Permachar
DISCLAIMER: this post definitely includes information from perplexity.ai pro search (Grok 4), deep research and labs that has not been 'properly' fact checked. It's a work in progress but I thought I'd give you the heads up and a good excuse why I've paused TLUDing for a while...
- Feedstock grown with regenerative principles integrating biochar in the soil eg. bamboo (fastest CO2 sequestering plant, high Si). A strong candidate is Moso (Phyllostachys
edulis) which is a running bamboo that can be pyrolysed for biochar that exhibits electrical properties that make it suitable for various applications, particularly in energy storage and
environmental remediation.
- according to the American Bamboo Society, the Botanical Classification of Bamboo is:
- **KINGDOM:** Plantae
- **PHYLUM (DIVISION):** Magnoliophyta
- **CLASS:** Liliopsida
- **SUBCLASS:** Commelinidae
- **ORDER:** Cyperales
- **FAMILY:** Gramineae (Poaceae)
- **SUBFAMILY:** Bambusoideae
- **TRIBE:** Bambuseae
- **SUBTRIBE:** bambusinae
- bamboo harvested, dried eg. <10% moisture content, chips->
- hopper->a- Pyrolysis eg. Joey/Cornell Uni 'Open source trough pyrolyser' (continuous)
- b-steam activation
- c-milled ?optimal particle size for Metal Organic Framework (MOF), OR 'abc' in an integrated industrial system, such as:
- https://www.bygen.com.au/
- hopper->a- pyrolysis to biochar->b- low temperature activation (LTA, uses 'gases') (more research eg. surface chemistry functionalisation, such as C-O-M
bonding sites)->c- milled (to custom particle size)
- minerals for MOF eg.Manganese Mn (highest redox potential of earth-abundant metals. Oz is world's third largest producer (3 million metric tons) in 2023 behind 1-South Africa
(7.2 million metric tons), 2-Gabon (4.6 million metric tons))
- fabrication of MOF for application-specific tunable chemistry
- **Functionalization of (Steam) Activated Biochar** (optional)
Further modify the activated biochar surface, if needed, to optimize functional groups for metal binding. This may involve mild chemical treatments to increase sites for C–O–M
coordination bonds.
- **Mixing with Metal Precursors and Organic Linkers**
Combine the **functionalized biochar** with **metal salts** (e.g. Manganese precursors) and **organic linkers** (e.g., dicarboxylates) in a suitable solvent. This prepares the
components for **self-assembly**
NOTE: ethanol or water can be used as a solvent.
A '**One pot**' process can possibly be used.
- **Self-Assembly under Controlled Conditions**
Allow spontaneous organization through coordination bonds and non-covalent interactions (e.g., hydrogen bonding) at mild temperatures (**room temperature** to 80°C) and controlled
pH. This forms the 3D MOF network within the 3D biochar matrix.
- **Post-Synthesis Treatments**
Wash, dry, and possibly further activate the assembled MOF to remove impurities and stabilize the structure. This may include drying at low temperatures or additional thermal
processing
- applications
- catalyst
- environmental remediation eg. organic pollutants, tetracyclines, dyes, heavy metal removal from water
- CO₂ Reduction and Gas Conversion eg. CO2 to fuel
- Hydrogen Evolution Reaction (HER) eg. ?photocatalyst, electrocatalyst, for Green Hydrogen production
- energy storage
- supercapacitors
- batteries eg. anode and possibly cathode for Na ion
- hybrid eg. redox flow batteries
- desalination
- perovskites
***
A NEW AI SEARCH STRATEGY
Here's an example of keyword clustering for 'Carbon based materials' I put together after initial keyword searches. I organised the keywords into clusters that can be directly entered into the
search query in perplexity.ai
1. keyword extraction considered 'important' from the initial searches
2. clustering of the keywords
3. keyword clusters fed back into follow up questions
4. Repeat 1, 2 and 3
Here they are:
- Carbon biology
- biochar
- biochar 3D matrix
- biochar surface chemistry and functionalised groups
- DNA self-assembly compared to Metal Organic Framework MOF self-assembly
- Carbon chemistry
- Carbon
- crystalline nano structures
- advanced Carbon materials
- steam activated activation
- temperature
- C-O-M bonds
- coordination bond mechanism
- perovskites
- Metal Organic Framework MOF
- template
- MOF topology
- earth-abundant metals for biochar MOF
- Fe-doped
- Manganese-doped
- Nickel-doped
- Na-doped
- synthesis parameters
- long-term stability mechanisms
- Carbon physics
- electrical parameters
- electrically conductive networks/pathways
- electrical conductivity EC
- capacitance
- cycling stability
- biochar
- pyrolysis
- temperature
- panel kilns
- TLUD
- rotary kilns
- trough pyrolyser
- Microwave Assisted Pyrolysis MAP
- GaN
- feedstock
- bamboo
- biochar
- Silicon-rich
- other feedstocks
- moisture content MC
- cellulose nano crystals CNC
Black gold digging section
- applications
- battery electrolyte anode cathode
- supercapacitors
- industry
- scalable production
- industrial processes
- manufacturing
Information about perplexity.ai
Lets pop the hood
- perplexity is a very clever chatbot - or - rather, a clever AI entity that has a long way to go before achieving AGI status
- here's some probing questions to ask it:
- what LLM does perplexity.ai 'Deep research' use and how does it work?
- Can perplexity.ai mine research behind paywalls?
- Does perplexity.ai store data from the academic research papers it reads?
- Why doesn't perplexity.ai use in-text page referencing?
- How is the perplexity.ai chatbot engineered?
- Is there a chatbot or LLM that provides inline page references of academic research papers?
- perplexity offers Pro, Deep Research or Labs searches. I've started using Labs with some interesting results generated.
- perplexity Labs is a new approach, creating tables, graphs, charts and possibly pictures, as well as deep research in between
- Web, academic, finance and social search categories are now available
An old research strategy that works extremely well
I go back to 'Problem Based Learning' research strategy that I learnt at University to cut through the smoke and mirrors of AI:
1 - define the problem->
2 - extract the data points (keywords)->
3 - build hypotheses based on the data points->
4 - create learning issues (questions) to test the hypotheses->
5 - compile a list of references to research the learning issues (perplexity.ai provides these plus additional web searches or even text books)->
6 - research the learning issues->
7 - refine existing hypotheses and build new hypotheses->
8 - create new learning issues to test the hypotheses->
9 - research the learning issues->
Repeat steps 7-9 until you are satisfied with the intel and have 'solved' the problem.
Here's some additional context:
-define the problem
- open a new 'thread' then start with open and general questions (LIs), based on testing a hypothesis, then 'zoom in' with every consecutive inquiry - just like a funnel
- the data points are the keywords. The LLM is finding existing and new relationships between the keywords
- follow references (in text with full references at the end) and hunt down research papers (if you can access them) and get to the primary source BUT often research paper references are
completely out of context, some of them probably just get their abstracts mined and some are behind pay walls.
- search results not peer reviewed - the user becomes the peer. This means you get to compare your own 'expert knowledge' to the results - what are the consistencies and contradictions? How can
they be resolved?
- use different internet search engines, as an additional layer for papers or ones not mentioned
- BUT what is the point of using AI if u need to fact check and evaluate it's sources? I suppose it needed to be done anyway for rigorous research but AI takes it to the next level of
confusion!!
- basically, if Intel is used for critical business decision making, check the sources and get to the primary sources, assuming they're not already primary sources. Also, could run some identical
Pro searches using different LLMs eg.Grok 4
- new knowledge, if it can be recognized with prior fact-checked knowledge, is also difficult to fact check because there's no prior history
That's all for now!
Write a comment