Why the Kernel calculation of bilayer TMD takes lots of time?more than 10 times the monolayer TMD

fen

unread,

Mar 5, 2026, 8:54:16 PMMar 5

to BerkeleyGW Help

Zhenglu Li

unread,

Mar 6, 2026, 1:55:49 AMMar 6

to fen, BerkeleyGW Help

If in the bilayer you double the Nv and Nc bands, the scaling is (Nc*Nv)^2, which is 16x more matrix elements to compute. If the cell size becomes bigger (e.g., longer c axis), there is another overhead.

Best,

Zhenglu

From: fen <lxtf...@gmail.com>
Sent: Thursday, March 5, 2026 5:54 PM
To: BerkeleyGW Help <he...@berkeleygw.org>
Subject: Why the Kernel calculation of bilayer TMD takes lots of time?more than 10 times the monolayer TMD

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.
To view this discussion visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/bd675429-6be0-4ac7-8607-f8b883c3e4d6n%40berkeleygw.org.

fen

unread,

Mar 6, 2026, 2:56:56 AMMar 6

to BerkeleyGW Help, Zhenglu Li, fen

Thancks for you reply, Professor Li. In my condition, I just expand the interlayer spacing and longthen the c size(25A to 30A). Then kernel calculation need 30 days after the adjustment. But the calculation without expanding only need 3.5 days. I don't know what caused this.

On Friday, March 6, 2026 at 2:55:49 PM UTC+8 Zhenglu Li wrote:

If in the bilayer you double the Nv and Nc bands, the scaling is (Nc*Nv)^2, which is 16x more matrix elements to compute. If the cell size becomes bigger (e.g., longer c axis), there is another overhead.

Best,

Zhenglu

From: fen <lxtf...@gmail.com>
Sent: Thursday, March 5, 2026 5:54 PM
To: BerkeleyGW Help <he...@berkeleygw.org>
Subject: Why the Kernel calculation of bilayer TMD takes lots of time?more than 10 times the monolayer TMD

-- You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group. To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@ berkeleygw. org. To view this

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.

To unsubscribe from this group and stop receiving emails from it, send an email to help+unsubscribe@berkeleygw.org.

Zhenglu Li

unread,

Mar 6, 2026, 12:49:17 PMMar 6

to fen, BerkeleyGW Help

The N_G (number of G vectors, ~ cell size) in BSE kernel also has a steep scaling (at least N^3). To understand the behavior, it may need more information regarding your system size, energy cutoff, and some details of your computational resources. Also, if the 30 days estimation is based on the first few-step prediction, it may not be accurate.

Best,

Zhenglu

From: fen <lxtf...@gmail.com>
Sent: Thursday, March 5, 2026 11:56 PM
To: BerkeleyGW Help <he...@berkeleygw.org>
Cc: Zhenglu Li <zhen...@usc.edu>; fen <lxtf...@gmail.com>
Subject: Re: Why the Kernel calculation of bilayer TMD takes lots of time?more than 10 times the monolayer TMD

Reply all

Reply to author

Forward