Robert Palgrave
@Robert_Palgrave
Followers
6,959
Following
1,479
Media
382
Statuses
4,059
Professor of Inorganic and Materials Chemistry at UCL. Director of UK National XPS Service @harwellxps
London, England
Joined November 2015
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
メイドの日
• 481103 Tweets
ONGSASUN CONFLICT
• 285999 Tweets
Whindersson
• 231794 Tweets
kyungsoo
• 155105 Tweets
メイド服
• 125643 Tweets
Madres
• 117135 Tweets
Joost
• 107534 Tweets
Mother's Day
• 78278 Tweets
#PLEDIS_TREAT_JUN_BETTER
• 71613 Tweets
#WHEREISJUN
• 71418 Tweets
Felipe Neto
• 59999 Tweets
トニオさん
• 53275 Tweets
Bannon
• 48141 Tweets
TREAT YOUR ARTISTS EQUALLY
• 44296 Tweets
Copa América
• 39907 Tweets
Gill
• 38177 Tweets
密漁海岸
• 33085 Tweets
Bitget
• 31404 Tweets
露伴先生
• 30182 Tweets
Palermo
• 21154 Tweets
San Martín
• 20931 Tweets
الامم المتحده
• 20581 Tweets
#CSKvsGT
• 19104 Tweets
JUN DESERVES BETTER
• 18341 Tweets
Trenes Argentinos
• 18123 Tweets
Thala
• 12759 Tweets
#IREvPAK
• 10790 Tweets
Magna Carta
• 10307 Tweets
$ATS
• 10175 Tweets
Pinned Tweet
Announcing a very special workshop this summer:
The Future of Materials Discovery - UCL, London 25-26 June
AI is going to change science in a big way. We have an unbelievable line-up of speakers to discuss what the future looks like
Please RT!
9
103
248
This exciting paper shows AI design of materials, robotic synthesis. 10s of new compounds in 17 days.
But did they?
This paper has very serious problems in materials characterisation. In my view it should never have got near publication. Hold on tight let's take a look 😱
🚀 Exciting News! Our A-Lab has revolutionized inorganic materials discovery! Combining ab-initio calculations, historical data from literature, and robotic testing, we're rapidly uncovering new compounds. 🤖
16
88
511
64
514
3K
Weird thing has happened in
#LK99
twitter storm - scientist's h-indices are being quoted, I think as a proxy for their reliability. This is not good, can't let it become normalised
Some examples:
Peter Higgs (Nobel prize) h=11
Grigori Perelman (solved Poincaré conjecture): h=10
In 2 days there are have been 4 studies that help explain LK-99's potential superconducting abilities.
These simulations converge on key properties that suggest a new class of SC materials, and help explain quirks of TK-99 we've seen so far.
Here is the easy-to-digest summary…
135
1K
6K
66
233
2K
Back in November, Google announced 2.2 milllion new materials.
Today, a paper in Chemistry of Materials from Ram Seshadri and Tony Cheetham dismantles that claim
Introducing GNoME: an AI tool that helped discover 2.2 million new crystals. 💎
Crystals are found in everything from the chips powering our phones to solar cells creating clean energy.
The model also better predicts the stability of new materials. 🧵
47
577
2K
18
368
1K
We have now completed our analysis of new materials reported in the Google Deepmind / Berkeley autonomous lab paper. My own initial analysis is in the quote tweet.
Happy to have worked with
@SchoopLab
to jointly put together a comprehensive analysis, now available on
@ChemRxiv
.…
This exciting paper shows AI design of materials, robotic synthesis. 10s of new compounds in 17 days.
But did they?
This paper has very serious problems in materials characterisation. In my view it should never have got near publication. Hold on tight let's take a look 😱
64
514
3K
20
215
940
The discovery of quasicrystals in 1982 literally redefined how we understand the solid state, and made a fool of one of the most famous chemists of all time.
On this day in 1982 Nobel Prize laureate Dan Shechtman (pictured left) discovered a crystal with 5-fold symmetry - a form that was thought to be impossible. His new crystals, called quasicrystals, caused uproar, and eventually textbooks had to be rewritten across the globe.
15
105
571
8
71
540
@Andercot
Nice post, many people don't realise this. But the energy cost needs consideration.
Going 110,000 ly in 12 years ship time for a small 10,000 kg ship costs 10^26J, or 10,000 times the current annual energy use of humanity
10
7
471
For every 10 likes this post gets, I'll ask the AI to make this Rietveld Refinement worse. 🤣
9
17
475
Initial thoughts from an experimental chemist:
Good news:
✅Cu doping calculated to shrink unit cell as reported
✅Band structure due to Cu 3d orbitals similar to other superconductors
✅Cu on Pb(1) site as reported is the one that's gives the exciting band structure
5
35
454
After the controversy on what is and isn't a perovskite - here's a chart to decide your own perovskite alignment (inspired by all the sandwich chart memes)
#perovskite
20
161
441
Obligatory mention that
@nature
charges you $11k to publish your article because of the quality of their peer review.
9
52
412
🚨Replication attempt🚨
Li Liu , Ziang Meng et al. from Beihang Univ. Beijing, publish a preprint attempting to replicate
#LK99
Thoughts below🧵
10
49
362
This thread shows our analysis of very serious problems in the Ceder group AI / Materials Design Paper
Read that one first to see the extent of the horror as this thread is just going to be me ranting
This exciting paper shows AI design of materials, robotic synthesis. 10s of new compounds in 17 days.
But did they?
This paper has very serious problems in materials characterisation. In my view it should never have got near publication. Hold on tight let's take a look 😱
64
514
3K
9
50
345
Latest replication attempt, and this one claims zero resistance. The paper is published as a video (obviously), link in Alex's thread, but here are my comments on the chemistry. 🧵
Well, here we have it! The first claimed replication of zero resistance in LK-99. As with the original pre-prints, the results are presented unconventionally (in a livestream on bilibili, a Chinese video sharing site). A few important caveats and notes in the thread
126
663
7K
3
21
314
Great news that John Goodenough has won the Nobel prize for Lithium ion batteries, here is some great advice he gave a few years ago to young material scientists
Ref: Chem. Mater. 2015, 27, 5149−5150
2
124
282
The
@BritishBakeOff
is back soon, so here's a handy ternary composition diagram of baking
#RealTimeChem
#gbbo
9
96
277
A new preprint on LK99 has caught some attention. Some solid state chemistry based thoughts....
Possible
Meissner
Effect
Near
Room
Temperature
In
Copper
Substituted
Lead
Apatite
229
360
4K
8
30
275
Sigh. Another one. It's actually Sb2Pb2O7.
ITS THE SAME PATTERN DID YOU NOT EVEN LOOK?
4
10
267
There is a lot of excitement about a possible new room temperature superconductor.
A lot of physicists have weighed in on the superconducting results but I haven't seen much comment on the material itself and how they made it. So here's my take 🧵
8/8 I cannot contain my excitement. It feels like January of 2020 with a huge wave coming that no one realizes yet, but in a much better way. What a time to be alive!! Check out the original paper:
126
1K
13K
10
59
268
The materials are substantially different. They really are.
We need an intervention now. I think I'm going to run an online course "XRD for Physicists"
Lesson 1 - just squinting at the pattern isn't valid analysis
Sign up below in comments
13
18
259
So they succeeded in making a 90 year old compound twice without realising and getting the composition wrong both times.
2
16
254
Erich Hückel (molecular orbital theory) h=11
Andrew Wiles (proved Fermat's Last Theorem) h=15
3
9
239
Only thing is, both Mg6MnO8 and Ni6MnO8 are known compounds and are cubic. And in fact solid solutions of these two compounds, which might be written Mg3MnNi3O8, were reported in 1995
1
6
230
For clarity, if you are not used to looking at PXRD - the residual (red line) should be as flat as possible. This residual is bigger than most of the peaks. There is no way this is a reliable refinement EVEN if it was combined with other characterisation. But as the sole form? No
2
7
218
The paper reports many 'new' compounds. The only characterisation they show is powder XRD. What no compositional analysis? Nope.
But maybe it's ok if they did a great job with the XRD analysis?
Oh.
9
12
219
Cheetham and Seshadri found "scant evidence for compounds that fulfil the trifecta of novelty, credibility, and utility"
Once again, just like the A-lab work that I wrote about earlier, a major problem here was compositional disorder
2
11
218
I've seen two posts on the Zhihu website that have more information on the synthesis of the new sulfur version of LK99, in the preprint by Wang et al.
Some more of the motivation for this synthesis is becoming clearer, and some pretty interesting points emerge.....
A new preprint on LK99 has caught some attention. Some solid state chemistry based thoughts....
8
30
275
5
23
211
@LoftusSteve
Embarrassing post Steve. Protestors attacked police, of course they will be arrested. What should the police do when attacked?
Revealing that you excuse violent conduct but want to ban signs you disagree with. Dark path you are on, luckily one always rejected by UK.
38
7
204
Using h-indices as a marker of legitimacy is phrenology for academia.
6
53
383
4
3
188
The headline conclusions:
It is likely that no new materials were discovered.
Most of the materials produced were misidentified, and the rest were already known
This happened because of problems in both the computational and the experimental portions of the work.
2
23
192
😬
BBC News - New material found by AI could reduce lithium use in batteries
11
12
183
The high throughput computational methods used by Google (and others) cannot handle compositional disorder, and that means they are severely limited in what they can achieve in the inorganic materials space.
C and R found that many material predictions "are highly unlikely to be…
1
8
185
Brussel sprout perovskites? It can only be the SSCG Christmas Meeting
2
20
174
Sorry did I say twice? I meant they made the same compound three times, because here is Sn2Sb2Pb4O13, or in reality, Sb2Pb2O7
4
9
171
A short explainer on compositional disorder.
Imagine a square grid. Place one circle on each intersection of the grid. The circles are arranged in an ordered pattern, this is similar to the order found in a crystal (although of course real crystals are 3D not 2D)
4
27
164
XRD fans look at that PXRD pattern of the 'new', 'hexagonal' Mg3MnNi3O8 two tweets ago. It looks very cubic doesn't it. Very Fm-3m.
We don't have the raw XRD data but looks likely this is a cubic pattern, and is in fact a solid solution reported 28 years ago.
1
2
163
OK this is getting tedious now - YET ANOTHER Sb2Pb2O7 misidentified. Same as the last three
2
5
157
How about this - MnAgO2. Again the fitting is dreadful. But worse, it was reported in 2021 and was in fact solved by another high throughput computational team and is in the ICSD (the database they used in this paper)
1
7
154
🚨PhD Position Available🚨
!Please RT!
I have a funded PhD position at UCL to start Sept 2023, for UK or INTERNATIONAL students to study photoemission spectroscopy of halide perovskite solar cells.
Deadline 26 Jan
@HarwellXPS
@UKSAF1
9
99
144
Meet Mg3MnNi3O8. A 'new' compound indexed as a hexagonal crystal system, with very interesting cation ordering
1
3
150
@JuliaHB1
@MikeJWake
You have 1000 hospital beds and 10,000 will eventually need them with covid. If all 10,000 show up on the same day, 9,000 don't get a bed. If the 10k are spread out over 10 months all get a bed. You can imagine the effect on death rates. You're welcome.
14
1
143
My personal view hasn't changed from my initial analysis. This paper is incorrect in its headline claims. No new materials were discovered. The best thing to happen here is for the paper to be retracted while these fundamental issues are fixed, both on the computational and…
7
14
147
Well I wonder, are there any known compounds of Sb, Pb and O. Why yes, one is Sb2Pb2O7, reported in 1933.
And it has exactly the same pattern as the two 'new' compounds above.
Collection Code 24246 on ICSD if you want to look
2
6
142
A thorough summary of progress towards unsupervised machine learning for XRD analysis. Very relevant for the recent A lab paper.
Great work by
@SodeAndy
,
@keeeto2000
and coworkers
3
21
140
Anyway - in summary - a failed reproduction attempt. Great detail on the synthesis, but seemingly only one attempt, didn't make thin film, but no SC properties detected.
2
5
135
For past 24hrs my twitter feed has been full of Nature editors making fun JACS for launching a new journal and I'm approaching my LD50 of irony
4
13
135
Luckily, Nature publishes the referee reports. Of the three referees, not one of them show any evidence of even having looked at the patterns. How shocking is that, in a paper reporting 10s of new compounds, that no one checked if they are actually real.
2
11
131
This exciting paper shows AI design of materials, robotic synthesis. 10s of new compounds in 17 days.
But did they?
This paper has very serious problems in materials characterisation. In my view it should never have got near publication. Hold on tight let's take a look 😱
64
514
3K
2
1
131
A few thoughts on the
@cedergroup
AI paper.
A few people have suggested my reason for this critique is that I don't like AI in science.
I actually thin AI will play a huge role in science in the coming years. Especially in Chemistry, including solid state chemistry
3
5
129
Now the bad news:
❌Cu is much more stable on the other Pb site, Pb(2)
✅ Clever chemistry can be used to prepare less stable structures but...
❌ reported high temp prep of
#lk99
would usually favour the most stable (thermodynamic) product
5
2
127
Look at these two patterns, for different 'new' materials. Hmm they look very similar don't they. They both contain Sb, Pb and O, and apparently one has Hf and the other has Fe...
1
5
127
UCL Chemistry has a position open for a Lecturer (Assistant Professor) in Inorganic and Materials Chemistry
Looking to start your own lab? Want to move your lab?
Link below, please RT, and to find out more have a read of the🧵
1
80
126
Aside from the cursed phrase "x-ray diffraction spectrum", (which I don't personally see as so terrible but some people HATE with a passion)..
this is NOT highly consistent. There are tonnes of extra peaks in the green spectrum, sorry, pattern, compared with the black
5
2
118
And of course the
#LK99
sample itself.
Sadly, it showed only semiconducting behaviour, and only weak diamagnetism.
In short, it isn't a superconductor.😪
2
5
111
A new paper on chemrxiv shows that one of the Ceder 'new materials' Ba6Ta2Na2V2O17, has a different structure to the one reported and predicted in their recent Nature paper
We have synthsized Ba6Ta2Na2V2O17 last year, which is accidentaly included in the new compounds report by
@cedergroup
last week. My refienment of the crystal structure leads different conclusion. Please check my manuscript in ChemRxiv.
2
12
80
3
14
109
Sunday morning quote:
"The first principle is that you must not fool yourself, and you are the easiest person to fool."
Richard Feynman
0
2
105
The computational predictions suffered as they generated structures with ordered cations, where in fact the same or very similar compounds are known with disordered cations.
This inability to deal with compositional disorder is an important limitation of the methods used here.
3
3
103
The
#LK99
patent reveals more about the chemistry of the new superconductor
The patent specifies a large range of compositions, far beyond the original papers. It also includes some analysis of the actual compositions of the material - some thoughts...🧵
3
21
97
👀
[1/N] Generative AI has revolutionized how we create text and images. How about designing novel materials? We at
@MSFTResearch
#AI4Science
are thrilled to announce MatterGen: our generative model that enables broad property-guided materials design.
👇
27
225
902
5
7
96
A very nice paper on autonomous labs: crystal growth and powder XRD identification of organic polymorphs, using computationally predicted structures.
Congratulations to
@amylunt
whose PhD work on autonomous solid state chemistry just came out in
@ChemicalScience
.
@erc_adam
/
@MIF_UoL
/
@LC_Mater_Design
3
26
143
4
11
96
The team behind
@uclchemlab_14
have put together a home practical chemistry course for SEVEN HUNDRED first year students to do university lab practicals remotely
Experiments include, Making Edible Glass, Titrating Turmeric, Building a colorimeter using LEGO and many more
Chocks away. 723 students registered
#Lab_14
. Part 1 of the Handbook has just been uploaded. The kits go out next week.
#butterflies
and jangling nerves.
2
4
40
3
18
93
What can we learn from this?
First, if you don't know how to characterise materials, please ask someone who does.
XRD does not tell you composition. It really doesn't.
2
6
93
I do not believe our comments are nit-picking or minor technicalities.
Identifying materials correctly is a cornerstone of inorganic chemistry.
Compositional disorder isn't a minor niche issue, it is a fundamental feature of materials, one that drives their functionality in…
3
1
92
Overall this is a great effort to get calcs done so quickly. I can't really comment much on the technical soundness of the calcs themselves but others certainly will!
3
0
87
I hope everyone is proud of me for not using lattice in this thread I feel this is real growth.
A short explainer on compositional disorder.
Imagine a square grid. Place one circle on each intersection of the grid. The circles are arranged in an ordered pattern, this is similar to the order found in a crystal (although of course real crystals are 3D not 2D)
4
27
164
5
0
81
@LoftusSteve
They're not attacking police, Steve, or being violent in any way. They are holding a sign you personally don't like.
35
0
78
Its likely that even if you had a SC phase, the many many grain boundaries in a compressed pellet would (I expect) make it look like a semiconductor.
2
2
78
One point that might have been missed is that the original Lee, Kim, Kwon preprint, the measurements were done on a thin film sample (produced by some mysterious thermal evaporation process).
Whereas Liu et al have measured on a compressed pellet
3
4
78
They do a great job in this paper of describing very clearly each step of the synthesis (with pictures!) They also measure the conductivity and magnetism of everything they can get their hands on (i.e. all the precursors and intermediates)
2
4
79
Exactly on the money. We do not know the composition or structure of
#LK99
with any confidence.
Looking at this
#LK99
structure more closely, and as others have pointed out in various ways, its not only the Cu atoms that are curious, but also the absence of a µ3-O. Most apatite structures have a channel of monovalent anions (Ca-F/OH, teeth!) etc.
4
9
145
2
7
78
Reviewer - the title says "Inorganic Materials". It's too general. You only make oxide powders.
Response: OK we removed the word "Inorganic" 🤣
Would love to have this confidence.
2
4
79
After the very early train from London, a group brunch before we head to the solid state Christmas meeting!
@SscgR
0
1
77
@PeakSquirrel
The standard of science is to present evidence to support your claims.
This is the only evidence provided of the formation of Ba2ZrSnO6.
Can you explain why I should be convinced by this?
5
6
78
When the predicted, ordered, compounds were synthesised experimentally, the known, disordered, compounds were produced instead.
This accounted for 2/3 of the whole list of compounds in the paper.
1
1
78
You might notice that one compound has no errors, this is CaFeP2O9. This seems to have been predicted and synthesised, but sadly was published just prior to the publication of the Nature paper
1
4
76
I believe these issues are crucial for high throughput materials prediction. Compositional disorder is so common in inorganic materials, that dealing with it is vital for inorganic materials predictions.
It doesn't matter how many inorganic materials you predict if a feature as…
1
3
75
In 14 years of marriage I am being trusted to make the tarka by myself for the first time
#wishmeluck
5
0
72
Today is 10 years since I joined
@UCLChemistry
as a lecturer! 🎂🔟
There's so many rewarding parts of this job it is hard to pick out the highlights. But some are:
➡️Working as head of outreach for 5 years, meeting a huge number of pupils, teachers, and members of the public
5
1
71
An editor of Cell, published by Elsevier, complains that not enough people are willing to do free peer review during a global pandemic.
Elsevier made a profit of £940 million in 2018, a 37% margin.
I am getting really frustrated by the number of scientists declining to peer-review manuscripts at the moment. Please step up and help us academic editors out. We don't have time either, especially to chase after 10s of names to get reviewers. Support the system.
586
259
2K
2
11
72
We finally have some EDX measurements of composition- summarised below. Two sets of LK99 are measured (fig 28, 41) First thing to note is how oxygen and copper rich they are compared with theory (shown right)
Second is how far the reactants are from the stated LK99 composition
2
7
71
Penn got community noted 😲
Katalin Karikó and Drew Weissman, Penn’s historic mRNA vaccine research team, win 2023
@NobelPrize
in Medicine
#nobelprize
1K
671
3K
2
11
71
Conclusions: this paper shows alternative Pb precursor is viable for producing this material.
Making the synthesis more Cu and P rich did not lead to an apatite structure, note the 1:1 synthesis is already Cu rich but P poor.
Maybe other ways to add P could be considered.
1
0
67
Check out our new preprint on correlation analysis in XPS.
XPS analysis is difficult as many different peak models can fit data equally well. Here we combine an underlying chemical model and large datasets to improve interpretation.
2
14
68
The latest test of our in operando battery electrochemistry monitored by XPS
@NaNexgen
@HarwellXPS
Using vacuum stable ionic liquids we can do electrochemistry inside the XPS machine, monitoring changes to electrode surface chemistry in real time during cycling.
2
6
66
This paper should be retracted while you work with subject experts in XRD analysis to make your AI Refinement process better. If that is fixed it is possible this is groundbreaking work that will change our field.
But at present it is not.
2
6
67
We have a new preprint out, and idea we've been working on for a while now...!
XPS measures the electronic structure, DFT calculates the electronic structure.
We combine these methods quantitatively - calculating the valence band spectrum and fitting a multi phase model...
Phase quantification of heterogeneous surfaces using DFT simulated valence band photoemission spectra
0
1
8
2
11
64