was in the national spotlight when two of its chemists announced the discovery of a powerful energy source that would solve the…

    In a paper published some time ago, Holmlid proposed that metal–graphite thin films could be used for storage, but also that large-scale storage is not recommended.

    Future interstellar rockets may use laser-induced annihilation reactions for relativistic drive
    Interstellar probes and future interstellar travel will require relativistic rockets. The problem is that such a rocket drive requires that the rocket…


    The most suitable future storage medium will probably be an assembly of thin metallic or graphitic films.


    Large-scale storage as H(0) is not recommended since the spontaneous nuclear reactions taking place in H(0) could give uncontrolled energy and radiation release.

    In theory there's not much preventing what you're suggesting, but it has never been done in practice, so it's unclear whether it can actually work for bulk storage (i.e. in amounts large enough to significantly affect the density of the storage material).

    It would really be interesting if extremely high-density materials (e.g. > 25 g/cm3) could be obtained by storing large enough amounts of H(0) within them. It would be conclusive and tangible proof that ultra-dense hydrogen actually exists.

    What about placing a reversible hydride upstream in the gas feedline? It could be externally heated and controlled automatically, and possibly not affect too much reactor calorimetry in the current (magicsound's) configuration.

    I notice that in the Santilli Magnegas patent he says that the carbon rod needs to be changed every 10 minutes in his prototype gas generator.

    I quickly skimmed through the patent, and after the introduction and theory, it describes an improvement which does not use carbon rods. The carbon is from the slurry into which they are immersed.


    [...] The sixth main novelty of this invention over pre-existing patents is the elimination of Carbon rods as electrodes and the use of high temperature resistant materials which do not necessarily release carbon under an electric arc. More specifically, extensive experimentation has established that, under a DC arc powered by 15 KW anodes composed by a tungsten rod of about 3/46 diameter and 2o length experience minimal consumption, with replacement needed over at least one month of operation.

    In the latest paper uploaded, Holmlid suggests that there are results related to ultra-dense hydrogen in the scientific literature, which might partially answer the above question:

    Decay-times of pions and kaons formed by laser-induced nuclear processes in ultra-dense hydrogen H(0)
    Laser-induced nuclear reactions in ultra-dense hydrogen H(0) (see review in Physica Scripta 2019) create mesons (kaons, pions) with up to 100 MeV thus velocity…


    Related results on ultra-dense hydrogen exist in the literature. A superconductive state consisting of very high-density hydrogen clusters in voids (Schottky defects) in palladium crystals has been studied experimentally by Lipson et al. [17]. This effect was discussed as due to Bose-Einstein condensation [18] or a Casimir effect [19]. Such hydrogen clusters may give increased nuclear fusion gains [20]. The close relation between these hydrogen clusters and ultra-dense hydrogen has been pointed out [21].

    References 17-21:

    [17]. A. Lipson, B. J. Heuser, C. Castano, G. Miley, B. Lyakhov, and A. Mitin, Transport and magnetic anomalies below 70 K in a hydrogen-cycled Pd foil with a thermally grown oxide, Phys. Rev. B 72, 212507 (2005). DOI: 10.1103/PhysRevB.72.212507

    [18] G. H. Miley, H. Hora, K. Philberth, A. Lipson, and P. L. Shrestha, in Low-Energy Nuclear Reactions and New Energy Technologies Source Book, eds. J. Marwan and S. B. Krivit, Vol. 2, p. 235-252 (American Chemical Society/Oxford University Press, Washington DC, 2009).

    [19] H. Hora, G. H. Miley, Maruhn–Greiner maximum of uranium fission for confirmation of low energy nuclear reactions LENR via a compound nucleus with double magic numbers. J. Fusion Energ. 26, 349 (2007).

    [20] X. Yang, G. H. Miley, K. A. Flippo, and H. Hora, Energy enhancement for deuteron beam fast ignition of a precompressed inertial confinement fusion target, Phys. Plasmas 18, 032703 (2011).

    [21] L. Holmlid, H. Hora, G. Miley, and X. Yang, “Ultrahigh-density deuterium of Rydberg matter clusters for inertial confinement fusion targets”. Laser Part. Beams 27, 529 (2009).

    I am pointing out that unlike other cavitation energy claims, cavitation here appears to be produced by flash-boiling water with heat inside a constricted tube, not water flow through a pump.

    I understand that the authors tried to replicate excess heat observations observed in commercial systems, but in my opinion from an experimental point of view a joule heater would decrease the number of variables involved and make it easier for others to reproduce the results.

    Alan Smith

    There have been cavitation systems in the past where the energy-producing cavitation was directly caused by the pump, but here the pump appears to be just a heating means, which is the confusing part in my opinion.

    If ordinary joule heaters could be used, something similar to Andrea Rossi's "small" E-cats from the early 2010s could be easily devised; actually that reminded me of those.

    I never had a look at the system in detail before, so sorry if I'm missing some details here. The refrigerant pump is just used to heat a tube to 150 °C or so in order to make cooling water in an outer jacket to flash evaporate and cause energy-producing cavitation?

    Why is a refrigerant pump needed at all? Couldn't this be more conveniently done with a cartridge heater or a band heater (or both)?