Full and approximate polaritonic ADC energies up to second order #153

BauerMarco · 2022-08-09T12:35:33Z

In the following polaritonic is always referred to by QED (Quantum Electrodynamics)

Features:

QED-MP up to second order from QED-HF reference and standard HF
QED-ADC up to second order from QED-HF
QED-ADC up to first order from standard HF
Method, which approximates QED-ADC up to second order from QED-HF

All features have been tested thoroughly, although only the methods based on a QED-HF reference should be used, since they are available up to second order.
One can also request QED-ADC(2) from a QED-HF reference with the polaritonic coupling only expanded up to first order, which is correct, but it is not sure, how useful this feature really is...This has to be tested in a future benchmark.

New Keywords:

coupl: Coupling strength to the cavity mode
freq: Energy of the cavity mode
qed_hf: Whether a QED-HF reference is provided or not (default is True)
qed_approx: Use standard or approximated method (default is False, but should be set to True after a proper benchmark)
qed_full_diag: If full diagonalization of the QED matrix is requested (with qed_approx = False), you should use this guess

Additional comments:

The tests do not use a QED-HF reference, but still use qed_hf = True, so they can still be tested with any of the possible backend options adcc provides.
As far as I know, there are two QED-HF reference implementations available from open source, which are the hilbert package developed by the DePrince group, which is a plugin to the psi4 programm package, and an other quite basic implementation in the Polaritonic section of the psi4numpy, which is meant for educational purposes, but can be used just as well as the hilbert package. I will provide some examples for the website, after the reviewing process is done.
I should mention, that the corresponding paper is not published yet, so I could not reference it at the appropriate places in the code.

…sis sets, but performance is garbage

…due to the doubles symmetry of the independent blocks

…avidson intermediate results can now be given to Lanczos after e.g. max_iter is reached, for arbitrary convergence, since Davidson crashes for high omega and/or coupling. Triplet multiplicity for n_states calculation still occurs.

…spersion)

… now)...probably some factors are still wrong

… ready for testing)

…lemented the full_diagonalization option, where a full QED guess, (3 * adc_matrix + 2) guesses, is build. Care that at the moment the H_1 two particle part of the phot/elec ph_ph coupling blocks has been commented out, since this results in the matrix, that can be tested with the dipole property only.

… account, while electronic terms (including vacuum coupling to the quantized field) are taken into account through second order.

…first order for qed_adc2.

…r the QED method have been added as well.

…, instead of always doubling it.

lgtm-com · 2022-08-09T12:52:44Z

This pull request introduces 4 alerts when merging 671183b into e58bc07 - view on LGTM.com

new alerts:

3 for Unused exception object
1 for Testing equality to None

maxscheurer · 2022-08-10T10:44:28Z

Thanks for the contribution! This will, however, take some time to digest, but there are probably ways to make the integration of QED less "invasive" in the existing code structure, right @mfherbst? 👍

…e adjusted dispatch routine in tdm

mfherbst · 2022-08-28T14:15:26Z

Cool! Thanks for the contribution @BauerMarco.

Indeed, quite a bit of code. Some looks like duplications that can be avoided. I have to admit that I lack the time to review the code without some explanation what's going on. Would be good to have a chat at some point.

…. Furthermore, qed parameters are now passed directly into the ReferenceState, easing up conditions.

BauerMarco · 2022-09-09T08:57:19Z

Cool! Thanks for the contribution @BauerMarco.

Indeed, quite a bit of code. Some looks like duplications that can be avoided. I have to admit that I lack the time to review the code without some explanation what's going on. Would be good to have a chat at some point.

I am available after the STC ends (22nd of september) and would very much appreciate to have a chat.

…n-polaritonic hf reference. Also missing indentation, leading to unbounded errors in backends init has been inserted.

…ity, due to an imaginary contribution on the diagonal of the final qed-matrix, which is requested as imaginary part of the cavity photon energy

qed approx methods now accounts for photon losses in the cavity via imaginary frequency

BauerMarco

initial review cycle

BauerMarco · 2022-10-24T15:58:05Z

adcc/AdcMatrix.py

@@ -130,21 +134,35 @@ def __init__(self, method, hf_or_mp, block_orders=None, intermediates=None,
        self.ndim = 2
        self.extra_terms = []

+        if self.reference_state.qed:


check with cvs, whether .qed or .is_qed or etc.

BauerMarco · 2022-10-24T15:58:18Z

adcc/AdcMatrix.py

        "adc0":  dict(ph_ph=0, ph_pphh=None, pphh_ph=None, pphh_pphh=None),  # noqa: E501
        "adc1":  dict(ph_ph=1, ph_pphh=None, pphh_ph=None, pphh_pphh=None),  # noqa: E501
        "adc2":  dict(ph_ph=2, ph_pphh=1,    pphh_ph=1,    pphh_pphh=0),     # noqa: E501
        "adc2x": dict(ph_ph=2, ph_pphh=1,    pphh_ph=1,    pphh_pphh=1),     # noqa: E501
        "adc3":  dict(ph_ph=3, ph_pphh=2,    pphh_ph=2,    pphh_pphh=1),     # noqa: E501
    }

+    qed_default_block_orders = {


merge dicts

BauerMarco · 2022-10-24T16:12:42Z

adcc/workflow.py

@@ -47,7 +50,8 @@ def run_adc(data_or_matrix, n_states=None, kind="any", conv_tol=None,
            n_guesses_doubles=None, output=sys.stdout, core_orbitals=None,
            frozen_core=None, frozen_virtual=None, method=None,
            n_singlets=None, n_triplets=None, n_spin_flip=None,
-            environment=None, **solverargs):
+            environment=None, coupl=None, freq=None, qed_hf=True,
+            qed_approx=False, qed_full_diag=False, **solverargs):


helper function for qed_full_diag

BauerMarco · 2022-10-24T16:25:48Z

adcc/backends/psi4.py

+            # and print the result
+            orben_a = np.asarray(self.wfn.epsilon_a())
+            orben_b = np.asarray(self.wfn.epsilon_b())
+            if all(orben_a == orben_b):


return all(orben_a == orben_b)

BauerMarco · 2022-10-24T16:28:15Z

adcc/backends/psi4.py

+            else:
+                print("This is an unrestricted calculation")
+                return False
+        else:


if isinstance(self.wfn, (psi4.core.RHF...
if isinstance(self.wfn, (psi4.core.UHF)):
else:
wavefunction object

BauerMarco · 2022-10-24T16:44:13Z

adcc/AmplitudeVector.py

@@ -217,3 +219,160 @@ def __radd__(self, other):
        else:
            ret = {k: other + tensor for k, tensor in self.items()}
            return AmplitudeVector(**ret)
+
+
+class QED_AmplitudeVector:


merge this class with AmplitudeVector

…long with other minor changes

Review cycle1

… are now supported

…object.

BauerMarco added 30 commits July 1, 2021 10:37

qed adc1

f886af7

qed-adc2 equations without adjustments to vector- and matrix class

b94d13f

added the lower order terms to all second order blocks

28235eb

full qed-adc2 converges for few states of all tested molecules and ba…

c64a6d6

…sis sets, but performance is garbage

adc1 expanded to also include doubly excited photonic contributions

c906dc9

still trying to fix the symmetry problem, which is probably not only …

adc8695

…due to the doubles symmetry of the independent blocks

still trying to fix the symmetry problem

3495f9b

symmetry seems to work fine with restricted triples or singles

bdee15d

check

7fdb127

check

ad0b144

check

9bdca70

enabled qed-adc1 from qed-hf reference (for single + double photon di…

c2c9a1a

…spersion)

started correcting for wrong factors (cancellations should be done by…

bf46dde

… now)...probably some factors are still wrong

some factors are still wrong

b3580ac

QED-ADC(2) for QED-HF reference should work now

380a582

included check for unrestricted or restricted reference

0c56424

cashed some stuff in matrix.py

77a96ea

corrected QED-ADC(2) from QED-HF reference (should be correct now and…

8179a2b

… ready for testing)

added the possibility to only take first order photonic coupling into…

001e2d3

… account, while electronic terms (including vacuum coupling to the quantized field) are taken into account through second order.

first_order_coupling option now also treats additive part to ERIs in …

7b8ac29

…first order for qed_adc2.

first_order_coupling option now also treats additive part to ERIs in …

1866ce1

…first order for qed_adc2.

uncommented missing H_1 terms in photonic coupling blocks

479ab98

corrected mistake within the qed pphh_ph (and vice versa) blocks

f1c9203

corrected sign error in ph_pphh photonic coupling blocks

ad3f8c7

full_diagonalize option included for qed-adc(1)

c60a4b0

added omega with corresponding factor to diagonal doubles space

0940196

added 1. order coupling blocks to first order coupling

b209615

corrected the scaling factor for omega in the doubles block

06c18fe

BauerMarco and others added 9 commits July 4, 2022 17:15

numerical inconsistency in approx function found and removed

28ca773

reintroduced approx method

89b12cc

no printout from LazyMp anymore

8dfbb93

started with qed test

aa44ebb

qed parameters can now also be set via the run_adc function. Tests fo…

cb369d1

…r the QED method have been added as well.

less printout and made the max subspace for qed a fully manual option…

0d7d166

…, instead of always doubling it.

final cleanup for initial commit to master

c25cc1d

Merge pull request #2 from BauerMarco/qed_merge_check

f214171

removed mp2_diffdm for qed, since it has not been tested thoroughly yet

671183b

lgtm corrections

ec2f2dd

BauerMarco added 2 commits August 10, 2022 18:41

making the linter happy

da9555a

removed numerical inconsistency of approx method, by reintroducing th…

9d5f74a

…e adjusted dispatch routine in tdm

adapted the test cases, so they also don't fail, if all tests are run…

1fabe94

…. Furthermore, qed parameters are now passed directly into the ReferenceState, easing up conditions.

BauerMarco and others added 3 commits October 4, 2022 16:14

apparently mp1 contribution got added twice in qed-mp2 energy from no…

9cec484

…n-polaritonic hf reference. Also missing indentation, leading to unbounded errors in backends init has been inserted.

qed approx method is now able to account for photon losses in the cav…

e8a032d

…ity, due to an imaginary contribution on the diagonal of the final qed-matrix, which is requested as imaginary part of the cavity photon energy

Merge pull request #3 from BauerMarco/qed_adc_complex_omega

45c7b8e

qed approx methods now accounts for photon losses in the cavity via imaginary frequency

BauerMarco commented Oct 25, 2022

View reviewed changes

BauerMarco and others added 8 commits October 27, 2022 14:41

qed adc runs with AmplitudeVector object, instead of QED_AmplitudeVector

af4b380

further code cleanup, due to full qed handling via AmplitudeVector, a…

529c6a6

…long with other minor changes

further cleanup

f9771b1

Merge branch 'adc-connect:master' into review_cycle1

21a1585

Merge pull request #4 from BauerMarco/review_cycle1

6edf332

Review cycle1

corrected qed_t1 denominator

9b7b618

arbitrary spatial orientations for polaritonic frequency and coupling…

5bbb3eb

… are now supported

make quasi-diabatic matrices accessible from resulting ExcitedStates …

889e213

…object.

BauerMarco mentioned this pull request Jul 27, 2024

Increase inheritable objects #174

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Full and approximate polaritonic ADC energies up to second order #153

Full and approximate polaritonic ADC energies up to second order #153

BauerMarco commented Aug 9, 2022

lgtm-com bot commented Aug 9, 2022

maxscheurer commented Aug 10, 2022

mfherbst commented Aug 28, 2022

BauerMarco commented Sep 9, 2022

BauerMarco left a comment

BauerMarco Oct 24, 2022

BauerMarco Oct 24, 2022

BauerMarco Oct 24, 2022

BauerMarco Oct 24, 2022

BauerMarco Oct 24, 2022

BauerMarco Oct 24, 2022

Full and approximate polaritonic ADC energies up to second order #153

Are you sure you want to change the base?

Full and approximate polaritonic ADC energies up to second order #153

Conversation

BauerMarco commented Aug 9, 2022

lgtm-com bot commented Aug 9, 2022

maxscheurer commented Aug 10, 2022

mfherbst commented Aug 28, 2022

BauerMarco commented Sep 9, 2022

BauerMarco left a comment

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment

BauerMarco Oct 24, 2022

Choose a reason for hiding this comment